Gibberellin and abscisic acid transporters facilitate endodermal suberin formation in Arabidopsis

The plant hormone gibberellin (GA) regulates multiple developmental processes. It accumulates in the root elongating endodermis, but how it moves into this cell file and the significance of this accumulation are unclear. Here we identify three NITRATE TRANSPORTER1/PEPTIDE TRANSPORTER (NPF) transporters required for GA and abscisic acid (ABA) translocation. We demonstrate that NPF2.14 is a subcellular GA/ABA transporter, presumably the first to be identified in plants, facilitating GA and ABA accumulation in the root endodermis to regulate suberization. Further, NPF2.12 and NPF2.13, closely related proteins, are plasma membrane-localized GA and ABA importers that facilitate shoot-to-root GA12 translocation, regulating endodermal hormone accumulation. This work reveals that GA is required for root suberization and that GA and ABA can act non-antagonistically. We demonstrate how the clade of transporters mediates hormone flow with cell-file-specific vacuolar storage at the phloem unloading zone, and slow release of hormone to induce suberin formation in the maturation zone. The authors identified a sub-clade of NPF transporters that orchestrates GA12 long-distance shoot-to-root translocation. Once in the phloem unloading zone, ABA and GA are loaded into pericycle vacuoles and then slowly released to induce endodermal suberin formation in the maturation zone.

14 is a vacuolar GA and ABA transporter. a, GA accumulation in NPF2.14-expressing and control oocytes exposed to the indicated GAs at 50 µM concentration for 60 min at room temperature at pH 5 for GA 4 (P = 0.0031) and GA 7 (P = 0.001) and at pH 6 for GA 9 (P < 0.0001) and GA 12 (P = 0.0004). Mean + s.e. (n = 5 for GA 4 , GA 7 and GA 9 and n = 6 for GA 12 ). Two-tailed t-test (***P < 0.005). b, GA accumulation at 20 h after direct injection of 23 nl of 8.2 mM membraneimpermeable GA 3 at pH 7.4 at 16 °C (t1) and after 60 min at room temperature at pH 5 (t2). Mean + s.e. (n = 6 single oocytes). Two-way analysis of variance (ANOVA) with Holm Sidak post hoc test to adjust for multiple comparisons (P ≤ 0.05). Different letters indicate statistically significant groups. c, GA accumulation in control oocytes or oocytes that express NPF2.14, NPF4.1 or both proteins exposed to 50 µM GA 3 at pH 5 for 60 min at room temperature and analysed by LC-MS/MS (liquid chromatography coupled to tandem mass spectrometry). Mean + s.e. (n = 7 single oocytes). Holm Sidak one-way ANOVA (P = 0.05). d, ABA accumulation in oocytes exposed to 50 µM ABA at pH 5 for 60 min at room temperature (n = 5 single oocytes). Two-tailed t-test (P = 0.05). e, Nitrate accumulation in oocytes exposed to 5 mM nitrate at pH 5 at room temperature for 60 min. Mean + s.e. (3 replicates of 5 oocytes analysed by analytical anion chromatography, n = 3). Holm Sidak one-way ANOVA (P = 0.05). f, Internal oocyte pH measured using three-electrode voltage clamp electrophysiology of control oocytes and NPF2.14-and NPF7.3-expressing oocytes. Oocytes were perfused at pH 7.4 for 5 min followed by perfusion at pH 5 for 5 min. Mean + s.e.
NPF7.3 has recently been shown to lower the cytoplasmic pH in X. laevis oocytes, which can indirectly influence the accumulation equilibrium of weak acids such as GA and ABA 34  Col-0 npf2.14 Col-0 Merge NPF2.14-YFP Vac-CFP PI npf2.14 NPF2.14 has a similar activity, we measured the intracellular oocyte pH using a proton-selective three-electrode voltage clamp setup. We showed that, unlike NPF7.3, NPF2.14 expression in the oocyte does not alter the internal oocyte pH (Fig. 1f). Another factor that can theoretically lead to a false-positive GA export result is alteration of membrane potential. Therefore, we measured the membrane potential of control and NPF2.14-expressing oocytes using a two-electrode voltage clamp setup. The membrane potential of control and NPF2.14-expressing oocytes were both approximately −15 mV (Fig. 1g). When oocytes were subjected to GA 4 for 60 min before membrane potential measurement, less GA was observed in NPF2.14-expressing oocytes than in control oocytes (Fig. 1g). Thus, NPF2.14 does not shift oocyte membrane potential. Many NPF proteins, including NPF2.14, contain the ExxE[K/R] motif 35 , which is involved in coupling substrate transport to the proton gradient across membranes 36 . Involvement of the ExxE[K/R] motif in NPF2.14-mediated effluxes would suggest antiporter function. To assess the involvement of the ExxE[K/R] motif, we generated C-terminal Yellow Fluorescent Protein (YFP) YFP-tagged NPF2.14 mutants substituted at each of the three charged residues with a polar but uncharged residue. The YFP-tag alone did not influence the apparent GA 4 transport by the wild-type NPF2.14 ( Fig. 1h). When any of the charged residues of the ExxE[K/R] motif was replaced with the polar uncharged Gln residue, no significant difference in GA 4 transport was observed compared with wild-type NPF2.14 ( Fig. 1h). Thus, the GA 4 transport mediated by NPF2.14 seems to be ExxE[K/R] motif independent.
To test whether the X. laevis oocyte GA transport data are physiologically relevant in planta, we isolated a homozygous T-DNA knockout line for NPF2.14. The single npf2.14 mutant did not show notable shoot or root growth phenotypes ( Supplementary Fig. 1a,b). Several NPF family members transport nitrate, including NPF6.3 37 . Thus, despite having shown that NPF2.14 does not transport nitrate in oocytes (Fig. 1d), we checked whether npf2.14 mutants display an impaired growth on low-nitrate media. npf2.14 T-DNA insertion mutants did not have a visible growth phenotype and did not differ from Col-0 plants under low-nitrate conditions ( Supplementary Fig. 1c). To determine whether NPF2.14 is involved in GA distribution and accumulation in the root, we tested whether the distribution of a fluorescently tagged GA 3 compound (GA 3 -Fl) was affected in the loss-of-function line. GA 3 -Fl has been developed in our lab to serve as a stable bioactive reporter to study GA movement/accumulation in planta 28 . Accumulation of GA3-Fl in the endodermis was visible in the Col-0 plants, as previously shown 28 . The npf2.14 mutants displayed a significantly stronger signal compared with the Col-0 control (Fig. 1i). This enhanced accumulation was restored to normal levels when expressing NPF2.14 driven by its native promoter (pNPF2.14:NPF2.14-GFP) on the background of the npf2.14 T-DNA line (Fig. 1i), indicating that loss of NPF2.14 affects GA3-Fl distribution in the plant. In agreement with this result, npf2.14 mutants accumulated significantly higher levels of GA 4 in their roots (Fig. 1j).
To test whether NPF2.14 has a dual-specificity function and can also import ABA, we tested the distribution of the fluorescently tagged ABA (ABA-Fl) in the roots. ABA-Fl has very low bioactivity, but can be utilized to estimate ABA movement in the plant 33 . In addition, we quantified endogenous levels of ABA in the roots. We found that similar to GA-Fl, npf2.14 mutants accumulated significantly high levels of ABA-Fl in their root endodermis cells ( Supplementary Fig. 2), but showed low levels of native ABA (extracted from the entire root) (Fig. 1j).
To study the subcellular localization of NPF2.14, we generated and imaged 35S:NPF2.14-YFP lines. Interestingly, NPF2.14 localized to the tonoplast vacuole membrane (Fig. 1k). Therefore, although we had hypothesized that NPF2.14 was a GA exporter that transported GA from inside the cytosol to the apoplast, the protein is instead a tonoplast-localized transporter. To the best of our knowledge, this is presumably the first report of a subcellular GA/ABA transporter.

