Next-generation ABACUS biosensors reveal cellular ABA dynamics driving root growth at low aerial humidity

The plant hormone abscisic acid (ABA) accumulates under abiotic stress to recast water relations and development. To overcome a lack of high-resolution sensitive reporters, we developed ABACUS2s—next-generation Förster resonance energy transfer (FRET) biosensors for ABA with high affinity, signal-to-noise ratio and orthogonality—that reveal endogenous ABA patterns in Arabidopsis thaliana. We mapped stress-induced ABA dynamics in high resolution to reveal the cellular basis for local and systemic ABA functions. At reduced foliar humidity, root cells accumulated ABA in the elongation zone, the site of phloem-transported ABA unloading. Phloem ABA and root ABA signalling were both essential to maintain root growth at low humidity. ABA coordinates a root response to foliar stresses, enabling plants to maintain foraging of deeper soil for water uptake.

Plant decision-making is distributed rather than centrally coordinated, but to survive and overcome stresses such as lack of water, responses must also be systemically coordinated. Abscisic acid (ABA) is a phytohormone that accumulates systemically under various local water stresses to coordinate responses over a complex and often-large morphology 1 . When roots experience low-water stress, for example, ABA closes the microscopic pores on leaves (stomata) to limit systemic water loss [2][3][4] . Interestingly, leaf water loss can cause changes in root growth responses and architecture: increasing transpiration genetically or through increased airflow produces larger root systems in Arabidopsis 5 and low relative humidity (RH) can promote root growth in many species [6][7][8] . Although a molecular mechanism remains elusive, it has been proposed that ABA, acting as a dehydration signal, could be coordinating these root growth responses 5,9 . The sites of ABA biosynthesis, metabolism and translocation are the subject of intensive research, but progress has been hampered by limitations in tools to quantify accumulation and depletion of ABA on a tissue/cellular scale where regulatory decisions controlling ABA dynamics are made 1,10 . The availability of sensitive reporters, particularly Förster resonance energy transfer (FRET) biosensors, for hormones, second messengers and metabolism is revolutionizing plant development, signalling and photosynthesis research 11 . Such biosensors are powerful tools to quantify metabolites in vivo at high spatiotemporal resolution 11 , including phytohormones under changing environmental conditions [12][13][14][15][16] . Direct ABA FRET biosensors 13,14 that do not require additional signalling components have broad application potential beyond ABA quantification in plant cells and subcellular compartments; for example, in ABA synthesizing pathogenic fungi 17 , in human granulocytes where ABA is a cytokine 18 , or in extracts from organisms where genetic modification is difficult using purified protein in vitro 19 . However, existing ABA FRET biosensors, ABAleons and Abscisic Acid Concentration and Uptake Sensors 1 (ABACUS1s) 13,14,20 lack the full complement of strengths in terms of the signal-to-noise ratio or affinity required to easily quantify ABA. Therefore, we engineered next-generation ABA biosensors and Article https://doi.org/10.1038/s41477-023-01447-4 of the PYL1 A190V mutant in purified protein assays (ABACUS1-2μ-ii; Fig. 1b, Extended Data Fig. 1 and Supplementary Data Tables 1 and 2).
We next incorrectly predicted that truncating the flexible fluorescent protein termini facing the sensory domain (edCitrine residues 1-229: edCitrineT9, edCerulean residues 8-238: T7edCerulean) would be sufficient to increase ratio change further (ABACUS1-2μ-iv; Fig. 1b, Extended Data Fig. 1 and Supplementary Data Table 1). Nonetheless, emission ratio change could be restored along with further affinity improvements by introducing either of two separate mutations (R143S, E141D) to a PYL1 region-the 'latch'-that is important for both PYL1-ABA and PYL1-PP2C interactions 25 .
The first mutation, PYL1 E141D, inspired by sequences of the high-affinity PYL8 and PYL9 ABA receptors, produced a high ratio-change sensor with our highest affinity, which we named ABA-CUS2-100n (K D (ABA): 98 nM, in vitro emission ratio change: +67%; Fig. 1b,d,e and Extended Data Fig. 1). The sidechain of PYL1 E141 faces out of the ABA binding pocket (Extended Data Fig. 2 and Supplementary Fig. 2), suggesting that the high-affinity mutation could affect the accessibility of the pocket for ABA and cause faster ABA binding rather than strengthening the interaction between the pocket and ABA. Alternatively, PYL1 E141D could strengthen the PYL1 and ABI1aid interdomain interaction after ABA binding, thereby causing slower release of ABA. Introducing E141D into ABACUS1-2μ or ABACUS1-2μiii, which do not contain fluorescent protein truncations, did not match the affinity or ratio change of ABACUS2-100n (Supplementary Table 2). This may indicate that shortening and rigidifying the linkers between the sensory domain and the FRET pair contributed to affinity and ratio change improvements. PYL1 R143S produced our highest ratio-change biosensor that has an ABA sensitivity suitable for in planta studies, which we named ABACUS2-400n (K D (ABA): 445 nM, in vitro emission ratio change: +71%;  Table 1). The sidechain of PYL1 R143 also faces out of the ABA binding pocket (Extended Data Fig. 2 and Supplementary Fig. 3). However, its backbone forms a water-mediated interaction with ABA, PYL1 P115 and ABI1 W300 deployed them to dissect cellular ABA dynamics and mobilization in response to foliar humidity stress, and to establish a systemic role for ABA in maintaining local root growth in response to a distant shoot stress.

