GABA signalling modulates stomatal opening to enhance plant water use efficiency and drought resilience

1The regulation of stomatal pore aperture is a key determinant of plant productivity and drought resilience, and profoundly impacts climate due to its influence on global carbon and water cycling–. The stomatal pore is delineated by a guard cell pair. Fine control of ion and water movement across guard cell membranes, via transport proteins, determines cell volume and pore aperture following opening and closing signals such as light and dark,, (Fig. 1). Due to their critical roles and their ability to respond to and integrate multiple stimuli, stomatal guard cells have become a preeminent model system for investigating plant cell signalling resulting in the elucidation of many critical pathways involved in plant biotic and abiotic stress tolerance–.

2GABA signalling in mammals relies upon receptor-mediated polarization of neuronal cell membranes,. Speculation that GABA could be a signal in plants is decades old, but a definitive demonstration of its mode of action remains elusive. GABA production in plants is upregulated by stress,. It is synthesised in the cytosol via the GABA shunt pathway, bypassing two stress-inhibited reactions of the mitochondrial-based tricarboxylic acid (TCA) cycle,. GABA is therefore well known as a stress-induced plant metabolite that is fed back into the mitochondrial TCA cycle to sustain cellular energy production,. The discovery that the activity of aluminium-activated malate transporters (ALMTs) can be regulated by GABA represents a plausible mechanism by which GABA signals could be transduced in plants, providing a putative—but unproven—novel signalling link between primary metabolism and physiology. Stomatal guard cells contain a number of ALMTs that impact stomatal movement and transpirational water loss–. Therefore, stomatal guard cells represent an ideal system to test whether GABA signalling occurs in plants.

3Significantly, here, we show that GABA does not initiate changes in stomatal pore aperture, rather it antagonises changes in pore size, which differentiates it from many of the signals known to regulate stomatal aperture–. Specifically, we find that GABA concentration increases under a water deficit and this reduces stomatal opening in an ALMT9-dependant manner. The anion channel ALMT9 is a major pathway for mediating anion uptake into the vacuole during stomatal opening; GABA signal transduction via ALMT9 leads to reduced transpirational water loss, increased water use efficiency (WUE) and improved drought resilience. As such, even though guard cell signalling is relatively well defined,, this study has been able to uncover another pathway regulating plant water loss. Furthermore, by revealing a mechanism by which GABA acts in stomatal guard cells, we demonstrate that GABA is a legitimate plant signalling molecule.

4To validate whether GABA is a physiological signal that modulates stomatal pore aperture, our initial experiments used excised Arabidopsis thaliana epidermal peels where stomatal guard cells are directly accessible to a chemical stimuli,–. When exogenous GABA or its analogue muscimol were applied under constant light or dark conditions, neither elicited a change in stomatal aperture (Fig. 2a, b; Supplementary Fig. 1a, b). Interestingly though, we found that both compounds suppressed light-induced stomatal opening and dark-induced stomatal closure (Fig. 2a, b; Supplementary Fig. 1a, b). We then fed intact leaves with an artificial sap solution through the detached petiole with or without the addition of GABA or muscimol and examined whether this affected gas exchange rates. We found, in the GABA and muscimol fed leaves, that the increase in water loss (transpiration) stimulated by a dark-to-light transition was dampened compared to leaves fed just the artificial sap solution due to reduced stomatal conductance (Fig. 2c; Supplementary Figs. 1c, d and 2a). This is consistent with the reduced extent of stomatal opening that we observed in epidermal peels in the presence of GABA or muscimol upon a dark-to-light transition (Fig. 2b; Supplementary Fig. 1a). The gas exchange values of fed leaves were used to calculate instantaneous intrinsic WUE (iWUE) and WUE (ratios of carbon gained through photosynthesis per unit of water lost), which are key traits underpinning drought tolerance in plants, and both values were greater (i.e. improved) in GABA fed leaves (Fig. 2d; Supplementary Fig. 2a–c).

5To examine whether GABA or muscimol can modulate stomatal aperture beyond the response to light and dark, we examined their impact on a range of opening and closing signals using epidermal peels of Arabidopsis. We found both GABA and muscimol inhibited abscisic acid- (ABA, 2.5 μM) or H2O2-stimulated stomatal closure and coronatine-induced opening (Supplementary Fig. 3a, c, e, f),. However, stomatal pores were fully closed in response to high concentrations of ABA (25 µM) (Supplementary Fig. 3b, d) or exogenous calcium in the presence of GABA or muscimol (Supplementary Fig. 3g), which indicated stomatal closure could occur in epidermal peels in the presence of GABA when the closing signal was of sufficient magnitude.

6We tested whether our results could be explained by GABA or muscimol treatment permanently locking guard cells in a closed (or open) state and preventing further change in stomatal pore aperture, which would argue against GABA being a physiological signal. We did this by incubating epidermal peels in GABA or muscimol, then removing this treatment and performing a light or dark transition. As would be expected from viable cells, after removal of the GABA or muscimol treatment, we found that stomatal guard cells responded to a light treatment by opening the pores (Supplementary Fig. 4a, b) or to a dark treatment by closing pores (Supplementary Fig. 4c, d). Collectively, these data again indicate that GABA signals would likely act to modulate stomatal aperture in the face of a stimulus rather than stimulating a transition itself.

7To test whether GABA is a universal modulator of stomatal control, we explored whether GABA or muscimol treatment of epidermal strips attenuated stomatal responses of other plant species to light or dark transitions, including the dicot crops Vicia faba (broad bean), Glycine max (soybean) and Nicotiana benthamiana (tobacco-relative) and the monocot Hordeum vulgare (barley) (Supplementary Fig. 5). The widespread inhibition of stomatal pore aperture changes suggests that GABA has the potential to be a universal ‘brake’ on stomatal movement in plants, including valuable crops.

