Continuous root xylem formation and vascular acclimation to water deficit involves endodermal ABA signalling via miR165

Roots are responsible for water uptake and transport through their xylem strands. Although well-known responses to water-limiting conditions include a rapid induction of abscisic acid (ABA) levels (Xiong and Zhu, 2003), stomata closure and growth adjustments (Osakabe et al., 2014), little is known about how the water transporting tissue itself is affected. Understanding how and by what mechanisms xylem development adjusts to water deficit raises the possibility of developing crop-breeding strategies to produce varieties that are better able to tolerate limited water availability.

In primary development, the xylem tracheary elements form as thin protoxylem strands with annular or spiral secondary cell walls (SCWs) or as wider metaxylem with reticulated or pitted SCWs. In the root of Arabidopsis thaliana, xylem identity is determined by transcription factors (TFs) of the class III homeodomain leucine zipper (HD-ZIP III) family (Carlsbecker et al., 2010). The HD-ZIP III gene family encompasses PHABULOSA (PHB), PHAVOLUTA (PHV), REVOLUTA (REV), Arabidopsis thaliana HOMEOBOX GENE 8 (ATHB8) and CORONA (CNA/ATHB15), which are post-transcriptionally controlled by microRNA165 (miR165) and miR166 (Reinhart et al., 2002; Rhoades et al., 2002). Genes encoding these miRNAs are transcriptionally activated in the endodermis, which surrounds the vasculature. miR165/166 then moves into the vasculature to restrict HD-ZIP III levels (Carlsbecker et al., 2010; Miyashima et al., 2011). This leads to lower peripheral HD-ZIP III levels, resulting in protoxylem formation, and higher central levels, resulting in metaxylem formation. Thus, the wild-type xylem axis develops two protoxylem strands flanking three central metaxylem strands (Fig. 1A), a pattern that is sensitive to miR165/166 level changes (Carlsbecker et al., 2010; Miyashima et al., 2011; Müller et al., 2016). Interestingly, miR165/166 levels are affected by abiotic stresses in a wide range of species (Eldem et al., 2012; Giusti et al., 2017; Kantar et al., 2010), suggesting that ABA-induced conditions confer developmental changes via miR165/166. Here, we address the role of ABA in Arabidopsis root xylem formation and in its crosstalk with miR165-HD-ZIP III developmental regulation, and the effects of water limitation on xylem morphology.

To examine the potential importance of ABA for root xylem development, we analysed ABA deficient and signalling mutants (Fig. 1). ABA2 and ABA3 encode proteins involved in the final two steps in ABA biosynthesis, and aba2-1 and aba3-1 therefore have substantially reduced basal ABA levels (Léon-Kloosterziel et al., 1996). Tracing the protoxylem and metaxylem strands of primary roots of 5-day-old aba2-1 and aba3-1 mutants revealed that 40-60% of roots analysed displayed discontinuous or absent xylem strands, primarily metaxylem (Fig. 1B,C and Figs S1,S2). No other apparent anatomical differences were detected, apart from the quiescent centre (QC) divisions previously described by Zhang et al. (2010), and both mutants had significantly shorter roots than wild type (Fig. S2C). Treatment for 72 h with 5 nM ABA was sufficient to restore the xylem defects of aba2-1 (Fig. 1B,C), whereas higher concentrations resulted in extra protoxylem strands next to the ordinary and seen as double protoxylem (Fig. 1C; see also below). Furthermore, wild-type plants treated for 72 h with the ABA biosynthesis inhibitor fluridone (Flu, 1 µM; Bartels and Watson, 1978) displayed discontinuous xylem (Fig. 1D,E), phenocopying ABA-biosynthesis mutants. Supplementing with 250 nM ABA alleviated these effects (Fig. 1E). Similarly, abi1-1, which carries a gain-of-function mutation in the ABI1 PP2C co-receptor phosphatase that renders the plant partially ABA insensitive (Leung et al., 1994; Meyer et al., 1994), displayed discontinuous metaxylem in both Col-0 and Ler backgrounds (Fig. 1F,G, Fig. S2D,E). Neither in mutants nor upon Flu application is it likely that ABA levels or signalling are completely suppressed, possibly explaining the partial penetrance of the xylem phenotype; it is further likely that non-related pathways act redundantly with ABA to ensure xylem continuity. Nonetheless, our results suggest that ABA synthesis and signalling are crucial for proper root xylem development.

