Earlier Degraded Tapetum1 (EDT1) Encodes an ATP-Citrate Lyase Required for Tapetum Programmed Cell Death^,

The ATP-citrate lyase, Earlier Degraded Tapetum1 (EDT1), plays a fundamental role in rice pollen development and is involved in cellular metabolism and tapetum programmed cell death.

Abstract

In flowering plants, the tapetum cells in anthers undergo programmed cell death (PCD) at the late meiotic stage, providing nutrients for further development of microspores, including the formation of the pollen wall. However, the molecular basis of tapetum PCD remains elusive. Here we report a tapetum PCD-related mutant in rice (Oryza sativa), earlier degraded tapetum 1 (edt1), that shows complete pollen abortion associated with earlier-than-programmed tapetum cell death. EDT1 encodes a subunit of ATP-citrate lyase (ACL), and is specifically expressed in the tapetum of anthers. EDT1 localized in both the nucleus and the cytoplasm as observed in rice protoplast transient assays. We demonstrated that the A and B subunits of ACL interacted with each other and might function as a heteromultimer in the cytoplasm. EDT1 catalyzes the critical steps in cytosolic acetyl-CoA synthesis. Our data indicated a decrease in ATP level, energy charge, and fatty acid content in mutant edt1 anthers. In addition, the genes encoding secretory proteases or lipid transporters, and the transcription factors known to regulate PCD, were downregulated. Our results demonstrate that the timing of tapetum PCD must be tightly regulated for successful pollen development, and that EDT1 is involved in the tapetum PCD process. This study furthers our understanding of the molecular basis of pollen fertility and fecundity in rice and may also be relevant to other flowering plants.

During the male gametogenesis in flowering plants, anther walls are built with four structure layers, from the interior to exterior: the tapetum, middle layer, endothecium, and epidermis (Zhang et al., 2011; Walbot and Egger, 2016). The tapetum cells comprise the innermost sporophytic layer and play an important role in pollen development by degrading callose to release microspores from tetrads. Meanwhile, tapetum cells can provide the necessary nutrients for microspore development and for storing and transporting the sporopollenin precursors for sexine formation (Yi et al., 2016). The tapetum cells undergo programmed cell death (PCD) after meiosis of microspore mother cells. Many studies have shown that the tapetum PCD must occur at the right time during anther development. Either early or belated turn-on of the PCD would lead to male sterility (Ko et al., 2014; Yi et al., 2016).

The tapetum PCD is a complex cellular process and is completed in a short time, involving turn-on or turn-off of expression of many specific genes. According to previous reports, two basic helix-loop-helix (bHLH) transcription factors, TAPETUM DEGENERATION RETARDATION (TDR) and ETERNAL TAPETUM 1 (EAT1), promote tapetum PCD (Li et al., 2006; Niu et al., 2013). TDR directly binds to the promoter region of OsCP1 that encodes a putative Cys protease referred to the release of microspores from the tetrads (Lee et al., 2004), and to the lipid transfer protein OsC6, which has lipid-binding activity and is essential for pollen wall development (Zhang et al., 2010). The eat1 mutant decreases the expression of two aspartic protease genes, AP25 and AP37, which promote cell death in yeast (Saccharomyces cerevisiae) cells and plants (Niu et al., 2013). In addition, many studies have established that the PCD process of tapetum is associated with accumulation of reactive oxygen species (ROS; Van Hautegem et al., 2015; Yi et al., 2016). Punctual production of ROS is critical for proper beginning of the tapetum PCD. MADS3 plays a key role in the modulation of late rice (Oryza sativa) anther development by regulating ROS homeostasis through interacting with the ROS-scavenging protein MT-1-4b (Hu et al., 2011). The ROS scavenger protein COX11 is a nuclear-encoded mitochondrial protein that can interact with the tapetum mitochondrial protein WA352 and determine cytoplasmic male sterility in rice (Luo et al., 2013). OsMT2b, coding another ROS scavenger protein, is highly expressed in early tapetum development and interacts with DEFECTIVE TAPETUM CELL DEATH 1 (DTC1) to inhibit ROS scavenging activity (Yi et al., 2016). When ROS bursts out, it can trigger premature tapetum PCD and lead to pollen abortion (Luo et al., 2013). In Arabidopsis (Arabidopsis thaliana), loss or gain of function of the RESPIRATORY-BURST OXIDASE HOMOLOG E (RBOHE) results in delayed or precocious tapetum degeneration by affecting ROS levels (Xie et al., 2014).

The accurate timing of tapetum degradation provides microspores with substances and energy for their further development. The pollen exine, composed of sporopollen, is considered to be one of the main protective barriers for microspores/pollens against various biological and environmental stresses (Ahlers et al., 2003; Li et al., 2010; Zhang et al., 2010). Moreover, limited evidence showed that sporopollenin is composed of lipid, phenolic hydroxyl groups, and aliphatic polyhydroxy compounds (Ahlers et al., 2000). Acetyl-coenzyme A (acetyl-CoA) is a fundamental building block of carbon metabolism in eukaryotes (Bazilevsky et al., 2019) and an important raw material for lipid synthesis in all organisms (Oudejans et al., 1983; Page et al., 1994). Acetyl-CoA is a central intermediate that mediates energy flowing and material circulation, and its molecule synthesis is essential for growth and adaptation to stress (Morrish et al., 2010; Cai et al., 2011). In plants, acetyl-CoA is found in at least five different subcellular compartments with differential roles: as a substrate for acetylation in the nucleus, for de novo fatty acid biosynthesis in the plastids, incorporated into the tricarboxylic acid cycle in the mitochondria, for production of fatty acids by β-oxidation and conversion of fatty acids into sugars in the peroxisomes, and for biosynthesis of a large battery of important compounds in the cytosol (Gengenbach et al., 2001; Fatland et al., 2002, 2005; Davidj et al., 2009; Morrish et al., 2010; Kong et al., 2017, 2018; Li et al., 2019).

Although acetyl-CoA is at the center of metabolism, it is membrane impermeable and must be synthesized separately in these subcellular compartments (Chen et al., 2012). In cytosol, acetyl-CoA is synthesized by ATP-citrate lyase (ACL), which catalyzes the ATP-dependent conversion reaction of citrate and coenzyme A to form acetyl-CoA and oxaloacetic acid (Fatland et al., 2002; Xing et al., 2014). In plants, ACL is a cytoplasmic heterologous protein consisting of two subunits (ACLA and ACLB), probably in an A₄B₄ configuration (Fatland et al., 2002). ACLA and ACLB are encoded by three and two genes, respectively, in Arabidopsis (Fatland et al., 2002; Hu et al., 2015). Previous study has shown that plants with reduced ACL activity achieved using antisense RNA technology have a complex, bonsai phenotype coupled with reduced accumulation of cuticular wax and excessive accumulation of starch and anthocyanin (Fatland et al., 2005). These findings indicate that an adequate ACL-produced cytosolic acetyl-CoA pool is essential for normal plant growth and development and that acetyl-CoA produced in other organelles could not compensate for deficiency in this pool (Fatland et al., 2005). In addition, it was also found that high levels of ACLA and ACLB mRNA transiently accumulated in the tapetum cells in anthers of Arabidopsis during flower bud development (Fatland et al., 2002), suggesting that ACL may play a role in regulating the development of anthers. However, it is not clear whether ACL is involved in the control of tapetum PCD, especially in rice.

