Rice SIZ1, a SUMO E3 ligase, controls spikelet fertility through regulation of anther dehiscence


  • Saminathan Thangasamy,

    1. Molecular and Biological Agricultural Sciences, Taiwan International Graduate Program, National Chung-Hsing University – Academia Sinica, Taipei, Taiwan
    2. Graduate Institute of Biotechnology, National Chung-Hsing University, Taichung, Taiwan
    3. Institute of Plant and Microbial Biology, Academia Sinica, Taipei, Taiwan
    Search for more papers by this author
  • Cian-Ling Guo,

    1. Institute of Plant and Microbial Biology, Academia Sinica, Taipei, Taiwan
    Search for more papers by this author
  • Ming-Hsiang Chuang,

    1. Institute of Plant and Microbial Biology, Academia Sinica, Taipei, Taiwan
    Search for more papers by this author
  • Ming-Hsing Lai,

    1. Crop Science Division, Taiwan Agricultural Research Institute, Wufeng, Taichung, Taiwan
    Search for more papers by this author
  • Jychian Chen,

    1. Molecular and Biological Agricultural Sciences, Taiwan International Graduate Program, National Chung-Hsing University – Academia Sinica, Taipei, Taiwan
    2. Institute of Molecular Biology, Academia Sinica, Taipei, Taiwan
    3. Department of Life Sciences, National Chung-Hsing University, Taichung, Taiwan
    Search for more papers by this author
  • Guang-Yuh Jauh

    1. Molecular and Biological Agricultural Sciences, Taiwan International Graduate Program, National Chung-Hsing University – Academia Sinica, Taipei, Taiwan
    2. Institute of Plant and Microbial Biology, Academia Sinica, Taipei, Taiwan
    3. Biotechnology Center, National Chung-Hsing University, Taichung, Taiwan
    Search for more papers by this author

Author for correspondence:
Guang-Yuh Jauh
Tel: +886 2 27893173
Email: jauh@gate.sinica.edu.tw


  • Sumoylation, a post-translational modification, has important functions in both animals and plants. However, the biological function of the SUMO E3 ligase, SIZ1, in rice (Oryza sativa) is still under investigation.
  • In this study, we employed two different genetic approaches, the use of siz1 T-DNA mutant and SIZ1-RNAi transgenic plants, to characterize the function of rice SIZ1.
  • Genetic results revealed the co-segregation of single T-DNA insertional recessive mutation with the observed phenotypes in siz1. In addition to showing reduced plant height, tiller number and seed set percentage, both the siz1 mutant and SIZ1-RNAi transgenic plants showed obvious defects in anther dehiscence, but not pollen viability. The anther indehiscence in siz1 was probably a result of defects in endothecium development before anthesis. Interestingly, rice orthologs of AtIRX and ZmMADS2, which are essential for endothecium development during anther dehiscence, were significantly down-regulated in siz1. Compared with the wild-type, the sumoylation profile of high-molecular-weight proteins in mature spikelets was reduced significantly in siz1 and the SIZ1-RNAi line with notably reduced SIZ1 expression. The nuclear localization signal located in the SIZ1 C-terminus was sufficient for its nuclear targeting in bombarded onion epidermis.
  • The results suggest the functional role of SIZ1, a SUMO E3 ligase, in regulating rice anther dehiscence.


Rice (Oryza sativa) is one of the most important food crops for more than one-half of the world’s population. The grains of most crops are the major food resources for most animals, including humans, and their production depends on successful sexual reproduction. The stamen is the male reproductive organ; pollen is one of the major routes of gene flow in nature through cross-pollination and is the major concern in the spread of genetically modified plants (Ma, 2005). The microspore initiates and undergoes a series of differentiation programs in the anther to result in the male gametophyte, pollen grain, for double fertilization. A comprehensive understanding of the functions of the genes and mechanisms involved in anther and pollen development will greatly benefit our knowledge of basic plant development, agriculture and the environment.