NPF2.14 regulates suberin formation in the root endodermis
To characterize NPF2.14 expression patterns in the plant, we generated NLS-YFP and GUS reporter lines driven by the NPF2.14 promoter. Confocal imaging of the NLS-YFP lines indicated that NPF2.14 is expressed only in the pericycle of the root mature zone, mainly at the phloem poles and not in the meristematic zone ( Fig. 2a and Supplementary Fig. 3a). In addition, GUS staining showed expression in the shoot vasculature in seedlings ( Supplementary Fig. 3b). To test whether GA or ABA affects NPF2.14 expression patterns, we examined pNPF2.14:GUS lines after an exogenous treatment of GA 3 (5 µM) or ABA (1 µM). While GA did not affect pNPF2.14:GUS staining, we observed a stronger staining in the ABA treatment, yet expression remained in the vasculature (Supplementary Fig. 4). In the mature stages, the NPF2.14-driven reporter was expressed in the periderm (Supplementary Fig. 3c). The pericycle is a deep layer of post-embryonic meristematic cells encircling the vascular tissue 38 . In the root, it is required for lateral root emergence 39 , xylem loading 40 and phloem unloading 41 . At later stages, it gives rise to the periderm, which serves as the outer protective layer when the surrounding tissue is sloughed off 38 . Both the endodermis and the cork, which is the outermost cell layer of the periderm, are suberized tissues 42 .
The expression of the reporter driven by the NPF2.14 promoter in a tissue that undergoes suberization, taken together with the ability of NPF2.14 to transport ABA, which has been previously shown to regulate suberin deposition 43 , led us to hypothesize that this transporter might facilitate root suberization.
To test this, we analysed suberization in npf2.14 T-DNA mutants using Nile red and Fluorol yellow, which are suberin dyes 44,45 . Suberization commences in the endodermis of the upper part of the maturation zone of the root and, as the plant matures more cells undergo suberization 46 . Quantification of Nile red and Fluorol yellow fluorescence intensity in the uppermost part of 5-day-old roots revealed that the mutant npf2.14 plants had significantly lower levels of endodermal suberin than Col-0 plants ( Fig. 2b and Supplementary Fig. 5). In addition, Col-0 plant roots showed a typical pattern of suberin formation with a non-suberized zone, followed by a suberizing zone where only patches of endodermal cells are suberized (patchy suberization) and a continuous suberized zone. We found significant reduction in the continuously suberized zone for npf2.14 T-DNA and an additional CRISPR npf2.14 allele we generated (Fig. 2c). Suberin levels remained lower compared with Col-0 at later stages of development, including 10-day-old roots and 3-week-old hypocotyls ( Supplementary Fig. 6), showing that the reduction of suberin deposition in the endodermis is stable over time and that also cork suberin is affected.
Suberin is a complex polyester based on glycerol and long-chain α,ω-diacids and ω-hydroxyacids 47 , and is primarily found in structures such as the periderm, endodermis and seed coat 30 . To examine changes in root suberin composition between npf2.14 and Col-0, we analysed their suberin monomer profiles via gas chromatography-mass spectrometry (GC-MS). We found significant reductions in ferulic acid, the predominant aromatic component of suberin, as well as in C22 fatty acid and C18:1(9) ω-hydroxyacid-two of the most abundant suberin building blocks in the Arabidopsis root endodermis (Fig. 2d). In addition, C18 ester was lower in npf2.14 roots. These reductions accompanied by lower levels of other monomers resulted in 35% less total suberin contents in the mutant roots (Fig. 2d). Overall, these findings provide several lines of evidence that alteration of NPF2.14 has a substantial effect on suberin deposition in the root endodermis.
Similar to the npf2.14 mutant, npf3.1 mutant plants, which have been shown to have impaired GA and ABA delivery to the endodermis 16 , displayed reduced suberization levels compared with Col-0 plants ( Supplementary Fig. 7). GA or ABA treatments completely rescued the npf2.14 mutant suberization levels (Fig. 2b).
Endodermal suberization is regulated by ABA perception, both under normal and stress conditions 43,48 . ABA treatment was previously reported to induce endodermal suberization 49 . In agreement with       b,c, Hormone uptake in control and NPF2.12-and NPF2.13-expressing Xenopus oocytes. Mean + s.e. b, Oocytes were exposed to 50 µM GA 1 at pH 5 (n = 5 single oocytes), 50 µM GA 4 at pH 5.5 (n = 5) or 50 µM GA 12 at pH 6 (n = 5). c, Oocytes were exposed to 50 µM ABA at pH 5 (n = 5). Hormone uptake analysed by LC-MS/MS. Holm Sidak one-way ANOVA (P = 0.05). d, Oocytes were exposed to 50 µM GA 1 at pH ranging from 5 to 7. Mean + s.e. (n = 5). e. IV (Current-Voltage) curve of GAinduced currents for mock (n = 10), NPF2.12 (n = 9) and NPF2.13 (n = 8) expressing oocytes exposed to 500 µM GA 3 at pH 5. Currents were measured using twoelectrode voltage clamp electrophysiology over a range of membrane potentials from +20 to −140 mV. this report, ABA significantly upregulated the Nile red fluorescence intensity in the root endodermis and rescued the reduced suberization observed in the aba2-1 mutant (AT1G52340), which is deficient in ABA biosynthesis 49 (Fig. 2e). To the best of our knowledge, GA has not been previously associated with endodermal suberization. To establish whether GA regulates root suberization in Arabidopsis seedlings, we quantified suberization levels in GA-treated wild-type and ga1-13 mutant plants using the Nile red dye. GA1 (AT4G02780) catalyses the first committed step in the GA biosynthetic pathway 50 . The GA biosynthesis mutant ga1-13 displayed a significant reduction in suberization levels, which was restored by the exogenous application of GA 3 (Fig. 2f). Furthermore, while ABA treatment, but not GA treatment, can rescue and induce the reduced suberin levels of aba2-1, both ABA and GA can rescue the reduced suberin levels of ga1-13 ( Fig. 2e-f). Notably, GA treatment could only rescue the suberin back to wild-type levels and did not induce it to a higher level as ABA treatment did. GA and ABA have long been thought to have completely antagonistic functions 51 . Our results indicate that this antagonistic activity is more complex and that GA and ABA induce root suberization. Taken together, our data suggest that NPF2.14 is a pericycle-specific GA and ABA transporter localized to the tonoplast and involved in regulating GA and ABA accumulation in the endodermis to promote suberization.
We hypothesized that due to this proximity on the phylogenetic tree, NPF2.12 and NPF2.13 might also contribute to GA and ABA accumulation in the endodermis. Both transporters were previously characterized as low-affinity nitrate transporters 52,53 and more recently, were shown to promote GA import activity in heterologous systems 14,34 . Neither have been characterized in plants as GA or ABA transporters. To test for direct GA transport activity of NPF2.12 and NPF2.13, we performed X. laevis oocyte-based transport assays. Oocytes expressing NPF2.12 or NPF2.13 accumulated significantly higher levels of GA 1 , GA 3 , GA 4 , GA 7 , GA 9 , GA 12 , GA 19 and GA 24 compared with control oocytes over the course of 60 min ( Fig. 3b and Supplementary Fig. 9a). This suggests that NPF2.12 and NPF2.13 are promiscuous GA importers. Both NPF2.12-and NPF2.13-expressing oocytes also had higher levels of ABA accumulation than controls (Fig. 3c). NPF2.12 and NPF2.13 both contain an ExxE[K/R] motif, which couples substrate transport to the proton gradient 36 . To test whether NPF2.12 and NPF2.13 substrate transport activity is coupled with external proton concentration, NPF2.12-and NPF2.13-expressing oocytes were exposed to membrane-impermeable GA 1 in solutions with pH ranging from 5 to 7 in 0.5 pH unit increments. In both NPF2.12-and NPF2.13-expressing oocytes, GA 1 accumulation increased as pH was lowered (Fig. 3d). These data indicate that GA transport by NPF2.12 and NPF2.13 is likely proton-coupled.
To investigate whether the transport of GA by NPF2.12 and NPF2.13 is electrogenic, we used two-electrode voltage clamp electrophysiology to evaluate oocytes that express the transporters. Subtracting the currents elicited by oocytes at different membrane potentials in the absence of GAs from the currents elicited by oocytes in the presence of 500 µM GA 3 revealed that GA 3 transport by both NPF2.12 and NPF2.13 is associated with negative currents relative to control oocytes (Fig. 3e). As the negative currents reflect a net positive influx of charges, this indicates at least a 2:1 proton:GA 3 stoichiometry if GA 3 is transported in its anionic form. However, it cannot be excluded that GA 3 may be transported in its neutral form. In non-clamped conditions, NPF2.13 displayed significantly higher uptake levels of GA 1 and GA 4 compared with NPF2.12. Similarly, currents elicited by NPF2.13-expressing oocytes were significantly higher compared with NPF2.12-expressing oocytes when oocytes were clamped at membrane potentials that mimic the membrane potential in the non-clamped uptake assays (Fig. 3e,f). However, at high negative membrane potential (−120 mV), GA 3 -induced currents in NPF2.12-expressing oocytes were of the same magnitude as those in NPF2.13-expressing oocytes (Fig. 3e,f). This suggests that the transport activity of NPF2.12 is sensitive to alterations in the membrane potential (Fig. 2b). In agreement with previous publications 52, 53 , we detected nitrate import into oocytes that expressed NPF2.12 or NPF2.13, but this transport did not interfere with the GA transport capabilities as GA accumulation was not affected in GA/nitrate competition assays ( Supplementary Fig. 9b). In continuation, we performed equimolar GA/ABA competition transport assays, which showed that ABA does not affect GA transport, whereas a slight but significant enhancement in ABA uptake was seen in NPF2.13-expressing oocytes when exposed simultaneously to GA 3 . The data show that NPF2.12 and NPF2.13 are multispecific towards nitrate, ABA and GA, and suggests that GA might enhance NPF2.13 ABA transport activity ( Supplementary  Fig. 10). Multispecificity has emerged as an inherent property of the NPF family that is suggested to enable the tantalizing integration of environmental information to the availability of different nutrients 18 . Exhaustive structure-function studies are needed to decipher the molecular basis of the selectivity of these transporters.
To elucidate whether these transporters are part of the GA transport mechanisms in the plant, we treated T-DNA knockout lines, which did not display any visible phenotypes ( Supplementary Fig. 11), with GA 4 -Fl. npf2.12 and npf2.13 single mutant plants treated with GA 4 -Fl displayed a significant reduction in accumulation in the endodermis compared with Col-0 plants (Fig. 3g), similar to reduced levels detected in npf3.1 mutants. To verify that the reduction in GA-Fl accumulation was not due to an off-target mutation, we repeated the test with an additional npf2.12 T-DNA insertion line (npf2.12-2) and obtained similar results ( Supplementary Fig. 12a). In addition, homozygous pNPF2.13:NPF2.13-YFP plants completely rescued the root GA 3 -Fl phenotype ( Supplementary Fig. 12b).

