The importance of soil drying and re-wetting in crop phytohormonal and nutritional responses to deficit irrigation

Soil drying and re-wetting (DRW) occurs at varying frequencies and intensities during crop production, and is deliberately used in water-saving irrigation techniques that aim to enhance crop water use efficiency. Soil drying not only limits root water uptake which can (but not always) perturb shoot water status, but also alters root synthesis of phytohormones and their transport to shoots to regulate leaf growth and gas exchange. Re-wetting the soil rapidly restores leaf water potential and leaf growth (minutes to hours), but gas exchange recovers more slowly (hours to days), probably mediated by sustained changes in root to shoot phytohormonal signalling. Partial rootzone drying (PRD) deliberately irrigates only part of the rootzone, while the remainder is allowed to dry. Alternating these wet and dry zones (thus re-wetting dry soil) substantially improves crop yields compared with maintaining fixed wet and dry zones or conventional deficit irrigation, and modifies phytohormonal (especially abscisic acid) signalling. Alternate wetting and drying (AWD) of rice can also improve yield compared with paddy culture, and is correlated with altered phytohormonal (including cytokinin) signalling. Both PRD and AWD can improve crop nutrition, and re-wetting dry soil provokes both physical and biological changes which affect soil nutrient availability. Whether this alters crop nutrient uptake depends on competition between plant and microbes for nutrients, with the rate of re-wetting determining microbial dynamics. Nevertheless, studies that examine the effects of soil DRW on both crop nutritional and phytohormonal responses are relatively rare; thus, determining the cause(s) of enhanced crop yields under AWD and PRD remains challenging.


Introduction
Plants in natural ecosystems, and rainfed or irrigated agriculture, are exposed to fluctuating soil water availability, characterized by repeated soil drying and re-wetting (DRW) cycles. These cycles affect both plant water uptake and the activity of soil microbes involved in nutrient cycling, thereby affecting soil nutrient availability. Nevertheless, many soil drying experiments reported in the plant physiological literature expose plants to a single drying cycle (often until loss of plant turgor), usually to evaluate relationships between soil and plant water status and physiological responses such as leaf growth and gas exchange. In contrast, the irrigation science community is concerned with using such information to optimize crop yields and water use efficiency (WUE) by understanding when and how much to irrigate. Exploiting our knowledge of how plants and soils respond to DRW cycles seems critical to optimize resource use efficiency in agriculture, especially given global concerns such as dwindling supplies of fresh water for irrigation (Elliot et al., 2014), and the limited availability and high cost of synthetic fertilizers to boost crop yields (Cordell et al., 2009).
Rapid and continued human population growth, combined with global economic growth, has placed increasing pressure on agricultural systems worldwide to deliver sufficient food. Where water resources permit, a solution adopted in many parts of the world has been to increase the land area under irrigated agriculture, since crop yields typically outstrip those of rainfed agriculture. Sometimes, this expansion of irrigation usage has been supported by farmers pumping aquifer water to supplement natural rainfall, and exploitation of aquifers has resulted in falling water table levels (Qin et al., 2013) and increased costs to the farmer of pumping water. Consequently, improving crop WUE (the harvested yield per unit of irrigation water supplied) while maintaining or even improving yield may conserve valuable water resources while allowing farmers to maintain their economic livelihoods.
Each day, a plant transpires more than its actual water content. When water resources are ample, complete replacement of crop evapotranspirational losses may be an appropriate management technique. Increasingly, due to scarcity of water resources, applying less water than crop evapotranspiration (termed 'deficit irrigation'; DI) has been implemented by various techniques that vary either the timing or the spatial distribution of irrigation (or both). Sustained DI applies a fraction of crop evapotranspiration to the entire rootzone throughout the growing season. Regulated DI applies a fraction of crop evapotranspiration to the entire rootzone, but varies the timing of deficit periods throughout the growing season to ensure that more water is supplied during sensitive phenological stages (e.g. during reproductive development) and less at other times. Partial rootzone drying (PRD) irrigates only part of the rootzone (e.g. one side of a row of vines or alternate furrows), although the fraction of crop evapotranspiration applied can vary. The irrigated and drying parts of the rootzone can remain fixed throughout the growing season (fixed PRD) or be regularly alternated (alternate PRD) at variable frequency. Alternate wetting and drying (AWD) is mainly applied to paddy rice, and involves alternate inundation and soil drying of the entire rootzone. These techniques have usually improved crop WUE, and sometimes either actual and/ or economic yields (Fereres and Soriano, 2007;Dodd, 2009;Davies et al., 2011), but results may be inconsistent and there is still a knowledge gap between our understanding of agronomic responses that integrate crop performance over the entire growing season, and of physiological responses that may be restricted to specific tissues at specific times of the growing season.
The irrigation frequency applied in such techniques will determine the degree and depth of soil drying, and subsequent plant physiological responses to both drying and re-irrigation. Interestingly, both alternate PRD and AWD explicitly impose DRW cycles, and the agronomic impacts of these cycles can be substantial. Figure 1 compares the yield response of plants supplied with the same irrigation volumes, but exposed to either fixed (part of the rootzone allowed to dry and the other part irrigated) or alternate (drying and irrigated parts of the rootzone are frequently swapped) PRD. Alternate PRD consistently outyielded fixed PRD, and it is tempting to speculate that physiological responses to repeated DRW cycles during alternate PRD are responsible.