In this study, a male sterile mutant, named earlier degraded tapetum (edt1), was identified. Map-based cloning revealed that EDT1 encodes a subunit of ACL that is essential for energy metabolism. Our results indicated that EDT1 interacts with ACLB-1 and can form a heteromultimeric protein to maintain the stability of cellular lipid metabolism and ROS homeostasis. We demonstrated that EDT1 plays an essential role in rice pollen development by regulating tapetum PCD.

RESULTS

Phenotypic Characteristics of the edt1 Mutant

The plant height of edt1 (77.2 ± 2.6 cm) was reduced compared to the wild type (98.5 ± 3.3 cm), while other traits (such as leaf number and tiller number) showed no obvious difference between the wild type and edt1 during vegetative growth and before the heading stage (Fig. 1A). At the anthesis stage, however, the anthers of edt1 were smaller in size and appeared white compared to wild-type anthers (Fig. 1, B and C). The mutant anthers could not be stained by I₂-KI and failed to produce any viable pollen (Fig. 1D). To investigate the developmental defect of the edt1 pollen grains, we performed carmine acetate dyeing experiments on edt1 and wild-type pollen of different stages. In wild-type anthers, the tetrad broke down after meiosis and microspores were released (Fig. 2, A and B). The released microspores developed further to form bicellular pollen grains, and eventually formed mature pollen grains (Fig. 2, C–E). In contrast, the edt1 mutant appeared normal only at meiosis through the young microspore stage (Fig. 2, F and G), but the released edt1 pollens failed to develop further from the later microspore stage to the early bicellular pollen stage and eventually underwent abortion (Fig. 2, H–J).

Figure 1.

Morphological comparison between rice wild type (WT) and earlier degraded tapetum 1 (edt1) mutant. A, Phenotypes of wild-type and edt1 mutant plants after bolting. B, Spikelets of wild-type and edt1 mutant plants after removal of the lemma and palea. C, Anthers of wild-type and edt1 plants at the heading stage. D, Anthers of wild type and edt1 stained with IKI solution. Note that staining of starch-rich pollen grains was seen only in the wild-type anther. Bars = 15 cm (A), 0.7 mm (B), and 0.5 mm (C and D).

Figure 2.

Microspore and tapetum development in the wild type and the edt1 mutant. A to J, Pollen grains of the wild type (A–E) and edt1 (F–J) at different stages were stained with carmine acetate. Grains at stages 8b (A and F), 9a (B and G), 9b (C and H), 10 (D and I), and 14 (E and J) are shown. Note the abnormal development and degradation of microspores in edt1 (arrowheads). DM, degraded microspores. Bars = 20 µm. K to R, Transverse section analysis of wild-type (K–N) and edt1 (O–R) anthers at different developmental stages (8–10) stained with toluidine blue. The images are of cross sections through single locules at stages 8a (K and O), 9a (L and P), 9b (M and Q), and 10 (N and R). Bars = 15 μm. DT, degraded tapetum; GP, germination pore; MC, meiotic cell; MSP, microspores; T, tapetum.

To further characterize edt1 for the pollen abortion, we observed detailed structural changes by cross-sectioning anthers of different stages. During meiosis, dyads were formed at stage 8 in wild-type anthers, and tapetum cells differentiated to form large vacuoles (Fig. 2K). Starting from stage 9, the tapetum concentrated faster and the cytoplasm stained heavily, and young microspores were separated from the tetrad and released into the anther locules (Fig. 2L). At early stage 10, microspores expanded and vacuolized (Fig. 2M), and then tapetum cells rapidly degenerated at late stage 10 (Fig. 2N). At stage 8, no substantial differences were observed between edt1 and the wild type (Fig. 2O). At stage 9, there was no morphological difference in microspores, but the edt1 tapetum appeared to be less concentrated and faintly stained (Fig. 2P). At early stage 10, edt1 microspores were partially vacuolized and mostly irregular in shape (Fig. 2Q). At late stage 10, the broken tapetum clearly surrounded the degraded microspores in the mutant anther chamber (Fig. 2R). These results suggest that the developmental defect of edt1 pollens may be associated with the change in tapetum degradation.

The edt1 Mutant Exhibits Precocious Tapetum PCD and an Abnormal Pollen Wall

To better understand the developmental disorder of the edt1 pollen, we observed the edt1 and wild-type anthers at the subcellular level by transmission electron microscopy (TEM). At stage 8, there was no substantial difference in anthers between the wild type and edt1, and the tapetum cells differentiated to form large vacuoles normally (Fig. 3, A–D). By late stage 9, the cytoplasm of wild-type tapetum cells was concentrated, and the Ubisch bodies were uniformly distributed on the inner surface of the wild-type tapetum (Fig. 3E). However, the tapetum structure was loosened and cell contents were reduced in edt1, and no normal Ubisch body was observed (Fig. 3F). In addition, mitochondria were clear and electron dense in the wild type, whereas mitochondrial degradation residues were seen in edt1 at this stage (Fig. 3, G and H). At stage 10, the cytoplasm of wild-type tapetum cells was further condensed into stripes and degraded substantially (Fig. 3I), while the tapetum cells in edt1 condensed irregularly and surrounded the vacuolized microspores (Fig. 3J). At stage 11, in both the wild type and edt1, the content of tapetum was degraded almost completely, but the remaining cell wall was not degraded (Fig. 3, K and L). The tectum and nexine I of the edt1 pollen wall became very thin, while the intine of pollen was thickened, and the endexine II was faulty and irregularly distributed compared to the wild type (Fig. 3, M and N). The mutant pollen wall did not develop, and starch granules observed in wild-type microspores at the mature stage were not present in the mutant (Fig. 3, O and P). Based on these observations, we speculate that earlier mitochondrial degradation may trigger precocious tapetum cell collapse and subsequently lead to defective pollen walls.

Figure 3.

Transmission electron microscopy of anthers in the wild type (WT) and the edt1 mutant at different developmental stages. A and B, Stage 8 anthers from the wild type (A) and edt1 (B). C and D, Close-ups of the boxed areas in A and B, respectively. E and F, Stage 9 anthers from the wild type (E) and edt1 (F). G and H, Close-ups of E and F, respectively, showing mitochondria (Mt) and other organelles. I and J, Stage 10 anthers from the wild type (I) and edt1 (J). K and L, Stage 11 anthers from the wild type (K) and edt1 (L). Arrows in H indicate the degraded mitochondria and the arrowheads in E and K indicate the Ubisch bodies (Ub). M and N, Transverse sections of the microspore walls from the wild type (M) and edt1 (N) at stage 10, showing tectum (Te), nexine I (Ne), bacula (Ba), endexine II (End), and intine (In). O and P, Transverse sections of the microspores of the wild type (O) and edt1 (P) at stage 12. Ba, bacula; Cy, cytoplasm; DEnd, defective endexine II; DNe, defective nexine I; DTe, defective tectum; DP, degraded pollen; DT, degraded tapetum; End, endexine II; In, intine; MP, mature pollen; Msp, microspores; Mt, mitochondrion; Ne, nexine I; T, tapetum; Te, tectum; Ub, Ubisch body. Bars = 3 μm (A–J, M, and N), 0.5 μm (K and O), and 1.5 μm (L and P).