NPF2.12 and NPF2.13 regulate root suberization
To further elucidate the biological function of NPF2.12 and NPF2.13, we generated NLS-YFP and GUS reporter lines to map NPF2.12 and NPF2.13 expression patterns. Confocal microscopy of plants expressing NLS-YFP driven by NPF2.12 and NPF2.13 native promoters revealed that in the root, NPF2.12 was expressed in the pericycle of the whole root and subsequently in the periderm of mature plants; NPF2.13, on the other hand, was expressed only in the shoot ( Fig. 4a and Supplementary Fig. 13). Analysis of pNPF2.12:GUS and pNPF2. 13:GUS lines showed that the two transporters are expressed in the shoot vasculature (Fig. 4b). GA or ABA treatment did not have a major effect on NPF2.12 and NPF2.13 expression level, as detected by qPCR and GUS reporter lines ( Supplementary Fig. 14). The fact that the NPF2.13 translational fusion construct (pNPF2.13:NPF2.13-Venus) introduced into the npf2.13 mutant background rescued the root GA 3 -Fl accumulation phenotype Article https://doi.org/10.1038/s41477-023-01391-3 indicated that the promoter region we cloned is sufficient and that NPF2.13 expression is restricted to the shoot ( Supplementary  Fig. 12b). Similar to npf2.14 mutants, npf2.12 and npf2.13 mutant plants displayed a reduction in suberization. Mutant roots stained with Nile red or Fluorol yellow showed a weaker fluorescence intensity than those in Col-0 plants ( Fig. 4c and Supplementary Fig. 15). In agreement,   Fig. 16). Similar to npf2.14 mutants, we detected a reduction in endodermal suberin levels of npf2.12 and npf2. 13 10-day-old roots and lower levels of cork suberin in 3-week-old npf2.12 npf2.13 double-mutant hypocotyls stained with Fluorol yellow (Supplementary Fig. 17).
To test whether NPF2. 12 and NPF. 13 have partially redundant activities, we generated the npf2.12 npf2.13 double mutant. The phenotype of the npf2.12 npf2.13 double-mutant line was not enhanced compared to the single npf2.12 and npf2.13 mutants (Fig. 4c,d). We further generated additional double and triple mutant combinations with the npf2.14 mutant via CRISPR genome editing (as NPF2.14 is genetically linked to NPF2.13) and tested their activity. We analysed suberin patterning for all genotypes and found that the phenotypes of the higher-order mutant knockouts were not enhanced compared to the single mutants ( Supplementary Fig. 18). This result is in line with the fact that there is limited overlap among the three genes in terms of expression pattern or protein localization. Notably, GA 3 or ABA treatment completely rescued npf2.12 low-suberin phenotype. The phenotypes of npf2. 13 and of the npf2.12 npf2.13 double mutant were completely rescued by ABA and largely rescued by GA 3 (Fig. 4c). Quantification of suberin monomer content revealed that both npf2.12 and npf2.13 mutant roots accumulated ~40% less total suberin contents compared with the Col-0, attributed to lower levels of vanillic and ferulic acids, C22 fatty acid, C20 fatty alcohol, and C18:1(9), C22 and C26 ω-hydroxyacids (Fig. 4e).