8Stomatal control is explicitly linked with the regulation of plant water loss, which impacts the survival of plants under drought; the wider the stomatal aperture, the greater the water loss of plants, the poorer the survival of plants under a limited water supply, as excessive water use by the plant diminishes the availability of stored soil water. The observation that the stress-induced metabolite GABA reduces plant water loss and improves WUE (Fig. 2d; Supplementary Fig. 2c)—key factors underpinning drought tolerance—implicates GABA as novel signal regulating plant drought resilience. Therefore, to examine the hypothesis that endogenous GABA concentration increases under a water deficit and acts as a signal, we first determined whether we could replicate the previously reported increases in GABA accumulation under drought,, (Fig. 3; Supplementary Fig. 6). In wild-type plants, a drought treatment was applied by withholding watering, which resulted in the gradual depletion of soil gravimetric water and a reduction in leaf relative water content (RWC) (Supplementary Fig. 6a, b). We found that GABA accumulation in drought stressed leaves increased by 35% compared to that of well-watered leaves (water versus drought at 7 days: 1.07 ± 0.08 versus 1.44 ± 0.11 nmol mg−1 FW) (Supplementary Fig. 6c).

9To investigate whether GABA has a role during drought, we obtained Arabidopsis T-DNA insertional mutants for the major leaf GABA synthesis gene, Glutamate Decarboxylase 2 (GAD2). Both gad2-1 and gad2-2 had >75% less GABA accumulation in leaves than in wild-type plants, whilst GABA concentrations in roots were unchanged (Fig. 3a; Supplementary Fig. 6d–f). Furthermore, leaves of gad2 plants did not accumulate additional GABA under drought conditions unlike wild-type controls where GABA increased by 45% after 3 days, and was maintained at this elevated level after 7 days of drought (Fig. 3a). Under standard conditions, both gad2 mutant lines exhibited greater stomatal conductance and wider stomatal pores than wild-type plants (Fig. 3b; Supplementary Fig. 6g), whereas stomatal density was identical to wild type (Supplementary Fig. 6h). The application of exogenous GABA to gad2 leaves inhibited stomatal pore aperture changes in response to light treatments (Supplementary Fig. 6i, j), indicating that gad2 stomata would be competent in a GABA response if sufficient GABA was present. Furthermore, the aperture of GABA pre-treated gad2 stomata after a dark-to-light transition were statistically insignificant from non-GABA treated wild-type stomata (Supplementary Fig. 6j), which is consistent with GABA playing a role in modulating opening of wild-type stomata under non-stressed conditions. It has been shown previously that both GAD2 transcription and GABA accumulation exhibit diurnal regulation; GABA usually peaks at the end of the dark cycle prior to stomatal opening and reaches a minimum when stomatal conductance is at its maximum near subjective mid-day. However, during stress, both GAD2 transcript abundance and GABA accumulation remain high. This suggests GABA may further minimise stomatal opening under stress and contribute to drought tolerance.

10Under drought, the leaf RWC of gad2 plants lowered more quickly than in wild type (Fig. 3c). Transcriptional profiles of key ABA-marker gene (RD22) and GABA-related genes (other than GAD2) were similar in wild type and gad2 lines, although RD29A was significantly higher in gad2-1 than wild type and gad2-2 on day 0 and day 7 of the drought treatment (Supplementary Fig. 7), which is consistent with the lower RWC of gad2-1 after 7 days (Fig. 3c). These results confirm that GAD2 is critical for leaf GABA production under stress, and suggests that GABA itself may regulate plant water loss and drought tolerance.

11Histochemical staining corroborated that GAD2 is highly expressed in leaves, particularly in guard cells (Supplementary Fig. 8a, b). GAD2 is a cytosolic enzyme; to examine if cytosolic GABA biosynthesis within the guard cell was sufficient to modulate transpiration we expressed—specifically in the guard cell—a constitutively active form of GAD2 (GAD2Δ) that has a C-terminal autoinhibitory domain removed, (Fig. 4a). This led to a large increase in leaf GABA accumulation (Fig. 4b) and to complementation of the steady-state stomatal conductance and aperture phenotypes of gad2 plants to wild-type levels (Fig. 4c; Supplementary Fig. 8c, d). At the same time, no change in stomatal density or leaf ABA accumulation was detected under standard conditions (Supplementary Fig. 8e, f), suggesting the complementation of the gad2 phenotype was due to the restoration of GABA synthesis in the guard cell. Other phenotypes restored to wild-type levels by guard cell-specific expression of GAD2Δ included the exaggerated stomatal opening and closure kinetics and decreased instantaneous iWUE/WUE of gad2-1 (Fig. 4d–f; Supplementary Fig. 8g–i). The drought sensitivity of gad2, compared to wild type, was also abolished by guard cell-specific expression of GAD2Δ (Fig. 4g, h). This demonstrates GABA synthesis in guard cells was sufficient to modulate stomatal movement, regulate water loss and improve drought resilience.

12To examine whether GABA metabolism can be modulated to improve drought resilience beyond wild-type levels, GAD2Δ was expressed specifically in the guard cells of wild-type Arabidopsis plants (Fig. 5a), this resulted in leaf GABA concentrations being increased to beyond wild-type levels (Fig. 5b). The steady-state stomatal conductance of the GABA overproducing transgenic plants in standard and drought conditions was lowered compared to wild-type plants (Fig. 5c). Consistent with this, the plants overexpressing GAD2Δ in the wild-type background maintained higher leaf RWC than wild-type plants after 10 days of drought treatment (Fig. 5d, e). Furthermore, a greater percentage of plants overexpressing GAD2Δ in the wild-type background survived following re-watering after a 12-day drought treatment (Supplementary Fig. 9). As such, we show here that GABA overproduction can reduce water loss and improve drought resilience.

13Our data show that although guard cell synthesised GABA can rescue the gad2 phenotype, it is clear that exogenously applied GABA can also modulate stomatal movement (e.g. Fig. 2 for wild type or Supplementary Fig. 6i, j for gad2). It is known that GABA can pass the membrane through a variety of transporters–, so it is unclear whether the site of guard cell GABA action is from the apoplast or cytoplasm. We expressed GAD2Δ specifically in the spongy mesophyll, adjacent to the abaxial stomatal layer, to test whether it could complement gad2 (Supplementary Fig. 10a, b). This resulted in a significant increase in leaf GABA, but no change in stomatal conductance (Supplementary Fig. 10c, d). As such, unlike guard cell-specific expression, GAD2Δ in the spongy mesophyll was insufficient to complement the gad2-1 phenotype.

