The wild-type rice Dongjin (O. sativa L. cv. japonica) was used in this work. The OsABIL2 and OsSAPK10 overexpression transgenic lines were in the Dongjin background, and the detail of OsABIL2 overexpression transgenic lines were introduced in previous study (Li et al., 2015). The auxin-related transgenic rice lines DR5-GUS (ZH11 background), OsPIN1b-GUS, OsPIN1c-GUS, OsPIN2-GUS, OsPIN5a-GUS, and OsPIN10a-GUS were from a previous study (Wang et al., 2009). For polar auxin transport assays by crown root GUS staining, the F₁ hybrid of OsABIL2-OE X DR-GUS and DJ X DR5-GUS lines was used.

For physiological analysis, the rice seeds were imbibed for 2 days at 30°C, and the germinated seeds were then transferred to bottom-cut 96-well PCR plates. Seedlings were grown on water with a 16-h (light, 28°C)/8-h (dark, 25°C) rhythm for the indicated days, which was defined as normal conditions in this study.

For overexpressing SAPK10, the genomic sequence of Os03g0610900 was cloned into the binary vector pCAMBIA1306 fused with a FLAG-tag at the C-terminus. The primers used for cloning OsSAPK10 are shown in Supplementary Table 1.

For measurement of SAPK10-OE expression level, roots of the 7-day-old wild-type and transgenic seedlings grown under normal conditions were cut, and the samples were frozen in liquid nitrogen and stored at -80°C for RNA extraction.

For root hair induction assays, the 5-day-old normal grown seedlings were transferred to solutions with or without the following additions: ABA (100 mM dissolved in ethanol and in order to add the same amount of ethanol between treatment group and mock, we gradient dilution ABA to 20 mM, 10 mM, 5 mM, 2 mM, and 1 mM solution), NAA (100 mM dissolved in ethanol, and in order to add the same amount of ethanol, we gradient dilution NAA to 5 and 1 mM solution), the ABA biosynthesis inhibitor fluridon (FLU) (100 mM dissolved in ethanol), the auxin efflux inhibitor NPA (100 mM dissolved in DMSO), and the control added the same amount of ethanol and/or DMSO. After 24 h, the crown roots were fixed in FAA solution (ethanol: acetic acid: 37% formaldehyde: H₂O = 50:5:10:35, v/v), and then photographed with a stereoscopic microscope (Carl Zeiss Discovery V20).

For β-glucuronidase (GUS) staining, crown roots were immersed in the GUS staining solution (1 mg/ml X-glucuronide in 100 mM sodium phosphate, pH 7.2, 0.5 mM ferricyanide, 0.5 mM ferrocyanide, and 0.1% Triton X-100), briefly subjected to a vacuum, and then incubated at 37°C in the dark. The stained plant roots were photographed using Carl Zeiss Discovery V20 stereomicroscope.

Crown roots of the 6-day-old seedlings were cut and fixed in FAA solution overnight. The fixed samples were dehydrated in an ethanol series of 50, 60, 70, 80, 90, and 100% ethanol for 1 h each, and then treated with ethanol: tert-butyl alcohol (v/v), 3:1, 1:1, and 1:3 for 1 h each. Finally, the samples were kept in pure tert-butyl alcohol. After vacuum freeze-drying, the samples were sputter-coated with Au-Pd and further visualized with a scanning electron microscope (Hitachi TM3000). For observation of root hair initiation, we screened the root tip from the root apex to the root hair zone.

Root hairs were observed, and images were captured with a Discovery V20 Stereomicroscope (Carl Zeiss). The zone of root hair growth is 4–5 mm from the root apex. At least 15 crown roots were measured, and 20 root hairs of each root were measured. Root-hair length measurements were performed with ImageJ¹.