This paper reviews changes in plant physiological responses (emphasizing root to shoot phytohormonal signalling affected by soil water availability; Davies and Zhang, 1991;Dodd, 2005) and soil nutrient availability induced by DRW cycles. Whether these changes in soil nutrient availability affect crop nutrition and root to shoot signalling of soil water availability is also explored. Although it may be premature to attribute the yield response identified in Fig. 1 to specific processes, further understanding of soil-plant interactions may inform management decisions that aim to improve the efficiency of irrigated agriculture.
the regulation of these processes may improve agronomic and breeding approaches to tune their sensitivity to mitigate the impacts of water deficits on crop yields. One of the most enduring controversies in plant water relations (Kramer, 1988;Passioura, 1988a;Christmann et al., 2007) is whether leaf growth and gas exchange are regulated hydraulically [via changes in leaf water potential (Ψ leaf ) or turgor] or chemically [via changes in ionic and phytohormonal signals originating from the roots or shoots). Several lines of evidence suggest that trying to distinguish the effects of each in the shoot is somewhat artificial, as the effect of the plant hormone abscisic acid (ABA) on stomatal aperture is mediated by Ψ leaf (Tardieu and Davies, 1992), and ABA affects stomatal conductance (g s ) both directly by acting on the stomatal guard cells and also indirectly by decreasing leaf hydraulic conductance (Pantin et al., 2013). Nevertheless, it is useful to consider that both forms of regulation can alter crop responses to soil DRW.
One of the earliest pieces of evidence provided for the existence and importance of root to shoot chemical signals were experiments that frequently measured leaf water relations, leaf growth, and gas exchange in response to soil drying. In some experiments, g s and leaf growth declined before any change in Ψ leaf (Blackman and Davies, 1985;Henson et al., 1989). Indeed, many plant species (termed 'isohydric') maintain Ψ leaf as the soil dries as a result of stomatal closure . Nevertheless, the relationship between g s and Ψ leaf in a single species can be positive, negative, or even absent according to the duration of soil drying (cf. Dodd et al., 2006;Kudoyarova et al., 2007), suggesting that stomatal closure cannot always be attributed to decreased Ψ leaf .
Another set of experiments applied a pneumatic pressure to the roots of plants to prevent any loss of leaf turgor during soil drying. Root pressurization was unable to prevent or reverse stomatal closure and leaf growth inhibition in some (Gollan et al., 1986;Passioura, 1988b) but not all (Saliendra et al., 1995;Fuchs and Livingston, 1996;Mencuccini et al., 2000;Yao et al., 2001) species. Stomatal closure when plants were maintained at full turgor has been interpreted as evidence for the existence of root-supplied chemical signals regulating shoot responses, while reversal of stomatal closure by applying pressure to the roots suggests the primacy of leaf water relations in regulating shoot responses. Clearly, both types of signal can play important regulatory roles as the soil dries and either may be more or less effective in different species and according to the time of day (Mencuccini et al., 2000).
Unlike this inconsistency in the effects of Ψ leaf , there is surprising consistency in some phytohormonal responses to soil drying, as assessed by determining xylem sap composition of plants exposed to drying soil. A brief literature survey (four separate studies: Masia et al., 1994;Hansen and Dörffling, 2003;Megat Wahab, 2007;Alvarez et al., 2008) revealed that xylem concentrations of ABA (a potent antitranspirant) increased 2-to 100-fold as the soil dried, while concentrations of the cytokinins zeatin and zeatin riboside declined compared with well-watered plants (Perez-Alfocea et al., 2011). In contrast, xylem ionic (nitrate, phosphate, sulphate, chloride, calcium, sodium, magnesium, and potassium) responses were much more variable (in another five studies: Munns and King, 1988;Trejo, 1994;Goodger et al. 2005;Megat Wahab 2007;Ernst et al., 2010). This suggests that some xylem phytohormone concentrations provide a more consistent 'measure' of the degree of soil drying than xylem ion concentrations. If there is no change in xylem loading (movement of ions and phytohormones from root cells into the xylem lumen) as xylem sap flow rates decrease (as a result of stomatal closure), it is predicted that the concentrations of all xylem sap constituents should increase as the soil dries (Jackson, 1993). However, increases and decreases in xylem sap phytohormone and ion concentrations indicate pronounced changes in xylem loading as the soil dries, although considerable additional work is needed to establish that changes in xylem sap composition can actually regulate leaf growth and gas exchange (Dodd, 2005). Furthermore, multianalyte hormone measurements of xylem sap composition (e.g. Albacete et al., 2008;Alvarez et al., 2008) will doubtless demonstrate other changes in xylem phytohormone composition as the soil dries.
While different soil drying episodes (varying in intensity and duration) induce variable changes in xylem sap composition and Ψ leaf , plants with low Ψ leaf generally show rapid (within hours) recovery of Ψ leaf and leaf growth in response to re-wetting the soil (Loewenstein and Pallardy, 2002;Parent et al., 2009). Recovery of Ψ leaf and leaf elongation upon rehydration paralleled increases in root hydraulic conductivity and were successfully modelled by determining water transfer following rehydration (Parent et al., 2009). Following rewetting, Ψ leaf and leaf elongation recovered more rapidly in maize lines with higher endogenous ABA concentration due to their higher root hydraulic conductivity.