14To further probe the role of guard cell synthesised GABA, we expressed full-length GAD2 under the guard cell-specific promoter (gad2-1/GC1::GAD2) (Supplementary Fig. 11a). This form of GAD2 requires activation by Ca2+/calmodulin or low pH to synthesise GABA. Interestingly, guard cell-specific expression of full-length GAD2 failed to complement the high stomatal conductance of the gad2-1 line to wild-type levels under standard conditions, whereas its constitutive expression (driven by pro35S-CAMV) did (Supplementary Fig. 11b, f, g). Under drought, the gad2-1/GC1::GAD2 lines increased GABA production, reduced their stomatal conductance significantly more than that of gad2-1 plants and had a comparable leaf RWC to wild-type plants following 5 days of drought (Supplementary Fig. 11c–e). This suggests that activation of full-length GAD2 via its regulatory domain is important in stimulating GABA production under drought in guard cells.

15We extended our investigation of GABA’s site of action through an epidermal peel experiment. We compared the effects of exogenously applied muscimol or muscimol-BODIPY, a muscimol molecule conjugated with a BODIPY fluorophore, which is active against GABA targets in plants and animals, but lacks cell-membrane permeability,. We found that unlike muscimol, membrane impermeable muscimol-BODIPY was unable to inhibit stomatal opening or closure (Supplementary Fig. 12). This result—alongside the differential effects of gad2 complementation by full-length GAD2 when expressed constitutively or solely in the guard cell (Supplementary Fig. 11a–e)—provides further evidence that GABA is likely to pass the plasma membrane and that it acts from the cytosol, consistent with our feeding assays (e.g. Fig. 2c). Collectively, the data in this section demonstrate that guard cell-specific cytosolic GABA accumulation is sufficient and necessary for controlling stomatal aperture and transpiration under drought, but suggests a role for other cell types in fine-tuning GABA signals under standard conditions.

16ALMTs are plant-specific anion channels that share no homology to Cys-loop receptors except a region of 12 amino acid residues predicted to bind GABA in GABAA receptors,. In animals, ionotropic GABA receptors are stimulated by GABA; in contrast, anion currents through ALMTs are inhibited by GABA,. There are a number of ALMTs expressed in guard cells that contain the putative GABA binding motif and have the potential to transduce the GABA signal, with most having been shown to have a role in regulating stomatal movement–,. For instance, ALMT12 (also called QUAC1, quickly-activation anion conductance 1) is a plasma membrane localised anion channel, which moves anions out of the guard cell during guard cell closure.

17Under the conditions tested here, the impact of GABA on stomatal closure appears to be limited to epidermal peels, it is not seen in intact leaves. However, epidermal peels still represent an assay system that can be used to test whether ALMT might transduce the inhibitory effect of GABA on closure. We observed that, unlike wild-type plants, stomatal closure in almt12 knockouts was insensitive to GABA or muscimol when transitioning from light-to-dark (Supplementary Fig. 13a). In contrast, stomatal opening of almt12 lines showed wild-type-like sensitivity to GABA or muscimol when transitioning from dark to light (Supplementary Fig. 13b). These data indicate that ALMT12 is a plasma membrane GABA target that affects stomatal closure in response to dark—in epidermal peels at least.

18However, if GABA inhibition of ALMT12/QUAC1 played a significant role during drought, then the resulting inhibition of closure would translate into an increase in water loss compared to wild-type plants during closure. As we found no evidence that GABA had an effect on closure in intact leaves, under a light-to-dark transition as measured by stomatal conductance or transpiration (Fig. 2c; Supplementary Fig. 2a), and the fact that GABA accumulation led to a net decrease in water loss and improvement in drought resilience, ALMT12 is unlikely to be a major target contributing to this outcome. We therefore focused on tonoplast-localised ALMTs that are involved in stomatal pore opening, as this is the process where GABA has its predominant affect in intact leaves.

19ALMT9 is the major tonoplast-localised channel involved in anion uptake into guard cell vacuoles during stomatal opening, but has no documented role in closure. We hypothesised that GABA might target and inhibit ALMT9 activity to reduce the rate or extent of stomatal opening. We initially attempted in vitro electophysiological studies to examine the impact of GABA on ALMT9-induced currents, but were unable to consistently detect stable currents following heterologous expression in either Xenopus laevis oocytes or tobacco mesophyll cells,. Therefore, we examined the potential regulation of ALMT9 by GABA by focusing solely on in planta studies as it is difficult to faithfully replicate regulatory pathways from guard cells in heterologous systems, e.g.–. In the first instance, we independently crossed two almt9 alleles (almt9-1 and almt9-2) with gad2-1. We found that, similar to gad2, both double mutants (gad2-1/almt9-1 and gad2-1/almt9-2) maintained low GABA accumulation in their leaves (Fig. 6a, b; Supplementary Fig. 14a, e). However, both gad2-1/almt9-1 and gad2-1/almt9-2 had wild-type-like stomatal conductance and aperture unlike gad2-1 where both these parameters are high (Fig. 6c, d; Supplementary Fig. 14d, f). Furthermore, guard cell-specific complementation of gad2-1/almt9-1 by GAD2Δ did not alter stomatal conductance (Supplementary Fig. 14a–d). Collectively, these data are consistent with ALMT9 being required for GABA to regulate gas exchange via stomatal control. An interesting additional observation was that the loss of ALMT9 in gad2-1 also resulted in ABA inducing stomatal pore closure to wild-type levels (Supplementary Fig. 14g–j), indicating that, although ALMT9 is a channel that regulates stomatal opening, it can influence the extent to which stomatal pores close under certain conditions (in epidermal peels at least). The incomplete stomatal closure of gad2 coupled to its greater stomatal opening may further contribute to its drought sensitivity. These findings are consistent with the regulation of stomatal aperture being a dynamic equilibrium between the pathways that regulate stomatal opening and closure, with stomatal aperture being weighted towards a particular state dependent upon the dominant stimuli,.