To analyze IAA concentrations, firstly, according to the root hair phenotype, we divided the root tips of rice crown roots into three regions: short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip) (Supplementary Figure 3). Each region was excised from at least 96 five-day-old seedlings treated with 0.5 μM ABA for 24 h, or not treated as a control. Samples were separated under Carl Zeiss Discovery V20 stereomicroscope and immediately frozen in liquid nitrogen, afterward stored at -80°C. Samples were powdered in liquid nitrogen and at least 30 mg powder homogenized in 750 μl of cold extraction buffer 1 [methanol: H2O: acetonitrile = 80:19:1 (v/v), 10 ng/ml d5-IAA], then shade and shake at 4°C 16 h (300 rpm). Samples were centrifuged at 4°C at 13,000 rpm for 10 min, the supernatant is transferred to another Eppendorf tube. Add 450 μl of cold extraction buffer 2 [methanol: H2O: acetonitrile = 80:19:1 (v/v)] into precipitate, shade and shake at 4°C 4 h (300 rpm), centrifuged at 4°C at 13,000 rpm for 10 min, combine two supernatants, transfer supernatant through 0.22 μm filter to a new Eppendorf tube. Then each sample was dried using nitrogen flow and re-dissolved in 200 μl of 30% cold methanol 3–6 h. Quantification was performed in an ABI 4000Q-TRAR LC-MS system (Applied Biosystems, United States) with stable, isotope-labeled auxin as the standard (OlChemIm, Czech Specials) according to a method described previously (Liu et al., 2012).

The qRT-PCR assays were carried out as described previously with small modifications (Zhang et al., 2009). The primers for qRT-PCR were designed with NCBI Primer-Blast² to avoid the homologous regions, shown in Supplementary Table 1. Total RNA was extracted with the Plant RNAprep Kit (Tiangen). About 2 μg DNase-treated RNA was used for reverse transcription (M-MLV reverse transcriptase, TaKaRa). The RT-PCR amplifications were carried out with a Bio-Rad CFX system, and PCR products were monitored with SYBR green dye. The expression level was normalized by the expression of OsACTIN1 (LOC_Os03g50885, internal control), and RT-qPCR results were analyzed by the 2^-ΔΔCT method using Bio-Rad CFX Manager 3.1 software.

According to the root hair phenotype, we divided the root tips of rice crown roots into three regions: short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip). Each region was excised from at least 96 five-day-old seedlings treated with 0.5 μM ABA for 24 h, or not treated as a control. Samples were separated and immediately frozen in liquid nitrogen, and afterward stored at -80°C until RNA isolation. Total RNA was extracted from each tissue using TRIzol (Invitrogen). RNA purity was checked with a NanoPhotometer spectrophotometer (IMPLEN, Munich, Germany) and RNA integrity was assessed with an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA, United States).

Two biological replicates were used for RNA-seq experiments for each tissue type. A total amount of 3 μg RNA per sample was used as input material for the RNA sample preparations. The samples were sent to Beijing Novogene Bioinformatics Technology Co., Ltd. (Beijing, China). The sequencing libraries were constructed using a NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, Ipswich, MA, United States) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly(T) magnetic beads. Quality of these libraries was assessed with an Agilent 2100 Bioanalyzer system. The clustering of the index-coded samples was performed with a cBot Cluster Generation System using a TruSeq PE Cluster Kit v3-cBot-HS (Illumina, San Diego, CA, United States) according to the manufacturer’s instructions. After cluster generation, the libraries were sequenced on an Illumina Hiseq 4000 platform and 150-bp paired-end reads were generated.

In this study, the genes with a significant (P ≤ 0.05) fold change > 2 or < -2 between the control and ABA treatment were defined as differentially expressed genes responding to ABA stimulation.

Statistical analysis was performed using IBM SPSS 20.0 software. One-way analysis of variance was used, and comparison between the two groups was performed using the least significant difference (LSD) test. Differences were considered significant if P < 0.05.

Gene ontology analysis was carried out using the Singular Enrichment Analysis (SEA) tool offered by agriGO (Du et al., 2010) at default settings of Fisher t-test (P < 0.05), False Discovery Rate (FDR) correction by Hochberg method and five minimum number of mapping entries against the rice-specific precomputed background reference.