In contrast, stomatal responses to re-wetting the soil can be significantly delayed. Following prolonged soil drying, there can be long-term (several days) residual suppression of g s (Fig. 2). These 'after-effects' on stomatal responses (Davies and Kozlowski, 1977;Dörffling et al., 1977) have been associated with elevated xylem ABA concentrations (Correia and Pereira, 1994;Loewenstein and Pallardy, 2002). Even though xylem ABA concentration sometimes recovers to the levels of well-watered plants while stomata remain partially closed (Fig. 2), rapid recovery of Ψ leaf yet slower recovery of g s may be considered as further evidence for the existence of root to shoot signals regulating stomatal responses. Interestingly, stomatal conductance of rewatered plants appears to be more sensitive to xylem ABA than in droughted plants (Loewenstein and Pallardy, 2002), although it is unknown whether this is related to elevated apoplastic ABA concentrations around the guard cells, as a result of changes in catabolism of xylemsupplied ABA (Gowing et al., 1993) and/or sequestration of ABA into the mesophyll (Wilkinson and Davies, 1997). An alternative explanation is that rewatering also changes root to shoot signalling of other phytohormones, which may alter stomatal sensitivity to ABA (Dodd et al., 1996).
Rewatering bare-rooted mandarin seedlings, that had been transplanted to dry sand for 24 h, caused exponential decreases in both xylem ABA and 1-aminocyclopropane-1-carboxylic acid (ACC) concentrations (Tudela and Primo-Millo, 1992). Peak ACC concentrations were attributed to ACC export from the root system (after root ACC accumulation during exposure to drying soil) following the resumption of water uptake from the root system. Following rewatering, stomatal re-opening was achieved more rapidly in plants which had been pre-treated with a soil drench of fluridone (an ABA biosynthesis inhibitor) that diminished both root and xylem sap (but not leaf) ABA concentrations (Gomez-Cadenas et al., 1996). However, fluridone also attenuated rehydration-induced changes in root, xylem sap, and leaf ACC concentrations, and leaf ethylene evolution (Gomez-Cadenas et al., 1996), consistent with evidence that ethylene can inhibit stomatal opening. In contrast, 1-methylcyclopropene treatment (which inhibits ethylene binding to receptors) prevented the stomata of older wheat leaves from opening in response to rehydration (Chen et al., 2013), although the dynamics of xylem and leaf phytohormone responses were not measured. Further work is needed to evaluate the role(s) of both ABA and ethylene in regulating physiological responses to rewatering.
Changes in xylem cytokinin concentrations and delivery may also occur following rewatering. Rewatering sunflower plants (that had reached a Ψ leaf of -1.0 MPa) caused an exponential decline of xylem ABA concentration (reaching a plateau after 4 h) while xylem zeatin riboside concentration rapidly increased (~5-fold) within 6 h of rewatering (Hansen and Dörffling, 2003). Cytokinin-mediated antagonism of ABA-induced stomatal closure was hypothesized to assist stomatal re-opening. The potential physiological importance of cytokinins was further suggested in experiments with Kentucky bluegrass (Poa pratensis) that had dehydrated to 40% leaf relative water content, such that rewatering accompanied by foliar sprays of a synthetic cytokinin (10 μM 6-benzylaminopurine) promoted stomatal re-opening (to a greater extent than rewatering alone), although supplementary cytokinins had no significant effect on photosynthetic recovery (Hu et al., 2013). Interestingly, foliar sprays of 50 mM potassium chloride also significantly enhanced the recovery of g s . This effect may also be cytokinin mediated, as potassium deficiency decreases plant cytokinin status (Wang et al., 2012a), although a direct relationship between foliar potassium sprays and enhanced foliar cytokinin synthesis remains to be demonstrated. Enhanced xylem cytokinin concentrations following rewatering may occur in response to increased root water potential (Ψ root ) (Hansen and Dörffling, 2003) and/or altered nutrient uptake.
While there are multiple phytohormonal mechanisms that may regulate stomatal responses following rewatering, alternative explanations may also apply. Severe dehydration causes membrane injury, and it was postulated that restoration of ion transport across the guard cell plasma membrane may be delayed following rewatering (Loewenstein and Pallardy, 2002). While this theory may be supported by accelerated recovery of g s if rewatering is accompanied by foliar potassium spraying (Hu et al., 2013), detailed ultrastructural measurements are required to establish the existence of guard cell membrane injury (and recovery).
In general, there have been relatively few studies of changes in xylem sap composition following rewatering of dry soil (in contrast to multiple studies that have considered responses to drying soil). These studies should consider effects of soil and root water status on water, nutrient, and hormone uptake from the soil, hormone synthesis and metabolism in the roots, and xylem loading of both nutrients and hormones. Whether whole-plant responses to rewatering can help interpret responses to rewatering only part of the rootzone (as occurs in alternate PRD) is considered next. Relative changes in stomatal conductance (a) and xylem ABA concentration (b) of plants that were exposed to soil drying and then re-watered at time 0 h (compared with well-watered plants=100% and 1.0, respectively). Data are re-plotted from Gomez-Cadenas et al. (1996) (Mandarin, filled circles) and Loewenstein and Pallardy (2002)

Responses to alternate partial rootzone drying (PRD)
PRD was originally conceived in the laboratory, to disentangle the effects of hydraulic and chemical signals on shoot physiology. When maize plants were grown with their roots split between two soil compartments, and only one compartment was watered, g s declined while Ψ leaf was maintained (Blackman and Davies, 1985), suggesting that g s was responding to chemical signals generated in the roots in the dry compartment. This concept was subsequently applied to grapevine in the field by placing drip-emitters on separate feeder lines either side of the vine row (Dry et al., 1996) or watering every other furrow of maize plants (Kang et al., 1997) to exploit chemical signalling to limit excessive vegetative vigour (grape) and luxury transpiration (maize), thereby improving crop WUE.