NPF2.12 and NPF2.13 regulate shoot-to-root GA translocation
Our previous work showed that GA 12 , although not bioactive, is the primary GA form transported over long distances through the vasculature in Arabidopsis thaliana 12 . GA 12 can move through the xylem in a root-to-shoot manner and in the phloem in a shoot-to-root direction to regulate adaptive plant growth 12,13 . However, the mechanism regulating this process remains unknown 21 . NPF2.13 was expressed strictly in the shoot (Fig. 4a,b), yet the knockout led to a phenotype in the root endodermis (Fig. 4c). This led us to hypothesize that the transporters are involved in the long-distance shoot-to-root translocation of GA. To test whether NPF2.12 and NPF2.13 facilitate shoot-born GA loading into the phloem, we quantified GA content in phloem exudates collected from leaf petioles. The double npf2.12 npf2.13 loss-of-function mutant showed a striking reduction in GA 12 content in the collected phloem exudates (Fig. 5a). Other GA metabolites (GA 15 , GA 24 , GA 9 , GA 4 , GA 34 and GA 51 ) were not significantly reduced (Fig. 5a). ABA levels showed a mild decrease, significant when compared to Col-0 using two-tailed t-tests, but not significant when using Dunnett's multiple comparisons test (P ≤ 0.05). The results imply that NPF2.12 and NPF2.13 regulate GA and possibly ABA, loading into the shoot phloem. In agreement, quantification of active GA 4 (downstream of GA 12 in the biosynthesis pathway) content in the root showed a significant reduction in the npf2.12 and npf2.13 single and double-mutant lines (Fig. 5b). ABA content was also significantly lower in mutant roots compared with Col-0 (Fig. 5b).
To further test this hypothesis, we examined the expression pattern of the two transporters using reporter lines. In mature rosette, NPF2.12 and NPF2.13 were expressed in the shoot apex and the main vascular vein (Fig. 5c). Cross-sections of pNPF2.12:GUS or pNPF2.13:GUS leaf petioles showed that both genes were expressed in the phloem companion cells (Fig. 5c). Next, we investigated whether these transporters are involved in long-distance GA movement from the shoot to the root. For this purpose, 16-day-old plants were grown on paclobutrazol, a GA biosynthesis inhibitor, for 4 d to deplete the plants of native GA and GA 12 was applied to a single leaf. We then quantified the abundance of the DELLA growth repressing protein REPRESSOR OF GA1-3 (RGA, AT2G01570) in the root. DELLA proteins are central inhibitors of GA-regulated processes and GA relieves their inhibiting activity by activating their degradation 54 . Time-course experiments in Col-0 plants showed a significant reduction in RGA accumulation in the root after GA 12 treatment, indicating GA 12 movement from the shoot to the root (Fig. 5d). On the other hand, in the npf2.12 npf2.13 double mutant, RGA abundance remained stable (Fig. 5d), signifying a reduced GA accumulation in the root.
We hypothesize that once GA 12 is translocated to the roots, it can be converted to the bioactive GA 4 by the GA20ox and GA3ox enzymes, which are expressed in the root 55 . Profiling the expression pattern of the GA3ox promoters (AT1G15550, AT1G80340) catalysing the last step in bioactive GA 4 hormone synthesis showed that expression of these enzymes is restricted to the stele as previously reported 55 (Fig. 5f). Together, these results imply that NPF2.12 and NPF2.13 function in GA 12 loading into the phloem for long-distance transport from the shoot to the root, with conversion to GA 4 taking place in the root stele.

Hormone storage in the phloem unloading zone facilitates suberization
To broaden our knowledge of GA and ABA distribution in the root and how it affects suberization, we created a mathematical model to simulate hormone distributions within the root cross-section, extending a modelling framework previously developed to study auxin transport 56 . Using a multicellular template segmented from a root-cross-sectional image ( Supplementary Fig. 21), we incorporated into the model experimentally observed transporter distributions: NPF3.1 on endodermal cell membranes 16,29 , NPF2.12 on pericycle cell membranes (Figs. 3i, 4a) and NPF2.14 on pericycle tonoplasts (Figs. 1k and 2a). The model simulated active hormone transport via NPF3.1, NPF2.12 and NPF2.14, passive hormone transport across both plasma membrane and tonoplast, hormone synthesis and degradation, and hormone diffusion within the apoplast with significantly reduced diffusion in the endodermal apoplast due to the presence of the Casparian strip. To parameterize the model, permeabilities associated with each passive and active transport component were estimated using the oocyte data, and transport rates were then specified on the basis of established pH and membrane potential values for plant cells [56][57][58] (Supplementary Table 5). An important factor is the source of the hormone. ABA2 and AAO3, which encode enzymes necessary for ABA biosynthesis, were previously shown to be expressed in the vasculature 59 . Bioactive GA 4 is also synthesized at high levels in the Arabidopsis stele 55 (Fig. 5f). Considering these data, together with the phloem unloading zone 41 (the docking belt for the long-distance shoot-to-root transported hormones), led us to specify the stele as the source of active GA and ABA in the model.
We first used the model to test the hypothesis that the discovered clade of transporters is necessary and sufficient to explain the observed endodermal hormone accumulation. With all transporters present, the model predicts high levels of both cytoplasmic and vacuolar hormones in the endodermis (Fig. 6a for GA and Supplementary Fig. 22), consistent with the wild-type GA-Fl and ABA-Fl observations ( Fig. 1i and Supplementary Fig. 2) 16 and where suberization later occurs. Mutations in npf3.1 and npf2.12 were predicted to have reduced endodermal hormone cytoplasmic concentrations ( Fig. 6a and Supplementary Fig. 22), in agreement with the loss of endodermal GA-Fl in these mutants (Fig. 3g) 16 and explaining their reduced suberization (Fig. 4c, and Supplementary Figs. 7 and 15-18). We also considered NPF2.13 which is not expressed in the root (Fig. 4a) but contributes to the long-distance translocation of GA 12 . Reducing long-distance translocation via the Article https://doi.org/10.1038/s41477-023-01391-3 npf2.13 mutation can be simulated by reducing the stele-specific synthesis rate, which leads to reduced predicted hormone concentrations throughout the root cross-section ( Fig. 6a and Supplementary Fig. 22), again providing an explanation for the reduction in suberization ( Fig. 4c and Supplementary Fig. 15-18). We concluded that NPF2.12, NPF2.13 and NPF3.1 all play distinct and necessary roles in creating the hormone accumulation within the endodermis that mediates suberization.
We then used the model to investigate the role of the tonoplast-localized NPF2.14. In contrast to the npf2.12, npf2.13 and npf3.1 mutations, simulations of the npf2.14 mutation predicted endodermal cytoplasmic hormone concentrations that are higher than those in the wild type ( Fig. 6a and Supplementary Fig. 22), in agreement with the GA-Fl accumulation at the elongation zone (Fig. 1i). Why npf2.14 exhibits reduced suberization when endodermal accumulation is higher remains unclear. However, the model predictions revealed that the vacuolar concentrations in the npf2.14 pericycle are much lower than in the wild type ( Fig. 6a and Supplementary Fig. 23). These predictions led us to hypothesize that NPF2.14 regulates hormone levels both inside and outside of vacuoles that could provide a hormone source in the maturation zone, where the hormone is no longer supplied by the phloem yet is required for suberization. To test this hypothesis, we simulated the hormone dynamics after cells leave the phloem unloading zone, and found that the presence of NPF2.14 leads to higher predicted endodermal cytoplasmic concentrations as cells mature (Fig. 6b and Supplementary Fig. 24). This suggests that the tonoplast-pericycle-localized NPF2.14 ensures that endodermal hormone concentrations are at levels necessary to mediate suberization. Thus, on the basis of the model, we propose a pericycle-specific slow-release GA and ABA mechanism that explains how the two hormones are loaded into the pericycle vacuoles at the phloem unloading zone 41 and released from these vacuoles later on when the cells are mature (Fig. 6b). In this mechanism, GA and ABA unloaded from the phloem are transported into the pericycle and loaded into the vacuole to form a storage pool. When these cells reach the maturation zone, the GA and ABA that were stored in the pericycle vacuoles are transported by NPF3 into the endodermis to induce suberization.
In conclusion, the mathematical model revealed that the discovered clade of transporters is sufficient to explain the observed hormone distributions, with NPF3.1, NPF2.12 and NPF2.13 playing distinct and necessary roles for endodermal accumulation. Furthermore, the model revealed that the tonoplast-localized NPF2.14 facilitates vacuolar hormone accumulation within the pericycle, providing a source of hormones and enabling the cross-section to maintain high endodermal hormone levels after cells leave the phloem unloading zone 41 . Thus, the model predictions provide mechanistic explanations for the suberization phenotypes observed in the NPF mutants.

