Abscisic acid, 98% (ABA, cat. no: 342401000), ABA bovine serum albumin (ABA-BSA, cat. no: CSB-MC00421b0105), ABA precursor abscisic aldehyde (ABA aldehyde, cat. no: SC-479414), β-D-glucopyranosyl abscisate (ABA glucosyl ester, ABA-GE, cat. no: HY-111974-1) and 1-aminocyclopropane-1-carboxylic acid (ACC, cat. no: A1178) were purchased from Thermo Fisher Scientific (Waltham, MA, USA), Cusabio (Houston, TX, USA), Bio-Connect (Huissen, The Netherlands), Sanbio B.V. (Uden, The Netherlands) and TCI (Tokyo, Japan), respectively. Bovine serum albumin (BSA, 9998) was purchased from Cell Signaling Technology (Danvers, MA, USA). The mouse anti-ABA monoclonal antibody (cat. no: MBS7041609) was acquired from MyBioSource (San Diego, CA, USA). Sodium acetate and Tween^®20 were purchased from Merck/Sigma-Aldrich (Darmstadt, Germany). NaCl and 2-(4-(2-hydroxyethyl)-1-piperazinyl)ethanesulfonic acid (HEPES) were purchased from Carl Roth (Karlsruhe, Germany), whereas dimethyl sulfoxide (DMSO) was acquired from Thermo Fisher Scientific. N-hydroxysuccinimide (NHS), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), 10 mM of sodium acetate, pH 5.0, 1 M of ethanolamine, pH 8.0, and HCl, pH 1.5, were obtained from Cytiva (Uppsala, Sweden) as well as CM5 sensor chips and the Biacore™ 8K instrument. Polypropylene 96-well plates (651201) used to deposit samples for Biacore™ experiments were acquired from Greiner Bio-One (Kremsmünster, Austria).

Tomato (Solanum lycopersicum, var. Ailsa Craig) seeds were sown in soil in growing trays and moistened with deionized water. Approximately two weeks after germination, the seedlings were transplanted onto individual rockwool blocks and watered with nutrient solution (electrical conductivity (EC): 2.5–3 dS/m, pH: 5.5–6; [41]). The plants were grown in a climate-controlled growth chamber with a set temperature of 26 °C and a constant relative humidity of 55%. The plants were grown under a long-day photoperiod with 16 h of light and 8 h in the dark, and a light intensity at the shelf level of 150 µmol/(m²s¹) (VS12 LED, Sunritek, Shenzhen, China).

The plants were sampled once they reached leaf stage eight (BBCH 18 [42]). The xylem sap sampling method consisted of cutting the top part of the plant close to the surface, right below the cotyledons. The sampling of xylem sap relied on active root pressure pushing out the sap and forming droplets on top of the cut stem. The first droplet was discarded due to a possible contamination of broken cells caused by cutting the stem. The sap was collected by pipetting it off the stem and depositing it in Eppendorf tubes placed on ice. This process was repeated for one hour, stopping before the wound response could set in and alter the xylem sap matrix. After collection, the tubes were stored at −20 °C.

Twenty plants at leaf stage eight were used for the stress experiments. The drought and salt stress experiments each included ten plants, with five plants receiving the stress treatment and five plants serving as controls. Until the start of the abiotic stress, both treatment groups received the same nutrient solution with an EC of 2.5–3 dS/m and a pH of 5.5–6. For the drought treatment, the treated plants were not irrigated anymore, while the control plants continued to be irrigated throughout the experiment. For the salt stress treatment, the control nutrient solution (EC 2.5–3 dS/m) was supplemented with NaCl to reach severe salinity (EC 10 dS/m) according to Holsteens et al. [41], whereas the control plants continued to receive the original nutrient solution. For both stress treatments, the abiotic stress period lasted for 48 h, after which the sampling method described above was employed to collect the xylem sap. Each sample was considered a biological replicate, without pooling the samples.