Results
In ABAleons and ABACUS1 biosensors, ABA sensory domains are connected by linkers to a pair of fluorescent proteins 13,14 (FP) (Supplementary Fig. 1). The orientation and distance between these FPs determine the transfer of excitation energy via FRET from a donor FP to an acceptor FP 21 . Ligand-induced conformational changes in sensory domains alter the relative positions of the FPs, which can be detected by exciting the donor and measuring a change in relative acceptor and donor emissions, hereafter referred to as emission ratio change.
ABAleons are negative ratio change biosensors that are sensitive to endogenous ABA concentrations, but have poor signal-to-noise ratios (small emission ratio change) 13,20 . ABACUS1s have a positive ratio change with high signal-to-noise ratio but poor sensitivity for endogenous ABA 14 . Ideal biosensors are also orthogonal, with minimal interaction with endogenous signalling and vice versa. ABAleons have strong ABA hyposensitivity phenotypes, while ABACUS1s have minor ABA hypersensitivity phenotypes 13,14,20 . We used ABACUS1-2μ as the basis to engineer next-generation biosensors with high sensitivity, emission ratio change and orthogonality (Extended Data Fig. 1), screening dozens of ABACUS variants in yeast lysate or as purified proteins (Supplementary Data Tables 1 and 2). ABACUS1-2μ has a dissociation constant for ABA (K D ) of ~1.1-1.8 μM (ref. 14) and consists of an N-terminal FRET acceptor (edCitrine), an attB1 linker, a sensory domain consisting of a mutated pyrabactin resistant 1 like 1 (PYL1 H87P) ABA receptor, an L52 linker, a truncated protein phosphatase 2C (PP2C) co-receptor, abscisic acid insensitive 1 aba interacting domain (ABI1aid), an attB2 linker and a C-terminal FRET donor (edCerulean) (Extended Data Fig. 1) 14 . We introduced a binding site mutation into ABACUS1-2μ (PYL1 A190V (residue numbering according to position in wildtype sequence)) that is known to increase the ABA affinity of PYL1 (ref. 22). The resulting ABACUS1-2μ-i had increased affinity but reduced emission ratio change in vitro ( Fig. 1a,b, Extended Data Fig. 1 and Supplementary Data Table 1).
Alphafold2 predicts nlsABACUS2-100n and nlsABACUS2-400n structures with pockets that could still accommodate ABA (Extended Data Fig. 2). When compared with the structure of wildtype PYL1 bound to ABA 26 , the largest changes are at the pocket entrance, which is also the binding interface between PYL1 and ABI1aid moieties (Extended Data Fig. 2). Nonetheless, due to the limitations of Alphafold2 in predicting the effects of individual mutations and ligand-binding dynamics, we cannot yet discriminate which aspects of sensor behaviour are improved in these successful biosensors. For example, it remains unclear how the PYL1 E141D and R143S 'latch' mutations in combination with linker changes affect on-and off-rates for PYL1-ABA and PYL1-ABA-ABI1aid interactions.
Similar to ABACUS1 (ref. 14), in vitro assays against other phytohormones, salts and ABA-related compounds demonstrated that ABACUS2-100n is highly specific for ABA and the ABA agonist pyrabactin, although with a smaller ratio change for the latter ( Supplementary  Fig. 4).
Previously, severe silencing prevented us from obtaining strongly fluorescent plants expressing ABACUS1 biosensors under the control of a 641 bp UBQ10 promoter in the Arabidopsis thaliana wildtype background Col-0 (ref. 14). Switching to the p16 promoter, previously found to improve expression of nlsGPS1 biosensors 12 , allowed us to obtain fluorescent Col-0 plants expressing the nlsABACUS1-2μ biosensor 14 (Supplementary Table 3). Addition of nuclear localization signals (nls) to these sensors allowed easy discrimination of the fluorescence of neighbouring cells and the exclusion of non-nuclear background and autofluorescence during image processing 11 . After screening constitutive promoters for expression of nlsABACUS2 in Nicotiana benthamiana transient expression assays ( Supplementary Fig. 5), we selected the 1,500 bp UBQ10 promoter 27 and obtained strongly fluorescent Arabidopsis Col-0 plants expressing nlsABACUS2 biosensors (Supplementary Table 3). To accelerate ratiometric image processing for these and other nuclear targeted FRET biosensors, we developed an improved and no-cost comprehensive image analysis toolset to quickly analyse confocal stacks in three-dimensional (3D)/4D, allowing users to robustly quantify and visualize nuclear emission ratios (FRETENATOR 1.5; Supplementary Methods and ref. 28).
In Col-0, nlsABACUS2-400n and nlsABACUS2-100n respond strongly to exogenous ABA, saturating at lower concentrations than nlsABACUS1-2μ ( Fig. 1f and Extended Data Fig. 3), consistent with their improved sensitivity. The ABACUS2 emission ratio changes are significantly larger than those of ABACUS1-2μ or other ABA sensors (ABAleonSD1-3L21) ( Fig. 1f and Supplementary Data Fig. 6) 13 . Interestingly, both sensors showed a saturated response at lower concentrations of exogenous ABA than they do in vitro, perhaps indicating that the ABA import mechanisms in these cells are concentrative at these ABA treatment levels.
To determine how the 5-25-fold higher affinity of nlsABACUS2 with orthogonalizing mutations affected ABA responses, we examined germination, lateral root development and primary root growth phenotypes known to be sensitive to ABA (Fig. 1g,h, Extended Data Fig. 4 and Supplementary Fig. 7). Without exogenous ABA, nlsABACUS2-400n lines germinate normally and nlsABACUS2-100n lines are delayed; however, all nlsABACUS2 lines display hypersensitive germination inhibition by exogenous ABA (Fig. 1g and Supplementary Fig. 7). If germination time is synchronized, primary root growth is normal in most ABACUS2 lines without exogenous ABA (Extended Data Fig. 4). However, primary root growth is hypersensitive to exogenous ABA for 3 and 6 d for nlsABACUS2-100 line 7 (Extended Data Fig. 4) and lateral root number is hypersensitive to exogenous ABA in all lines except nlsABACUS2-400n line 1 (Fig. 1h). Together, these ABA hypersensitivity phenotypes suggest that the ABACUS2 PYL1s remain somewhat active in planta despite the PYL1 S112A orthogonalizing mutation. The milder phenotypes of ABACUS2-400n expressing lines, particularly without exogenous ABA, are expected owing to their lower-affinity PYL moiety. For future investigations, using nlsABACUS2-400n (particularly line 1) may be preferable to using nlsABACUS2-100n lines if phenotypes are a concern, as long as the relevant ABA dynamics fall in the appropriate detection range.
As for ABACUS1-2μ, both ABACUS2 sensors were reversible in vitro and in planta, indicating that they can be used to track ABA accumulation and depletion (Fig. 2a,b, Extended Data Fig. 5 and Supplementary  Fig. 8).
The availability of sensitive reporters for other phytohormones such as auxin revolutionized plant developmental biology by revealing localized activity of a key hormone for morphogenesis and patterning 11 . Similarly, sites of ABA accumulation may give insights into developmental regulation and stress responses. Therefore, we used nlsABACUS2s to determine the distribution of ABA in whole Arabidopsis seedlings at the cellular scale ( Fig. 2f-i, Supplementary Fig. 12 and Extended Data Fig. 6). Untreated nlsABACUS2 seedlings had high emission ratios in the root meristem and elongation zones and low emission ratios in the mature root (Extended Data Fig. 6), differences that were not apparent with the first-generation ABA sensors, possibly due to differences in affinity, localization, signal-to-noise ratio or experimental conditions 14,30 .