To understand whether there is any change in PCD in the edt1 tapetum cells, we investigated DNA fragmentation using terminal deoxynucleotidyl transferase-mediated deoxy-UTP nick-end labeling (TUNEL) assays on wild-type and edt1 anthers at stages 6–9 (Supplemental Fig. S1). In wild-type tapetum cells, TUNEL-positive signals (yellow colored signals) did not appear until stage 8 (during late meiosis; Fig. 4, A–C), and a strong TUNEL signal was detected at stage 9 (the young microspore stage; Fig. 4D). There was no TUNEL positive signal seen in the edt1 tapetum cells at stage 6 (Fig. 4E). However, slightly fragmented DNA signal appeared at stage 7 (early meiosis; Fig. 4F) and the strongest TUNEL signal was seen at stage 8 in edt1 (Fig. 4G). Then the TUNEL signal disappeared in the later stages (Fig. 4H). The results show that precocious tapetum PCD occurred in edt1.

Figure 4.

DNA fragmentation in wild-type (WT) and edt1 mutant anthers. A to H, Detection of nuclear DNA fragmentation by TUNEL assay in anthers of the wild type (A–D) and edt1 (E–H) at stages 6 (S6) through 9 (S9). Nuclei were stained with propidium iodide as red fluorescence. The yellow fluorescence is the merged signal from TUNEL-positive nuclei staining (green) and propidium iodide staining (red). The arrows in C, D, F, and G indicate TUNEL-positive signals in tapetum cells. Bars = 50 μm. I to L, Detection of nuclear DNA fragmentation by comet assays. DNA damage level was assessed in wild-type (I and K) and edt1 (J and L) anthers at S6 (I and J) and S7 (K and L). Bars = 20 μm. M, Quantification of the DNA damage in S6 through S9. The extent of DNA damage in each nucleus is indicated by the units 0, 1, 2, or 3. An increased unit correlated with a larger comet tail and a smaller comet head, as illustrated in the inset in M. The final DNA damage value was obtained by summing the damage units of 50 nuclei per slide. Data are the mean ± sd (n = 3). The asterisks indicate statistical significance at P ≤ 0.05 by Student’s t test.

We quantified the DNA fragmentation using comet assay and compared the extent of DNA damage between wild-type and edt1 anthers. At stage 6, DNA damage in edt1 anthers was similar to that in the wild type (Fig. 4, I and J). However, at stages 7–9, the mutant anthers had higher levels of DNA damage than the wild-type anthers (Fig. 4, K–M). The results from the comet assay once again confirmed that the tapetum cells of edt1 were degraded earlier.

EDT1 Encodes ACLA

To isolate the EDT1 gene, we first mapped the edt1 locus between the two insertion-deletion markers B3 and B4 on chromosome 11. To narrow down the region further, 742 mutant plants were identified from a F2 population and used to anchor the locus to a region of 543 kb between markers B9 and B10 (Fig. 5A). Sequencing and comparison analysis identified a deletion of 147 kb in edt1 (Supplemental Fig. S2). This deletion contains eight annotated genes and seven genes encoding putative expressed proteins (Rice Genome Annotation Project, http://rice.plantbiology.msu.edu/; Supplemental Table S2). Quantitative real-time PCR showed that Os11g0696200, a gene encoding ACL, has the highest expression in anthers (Supplemental Fig. S3). Thus, Os11g0696200 was considered a candidate gene for EDT1 (Fig. 5B).

Figure 5.

Isolation of the EDT1 gene. A, Fine mapping of the edt1 locus. B1 to B10, D8, and D13 are markers developed in this work (Supplemental Table S1). The edt1 locus was localized to a 543-kb region between the two markers B9 and B10. The edt1 genome contains a 147-kb deletion in which the EDT1 gene is located. The number of recombinants is indicated below the map. CEN, centromere. B, Genomic structure of EDT1 and a point mutation in the c-edt1 mutant. The c-edt1 genome contains an extra base A in the marked site in exon 3. Lines indicate introns and boxes indicate exons. The ATP-grasp domain and the citrate-binding domain in EDT1 are marked with dark blue and light blue, respectively. Anthers of the wild type (C), edt1 (E), a transgenic pEDT1-EDT1 line (G), and a mutant c-edt1 created by CRISPR-Cas9 (I). IKI staining of the pollen grains from wild-type (D), edt1 (F), pEDT1-EDT1 (H), and c-edt1 (J) plants. K and L, TEM images showing the morphology of tapetum cells (K) and the microspore wall (L) of the pEDT1-EDT1 plant. The arrowhead in K indicates the Ubisch bodies (Ub). M and N, Transverse sections of tapetum cells (M) and the microspore wall (N) in c-edt1. Ba, bacula; Cy, cytoplasm; DNe, defective nexine I; DTe, defective tectum; In, intine; Ne, nexine I; T, tapetum; Te, tectum; Ub, Ubisch body. Bars = 1 mm (C, E, G, and I), 20 μm (D, F, H, and J), 3 μm (K and M), and 0.5 μm (L and N).

We performed functional complementation and clustered regularly interspaced short palindromic repeats -CRISPR associated protein 9 (CRISPR-Cas9) experiments to verify whether Os11g0696200 is responsible for EDT1 (Fig. 5, C–N). The complemented lines (pEDT1-EDT1) carrying a genomic sequence of Os11g0696200 were able to restore pollen fertility in edt1 to a level similar to that observed in the wild type (Fig. 5, C–H; Supplemental Table S3). In the TEM observation, the cytoplasm of tapetum cells at stage 9 was concentrated and the Ubisch bodies could be found on the inner surface of the complemented line’s tapetum, as seen in the wild type (Fig. 5K). The pollen wall of complemented lines was also thickened and did not stagnate like edt1 (Fig. 5L). In addition, the EDT1 protein level also recovered to the wild-type level (Supplemental Fig. S4). Subsequently, we knocked out Os11g0696200 by the CRISPR-Cas9 method and produced a mutant plant designated c-edt1. The c-edt1 mutant had a single base insertion between nucleotides 1,306 and 1,307 (on the third exon), causing a frame shift and resulting in a premature translation termination at the C terminus (Fig. 5B). The TEM observation showed that the ultrastructure of the c-edt1 tapetum at stage 9 was loose and its contents reduced, the Ubisch body had almost disappeared, and the pollen wall was thin, all mimicking the edt1 phenotype (Fig. 5, M and N). Similar results were also obtained in the EDT1 knockdown lines (Supplemental Fig/ S5; Supplemental Table S3). Taken together, these observations led us to conclude that Os11g0696200 is EDT1.

EDT1 consists of 11 exons and 10 introns and has a full-length coding DNA sequence of 1,272-bp. The open reading frame of EDT1 encodes a putative A subunit of ACL. EDT1 is 423 amino acids long and contains an ATP-grasp domain at the N terminus and a citrate-binding structure at the C terminus (Fig. 5B; Supplemental Fig. S6). A phylogenetic relationship analysis of the ACL proteins revealed that eight species are closely related in evolution, including Arabidopsis, Brachypodium distachyon, Oryza brachyantha, Oryza sativa, Sorghum bicolor, Setaria italica, Triticum urartu, and Zea mays. In Arabidopsis, ACLA was encoded by three members, ACLA-1, ACLA-2, and ACLA-3, among which ACLA-1 and ACLA-2 have the closest genetic relationship, and there are also great similarities in sequence (89% identity). ACLB is encoded by two members, ACLB-1 and ACLB-2. However, in rice, there are only two members encoding ACLA (ACLA-2, referred to herein as EDT1, and ACLA-3) and one gene encoding ACLB (ACLB-1; Supplemental Fig. S7).