To test the effects of elevated ABA levels, we treated 2-day-old wild-type seedlings for 72 h with varying concentrations of ABA. Root growth was compromised by ABA concentrations of 1 µM and above, as previously reported (Ghassemian et al., 2000), but was not affected by 0.5 µM ABA (Fig. S3A). In 60% of the roots, 0.5 µM ABA caused double protoxylem: the extra strand forming either aligned with or at an angle to the xylem axis (Fig. 1H,I, Fig. S1D). Furthermore, metaxylem SCWs became reticulate as opposed to pitted (Fig. 1H, Fig. S1D), and extra metaxylem appeared next to the xylem axis or prematurely in central positions (Fig. 1H, Fig. S1D). Higher ABA concentrations produced similar results (Fig. 1I). The abi1-1 mutation prevented double protoxylem and reticulate metaxylem formation beyond what was already present in this mutant (Fisher's exact test, P<0.05) (Fig. 1G), suggesting that the ABA-induced xylem morphology changes were mediated via canonical ABA signalling, as defined by Cutler et al. (2010).

To determine whether the observed xylem morphology changes were the result of cell fate changes, we analysed the effect of ABA on the pARABIDOPSIS HISTIDINE PHOSPHOTRANSFER PROTEIN6::GFP (pAHP6::GFP) reporter, which normally marks protoxylem plus neighbouring pericycle cell files within the meristem (Mähönen et al., 2006), and pTARGET OF MONOPTEROS5::n3GFP, which marks the immature xylem axis (De Rybel et al., 2013). Following 24 h of 0.5 µM ABA treatment, xylem changes were visible in 60% of 4-day-old plants (Fig. S3B), and already after 12 h treatment we detected pAHP6::GFP in double protoxylem cell files in 9/19 (47%) roots (Fig. 1J). Similarly, a 48 h ABA treatment increased the number of pTMO5::n3GFP-expressing cells to six (n=13/24, 54%), as opposed to five in mock-treated plants (n=18/21, 85%) (Fig. S3C). In addition, expression of the protoxylem-differentiation marker pVASCULAR NAC DOMAIN7::YFP (Kubo et al., 2005) expanded upon short-term high-concentration ABA treatment (Fig. S3D). These results suggest that ABA affects xylem cell fate.

ABA is predominantly localized to the endodermis (Ondzighi-Assoume et al., 2016). To determine whether ABA locally or non-cell autonomously influences xylem cell fate, we used UAS::abi1-1 (Duan et al., 2013), crossed with GAL4 enhancer-trap lines, to reduce ABA signalling tissue specifically (Fig. 2A-E and Fig. S4). Assessing the xylem-pole pericycle J0121≫abi1-1, and the columella and lateral-root cap specific J3411≫abi1-1 transactivation lines revealed that neither affected xylem morphology (Fig. 2F; Fig. S4). However, both Q2500≫abi1-1, which is active in pericycle, endodermis and weakly in cortex, and the ground tissue-specific J0571≫abi1-1 line displayed discontinuous xylem and double protoxylem, similar to abi1-1 mutants, whereas the stele-specific Q0990≫abi1-1 line did not affect xylem development (Fig. 2B-D,F,I). Intriguingly, the epidermal/lateral root cap J0951≫abi1-1 line also displayed discontinuous xylem strands and double protoxylem (Fig. 2E,F). Furthermore, two independent homozygous endodermis/QC-specific pSCARECROW::abi1-1 (pSCR::abi1-1) lines (Duan et al., 2013) displayed xylem breaks (Fig. 2G,H), supporting the notion that reducing ABA signalling in the endodermis is sufficient to alter normal xylem development.

To assess whether tissue specific ABA signalling inhibition would influence the effect of ABA treatment, we analysed the response of the cell-type-specific abi1-1 driver lines to 0.5 µM ABA treatment. Stele-specific ABA signalling reduction by Q0990≫abi1-1 was not sufficient to affect the formation of double protoxylem resulting from the ABA application (Fig. 2I). In contrast, the ground-tissue driver line J0571≫abi1-1 displayed significantly reduced sensitivity to the ABA treatment (Fisher's exact test, P<0.05) (Fig. 2I). These results suggest that ABA predominantly affects stele vascular patterning non-cell-autonomously.

