Abiotic stress modulates root patterning via ABA-regulated microRNA expression in the endodermis initials

Arabidopsis primary xylem is composed of two cell types, a peripheral protoxylem (PX) and central metaxylem (MX) (De Rybel et al., 2016). Specification of the xylem axis is regulated by auxin-cytokinin (CK) feedback loops. Auxin induces vascular specification through MONOPTEROS (MP) [also known as AUXIN RESPONSE FACTOR 5 (ARF5)] (Hardtke and Berleth, 1998). In PX provascular cells, MP induces expression of the CK signaling inhibitor ARABIDOPSIS HISTIDINE PHOSPHOTRANSFER PROTEIN 6 (AHP6) (Bishopp et al., 2011; Mahonen et al., 2006).

In parallel, the radial patterning of PX and MX is regulated by the GRAS transcription factors SHORTROOT (SHR) and SCARECRAW (SCR), which, in the endodermis initials, induce the expression of two microRNAs (miRNAs) miR165a and miR116b. miR165a and miR166b move into the stele, presumably forming a local gradient, and restrict the expression of the Class III HOMEODOMAIN LEU-ZIPPER (HD-ZIPIII) family transcription factors PHABULOSA (PHB), REVOLUTA (REV), PHAVOLUTA (PHV), ATHB8 and ATHB15 [also known as CORONA (CNA)] to MX provascular cells. The low and high expression levels of HD-ZIPIII proteins in PX and MX provascular cells, respectively, are required for PX and MX specification (Carlsbecker et al., 2010; Miyashima et al., 2011).

The expression of HD-ZIPIIIs depends on local auxin biosynthesis in the root and its concentration by polar auxin transport (Bishopp et al., 2011; Muraro et al., 2014; Ursache et al., 2014). PHB induces local CK biosynthesis by inducing the expression of ISOPENTENYL TRANSFERASE 7 (IPT7) (Dello Ioio et al., 2012; Sebastian et al., 2015) and suppresses auxin signaling by induction of IAA20 and IAA30 auxin signaling inhibitors (Müller et al., 2016). Following specification in the meristem, the differentiation and maturation of PX and MX depend on the key transcriptional regulators VASCULAR DEPENDENT NAC DOMAIN 7 (VND7) and VND6, respectively (Kubo et al., 2005).

Abscisic acid (ABA) is a central regulator of abiotic stress responses in plants. ABA levels in roots increase rapidly in response to drought and salt stress (Audran et al., 1998; Geng et al., 2013; Qin and Zeevaart, 1999; Schachtman and Goodger, 2008). ABA signaling functions via intracellular receptors called PYRABACTIN RESISTANT1/PYR1 LIKE (REGULATORY COMPONENTS OF ABA RECEPTORS) [PYR1/PYL (RCAR)]. Upon binding ABA, PYR1/PYL sequester type 2C protein phosphatases (PP2C) leading to the activation of SUCROSE NON FERMENTING RELATED SERINE/THREONINE PROTEIN KINASE (SnRK) family kinases that phosphorylate ion channels and transcription factors (Fujii et al., 2009; Ma et al., 2009; Park et al., 2009; Parry et al., 2009). Different members of the PYR/PYL/RCAR family influence ABA signaling in the root (Antoni et al., 2013; Xing et al., 2016; Zhao et al., 2014). In addition, tissue-specific expression of the PP2C dominant mutant ABA INSENSITIVE 1 (abi1^G180D) inhibits ABA responses in roots (Duan et al., 2013; Geng et al., 2013).

In the primary root, ABA inhibits cell division of the quiescent center (Zhang et al., 2010). Inhibition of lateral root (LR) formation by ABA takes place at either the positioning/initiation or emergence stages (De Smet et al., 2003; Orman-Ligeza et al., 2018) at high or low ABA concentrations, respectively. Tissue specific expression of abi1-1 indicated that, at post-emergence, ABA-induced LR quiescence involves non-cell autonomous signaling from the primary root endodermis (Duan et al., 2013). scr and shr seedlings are ABA hypersensitive, and ABA response genes are direct targets of SCR (Iyer-Pascuzzi et al., 2011). In the stele, VND7 cluster secondary wall-associated genes are upregulated in response to various stress stimuli, but although patterning regulators such as SCR are associated with stress responses, changes in radial patterning have not been observed upon stress (Iyer-Pascuzzi et al., 2011). Thus, the function of ABA in root patterning is not fully understood.

Our results demonstrate that osmotic stress and ABA influence xylem patterning by regulating levels of miR165a/166b and, as a result, HD-ZIPIII RNA and concomitantly VND7 levels. Regardless of known negative effects of ABA on auxin responses, our analysis shows that the effects of ABA and auxin on xylem differentiation function via separate pathways. Moreover, our results show that ABA-induced LR quiescence at post-emergence stage is associated with an increase in miR165a and a reduction in PHB levels. During the preparation of this work, a complementary study demonstrating the effect of ABA on xylem patterning in the root was published (Ramachandran et al., 2018). This study demonstrated that osmotic stress via ABA regulates PX differentiation and MX continuity by affecting miR165 expression.