Discussion
In this work, we identified NPF2.14, a previously uncharacterized transporter, as a dual-specificity GA and ABA vacuolar transporter. To the The model predicts that NPF2.14 leads to higher endodermal cytoplasmic concentrations, the endodermis being the region where suberin forms (pale red region).
Article https://doi.org/10.1038/s41477-023-01391-3 best of our knowledge, NPF2.14 is presumably the first known subcellular GA/ABA transporter. We showed that NPF2.14 is expressed in the pericycle to facilitate endodermal root suberization. Oocyte experiments showing that NPF2.14 exports GA from the cytosol, combined with NPF2.14's localization to the tonoplast, indicate that NPF2.14 transports GA and ABA from the cytosol into the vacuole. The results presented here suggest that the pericycle serves as a buffer zone, regulating the transitions of hormones from the stele to the endodermis (Fig. 7). The stele acts as the source of bioactive GA and ABA in the root 55,59 . Both hormones have been shown to accumulate and affect their respective responses specifically in the endodermis 25,28,60,61 .
In addition, it appears that the GA and ABA do not simply flow through the pericycle but rather are loaded into the vacuole by NPF2.14 to form a reservoir for later developmental stages. We propose that the high levels of GA and ABA present in the phloem unloading zone are taken into the pericycle by NPF2.12. Once in the pericycle cells, NPF2.14 facilitates their import into the vacuole for storage. We speculate that a slow-release mechanism feeds the differentiating cells with GA and ABA, thus allowing suberin formation in the mature root (Fig. 7). Vacuoles have been proposed to act as storage, modification or degradation compartments for plant hormones 62 . It is possible that tonoplast-localized NPF2.14 mediates the hormonal homeostasis  Once in the pericycle cytoplasm, NPF2.14 imports the hormones into the vacuole to form a reservoir that will be available at later stages. When the root elongates over time and the cells that accumulated high levels of GA and ABA in the vacuoles mature, the hormones are exported out of the pericycle vacuole and imported into the endodermis by NPF3.1 to induce suberization.
Article https://doi.org/10.1038/s41477-023-01391-3 balance that is needed for the proper execution of the developmental plan in the neighbouring cell file. At this point, it is not clear whether GA and ABA are stored in their bioactive form in the vacuole. Confocal imaging of the NLS-YFP lines indicated that both NPF2.12 and NPF2.14 are expressed in the pericycle, mainly at the phloem poles (Figs. 2a and 4a). It is therefore possible that the hormone uptake is not carried out uniformly throughout the pericycle ring, but rather amplified at the pericycle phloem pole cells. If so, do the two hormones retain polar distribution in the pericycle and in the subsequent endodermis layer? Or do the hormones have the ability to move within the two cell file rings? Future mathematical models and genetic work is required to address these questions.
The significantly reduced GA 12 content in npf2.12 npf2.13 double knockout phloem extracts implies that these transporters are required for GA 12 loading into the phloem and translocation of GA 12 from the shoot to the root. Thus, we hypothesize that NPF2.12 and NPF2.13, which are plasma membrane-localized importers expressed in the shoot phloem companion cells, are part of the long-sought mediators of long-distance GA shoot-to root translocation. Our results agree with the previous finding that GA 12 is the main form of GA that is transported long distances through the plant 12 . In addition, we showed that GA3ox1 and GA3ox2, which catalyse the final step of active GA production, are expressed in the root stele 55 (Fig. 5f). Thus, once GA 12 docks at the root stele, it is converted by GA3ox1 and GA3ox2 into the bioactive GA 4 form, which is then delivered from the stele to the endodermis by NPF2.12 and NPF3.1 (Fig. 6c).
It is intriguing that NPF2.12 and NPF2.13 act as GA 12 shoot-to-root transporters and also promote the delivery of the bioactive forms of GA and ABA to the endodermis. In both cases, a plasma membrane import activity is involved, but the substrate specificity differs. Since we showed that NPF2.12 and NPF2.13 can transport both intermediates of and the bioactive forms of GA (Fig. 3b, and Supplementary Figs. 9 and 10) and that GA 12 is present at high levels in the shoot phloem (Fig.  5a), we speculate that GA 12 is the primary substrate of the transporters. In the root stele, NPF2.12 recognizes bioactive ABA and GA 4 , which are present in high concentrations due to being synthesized there 59 .
The presented work show apparently for the first time that GA deficiency results in lower endodermal suberization and can be complemented by GA or ABA. This result is interesting at multiple points of view. First, it may explain the physiological importance of GA accumulation in the endodermis. Second, it suggests that GA and ABA function non-antagonistically to promote endodermis suberization. At this stage, it is not clear whether GA promotes endodermis suberization directly by, for example, direct binding and activation of suberin biosynthesis factors by DELLA or DELLA co-partners, or whether suberization is a secondary effect of maturation or signalling through the ABA components. Our result in oocytes suggesting that GA enhances ABA import may point to another possibility where there is co-activity in regulating suberin formation at the transport level of the two hormones. While GA and ABA are considered antagonistic hormones 63 , it is possible that that the two hormones can act non-antagonistically to promote root suberization by dual-transport activity of the hormones. How GA promotes ABA import biochemically from the transport structure/function point of view is not clear at the moment. The non-antagonistic results between the two hormones agree with the growth defects displayed by biosynthesis mutants of these hormones, which result in small dark green plants [64][65][66] . Thus, it could be that in a specific context where GA and ABA transport plays a role, the two hormones act synergistically.
Plants were sown on vertical plates containing 0.5× Murashige-Skoog (MS) medium, 1% sucrose and 0.8% agar (pH 5.7), stratified for 2 d at 4 °C in the dark, then transferred to growth chambers (Percival CU41L5) at 21 °C and 100 µE m −2 s −1 light intensity under long day light (16 h light/8 h dark). All plants in suberin quantification experiments were grown on 0.5× MS medium, 1% sucrose and 0.8% agar (pH 5.7) for 3 d and subsequently moved to 0.5× MS medium without sucrose supplementation due to phenotype masking by the sucrose treatment. For low-nitrate experiments, plants were sown on nitrate-free MS with vitamins (Caisson labs MSP07-50LT), which was supplemented with 0.01 mM (low nitrate) or 10 mM (high nitrate) KNO 3 .