Expression Pattern and Subcellular Localization of EDT1

Reverse-transcription quantitative PCR (RT-qPCR) analysis of wild-type plants showed that EDT1 expression was lower in the culm and leaves, but higher in spikelets, including paleas, lemmas, pistils, and anthers (Fig. 6A). A detailed analysis of anthers indicated that expression of EDT1 increased gradually during early anther development, reached a peak level at stage 9, and then declined toward maturation (Fig. 6A). The developmental expression pattern of EDT1 in anthers, together with the pollen sterility in the edt1 mutant, suggests the importance of its role in pollen development. Next, we investigated promoter activity of EDT1 by producing pEDT1-GUS transgenic plants and analyzing GUS activity in spikelets. GUS staining showed that the signal began to appear in anthers at stage 8, became stronger at stage 9, then weakened at stage 10 and was faint at stage 11 (Fig. 6B). Cross-section analysis of GUS-stained locules in stage10 and in situ RNA hybridization with wild-type floral sections revealed that EDT1 is highly expressed in the tapetum cell layer (Fig. 6, C–F). In addition, there was GUS staining in paleas and lemmas (Fig. 6B). This result is consistent with the RT-qPCR analysis.

Figure 6.

Expression pattern of EDT1 and protein subcellular localization. A, RT-qPCR analysis of EDT1 expression in vegetative organs (culm and leaf) and reproductive organs (glume, lemma, palea, pistil, and anthers) at different times from stage 5 (S5) to stage 11 (S11), with ubiquitin (UBQ) used as an internal control. B, GUS staining of anthers at different developmental stages in the pEDT1-GUS transgenic line. Bar = 2 mm. C, GUS staining in a cross section of a stage 10 anther locule from pEDT1-GUS transgenic plants (as indicated by the black horizontal bar in B). BS7, before stage 7. Bar = 50 μm. D to F, In situ hybridization for EDT1 mRNA in wild-type anthers at stages 8 (D) and 10 (E) with antisense or sense (F) dig-labeled probes of EDT1. Msp, microspore; T, tapetum. Bars = 50 μm. G, Subcellular localization of the EDT1-GFP fusion protein in rice protoplasts (top) and N. benthamiana epidermal cells (bottom). GFP signals of EDT1-GFP fusion proteins were localized in the nucleus and the cytoplasm of rice protoplasts and tobacco epidermal cells. Confocal microscopy was used to observe the fluorescence signals of GFP (green fluorescence), and NLS-RFP (red fluorescence) was used as a nuclear localized control. Yellow fluorescence indicates merged images. Bars = 20 μm (top) and 10 μm (bottom).

To study the subcellular localization of EDT1, we fused the full-length EDT1 coding region to GFP under control of the 35S promoter and transiently expressed the construct in mesophyll protoplasts of rice leaves. We found the fusion protein in both the nucleoplasm and the cytoplasm (Fig. 6G). In order to verify this result, we transfected Nicotiana benthamiana epidermal cells with the fusion gene, and the GFP fluorescence again was observed in the nucleoplasm as well as in the cytoplasm (Fig. 6G). Unlike EDT1, ACLB-1-GFP fusion proteins only localized in the cytoplasm in rice protoplasts (Supplemental Fig. S8).

ACL Is a Heteromultimeric Enzyme in Rice

It was previously reported that ACL in Arabidopsis is a heterologous enzyme consisting of two essential subunits, ACLA and ACLB. The entire ACL protein is a heteromeric complex, which corresponds to a heterodimer with an A₄B₄ configuration (Fatland et al., 2002). We performed a yeast two-hybrid (Y2H) assay to determine whether ACLA and ACLB interacted with each other in rice. Yeast strains coexpressing EDT1 and ACLB-1 grew normally on stringent selection media, confirming the interaction between EDT1 and ACLB-1 (Fig. 7A). In addition, ACLB-1 could interact with itself (Fig. 7A). The interaction in vitro between these two proteins was further confirmed using pull-down assays. Immunoblotting showed that the glutathione S-transferase (GST)-EDT1 and maltose-binding protein (MBP)-ACLB-1 protein aggregations were clearly observed in the sample group, while they were both absent from the control group (Fig. 7B). Additionally, the 1% inputs of the two groups were detected by the anti-MBP antibodies (Fig. 7B). Using the same method, we also detected self-interaction of EDT1 in vitro (Fig. 7C). For additional verification, we conducted bimolecular fluorescent complementation (BiFC) assays in tobacco epidermal cells. Yellow fluorescent protein (YFP) signals were detected in the cytoplasm of cells expressing either YFP^N-ACLB-1/YFP^C-EDT1 or YFP^N-EDT1/YFP^C-ACLB-1, confirming interaction of the two proteins in vivo (Fig. 7D). Also, we detected that ACLB-1 interacts with itself, and the YFP signals (YFP^N-ACLB-1/YFP^C-ACLB-1) were found to exist in the cytoplasm (Fig. 7D). Interestingly, EDT1 could also bind to itself, and YFP signals (YFP^N-EDT1/YFP^C-EDT1) were found in the cytoplasm and nucleus, which is consistent with previous subcellular localization (Fig. 7D). Based on the BiFC observation, we speculate that ACL is a cytoplasm-functioning enzyme, although EDT1 is localized in the cytoplasm and nucleus. Taken together, these molecular data provide reliable evidence of the physical interaction between EDT1 and ACLB-1.

Figure 7.

EDT1 interacts with ACLB-1. A, Y2H assays examining interaction between EDT1 and ACLB-1. DDO, double drop-out medium, control medium (synthetic dropout medium [SD]/-Trp-Leu); QDO, quadruple drop-out medium, selective medium (SD/-Trp-Leu-His -Ade); Prey, activation domain, representing the pGADT7 vector; Bait, binding domain, representing the pGBKT7 vector. B, In vitro pull-down assay of recombinant MBP-ACLB-1 using bead-coupled GST-EDT1. C, In vitro pull-down assay of recombinant protein GST-EDT1 using bead-coupled MBP-EDT1. D, BiFC assay showing self-interaction of EDT1 and ACLB-1, respectively, and interaction between EDT1 and ACLB-1 in the cytoplasm of N. benthamiana leaf cells. YFP, yellow fluorescent protein. YFP^N, p2YN vector; YFP^N-EDT1, p2YN-EDT1 vector; YFP^N-ACLB-1, p2YN- ACLB-1 vector; YFP^C, p2YC vector; YFP^C-EDT1, p2YC-EDT1 vector; YFP^C-ACLB-1, p2YC- ACLB-1 vector. Bars = 20 μm.

Subsequently, we induced and purified EDT1 and ACLB-1 proteins with MBP tags in vitro and then measured the ACL activity using spectrophotometric assays, which detected the content of ACL-catalyzed oxaloacetate (OAA) by coupling to the oxidation of NADH catalyzed by malate dehydrogenase. As a result, using EDT1+ACLB1 as the enzyme at a concentration of 250 pmol achieved the maximum activity, which then remained stable (Fig. 8A). It was found that only when both EDT1 and ACLB-1 were present was the content of NADH obviously decreased, indicating that functioning of ACL requires coexistence of both ACLA and ACLB (Fig. 8B). Next, the contents of citric acid and OAA, which are substrate and product, respectively, of ACL were analyzed by ultra-performance liquid chromatography. In edt1 anthers, the level of citric acid was obviously high at stages 8–10 (Fig. 8C), but the content of OAA decreased significantly at stage 9 (Fig. 8D). Collectively, our in vitro enzymatic activity assays showed that EDT1 does function as an ACL and catalyzes production of acetyl-CoA and OAA from citrate.

