The strains of A. thaliana used in this study were all Columbia-0 (Col-0) ecotype. A. thaliana mutants atg2 (SALK_076727), atg5 (SAIL_129_B07), atg7 (SAIL_11_H07), atg10 (SALK_084434C), and rboh (rbohd, CS9555; rbohf, CS9557 and rbohdf, CS9558) were obtained from the Arabidopsis Information Resources Centre. Seeds of plants expressing 35Spro:GFP-ATG8e and ATG8epro:GUS (Chen et al., 2015) were kindly provided by Prof. Shi Xiao (Sun Yat-sen University). All Arabidopsis seeds were surface sterilized in 20% bleach [1% (v/v) NaClO] and 0.01% Triton X-100 for 12 min, washed five times with sterile water, and stored at 4°C for 3 days. Then, the seeds were sown on plates containing half-strength MS medium with 1% sucrose and 0.8% (w/v) agar. The plates were vertically placed in growth chambers with a 16-h-light/8-h-dark cycle at 22°C.

All waterlogging treatments were performed in three independent biological replicates. Waterlogging treatment was carried out as described previously, with minor modifications (Wang et al., 2016). Briefly, 1-week-old seedlings were transferred to pots containing a stagnant liquid MS medium with 1% sucrose. The stagnant solution contained 0.2% (w/v) dissolved agar to prevent convective movement and was deoxygenated (dissolved O₂, <0.5 mg/L) by flushing with N₂ gas (Wiengweera et al., 1997; Yamauchi et al., 2017). The plant roots were completely immersed in stagnant solution, and leaves protruded 0.5 cm above the surface of the solution. The plants were placed in chambers at 22°C under a 16-h photoperiod. In order to calculate root length and survival rates after waterlogging, more than 20 plants of each genotype were waterlogged. Survival rates were based on the ability to produce new leaves and continue to grow.

Acid dyes, i.e., monodansylcadaverine (MDC) and GFP-ATG8 fusion proteins, are often used to detect autophagy in plants (Contento et al., 2005). Seedlings (1 weeks old) expressing GFP-ATG8e fusion protein were waterlogged with stagnant solution for the indicated times. GFP fluorescence was observed and photographed using a Leica SP8 laser scanning confocal microscope (Leica, Germany), and the excitation and emission wavelengths were 488 and 507 nm, respectively.

Monodansylcadaverine staining was carried out as described previously (Contento et al., 2005). Briefly, seedlings (1 weeks old) were waterlogged for the indicated times and subsequently incubated in 0.05 mM MDC (Sigma) in PBS for 10 min, followed by three washes with PBS at room temperature. Seedling were observed and photographed with a Leica SP8 laser scanning confocal microscope using a DAPI-specific filter. The excitation and emission wavelengths for MDC were 345 and 455 nm, respectively.

Each sample consisted of three independent biological replicates. Three roots per experiment were analyzed microscopically. Five optical sections per root were counted. The thickness of a section is 1 μm. The GFP-ATG8e-labeled and MDC staining punctate structures were counted using the Adobe Photoshop software¹.

We used diaminobenzidine (DAB) staining to detect H₂O₂ in situ as described previously, with minor modifications (Yang et al., 2014). Seedlings (1 weeks old) were incubated in 0.5 mg/mL DAB (Sigma) dissolved in 50 mM Tris–HCl (pH 5.0) for 2 h. The seedlings were rapidly washed three times with water and observed on a differential interference contrast microscope (Nikon 80i Eclipse). The quantitative measurement of staining intensity was carried out using Image pro plus 6.0.

Nitroblue tetrazolium (NBT) staining was performed to detect superoxide anion as described previously (Dunand et al., 2007). Briefly, seedlings (1 weeks old) were steeped in 2 mM NBT in 20 mM phosphate buffer (pH 6.1) for 15 min. The seedlings were transferred to distilled water and observed on a differential interference contrast microscope (Nikon 80i Eclipse). The quantitative measurement of staining intensity was carried out using Image pro plus 6.0.