One of the difficulties of applying PRD is that although soil drying is needed to alter root metabolism to produce chemical signals in the roots, water uptake from those roots is also diminished. Thus chemical signal production is increased while signal transport (from those roots) is decreased, and the relative sensitivity of those two processes to soil drying will determine whether root to shoot chemical signalling is actually enhanced by PRD. While both processes can be measured (e.g. Dodd et al., 2010), ultimately sap flux from roots in drying soil may cease entirely, thus allowing stomatal reopening as the duration of PRD increases (Khalil and Grace, 1993;Stoll et al., 2000). A practical solution to this problem has been to alternate the wet and dry parts of the rootzone regularly, to ensure that some roots remain in drying (not completely dry) soil, allowing continued signal production and transport to the shoot.
Given this conceptual basis, it is surprising that few studies have attempted to measure xylem sap composition during alternating PRD cycles, even in containerized plants. When tomato plants were grown in split pots, alternating the wet and dry soil compartments had no effect on Ψ leaf compared with plants supplied with the same irrigation volume but with the wet and dry soil compartments remaining fixed (Dodd et al., 2006). However, within 1 h of alternating the irrigation, leaf xylem ABA concentration ([X-ABA] leaf ) significantly increased and g s further declined, effects which persisted throughout the entire photoperiod but not beyond. Enhanced [X-ABA] leaf following re-watering the originally dry soil compartment probably mobilized root-sourced ABA to the transpiration stream. It was suggested that soil drying caused root ABA accumulation by enhancing root ABA biosynthesis, but then reduced water uptake (and sap flow) limited transport of this additional ABA to the shoots. Nevertheless, increased [X-ABA] leaf following rewatering (irrigation alternation) of PRD plants is not universal (Wang et al., 2012c;Pérez-Pérez and Dodd, 2015). Hydraulic redistribution of water from irrigated to drying roots during a drying cycle (Caldwell and Richards, 1989;Burgess et al., 1998) may limit root ABA accumulation by mitigating decreases in Ψ root (Puertolas et al., 2013). The magnitude and duration of any ABA pulse following rewatering probably depends on the dynamics of water uptake by the previously dried roots, the size of the ABA pool accumulated within these roots, and possible losses of ABA from the root system via metabolism and ABA efflux to the soil after rewatering. Should rewatering occur during the photoperiod, Ψ root will be lower than soil water potential (Ψ soil ), thus any root exudation into the rhizosphere must occur in opposition to massive plant water uptake from the soil. Theoretically, an efflux of ABA-loaded water to the soil can only happen when soil water is redistributed from wet to dry soil during the night via the plant (Fig. 3).
Increased [X-ABA] leaf following alternation of the wet and dry parts of the root system may depend on the severity of soil drying. Alternating wet and dry sides after 3 d of PRD had no effect on [X-ABA] leaf or g s , and further drying of part of the soil was needed to liberate an ABA pulse following rewatering (Dodd et al., 2006). More recent experiments in the same soil type failed to demonstrate this ABA pulse, and maintaining or alternating the wet and dry parts of the rootzone resulted in similar relationships between [X-ABA] leaf , Ψ leaf , and whole-pot soil water availability (Pérez-Pérez and , even though a similar degree of soil drying occurred to that in the previous study (Dodd et al., 2006). The chief difference between the two studies was the depth of the soil profile (13 cm versus 30 cm in Pérez-Pérez and versus Dodd et al., 2006 but, paradoxically, root ABA accumulation was attenuated when there were more pronounced vertical gradients in soil moisture than a more homogenous soil moisture distribution (Puertolas et al., 2013). Furthermore, roots apparently accumulate more ABA (at the same soil water content) when PRD is imposed laterally rather than vertically (Puertolas et al., 2015), perhaps due to differences in ages of the roots sampled. Both these observations suggest that growing PRD plants in long columns (rather than shorter pots) should have prevented a stimulation of [X-ABA] leaf following irrigation alternation. A potential explanation is that when root systems are constrained within pots, those growing at the pot's edge will be exposed to different environmental conditions from those embedded in the substrate; thus, differences in surface area to volume ratio between container sizes may result in different root system ABA accumulation at the same bulk soil water content (Pérez-Pérez and . Clearly, further studies are needed to understand the factors affecting root ABA accumulation and export of ABA to the shoot during PRD, to identify conditions under which alternating the wet and dry parts of the root system stimulates ABA transport to the shoot. Changes in ABA export following alternation of wet and dry parts of the root system are probably of physiological significance in field-grown plants. On the day following re-irrigation, whole-plant water use of grapevines was much more conservative when plants were grown with PRD than conventional irrigation (Collins et al., 2010). Nevertheless, relatively few studies have measured changes in xylem sap composition in response to alternate PRD in the field. Diurnal measurements revealed that mid-morning (09:00-12:00 h) xylem ABA concentration of PRD grapevine was approximately double that of well-watered plants (Stoll et al., 2000), although the timing of these measurements relative to the alternation of wet and dry parts of the rootzone was not specified. Prior to irrigation, field-grown tomato plants exposed to PRD and DI had similar [X-ABA] leaf throughout a diurnal cycle (Topcu et al., 2007). Following re-watering, PRD and DI plants only had a similar [X-ABA] leaf during the early morning (before 10:00 h), with higher [X-ABA] leaf detected in PRD plants during the afternoon. Similar diurnal measurements in grapevine following alternation of wet and dry sides revealed some significant differences in [X-ABA] leaf between PRD and DI plants, but there was no consistent response throughout the day (Romero et al., 2012). In lemon trees grown in the same geographical region, [X-ABA] leaf did not differ between PRD and DI treatments either before or immediately after (<12 h) rewatering the dry side of PRD trees, yet crop WUE was still greater in PRD trees (Pérez-Pérez et al., 2012). Paired measurements of [X-ABA] leaf and soil moisture availability are necessary to determine whether different irrigation techniques induce genuine differences in chemical signalling.