20To further test whether ALMT9 transduces GABA signalling, we examined the effect of GABA on regulating stomatal opening in almt9 mutant plants. In wild-type plants, we previously showed that light-induced stomatal opening was inhibited by exogenous GABA (Fig. 2a) or muscimol (Supplementary Fig. 1a). In almt9 lines, exogenous GABA or muscimol did not antagonise stomatal opening (Fig. 7a, b; Supplementary Fig. 15a, b), whereas dark-induced stomatal closure in almt9 retained its GABA sensitivity (Fig. 7c, d; Supplementary Fig. 15c, d). These results are consistent with GABA reducing stomatal opening via negative regulation ALMT9-mediated Cl– uptake into guard cell vacuoles. Furthermore, it strongly indicates the corollary of this finding, that the higher stomatal conductance phenotype of gad2 is the result of greater ALMT9 activity due to its lack of inhibition by GABA.

21We tested this hypothesis by attempting to complement almt9 plants with either the native channel or a site-directed ALMT9 mutant (ALMT9F243C/Y245C). The mutations within ALMT9F243C/Y245C are in the 12 amino acid residue motif that shares homology with a GABA binding region in mammalian GABAA receptors,. Mutations in the aromatic amino acid residues in this motif have been shown for other ALMTs to result in active channels that are not inhibited by GABA when tested in heterologous systems, (Fig. 8; Fig. 9). However, no in planta tests have been conducted to date—for any ALMT—to determine whether mutations in this region result in a transport competent protein that lacks GABA sensitivity. Here, we observed that ALMT9 and ALMT9F243C/Y245C had similar expression in almt9-2 complementation lines and the mutations (in ALMT9F243C/Y245C) did not alter the membrane localisation with both versions of the ALMT9 protein being clearly present on the tonoplast (Supplementary Fig. 16). Further, we found that similar to almt9 lines, almt9-2 expressing ALMT9F243C/Y245C was insensitive to GABA during a dark-to-light transition assayed on epidermal peels and detached leaves, for stomatal opening and stomatal conductance, respectively; this contrasts the GABA sensitivity of wild-type plants and plants expressing native ALMT9 in the almt9-2 background (Figs. 2c and 8a, b, d, f; Supplementary Fig. 17a, b, d, e, g, h). Furthermore, instantaneous iWUE/WUE of almt9-2 was improved by native ALMT9 complementation, but not ALMT9F243C/Y245C (Fig. 8c, e, g; Supplementary Fig. 17c, f, i). Steady-state stomatal conductance and aperture of ALMT9F243C/Y245C lines were also significantly greater than that of wild-type and almt9 lines and were insignificant from gad2-1 under standard conditions (Fig. 9; Supplementary Fig. 17j). This result indicates that we successfully complemented almt9 with an active, but GABA-insensitive form of ALMT9, and that this increased transpirational water loss over wild-type levels. These data are completely consistent with ALMT9 being a GABA target that regulates plant water loss, even under non-stressed conditions, through modulation of ALMT9 activity. The GABA effect is then amplified under a water deficit when GABA concentration increases. We propose that GABA accumulation has a role in promoting drought resilience by reducing the amplitude of stomatal re-opening each morning, which minimises whole plant water loss. As such, the GABA–ALMT pathway is a strong candidate for constituting the ABA-independent stress memory of a decreased soil water status that has been previously proposed without mechanistic attribution,.

22The data in this manuscript have unveiled a GABA signalling pathway in plants, which can be summarised by the simplified models presented in Fig. 10. We propose that cytosolic GABA signals, generated by GAD2, modulate stomatal opening, WUE and drought resilience transduced through negative regulation of ALMT9 activity (Fig. 10).

23Collectively our use of leaf feeding, knockouts, complementation and point mutagenesis strongly suggests ALMT9 is an essential and major component transducing GABA signalling in guard cells during well-watered and drought conditions. As has become evident for other guard cell based signalling pathways through their examination over time–, we are cognizant of the potential that other GABA response elements, including other ALMT, may be involved in transducing and fine-tuning this signalling pathway. Our finding that GABA does not impact stomatal closure in epidermal peels of almt12 knockouts infers a potential role for this plasma membrane localised ALMT12 in transducing guard cell GABA signals. The fact that light-induced stomatal opening and dark-induced stomatal closure was completely GABA insensitive in almt9xalmt12 knockouts (Supplementary Fig. 18) suggests that both channels have the potential to transduce the major effects of GABA in guard cells.

24However, it is interesting that GABA inhibition of stomatal opening was consistently seen between epidermal peel assays and leaf feeding, whilst GABA only inhibited stomatal closure during isolated epidermal peel experiments, but not when it was fed to leaves. This suggests that GABA acts through ALMT12 on processes associated with stomatal closure, but in the context of an intact leaf this phenotype is lost, which is likely due to the loss of functional epidermal and/or mesophyll cells. This is consistent with the growing body of evidence that indicates stomatal aperture experiments on isolated epidermal peels require validation via studies on intact leaves to avoid overinterpreting potential artifacts from this reductionist system. However, it also means we cannot fully rule out whether GABA inhibition of stomatal closure does have a role under certain physiological scenarios that are yet to be identified. Therefore, in future studies, it would be pertinent to examine whether ALMT12-dependent GABA inhibition of stomatal closure has a physiological role in transducing GABA signals in conditions not examined here, and, more broadly, whether other ALMTs or additional elements are involved in GABA signal transduction.