For 2′,7′-dichlorofluorescin diacetate (DCFH-DA) staining to detect ROS (Soh, 2006; Kalyanaraman et al., 2012), 1-week-old seedlings were incubated at 37°C in 10 μM staining solution for 30 min. The DCFH-DA (sigma) was diluted from a 10 mM stock in ethanol using PBS. The stained seedlings were rapidly washed with PBS three times and were observed on a Leica SP8 laser scanning confocal microscope with an excitation wavelength of 488 nm and an emission wavelength of 525 nm. The quantitative measurement of fluorescence intensity was carried out using Image pro plus 6.0.

The content of H₂O₂ was measured as described previously (Patterson et al., 1984). The content of superoxide anion was determined according to hydroxylamine method (Elstner and Heupel, 1976).

For the GUS staining assay, 1-week-old seedlings were soaked in GUS assay buffer [0.1 M phosphate buffer (pH 7.0), 5 mM K₄Fe(CN)₆, 5 mM K₃Fe(CN)₆, 0.1% Triton X-100, and 0.5 mg/mL X-Gluc] at 37°C in the dark for 9 min and were then incubated in 75% ethanol. Seedlings were photographed with a differential interference contrast microscope (Nikon 80i Eclipse).

One-week-old Arabidopsis roots (1.0 g) prepared for enzyme activity were homogenized in an extraction solution containing 50 mM Na₂HPO₄–NaH₂PO₄ buffer, 0.2 mM EDTA, and 2% insoluble polyvinylpyrrolidone. The homogenate was centrifuged at 12,000 rpm for 20 min and the supernatant was used to determine enzyme activity. The activity of SOD, APX, and CAT was determined by using SOD assay kit, APX assay kit, and CAT assay kit purchased from Nanjing Jiancheng Bioengineering Institute according to manufacturer’s directions (Hussain et al., 2016).

Total RNA was extracted from 1-week-old Arabidopsis roots using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. The isolated RNA was reverse transcribed using the PrimeScript RT reagent kit with gDNA Eraser (Takara). Reverse transcription quantitative PCR (qRT-PCR) was performed with SYBR Premix Ex Taq II (Takara) using a StepOne Plus real-time PCR system (Applied Biosystems). The program for qRT-PCR was 95°C for 30 s, 40 cycles of 95°C for 5 s, and 60°C for 30 s. Each reaction had three biological and three technical replicates. ACTIN2 was used as the internal control and transcription levels were calculated according to the 2^-ΔΔCt formula. Information of genes included in this study are listed in Supplementary Table S1. Primers used for qRT-PCR analysis are listed in Supplementary Table S2.

The general procedures of preparing conventional samples for transmission electron microscopy (TEM), described previously, were followed with minor modifications (Xu et al., 2013). Briefly, middle areas of root sections (1 mm³) were cut from seedlings (1 weeks old) and fixed immediately in 2.0% paraformaldehyde and 2.5% glutaraldehyde in PBS overnight at 4°C. The tissues were rinsed three times using PBS, post-fixed in 1% osmium tetroxide for 4 h, rinsed three times using PBS again, dehydrated in a graded acetone series, and embedded in SPI-PON 812 resin (SPI Supplies). For TEM, the ultrathin sections were contrasted with uranyl acetate and lead citrate and were observed directly under the TEM (Hitachi H-7650, Japan) at 80 kV. For statistical analysis of the autophagic structures, five roots for per experiments and three cells from every root/root tissue were analyzed microscopically.

Total protein was extracted from 1-week-old Arabidopsis roots using an ice-cold RIPA lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) supplemented with 1 mM phenylmethanesulfonyl fluoride (PMSF; Roche). The homogenates were placed on ice for 30 min and were then centrifuged at 4°C for 15 min at 12,000 rpm. The supernatant was transferred to a new microfuge tube before electrophoresis.