We were initially surprised that cotyledon emission ratios were not higher, as mass spectroscopy data indicate that leaves and aerial organs contain more ABA than roots in many species 9,31 . Due to the imaging modality, epidermal cells make up the bulk of segmented nuclei in our whole-plant images, but internal tissues have higher emission ratios, indicating higher ABA levels ( Supplementary Fig. 13). Using PP11, an inert compound that reduces optical scattering by filling mesophyll air-spaces 32 , to image deep into cotyledons revealed that mesophyll cells have higher emission ratios than the epidermis, and vascular strands have very high emission ratios (Extended Data Fig. 7).
High ABA in the shoot vasculature is notable, as the phloem companion cells are a key site for ABA biosynthesis 33,34 and ABA is thought to be transported in the phloem 9 . The phloem transports sugars, hormones and other metabolites from shoot to root, where it can be unloaded via the phloem-pole pericycle cells in the root elongation zone from two distinct vascular poles 35 . nlsABACUS2 roots show high emission ratios in the elongation zone and vascular tissues (Fig. 2h, Supplementary Fig. 14 and Extended Data Fig. 6). We used single-plane illumination microscopy (SPIM) to examine whether phloem-sourced ABA could be unloaded in the elongation zone ( Fig. 3a,b and Supplementary Fig. 15). In untreated roots, nlsABACUS2-400n emission ratios were higher in two poles of the vascular tissues, as would be predicted for a phloem-transported hormone (Fig. 3b). Root emission ratios increased in the elongation zone and transition zone following shoot ABA treatment ( Fig. 3a,b). The lack of mature root emission ratio increases is consistent with elongation zone ABA accumulation being sourced from the vasculature and these ABA dynamics match those of shoot-applied fluorescent dyes that are unloaded from the phloem 35 .
Exogenous ABA causes concentration-dependent promotion or inhibition of root growth 36 , so ABA from the phloem must be tightly regulated independently of local biosynthesis. The abscisic acid 8′-hydroxylases CYP707A1-4 catabolic enzymes have been implicated in eliminating ABA after stress 37,38 . CYP707A1 and CYP707A3 are the isoforms most expressed in the root 29 and cyp707a1cyp707a3 double mutants 39 displayed a strong over-accumulation of ABA in the root tip ( Supplementary Fig. 16). Exogenous ABA pulsing using the RootChip microfluidics system 12,40 revealed larger emission ratio increases in cyp707a1cyp707a3 and a larger elimination half-life (Fig. 3c). While these enzymes are critical in preventing over-accumulation of ABA in the root tip, other ABA depletion mechanisms must also contribute to the ABA elimination as there is still a gradual reduction in cyp707a-1cyp707a3 nlsABACUS2-400n emission ratios following an ABA pulse (Fig. 3c).
ABA has numerous roles in protecting plants from abiotic stress, particularly osmotic and ionic stresses. During salt stress, root ABA responses mediate endodermal cell wall suberization 41,42 , limiting ion and water flow to protect the plant; however, it is currently unclear which cells accumulate ABA. High-resolution imaging of ABACUS2s gave us an unparalleled view of the ABA accumulation after a 6 h 100 mM NaCl stress (Fig. 3d,e and Extended Data Fig. 8), allowing us Article https://doi.org/10.1038/s41477-023-01447-4 to quantify which tissues accumulate ABA. Under salt stress, the stele (a site of ABA biosynthesis) and endodermis (a site of ABA-dependent protective responses) of the differentiation/maturation zones accumulate ABA to a higher concentration than the surrounding epidermis and cortex tissues (Extended Data Fig. 8).
Confident that we could image and detect cell type-specific ABA accumulations, we decided to investigate the effect of humidity on plant ABA levels and responses in detail. A 6 h humidity drop increased emission ratios in stomata and pavement cells expressing nlsABA-CUS2-400n (two-way ANOVA: Humidity factor P = 0.0165; Fig. 3f and Extended Data Fig. 9), which coincided with a decreased stomatal aperture (Extended Data Fig. 9). Leaf humidity increases trigger expression of ABA catabolic genes CYP707A1 and CYP707A3 (ref. 38) and nlsABACUS2-400n emission ratios decreased following a humidity increase (two-way ANOVA: Humidity factor P < 0.0001; Fig. 3g) and stomata opened (Extended Data Fig. 10). Remarkably, nlsABACUS2-400n emission ratios responded to humidity changes in both pavement cells and stomatal cells (Fig. 3f,g and Extended Data Figs. 9 and 10). ABA famously closes stomata and along with the vasculature, stomata have been proposed as sites of ABA biosynthesis 33,38,43 , but little attention has been paid to whether pavement cells accumulate ABA. Such broad ABA increases may indicate a systemic response that travels beyond the tissues responsible for fast local responses.
As foliar ABA levels increase following a humidity stress and foliar ABA can be transported to the root (Fig. 3a,b) 8, 9, 9, 9, 9, 9 biologically independent roots from left to right). A Holm-Sidak post hoc test was used for multiple comparisons. root growth and development. Leaf transpiration rates can affect root growth and morphology through an uncharacterized mechanism 5 ; however, root plasticity is strongly ABA regulated under salt and other local water stresses 44,45 . We developed a system where leaves could be exposed to low humidity and roots would remain hydrated ( Supplementary Fig. 17) and maintain robust growth (Fig. 4a). Remarkably, the ABA biosynthesis mutant aba2 suffered a strong root growth inhibition under low humidity (Fig. 4a), implying that ABA signalling functions to maintain root growth when foliar humidity is low, a scenario common in irrigation agriculture.
nlsABACUS2-100n roots displayed increased root emission ratios at low humidity, which were particularly prevalent in the elongation zone, the site of phloem unloading and a tissue critical for root growth (Fig. 4b,c). We took a targeted genetic approach to determine whether increases in root ABA are critical for plants to increase/maintain root growth at low humidity. ABA responses rely on the activity of the SnRK2 kinases SnRK2.2, SnRK2. 3 (Fig. 4d). Complementation of the snrk2.2snrk2.3 mutant specifically in the root meristem with RCH1pro::SnRK2.2 (ref. 50) allowed plants to maintain root growth under a humidity stress, indicating that local ABA signalling is required to regulate root growth as humidity varies (Fig. 4d).
ABA synthesized in the phloem companion cells 33 is likely to be transported to phloem sinks including the root elongation zone 35 . We posited that the root induction of ABA accumulation at low foliar humidity might be phloem sourced so we performed targeted ABA depletions by controlled induction of CYP707A3 overexpression (Fig. 4e). Regardless of whether ectopic ABA depletion was restricted to phloem-loading companion cells (SUC2pro::XVE>>CYP707A3) or was ubiquitous (UBQ10pro::XVE>>CYP707A3), root growth was inhibited at low foliar humidity (Fig. 4e). Taken together, our results indicate that phloem ABA and root tip ABA signalling regulate root growth during a distal humidity stress in leaves.