25ALMT activity appears to be regulated by a suite of factors including anions, (Al3+ for ALMT1), pH, ATP, voltage and GABA. As such, it is becoming clear that ALMTs have the potential to act as a key signalling hub in a variety of physiological processes. Following on from this study, leading on from the observed GABA modulation of ABA, H2O2 and coronatine effects on stomata, the investigation into cross-talk between GABA and other signals for ALMT9, in particular, and ALMTs, in general, provides the basis for future research areas. Such studies will be able to resolve questions such as ‘whether GABA can act directly on guard cell ALMTs?’, as appears to occur for wheat ALMT1,, or ‘whether other signalling intermediates are also involved?’. GABA inhibition of the wheat ALMT1 anion conductance was recently found to occur from the cytosol only, by reducing the open probability of the channel to anions. However, that study was unable to determine whether this occurred through permeation of uncharged GABA through the ALMT pore or through GABA binding modifying channel structure. Cytosolic GABA inhibition was dependent upon the putative GABA binding residue F213 (equivalent to F243 in ALMT9, which is also predicted to face the cytosol),. Our study therefore highlights the real need to definitively determine whether GABA binds to ALMTs or whether the identified amino acid residues affect GABA sensitivity independent of anion permeability through other means. For instance, future studies should address whether GABA permeability of ALMTs has a role in signal transduction in guard cells and the regulation of other physiological processes. These later questions would be aided by the determination of GABA concentrations in different cell types and compartments to further understand the co-ordination of GABA signalling across membranes, leaves and other organs, and this could be achieved through the deployment of novel GABA sensors, as recently used in animal tissues,.

26GABA concentration oscillates over diel cycles and increases in response to multiple abiotic and biotic stresses including drought, heat, cold, anoxia, wounding pathogen infection and salinity. ALMTs have been implicated in modulating multiple developmental and physiological processes in plants–,– including those underpinning nutrient uptake and fertilization that are affected by GABA,,. Therefore, the discovery that GABA regulates ALMT to form a physiologically relevant signalling mechanism in guard cells is likely to have broad significance beyond stomata, particularly during plant responses to environmental transitions and stress.

27GABA’s effect on stomata appears to be conserved across a large range of crops from diverse clades including important monocot and dicot crops (Supplementary Fig. 5), indicating that GABA may well be a stomatal signal of economic significance. As we find that the genetic manipulation of cell-type specific GABA metabolism can reduce water loss leading to improved drought performance, our work opens up alternative ways for manipulating crop stress resilience. This statement is tempered in the knowledge that GABA modulated stomatal signalling in the face of another signal and did not stimulate changes in stomatal aperture itself. GABA’s role appears to be that of fine-tuning stomatal aperture. Our data suggest that GABA modulated stomatal movement occurs in response to light and dark and low concentrations of signal intermediates, but in the face of a strong stress stimulus its affects may be overridden. As such, GABA may well provide a direct link between the metabolic status of the cell—GABA being produced in the cytosol in times of stress as a bypass of several reactions of the TCA cycle—to regulate and sustain a certain physiological process prior to it being shut down via a more severe stress response pathway. More broadly, this study also provides proof that GABA is a plant signalling molecule and not just a plant metabolite,, and in so doing, we conclude that GABA is an endogenous signalling molecule beyond the animal and bacterial kingdoms, enacted through distinct and organism specific mechanisms.

28All experiments were performed on A. thaliana were in the Columbia-0 (Col-0) ecotype background, unless stated. Arabidopsis wild type, T-DNA insertion mutant and other transgenic plants were germinated and grown on ½ Murashige and Skoog (MS) medium with 0.8% phytagel for 10 days before being transferred to soil for growth in short-day conditions (100–120 µmol m−2 s−1, 10 h light/14 h dark) at 22 °C. The T-DNA insertion mutant gad2-1 (GABI_474_E05) and gad2-2 (SALK_028819) were obtained from the Arabidopsis Biological Resource Centre (ABRC). gad2-1 was selected using primer sets:

29gad2_LP1 (5′-TATCACGCTAACACCTAACGC-3′), gad2_RP1 (5′-TTCAAGGTTTGTCGGTATTGG-3′) and GABI_LB (5′-GGGCTACACTGAATTGGTAGCTC-3′) for removing the second T-DNA insert; gad2_LP2 (5′-ACGTGATGGATCCAGACAAAG-3′), gad2_RP2 (5′-TCTTCATTTCCACACAAAGGC-3′) and GABI_LB for isolation of the GAD2 (At1g65960) T-DNA insertion. gad2-2 was selected using primer sets: gad2-2_LP (5′-AGTTGTATGAAAGTTCATGTGGC-3′), gad2-2_RP (5′-TCGACCACGAGATTTTAATGG -3′) and SALK_LB (5′-ATTTTGCCGATTTCGGAAC-3′). almt9-1 (SALK_055490), almt9-2 (WiscDsLox499H09)), almt12-1 (SM_3_38592) and almt12-2 (SM_3_1713) were selected as described previously,. The double mutant lines gad2-1/almt9-1, gad2-1/almt9-2, almt9-2/12-1 and almt9-2/12-2 were obtained, respectively, from crossing the respective mutants. The mesophyll enhancer-trap line JR11-2 in the Col-0 background was kindly provided by K. Baerenfaller (ETH Zurich). JR11-2 (Col-0) and gad2-1/JR11-2 were segregated from crossing gad2-1 with JR11-2. JR11-2 was selected using primer sets: JR11-2_LP (5′-TTATTTAGGGAAATTACAAGTTGC-3′), JR11-2_RP (5′-AGACACATTTAATAACATTACAACAAA-3′) and JR11-2_LB (5′-GTTGTCTAAGCGTCAATTTGTTT-3′). All experiments were performed on stable T3 transgenic plants or confirmed homozygous mutant lines. The other plants V. faba, N. benthamiana and G. max were grown in soil in long-day conditions (400 µmol m−2 s−1, 16 h light/8 h dark, 28 °C/25 °C). H. vulgare (barley) cv. Barke was grown in a hydroponic system with half-strength Hoagland’s solution in long-day conditions (150 µmol m−2 s−1, 16 h light/8 h dark, 23 °C).