Figure 8.

In vitro ACL activity of EDT1 and determination of citric acid and OAA. A and B, In vitro enzymatic activity of the recombinant EDT1 protein determined using spectrophotometric assays. The reaction time was 10 min and relative absorbance at 340 nm. The first four columns in A represent EDT1+ ACLB-1, the fifth column represents EDT1 only, and the last three columns represent the positive control. C, Citric acid content in wild-type and edt1 anthers at stages 8–10 (S8–S10). D, OAA content in wild-type and edt1 anthers at stage 9. Error bars indicate the mean ± sd (n = 3). *P ≤ 0.05 and **P ≤ 0.01; Student’s t test.

Metabolic Changes in edt1 Mutant Anthers

Based on the observed defects of the mitochondria, Ubisch bodies, and pollen walls in edt1 (Fig. 3), and on the fact that EDT1 encodes an ACL, a key enzyme in the citric acid transport system transporting acetyl-CoA from the mitochondria to the cytoplasm, we hypothesized that EDT1 might be involved in metabolic disorders in anthers. Then, we analyzed the presence of superoxide anion in wild-type and edt1 anthers using nitroblue tetrazolium (NBT). In both wild-type and edt1 anthers, the level of superoxide anion was obviously high at stages 8 and 9, and then decreased in the wild-type anthers but remained high in edt1 through stages 10 and 11 (Fig. 9A). Measurements of superoxide anion levels in both wild-type and edt1 anthers showed results consistent with those from NBT-staining (Fig. 9B). Finally, transverse section analysis showed that the superoxide anion staining signals were localized within the tapetum cells (Supplemental Fig. S9). Staining of hydrogen peroxide with 3,3′-diaminobenzidine (DAB) showed no obvious differences in the level of hydrogen peroxide between the wild type and edt1 (Supplemental Fig. S10). Considering the mitochondrial defects observed in edt1, we then used ultra-performance liquid chromatography to detect levels of ATP, ADP, and AMP in wild-type and edt1 anthers at different developmental stages. The result showed that the amount of ATP and energy charge in the edt1 mutant was significantly lower than that in the wild type at stages 8 and 9 (Fig. 9, C and D). This suggests that insufficient production of energy in the edt1 mutant anthers eventually led to anther abortion. Since the product of ACL is acetyl-CoA, which is a precursor material for the synthesis of fatty acids, ketones, and other energy substances, we subsequently extracted the fatty acids by rapid chloroform extraction from wild-type and edt1 anthers and detected contents of various fatty acids using gas chromatography-mass spectrometry. The results showed that almost all of the fatty acids in edt1 were decreased. Among them, the contents of C16:0 (palmitic acid), C18:0 (stearic acid), C18:3N3 (alpha-linolenic acid), and C18:2N6C (linoleic acid) in edt1 anthers were significantly lower than in the wild type (Fig. 9E). Thus, we conclude that the cytoplasmic enzyme ACL has an important role in the synthesis of downstream fatty acids and the formation of pollen coat.

Figure 9.

Detection of metabolic changes and expression analysis of genes related to pollen development in edt1. A and B, Analyses of superoxide anion levels by NBT staining (A) and WST-1 (Na, 2-[4-iodophenyl]-3-[4-nitrophenyl]-5-[2, 4-disulfophenyl]-2H-tetrazolium; B) production in wild-type and edt1 anthers from early stage 8 to stage 11. Bars = 0.5 mm. C and D, Comparisons of the amount of energy charge (C) and ATP (D) between wild-type and edt1 anthers from stages 8 to 11. Energy charge = ([ATP] + 0.5[ADP])/([ATP] + [ADP] + [AMP]). E, Comparison of the fatty acid content between wild-type and edt1 anthers at the heading stage. F, Alteration in the expression of key regulatory genes involved in pollen development in edt1 at stage 9 by RT-qPCR, with UBQ as a control. Data are presented as the mean of three biological replicates ± sd. Error bars in B to F indicate the SD (n = 3). Student’s t test was used for statistical analysis (*P ≤ 0.05; **P ≤ 0.01). G, A proposed model for the role of EDT1 in tapetum PCD. EDT1 encodes ACLA, which can catalyze citric acid (transported from mitochondria) to form OAA and cytosolic acetyl-CoA. OAA is replenished into the mitochondria as part of anaplerotic reactions for TCA cycle. In the cytosol, acetyl-CoA is a precursor substance for fatty acids, which will contribute to pollen formation. In the edt1 mutant, the defective ACL caused excessive accumulation of citric acid and impaired OAA replenishment, further leading to mitochondrial dysfunction. Abnormal degradation of mitochondria causes oxygen stress and insufficient energy, and the tapetum eventually degrades prematurely. Nutrients in the anomalously degraded tapetum (like cytoplasmic liposomes and sporopollenin synthesis precursors) cannot be used normally for the formation of pollen grains, and coupled with the effects on the synthesis of cytosolic acetyl-CoA, it results in defects in pollen formation. Ba, bacula; Ep, epidermis; En, endothecium; In, intine; Lo, locules; M, mitochondria; ML, middle layer; Ne, nexine I; PM, plasmalemma; T, tapetum; TCA, TCA cycle; Te, tectum.

In order to further explore the role of EDT1 in rice anther development, we used RT-qPCR in wild-type and edt1 anthers to examine expression of the genes reported to participate in rice anther development. The expression of OsTDR and EAT1 (expressed in stages 8–9) in the mutant at stage 9 was downregulated, but there was no obvious change in MSP1 (MULTIPLE SPOROCYTE 1) and bHLH142 (bHLH transcription factor; both MSP1 and bHLH142 are expressed before stage 8) between edt1 and the wild type (Fig. 9F). In the mutant, CYP704B2 (Cytochrome P450 Family Member) and DPW (DEFECTIVE POLLEN WALL), related to fatty acid synthesis in the pollen cuticle, exhibited a significant reduction in expression (Fig. 9F). Besides, the lipid transporter gene C6 (lipid transfer protein gene) was also downregulated in edt1, indicating that there is indeed a defect in the pollen wall (Fig. 9F). The expression level of protease genes CP1 (CYS PROTEASE 1) and AP25 (ASPARTIC PROTEASE), which regulate the PCD process in rice tapetum, also declined (Fig. 9F). Overall, the dysfunction of EDT1 resulted in excessive accumulation of ROS in the anthers and reduced energy charge, which may lead to the earlier occurrence of PCD. Furthermore, there were decreases in fatty acid content and developmental disorder of edt1 anthers (Fig. 9G), indicating that EDT1 plays an extremely important role in anther development.