Variability in the responses of [X-ABA] leaf to drying and rewetting cycles during PRD implies that it may be possible to develop models as a conceptual framework to determine circumstances in which alternation of PRD irrigation enhances chemical signalling in planta. ABA is produced throughout the plant as a function of local water potential . To model a root-induced hormonal signal it is important to know Ψ root , which is in turn a function of the distribution of Ψ soil , the transpiration demand, and the hydraulic conductivity of the plant (Javaux et al., 2013). When soil moisture is non-uniformly distributed (as in PRD experiments), different gradients in Ψ soil will develop and this will redistribute water within the soil or, during the night, through the root system that acts as hydraulic bridges between wet and dry parts in the soil (Stoll et al., 2000). Further, root water uptake from the dry part of the soil may be compensated by uptake from the wetter parts of the rootzone if Ψ soil is non-limiting, thereby preventing local (rhizosphere) soil moisture depletion (Javaux et al., 2013). Nevertheless, even if the soil water is ample, non-uniform soil moisture distribution decreases xylem water potential (if not bulk Ψ leaf ; Blackman and Davies, 1985;Sobeih et al., 2004) more than when soil moisture is uniform (Huber et al., 2014). This can impact on ABA dynamics during PRD.
One method to discriminate between hydraulic and hormonal responses can be simulation studies. Integrating root-induced hormonal signals and stomatal regulation as a function of signal concentration and Ψ leaf (Tardieu and Davies, 1993) within a functional-structural plant model showed that an isohydric response, as observed in maize, could not be adequately described with a root-induced ABA signal (Lobet et al., 2013). Another simulation study (Huber et al., 2014), where the same relationship for a hormonal signal was used, but implemented in a 3D coupled soil and root water flow model (Javaux et al., 2008), illustrated the impact of non-uniformly distributed soil water on Ψ root , resulting hormone concentrations, and subsequent stomatal regulation. Although this study simulated realistic hormone concentrations in the shoot, generally the stomatal reaction to non-uniform soil drying was adequately described by plant hydraulics alone. However, this study did reproduce effects such as the previously described 'pulse' of xylem ABA following rewetting of a previously dry soil (Dodd et al., 2006). This was attributed to the cessation of water uptake when soil became too dry, which trapped hormones in the roots, which were only transported when this part of the soil was rewetted.
These simulation studies only consider the impact of ABA on g s ; however, ABA's influence on root hydraulic conductivity (Parent et al., 2009) will directly affect the production of ABA due to changes in Ψ root . Root water potential will also be affected by hysteretic hydraulic properties of the rhizosphere, which can turn hydrophobic upon drying and can remain in that state for >1.5 h after rewetting (Carminati et al., 2010;Zarebanadkouki and Carminati, 2014). This leads to a simultaneous reduction in root water uptake and could prolong both the production and transport limitation of root-sourced ABA. Thus rhizosphere properties could directly affect the arrival time and intensity of an ABA pulse following rewatering part of the rootzone.
Since changes in root to shoot chemical signalling following alternation of wet and dry soil compartments during PRD may be transient, they may have limited physiological effects on plant performance when integrated over its life cycle. Furthermore, the logistical challenges of investigating root to shoot signalling in the field mean that it may be more illuminating to determine integrative responses such as cumulative whole-plant water use. At the single-leaf scale, measurements of carbon isotope discrimination (Δ 13 C) in C 3 species can indicate the degree of stomatal closure throughout the leaf lifespan, with greater stomatal closure and WUE indicated by greater (less negative) Δ 13 C (Condon et al., 2004).
Although alternate PRD and DI had statistically similar effects on foliar Δ 13 C of field-grown grape (de Souza et al., 2003) and glasshouse-grown tomato (Wang et al., 2010a), the latter study indicated consistently greater Δ 13 C of shoot material of PRD plants. While these measurements were correlated with increased whole-plant WUE of PRD-grown tomato, it is less certain whether these effects can be attributed to phytohormonal or nutritional effects, as foliar Δ 13 C increased with increased specific leaf N content (Wang et al., 2010a). Further studies in tomato and potato indicated variable effects of PRD (relative to DI) on foliar Δ 13 C according to N fertilization rates, with higher foliar Δ 13 C of PRD plants at low (70 mg N kg −1 soil) N fertilizer rates but lower foliar Δ 13 C at high (200 mg N kg −1 soil) fertilizer rates (Yan et al., 2012;Sun et al., 2013). These differences reflected variation in xylem ABA concentration in tomato which probably effected stomatal closure (Yan et al., 2012). While such isotopic measurements have repeatedly demonstrated good correlations with agronomic measurements of WUE, the importance of soil DRW cycles has yet to be illustrated by specific comparisons of Δ 13 C in plants exposed to alternate and fixed PRD. Although such measurements usefully demonstrate seasonal responses, they cannot inform the management of PRD irrigation to determine the optimum time to alternate wet and dry parts of the root system.
Despite these experimental and (more recently) simulation studies of ABA dynamics during alternate PRD, and meta-analyses showing the agronomic benefits of the technique compared with conventional DI (Dodd, 2009) and fixed PRD (Fig. 1), no studies have satisfactorily demonstrated how altered phytohormone signalling during PRD alters crop yields. ABA-induced stomatal closure may limit photosynthetic carbon gain even while improving instantaneous leaf WUE, although improved leaf water relations of ABA-overexpressing transgenics allowed greater leaf extension (Thompson et al., 2007). Paradoxically, the limitation of ABA transport from roots during prolonged drying cycles in PRD (Dodd et al., 2010) and the consequent decline in xylem ABA concentration (which potentially allows higher photosynthesis), rather than any transient pulse in xylem ABA concentration following alternation of wet and dry parts of the root system (Dodd et al., 2006), may be responsible for the enhanced yield of PRD plants compared with DI plants. Simulations that model the impacts of ABA signalling on whole-plant carbon gain are likely to resolve this uncertainty.