Discussion
A series of local and systemic responses are required for plants to respond to varying water availability. Phenotypic data suggest that plant roots can respond to local osmotic differences through ABA, for example, by growing towards water (hydrotropism) 50 , but determining whether ABA levels vary across cells in a tissue has been experimentally challenging 14,20,30 . nlsABACUS2 allows ABA patterns across whole plants to be quantified at cellular resolution. nlsABACUS2 indicates that ABA is relatively low in the cotyledon epidermis, which agrees with ABAleon imaging 30 . The segmentation of nuclear localized ABACUS2 allows easier quantification of sub-epidermal tissues with reduced autofluorescence artefacts, revealing high ABA in the mesophyll and vasculature of cotyledons. ABACUS2 sensors also reproduce the rootward transport of shoot-applied ABA, as previously demonstrated with ABAleon 13 . For basal ABA levels in roots, ABACUS1 biosensors did not indicate strong celullar patterns, presumably due to their relatively low affinity. Although ABAleons and ABACUS2 biosensors both have higher affinity, the basal ABA patterns for the two sensor families differ markedly in roots. ABAleon biosensors imply high ABA at the root-hypocotyl junction 13,30 , which decreases gradually rootward. nlsABACUS2 biosensors indicate low ABA at the root-hypocotyl junction, which increases rootward to a maximum at the meristem and elongation zone. These contrasts are potentially due to differences in biosensor properties (for example, signal-to-noise ratio and brightness), image acquisition and analysis (for example, segmentation of confocal images of nuclear targeted biosensors reduces autofluorescence artefacts), experimental conditions or developmental staging.
The high affinity, signal-to-noise ratio and nuclear localization of ABACUS2 also allowed endogenous ABA increases and decreases to be quantified at the cellular scale in shoots and roots. In a parallel study 51 , a local increase in root ABA in response to root growth through air spaces without an increase in foliar ABA levels was reported 51 . Similarly, we have shown that salt stress induces ABA accumulation in the tissues where a protective response is required-the root endodermis. However, plant roots can also induce systemic ABA accumulation. During soil drying, both sulfate and CLE25 peptides can be transported from the root to induce foliar ABA accumulation, closing stomata and limiting water loss 2-4 . During drought, some of this shoot-derived ABA is also transported down to the root to promote and maintain root growth, allowing more access to soil water 9 . That foliar tissues can sense water loss has long been known, as plants quickly regulate their stomatal aperture in response to an increased vapour pressure deficit-a process enhanced by foliar ABA accumulation 13,52 . Here we show, with cellular resolution afforded by nlsABACUS2 biosensors, that foliar drying can also regulate root ABA accumulation and that this root ABA is important to maintain root growth under stress. This demonstrates that the root and shoot can each systemically regulate each other's responses to stresses that may only be experienced locally, providing a robust system to maintain plant water status.