30For guard cell-specific complementation, the constitutively active form of GAD2 with a truncation of the calmodulin binding domain (GAD2Δ), and the full-length GAD2 coding sequence (GAD2) was driven by a guard cell-specific promoter GC1 (−1140/+23), as designated GC1::GAD2Δ and GC1::GAD2, respectively. PCR reactions first amplified the truncated GAD2Δ with a stop codon and GC1 promoter (GC1) separately using Phusion® High-Fidelity DNA Polymerase (New England Biolabs) with the primer sets: GAD2_forward (5′-CACTACTCAAGAAATATGGTTTTGACAAAAACCGC-3′) and GAD2_truncated_reverse (5′-TTATACATTTTCCGCGATCCC-3′); GC1_forward (5′-CACCATGGTTGCAACAGAGAGGATG-3′) and GC1_reverse (5′-ATTTCTTGAGTAGTGATTTTGAAG-3′). This was followed by an overlap PCR to fuse the GC1 promoter to GAD2Δ (GC1::GAD2Δ) with the GC1_forward and GAD2_truncated_reverse primer set. The same strategy was used to amplify GC1::GAD2Δ without a stop codon (GC1::GAD2Δ-stop), GC1::GAD2 and GC1::GAD2 without a stop codon (GC1::GAD2-stop) with different primer sets: (1) GC1::GAD2Δ-stop amplified with GAD2_forward and GAD2_truncated-stop_reverse (5′-TACATTTTCCGCGATCCCT-3′); (2) GC1::GAD2 amplified with GAD2_forward and GAD2_reverse (5′-TTAGCACACACCATTCATCTTCTT-3′) and (3) GC1::GAD2-stop amplified with GAD2_forward and GAD2-stop_reverse (5′- CACACCATTCATCTTCTTCC-3′). The fused PCR products were cloned into the pENTR/D-TOPO vector via directional cloning (Invitrogen). pENTR/D-TOPO vectors containing GC1::GAD2Δ or GC1::GAD2 were recombined into a binary vector pMDC99 by an LR reaction using LR Clonase II Enzyme mix (Invitrogen) for guard cell-specific complementation, after an insertion of a NOS Terminator into this vector. A pMDC99 vector was cut by PacI (New England Biolabs) and ligated with NOS terminator flanked with PacI site using T4 DNA ligase (New England Biolabs). This NOS terminator flanked with PacI site was amplified with primer set: nos_PacI_forward (5′-TACGTTAATTAAGAATTTCCCCGAT-3′) and nos_PacI_reverse (5′-GCATTTAATTAAAGTAACATAGATGACACC-3′) and cut by restriction enzyme PacI before T4 DNA ligation. GC1::GAD2Δ-stop and GC1::GAD2-stop were recombined from the pENTR/D-TOPO vector into a pMDC107 vector that contained a GFP tag on the C-terminus (GC1::GAD2Δ-GFP and GC1::GAD2-GFP).

31To create GAD2 complementation driven by a constitutive 35S promoter, the full-length GAD2 was also amplified using primer set GAD2_forward2 (5′-CACCATGGTTTTGACAAAAACCGC-3′) and GAD2_reverse and cloned into pENTR/D-TOPO vector via directional cloning (Invitrogen), followed by an LR reaction recombinant into pMDC32. For mesophyll specific complementation, GAD2Δ with a stop codon was amplified with the GAD2_forward2 and GAD2_truncated_reverse primer set, and cloned into the pENTR/D-TOPO vector, followed by an LR reaction recombined into the pTOOL5 vector (UAS::GAD2Δ).

32For almt9-2 complementation, the pART27 binary vector containing the ALMT9 coding sequence was used for native ALMT9 complementation driven by the 35S promoter, and also used as a template for a site-direct mutagenesis PCR to replace F243 and Y245 of ALMT9 with two cysteines (ALMT9F243C/Y245C) using the primer sets: ALMT9_DoubleF (5′-GTTTAGGTGTTAATATGTGTATCTGTCCTATATGGGCTGGAGAGG-3′) and ALMT9_DoubleR (5′-CCATATAGGACAGATACACATATTAACACCTAAACTAACACCAGCACC-3′).

33For GAD2 expression analysis, a 1 kb sequence upstream of the GAD2 start codon was designated as the GAD2 promoter (pGAD2) and amplified using primer set proGAD2_F (5′-ATTTTGAATTTGCGGAGAATCT-3′) and proGAD2_R (5′-CTTTGTTTCTGTTTAGTGAAAGAGAA-3′). The pGAD2 PCR product was cloned into pCR8/GW/TOPO via TA cloning and recombined via an LR reaction into the pMDC162 vector containing the GUS reporter gene for histochemical assays. The binary vectors, pMDC32, pMDC99, pMDC107, pMDC162, pTOOL5 and pART27 carrying sequence-verified constructs, were transformed into Agrobacterium strain AGL1 for stable transformation in Arabidopsis plants.

34Soil-grown Arabidopsis (5–6-week-old) were used for stomatal aperture and density measurements. Two-to-three-week-old soybean, broad beans and barley and 5–6-week-old tobacco were used for stomatal aperture assays. Epidermal strips from Arabidopsis, soybean, faba bean and tobacco were peeled from abaxial sides of leaves, pre-incubated in stomatal pore measurement buffer containing 10 mM KCl, 5 mM L-malic acid, 10 mM 2-ethanesulfonic acid (MES) with pH 6.0 by 2-amino-2-(hydroxymethyl)-1,3-propanediol (Tris) under light (200 µmol m−2 s−1) or darkness and transferred into stomatal pore measurement buffer with blind treatments as stated in the figure legend. For barley epidermal stomatal assays, a modified method was used: the second fully expanded leaf from 2-week-old seedlings was used as experimental material, leaf samples were first detached and bathed in a modified measurement buffer (50 mM KCl, 10 mM MES with pH 6.1 by KOH) under light (150 µmol m−2 s−1) for 1.5 h or darkness for 1 h, then pre-treated in the same buffer with or without 1 mM GABA for 0.5 h; after this pre-treatment, samples were incubated in continuous dark, light, light-to-dark or dark-to-light transition for an additional 1 h as indicated in the figure legend before leaf epidermal strips were peeled for imaging. For Arabidopsis stomatal density measurement, epidermal strips were peeled from abaxial sides of young and mature leaves, three leaves per plants, three plants per genotype. Epidermal strips for both aperture and density measurement were imaged using an Axiophot Pol Photomicroscope (Carl Zeiss) apart from the barley epidermal strips imaged using an Nikon Diaphot 200 Inverted Phase Contrast Microscope (Nikon). Stomatal aperture and density were analyzed using particle analysis (http://rsbweb.nih.gov/ij/).