Cell walls were stained with propidium iodide (PI), and the measurement of cortex cell lengths using the Cell-o-Tape ImageJ (Fiji) macro (French et al., 2012) allowed accurate identification of the end of the transition zone (ETZ) that outlines the termination of the root meristematic zone and the beginning of the rapid cell elongation/differentiation zone. The Cell-o-Tape analysis showed that following ABA treatments the ETZ was shorter, indicating that ABA induced earlier cell elongation and/or differentiation relative to the position of the quiescent center (QC) (Fig. S1). The accurate analysis of differentiation along the root longitudinal axis ensured that radial patterning was reproducibly analyzed at the same position.

The staining of root lignified secondary cell walls (SCW) with basic Fuchsin showed a significant increase after ABA treatments (Fig. S2A,B). Co-staining of roots with Calcofluor White, which labels cellulose, and basic Fuchsin enabled analysis of the position of xylem lignification relative to the ETZ, revealing the timing of differentiation. Lignified SCW in PX and MX appear at distance of 8 to 9 cells and 18 to 20 cells, respectively, indicating that their maturation is separated in time. Following ABA treatments, lignified SCW appeared in the PX at 5 to 6 cells away from the ETZ (Fig. 1A,B and Fig. S2C). Moreover, after ABA treatments additional files of PX appeared at the MX position at a distance of 8 to 9 cells from ETZ (Fig. 1A,C), indicating that ABA induces early differentiation of xylem.

In tomato (Solanum lycopersicum) variety M82, two PX files differentiate between 2.5-3 mm from the tip and an additional internal two PX files start to differentiate ∼5 mm from the tip (Fig. 1D-F and Fig. S3). Following treatment with 1 or 5 µM ABA, the external PX files start to differentiate 1.5 mm from the tip and the two internal files differentiate ∼2.2 mm from the tip (Fig. 1D-F and Fig. S3A-D). In addition, ABA causes differentiation of up to six PX files 5 mm from the tip (Fig. 1E and Fig. S3E,F). The identity of the PX cells in Arabidopsis was confirmed using the PX marker pS18:ER-GFP (Lee et al., 2006). Following ABA treatments, pS18:ER expression was also observed in the two MX provascular cell files adjacent to the PX, confirming their PX specification (Fig. 1G and Fig. S4A). Taken together, the analysis of Arabidopsis and tomato roots indicated that ABA primarily affects the timing of provascular cell differentiation. Next, we examined whether the early PX differentiation is associated with VND7.

We evaluated expression of pVND7::GFP-NLS (Kubo et al., 2005) after tissue clearing as previously described (Kurihara et al., 2015). VND7 expression began in the cell division zone PX provascular cells located 3 to 4 cells distal to the QC. The expression levels were low and stable in the first 8 to 9 cells and then gradually increased (Fig. 1H,I and Fig. S4B,C). Following treatment with 5 µM ABA, expression of VND7 in provascular PX appeared later (6 to 7 cells distal to the QC), but the expression levels increased significantly faster compared with control (Fig. 1H,I and Fig. S4C,D). Following ABA treatments, pVND7::GFP-NLS signal was also detected in MX provascular cells (Fig. 1H,J) and in the xylem pole pericycle cells in the elongation zone (Fig. S4B). The analysis of the pVND7::GFP-NLS signal distribution indicates that exogenous ABA treatment resulted in increased VND7 expression in the longitudinal and radial axes, which likely accounted for the earlier SCW formation.

Lignification of the Casparian strip (CS) in the endodermis is required for the formation of a functional diffusion barrier (Naseer et al., 2012). Fuchsin staining revealed that, in Col-0 seedlings, lignification of the PX and CS in the endodermis were synchronized. CS lignification occurred around 9 to 10 cells from the ETZ. Following ABA treatments, CS lignification moved to 6 to 8 cells from ETZ (Fig. 1C,K). Collectively these results indicate that differentiation of PX and endodermis in roots is synchronized, and ABA induced earlier differentiation of these tissue layers.

Next, the organization of vascular tissue in the maturation zone was examined in the region between 11 to 13 cells from ETZ. At this position, no MX differentiation occurs and two PX files are easily detected with PI staining (Fig. S1A). As mentioned above, ABA treatment induced the formation of additional PX cell files that were distinguished by their typical spiral cell wall pattern. Whereas wild-type (WT) plants constantly have two PX cell files, ABA treatments induced the appearance of one or two additional files, depending on ABA concentration. These extra PX files were on the inner side of the normal PX position and resulted from earlier differentiation of MX provascular cells (Fig. 2A,B and Fig. S5A).