Alternate wetting and drying (AWD)
While the effects of soil DRW during PRD are visualized by comparing plants exposed to alternate and fixed PRD (Fig. 1), the effects of DRW during AWD are compared against continuous inundation of rice paddies. The agronomic effects of AWD seem less predictable (reviewed in Price et al., 2013), possibly due to considerable variation in the phenological phases in the crop life cycle when AWD is applied, which may be related to the onset of the monsoon in some geographical areas. Thus AWD is applied post-transplanting to the onset of flowering in Bangladesh (Price et al., 2013), while in China it is maintained throughout the cropping cycle . However, another key factor is the extent to which the soil is allowed to dry during AWD.
Compared with plants maintained under continuous flooding, during soil drying cycles AWD decreased photosynthesis linearly with soil water potential (Fig. 4), with even mild soil drying (soil water potential, Ψ soil , of -15 kPa measured with a tensiometer at 15-20 cm depth) decreasing photosynthesis. Photosynthesis increased following rewatering (compared with continuously flooded plants), but only if soil drying was not too severe.
These photosynthetic changes were closely associated with foliar levels of both zeatin-type and isopentenyl-type cytokinins, with more severe soil drying (Ψ soil of -50 kPa measured with a tensiometer at 15-20 cm depth) decreasing leaf cytokinin levels and rewatering of plants exposed to mild soil drying increasing leaf cytokinin levels . It is less certain whether changes in root cytokinin biosynthesis (and subsequent export to the shoot) are involved. While mild soil drying had no effect on root cytokinin levels, severe soil drying increased concentrations of zeatin-type cytokinins and decreased concentrations of isopentenyl-type cytokinins: changes that were completely immune to rewetting the soil . Likewise, root indole acetic acid levels (which were increased ~25% by severe soil drying) showed no change the day after the soil was rewatered. This relative stability of root hormone concentrations contrasts with the relatively rapid changes in xylem hormone concentrations following re-wetting of drying soil (e.g. Fig. 2) and may reflect the relatively mild soil water deficits applied during AWD. However, further investigation of the relationship between root and xylem hormone concentrations is needed to determine possible regulation of xylem fluxes by root hormone biosynthesis during AWD.
In a series of papers, Yang and colleagues compared the effects of AWD and continuous flooding on the phytohormone profile of developing rice grains, in attempting to establish whether specific hormonal messages from the root system cause specific physiological responses. Although phytohormone concentrations and grain-filling rates of superior (early flowering) spikelets were similar in plants exposed to AWD and continuous flooding, inferior (late flowering) spikelets showed pronounced differences in grain hormone concentrations. Plants exposed to AWD (Ψ soil of -25 kPa measured with a tensiometer at 15-20 cm depth) had higher grain ABA concentrations throughout growth (Zhang et al., 2012) while grain cytokinin concentrations fluctuated with AWD cycles and the severity of soil drying . Rewatering enhanced grain cytokinin concentrations following moderate soil drying (-25 kPa) but had no effect following severe soil drying (-50 kPa) which generally decreased cytokinin concentrations by ~25%. Nevertheless, it remains to be demonstrated whether xylem sap phytohormones can regulate the concentrations detected in developing grains, and exogenous ABA and cytokinin treatments had similar effects to AWD on grain-filling rate (Zhang et al., , 2012, implying that grain hormone concentrations can act independently of rootsupplied chemical signals. These changes in phytohormone signalling during AWD will not only affect leaf gas exchange, but will also modulate grain filling and tiller dynamics. Later flowering inferior spikelets accumulated more ABA under AWD than continuous flooding, which was correlated with increased activities of enzymes involved in sucrose to starch conversion (higher sink strength) and grain-filling rate (Zhang et al., 2012). Phytohormonal changes induced by AWD may also regulate sink strength of tillers, as AWD sometimes increases the percentage of productive tillers (Xue et al., 2013). Establishing a role for root to shoot signalling in regulating such developmental responses requires further attention.
Soil nutrient availability responses to soil drying and re-wetting

The 'Birch effect'
The contribution of soil DRW cycles to increased mineral nutrient availability, first reported by Birch (1958), is well documented. The effect describes a pulse of organic matter decomposition and mineralization upon rewetting a dried soil, with a decline in rate over time. Different interacting soil processes regulate the occurrence and magnitude of the 'Birch effect' in determining nutrient availability. Briefly, more intense drying (such as that occurring at high temperatures or for long durations) induces more mineralization upon rewetting, and recurrent DRW cycles increase mineralization overall, compared with continuously moist soils (Jarvis et al., 2007), emphasizing the importance of soil moisture fluctuations in driving the underlying nutrient cycling processes. However, whether these fluctuations can be actively managed to enhance soil nutrient provision to plants has received little attention.
Both physical and biological processes cause and regulate the 'Birch effect' (Fig. 5). Physical processes during DRW events disrupt soil aggregates, which can be important causes of nutrient release. Slaking (physical breakdown) caused by DRW can release molybdate-reactive (inorganic) and -unreactive (organic) phosphorus (P) (MRP and MUP, respectively) (Haygarth et al., 1998). Reactive P is readily available to plants. Following rewetting, the extractable reactive P concentration was significantly higher (by up to 44%) in a sterilized, dried soil than a constantly moist soil (Bünemann et al., 2013), indicating the importance of physical processes. Furthermore, DRW can detach soil colloids thereby increasing MRP and MUP in leachate due to shrinkage and swelling of soil aggregates (Chepkwony et al., 2001;Blackwell et al., 2009). However, DRW can also stimulate the P adsorption capacity of some soils, temporarily reducing P availability, which recovers to initial levels following rewetting (Haynes and Swift, 1985).