Data visualization and statistical analysis
Unless otherwise stated, data were processed as Pandas dataframes in Python, using statsmodel/Graphpad Prism for statistics and Seaborn/ matplotlib/Excel for plotting. All statistical tests are described in the figure legends, along with sample size. Statistics tables are provided in Supplementary Table 5.

Generation of ABACUS affinity and orthogonality variants
Single amino acid mutations of the PYL1 domain of ABACUS1 in the pDRFLIP38-ABACUS1-2μ vector 14 were performed using the QuikChange II XL (Agilent) site-directed mutagenesis kit according to manufacturer instructions. All primers used for site-directed mutagenesis are listed in Supplementary Table 4.

Generation of ABACUS ratio-change variants
The edCitrine present in ABACUS1 variants 14 was exchanged with a codon-diversified version for optimal expression in yeast and to allow PCR-based cloning methods. The synthetic DNA fragment containing the codon-diversified edCitrine was introduced in the ABACUS yeast expression vectors using the In-Fusion kit (Takara Bio) according to manufacturer instructions.
The poly-proline screen variants, which included substitution of the attB1 and attB2 linkers of ABACUS1 with 1-3 proline residues, and the fluorescent protein truncations were obtained using the In-Fusion kit according to manufacturer instructions. All primers used for In-Fusion cloning are listed in Supplementary Table 4.

Fluorescence analysis and titration with (+)-ABA of protein-purified cell lysate
Yeast cell cultures (optical density (OD) 600 ≈ 0.6) containing yeast expression vector pDRFLIP38-ABACUS1-2μ or variants were centrifuged at 4,000 g for 10 min, washed once in 1 ml 50 mM MOPS buffer (pH 7.4), transferred to 1.5 ml micro-centrifuge tubes and centrifuged again at 10,000 g for 1 min. The supernatant was discarded and 1 ml of chilled glass bead slurry (50 mM MOPS pH 7.4, 0.1% Triton X-100 and 50% v/v 0.5 mm zirconia/silica beads (Thistle Scientific)) was added to the yeast pellet inside each tube. The tubes were then vortexed at maximum power at 4 °C for 5 min. The tubes were then centrifuged at 14,000 revolutions per minute at 4 °C for 10 min. The supernatant was transferred to previously prepared HisPur cobalt spin columns (0.2 ml; Thermo Fisher). Protein purification was performed following manufacturer instructions. The subsequent first elution from the purification column was diluted in 50 mM MOPS solution. The tubes were briefly vortexed and 100 μl of diluted eluate was transferred to 96-well flat-bottom clear microplates (Greiner). A serial dilution of (+)-ABA (Cayman Chemical) was made using a 4.5 mM stock solution in ethanol and sequentially diluting it in 50 mM MOPS solution. A 50 μl volume of each (+)-ABA dilution was added to 100 μl of sensor eluate. The sample's fluorescence emission was recorded using a SpectraMax i3x microplate reader (Molecular Devices), scanning from 470 to 550 nm after excitation at 430 nm with a bandwidth of 5 nm. Ratio was calculated by dividing emission at 525-535 nm by emission at 480-490 nm. The data produced were analysed using GraphPad Prism to determine the K D and ratio change of each sensor, assuming the Hill function with a single binding site.

Reversibility testing in vitro
ABACUS2 protein was purified using HisPur cobalt spin columns as before, then loaded onto Zeba spin desalting columns with mock or 100 μM ABA, according to manufacturer instructions. On the column, two washes were performed before eluting. Post elution, protein was treated with mock or 100 μM ABA and fluorescence emission was recorded on the SpectraMax i3x as above.

Structure prediction
nlsABACUS2-100n and nlsABACUS2-400n structures were predicted (for illustrative purposes only) using the ColabFold 1.5 notebook, based on Alphafold2, using MMseqs2 for homology detection and multiple sequence alignment pairing 53 . Of the five highest-ranked predictions by pLDDT, the prediction with the best PAE scores for PYL1 and ABI1aid were used. Structural validation and confidence measures are shown in Supplementary Figs. 2 and 3.

Plant transformation
Arabidopsis thaliana plants (Columbia, Col-0 background) were transformed by the floral dip method 58 and successful transformants were identified by FAST RED screening 59 or hygromycin selection. Full details of the Arabidopsis germplasm are available in Supplementary Table 3.

Salt treatment
Seeds were surface sterilized with 96% ethanol, then sown on ½ Murashige and Skoog (MS) 60 with 0.05% MES (pH 5.7, adjusted with KOH) in 0.8% agar plates, sealed with micropore tape, then stratified for 4 d at 4 °C. Plants were grown for 5 days afer germination (DAG) before a 5.5 h treatment. Treatment consisted of transfer to ½ MS plates containing 100 mM NaCl (Merck) or fresh ½ MS with 0.05% MES (pH 5.7, adjusted with KOH) for mock.

Fluridone treatment
Seeds were surface sterilized with 96% ethanol, then sown on ½ MS with 0.05% MES (pH 5.7, adjusted with KOH) in 0.8% agar plates, sealed with micropore tape, then stratified for 4 d at 4 °C. Plants were grown for 5 DAG before a 24 h treatment. For treatment, plants were transferred to ½ MS with 0.05% MES (pH 5.7, adjusted with KOH), containing 0.4 μM fluridone (Merck, 45511) or an ethanol mock.

β-estradiol induction of ABA biosynthesis/catabolism
Seeds were surface sterilized with 96% ethanol, then sown on ½ MS with 0.05% MES (pH 5.7, adjusted with KOH) in 0.8% agar plates, sealed with micropore tape, then stratified for 4 d at 4 °C. Plants were grown for 5 DAG before a 24 h treatment. Treatment consisted of transfer to ½ MS with 0.05% MES (pH 5.7, adjusted with KOH), containing 10 μM β-estradiol or a dimethylsulfoxide (DMSO) mock.