CRISPR
To generate the CRISPR/Cas9 vector targeting NPF2.14, the MoClo system was implemented. Cloning of the NPF2.14-specific guide (guide sequence: ATTGTTGTCTCGTCGTTAAATCCG) into the system was done according to ref. 67.

Hypocotyl cross-sections
Sectioning and clearing were performed as previously described 44 . Hypocotyls (3-week-old) were fixed in 4% PFA for an hour, rinsed twice in 1× PBS embedded in 5% agarose and sectioned to 150 µM slices using a Leica VT1000S vibratome. Slices were cleared using a ClearSee solution for 5 d. Following clearing, sections were counterstained with 0.1% calcofluor white in ClearSee solution for 30 min. Next, the seedlings were washed in ClearSee for 30 min with gentle shaking. For imaging, sections were mounted directly in ClearSee and imaged using a Zeiss LSM 780 inverted microscope.

Hormone application
Hormone was added to the agar medium at concentrations indicated in the figure legends. Seedlings were either germinated on media or moved after germination to treatment plates. GA-Fl (5 µM) was applied in liquid MS media for 16 h before imaging. For ga1 experiments, both Col-0 and ga1 seeds were imbibed in sterile water containing 5 µM GA 3 for 16 h to induce uniform germination. Following imbibition, seeds were washed three times in sterile water to wash away excess GA and sown on MS plates.

Cloning of NPFs overexpression and reporter lines
NPF2.12 and NPF2.14 coding sequences were synthesized by Bio Basic, cloned into pENTR/D-TOPO (Invitrogen K2400) and subsequently cloned into the pH7YWG2 destination vectors using the LR Gateway reaction (Invitrogen 11791). NPF2.12 and NPF2.14 promoters were amplified with the primers listed in Supplementary Table 2 using a Phusion high-fidelity polymerase (New England Biolabs), cloned into pENTR/D-TOPO, and then cloned into pMDC7 vector for NLS-YFP reporters and pGWB3 vector for GUS reporters.

Imaging and analysis
Seedlings were stained in 10 mg l −1 propidium iodide (PI) for 5 min, rinsed and mounted in water. Seedlings were imaged using a laser Article https://doi.org/10.1038/s41477-023-01391-3 scanning confocal microscope (Zeiss LSM 780 inverted microscope), with argon laser set at 488 nm for fluorescein, 514 nm for YFP and 561 nm for PI excitation. Emission filters used were 493-548 nm for fluorescein derivatives, 508-570 nm for YFP and 583-718 nm for PI emission. Image analysis and signal quantification were done with the measurement function of ZEN lite software. The number of quantified biological repeats and sampling points is indicated for each graph in figure legends. All statistical analyses and graphs were made using GraphPad Prism v8.

Root length characterization
For root length measurements, seedlings were imaged using a Zeiss Stemi 2000-C stereo microscope and measured using ImageJ software (http://rsbweb.nih.gov/ij/index.html).
For cross-sectioning of GUS-stained leaf petioles, after clearing in 70% ethanol, the samples were fixed in FAA solution (3.2% formaldehyde, 5% acetic acid, 50% ethanol) for 30 min and kept overnight at 4 °C. The samples were then dehydrated in an ethanol gradient ranging from 50% to 96% and incubated in 2% eosin overnight at 4 °C. After several washes in 96% ethanol, the samples were progressively rehydrated in ethanol/HISTO-CLEAR II (Electron Microscopy Sciences) solution, incubated in 50% HISTO-CLEAR II 50% PARAPLAST PLUS (McCormick Scientific) at 60 °C for 2 h and embedded in 100% PARAPLAST PLUS. Paraffin-embedded samples were cross-sectioned with a LEICA RM2155 microtome and imaged using a Leica Leitz Dmrb microscope.

Nile red suberin staining, imaging and quantification
Nile red suberin staining was performed as previously described 44 . Briefly, 5-day-old seedlings were fixed in paraformaldehyde for 1 h under gentle agitation and washed twice in phosphate-buffered saline at pH 7.4. Plants were covered in filtered 0.05% Nile red (Acros Organics, 7385-67-3) solution dissolved in ClearSee for 16 h. Following staining, plants were washed three times in ClearSee for 30 min per wash. Next, plants were counterstained with 0.1% calcofluor white (Glentham Life Sciences, 4404-43-7) dissolved in ClearSee for cell wall imaging. After 30 min, plants were washed in ClearSee for 30 min. Plants were mounted directly in ClearSee on slides and imaged with a Zeiss LSM 780 confocal microscope. Images were taken from the upper part of the root and under the root-hypocotyl junction with an argon laser set at 514 nm for Nile red excitation and 405 nm for calcofluor excitation. Emission filters used were 561-753 nm filter for Nile red and 410-511 nm filter for calcofluor emission. Fluorescence intensity was assessed from 5 endodermal cells per root using the Zen software.
For root patterning, following Nile red staining, roots were imaged using an AxioZoom 16 Zeiss binocular microscope. Root length and length of continuously and patchily suberized zones were measured using ImageJ. Percent of suberized area was normalized to total root length. All statistical analyses and graphs were made using GraphPad Prism v8.

RT-qPCR
Total RNA was isolated from the indicated plant materials using RNeasy Plant mini kit (QIAGN 74,904). DNA was removed by RQ1 RNase-free DNase (Promega M6101). Total RNA (2 µg) was converted to complementary DNA (cDNA) using M-MLV Reverse Transcriptase (Promega M1701) with oligo(dT)15 primer according to manufacturer protocols. RT-qPCR was performed with 40 ng cDNA in a final volume of 10 µl with Fast SYBR Green Master Mix (ABI 4385612) using the Step One Plus system and software (ABI). The reaction conditions included 40 amplification cycles (3 s at 95 °C, 30 s at 60 °C). Three technical repeats were performed for each cDNA sample, and at least three biological repeats were used for each treatment. Relative quantification was calculated using the ΔΔCt method, with PP2A used as the reference gene. Primers are specified in Supplementary Table 3.