To understand how plants adapt to environmental stresses, we tested whether the stress-responsive hormone ABA affects root hair growth in the Dongjin (DJ) cultivar of rice. Crown roots of the 5-day-old seedling were treated with exogenous ABA at concentrations of 0.1, 0.5, 1, and 2 μM for 24 h. As shown in Figure 1A, in the seedlings treated with 0.1 μM ABA, the root hair length was significantly less than in the control seedlings not treated with ABA. However, treatments with 0.5, 1, and 2 μM ABA significantly enhanced root hair elongation, indicating higher concentrations of ABA can promote root hair elongation, and we used the ABA concentrations of 0.5 and 2 μM for further studies.

We also used SEM to observe root hair morphology. We found that most root hairs initiated in the region about 3 mm from the root apex under normal growth conditions, but most root hairs initiated in the region about 0.2 mm from the apex in roots treated with 2 μM ABA for 24 h. We also further confirmed that the root hair length increased in response to ABA treatment (Figure 1B and Supplementary Figure 1). These results indicated that exogenous ABA not only enhances root hair length, but also promotes earlier root hair initiation. In addition, fluridone as an ABA biosynthetic inhibitor (10 μM) was used to investigate the effect of ABA on root tip responses under moderate water stress in rice and Arabidopsis (Yoshioka et al., 1998; Xu et al., 2013). The treatment of fluridon significantly repress the root hair length (Figures 1C,D). We measured the root hair length which located 4–5 mm distant from root apex in Figure 1D, this because considering the several ABA concentration treatment of root hair phenotype, we found that the best region to measure root hair length located at 4–5 mm distant from root tip.

To investigate whether ABA promotes root hair elongation through the major ABA signaling components, we constructed rice lines overexpressing SAPK10, which plays a positive role in the ABA signaling. Three of the SAPK10 overexpression lines showed ∼60-fold, ∼100-fold, and ∼130-fold increases in SAPK10 expression in the roots of the 7-day-old seedlings (Figure 2A). Then we measured the length of root hairs in these lines and found that the SAPK10 overexpression significantly enhanced root hair elongation (Figure 2B and Supplementary Figure 2). However, root hairs were rarely observed in the transgenic lines expressing OsABIL2, which is a negative regulator of rice ABA signaling (Figure 2C). The next to detect the root hair of SAPK10-OE and OsABIL2-OE response to ABA treatment, we would like to clarify the SAPK10-OE-1 was used in this study. Because this SAPK10-OE-1 line has moderate overexpression level of SAPK10 gene, which lead to moderate response to ABA treatment suitable for comparison with the wild type. While the other two lines SAPK10-OE-2, and -3 have much higher level expression of SAPK10 than the SAPK10-OE-1 line (Figure 2A), which lead to root hair elongation in the region very close to the root tip in the other two lines, rather than in the region 4–5 mm from the root tip after ABA treatment (Supplementary Figure 2). Considering these factors, we choose the SAPK10-OE-1 line for all other experiments for better comparison to the wild type after ABA treatments. With the 0, 0.1, 0.5, 1, and 2 μM ABA treatments (Figures 2C,D), the root hair length of the SAPK10-OE-1 line increased much more than that of the wild-type DJ. In contrast, ABA treatment did not increase root hair elongation of the OsABIL2-OE plants, which is a negative regulator of rice ABA signaling. This indicated that the effect of ABA on root hair elongation largely depends on the major PYR/PP2C/SnRK2 signaling pathway.