For the immunoblot analysis, total proteins were separated by 15% SDS–PAGE and were then electroblotted to a nitrocellulose membrane (Biosharp). Anti-AOX (agersera), anti-Actin (sigma), anti-ATG8a (Abcam), and anti-GFP (Abcam) antibodies were used in the immunoblotting analysis. Quantification of the protein signal was performed using Image pro plus 6.0 software.

For phenotypic analysis, 1-week-old Arabidopsis roots under normal or waterlogged conditions were stained using 10 μg/mL propidium iodide (PI; Sigma) and were observed on a Leica SP8 laser scanning confocal microscope with an excitation wavelength of 535 nm and an emission wavelength of 615 nm.

For FDA staining, 1-week-old Arabidopsis roots under normal or waterlogged conditions were stained with 5 ug/mL Fluorescein diacetate (FDA; Sigma) for 5 min and were washed three times using sterile water. The fluorescence signal was observed on a Leica SP8 laser scanning confocal microscope with an excitation wavelength of 488 nm and an emission wavelength of 525 nm. The quantitative measurement of fluorescence intensity was carried out using Image pro plus 6.0.

For trypan blue staining, 1-week-old seedlings were placed at room temperature in trypan blue staining buffer (0.4% solution in PBS) for 10 min, followed by washing three times using PBS. Seedlings were photographed with a differential interference contrast microscope (Nikon 80i Eclipse).

For DAPI staining, 1-week-old Arabidopsis roots under normal or waterlogged conditions were stained in PBS containing 1 μg/mL 4′,6-diamidino-2-phenylindole (DAPI; Sigma) for 30 min and washed three times with PBS. The fluorescence signal was observed on a Leica SP8 laser scanning confocal microscope with an excitation wavelength of 358 nm and an emission wavelength of 461 nm.

A TUNEL assay was performed using an In Situ Cell Death Detection Kit (Roche, Germany) according to the manufacturer’s instructions with a few modifications. In brief, 1-week-old Arabidopsis roots were fixed in 4% (w/v) paraformaldehyde in PBS (pH 7.4) for 20 min. Then, the fixed seedlings were washed for 30 min with PBS before immersing in permeabilization solution (0.1% Triton X-100, 0.1% sodium citrate) for 2 min. After washing the samples three times with PBS, 50 μL aliquot of TUNEL reaction mixture was then added to the sample, and it was incubated for 1 h (37°C) in a humidified atmosphere in the dark. After being rinsed three times with PBS, photographs were taken under a Leica SP8 laser scanning confocal microscope with an excitation wavelength of 488 nm and an emission wavelength of 530 nm.

All the experiments were run in triplicate or more unless otherwise indicated, and the results reported in this study are presented as the mean ± SD. Five roots were used for every kind of experiment (ROS assays, GUS assays, and cell death detection). Data were analyzed by a two-tailed Student’s t-test using GraphPad Prism 7.0. The significance levels are ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001.

To validate the system of waterlogging, we analyzed the relative expression of the hypoxia-responsive genes ADH1, PDC1, PDC2, SUS1, SUS4, LDH, hemoglobin 1 (HB1), hypoxia-responsive unknown protein 43 (HUP43), and LOB domain-containing protein 41 (LBD41) (Hunt et al., 2002; Mustroph et al., 2010) in wild-type roots at 0, 4, 8, 12, and 24 h after waterlogging. The expression of the hypoxia-responsive genes ADH1, PDC1, SUS1, HB1, HUP43, and LBD41 started to increase within 4 h and peaked at 4 h (Supplementary Figure S1), and expression of PDC2, SUS4, and LDH peaked at 8 h under waterlogged conditions (Supplementary Figure S1). These results demonstrate that the system is appropriate to study the hypoxia response of waterlogged Arabidopsis roots.