Transport assays in Xenopus oocytes
Coding sequences were cloned into the pNB1u vector, and complementary RNA (cRNA) was produced as previously described 34 . Xenopus oocyte assays were performed as previously described 34 . Defolliculated X. laevis oocytes (stage V-VI) were purchased from Ecocyte Biosciences, injected with 25 ng cRNA in 50.6 nl using a Drummond Nanoject II and incubated for 2-4 d at 16 °C in HEPES-based kulori (90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES (pH 7.4)) before use. Oocytes were pre-incubated in MES-based kulori (90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM MES (pH 5)) for 4 min and then transferred to phytohormone-containing MES-based kulori for 60 min. After washing three times in 25 ml HEPES-based kulori followed by one wash in 25 ml deionized water, oocytes were homogenized in 50% methanol and stored for >30 min at −20 °C. Following centrifugation (25,000 x g for 10 min 4 °C), the supernatant was mixed with deionized water to a final methanol concentration of 20% and filtered through a 0.22 µm filter (MSGVN2250, Merck Millipore) before analytical LC-MS/MS as described below. For nitrate assays, sodium chloride in kulori was substituted for equimolar sodium nitrate so as not to affect the membrane potential.

Quantification of phytohormone content by LC-MS/MS
Compounds in the diluted oocyte extracts were directly analysed by LC-MS/MS. The analysis was performed following a previously described method 16 with modifications. In brief, chromatography was performed on an Advance UHPLC system (Bruker). Separation was achieved on a Phenomenex Kinetex 1.7u XB-C18 column (100 × 2.1 mm, 1.7 µm, 100 Å) with 0.05% (v/v) formic acid in water as mobile phase A and acetonitrile with 0.05% formic acid (v/v) as mobile phase B. The gradients used for elution of GAs were 0-0.5 min, 2% B; 0.5-1. The mobile phase flow rate was 400 µl min −1 and column temperature was maintained at 40 °C. The liquid chromatography was coupled to an EVOQ Elite triple quadrupole mass spectrometer (Bruker) equipped with an electrospray ion source operated in positive and negative ionization mode. Instrument parameters were optimized by infusion experiments with pure standards. For analysis of GAs, the ion spray voltage was maintained at +4,000 V and −4,000 V in positive and negative ionization mode, respectively, and the heated probe temperature was set to 200 °C with probe gas flow at 50 psi. For ABA, the ion spray voltage was maintained at −3,300 V in negative ionization mode, and heated probe temperature was set to 120 °C with probe gas flow at 40 psi. Remaining settings were identical for all analytical methods with cone temperature set to 350 °C and cone gas to 20 psi. Nebulizing gas was set to 60 psi and collision gas to 1.6 mTorr. Nitrogen was Article https://doi.org/10.1038/s41477-023-01391-3 used as probe and nebulizing gas, and argon as collision gas. Active exhaust was constantly on. Multiple reaction monitoring was used to monitor analyte parent ion to product ion transitions for all analytes. Multiple reaction monitoring transitions and collision energies were optimized by direct infusion experiments. Detailed values for mass transitions can be found in Supplementary Table 4. Both Q1 and Q3 quadrupoles were maintained at unit resolution. Bruker MS Workstation software (v8.2.1) was used for data acquisition and processing. Linearity in ionization efficiencies was verified by analysing dilution series of standard mixtures. Sinigrin glucosinolate was used as internal standard for normalization but not for quantification. Quantification of all compounds was achieved using external standard curves diluted with the same matrix as the actual samples. All GAs were analysed together in a single method. GA 12 suffered from severe ion suppression when combined with the other GAs in the standard curve, thus quantification was not achieved for GA 12 .

Root suberin monomer profiling by GC-MS
Suberin monomers were extracted from Col-0 and mutant roots according to previously described protocols 69,70 . A sample volume of 1 µl was injected in splitless mode on a GC-MS system (Agilent 7693A Liquid Auto injector, 8860 gas chromatograph and 5977B mass spectrometer). GC was performed (HP-5MS UI column; 30 m length, 0.250 mm diameter and 0.25 µm film thickness; Agilent J&W GC columns) with injection temperature of 270 °C, interface set to 250 °C and the ion source to 200 °C. Helium was used as the carrier gas at a constant flow rate of 1.2 ml min −1 . The temperature programme was 0.5 min isothermal at 70 °C, followed by a 30 °C min −1 oven temperature ramp to 210 °C and a 5 °C min −1 ramp to 330 °C, then kept constant during 21 min. Mass spectra were recorded with an m/z of 40 to 850 scanning range. Chromatograms and mass spectra were evaluated using the MSD ChemStation software (Agilent). Integrated peaks of mass fragments were normalized for sample dry weight and the respective C32 alkane internal standard signal. For identification, the corresponding mass spectra and retention time indices were compared with the NIST20 library as well as in-house spectral libraries.

Xenopus oocyte injection-based efflux transport assays and competition assays
For injection-based export assays, on the second day of gene expression, oocytes were injected with 23 nl 8.2 mM in 98 mM KCl, 1 mM CaCl 2 and 10 mM HEPES (pH 7.4). T1 oocytes were left to heal for 10 min and then transport was evaluated as described above. T2 oocytes were left for approximately 20 h in HEPES-based kulori at 16 °C, followed by transport analysis.

Quantification of nitrate from oocytes by HPLC
Nitrate concentration in the oocyte extracts was quantified using a Dionex ICS-2100 anion exchange chromatography system (Thermo Fisher). The separation was done on a Dionex IonPac AG11-HC analytical column coupled to the AS11-HC guard column (Thermo Fisher). The columns were connected to a Dionex AERS 500 anion suppressor (Thermo Fisher). The analyses were performed under the following conditions: sample injection volume of 4.8 µl, column temperature of 30 °C, flow rate of 0.38 ml min −1 , isocratic eluent gradient using 30 mM KOH solution in QH 2 O, suppressor current of 29 mA and runtime of 15 min. Nitrate detection was done at 220 nm using a Dionex UltiMate 3000 (Thermo Fisher). QH 2 O water dilutions of Dionex Combined Seven Anion Standard (Thermo Fisher) were used to create a standard calibration curve. Accuracy and precision of the quantification was checked by including samples of potassium nitrate throughout the sequence.

pH measurements of oocyte lumen
Stabilization of pH was performed as described previously 34 . pH electrodes were pulled from borosilicate glass capillaries (KWIK-FIL TW F120-3 with filament) on a vertical puller (Narishige), baked for 120 min at 220 °C and silanized for 60 min with dimethyldichlorosilane (Silanization Solution I, Sigma Aldrich). Electrodes were backfilled with a buffer containing 40 mM KH 2 PO 4 , 23 mM NaOH and 150 mM NaCl (pH 7.5). The electrode tip was filled with a proton-selective ionophore cocktail (hydrogen ionophore I cocktail A, Sigma Aldrich) by dipping the tip into the cocktail. Oocytes, as described above, were placed in freshly made HEPES-based ekulori (2 mM LaCl 3 , 90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM HEPES (pH 7.4)) for at least 30 min before three-electrode voltage clamp experiments. Before each oocyte measurement, a pH calibration curve was made for each oocyte using 100 mM KCl pH 5.5, 100 mM KCl pH 6.5 and 100 mM KCl pH 7.5. Oocytes were clamped at 0 mV and perfused with HEPES-based ekulori pH 7.4, followed by MES-based ekulori (2 mM LaCl 3 , 90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM MES (pH 5)) and internal pH response was measured continuously as a function of external pH change.