35All stomatal conductance measurements were performed on 5–6-week-old Arabidopsis plants. The stomatal conductance determined by the AP4 Porometer (Delta-T Devices) was calculated based on the mean value from 2–3 leaf recordings per plant (Figs. 3b, 4c, 5c, 6c, d and 9; Supplementary Fig. 10d, 11d, g and 14c). The time-dependent stomatal conductance, transpiration and photosynthetic rate was recorded using LI-6400XT Portable Photosynthesis System (LI-COR Biosciences) equipped with an Arabidopsis leaf chamber fluorometer (under 150 µmol m−2 s−1 light with 10% blue light, 150 mmol s−1 flow rate, 400 ppm CO2 mixer, ~50 % relative humidity at 22 °C) as indicated (Fig. 4d; Supplementary Fig. 8g, h).

36The analysis of Arabidopsis leaf ABA concentration followed a method as described previously. Briefly, >50 mg of ground fresh leaf samples were used to determine ABA concentration using an Agilent 6410 Series Triple Quad liquid chromatography (LC)-mass spectrometer (MS)/MS, equipped with Agilent 1200 series HPLC (Agilent Technologies) using a Phenomenex C18 column (75 mm × 4.5 mm × 5 μm) with a column temperature set at 40 °C. Solvents were nanopure water and acetonitrile, both with 0.05% acetic acid. Samples were eluted with a linear 15-min gradient from 10 to 90% acetonitrile. Compounds were identified by retention times and mass/charge ratio.

37Plants were germinated on ½ MS medium with 0.8% phytagel for 10 days in short-day conditions (100–120 µmol m−2 s−1, 10 h light/14 h dark) at 22 °C before being transferred to pots (size 2.5 inch diameter × 2.25 inch height, LI-COR Bioscience) with soil, containing coco peat/Irish peat (1:1 ratio). Prior to 10-day-old seedling transfer, all pots were weighed on an Ohaus ARA520 Adventurer Balance and soil was aliquoted into the pots within ±0.1 g between all replicates within an experimental run, randomly placed in growth cabinet and moved every other day in the same environmental conditions stated above. The starting weight varied amongst experimental runs dependent upon soil moisture (from 75 to 78 g). The drought assay was performed on 5–6-week-old Arabidopsis plants (Figs. 3, 4g, h and 5; Supplementary Figs. 6a–c, 9 and 11c–e). All plants were well-watered (saturated) the night before the drought assay, but not watered again during the assay. During the drought assay, all plants were randomly moved around once a day to avoid any bias of uneven light distribution or air flow within the cabinet that may differentially affect water loss.

38At each sampling point, fresh weight of 2–3 leaves per plant was determined on an Ohaus Explorer E02140 balance (in Fig. 3a, Supplementary Figs. 6c and 11c, this occurred immediately after the rest of the leaf rosette was snap frozen in liquid nitrogen for later GABA measurement). Sampled leaves were then rehydrated to full turgid weight in ultrapure water overnight and measured after surface water was dried with paper towel. Dry weight was determined at 65 °C for 1 day. Leaf RWC was calculated as (Figs. 3c, 4g and 5d; Supplementary Figs. 6a and 11e)At each sampling point, fresh soil weight of the whole pot (Mwet) and dry soil weight after drying the soil (Mdry) at 105 °C for 3 days was measured using an Ohaus ARA520 Adventurer Balance (Supplementary Fig. 6b). Gravimetric soil water content (θg) of the whole soil in the pots was calculated as

39The stomatal conductance, transpiration and photosynthetic rate of the detached leaf feeding assay was recorded using either a LCpro-SD Portable Photosynthesis System (ADC Bioscientific) with 350 µmol m−2 s−1 light, 200 μmols s−1 flow rate and 400 ppm CO2 at 22 °C (Supplementary Fig. 1c, d) or LI-COR LI-6400XT (LI-COR Biosciences) with 200 µmol m−2 s−1 light, 150 µmols s−1 flow rate and 400 ppm CO2 at 22 °C (Figs. 2c and 8b, d, f; Supplementary Figs. 2a, b and 17a, b, d, e, g, h). The detached leaf was fed with artificial xylem sap solution modified as described, containing 1 mM KH2PO4, 1 mM K2HPO4, 1 mM CaCl2, 0.1 mM MgSO4, 1 mM KNO3, 0.1 mM MnSO4, 1 mM K-H-malate, pH 6.0 (KOH) with or without GABA or muscimol supplement as indicated, detached leaves were pre-fed under 150 µmol m−2 s−1 light to allow the uptake of treatments for 45–60 min before recording. iWUE and WUE were calculated based on the equation as described in ref..

40GABA concentration was determined using ultra performance LC (UPLC) as described previously. Briefly, GABA was extracted from samples using 10 mM sodium acetate and derivatized with the AccQ Tag Ultra Derivatization Kit (Waters). Chromatographic analysis of GABA was performed on an Acquity UPLC System (Waters) with a Cortecs or Phenomenex UPLC C18 column (1.6 μm, 2.1 × 100 mm). The gradient protocol for amino acids analysis was used to measure GABA with mobile solvents AccQ Tag Ultra Eluents A and B (Waters). Standard GABA solution was used for calibration ranging from 0 to 150 µM. The results were analyzed by Empower chromatography software version 3 (Waters).