Because miR165 is a well-known signal emanating from the endodermal cell layer to determine vascular patterning (Carlsbecker et al., 2010; Miyashima et al., 2011), we asked whether miR165 could respond to altered ABA levels. Indeed, treatment with 10 µM Flu for 24 h significantly reduced pMIR165A::GFP fluorescence intensity (Fig. 3A,B), indicating a requirement of ABA for proper MIR165A expression. Similarly, reducing endodermal ABA signalling lowered mature miR165 levels (Fig. 3C) and led to a significant increase in ATHB8 and REV transcript levels (Fig. 3D). Conversely, 4 h treatment with 50 µM ABA resulted in significant upregulation of pMIR165A::GFP (Fig. 3E,F), accompanied by a slight but significant increase in mature miR165 levels (Fig. 3G), and significantly reduced PHB and ATHB8 expression levels (Fig. 3H). We have previously shown that a weak miR165-induction line that suppresses HD-ZIP III transcript levels to a similar extent is sufficient to cause protoxylem formation in metaxylem position (Müller et al., 2016). Thus, these data suggest that miR165 constitutes a non-cell autonomous endodermis-derived signal affecting xylem development upon alterations in ABA levels.

To further substantiate the involvement of HD-ZIP III TFs in ABA-mediated xylem morphology changes, we took a genetic approach, reasoning that loss-of-function HD-ZIP III mutants would display less pronounced phenotypic effects if HD-ZIP III TFs are decisive for the xylem phenotypes under low ABA levels. To test this, we treated HD-ZIP III single and double mutants for 72 h with 1 µM Flu. Compared with wild type, all mutants displayed significantly reduced sensitivity to the Flu treatment in terms of frequency of occurrence of discontinuous metaxylem (Fisher's exact test, P<0.05) (Fig. S5A). Furthermore, the Flu treatment decreased root length in wild type, but significantly less in phb-13 (Fig. S6A). Thus, these results suggest that HD-ZIP III TFs are needed for the vascular aberrations inflicted by the Flu treatment. Conversely, if ABA suppresses HD-ZIP III expression, we expected an enhanced ABA response in HD-ZIP III mutants compared with wild type. We found that although rev-6 responded similarly to wild type after 72 h of 0.5 µM ABA treatment, all other tested mutants, and particularly athb8 and athb8 phb, were more sensitive to the ABA treatment than wild type (Figs S5A and S6B,C). These results are consistent with our previous findings that mutations in PHB and ATHB8 especially dictate formation of double protoxylem (Carlsbecker et al., 2010). The phenotypic enhancement of the HD-ZIPIII single and double mutant xylem phenotypes, which mimic higher order HD-ZIP III mutant phenotypes, suggests that ABA negatively affects multiple HD-ZIP III factors.

Taken together, perceived ABA levels affect MIR165A activity in the endodermis, leading to alterations in mature miR165 levels and changes in levels of certain HD-ZIP III TFs in the stele, which, at least in part, explains the altered xylem morphology observed under low or high ABA levels or signalling.

Treatment with exogenous ABA is an often-assumed proxy for abiotic stress, such as limited water availability, which normally rapidly induce elevated endogenous ABA levels (Xiong and Zhu, 2003). However, it is possible that exogenous ABA treatments confer different or non-natural effects. We therefore asked whether limited water availability would cause a similar phenotypic response as application of ABA. To reduce available water under in vitro growth, we overlaid agar plates with polyethylene glycol (PEG), following Verslues and Bray (2006), using three different PEG concentrations: 250, 400 and 550 g/l, thereby reducing the water potential to −0.5, −1 and −1.6 MPa, respectively (Fig. S7A). Growth for 72 h on 550 g/l PEG resulted in root growth inhibition in 6-day-old-plants, whereas lower concentrations did not (Fig. 4A). All three PEG concentrations caused double protoxylem and reticulate metaxylem, similar to treatments with ABA (Fig. 4B,C). After 24 h of growth on 550 g/l PEG, 23% displayed double protoxylem or reticulate metaxylem; after 48 h, 53% displayed these phenotypes (Fig. S7B). In line with the double protoxylem formation, pAHP6::GFP expanded in 9/22 plants after 12 h growth on 550 g/l PEG (Fig. 4D). Similarly, the number of pTMO5::n3GFP-expressing cells increased to six after 48 h of growth on 550 g/l PEG (10/23, 43%) compared with mock, which had five GFP-expressing cells (17/17, 100%) (Fig. S7C). Thus, these results suggest that reduced water availability affects xylem cell identity similarly to exogenous ABA treatment.