Male sterility caused by various defects, including anther indehiscence, has adverse effects on agricultural productivity and diminishes significantly the crop yield. The study of male sterility is an excellent way to understand the molecular mechanisms essential for the complicated sexual reproduction process. The final step of anther development is anther dehiscence, the opening of the anther wall, to release mature pollen grains for pollination, fertilization and seed production. For successful anther dehiscence, the indispensable processes are the timing of dehiscence, formation of secondary wall thickening, degeneration of various anther tissues, changes in carbohydrate metabolism and movement of water out of the anther (Goldberg et al., 1993; Scott et al., 2004; Ma, 2005). The anther dehiscence process is initiated after the formation of tetrads and involves three unique tissues: endothecium, septum and stomium. The stomium tissue, constituting a single layer of specialized cells, is the last breakage site for anther dehiscence (Keijzer, 1987). In addition to anatomical observations, recent genetic studies have highlighted the molecular functions of several genes regulating anther dehiscence, such as receptor-like protein kinase (Mizuno et al., 2007) and polygalacturonase (Gorguet et al., 2009), that are essential for stomium opening and male fertility.

Plant hormones play important roles in anther dehiscence. For example, the delayed dehiscence1 mutant, defective in the enzyme 12-oxophytodienoate reductase in jasmonic acid (JA) biosynthesis, shows abnormal stomium degeneration and timing of anther dehiscence (Sanders et al., 2000). DEFECTIVE IN ANTHER DEHISCENCE1 (DAD1) encodes the chloroplastic phospholipase A1, which participates in JA biosynthesis; its corresponding mutant exhibits defects in flower opening, anther dehiscence and pollen maturation (Ishiguro et al., 2001). Mutation of another JA biosynthesis pathway gene, ALLENE OXIDE SYNTHASE (AOS), also reveals delayed anther dehiscence and, consequently, causes male sterility (Park et al., 2002). In tobacco, ethylene regulates the timing of anther dehiscence instead of affecting ovule and anther development. Tobacco plants carrying an Arabidopsis mutated etr1-1 allele show delayed degeneration and dehydration of stomium cells, which alter the synchronization for pollination (Rieu et al., 2003). Two ethylene receptors from petunia, PhERS1 and PhETR2, regulate the synchronization of anther dehiscence with flower opening (Wang & Kumar, 2007). Auxin coordinates filament elongation, pollen maturation and anther dehiscence (Cecchetti et al., 2008). The herbicide glyphosate inhibits anther dehiscence in cotton by inducing changes in auxin-mediated microtubule orientation from longitudinal to transverse and secondary wall modifications in endothecium cells (Yasuor et al., 2006). Recent studies have also suggested that other internal and external factors affect anther dehiscence, including high temperature (Matsui et al., 2001) and the availability and/or movement of water in rice (Bots et al., 2005), mineral nutrient potassium in barley (Rehman & Yun, 2006) and the accumulation of flavonoids in Arabidopsis (Thompson et al., 2009).

Several transcription factors are involved in anther dehiscence. For example, ANTHER INDEHISCENCE1 (AID1) encodes a MYB protein, and its mutation results in partial to complete spikelet sterility because of defective pollen development, septum degradation and stomium breakage in rice (Zhu et al., 2004). Arabidopsis male sterile35, a MYB26 mutant, produces viable pollen, but also male sterility, as a result of nondehiscent anthers (Steiner-Lange et al., 2003), because MYB26 regulates secondary wall thickening in the endothecium and subsequent dehiscence by modulating the expression of several IRXs and NSTs (Yang et al., 2007). NAC (NO APICAL MERISTEM, ARABIDOPSIS TRANSCRIPTION ACTIVATION FACTOR AND CUP-SHAPED COTYLEDON), secondary wall thickening promoting factor 1 (NST1) and NST2 regulate the thickening of the secondary wall that is necessary for anther dehiscence (Mitsuda et al., 2005). In maize, ZmMADS2, highly expressed in endothecium and connective tissues before dehiscence and in mature pollen after dehiscence, accumulates apoptotic bodies during anther dehiscence and is required for anther dehiscence and pollen maturation (Schreiber et al., 2004). Accumulating studies have revealed important transcriptional regulation in anther dehiscence; however, the role of post-translational modification, such as sumoylation, in this process is unclear.