We previously showed that accumulation of ROS in wheat root cells was induced by waterlogging (Xu et al., 2013). To investigate whether ROS are involved in the hypoxic response of waterlogged A. thaliana, we investigated the generation of ROS in wild-type roots subjected to waterlogging. DAB, NBT, and DCFH-DA staining were used to detect H₂O₂, superoxide anion, and ROS in situ. Pronounced accumulation of H₂O₂, superoxide anion (Figure 1A), and ROS (Figure 1B) was detected in the meristematic and elongation zones of the primary root under waterlogging conditions. In contrast, plants grown under normal conditions had relatively low ROS levels (Figure 1A–D). We also found that the ROS in the root stele seemed to be higher. To further confirm this, we determined the level of H₂O₂ and superoxide anion in wild-type roots under waterlogging conditions. The content of H₂O₂ in Arabidopsis roots was significantly upregulated within 24 h (Figure 1E). The content of superoxide anion started to increase within 4 h and peaked at 12 h (Figure 1F). These findings suggest that waterlogging induced accumulation of ROS.

RBOHD and RBOHF are the most in-depth studied enzymatic ROS-generating systems, and reports of their participation in various plant processes have increased considerably in recent years. To obtain genetic evidence of the response of ROS to waterlogged conditions, we obtained rbohd, rbohf, and rbohdf mutants and characterized their responses to waterlogging. The results showed that waterlogging markedly suppressed the growth of all plants. Of note, the double mutants and the two single mutants grew more slowly than the wild type (Supplementary Figure S2A). The root length of rbohdf was clearly less than that of wild type, rbohd and rbohf after exposure to waterlogged conditions (Supplementary Figure S2B). Moreover, we used DAB and NBT staining to detect ROS accumulation in roots of wild type and rboh mutants. DAB and NBT staining results showed that the root tips of rbohd, rbohf, and rbohdf mutants accumulated less H₂O₂ and superoxide anion than those of the wild type after waterlogging (Supplementary Figures S2C–E). Together, these results indicate that both RBOHD and RBOHF are required for waterlogging induced ROS generation.

The fact that AOX can use reductant in excess of either the Cyt pathway capacity or the rate of ATP use suggests that AOX may play an important role in reducing the generation of ROS (Umbach et al., 2005). To test whether waterlogging affected the expression of the AOX protein, total endogenous protein extracted from wild-type roots was immunoblotted with specific monoclonal anti-AOX antibody. The levels of the AOX protein increased upon waterlogging and accumulated to the highest level at 8 h (Figure 2A). As shown in Figure 2B, the gray values of the total AOX protein and the control are significantly different after treatment for 4, 8, and 12 h.

In A. thaliana, there are five AOX genes: AOX1a, AOX1b, AOX1c, AOX1d, and AOX2 (Giraud et al., 2008). Because the expression of these five AOX genes in response to various stresses is complex, we investigated their transcription levels in wild-type roots after waterlogging. qRT-PCR analyses showed that only the transcription level of AOX1a was significantly increased. The transcription level of AOX1a started to increase within 4 h, and peaked at 8 h (Figure 2C), which was consistent with the change in the AOX total protein level (Figure 2A). In conclusion, waterlogging may be affecting only the expression of AOX1a.

To counteract oxidative stress in plant cells, three enzymes, SOD, APX, and CAT, are crucial. To further investigate enzymatic mechanisms that scavenge overproduced ROS, we measured the activities of SOD, APX, and CAT in wild-type roots during waterlogging. The activities of SOD, CAT, and APX in Arabidopsis roots were significantly upregulated within 24 h (Figure 2D–F). The activities of SOD, CAT, and APX started to increase within 4 h and peaked at 8 h (Figure 2D–F). Together, these findings indicate that activities of antioxidant enzymes were enhanced under waterlogged conditions.