Next, we asked whether the formation of the extra PX files depended on the core ABA signaling cascade. The roots of both the quadruple ABA receptor mutant (pyr1;pyl1;pyl2;pyl4) (Park et al., 2009) and abi1-1, which have a dominant mutation of PP2C phosphatase (Leung et al., 1994), had normal PX organization consisting of two files under control conditions. Following ABA treatments, however, fewer files of extra PX were formed (Fig. 2A,B and Fig. S5B,C). ABA signaling and responses can be induced in transgenic plants independently of exogenous ABA application by expressing an engineered PYR1 ABA receptor. In plants that express PYR1^{Y58H/K59R/V81I/F108A/S122G/F159L} (PYR1^MANDI), mandipropamid triggers ABA responses (Park et al., 2015). Treatment of 35S::PYR1^MANDI seedlings with mandipropamid induced a similar but stronger effect on root morphology than did ABA, resulting in earlier differentiation of root hairs and vascular tissues compared with control (Fig. S6A). Differentiation of MX provascular cells into PX occurred more efficiently following treatment with 5 µM mandipropamid than with ABA, and some 35S::PYR1^MANDI roots treated with mandipropamid had five PX files (Fig. 2C and Fig. S6B). These results indicate that the core ABA signaling is required for ABA effects on the timing of PX differentiation and PX/MX specification.

As ABA mediates the responses to osmotic and salt stress, we asked whether either stress induces the formation of additional PX. No effect on PX differentiation was observed following incubation of roots for 24 h in 25, 50 or 100 mM NaCl (Fig. S7A). Similar to the effect of ABA, incubation in 60 g/l and 120 g/l polyethylene glycol 8000 (PEG 8000) for 24 h, which results in osmotic stress, induced the appearance of one or more additional files of PX in 80% of WT roots (Fig. 2D,E and Fig. S7B). PEG 8000 treatments also caused disordering of SCW in PX, leading to a variety of xylem patterns (Fig. S7C). To examine whether the influence of PEG 8000 on xylem differentiation was ABA-dependent, treatments were carried out using pyr1;pyl1;pyl2;pyl4 and the aba2-1 ABA biosynthesis mutant (González-Guzmán et al., 2002). Following incubation in 60 g/l PEG 8000, only one extra file of PX appeared in ∼20% of pyr1;pyl1;pyl2;pyl4 and aba2-1 roots (Fig. 2E and Fig. S7E). The reduced number of cells in xylem files in the ABA signaling and biosynthesis mutants suggests that PEG 8000 alters xylem differentiation in the root in a manner that depends on ABA biosynthesis and core ABA signaling. Incubation with 0.2 M sorbitol, which causes osmotic stress, similar to PEG8000, induced formation of additional xylem files in three out of nine seedlings (Fig. S7D).

To determine whether the meristem is required for ABA-induced PX differentiation, we excised the root tip at distances longer than 500 µm or shorter than 200 µm (Fig. 3A), which prevented or allowed, respectively, regeneration (Efroni et al., 2016; Sena et al., 2009). Excision of the root at distances longer than 500 µm from the tip prevented regeneration and abolished the ABA-induced PX differentiation from MX provascular cells. In contrast, when less than 200 µm was excised, the ABA treatment induced formation of ectopic PX (Fig. 3B and Fig. S8A). Excision at distances shorter than 200 µm, which removed the stem cell niche, likely allowed the respecification of cell identities required for the ABA-induced PX repatterning.

The lack of PX repatterning in ABA-treated roots that were excised at a distance longer than 500 µm from the tip suggests that cell respecification is crucial for the ABA effect. An alternative scenario that does not involve cell respecification is that there is a loss of a component or components required for maturation and radial patterning when the root is excised at distances longer than 200 µm from the tip. To distinguish between these scenarios, we examined whether cells that exit the meristematic zone respond to ABA. To this end, the position of the elongation zone initiation was labeled immediately before incubation in ABA (Fig. S8B, asterisk). With or without ABA treatment, the cells in the elongation zone did not form additional PX files, whereas MX was formed (Fig. 3C and Fig. S8C). This result indicates that ABA regulates PX/MX specification in the meristem and suggests that the effects of ABA on vascular patterning do not result from long-distance signaling between the meristem and the maturation zone.

Enhancer trap lines with tissue-specific expression of abi1-1 (Duan et al., 2013) were used to determine the tissue specificity of the ABA effect on PX differentiation. The ABA-induced differentiation of MX provascular cells into PX was repressed when abi1-1 was expressed in the endodermis in enhancer trap lines (Fig. 3D and Fig. S9: J0571>>abi1-1 and Q2500>>abi1-1) and pSCR promoter driven abi1-1 lines (Fig. 3F,G). In comparison, expression of abi1-1 in the stele or in control lines harboring abi1-1 without a driver promoter did not repress the ABA-induced ectopic PX files differentiation (Fig. 3D,E and Fig. S9: Q0990>>abi1-1). Expression of abi1-1 in the columella also had mild inhibitory effects (Fig. 3D and Fig. S9: J3411>>abi1-1). Taken together, our results indicate that the ABA signaling that induces formation of additional PX cells takes place primarily in endodermal precursor cells in the meristem.