The quantification of ABA in xylem sap samples using the Biacore™ SPR immunoassay was performed based on the established ABA calibration curve in buffer, by following the corresponding Biacore™ application guide [45]. The xylem sap samples, collected from plants subjected to abiotic stress, were centrifuged at 2000× g for one minute and diluted two-fold in running buffer. These samples were analyzed following the same conditions and protocol as explained in Section 2.4. Reference and blank subtraction were not applied for these measurements. Although a reference surface (Fc1) is highly valuable in many SPR applications, it is not recommended for concentration measurements, as errors can be caused due to the difference in nonspecific binding of the sample matrix between the active (Fc2) and reference (Fc1) surface [45]. Additionally, blank xylem subtraction was not possible either because xylem sap samples collected from control plants also contain endogenous ABA.

The UPLC-MS/MS method, as described by Turečková et al. [21] with some modifications, was used as a benchmark to quantify endogenous levels of ABA and its catabolites in the matching tomato plant xylem sap samples from the Biacore™ SPR immunoassay experiments. Briefly, tomato xylem sap was homogenized using a vortex, and 5 μL of each sample was transferred into vials, then spiked with 3 μL (3 pmol) of a labelled internal standard mixture (including (−)-7’,7’,7’-²H₃-phaseic acid; (−)-7’,7’,7’-²H₃-dihydrophaseic acid; (-)-8’,8’,8’-²H₃-neophaseic acid; (+)-4,5,8’,8’,8’-²H₅-ABAGE; (−)-5,8’,8’,8’-²H₄-7’-OH-ABA and (+)-3’,5’,5’,7’,7’,7’-²H₆-ABA). The samples were then diluted with 30% (v/v) methanol up to a final volume of 30 μL. A 3 μL aliquot of each sample was injected onto a reversed-phase column (Acquity UPLC BEH C18, 1.7 μm, 2.1 × 150 mm, Waters, Milford, MA, USA) and analyzed by UPLC-MS/MS using a Xevo TQ-XS triple quadrupole mass spectrometer (Waters).

All data were analyzed and visualized using GraphPad (Version 8.0.2, Delta Prism). For all experiments, due to small sample sizes, the normality of the residuals was tested using both the D’Agostino-Pearson omnibus K² and the Shapiro–Wilk W normality tests [46,47]. In the first analysis, the RU signal values of both ABA analogues were individually compared to those of ABA in running buffer supplemented with 2% DMSO (n = 4). None of the data sets passed the normality tests, so the significant differences were confirmed by a Mann–Whitney U test for the distribution-free assumption (p-value < 0.05) [48]. In the second analysis, the ABA concentration measurements of the abiotic stress-treated plant samples were compared with their respective controls. All data sets were confirmed to pass both normality tests except for the high salinity data set with the Biacore™ platform and with the UPLC-MS/MS. For reasons of uniformity, the significant difference for all data sets was confirmed by a Mann–Whitney U test for the distribution-free assumption (p-value < 0.05). In the third analysis, correlation plots were constructed between the Biacore™ and the UPLC-MS/MS techniques. All data sets passed the normality tests, so the Pearson correlation coefficient was used to compute the correlation between the two techniques (p-value < 0.05) [49].