DISCUSSION

Mutation of EDT1 Leads to Abnormal Development of Tapetum and Pollen in Rice

ACL is a cytosolic enzyme that catalyzes production of acetyl-CoA from citrate: citrate + CoA + ATP → OAA + acetyl-CoA + ADP + Pi (Fatland et al., 2002; Xing et al., 2014). Acetyl-CoA is an important component of the endogenous biosynthesis of fatty acids and cholesterol. Acetyl-CoA is also involved in acetylation and isoprenoid-based protein modification (Chypre et al., 2012). However, there are very few studies on the mechanism of action of ACL in plants. We report here the characterization of a tapetum PCD-involved protein, EDT1, predicted to be ACLA. The edt1 mutant exhibited precocious tapetum degeneration, as evidenced by the loose tapetum structure, degraded mitochondria, and significantly increased DNA fragmentation in tapetum cells (Figs. 3, G and H, and 4). We hypothesized that the accumulation of citric acid in the mitochondria might lead to a drop in pH and H⁺ stress, and finally cause mitochondrial degradation (Simpson, 1967; Wright et al., 1982). In addition, we detected a significant decrease in OAA content in edt1 at stage 9 (Fig. 8D). In previous reports, the amount of OAA directly affected the rate of the tricarboxylic acid cycle (Kappelmann et al., 2016; Latorre-Muro et al., 2018).

ROS is the signal that promotes PCD in animals and plants (Hao et al., 2003; Wiseman, 2006; Doyle et al., 2010). In plants, ROS mainly includes superoxide anion, hydrogen peroxide, and hydroxyl radicals and plays central roles in plant cell death (Moeder et al., 2002; Overmyer et al., 2003; Bouchez et al., 2007). When the level of ROS increases in anthers, the tapetum degrades prematurely. Conversely, the tapetum delays degradation in response to reduction of ROS (Xie et al., 2014). However, it is not clear how the citrate lyase affects rice tapetum PCD and homeostasis of ROS. In this study, we detected that the contents of superoxide anion were significantly increased in the mutant compared to the wild type (Fig. 9, A and B). The normal growth of cells requires energy consumption; however, the edt1 mutant has significantly reduced energy production during stages 8–9 (Fig. 9, C and D). Thus, we conclude that the premature degradation of the tapetum in the edt1 mutant is a result of both the increase in superoxide anion and the decrease in energy charge associated with mitochondrial defects. However, how plants regulate ROS accumulation to a desirable level at different developmental stages and in different cell types or organs needs further exploration.

In this study, RT-qPCR experiments were performed to verify that the dysfunction of EDT1 can lead to abnormal anther development. The expression of OsTDR and EAT1 was obviously reduced in the mutant (Fig. 9F), which seems to conflict with a previous report (Li et al., 2006). Take OsTDR, for example, which is supposed to be upregulated in edt1 as the loss of function of OsTDR leads to delayed tapetal PCD. It is possible that downregulation of OsTDR in edt1 is a premature PCD-associated universal effect. Alternatively, OsTDR, as a transcription factor, may function in a different pathway and with a different mechanism. We exclude the possibility that early triggered PCD in edt1 is realized through OsTDR. A solid link between EDT1 and OsTDR or other transcription factors needs more study.

The tapetum, as the innermost layer of the anther wall, is in direct contact with microspores and plays a crucial role in pollen development (Phan et al., 2011; Niu et al., 2013). Degradation of the tapetum at a programmed specific time point provides plenty of nutrients, such as flavonoids, alkanes, elaiosomes, cytoplasmic liposomes, sporopollenin synthesis precursors, and signal substances for the development of pollen (Lallemand et al., 2013; Liu and Fan, 2013; Ischebeck, 2016). In addition, the Ubisch bodies are uniformly distributed on the inner surface of the tapetum, which are the media that transport tapetum-derived lipidic sporopollenin precursors to the microspores (Wang et al., 2003). The chemical composition of the pollen outer wall is mainly anticorrosion sporopollenin substances, which are formed by the polymerization of some long-chain fatty acids, oxidizing aromatic rings, and phenylpropionic acid substances (Ischebeck, 2016). In Arabidopsis, two flavonoid transporter mutants, tt12 and tt19, have reduced the amount of flavonoids in the cytosol and their pollens are more sensitive to UV-B irradiation than wild-type pollens on germination (Hsieh and Huang, 2007). In our study, there was almost no normal Ubisch body in the edt1 tapetum, and the edt1 microspore walls were also defective and eventually degraded (Fig. 3), likely as a result of premature degradation of the tapetum. On the other hand, the product of ACL, cytoplasmic acetyl-CoA, is the substrate for synthesis of the downstream fatty acids and other compounds, such as alkanes, cuticle, suberin, flavonoid derivatives, malonate, and isoprenoids (Gengenbach et al., 2001; Sasaki and Nagano, 2004; Hayashi and Satoh, 2006; Li et al., 2019). In this study, we found that the level of lipids in mature anthers of the edt1 mutant was significantly lower than that in the wild type, especially the content of C16:0 (palmitic acid), C18:0 (stearic acid), C18:3N3 (alpha-linolenic acid), and C18:2N6C (linoleic acid; Fig. 9E). According to recent reports, lack of the first three fatty acids in the pollen coat can cause rapid dehydration of pollen grains (Xue et al., 2018). Therefore, lower content of fatty acids in edt1 may be another important cause of the microsporogenesis defect.

EDT1 Interaction with ACLB-1 to Form a Heteromultimeric Protein in Rice

In vertebrates, ACL is a homotetramer of four identical subunits that was named ACLY, and it is highly regulated at the transcriptional and posttranslational level, including by phosphorylation (Sato et al., 2000; Chypre et al., 2012). In this study, we verified, by a series of assays, that EDT1 could interact with ACLB-1, and we also detected that EDT1 and ACLB-1 can separately interact with themselves (Fig. 7). In the enzyme activity analysis, ACL showed enzyme activity only when EDT1 and ACLB-1 were added to the reaction (Fig. 8). These results together suggest that ACL in rice functions in the form of A_nB_n as in Arabidopsis (Fatland et al., 2005), although they do not have the closest genetic relationship and the number of genes encoding ACLA and ACLB are also different between the two species (Supplemental Fig. S7). The differences in ACL structure between animals and plants may be due to the fusion of ACLA and ACLB subunits in the early stages of evolution to adapt to different lifestyles and environments (Griffiths et al., 2012).

It is interesting that the two genes EDT1 and Os12g0566300 both code for ACLA in rice. The edt1 mutant exhibits obvious defects in reproductive growth (anther development in particular) compared with the wild type (Fig. 1). RT-qPCR analysis indicated that EDT1 was specifically expressed in spikelets, especially in anthers (Fig. 6A). However, according to the Web site expression data, Os12g0566300 is a constitutively expressed gene (http://ricexpro.dna.affrc.go.jp/). It is possible that the two genes may play their role in different tissues independently. In rice, ACLB is only encoded by one gene, ACLB-1, and biallelic knockout plants could not be obtained in our transformation with the CRISPR-Cas9 system. In addition, EDT1 protein is localized in the nucleus and cytoplasm (Fig. 6C), whereas ACL is a cytoplasm-localized enzyme and ACLB-1 interacts with itself also in the cytoplasm (Fig. 7C). In consideration of ACL enzyme activity experiments in vitro and changes in the content of the corresponding substrate and product (Fig. 8), we speculate that the cytoplasmic localization of EDT1 is important for formation of the heteromultimeric enzyme with ACL and thus may be responsible for the phenotype. On the other hand, EDT1 is also critical for the lipid metabolism process using acetyl-CoA as a synthetic substrate. The different localization of the two proteins suggests that EDT1 not only interacts with ACLB-1 to form ACL to catalyze the ATP-dependent conversion of citrate and coenzyme A to acetyl-CoA and oxaloacetic acid, but also may participate in other pathways and have other unknown functions that need further exploration. In conclusion, we have demonstrated that loss of function of EDT1 is involved in the process of early tapetum degradation and causes metabolic disorders in rice anthers through the action of mitochondria. These results reveal the importance and irreplaceability of cytosolic acetyl-CoA. In a previous report, significantly increased citrate contents under mild drought stress or abscisic acid treatment in fruits can reduce ACL gene expression (Hu et al., 2015). Therefore, ACLs may also participate in other unknown regulatory networks. In animals, ACL is phosphorylated by multifarious kinases, which can have effects on the kinetic properties and stability of the enzyme (Pierce et al., 1981). However, the significance of the potential phosphorylation events on ACL activity or stability in plants is unknown, and this study opens the window for the further study of rice ACL function.