Biological processes contributing to the 'Birch effect' include the death of soil microbes due to desiccation upon drying and lysis upon rapid rewetting (caused by osmotic shock), releasing their nutrient constituents into the soil solution. This effect contributed the majority (88-95%) of the increased water-extractable MUP from two fertilized pasture soils following DRW . In many studies, the quantities of organic P released into solution after drying were greater than those of inorganic P Haygarth, 2001, 2003;Blackwell et al., 2009; but see Butterly et al., 2011a), due largely to the contributions from the microbial biomass and soil organic matter (SOM) (Blackwell et al., 2009;Butterly et al., 2009). Microbial cell walls also become incorporated into the SOM, and mineralization of labile forms releases inorganic P, which is available for plant uptake and immobilization by both surviving and new microbes (Blackwell et al., 2010). Following rapid rewetting of dry soil, microbial biomass can recover over short time scales (e.g. within 6 h, based on measuring ATP synthesis) (De Nobili et al., 2006). Enhanced microbial assimilation of P suggests that any benefits to plants from increased availability of P following DRW events may be short-lived.
Physical and biological processes causing and regulating the 'Birch effect' are coupled. Soil DRW induces structural changes which release SOM from microaggregates. No longer protected, this SOM is more readily mineralized by soil microbes (Chepkwony et al., 2001;Navarro-Garcia et al., 2012). After drying clay soils, fractions of plant-available P (both MRP and sodium bicarbonate-extractable P) increased (as a proportion of total P), a result of structural changes in SOM and physical disturbance of microbial cells (Soinne et al., 2010).
Initial nutrient availability can affect the outcome of soil DRW events on nutrient release, and the important role of the soil microbial biomass in driving observed increases in nutrient availability following DRW has been substantiated by isotope studies. Following application of 32 P to a silt loam as an inorganic P source, DRW decreased the specific activity of 32 P in soil extracts due to dilution by mineralized 31 P, yet soilextractable P concentration and wheat P uptake and yields increased (Chepkwony et al., 2001). Mineralized P following DRW makes a greater contribution to total extractable P at higher inorganic fertilizer rates (supplied as monocalcium phosphate), since increased soil P adsorption during soil drying results in less adsorption of the mineralized P after rewetting (Chepkwony et al., 2001). Thus initial soil P status affects the impacts of DRW on P availability.
The intensity and rate of DRW also affect the magnitude of the 'Birch effect'. Dissolved organic C and N concentrations in leachates following DRW increased from fertilized but not unfertilized grassland soils, and the increase in dissolved inorganic N in leachate caused by DRW was most pronounced from the fertilized soil (Gordon et al., 2008).
Decreased microbial biomass N, particularly in the fertilized grassland soils, further suggests nutrient release from microbial cell lysis. Although the fertilized soil reached a lower Ψ soil (-5.74 MPa) than the unfertilized soil (-4.77 MPa) at the same soil water content, greater nutrient release from the fertilized soil was attributed to its lower fungi to bacteria ratio and lower nutrient retention capability. In both soils, it was expected that bacteria were inactive (previously reported at Ψ soil less than -1.0 MPa to -1.5 MPa; Swift et al., 1979) while fungi remained active (previously reported at Ψ soil of -15 MPa or even below -40 MPa, according to species ;Swift et al., 1979) (Fischer, 2009). In another study, upon drying to -1.5 MPa (widely recognized as the permanent wilting point for plants), microbial respiration was only 5-8% of the maximum respiration (Fischer, 2009), suggesting that microbial biomass was also adversely affected by the low Ψ soil . This supports the importance of the role of microbial community composition, structure, and function in nutrient dynamics under soil DRW cycles.
Soil carbon status largely determines the energy supply to microbes, regulating mineralization and the extent of the nutrient pulse. Microbial biomass size and composition also has a major regulatory role in determining the significance of the 'Birch effect' to nutrient availability in different soils (Jarvis, 2007;Gordon et al., 2008;Blackwell et al., 2010). Nevertheless, microbial immobilization of P released by DRW, indicated by the strongly reduced C:P ratio of the microbial biomass in the DRW soils compared with control soils, can result in some soils showing no change in reactive P concentrations following DRW (Butterly et al., 2011b). Understanding these transformations upon DRW and the competition between soil microbes and plants is necessary to determine the fate of released nutrients in different soils and management systems. Drying and re-wetting effects during deficit irrigation | 2247 Interacting physical and biological soil dynamics under DRW pose a challenge to determining the effects of irrigation management on nutrient availability, which is further complicated by the large variability in soil type and DRW methodologies but also nutrient analyses across different studies. To compare nutrient dynamics under different DRW cycles in different studies, it is necessary to relate soil water contents and Ψ soil via a moisture release curve, yet this characterization is often omitted from studies. Further work should prioritize determining thresholds of drying and rewetting intensities and rates at which soil nutrient availability increases, and quantifying the contributions of physiochemical processes and the microbial biomass in different soils and management conditions.

Deficit irrigation techniques and plant nutrient acquisition
Despite the frequently reported increase in N and P availability upon soil DRW, effects of fluctuating water regimes (implied in the DI techniques described above) on plant nutrient acquisition has received limited attention, particularly for P. However, some understanding of the nutritional effects of DI has been achieved.