Membrane potential measurements
Membrane potentials of oocytes were measured in ekulori (90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM MES, 2 mM LaCl 3 (pH 5)) using the automated two-electrode voltage clamp system, Robo-ocyte2 (Multi Channel Systems), with electrodes backfilled with 1 M KCl and 1.5 M potassium acetate. All oocytes were measured using the same electrodes with a resistance of 280-350 kΩ. The experiment was terminated when the resistance of one of the electrodes shifted to approximately 600 kΩ.

Two-electrode voltage clamp electrophysiology
The electric signal elicited by GA treatment of oocytes was measured in ekulori (90 mM NaCl, 1 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM MES, 2 mM LaCl3 (pH 4.5 or pH 5)) using Roboocyte2 (Multi Channel Systems), with electrodes (resistance 280-1,000 kΩ) backfilled with 1 M KCl and 1.5 M potassium acetate. Oocytes were clamped at −60 mV, and IV (Current-Voltage) curves were obtained before and after substrate addition. Substrate dependent currents were calculated by subtracting currents before addition of substrate from currents after addition of substrate.

Root hormone quantification
Hormone extraction and analysis was performed as previously described 33 . Standards (both labelled and non-labelled) were obtained from Olchemim and the National Research Council (NRC-CNRC, Canada). Standard-grade solvents methanol, acetic acid (LiChrosolv, Sigma Aldrich), acetonitrile ( J.T.Baker), formic acid (Honeywell Fluka, Thermo Fisher) and deionized water (Milli-Q, Synergy-UV millipore system) were used for sample preparation. Briefly, root tissue frozen in liquid nitrogen was grounded using a mortar and pestle. Around 200 mg of root sample was measured from ground powder and extracted with ice-cold methanol:water:formic acid (15:4:1 v/v/v) added with deuterium-labelled internal standards. Similar concentrations of internal standards of abscisic acid and gibberellin (GA 4 ) were added into samples and calibration standards. The samples were purified using Oasis MCX SPE cartridges (Waters) according to manufacturer protocol. The samples were injected on an Acquity UPLC BEH C18 column (1.7 µm, 2.1 × 100 mm (Waters); with gradients of 0.1% acetic acid in water or acetonitrile) connected to an Acquity UPLC H class system (with Waters Acquity QSM, FNR sample manager and PDA) coupled with a UPLC-ESI-MS/MS triple quadrupole mass spectrometer (Xevo TQ-S (Waters), equipped with an ESI probe) for identification and quantification of hormones. The hormones were measured using an MS detector, both in positive and negative mode, with two MRM (Multiple Reaction Monitoring) transitions for each compound. External calibration curves were constructed with hormone standards added with internal standards, used for quantification and calculated using Target Lynx (v4.1; Waters) software by Article https://doi.org/10.1038/s41477-023-01391-3 comparing the ratios of MRM peak areas of the analyte to those of the internal standard.

Phloem extract and hormone quantification
Rosette leaves of 5-week-old Col-0 and npf2-12 npf2-13 mutant plants (before bolting) were cut with a razor blade at the base of the petiole, and each leaf was dipped in a tube containing 80 µl of exudation buffer (50 mM potassium phosphate buffer (pH 7.6), 10 mM EDTA). Exudation was carried out for 3 h in the dark at high humidity to limit transpiration. Exudation of 75 leaves was regrouped and concentrated under vacuum centrifugation. Hormone contents of phloem exudates were determined using UPLC-MS/MS (Waters Quattro Premier XE). Concentrated residue of phloem sap was resuspended in 80% methanol-1% acetic acid including 17-2 H 2 -labelled GA internal standards (Olchemim), mixed and passed through an Oasis HLB column . The dried eluate was dissolved in 5% acetonitrile-1% acetic acid, and the GAs were separated by UHPLC chromatography (Accucore RP-MS column 2.6 µm, 100 × 2.1 mm; Thermo Fisher) with a 5 to 50% acetonitrile gradient containing 0.05% acetic acid, at 400 µl min −1 over 22 min. The concentrations of GAs in the extracts were analysed with a Q-Exactive mass spectrometer (Orbitrap detector; Thermo Fisher) by targeted SIM (Selected Ion Monitoring) using embedded calibration curves and the Xcalibur 2.2 SP1 build 48 and TraceFinder programmes.

Grafting assays
Grafting was performed without collars on water imbibed 0.45 µM MCE membrane (Millipore) between hypocotyls of rootstocks and scions of 6-day-old seedlings grown on 1× MS agar plate. Grafted seedlings were then kept vertically to recover for 5 d under constant humidity. Successful grafts were transferred onto ½× MS agar plates and grown under a 16 h photoperiod at 22 °C. Root growth was measured every day for 3 d with ImageJ (https://imagej.nih.gov/ij/download.html). Nile red suberin staining fluorescence intensity was assessed as previously described, in roots of 13-day-old grafted seedlings 2 d after transfer onto ½× MS agar plates.

DELLA degradation assays
Seedlings (12-day-old) were transferred to 1× MS agar modified medium without nitrogen (bioWORLD plant media) supplemented with 0.5 mM KNO 3 and 1 µM paclobutrazol (Sigma). At 4 d after transfer, a drop of GA 12 (5 µl at 1 µM) was placed on one of the first two leaves formed. Roots were collected at 6, 12 and 24 h after adding GA 12 . Total proteins were extracted in 2× SDS-PAGE sample buffer and separated on 10% SDS-PAGE gel. After transfer onto membranes, immunoblots were performed using a 2,000-fold dilution of anti-RGA (Agrisera) and a 10,000-fold dilution of peroxidase-conjugated goat anti-rabbit (Thermo Fisher). Signals were detected with Fusion FX (Vilber) using Immobilon Forte Western HRP Substrate (Millipore). The blot was subsequently stained with Coomassie blue. Quantification of the signals was performed using ImageJ.

Mathematical model
Root templates were segmented from an experimental image using the CellSeT image analysis tool 71 (Supplementary Fig. 21). We used CellSeT to manually assign a cell type to each cell and then read the geometrical and cell-type data into a tissue database (based on the OpenAlea tissue structure 72 ), extending the data structure to incorporate vacuolar compartments within each cell. The geometrical, topological and transporter-distribution data were used to form a system of ordinary differential equations to describe the GA transport, synthesis and degradation within the multicellular root cross-section. Parameters associated with the passive and transporter-mediated transport components were estimated using the oocyte data (Figs. 1a and 3b) and the remaining parameter values were obtained from the literature (Supplementary Table 5). These ordinary differential equations were simulated using the solve_ivp package in Python 3.6.5. Full details of the model equations and assumptions are provided as Supplementary text.

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

Data availability
All the data supporting the findings of this study are available within the Article and its Supplementary Information. Source data are provided with this paper. The python code used to produce the model results is available at: https://gitlab.com/leahband/ ga_aba_transport_rootcrosssection_model.