To investigate the role of waterlogging in autophagosome formation in plant cells, we used Arabidopsis plants expressing the GFP-ATG8e fusion protein as a marker (Contento et al., 2005). Seedlings expressing GFP-ATG8e were waterlogged for 4 and 8 h and GFP fluorescence was analyzed using confocal microscopy. In contrast to the root cells of the control, there were more GFP-ATG8e labeled dots and ring-like structures in wild-type root tips and mature root cells subjected to waterlogging (Figure 3A,B). At the beginning of autophagy, GFP-ATG8e that appeared in the shape of dots formed autophagosomes with other autophagy-associated proteins and subsequently fused with the vacuole. After entering the vacuole, GFP-ATG8e was sheared into free GFP and ATG8 hydrolysis fragments by acidic hydrolases. Since the GFP protein is very stable under acidic conditions, it is possible to determine the intensity of autophagy by determining the degradation of GFP-ATG8e (Ichimura et al., 2004). Thus, the appearance of free GFP can be used to monitor the extent of autophagy. The waterlogging treatment enhanced the degradation of GFP-ATG8e, as measured by the accumulation of free GFP, in comparison to the control (Figure 3C).

To investigate autophagy characteristics upon waterlogging, we examined ATG8e expression in Arabidopsis roots using seedlings expressing GFP-ATG8e and ATG8epro-GUS. PI staining was used to distinguish the structure of Arabidopsis roots. In contrast to the control, there were more GFP-ATG8e-labeled dots and ring-like structures initially observed in the root stele cells after 4 h of waterlogging followed by epidermis of the root after 8 h of waterlogging (Figure 4A). After waterlogging, the appearance of GFP-labeled autophagosome-like structures was significantly increased in the wild-type root stele cells (Figure 4B). GUS staining revealed that the expression of ATG8e initially intense in the root stele cells after 4 h of waterlogging followed by other parts of the root after 8 h of waterlogging (Figure 4C). This result shows that waterlogging-activated autophagy initially occurs in the root stele. To further confirm this, we used TEM to examine ultrastructure of Arabidopsis root. Analysis of cross sections by TEM revealed autophagy-related structures increased in the root stele after 4 and 8 h of waterlogging treatment and in epidermis of the root after 8 h of waterlogging treatment (Supplementary Figure S3), whereas no obvious difference was detected in other parts of the root (Supplementary Figure S3).

Previous studies have implied the function of autophagy under stress conditions (Bassham et al., 2006; Han et al., 2011). To investigate the role of autophagy in the response to waterlogging, we transferred 1-week-old atg mutant plants (atg2, atg5, atg7, and atg10), alongside wild-type plants, to hypoxic conditions induced by waterlogging. We analyzed the atg mutants that showed enhanced hypersensitivity to waterlogging, waterlogging + 1% alcohol, and waterlogging + 1% perhydrol (Supplementary Figure S4A). As shown in Supplementary Figure S4B, the primary roots of atg mutants were significantly shorter than those of that wild type after waterlogging and the waterlogging + 1% alcohol treatment. During the course of waterlogging + 1% perhydrol stress, most of the wild-type plants survived, but more than half of the atg mutants died (Supplementary Figure S4C).

To further evaluate the function of autophagy in the waterlogging response, we transferred 1-week-old wild-type and atg mutant seedlings (atg5 and atg7) to waterlogged conditions, and then MDC staining was used to observe autophagosomes. Accumulation of MDC-stained autophagosomes was observed in root tips and mature root cells of wild-type plants subjected to waterlogging (Supplementary Figure S4D), whereas the number of autophagosomes was significantly decreased in the atg mutant root cells after waterlogging (Supplementary Figure S4E).

Under waterlogged incubation conditions, accumulation of ROS and autophagy are induced. To investigate the relationship between ROS and autophagy, qRT-PCR was used to quantify the expression levels of ATG genes in the wild-type and rboh mutants exposed to waterlogging. In comparison to the wild type, the expression of ATG genes was significantly downregulated in the rboh mutants (Figure 5A). To further confirm the qRT-PCR data, we observed autophagosomes of rboh mutants following waterlogging using MDC staining. Upon waterlogging exposure, the appearance of autophagosome-like structures was decreased in the rbohd and rbohf mutants in comparison with the wild type (Figure 5B,C). We also used TEM to monitor the autophagic activity after waterlogging. The results showed that lower levels of autophagic activity were detected in the rboh mutants than in wild type after waterlogging (Supplementary Figures S5A,B). These data demonstrated that the occurrence of autophagy is significantly decreased in rboh mutants after waterlogging.