In the root meristem, auxin signaling upregulates the expression of the CK signaling inhibitor AHP6 to specify PX (Bishopp et al., 2011). AHP6 is expressed in the PX and adjacent pericycle cells (Mahonen et al., 2006). Imaging of cleared roots demonstrated that low levels of pAHP6::GFP-ER (AHP6) were also present in MX provascular cells with and without ABA treatment. Treatments with 1 µM ABA had no effect on the expression level or pattern of AHP6. Treatments with 5 µM ABA lowered the overall average AHP6 expression but increased the variance and were not statistically significant (Fig. 4A,F and Fig. S10A). Given that ABA treatments did not result in increased AHP6 expression in the MX provascular cells, it is unlikely that ABA affects xylem pattering via MP-mediated auxin signaling. To further explore the interaction between auxin and ABA in regulation of PX differentiation we examined the effect of ABA in seedlings treated with auxin signaling and transport inhibitors, and in auxin signaling mutants (Figs S11 and S12). Treatments with the auxin signaling inhibitor auxinole (Hayashi et al., 2008) strongly inhibited the auxin response (Fig. S11A-D) and the inhibition was slightly enhanced by co-treatments with ABA (Fig. S11C,D). These data agree with a known inhibitory effect of ABA on auxin responses (He et al., 2012; Shani et al., 2017). Treatments with auxinole completely abolished xylem formation, whereas treatments with 1-naphthylphthalamic acid (NPA) had no effect. Co-treatments with auxinole and ABA induced formation of a single file or two files of mature PX cells (Fig. S11E,F). Similar to auxinole, effects of ABA on xylem differentiation were reduced in non-aborted tir1;afb2;afb4;afb5 quadruple auxin receptor mutant roots (Prigge et al., 2016). Conversely, ABA effects were enhanced in 35S::Iaa6m1m2 plants, which display a constitutive auxin response (Li et al., 2011). Moreover, qPCR analysis indicated that a 3 h ABA treatment did not affect RNA levels of three auxin biosynthesis genes, TAA1, YUCCA3 or YUCCA9, and caused an ∼2-fold reduction in YUCCA5 mRNA levels. Taken together, the analysis of ABA-auxin interaction (Figs S11 and S12) indicated that ABA and auxin do not function in the same pathway but rather have synergistic effects on xylem differentiation, providing an explanation for why AHP6 was not upregulated by ABA.

Next, we examined whether ABA affects the expression of miR165a/166b and the HD-ZIPIII targets of these miRNAs (De Rybel et al., 2016). ABA treatments induced significant increases in the expression of pmiR165a::GFP and pmiR166b::GFP (miR165a/166b) (Carlsbecker et al., 2010) in the endodermis and cortex, respectively (Fig. 4B,C,G,H and Fig. S10B,C). Similarly, miR165a expression was induced by mandipropamid in PYR1^MANDI plants (Fig. 4D,I and Fig. S10D), indicating that enhancement of miR165a expression involved core ABA signaling. miR165/166 levels were previously shown to be negatively regulated by ZLL (Liu et al., 2009; Yu et al., 2017; Zhu et al., 2011). In roots of zll/pZLL::YFP:ZLL (Tucker et al., 2008) YFP:ZLL is expressed in the stele, with the highest expression in procambial cells and lower levels in xylem cells (Fig. 4E and Fig. S10E). Treatments with 5 µM ABA caused reduction in YFP:ZLL expression levels (Fig. 4E,J and Fig. S10E), likely leading to increased levels of free miR165a/166b in the stele. Taken together, ABA increases miR165a/166b levels by inducing their expression and repressing the expression of their suppressor ZLL.

As miR165a and miR166b target mRNAs encoding HD-ZIPIII family proteins, including PHB, we asked whether ABA induced an altered expression pattern of pPHB::GFP_ER (PHB) transcription and pPHB::PHB-GFP (PHB) reporters (Miyashima et al., 2011). For accurate observation tissue clearing was used. The expression pattern and levels of PHB transcript reporter were similar in control and ABA treatment (Fig. 4K,M and Fig. S10F), indicating that ABA does not influence the transcription of PHB. In contrast, ABA treatment resulted in reduction in the levels of PHB protein reporter. In control seedlings, PHB was detected in the central part of stele, excluding the pericycle. Following incubation with ABA, PHB levels were lower and more restricted (Fig. 4L,N and Fig. S10G). Importantly, reduction in PHB levels were already detected following 3 h and 5 h treatments with ABA, suggesting rapid effect of ABA on PHB mRNA levels (Fig. 4O). Taken together, the analyses of ABA signaling with the data on miR165a/miR165b, PHB and PHB expression indicate that ABA signaling in endodermal precursors induces expression of miR165a/166b and reduces the level of their negative regulator ZLL, and that these miRNAs post-transcriptionally repress the expression of PHB and possibly other HD-ZIPIII-encoding genes in meristematic cells.