Partial rootzone drying increased soil N availability (compared with DI) in several studies, with increased plant N concentration attributed to PRD stimulating the microbial biomass and mineralization of soil organic N, enhancing its availability (Wang et al., , 2010b(Wang et al., , 2012b. For containerized potato grown in peat , PRD increased leaf, stem, and tuber total N concentrations by 17, 35, and 24%, respectively, compared with plants exposed to standard DI. When tomato was grown in a sandy loam (Wang et al., 2010b), PRD increased plant N accumulation by 16% compared with plants under standard DI. Mineralization of 15 N-labelled straw was 25% greater in the PRD treatment, providing an enhanced pool of inorganic N available for uptake. These results were conserved across contrasting soil types and different crop species, probably due to the more intense soil drying of part of the pot during PRD (even though the same irrigation volumes were supplied in both PRD and DI), which affects microbial processes and nutrient availability.
Determining the relative contributions of plant and microbial processes to N acquisition during DRW remains an important challenge to address. Benefits of PRD on crop N uptake may not always be immediate, but rather may become apparent after several DRW cycles (Wang et al., 2012b). Higher shoot N concentrations following PRD may be due to DRW cycles stimulating mineralization of SOM, but additional data on soil N availability would be valuable. Future research could achieve this by labelling soil amendments with both N and P isotopes to trace nutrient movement in different soil pools, with repeated xylem sap sampling from an intact plant (Netting et al., 2012) determining nutrient uptake dynamics.
Measurements of xylem nutrient fluxes may provide a more sensitive indicator of momentary plant sensing of soil nutrient availability than bulk leaf nutrient concentrations. Root to shoot transfer of nutrients via the transpiration stream may be particularly valuable to detect how the intensity and rate of DRW regulates plant nutrient uptake. Root xylem phosphate concentration and Ψ root were significantly higher in tomato plants after 13 d of PRD (compared with DI) but did not differ thereafter (Wang et al., 2012c), suggesting that the lower Ψ root of DI plants may have inhibited P acquisition (Wang et al., 2012c). However, xylem sap was not collected at an appropriate flow rate (matching whole-plant transpiration to avoid concentration of xylem sap constituents) despite being critical to comparing xylem sap composition of plants grown at different Ψ root (Dodd et al., 2008). It is uncertain whether these differences in xylem phosphate concentration result from changes in nutrient availability within the soil or variation in plant nutrient uptake ability (as affected by Ψ root ).
Possible explanations for the inhibited nutrient uptake include reduced ion diffusion in the soil solution, decreased root nutrient uptake capacity mediated by low Ψ root , and the increased mechanical impedance of dry soil limiting root growth and thus the soil volume explored for nutrient acquisition (Misra and Tyler, 2000). Nevertheless, enhanced root growth has frequently been reported in plants following DRW (e.g. Engels et al., 1994) and specifically under PRD (e.g. Mingo et al., 2004;Wang et al., 2012b), which probably enhances root nutrient acquisition.
Nutritional benefits of PRD may not be restricted to nutrient uptake, but also may affect intraplant nutrient allocation. Although PRD and DI had similar N uptake, PRD plants allocated more N to fruits relative to total biomass N (Topcu et al., 2007). This higher N harvest index has great agronomic and economic importance, and it is likely that phythormone signalling within the plant will regulate intraplant N distribution (Pons et al., 2001).

Conclusions
While this review has emphasized phytohormonal and nutritional responses as being largely independent mechanisms contributing to crop yields under different deficit irrigation treatments (eg. Fig. 1), an abundant literature shows that variation in crop nutrient status affects phytohormonal signalling (Kudoyarova et al., 2015). For example, low P supply increases not only xylem ABA concentrations but also stomatal sensitivity to ABA (Radin, 1984). Quantifying xylem sap ionic and phytohormone composition in response to factorial combinations of irrigation (e.g. PRD versus DI) and fertilizer applications may account for physiological responses such as stomatal closure, and greater attention should be given to understanding the sensitivity of leaf tissues to these changes in root to shoot signalling.
While much of the deficit irrigation literature (especially PRD studies) has emphasized that elevated ABA status can increase crop WUE, decoupling of stomatal and xylem ABA responses to the re-wetting of dry soil (Fig. 2) suggests it is necessary to understand the temporal dynamics of other long-distance phytohormonal signals. While single-leaf measurements may be relatively straightforward, gradients of both stomatal conductance (Soar et al., 2004) and phytohormone concentrations (Havlova et al., 2008) within the canopy suggest that paired measurements of whole-plant gas exchange, and root system ionic and phytohormone export, are necessary to further our understanding of physiological and agronomic responses to different irrigation treatments. These measurements are particularly challenging in field environments due to the difficulties in collecting xylem sap (at flow rates approximating whole-plant transpiration) following whole-plant gas exchange measurements (Steduto et al., 2002), but they are technically feasible for whole plants grown in a pressure chamber and exposed to a single drying cycle in the laboratory (Jarvis and Davies, 1997).
Active management of DRW cycles may not only provide an opportunity to enhance carbon gain (Fig. 4) and yield (Price et al., 2013) relative to conventional management systems, but may also increase soil nutrient availability and crop uptake. More work is required for understanding the fundamental soil properties and mechanisms affecting changes in soil nutrient availability in diverse soils under different patterns of DRW. Gains in crop water and nutrient use efficiency achieved by exploiting these different deficit irrigation treatments will ensure continued research efforts in this area, but our improved understanding of fundamental plant and soil science processes must be better integrated with management decisions made by irrigation practitioners.