Leaf humidity treatments for leaf imaging
nlsABACUS2-400n seeds were surface sterilized with 96% ethanol, then stratified for 4 d at 4 °C in sterile deionized water before sowing on F2 Levington's compost. Plants were grown (120 μE, 22 °C for 18 h; 0 μE, 18 °C for 6 h) for 15 DAG before humidity treatment. Plants were germinated under a clear plastic propagator lid, which was removed at 4 DAG.
To increase humidity, the chamber was set to 60% RH and humidity increased by placing a propagator lid over the plants for 6 h before imaging. Humidity and temperature were measured at leaf height above compost at ~95% RH and 22 °C for treatment, and at ~82% RH and 22 °C for mock. Humidity and temperature were measured using a BME280 sensor.
To decrease humidity, the chamber was set to 40% RH and plants were grown with a propagator lid until treatment. For treatment, compost was covered with acetate to slow evaporation and the lid was removed for 6 h before imaging. Humidity and temperature were measured at leaf height at ~76% RH and 22 °C for treatment, and at ~95% RH and 22 °C for mock. Humidity and temperature were measured using a BME280 sensor.

Peristomatal distance measurement
Stomatal aperture is challenging to measure from confocal images, but correlates strongly with peristomatal groove distance 61 , which we measured in our nlsABACUS2-400n humidity treatment confocal stacks. The line tool in Fiji was used to measure distance using a transmitted-light channel.

Foliar humidity treatment for root imaging
An 8 ml volume of ½ MS with 0.05% MES (pH 5.7, adjusted with KOH) in 0.8% agar was poured into a Nunc Lab-Tek II chambered coverglass (155360, Thermo Fisher) and allowed to set. Half of the agar was aseptically removed and seeds were placed on the agar next to the coverslip to allow plant roots to grow vertically between the agar and the coverslip (Supplementary Fig. 17). Chambers were sealed three times with micropore tape and stratified for 4 d and then plants were grown to 6 d post stratification in a long-day chamber. For the humidity treatment, imaging chambers were opened, a piece of folded acetate was placed over the agar to prevent direct evaporation and aerial tissues were exposed to the 40% RH, 22 °C chamber for 6 h (Supplementary Fig. 17). Mock treatment involved opening the chamber, applying a smaller piece of acetate and resealing before returning to the growth chamber. The smaller acetate application acts as a control for any mechanical perturbation, but still retains a large area for water exchange between the agar and air, so the chamber remains humid and equilibrates quickly.

Foliar humidity treatment for root growth assays and β-estradiol pretreatment
An 80 ml volume of ½ MS with 0.05% MES (pH 5.7, adjusted with KOH) in 0.8% agar was poured into a 10 cm square plate and allowed to set. Agar (2.5 cm) was aseptically removed from one side and seeds were placed on the agar next to the back of the plate to allow plant roots to grow vertically between the agar and plate (Supplementary Fig. 17). Plates were sealed three times with micropore tape, stratified for 4 d and then plants were grown for 6 d post stratification in a long-day chamber. Immediately before treatment, the position of the primary root was marked on the plate with a razor blade and a dissecting microscope. Plants were imaged with a flatbed scanner immediately at the end of the humidity treatment, allowing growth during the treatment to be assayed.
For the humidity treatment, plates were opened, a piece of folded acetate was placed over the agar to prevent direct evaporation and plants were exposed to the 40% RH, 22 °C chamber for 7 h (Supplementary Fig. 17). Mock treatment involved opening the plates, applying a smaller piece of acetate and resealing before returning to the growth chamber. The smaller acetate application acts as a control for any mechanical perturbation, but still retains a large area for water exchange between the agar and air, so the plate remains humid and equilibrates quickly. Immediately following humidity treatment, plates were scanned with an EPSON flatbed scanner at 1,200 dpi and saved as .tif files. Root growth was then assayed with ImageJ.

RootChip microfluidics treatments
The RootChip-8S device was used for ABA pulsing as described previously 12,40 . Arabidopsis seeds were germinated on the bottom 5 mm of 10 μl pipette tips filled with solidified growth medium (½ MS, 0.05% MES (pH 5.7, adjusted with KOH), 1% agar). After 4-7 d, pipette tip seedlings were transferred to the polydimethylsiloxane RootChip-8S device under aseptic conditions. A peristaltic pump was used (DNE; volumetric flow rate in each channel, 5 ml min −1 ) to perfuse the roots with ¼ MS (pH 5.7) liquid media. The dead volume was assessed, and it took approximately 12 min for media to pass through the tubing to reach the root, which was accounted for when plotting the ABA treatments. Imaging was performed on an inverted Leica SP8 with a ×20 dry 0.70 HC PLAN APO objective. Sequential scanning with a 448 nm laser was used to excite the edCerulean (for edCerulean and edCitrine FRET emission) and 514 nm lasers were used to excite edCitrine (for edCitrine emission, acting as an expression control). Emission settings were 460-490 nm for Cerulean and 520-550 nm for edCitrine.

ABA hypersensitivity germination assays
Seeds were surface sterilized, placed on large agar plates with ½ MS, 0.05% MES (pH 5.7, adjusted with KOH) and 0.8% agar with or without 1 μM ABA and stratified for 4 d. After transfer to a growth chamber, a dissecting microscope was used to score germination daily. Seedling emergence from the endosperm was used to score germination.

ABA hypersensitivity primary and lateral root growth assays
Seeds were surface sterilized, placed on large agar plates with ½ MS, 0.05% MES (pH 5.7, adjusted with KOH) and 0.8% agar vertically in a growth chamber. At 6 DAG, seedlings of approximately equal length were transferred to mock or 10 μM ABA plates. Root tip positions were marked with a pen and plates were placed vertically in the growth cabinet for 3 and 6 d before imaging on a flatbed scanner. Primary root growth was measured from the length at transfer to the root tip with the segmented line tool of Fiji. Total visible lateral root count was assayed from the scanned images of mock-treated roots at 6 d post transfer, with the multipoint tool in Fiji, including lateral roots initiating both before and after transfer.