Previous findings on the role of primary root endodermis ABA signaling in inhibition of LR growth (Duan et al., 2013) promoted us to examine whether, and how, ABA regulates miR165a and PHB during LR development. Weak expression of miR165a in cleared roots was first detected at stage IV in outer layer 2 precursors of endodermis, cortex and QC (Fig. 5A) of lateral root initials (LRIs), in which SCR is expressed (Goh et al., 2016). miR165a expression increased in a linear fashion up to the emergence stage and then an exponential increase in miR165a expression was detected at the post-emergent LRIs (Fig. 5A,B). At stage VII, and more significantly at the emergence and post-emergence stages, miR165a expression dropped down in QC precursor cells and was restricted to endodermal precursors (Fig. 5A, white arrows). Treatments with 1 µM ABA significantly increased miR165a expression levels specifically at the emergence and post-emergence stages, coinciding with the ABA-induced LR quiescence at emergence stage (De Smet et al., 2003) (Fig. 5A,B). Furthermore, the ABA treatments also inhibited the reduction in miR165a levels in the QC precursor cells (Fig. 5C,D). Importantly, enhancer trap lines Q2500 and J0571, which drive the expression of abi1-1 in the primary root endodermis (Duan et al., 2013; Fig. S9) were detected in endodermis and QC precursor cells in LRIs (Fig. 5E). Thus, it is highly likely that inhibition of ABA-induced LR quiescence in enhancer trap lines Q2500 and J0571 (Duan et al., 2013) was associated with expression of abi1-1 in LRI endodermal precursor cells rather than the endodermis of the main root. It has been shown that ABA inhibits LR growth following emergence, likely through an auxin-independent mechanism (De Smet et al., 2003). Thus, there is a tight correlation between the endodermal/QC precursor cell ABA signaling-induced quiescence of LR growth and the increase in miR165a levels specifically at the LRI emergence stage.

In agreement with previously published data (Hawker and Bowman, 2004), PHB expression was already detected in the inner layer (IL) of stage II LRIs and in the progeny of this layer during later stages of LR development in cleared roots (Fig. 6A). Coinciding with the increase in miR165a levels, the expression domain of PHB became confined to up to five layers of IL progeny cells in the mid region of emergence stage LRI and restricted from endodermal and pericycle layers (Fig. 6B,D, region 2). Hence, PHB and miR165a form complementary expression domains in developing LRIs. Following ABA treatments and coinciding with the increase in miR165a levels, the overall levels and the number of cells that express PHB were reduced at the emergence stage (Fig. 6B-D). Taken together, these results strongly suggest that ABA may regulate post-emergent LR growth by regulating the levels of miR165a, and hence PHB. VND7 expression was first detected at post-emergence LRIs but became expanded only in LRs (Fig. 4E), suggesting that PHB likely inhibits VND7 expression during early LRI development.

To further examine function of HD-ZIPIII in LR development we examined the root length and LR development in a triple phb;phv;cna mutant (Prigge et al., 2005) and LR development in the gain-of-function phb1-d mutant (McConnell et al., 2001). Following treatments with 0.5 µM ABA, root lengths of Col-0 and phb;phv;cna were significantly reduced (P≤0.005, Student's t-test). Interestingly, following ABA treatment the phb;phv;cna were slightly but significantly shorter (P≤0.05. Student's t-test) (Fig. S13A). The ABA treatments caused significant reduction in LR density in both Col-0 and the phb;phv;cna seedlings (P≤0.005, Student's t-test) but no differences were detected between the WT and mutant plants (Fig. S13B). phb1-d mutants developed very short primary roots, with seedlings developing not more than one LR under control conditions. Interestingly, some phb1-d seedlings developed a LR following a treatment with 1 µM ABA (Fig. S13C). However, owing to the short primary root length and the very small number of LR it is difficult to know whether the development of LR at 1 µM ABA did indeed reflect a lower sensitivity of the mutant to the hormone during LR development.

Quantitative RT-PCR analyses were carried out to examine whether ABA regulates the expression of all five HD-ZIPIII-encoding genes. PYL4 and PP2C are known to be down- and upregulated by exogenous ABA treatments, respectively, and were used as controls in all the experiments for monitoring the efficiency of the ABA and mandipropamid treatments (Fig. S14A-C). In Col-0 seedlings, a 1 h 10 µM ABA treatment resulted in significant reductions in the levels of all five HD-ZIPIII-encoding genes (Fig. 7A). Similarly, a 3 h treatment of PYR1^MANDI plants with mandipropamid induced significant reductions in the levels of all five mRNAs (Fig. 7B). Furthermore, the effects of ABA on mRNA levels of all five HD-ZIPIII genes were reduced in abi1-1 and pyr1;pyl1;pyl2;pyl4 ABA signaling mutants compared with effects in the WT seedlings (Fig. 7C,D). The changes in expression levels of all 5 HD-ZIPIII genes at 24 h were statistically significant, but reduced compared with the 1 h treatment (Fig. S14D). The quantitative RT-PCR data indicated that elevated ABA levels induce a rapid decrease in HD-ZIPIII-encoding mRNAs via core ABA signal transduction, in line with the changes in expression levels of PHB-GFP and miR165a/166b. In summary, the developmental and expression analyses indicated that, upon drought stress, increases in ABA levels lead to increases in miRNA165a/166b, which cause repression of all five HD-ZIPIII mRNAs. The reduction in the levels of HD-ZIPIII proteins lead in turn to earlier differentiation of MX provascular cells into PX in primary root and likely quiescence at the emergence stage in LRI.