In addition, according to a previous study, overexpression of ACL in Arabidopsis resulted in a 30% increase in wax on stems, while overexpression of a chimeric homomeric ACL in dandelion roots resulted in a 4-fold increase in rubber content (Xing et al., 2014). This line of research may eventually be used to increase rubber production through genetic modification. Moreover, some molecules derived from acetyl-CoA (like polyester polyhydroxybutyrate) have a high industrial value in the pharmaceutical and chemical industries (Mazur et al., 2009; Xing et al., 2014), and controlling carbon flux through acetyl-CoA is an important topic for many biotechnology applications (Balcke et al., 2017). This study shows that there is a close relationship between ROS balance, fatty acid metabolism, and tapetum PCD, which lays a theoretical foundation for further research on the complex tapetum development process. Considering that rice is a major crop, this study may provide a molecular basis for controlling pollen fertility and fecundity, so that it can be applied to production practices to increase crop yield.

MATERIALS AND METHODS

Plant Material

The rice (Oryza sativa) edt1 mutant was isolated from a tissue culture-derived population of the indica rice variety IR64. A mapping population was created by crossing edt1 with the wide-compatibility japonica variety 02428. In the F2 and F₃ progenies, male sterile plants were selected for linkage analysis and fine mapping. For fine mapping of the edt1 locus, bulked-segregant analysis was used and molecular markers were designed by comparison of the local genomic sequence differences between 93-11 (O. sativa indica) and Nipponbare (O. sativa japonica) available at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/). All plants were grown in the paddy field during the natural rice growing season at Nanjing Agricultural University, Nanjing, China. Primers used in the mapping are listed in Supplemental Table S1.

Anther Anatomy Analysis and Development Staging

Fresh rice spikelets and anthers at various developmental stages treated by iodine-potassium iodide (I₂-KI), NBT, or DAB staining were photographed with a Nikon AZ100 stereoscope. NBT is a widely used indicator that can be oxidized by superoxide anion and form a dark-blue formazan precipitation (Liszkay et al., 2004). Since defects in the tapetum often cause a delay in microsporogenesis, anther development stages cannot be precisely determined based on either microspore developmental stages or tapetum status (Sanders et al., 1999). We then determined anther stages based on spikelet length: stage 5, 2.5–3 mm; stage 6, 3.2–3.8 mm; stage 7, 4–4.8 mm; stage 8, 5–5.8 mm; stage 9, 6–7.5 mm; stage 10, 7.8–8.5 mm; stage 11, 8.8–10 mm; stage 12, 8.8–10 mm, with light-green lemma; stage 13, 8.8–10 mm, with green lemma before heading stage; stage14, mature pollen at heading stage. For pollen analysis and semithin cross sectioning, spikelets of different stages were fixed in formaldehyde-acetic acid alcohol solution containing a 5:5:63:27 [volume] mixture of 37% (v/v) formalin, acetic acid, ethanol, and water. For TEM analysis, anthers of different stages were fixed in 2.5% (v/v) glutaraldehyde for 24 h and postfixed by 1% (w/v) OsO4 in phosphate-buffered saline, pH 7.2. After dehydration through an ethanol series, samples were placed in a Spurr’s resin and then ultrathin sectioned. Furthermore, sections were double stained with 2% (w/v) uranyl acetate and 2.6% (w/v) lead citrate aqueous solution, then examined with a JEOL 100 CX electron microscope (JEOL).

TUNEL Assays

To examine tapetum degradation in edt1, TUNEL assays were performed using the Dead End Fluorometric TUNEL Kit (Promega). Preparation of the anther paraffin sections and the experimental procedure were conducted according to the supplier's instructions (Promega). The green fluorescence of fluorescein (TUNEL signal) and red fluorescence of propidium iodide were analyzed at 488 nm/510 nm and 530 nm/640 nm on the excitation/emission spectrum under a confocal laser scanning microscope (Zeiss; Li et al., 2006).

Comet Assay for DNA Damage

The comet assay was performed using the comet assay Kit (Trevigen) with minor modifications. Anthers at different developmental stages were minced with a scalpel in a petri dish containing 1 mL of ice-cold 1× phosphate-buffered saline and 20 mm EDTA. A 100-mm nylon mesh filter (Millipore) was used to filter the mixture, and then 50 μL of the nuclei mixed with 300 mL of 1% (w/v) low-melting agarose (preincubated at 37°C) was taken and pipetted onto slides. After a series of incubations, lysis, and washes under 4°C conditions, slides were electrophoresed at 25 V in 1× TBE for 30 min and then immersed in 70% (v/v) ethanol for 5 min. The slides were stained with SYBR green (1:10,000 dilution; Bio-Rad) and then examined with a Zeiss ImagerA2 microscope. The CometScore software (http://www.autocomet.com) was used to evaluate the percentage of DNA in each comet tail. The degree of DNA damage of different samples was calculated according to the method described by Wang and Liu (2006).

Genetic Complementation

Functional complementation experiment was performed to determine whether the candidate gene Os11g0696200 was responsible for the pollen sterility in edt1. A pCUbi1390 binary plasmid containing a 7.6-kb wild-type genomic DNA fragment of Os11g0696200, harboring an ∼3-kb promoter region, an ∼3.2-kb genic region, and an ∼1.5-kb downstream region, was introduced into the seed-derived edt1 calli using the Agrobacterium-mediated transformation method (Yukoh et al., 1994). Calli were induced from seeds of heterozygous plants (edt1/IR64 or edt1/02428), and transgenic calli or plants in the edt1 background were isolated by genotyping. The primers used for PCR are listed in Supplemental Table S1.

CRISPR-Cas9 and RNAi Vector Construction

For the CRISPR-Cas9 vector, a 20-bp target site specific for EDT1 was synthesized, fused with the Aar I linearized intermediate vector SK-Grna, and then introduced into the CRISPR-Cas9 binary vector pCAMBIA1305 (Sun et al., 2017) to generate a knock-out construct. For the RNAi vector, a 207-bp fragment from the 5′UTR of EDT1 was amplified with specific primers and cloned into the LH-FAD2-1390RNAi vector to generate a knock-down construct. Both the knock-out and knock-down constructs were transformed into the variety Nipponbare (O. sativa japonica) via an Agrobacterium tumefaciens-mediated transformation system (Yukoh et al., 1994). Transgenic plants were regenerated from transformed calli by selection on hygromycin-containing medium. Primers used in the plasmid construction are listed in Supplemental Table S1.

Phylogenetic Analysis

A phylogenetic tree was constructed to understand the phylogenetic relationship between the ACL proteins. The EDT1 full-length amino acid sequence was used to search for homologous sequences in the public database National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/). Target sequences were aligned using ClustalX and the phylogenetic analysis was performed with MEGA 5.0 software using the neighbor-joining method with pairwise deletion and 1,000 bootstrap replicates.