The similarities in phenotypes between plants exposed to ABA and low water availability prompted us to investigate whether similar molecular signalling was involved. Reducing ABA signalling using pSCR: :abi1-1 was sufficient to suppress the xylem phenotype effects of growth on 400 g/l PEG (Fisher's exact text, P<0.05) (Fig. 4E), indicating that endodermal ABA signalling is similarly required for the response to water limitations as for the response to ABA. Furthermore, growth on 550 g/l PEG resulted in a significant upregulation of pMIR165A::GFP after 6-8 h (Fig. 4F,G), whereas the mRNA levels of PHB, CNA, ATHB8 and REV were reduced after 6 h treatment (Fig. 4H). Moreover, the effects of growth on 250 g/l PEG were greatly enhanced in phb such that 61% of the phb seedlings displayed double protoxylem compared with 25% for wild type (Fig. S7D). Thus, as summarized in the model in Fig. S8, these results suggest that water deficit invokes alterations in xylem morphology mediated by an ABA-transmitted signal in the endodermis that enhances the levels of miR165, which in turn negatively affects HD-ZIP III TF levels within the stele; this results in changes in xylem development.

Plant survival hinges on adequate responses towards adverse conditions. Here, we show that exposure to water-limiting conditions causes major changes to the root xylem morphology. Similar changes have also been observed in stems of poplar, where water limitation resulted in a decrease in vessel diameter, an increase in vessel number and increase in vessel wall thickness, providing an increased tolerance to cavitation and thereby conferring resistance to water deficit (Awad et al., 2010; Arend and Fromm, 2007). Similarly, soybean plants under water stress increase the number of metaxylem strands, thereby enhancing root hydraulic conductivity (Prince et al., 2017). Upon ABA and PEG treatments, we found that, apart from the formation of extra protoxylem, the metaxylem SCW pattern shifted from pitted to reticulate. A recent study investigated the anatomical characteristics of maize leaf xylem cells upon dehydration and concluded that embolisms formed at a higher rate in metaxylem compared with protoxylem, correlating with the highly hydrophobic SCW pattern of metaxylem cells (Ryu et al., 2016). It will be interesting to test whether the vascular alterations in the Arabidopsis root grown under water deficit conditions similarly facilitate water transport.

Cell-type-specific transcriptomic changes have previously been observed in response to various stresses (Dinneny et al., 2008; Geng et al., 2013; Iyer-Pascuzzi et al., 2011). In this study, we found that reducing ABA signalling either in the ground tissue or the epidermis affects stele cell morphology. As the endodermis is the predominant localization for ABA (Ondzighi-Assoume et al., 2016), we focused on signalling mechanisms from endodermis into the stele, but the effects of reduced epidermal ABA signalling indicate the presence of yet another non-cell-autonomous molecular mechanism through which ABA affects xylem development. The importance of the endodermis as a signalling hub in mediating stress responses is emphasized by studies showing massive endodermal transcriptome responses and its influence on lateral root formation under salt stress (Duan et al., 2013). The endodermis coordinates cell elongation among tissue layers and undergoes a specific differentiation program, subjected to influences from cues such as nutrient and water availability, to become a barrier between tissues directly exposed to the surrounding environment and those allowing systemic movement of molecules (Doblas et al., 2017; Geldner, 2013). This provides a unique opportunity for the endodermis to act as a signalling centre in the root, within the root meristem, as well as in more differentiated tissues, relaying external stimuli to the interior of the plant. In our study, we contribute with an example of such a signal, miR165, that directly affects the formation of the water-transporting tissue in response to water availability.

Seeds were surface sterilized using chlorine gas for 1-2 h or in 15% bleach solution, imbibed and stratified for 48 h at 4°C, and plated on square plates containing 0.5×MS with 1% (w/v) phytagel or 0.5% (w/v) gelrite, and 0.5% sucrose. For treatments, the plants were germinated on media containing sucrose and then transferred to the treatment plates without sucrose. For experiments in which plants had to be transferred or for collection of roots for expression analysis, seedlings were grown on 25 µm pore Sefar Nitex 03-25/19 mesh as previously described (Müller et al., 2016). The plates were placed vertically in a Sanyo MLR growth cabinet at 22°C and 18 h light and 6 h darkness. For ABA (Sigma) and fluridone (Sigma) treatments, stock solutions of 50 mM and 10 mM in 95% ethanol were used to make media supplemented with different ABA and Flu concentrations. For simulating water-limiting conditions the PEG-overlay method was used (Verslues and Bray, 2006). Briefly, 40 ml of medium without sucrose was poured and allowed to solidify. The solid medium was overlaid with liquid 0.5×MS containing different concentrations (250, 400 or 550 g/l) of PEG 8000 (Sigma). The plates were left overnight and excess PEG solution was then discarded before transfer of plants to the plates. Water activity of the different PEG overlay media was measured using Aqualab Pre Water Activity Meter and results were used to calculate the water potential.