The process of sumoylation conjugates Small Ubiquitin-related MOdifier (SUMO) to substrate proteins through reversible post-translational modification, and is highly regulated by the environment in yeast and animal systems (Johnson, 2004). Similar to ubiquitination, sumoylation also involves a cascade of biochemical processes with the enzymes SUMO E1 activating enzyme, E2 conjugating enzyme and E3 ligase enzyme (Hay, 2005). In metazoans and yeast, sumoylated proteins play vital roles in diverse cellular events, including cell cycle progression and mitosis, DNA repair, chromatin stability, cell division, nuclear targeting, innate immunity and transcriptional regulation (Johnson, 2004; Gill, 2005). Two members of the Protein Inhibitor of Activated STAT (PIAS) group of SUMO E3 ligases (SIZs) are in yeast (SIZ1, SIZ2), one member is in Arabidopsis (AtSIZ1) and two members are in rice (OsSIZ1 and OsSIZ2) (Miura et al., 2007a). In Arabidopsis, sumoylation by AtSIZ1 is implicated in the regulation of phosphate starvation by modifying the transcription factor PHR1 to positively regulate the expression of the downstream genes AtIPS1 and AtRNS1 (Miura et al., 2005), in low-temperature stress tolerance by altering ICE1 transcription factor (Miura et al., 2007b), and in the process of plant growth and development through hormone signaling. In addition, sumoylation participates in salicylic acid (SA)-independent basal thermotolerance (Yoo et al., 2006), SA-mediated innate immunity (Lee et al., 2007), SA-mediated floral repression by changing FLC chromatin structure (Jin et al., 2008), negative regulation of abscisic acid (ABA) signaling by sumoylating ABI5 (Miura et al., 2009) and drought responses (Catala et al., 2007). All characterized SUMO E3 ligases contain conserved domains indispensable for DNA binding, sumoylation and nuclear localization. Arabidopsis AtSIZ1 is expressed in all plant tissues, and its mutant, atsiz1-3, produces reduced growth and low tolerance to drought with significant down-regulation of many genes participating in the brassinosteroid biosynthesis/signaling and drought tolerance/JA signaling pathways, respectively (Catala et al., 2007). In addition, AtSIZ1 regulates cell growth and plant development in Arabidopsis by regulating SA accumulation (Miura et al., 2010). Recently, OsSIZ1 and OsSIZ2 have been suggested to function as SUMO E3 ligases to induce protein sumoylation under different stress conditions, including low and high temperature, NaCl and ABA (Park et al., 2010); nevertheless, their biological functions in rice are still under investigation.

Sumoylation plays diverse functions in yeast, metazoans, animals and Arabidopsis; however, the role of SUMO E3 ligase in rice is still unclear. In this study, we investigated the functional characterization of a single recessive rice SUMO E3 ligase mutant, siz1, regulating spikelet sterility, plant height and tillering phenotypes. Interestingly, similar phenomena were found in SIZ1-RNAi transgenic lines with significantly down-regulated expression of SIZ1. Comprehensive morphological investigation of siz1 and SIZ1-RNAi anthers revealed that the spikelet sterility was caused by defective anther dehiscence instead of pollen viability. Moreover, the expression of a few genes involved in secondary wall thickening in the endothecium during anther dehiscence in other species and their rice orthologs were down-regulated in the siz1 mutant. We discuss the possible mechanisms of SIZ1 in the regulation of anther dehiscence.

Materials and Methods

Plant materials and growth conditions

Seeds of rice (Oryza sativa L. cv Japonica) cv Tainung67 (wild-type, Wt), siz1 and SIZ1-RNAi were germinated, and 3-wk-old seedlings were transplanted under genetically modified organism-free field conditions at the Taiwan Agricultural Research Institute, Taichung, and under glasshouse conditions (30°C and 20°C during the day and night, respectively, with a 14 h light : 10 h dark cycle) at Academia Sinica in Taipei.

Generation of SIZ1-RNAi transgenic plants

The inverted repeat region of SIZ1 (Fig. 4a and Supporting Information Fig. S1a) was amplified by the use of primer sets (Table S1) and cloned into the pANDA RNAi-expression vector (Miki & Shimamoto, 2004) with a gateway cloning kit (Invitrogen, Carlsbad, CA, USA). Agrobacterium-mediated transformation of rice wild-type calli was performed as described by Sallaud et al. (2003). Plantlets regenerated from transformed calli were selected for hygromycin resistance, and regenerated transgenic plants were moved to an isolated field for phenotypic observation. The transgenic seedlings were confirmed by genotyping and the expression of SIZ1. The locus number for SIZ1 from the TIGR version 6.1 rice genome annotation database (http://rice.plantbiology.msu.edu/) is Os05g03430.