Auxin has a positive effect on root hair elongation without affecting the determination of root hair cell fate in Arabidopsis (Masucci and Schiefelbein, 1994, 1996; Pitts et al., 1998; Cho and Cosgrove, 2002). Therefore, we asked whether the ABA-regulated root hair elongation in rice also requires auxin transport. Then we used the auxin efflux inhibitor NPA to examine this. The 5-day-old wild-type seedlings were treated with ABA, NPA, or ABA plus NPA for 24 h. As shown in Figure 3A, ABA plus NPA treatment inhibited root hair elongation as compared with ABA alone. Similarly, in the SAPK10-OE line, ABA plus NPA also severely inhibited the root hair elongation as compared with the wild type shown in Figure 3B. We then examined whether additional auxin can rescue the reduced root hair elongation in the OsABIL2-OE, using the 5-day-old seedlings of OsABIL2-OE treated with 0.1 μM or 0.5 μM NAA for 24 h. As shown in Figure 3C, compared with Mock, 0.5 μM of NAA can significantly rescue the root hair length of OsABIL2-OE. Taken together, these results indicated that auxin acts downstream of ABA to mediate the ABA-induced root hair elongation.

To address whether ABA influences local auxin concentrations, we used a DR5-GUS rice line, which has been widely used as a marker line for monitoring endogenous auxin levels in plants (Yamamoto et al., 2007). The 5-day-old DR5-GUS seedlings were treated with 0, 0.1, 0.5, 1, or 2 μM ABA for 24 h, then we stained their roots for GUS activity. We found that DR5 was widely expressed in the crown roots, especially in their steles, and with increased concentrations of exogenous ABA, the GUS staining gradually accumulated close to the root tip and in the outer layers of the roots (Figure 3D). With a higher magnification, we observed that the region with the strongest GUS signal produced the longest root hairs (Figure 3E), suggesting that the ABA-promoted local auxin accumulation may lead to the specified root hair elongation.

We further tested whether the ABA-induced auxin response is dependent on ABA signaling by crossing the DR5-GUS into the OsABIL2-OE background; the wild type crossed with DR5-GUS (ZH11 background) was used as a control. The GUS staining showed that ABA treatment did not cause auxin redistribution in the OsABIL2-OE crown roots (Figure 3F), indicating that ABA signaling promotes auxin accumulation in the root tip to induce root hair elongation. The DR5-GUS staining patterns look different between Figures 3D,F. There may be two reasons. First, in Figure 3F, F₁ seeds were used for the observation, which contains half dosage of the reporter gene GUS. Therefore, the GUS staining apparently is weaker in Figure 3F than in Figure 3D. In general, the relative staining density and the trends of DR5-GUS expression pattern between Figures 3D,F are similar.

Auxin redistribution is mainly controlled by auxin transporters, which include the influx transporters of the AUXIN1/LIKE AUX1 (AUX1/LAX) family (Kerr and Bennett, 2007; Swarup et al., 2008; Zhao et al., 2015), and efflux transporters of the PIN-FORMED (PIN) families (Petrasek et al., 2006; Wang et al., 2009). To test how ABA signaling regulates auxin transport, we investigated the ABA-regulated expression of the major genes involved in auxin transport. We found that the mRNA levels of OsPINs and OsAUX1 were induced by treatment with 2 μM ABA at 6 and 24 h (Figure 4A).