For phenotypic and expression analysis, plants were analysed 5 days after germination, unless otherwise specified. Mutants used in this study include aba2-1 and aba3-1 (Léon-Kloosterziel et al., 1996), abi1-1C (Kanno et al., 2012), two lines of pSCR::abi1-1 (Duan et al., 2013), UAS::abi1-1 (Duan et al., 2013) in Col-0, abi1-1 and rev-6 in Ler (Leung et al., 1994), and phb-13, athb8-11, cna-2, phb-13 phv-11, cna-2 phb-13, cna-2 phv-11, athb8-11 phb-13, athb8-11 phv-11 and athb8-11 cna-2 (Prigge et al., 2005) in Col er2. For tissue-specific expression of abi1-1, UAS::abi1-1 plants were crossed to Haselhoff driver lines (Haseloff, 1999) and F1 plants were analysed. Reporter lines used were pAHP6::GFP (Mähönen et al., 2006), pTMO5::n3GFP (De Rybel et al., 2013), pVND7:NLS:YFP (Kubo et al., 2005) and pMIR165A::GFP (Carlsbecker et al., 2010).

For analysis of primary root xylem defects, seedlings were mounted in 8:2:1 chloral hydrate: glycerol: water (w/v/v) solution and the cleared roots were visualized using a Zeiss Axioscope A1 microscope at 40× magnification, with differential interference contrast (DIC) optics using an Axio Cam ICc5 camera. Each experiment was repeated at least twice. Phenotype quantification was carried out by assessing the vascular defects in each root and representing the frequency of a certain defect in a population. If roots exhibited more than one type of vascular defect, it was quantified in a separate category as individuals exhibiting a combination of phenotypes.

Basic fuchsin staining for lignin was performed according to Mähönen et al. (2000), and confocal imaging was carried out using Zeiss 780 inverted Axio Observer with supersensitive GaAsP detector. The brightness and contrast of fuchsin-stained longitudinal and transverse section images were modified using Photoshop (Adobe) for representation of xylem defects. For analysis of GFP-reporter lines, roots were mounted between two coverslips in 40 µM propidium iodide solution and imaged immediately. Confocal analysis was carried out using either a Zeiss LSM 780 or a Zeiss LSM 800 microscope. In each experiment, image acquisition settings were kept constant for mock and respective treatments, and images were acquired at the depth of the QC cells. For quantification of GFP fluorescence intensity, the histogram function in ZEN Black edition software (Zeiss) was used. For pMIR165A::GFP quantification, a rectangular area of fixed size was drawn around endodermal cells 9-15, counted from QC, in every root meristem. Another equally sized rectangle was drawn in a region not expressing GFP and used to calculate the background. Background intensities were subtracted from GFP intensities and intensity values were represented in a box plot diagram.

For mRNA transcript quantification analysis, whole roots were collected. For each experiment, three biological replicates consisting of ∼125 individuals each were harvested in RLT buffer from Qiagen Plant Mini Kit and homogenized using Fast prep homogenizer (Savant) at speed 6 for 20 s. RNA extraction and on-column DNase digestion was performed according to the manufacturer's instruction. cDNA was synthesized using iScript cDNA synthesis kit using 500 or 1000 ng of RNA (quantified using Qubit Fluorometer 2.0). For qRT-PCR, iQ SYBR Green Supermix was used in 20 µl reactions and reactions were run in an iCycler iQ Real-Time PCR (Bio-Rad) instrument. Three-step PCR reactions with three technical replicates were carried out with an annealing temperature of 60°C. Results were normalized following the method of Pfaffl (2001) to the reference gene ADENINE PHOSPHORIBOSYL TRANSFERASE 1 (APT1), which was found to be stable and unaffected by ABA treatments (data not shown).

For quantification of mature miRNA levels, whole roots from three biological replicates were collected in 450 µl lysis/binding buffer from the miRvana miRNA Isolation kit (Ambion). Homogenization was carried out as mentioned above and RNA extraction according to manufacturers' instructions. cDNA synthesis was performed using Superscript IV Reverse Transcriptase (Invitrogen) following the stem-loop pulsed reverse transcription protocol (Varkonyi-Gasic et al., 2007). For the reverse transcription of mature miRNA and reference gene, stem loop primers specific for miR165 and primers for GAPDH (GAPDH qR) were used in the same reaction (see Table S1). For qRT-PCR, SYBR Green Supermix (Bio-Rad) was used and reactions were performed using the published conditions (Varkonyi-Gasic et al., 2007).

We thank A. Minina for technical advice, J. R. Dinneny, T. Hirayama, M. Prigge, A.-P. Mähönen, K. Ohashi-Ito, B. de Rybel and Nottingham Arabidopsis Stock Centre for seeds.