In order to further investigate the interrelation between ROS and autophagy, we analyzed the mRNA expression level of biosynthesis and scavenging genes of ROS in the wild-type and atg mutants upon waterlogging. In comparison to the wild type, the transcription level of RBOHD and RBOHF was notably increased in the atg mutants at 0 and 4 h of waterlogging (Figure 6A). However, the expressions of SOD, CAT1, APX1, and APX2, which constitute the enzymatic scavenging system to eliminate ROS in plants, were significantly reduced in the atg mutants in comparison to the wild type (Figure 6A). In that case, the atg mutant plants exhibited excessive accumulation of ROS compared with the wild type under waterlogging stress. In order to further verify higher levels of ROS in atg mutants, we transferred 1-week-old atg mutant plants, alongside wild-type plants, to DCFH-DA dye. DCFH-DA staining showed higher accumulation of ROS in atg mutant plants under waterlogging conditions, in comparison to wild type (Supplementary Figures S6A,B). In comparing the DAB and NBT staining in root tips of the wild-type and atg mutants, we found the H₂O₂ and superoxide anion levels were both greatly elevated in the atg mutants after waterlogging treatment, compared with those of the wild type (Figure 6B–D). Waterlogging treatment significantly enhanced the production of ROS in both wild-type and atg mutants (Figure 6B–D and Supplementary Figure S6). In summary, accumulation of ROS is increased in atg mutants after waterlogging.

Arabidopsis hypocotyls form lysigenous aerenchyma as a consequence of cell death of secondary xylem during hypoxia (Mühlenbock et al., 2007). To investigate the PCD process upon waterlogging, 1-week-old wild-type seedlings were treated with PI, which cannot pass through a live cell membrane, but can cross a damaged cell membrane and stain the nucleus. PI staining showed that cell death started at 24 h in the primary root, whereas under normal conditions it was hardly detected (Figure 7A). We used FDA staining to show the effects of waterlogging on cell viability in the meristematic and elongation zones of wild-type roots after waterlogging for 24 and 48 h. Arabidopsis plants in the control showed strong FDA fluorescence, whereas weak fluorescence was detected under waterlogging conditions (Figure 7B,C). Typical characteristic of PCD is the occurrence of morphological changes in the nucleus and the cleavage of genomic DNA at internucleosomal sites by endogenous nucleases. To further determine whether the root cell death induced by waterlogging is a kind of PCD, we used DAPI staining and TEM to show the effects of waterlogging on the chromatin condensation of wild-type roots. Confocal images and TEM images showed that a marked increase in DAPI fluorescence and condensed and moon-shaped nuclei were detected under waterlogging conditions (Figure 7D). We also investigated the internucleosomal fragmentation of the DNA by TUNEL assay. TUNEL-positive signals were detected in the meristematic zone to the elongation zone from root tips after waterlogging treatment (Figure 7E). These results indicate that waterlogging induced cell death in Arabidopsis roots is a kind of PCD.

To examine whether autophagy is also involved in the PCD process in Arabidopsis roots, 1-week-old atg mutant plants were waterlogged. FDA staining was used to observe cell viability of waterlogged wild-type and atg mutants. The fluorescence intensity of atg mutants was obviously weaker than that of the wild type in the waterlogged treatment (Figure 8A). However, the fluorescence intensity of wild-type and atg mutants grown under normal conditions was fairly similar (Figure 8B). Waterlogging triggered cell death, as indicated by the blue color, in root of the wild-type and atg mutants (Figure 8C). In comparison to the wild type and controls, much higher levels of cell death were observed in the roots of atg mutants after waterlogging (Figure 8C). To further evaluate the role of autophagy in PCD, we examined cell death in Arabidopsis root subjected to waterlogging. In comparison to the wild type, cell death was significantly increased in the atg mutants after exposure to waterlogging. And cell death was mainly concentrated in root stele (Figure 8D). In conclusion, autophagy has effect on PCD in Arabidopsis roots during waterlogging.