To examine the spatial expression of these genes, we divided the root tip into three zones according to the root hair distribution and DR5-GUS activity: the short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip) (Figure 4B and Supplementary Figure 3). First, we tested expression of the ABA marker genes OsLEA3, OsBZ8, OsZEP1, and OsRD22 after ABA treatments (Figure 4C), and the result showed that all of the ABA marker genes (except OsRD22 in Tip) were induced in the three different zones. We then detected the expression of OsPINs in the three zones and found that most of the OsPINs were induced by the ABA treatment (Figure 4D). Second, we used the GUS reporter lines driven by various OsPIN promoters (OsPIN1b-GUS, OsPIN1c-GUS, OsPIN2-GUS, OsPIN5a-GUS, and OsPIN10a-GUS) (Wang et al., 2009) to visualize the OsPIN expression pattern. We found that the expression of OsPIN1b-GUS, OsPIN1c-GUS, and OsPIN5a-GUS was not dramatically altered by ABA treatment, but the expression of OsPIN2-GUS and OsPIN10a-GUS was significantly enhanced by ABA treatment, especially in the outer layers of the root (Figure 4E). In this assay, we indeed found that there were some different between real-time RT-PCR and GUS staining. In general, real-time RT-PCR or promoter-GUS has their own advantages and disadvantages. First, real-time RT-PCR cannot distinguish the specific expression tissue, which may mask some tissue specific signals. Second, the promoter-GUS approach has a defect on the specificity of promoter, because some times, ones do not known what the exact promoter size in plants. Therefore, the combination of these two approaches will show the expression pattern more accurately. This assay we wanted to show the genes expression fold change used the real-time RT-PCR, and GUS staining to show tissue specificity. In Arabidopsis, PIN2 is expressed in the epidermal cells (Wisniewska et al., 2006). We also checked the rice microarray expression database (RiceXPro³, OsPIN2 Locus ID: Os06g0660200, OsPIN10a Locus ID: Os01g0643300), and found that OsPIN2 and OsPIN10a were especially expressed in the epidermal cells (Supplementary Figures 4A,B). Therefore, we concluded that ABA-promoted root hair elongation likely requires functional basipetal auxin transport.

To obtain a global view of the differential expression of genes related to ABA-induced root hair elongation, the 5-day-old seedlings were treated without or with 0.5 μM ABA for 24 h, then the three different zones (SRH, LRH, and Tip) of root tips were collected to analyze transcript profiles (Supplementary Figure 3). We conducted two biological replicates and analyzed their repeatability by calculating the Pearson correlation coefficient (Supplementary Figure 5A), which indicates these results are consistent and repeatable.

To confirm the result that ABA promotes auxin transport and to validate the RNA-seq accuracy, we used RNA-seq to analyze the expression of OsPINs, OsAUX, and PINOID (PID). Arabidopsis PID encodes a serine/threonine protein kinase that regulates auxin redistribution through control of the subcellular localization of PINs (Robert and Offringa, 2008). As shown in Supplementary Figure 5B, ABA treatment promotes auxin transport by enhancing expression of OsPINs, OsAUX, and OsPID, especially in the Tip and LRH zones, consistent with the qRT-PCR (Figure 4D) and OsPINs-GUS staining results (Figure 4E). The results also showed a high correlation (R² = 0.84, Supplementary Table 2) between the RNA-seq and qRT-PCR data.

In addition to increasing auxin transport to the LRH zone, enhanced local auxin biosynthesis or reduced local auxin degradation could also increase local auxin concentrations, as indicated by the enhanced DR5-GUS expression in the LRH zone. Therefore, we first used our RNA-seq data to measure the expression levels of genes involved in auxin biosynthesis, including OsYUCCAs (Yamamoto et al., 2005), OASA (Kanno et al., 2004), OASB, OsNIT, OsAO, and OsAMI. These data showed that almost all of these auxin biosynthesis genes were upregulated after ABA treatment in the root tips (Figure 5A and Supplementary Table 4) and that OsAMI1 (∼3.6-fold), OsYUCCA4 (∼1.8-fold), and OsYUCCA1 (∼1.4-fold) were upregulated in the LRH zone.

We then checked the expression of the genes encoding IAA-amido synthetases of the GH3 family, key enzymes involved in the conversion of active IAA to an inactive form via conjugation of IAA with amino acids, such as Asp, Ala, and Phe (Fu et al., 2011). We also examined the genes encoding members of the UDP glycosyltransferase (UGT) family and dioxygenase for auxin oxidation (OsDAO) (Bowles et al., 2006; Zhao et al., 2013; Kasahara, 2016). In rice, 10 of the 14 genes involved in IAA inactivation were upregulated in the tip zone, and 9 of them were also upregulated in the LRH but not expressed in the SRH (Figure 5B and Supplementary Table 4), suggesting that the enhanced auxin transport and biosynthesis in the Tip and LRH zones of rice roots are the major reasons for local auxin accumulation, and the enhanced auxin inactivation is most likely caused by negative feedback regulation.