Confocal imaging
An upright Leica SP8-Fliman confocal microscope was used for most biosensor imaging. An inverted Leica SP8-iphox confocal microscope was used for RootChip imaging, cyp707a1cyp707a3 imaging and PP11 cotyledon imaging. All images were acquired as z-stacks in 16 bit mode, with a ×10 or ×20 dry 0.70 HC PLAN APO objective. Samples were mounted in ¼ MS (pH 5.7, adjusted with KOH).

Lightsheet microscope setup
Lightsheet microscopy was performed using a custom-built laser scanning lightsheet microscope. The design is based on an openspim geometry 62 with dual side illumination and dual side detection. Water immersion objectives were mounted horizontally (Nikon ×10, 0.3 NA for excitation, Olympus ×20 1.0 NA for detection), with the sample suspended from the top in an agarose-filled fluorinated ethylene propylene tube. For sample placement as well as for imaging, the sample can be moved between the objectives well as rotated with piezo-driven stages (Nanos LPS-30, Nanos RPS-LW20). Image stacks were acquired by moving the sample through the stationary imaging plane. Lasers (445 nm and 488 nm; Omicron LuxX 445-100, Omicron LuxX 488-200) were used for excitation and combined in an Omicron LightHub 6 with dual fibre output. The fibre output was collimated, galvo scanned (Galvo system, Thorlabs GVSM002-EC/M) and magnified, resulting in a scanned light sheet with typical full-width at half-maximum <5 μm. Two sCMOS cameras (Hamamatsu Orca Flash 4) with 6.5 × 6.5 μm 2 pixel size were used for detection. Two motorized filter wheels (Cairn OptoSpin) with bandpass filters (Semrock FF01-480/17, Semrock FF01-532/18) allow the recording of specific fluorescence bands. The microscope was controlled by a custom software developed in LabVIEW (National Instruments). Data were streamed to disk and converted to TIFF files directly after acquisition, resulting in image voxel sizes of 1 μm 3 .

Lightsheet imaging
The plants were grown suspended in a cut 10 μl pipette tip as in ref. 63 (in ½ MS, 0.05% MES (pH 5.7, adjusted with KOH) and 0.5% agarose within fluorinated ethylene propylene tubes (i.d. 0.8 mm)). They were illuminated from 2 sides while 3 fluorescent channels were recorded sequentially (Ch1: Exc 445 nm, Em 480/17; Ch2: Exc 488 nm, Em 532/18; Ch3: Exc 445 nm, Em 532/18). Typical excitation powers set in the software were 10%-50% for 445 nm excitation and 1-3% for 488 nm excitation. Camera exposure time was set to 100 ms per plane for all channels. Multiple viewpoints (60° rotation increments) were recorded for each timepoint and combined in Fiji 64 using the Multiview reconstruction plugin 65 before further analysis. Foliar ABA treatment was performed by pipetting 5 μM ABA onto the cotyledons, which were isolated from the roots.

'FRETENATOR 1.5' toolset development
A fast yet flexible analysis pipeline was required to analyse biosensor data. Because the biosensors used in this paper are nuclear localized, the pipeline was designed for punctate nuclear segmentation and analysis was performed on a per nucleus basis. The toolset consisted of two plugins. 'FRETENATOR segment and ratio 1.5' was used to segment punctate structures, perform ratio calculations and export the data as images and as a results table. 'FRETENATOR ROI labeller' was used to assign specific labels to the regions of interest (ROI) produced by 'FRETENATOR Segment and ratio' and export this information to the results table.

Development of FRETENATOR segment and ratio v.1.5
Fiji 64 , an open-source multiplatform widely adopted ImageJ 66,67 distribution, was chosen as platform to allow the greatest flexibility to users. All plugins were developed in Jython using CLIJ/CLIJ2 (ref. 68) to perform image processing directly on the graphics card. On computers with dedicated graphics cards, this allows fast analysis and modification of the segmentation settings that can be performed through a graphical user interface ( Supplementary Fig. 18) with near-real-time segmentation previews. All code is freely available at https://github.com/JimageJ/ ImageJ-Tools, along with installation and usage tutorial videos.
Segmentation steps are illustrated in Supplementary Fig. 19. Preprocessing consists of extracting the segmentation channel, applying a 3D difference of Gaussian filter to smooth noise and remove background. An optional tophat filter allows further background subtraction. A choice of various automatic methods or manual thresholding is then used to generate a binary map.
An optional 3D watershed is used to split objects. Because 3D watershed can cause the loss of too many nuclei ROI or shrink them below their original size, we compare the watershed to non-watershed binary maps. A map of the 'lost nuclei' is generated, with these lost nuclei being added back later.
A 3D connected-components analysis is used to generate a label map of the watershed nuclei. As a watershed shrinks objects, the labelled objects are dilated (on zero-value pixels only), then multiplied by the original threshold image. This provides a good segmentation with split objects without object shrinkage.
To correct account for any 'lost nuclei' absent from the image, a connected-components analysis is run on the 'lost nuclei' map to generate labels which are supplemented back onto the first label map.
Once the segmentation is complete, voxels that are saturated on either the donor excited donor emission (DxDm) or the donor excited acceptor emission (DxAm) are excluded from analysis of both channels, and the emission ratio (DxAm/DxDm) is calculated for each ROI. The segmentation is also used to quantify position, size, donor intensity, acceptor FRET intensity, acceptor intensity, pixel count and image frame for each ROI, which are exported as a results table along with file name and ROI identifiers (Supplementary Fig. 20). The following outputs are produced upon plugin completion: Threshold stack, the Article https://doi.org/10.1038/s41477-023-01447-4 Label stack, Emission ratio stack, Emission ratio maximum Z-projection and Emission ratio nearest-point Z-projection. Please note, to halve the file size of exported images, emission ratio values are multiplied by 1,000 in exported image files, allowing the files to be saved as 16-bit images, instead of 32-bit images.
A log of segmentation settings is also created every time the 'FRETENATOR segment and ratio 1.5' plugin is run.