The phenotypes of a phb1-d (McConnell et al., 2001) mutant and a triple phb;phv;cna (Prigge et al., 2005) mutant with and without ABA treatments were analyzed to further examine the causal relationships between ABA signaling and the timing of SCW formation in PX and endodermis. Basic Fuchsin staining of phb;phv;cna roots indicated that the position of PX lignification is not significantly different from that of WT roots (Fig. S15A,B). phb1-d is a miR165a/166b-resistant mutant and should therefore display reduced sensitivity to ABA (Carlsbecker et al., 2010; Miyashima et al., 2011). Staining of phb1-d showed that mutant seedlings generally lack PX and that lignification of MX initiates 17 to 18 cells from the ETZ (Fig. S15A,B). About 10% of phb-1d seedlings developed one file of PX at distance of 7 to 8 cells from the ETZ (Fig. S15D). Cases of normal PX development in phb1-d have been reported previously (Miyashima et al., 2011). Thus, the PHB gain-of-function in phb1-d reduces but does not eliminate PX differentiation. In agreement, treatment of phb1-d seedlings with 1 µM ABA resulted in PX development and lignification 6 to 7 cells from the ETZ, indicating that ABA partially overcomes the miR165a/166b-resistant phenotype of pbh1-d plants (Fig. S15A,B). SCR and SHR regulate the expression of miR165a/166b and, similar to phb1-d, scr-3 and shr-2 mutant plants lack PX and develop MX also in the PX and pericycle cell files (Carlsbecker et al., 2010). Similar to phb1-d, MX differentiation takes place later in development and not where PX normally differentiates (Fig. S16A). These data suggest that when cells are not specified as PX they can undergo differentiation into MX later in their development. In both scr-3 and shr-2, PX differentiation could be rescued by treatments with 5 µM ABA, suggesting that ABA-induced increases in miR165a/166b levels are partially additive to SHR and SCR (Fig. S16A,B).

Interestingly, whereas SCW lignification in the MX of most phb1-d roots takes place 17 to 18 cells from the ETZ, the lignification of the CS in the endodermis occurs 8 cells from the ETZ, even earlier then in Ler roots (10 cells from the ETZ) (Fig. S15C,E). Thus, SCW formation in xylem and the CS are genetically separable. In contrast to Ler seedlings, the MX of phb1-d was visible with PI staining (Fig. S15E), indicating that maturation of the CS is compromised (Roppolo et al., 2011) and suggesting that PHB influences CS integrity.

In Ler roots, ABA induced formation of four PX files, but at lower frequencies compared with Col-0. In phb-1d roots, following ABA treatments no, one, two and rarely three PX files, but not four PX files, were observed (Fig. 8A,B,E). In scr-3 and shr-2, ABA induced the formation of only one or two PX files (Fig. 8D,E). Thus, in the miR165a/166b-resistant phb1-d, as well as in scr-3 and shr-2 roots, the effects of ABA were suppressed compared with WT. Conversely, ABA enhanced PX formation in the phb;phv;cna triple mutant, resulting in formation of five and six PX files (Fig. 8C,E).

We hypothesized that ABA-induced reductions in the levels of PHV, REV, CNA and ATHB8 partially overcome the phb1-d gain-of-function and enable differentiation of PX. qPCR was used to examine the expression of all five HD-ZIPPIII genes in phb1-d. As expected, PHB RNA levels are significantly increased in phb1-d compared with Ler. Interestingly, the levels of ATHB15, ATHB8 and PHV were also increase relative to Ler (Fig. S14E). In ABA-treated Ler seedlings, all five HD-ZIPIII RNA levels were reduced (Fig. 8F), indicating that ABA had a similar effect on HD-ZIPIII expression in Ler and Col-0 ecotypes. In ABA-treated phb1-d seedlings, levels of PHV, REV and ATHB8 RNAs were reduced compared with control, whereas PHB and also ATHB15 RNA levels were not affected by the ABA treatments (Fig. 8G).

ABA has been referred to as ‘the hidden architect of root structure’ (Harris, 2015). The general effect of ABA on the root is the enhancement of maturation, which is reflected in smaller cell number in the cell division and elongation zones, premature differentiation of PX and maturation of the CS. ABA regulates PX differentiation and likely LR emergence by regulating the levels of miR165a/166b (Fig. 9).