We also measured the endogenous IAA levels in the Tip, LRH, and SRH zones of the rice root. As shown in Figure 5C, auxin accumulated in the LRH zone, but not in the Tip, and the IAA concentration in the LRH zone after ABA treatment was significantly higher than in the untreated roots. We should note that the DR5-GUS staining shows strong signal in the root tip, indicating high concentration auxin in the root tip. However, we measured the IAA concentration in Tip, LRH, and SRH, and found the different IAA concentration between “Tip” and the typical root tip as shown in Supplementary Figure 3. We would like clarify here that the defined “Tip” here includes not only the classical root tip region but also the meristematic zone, transition zone, and root hair zone under the LRH, which covers a larger region. In addition, we used the roots from the 5-day-old seedling, and the majority of auxin may be from the shoot by transport.

To extend our analysis, we globally analyzed the genes differentially regulated in response to exogenous ABA in the three regions of root tips. We identified 1444, 838, and 855 genes that were differentially expressed before and after ABA treatment in Tip, LRH, and SRH, respectively. Of these, 1070, 353, and 477 genes were altered specifically in Tip, LRH, and SRH, respectively (Figure 6A). Further clustering analysis of these differentially expressed genes in each zone indicated that most of the ABA upregulated genes are in the Tip, followed by the LRH, and the fewest in the SRH (Figure 6B). These results suggest that the rice root tip cells rapidly sense and respond to ABA.

To understand the overall biological processes that occurred in the LRH zone in response ABA treatment, we performed gene ontology (GO) enrichment analysis of the 353 differentially expressed genes in the LRH; this identified 24 significantly (FDR < 0.05) enriched GO terms (Figure 6C) in two big categories. The first category includes GO terms involved in stress responses, such as “response to oxidative stress,” “response to chemical stimulus,” “response to stimulus,” and “response to stress.” The second category includes GO terms involved in “metabolic process,” such as “glycine metabolic process,” “small molecule metabolic process,” “primary metabolic process,” “cellular amino acid and derivative metabolic process,” “lipid metabolic process,” “cellular carbohydrate metabolic process,” and “polysaccharide metabolic process.” Then, we found 19 stress-responsive genes specifically responding to ABA in LRH (Supplementary Table 3). Nine of the 19 stress-responsive genes encode peroxidases, suggesting that reactive oxygen species have important roles in root hair growth in response to ABA treatment.

Because of the importance of auxin in root hair elongation, to understand whether these differentially expressed genes are related to ABA-regulated root hair elongation, we checked whether auxin affects the 353 genes that are differentially expressed in the LRH by comparing the genes to the auxin-regulated genes in the RiceXpro database (Sato et al., 2013). As shown in Figure 6D, among the 192 ABA upregulated genes, 57 (∼29.7%) were also upregulated by auxin (Table 1), and 56 of the 161 ABA downregulated genes were also repressed by auxin (Table 2). This suggests that auxin plays an important role in ABA-promoted rice root hair elongation, and the 113 genes co-regulated by both ABA and auxin in the LRH zone may function directly in regulating root hair elongation. We found that OsPIN9 and OsPP2C 59 are upregulated, further supporting the importance of polar auxin transport in root hair elongation.

To identify root hair-specific genes, we also screened the promoter regions of the 113 genes co-regulated by ABA and auxin (2000 bp upstream of the start codon) for the RHE sequence “WHHDTGNNN(N)KCACGWH” (where W = A/T, H = A/T/C, D = G/T/A, K = G/T, and N = A/T/C/G), as previously described (Won et al., 2009). We found 69 RHEs in 51 genes, with 13 genes carrying two or more RHEs (Table 3 and Supplementary Table 5). Because we lack root hair-specific gene expression data, we predicted whether these genes are specifically expressed in epidermal cells by searching the rice microarray expression database (RiceXPro) (Supplementary Table 5). These results indicated that 35 out of the 51 genes containing the RHE are highly expressed in epidermal cells in rice root tips.