Development of FRETENATOR ROI labeller
The ROI labeller is a follow-on tool for post-segmentation analysis where users can categorize the ROI in their segmented images (Supplementary Fig. 21). It currently works on single-timepoint 3D-label images, allowing users to visually assign labels to one of 10 categories. Results are either output to an existing results table or can be used to remeasure a chosen image.

FRETENATOR software compatibility
The majority of testing was performed on a 2017 Dell desktop (Windows 10, Intel i7-6700 CPU, 3.41 GHz, 32 GB RAM, Intel HD Graphics 4000/AMD Radeon R7 450) and a 2014 Gigabyte laptop (Ubuntu, Intel i7-4710Q, 2.5 GHz Quad core, 16 GB RAM, Nvidia GTX 860M 4 gb). We also regularly use the software on Windows, Linux and Mac machines of varying ages and specifications. Considerable speed increases are present on modern hardware with fast graphics memory. Dozens of Arabidopsis cotyledon z-stacks have been tested.

FRETENATOR 1.5 validation by comparison with Imaris 8.2
'FRETENATOR segment and ratio 1.5' analysis was compared to the commercial software Imaris 8.2 (https://imaris.oxinst.com/) for validation and to ensure comparable results. Buffer exchange and segmentation were performed as previously described 28,69 . Segmentation in Imaris was performed using the surfaces wizard on the AxAm channel, with background subtraction and object splitting. The XTMeanIntensityRatio Xtension was used for emission ratio calculation. FRETENATOR 1.5 and Imaris gave extremely close results in terms of both segmentation and quantification of emission ratio (Supplementary Fig. 22). As FRETENATOR 1.5 is free, quick to use and can be installed on old, low-specification computer hardware, FRETENATOR 1.5 was used for subsequent biosensor analysis.

Image analysis using FRETENATOR 1.5
All segmentation and labelling were performed with the 'FRETENATOR' plugins. Segmentation settings were optimized for each experiment but kept constant within each experiment. The AxAm channel was used for segmentation. Watershed was used for the dense nuclei of the root tip but switched off for leaf imaging. Difference of Gaussian kernel size was determined empirically due to different magnifications, resolutions and amount of noise. As a default, Otsu thresholds were used for segmentation, but in experiments where this gave poor segmentation, a manual threshold was used on the dataset (the same value for each image in the dataset).
For time courses, images were concatenated, registered using the 'Correct 3D drift' 70 and 'Manual drift correction' plugins in Fiji before analysis. To examine internal vs external tissues of cotyledons, segmentation of epidermal tissues was performed using 'EZ-Peeler' v. 1.16 (ref. 71) to generate separate surface and interior image stacks. Then further analysis using a constant threshold in FRETENATOR 1.5 was performed.
For lightsheet images, viewpoints were combined in Fiji 64 using the 'Multiview reconstruction' plugin 65 . Rolling ball background subtraction (Fiji: subtract background) was performed before processing with FRETENATOR 1.5.

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

Data availability
For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any author-accepted manuscript version arising from this submission. New plant lines are deposited at the Nottingham Arabidopsis Stock Centre (NASC IDs: N2111654-N2111668). Binary vectors for ABACUS2 plant transformation and ABACUS2 constructs in pENTR221-f1 are deposited at Addgene (IDs: 203725-203728). All data are deposited at the Cambridge data repository at https://doi.org/10.17863/CAM.96615.

Code availability
The FRETENATOR 1.5 image analysis toolset, as well as installation and usage instructions, are available at https://github.com/JimageJ/ ImageJ-Tools. Fig. 2 | Comparisons of Colabfold predictions of the nlsABACUS2-100n and 400n binding pockets with crystal structure of PYL1-ABA-ABI1 PDB 3KDJ. a) Comparison of the PYL1 ABA binding pocket and latch in wildtype crystal structure 3KDJ with ABACUS2 Colabfold structural prediction. The PYL1 R143S mutation (labelled in ABACUS2-400n) prediction has altered residue positions near the top of the pocket with a relatively unchanged inner pocket. b) Comparison of the PYL1 ABA binding pocket and latch in wildtype crystal structure 3KDJ with ABACUS2 Colabfold structural prediction. The PYL1 E141D (labelled in ABACUS2-100n) prediction has altered residue positions near the top of the pocket with a relatively unchanged inner pocket. c) The water mediated interaction between PYL1 R143, PYL1 P115 and ABI W300 is not reproduced in the ABACUS2-400n Colabfold structural prediction, which includes a PYL1 R143S. Green: 3KDJ protein; White: 3KDJ ABA; Blue: 3KDJ water molecule; Magenta: nlsABACUS2-400n prediction. d) The PYL1 E141D mutation may alter the 'latch' fold. Green: 3KDJ protein; White: 3KDJ ABA; Blue: 3KDJ water molecule; Magenta: nlsABACUS2-100n prediction.  Extended Data Fig. 9 | Decreasing foliar humidity increases leaf epidermal ABA levels and closes stomata. a) Peristomatal distance for 15 DAG nlsABACUS2-400n plants decreases following a 6-hour humidity reduction. Datapoints indicated the mean of distance of all stomata in a field of view of a confocal datastack. Each field of view contained between 8 and 16 measured stomata. Peristomatal distance taken to be a proxy for stomatal aperture 61 . Unpaired two tailed T-test p=0.0002, t=4.501, df=19. n=9,12 biologically independent leaves respectively. For boxplots, centre line indicates median; box limits indicate upper and lower quartiles; whiskers indicate the upper/lower adjacent values. b, c) Cropped representative images of nlsABACUS2-400n emission ratios in response to a 6-hour humidity decrease. Relative humidity (RH) indicates the measured humidity at leaf height during the treatments. Scale bar indicates 25 μm.