Similarly, it has been shown that osmotic stress induces formation of additional PX files via endodermal ABA signaling that enhances the expression of miR165a, thereby reducing RNA levels of all five HD-ZIPIII RNAs (Ramachandran et al., 2018). The same work also demonstrated that ABA is required for the continuity of the MX files. Our work adds several essential and distinctive findings. By analyzing both Arabidopsis and tomato roots we show that the primary effect of ABA is the timing of PX maturation, by affecting the levels of VND7 expression in provascular cells. This finding suggests that commitment of MX provascular cells to differentiate into MX takes place later in development and that in the meristem these cells maintain the plasticity to differentiate into PX. VND7 represses the expression of REV and PHB by directly binding to their promoters. REV represses PAL4, suggesting that VND7 induces expression of SCW genes by suppressing the expression of HD-ZIPIII proteins (Taylor-Teeples et al., 2015). The analysis of tomato roots also showed that the effect of ABA on xylem differentiation are evolutionarily conserved. In addition, we identified ABA receptors that are involved in regulating xylem differentiation.

Uniquely we show that ABA also affects miR165a/166b by regulating the expression level of their suppressor ZLL. High expression levels of ZLL in the procambium likely reduced the levels of free miR165a/166b in procambial cells, restricting their inhibitory effect to the xylem pole. Similarly, in the shoot meristem, ZLL is required for maintaining the stem cell niche and leaf polarity through suppression of miR165/166 expression (Liu et al., 2009; Tucker et al., 2008).

We have not observed expansion of the AHP6 expression domain by ABA, contrary to Ramachandran et al. (2018). Using clearing, we observed AHP6 expression outside the PX files in control untreated plants. Our results agree with the additive effects of ABA and auxin in PX differentiation, which indicate that ABA does not affect xylem differentiation via core auxin signaling. The analysis of ABA effects in the shr and scr background indicated that its effects are partially independent of their function. Our results also indicate that ABA regulates miR165/166 levels in endodermal precursor cells in both primary roots and LRIs.

It has been shown that SCR is required for the establishment of LR QC (Goh et al., 2016). Consistently, we discovered a specific expression pattern of miR165a in LRIs, which is regulated by ABA at the emergence stage, when ABA suppresses LR growth. These findings provide a possible mechanism for earlier findings on the non-cell autonomous role of endodermal ABA signaling which regulates LR growth (Duan et al., 2013). Alternatively, the ABA-induced upregulation of miR165a in the QC may be a consequence of a specific effect on LR meristem formation.

Our finding demonstrate that ABA regulates xylem patterning, CS maturation and possibly LR emergence via miR165/166 and HD-ZIPIII levels, and that ABA regulation of xylem differentiation is evolutionarily conserved.

Arabidopsis plants used in this study were in Col-0, Ler or C24 ecotypes. Analysis of abi1-1 enhancer trap lines and p35S::PYR1^MANDI and pMIR165a::GFP crosses were carried out at the F1 generation. Plants were grown on 0.5× Murashige & Skoog (MS) salt mixture plates under long-day conditions (16 h light/8 h dark). For more details, see supplementary Materials and Methods. See Table S2 for resources.

Seedlings at 5 days after germination (DAG) or 12 DAG (for lateral root experiments) were transferred under sterile conditions to fresh standard plates supplemented with +(-)cis, trans-abscisic acid (ABA; Duchefa A0941, 10 mM stock solution in ethanol); mandipropamid (Sigma-Aldrich, 32805, 10 mM stock solution in methanol); α-naphthaleneacetic acid (NAA; Sigma-Aldrich, N-0640, 100 mM stock in ethanol), auxinole (a gift from Ken-Ichiro Hayashi, Okayama University of Science, Japan, 50 mM in DMSO). For control treatments, plates were supplemented with solvents alone. All analyses were carried out 24 h after transfer to the new growth conditions.

Five DAG seedlings were transferred to standard plates supplemented with NaCl, sorbitol or PEG 8000 (Sigma-Aldrich, P5413) at indicated concentrations. PEG 8000-infused plates were prepared according to a previously described procedure (Verslues et al., 2006). In brief, plates containing 20 ml of standard agar media were covered with 30 ml liquid 0.5× MS media supplemented with PEG 8000 and incubated overnight at room temperature. To obtain final concentrations of 60 g/L and 120 g/L, plates were covered with 100 g/L and 200 g/L PEG 8000, respectively. The 500 g/L PEG 8000 stock solution in 0.5× MS media was filter sterilized and stored at 4°C in the dark. Liquid PEG 8000 media were removed just before seedling transfer, and plates were sealed with paraffin and transferred to growth chamber for 24 h. Importantly, for reproducible results, the PEG 8000 was freshly prepared. All analyses were carried out 24 h after transfer to the new growth conditions.

Seedlings were stained for 10 min with 100 μg/ml PI (Sigma-Aldrich, P4864, 1 mg/ml in doubly distilled water), washed for 1 min in water, and immediately subjected to imaging.

Roots were stained with PI. Cell length measurements were carried out using the semi-automated Cell-o-Tape macro for ImageJ (Fiji) (Band et al., 2012). All measurements were taken from the cortex cell layer, with numbering starting from the first initial adjacent to the QC. For more details see supplementary Materials and Methods.