To establish sensitive and specific detection of ABA in xylem sap, we first characterized the selected commercial anti-ABA antibody (ligand) for its interaction with the target (analyte) molecule. This was performed using the Biacore™ SPR platform, which enables the study of these interactions in a direct manner even when target molecules are as small as ABA (264.32 g/mol). In addition to ABA alone, we also included ABA-BSA in this characterization study, as this conjugate was used for the development of the selected antibody (since ABA alone is too small to elicit an immune response), and BSA as a negative control. Anti-ABA antibody was covalently immobilized on the sensor chip via amine coupling, and multiple concentrations of either ABA, ABA-BSA, or BSA were flown over this sensor chip. A schematic representation of the sensorgram obtained from such direct SPR immunoassay with ABA is shown in Figure 1, revealing four stages of a sample measurement: (1) the baseline, established in the running buffer; (2) the association phase, starting upon sample injection that shows ABA binding to the anti-ABA antibody in real time; (3) binding equilibrium and (4) dissociation of the bound ABA that takes place in the running buffer. The ABA response binding levels (RUs) were measured at a report point during the dissociation phase when only running buffer flowed over the sensor surface, 10 s after the end of sample injection, to avoid bulk RI contributions from the sample matrix, such as the xylem sap. In this context, it is important to note that data shown in Supplementary Figure S1 indicates the absence of nonspecific binding of the xylem sap matrix to the reference channel Fc1, either left blank or functionalized with an unrelated antibody to mimic the surface characteristics of the anti-ABA antibody functionalized surface of Fc2. Therefore, the obtained SPR signal originates only from bound ABA.

The actual SPR sensorgrams obtained from the direct interaction of anti-ABA antibody, immobilized on the chip surface, with ABA, ABA-BSA, and BSA are depicted in Figure 2A, B, C, respectively. The results are shown from one representative channel out of eight, after correction using the signal from Fc1 and the buffer blank. The kinetic rate constants (k_a and k_d) and the equilibrium dissociation constant (K_D) were derived by fitting the experimental data to the 1:1 Langmuir binding model [50]. The anti-ABA immobilization graph and the residuals graph for the ABA, ABA-BSA, and BSA sensorgrams are all provided in Supplementary Figure S2. Higher ABA concentrations resulted in a noticeable increase in signal due to the binding of ABA to the anti-ABA antibody on the surface. As can be seen from the sensorgram shape, the binding is transient and characterized by a fast association and dissociation phase. Importantly, a dissociation phase of two minutes was enough to remove all ABA from the anti-ABA antibody surface, which avoided the use of surface regeneration. The corresponding k_a and k_d values, as well as K_D, are reported in the table under Figure 2A. The dissociation constant K_D of 3.22 ± 0.12 × 10⁻⁷ M indicated a moderate affinity.

The interaction of ABA-BSA with the anti-ABA antibody produced a visually different sensorgram with slower binding kinetics (Figure 2B). The response signal was notably higher for ABA-BSA, which can be explained by its significantly higher molecular mass (66.5 kDa) compared to ABA (264 Da). When translated into molar concentrations, the maximum ABA concentration (i.e., 1000 ng/mL) corresponds to 3780 nM, whereas the maximum ABA-BSA concentration (i.e., 4000 ng/mL) corresponds only to 60.15 nM. Therefore, even though lower molecular concentrations were used for ABA-BSA, its larger mass resulted in higher response signals than ABA. Interestingly, there was no equilibrium plateau forming, nor any dissociation after injection, indicating the need for a regeneration step. The K_D value for ABA-BSA was also a hundred times lower than the K_D for ABA, due to the k_d value being a hundred times smaller, suggesting a strong interaction of the antibody with the ABA target in the nanomolar range. This observation can be explained by the fact that the anti-ABA antibody was raised against ABA-BSA. Importantly, the negative control experiment using BSA as analyte confirmed that the antibody response to ABA-BSA is due to the specific recognition of ABA and not due to nonspecific binding to the carrier protein BSA.