41A GUS histochemical assay was performed using the methods described previously. Three-to-four-week-old transgenic pGAD2::GUS plants were stained in buffer containing 50 mM Na phosphate pH = 7.0, 10 mM EDTA, 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide, 0.1% (v/v) Triton X-100 and 0.1% (w/v) X-Gluc (5-bromo-4-chloro-3-indolyl β-D-glucuronide) during a 1.5 h incubation at 37 °C in the dark. The stained plants were destained in 70% ethanol. GUS-stained plants were imaged using an Axiophot Pol Photomicroscope (Carl Zeiss).

42The fluorescence of fluorescent proteins in transgenic gad2-1/GC1::GAD2Δ-GFP, gad2-1/GC1::GAD2-GFP and WT/GC1::GAD2Δ-GFP plants was imaged by confocal laser scanning microscopy using a Zeiss Axioskop 2 mot plus LSM5 PASCAL and argon laser (Carl Zeiss). Sequential scanning and laser excitation was used to capture fluorescence via the LSM5 PASCAL from GFP (excitation = 488 nm, emission band-pass = 505–530 nm), chlorophyll autofluorescence (excitation = 543 nm, emission long-pass = 560 nm). The fluorescence of fluorescent proteins in the mesophyll protoplasts of transgenic almt9-2 complementation lines and N. benthamiana (Supplementary Fig 16c, d) was imaged using Nikon A1R Laser Scanning Confocal with DS-Ri1 CCD camera. Sequential scanning and laser excitation was used to capture fluorescence via the Nikon A1R Laser Scanning Confocal from GFP (excitation = 488 nm, emission = 525–575 nm), chlorophyll autofluorescence (excitation = 561 nm, emission = 595–645 nm).

43Reverse transcriptional PCR was determined by PCR amplification on cDNA synthesized from RNA extracted from plants as indicated. PCR amplified GAD2, Actin2, GFP, GAD2 mRNA, UAS::GAD2Δ and ALMT9 using Phire Hot Start II DNA Polymerase (Invitrogen) with primer sets:

44GAD2_rt_F (5′-ACGTGATGGATCCAGACAAAG-3′) and

45GAD2_rt-R (5′-TACATTTTCCGCGATCCCT-3′);

46Actin2_rt_F (5′-CAAAGGCCAACAGAGAGAAGA-3′) and

47Actin2_rt_R (5′-CTGTACTTCCTTTCAGGTGGTG-3′);

48GFP_rt_F (5′-GGAGTTGTCCCAATTCTTGTT-3′) and

49GFP_rt_R (5′-CGCCAATTGGAGTATTTTGT-3′);

50GAD2mRNA_rt_F (5′-ACGTGATGGATCCAGACAAAG-3′) and GAD2mRNA_rt_R (5′-TCTTCATTTCCACACAAAGGC-3′);

51UAS_GAD2_rt_F (5′-TCACTCTCAATTTCTCCAAGG-3′) and

52UAS_GAD2_rt_R (5′-CGGCAACAGGATTCAATCTTAAG-3′);

53ALMT9_rt_F (5′-AATACTCGAGAAACGGGAGAG-3′) and

54ALMT9_rt_R (5′-CATCCCAAAACACCTACGAAT-3′).

55Quantitative reverse transcription PCR was performed using KAPA SYBR FAST ABI PRISM kit (Kapa Biosystems) using a QuantStudioTM 12K Flex Real-Time PCR System (Thermo Fisher Scientific) to determine the expression levels of GAD1, GAD2, GAD3, GAD4, GAD5, GABA-T, ALMT9, ALMT12, RD29A and RD22 genes with primer sets:

56GAD1_qF (5′-TCTCAAAGGACGAGGGAGTG-3′) and

57GAD1_qR (5′-AACCACACGAAGAACAGTGATG-3′);

58GAD2_qF (5′-GTCTCAAAGGACCAAGGAGTG-3′) and

59GAD2_qR (5′-CATCGGCAGGCATAGTGTAA-3′);

60GAD3_qF (5′-CCGTTAGTGGCGTTTTCTCT-3′) and

61GAD3_qR (5′-TCTCTTTGCGTCTCCTCTGG-3′);

62GAD4_qF (5′-GTGTTCCGTTAGTGGCGTT-3′) and

63GAD4_qR (5′GTCTCCTCTGGCGTCTTCTT-3′);

64GAD5_qF (5′-TCAACCCACTTTCACTCTCA-3′) and

65GAD5_qR (5′-TTCCTTCTCTTAGCCTCCTT-3′);

66GABA-T_qF (5′-AGGCAGCACCTGAGAAGAAA-3′) and

67GABA-T_qR (5′-GGAGTGATAAAACGGCAAGG-3′);

68ALMT9_qF (5′-CAGAGAGTGGGCGTAGAAGG-3′) and

69ALMT9_qR (5′-GGATTTGAAGGCGTAGATTGG-3′);

70ALMT12_qF (5′-TTGACGGAACTCGCAGATAG-3′) and

71ALMT12_qR (5′-CGATGGAGGTTAGAGCCAAG-3′);

72RD29A_qF (5′-AAACGACGACAAAGGAAGTG-3′) and

73RD29A_qR (5′-ACCAAACCAGCCAGATGATT-3′);

74RD22_qF (5′-AGGGCTGTTTCCACTGAGG-3′) and

75RD22_qR (5′- CACCACAGATTTATCGTCAGACA-3′).

76Expression levels of each gene was normalised to three control genes—Actin2, EF1α and GAPDH-A—that were amplified with primer sets:

77Actin2_qF (5′-TGAGCAAAGAAATCACAGCACT-3′) and

78Actin2_qR (5′-CCTGGACCTGCCTCATCATAC-3′);

79EF1α_qF (5′-GACAGGCGTTCTGGTAAGGAG-3′) and

80EF1α_qR (5′-GCGGAAAGAGTTTTGATGTTCA-3′);

81GAPDH-A_qF (5′-TGGTTGATCTCGTTGTGCAGGTCTC-3′) and

82GAPDH-A_qR (5′-GTCAGCCAAGTCAACAACTCTCTG-3′).

83Further information on research design is available in the Nature Research Reporting Summary linked to this article.