Analyses were carried out on roots stained with PI, at a constant region in the maturation zone of the root, 3 or 4 cells distal from the first emerging root hair or 11 to 13 cells from ETZ. In control conditions, maturing PX cells at this region have a well-defined spiral pattern and MX differentiation could not be detected. Confocal imaging of vascular tissue was carried out using a Zeiss C-Apochromat 40×/1.20 W objective. The number of PX cell files was evaluated by 3D reconstruction (ZEN software, Zeiss) of z-stack confocal sections. Typical stacks consisted of 25-30 sections at 0.6 μm intervals.

Roots were fixed in 4% paraformaldehyde and cleared in ClearSee reagent for 1 week according to the published protocol (Kurihara et al., 2015). Tomato roots were fixed and cleared with ClearSee for 1 month. Primary cell walls were stained with Calcofluor White (100 μg/ml in ClearSee, Sigma-Aldrich, 18909) for 1 h at room temperature and washed overnight in ClearSee solution. For double staining with Calcofluor White and basic Fuchsin (Fluka, 47860), cleared seedlings were incubated in basic Fuchsin (0.05% w/v in ClearSee) for 24 h, washed for 1 h in ClearSee, then stained with Calcofluor White (100 μg/ml in ClearSee) for1 h, washed for 24 h in ClearSee solution, and mounted on slides in ClearSee. The staining of tomato roots was similar, except the incubation with Calcofluor White and basic Fuchsin was carried out for 24 h.

For quantification of PX, MX and CS lignification position, cleared roots were double stained with Calcofluor White and basic Fuchsin. Images were captured using a C-Apochromat 40×/1.20 W objective, 8-bit depth (Zeiss). Zeiss Plan-Apochromat 20×0.8 M27 objective was used for image acquisition of tile-scans in the middle sections of roots. Counts were done on cortex cells stained with Calcofluor White, from the ETZ to the position of first PX, MX or CS stained by basic Fuchsin. Fluorescence intensity in appropriate cell layers, as indicated in the text, was measured using Fiji.

Roots of DR5::Venus-NLS 5 DAG seedlings were excised under a dissecting microscope as previously described (Sena et al., 2009). Sectioning was carried out either 500-600 μm from root tip (complete removal of root meristem including QC, division and transition zones) or 150-200 μm from the tip (removal of QC and first initials in the division zone). After excision, seedlings were immediately transferred to 0.5× MS, 1% sucrose and 0.8% agar plates supplemented with 5 μM ABA or 0.1% ethanol for control, grown vertically for 24 h, and cleared with ClearSee.

Immediately after transfer to fresh plates, the position of the elongation zone in each root (Col-0) was labeled by placing a small granule of coal on the agar close to the root. Seedlings were grown vertically for 24 h, then the section of the root below the coal was removed and the upper root region was transferred to the clearing solution and subsequently analyzed.

12 DAG roots were transferred to standard plates supplemented with 1 µM ABA or solvent for 24 h, cleared and stained with Calcofluor White. Stages of LR development were determined according to Malamy and Benfey (1997).

Images were captured on Zeiss LSM 780-NLO laser-scanning confocal microscope. Image analyses were performed using Fiji and Zeiss ZEN10. Where necessary, channel dye separation was used for separating signal from autofluorescence. For more details, see supplementary Materials and Methods.

DII-Venus-NLS seeds were germinated and grown in 24-well plates on 0.5× MS, 1% sucrose and 0.8% agar medium for 5 days at an angle of 35° to enable roots to reach the bottom of the wells and continue growing between the medium and the bottom of the plate. ABA (5 μM), auxinole (25 μM) or DMSO was added to wells, and time-lapse imaging of root tips was carried out using Zeiss LD Plan-Neofluar 40×/0.6 objective for 80 min, with 5 min intervals, on 8-bit depth single confocal sections. Mean fluorescent intensity in root tips was quantified for each time point using Fiji (ImageJ). Fold change in fluorescence intensity was calculated compared with time point 0. Mean fold changes are based on measurements carried out on five or six seedlings per treatment.

Images were captured using a C-Apochromat 40×/1.20 W objective, 8-bit depth. Fluorescence intensity in appropriate cell layers, as indicated in text, was measured using Fiji.

Total RNA was extracted from roots and used for reverse transcription, which was in turn used for qPCR (see Table S1 for primers). At least three biological replicates were used for each genotype and treatment, and representative experiments are presented in figures. For more details, see supplementary Materials and Methods.

Stacked charts and box plots were prepared using Office Excel 2016. Statistically significant differences were determined using the Student's t-test, two-tail distribution, unequal variance. The n values, which represent different roots, and precision measures are specified in figures and figure legends.

We thank Annelie Carlsbecker, Keiji Nakajima, Sean Cutler, Philip Benfey, José Dinneny, Thomas Laux, John Bowman, Eilon Shani, Mark Estelle and Tako Demura for materials, and Eilon Shani, José Dinneny and members of the Yalovsky lab for helpful comments and discussions.