RNA Extraction and Quantitative Real-Time PCR Analysis

Total RNA was isolated from rice tissues using the RNA prep pure plant kit, and treated with DNase (TIANGEN). The first-strand complementary DNA (cDNA) was synthesized using Oligo (dT) 18 primer using 1 µg of RNA, and PrimeScript Reverse Transcriptase (TaKaRa) for reverse transcription. The rice ubiquitin (UBQ) gene was used as a normalizer control. Quantitative real-time PCR analysis was performed using the SYBR Green Mix (Bio-Rad) on an ABI 7500 real-time PCR system following the manufacturer’s instructions with three biological replicates. Primers used for RT-qPCR are listed in Supplemental Table S1.

Subcellular Localization of EDT1

The EDT1 cDNA was fused with GFP and inserted in the pAN580-GFP vector or pCAMBIA1305.1-GFP between the cauliflower mosaic virus 35S promoter and the nopaline synthase terminator (primers are listed in Supplemental Table S1). The resulting 35S-EDT1-GFP plasmids were transiently expressed in rice protoplasts or N. benthamiana epidermal cells (Zhao et al., 2013).

GUS Histochemistry Staining

A 4422-bp fragment upstream of the EDT1 transcriptional start codon was fused with the GUS reporter gene and introduced into pCAMBIA1381Z digested with BamH I and Hind III by In-Fusion (Takara Bio; primers used in the plasmid construction are listed in Supplemental Table S1). The proEDT1-GUS vector was transformed into Nipponbare by the Agrobacterium-mediated method (Yukoh et al., 1994). GUS activity was measured by staining spikelets at different stages of T1 transgenic lines as described previously (Zhou et al., 2011).

Determination of Superoxide Anion Levels

To measure superoxide anion, 50 mg freshly detached anthers at various developmental stages were incubated either in 1 mL K-phosphate buffer (20 mM, pH 6.0) containing 500 mm WST-1 (Na,2-[4-iodophenyl]-3-[4-nitrophenyl]-5-[2,4-disulfophenyl]-2H-tetrazolium, Dojindo) on a shaker for 8 h at 25°C in the dark, or in 10 mm K-citrate buffer (pH 6.0) containing 0.5 mm NBT (Xie et al., 2014). WST-1 can be reduced by superoxide to a formazan dye (Peskin and Winterbourn, 2000; Tan and Berridge, 2000; Schopfer et al., 2001). As the control, blanks without plant material were used. After incubation, the samples were measured at 440 nm with a microplate reader (Yi et al., 2016).

ACL Enzyme Activity Analysis

To measure ACL activity in vitro, the coding sequences of EDT1 and ACLB-1 were introduced into the pMAL-C2x vector (New England Biolabs) to construct the fusion proteins MBP-EDT1 and MBP-ACLB-1. The induced protein was purified and added to the reaction mixture (200 mm Tris-HCl, pH 8.4, 20 mm MgCl₂, 1 mm dithiothreitol, 10 mm ATP, 10 mm citrate, 0.2 mm CoA, 6 units of malate dehydrogenase, and 0.1 mm NADH). Blanks were performed by omitting coenzyme A or ATP. The final optical density value was determined by the microplate reader (Spectra Max.M3) at 340 nm and 28°C.

Y2H Assays

Y2H assays were performed using the MATCHMAKER two-hybrid system (Clontech). Various coding fragments were cloned into pGBKT7 (bait) and pGADT7 (prey) to construct BD-EDT1, BD-ACLB-1, AD-EDT1, and AD-ACLB-1, respectively (primers used in Y2H assays are listed in Supplemental Table S1). All constructs were cotransformed into the recipient yeast strain AH109, followed by incubation on SD/-Trp-Leu double dropout plates (control medium) for 3 d at 30°C. The interactions were tested on the SD/-Trp-Leu-His-Ade quadruple dropout selective medium plates.

Protein Pull-Down (In Vitro) Assay

The coding sequence of ACLB-1 and EDT1 was introduced into the pMAL-C2x vector (New England Biolabs) to construct the fusion protein MBP-ACLB-1 and MBP-EDT1, and EDT1 was cloned into the pGEX-4T-2 vector to make the GST-EDT1 fusion protein (primers used in the assays are listed in Supplemental Table S1). The pull-down assays were performed as described (Miernyk and Thelen, 2008) and were detected with anti-GST (Bio-Rad) or anti-maltose-binding protein (anti-MBP; New England Biolabs) antibodies at 1:5,000 dilution, and the second antibody was antimouse (1:5,000 dilution; Abmart). The enhanced chemiluminescence reagent (Bio-Rad) was used to test the immunoblot.

BiFC Assay

For the BiFC assays, the full-length cDNAs of EDT1 and ACLB-1 were cloned into the p2YN and p2YC vectors to construct the YFP^N-EDT1, YFP^C-EDT1, YFP^N-ACLB-1, and YFP^C-ACLB-1 fusion proteins, respectively (primers used in the assays are listed in Supplemental Table S1). The BiFC analyses were performed in 5- to 6-week old N. benthamiana epidermal cells with the silencing suppressor P19 strain, as described previously (Zheng et al., 2015). The YFP signals were observed under a confocal laser-scanning microscope (Zeiss).

Accession Numbers

Sequence data from this article for the mRNA, cDNA, and genomic DNA can be found in the GenBank/EMBL/Gramene data libraries or Web site under accession numbers Os11g0696200 (EDT1); Os12g0566300 (OsACLA-3); Os01g0300200 (ACLB-1); AT1G10670 (AtACLA-1); AT1G60810 (AtACLA-2); AT1G09430 (AtACLA-3); GRMZM2G107082 (ZmACLA-2); GRMZM2G034083 (ZmACLA-3); Os01g0293100 (bHLH142); Os02g0120500 (OsTDR); Os11g0582500); Os04g0670500 (OsCP1); Os04g0599300 (EAT1); Os01g0917500 (MSP1); Os03g0186900 (AP25); Os03g0167600 (DPW); Os03g0168600 (CYP704B2); and Os01g0201700 (MADS3).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Channel maps of the TUNEL experiment.

• Supplemental Figure S2. Deletion of a DNA fragment in edt1 as detected by PCR.

• Supplemental Figure S3. Expression analysis of the putative genes within the deleted region by RT-qPCR.

• Supplemental Figure S4. Detection of EDT1 protein content using immunoblotting.

• Supplemental Figure S5. Suppression of EDT1 by RNA interference.

• Supplemental Figure S6. Amino acid alignment of EDT1 and its homologs.

• Supplemental Figure S7. Phylogenetic tree of the ACL proteins.

• Supplemental Figure S8. Subcellular localization of ACLB-1 protein in rice protoplasts.

• Supplemental Figure S9. Paraffin section to localize NBT staining in anthers of the wild type and edt1.

• Supplemental Figure S10. DAB staining analysis of hydrogen peroxide production of anthers.

• Supplemental Table S1. Primers used in map-based cloning, RT-qPCR, and plasmid construction.

• Supplemental Table S2. Annotated genes in the deleted region based on the bacterial artificial chromosome clones of cv 9311 (http://rice.plantbiology.msu.edu/).

• Supplemental Table S3. Fertility statistics of various transgenic lines.