The specificity of the anti-ABA antibody towards ABA was further investigated through additional cross-reactivity studies with ABA analogues (i.e., namely ABA-related compounds, including the precursor ABA aldehyde and the glucose conjugate ABA-GE) and another small plant molecule, ACC (a precursor of the plant hormone ethylene). Experiments involving ABA and its analogues were conducted in running buffer supplemented with 2% DMSO, which was required to dissolve the analogues (the DMSO bulk shift corrections are presented in Supplementary Figure S3). In contrast, ACC was analyzed using the original running buffer. From the direct immunoassay performed as described above, the binding kinetic parameters were determined for ABA, ABA aldehyde, ABA-GE, and ACC as listed in Table 1. The results show that the anti-ABA antibody binds most strongly to its native target ABA (K_D: 4.66 ± 0.73 × 10⁻⁷ M) while exhibiting considerably lower affinity towards ABA aldehyde (K_D: 8.12 ± 0.83 × 10⁻⁶ M) and ABA-GE (K_D: 1.74 ± 0.31 × 10⁻⁵ M). Additionally, when all four were injected at a consistent concentration of around 150 ng/mL, both ABA aldehyde and ABA-GE produced significantly lower RU values on the SPR platform, in contrast to ABA. The negative control, ACC, showed no detectable interaction with the antibody under the tested conditions and, thus, none of the parameters could be determined. Kinetic binding curves with Rmax value and the related residual for each of the analytes are presented in Supplementary Figure S4.

Calibration curves were established in all eight channels of the CM5 chip by detecting ABA spiked into running buffer at concentrations ranging from 2 to 1024 ng/mL using a two-fold dilution (Figure 3 depicts a calibration curve from channel one). All ten concentrations were measured from low to high and in triplicate (n = 3). Due to the rapid k_d, no regeneration condition was required between cycles. The running buffer was supplemented with BSA to block the chip surface. The binding characteristics of ABA diluted in running buffer and ABA diluted in running buffer supplemented with BSA were found to be identical, indicating that BSA does not influence the binding of ABA to the anti-ABA antibody (Supplementary Figure S5). To build the calibration curves, the response at the report point during the dissociation phase was plotted against the corresponding ABA concentration. ABA quantification at this point, when running buffer flows over the chip surface, enables the selective detection of ABA that is specifically bound to the sensor surface. This report point eliminates matrix interference and the need for a calibration curve in the xylem sap matrix. Moreover, no Fc1 and buffer correction were performed as these calibration curves were to be used to derive the unknown ABA concentrations from xylem sap samples (see Section 3.3). A sigmoidal, four-parameter logistic regression was used to fit the calibration curve, for which the limit of detection (LOD) was calculated to be 1.36 ng/mL, which is below the lowest measured ABA concentration (2 ng/mL) [45]. It is important to note that the stability of the immunosensor chip was evaluated and found stable for at least three days and 77 cycles, with no noticeable initial decline (Supplementary Figure S6).

The established SPR Biacore™ immunoassay was further employed to quantify unknown ABA concentrations in xylem sap samples collected from control and abiotically stressed tomato plants, which had undergone drought and high salinity treatments. It is important to note that, for each tested category (i.e., control plants for drought treatment, control plants for high salinity treatment, as well as drought and high salinity treatments), the xylem sap was collected from five different plants, with each sample being considered a biological replicate. Each sample was then divided into two aliquots for parallel analysis using our newly established SPR Biacore™ immunoassay and UPLC-MS/MS. Figure 4 provides an overview of the results from the two assays, with ABA concentration values reported for both the Biacore™ SPR immunoassay and the UPLC-MS/MS. The treated plants under drought and high salinity stress exhibit a significantly higher ABA concentration than the control plants for both techniques. Additionally, the relative signal change between the stressed and control plants can be compared, defined as the ratio of the treatment signal over the control signal (displayed on top of the line). For both Biacore™ and UPLC-MS/MS, the fold changes for the high salinity treatments are higher than the fold changes for the drought treatments. When compared to each other, fold changes are slightly lower for the Biacore™ immunosensor than for UPLC-MS/MS.

Next, a correlation plot was created between both techniques (Figure 5), including a linear regression and a Pearson correlation coefficient. The slope of the linear regression with a value of 0.5782 indicates an underestimation of the ABA concentrations by the Biacore™ immunoassay compared to the UPLC-MS/MS, while the Pearson correlation coefficient with a value of 0.9972 shows a strong positive correlation between both.

The research data are available upon request.