DNA extraction, genotyping and DNA blotting

Genomic DNA was isolated from rice seedlings of the wild-type, T2 segregants and siz1 by the cetyl trimethylammonium bromide (CTAB) method (Murray & Thompson, 1980). Genotyping of T2 segregants involved the primer sets in Table S1. For Southern blot analysis, genomic DNA (20 μg) from each sample was digested with different restriction enzymes, separated on 0.8% agarose gel and transferred onto GeneScreen Plus (PerkinElmer, Zaventem, Belgium) as described in Ho et al. (2000). The β-glucuronidase (GUS) region from T-DNA (pTAG8; the vector used to generate the T-DNA mutant population) and Tos17 sequences from wild-type genomic DNA were PCR amplified with primers (Table S1) and labeled with α32P-dCTP using the Amersham Rediprime II random prime labeling system (GE Healthcare, Amersham, UK). Prehybridization, hybridization and washing steps followed the manufacturer’s protocol (Biochain Institute, Hayward, CA, USA). Images of membranes were developed after exposure to BioMax MS film (Carestream Health, Inc. Rochester, NY, USA) at −70°C for 6 d.

RNA extraction, RNA blotting, semi-quantitative and real-time reverse transcription-polymerase chain reaction (RT-PCR)

Total RNA was isolated from different tissues and stages of wild-type, siz1 and SIZ1-RNAi plants with the use of TRIzol reagent, according to the manufacturer’s protocol (Invitrogen). Ten micrograms of total RNA were separated on 1.2% agarose gel with formaldehyde and transferred into GeneScreen Plus (PerkinElmer) as described previously (Ho et al., 2000). The 3′-cDNA of SIZ1 was PCR-amplified with the primer sets (Table S1), and probe preparation, prehybridization, hybridization and washing steps were performed as described previously. For semi-quantitative RT-PCR, total RNA was treated with RNase-free rDNase I (USB Corp., Cleveland, OH, USA). In total, 3 μg of RNA from spikelets was employed to prepare cDNA with the use of ThermoScript reverse transcriptase (Invitrogen) at 65°C for 1.5 h, followed by RNaseH treatment and RT-PCR with the corresponding primer sets (Table S1). SIZ1–GUS fusion transcripts were obtained from total RNA of siz1 leaves using RT-PCR with the primers in Table S1.

Real-time RT-PCR was conducted with a 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA, USA) and SYBR Green (Applied Biosystems) according to the manufacturer’s instructions. Each real-time experiment with different primer pairs and/or different templates was repeated three times, and the expression levels of selected genes were normalized to the constitutive expression level of Elongation Factor 1 (EF1). Finally, the averaged relative gene expression levels were calculated by the 2−ΔΔCT (Livak & Schmittgen, 2001) method.

Protein extraction and analysis of protein sumoylation

Total proteins from wild-type, siz1 and several RNAi line plants were isolated from maturing spikelets before anthesis as described previously (Chaikam & Karlson, 2008). For sumoylation assay, 30 μg of total proteins from each plant, separated by 4–15% gradient sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE), were used for western blot analysis. The membranes were probed with anti-SUMO1 antibody (Ab5316; Abcam, Cambridge, MA, USA; Kurepa et al., 2003), followed by incubation with horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody (Thermo Fisher Scientific Inc., Waltham, MA, USA). The signals were detected using an enhanced chemiluminescence-plus detection kit (GE Healthcare Biosciences, Piscataway, NJ, USA).

Subcellular localization of SIZ1

The open reading frame of the SIZ1 C-terminus was subcloned in front of enhanced green fluorescent protein (eGFP) with the use of primers (Table S1) containing restriction sites, and was driven by the cauliflower mosaic virus 35S promoter (CaMV35S). As a positive control for nucleus, cDNA of Ethylene Response Factor 4 (ERF4 ) was cloned in front of monomeric red fluorescent protein (mRFP; L. Chang and G. Y. Jauh, unpublished). The in-frame fusion constructs were bombarded into onion epidermal cells as described previously (Wang et al., 2008). Bombarded onion epidermal cells were incubated in the dark for 16 h at room temperature, and fluorescent signals were observed using an LSM 510 Meta confocal laser-scanning microscope (Zeiss, Thornwood, NY, USA). All images were recorded at 1024 × 1024 pixels by line two averaged scanning.

Preparation of rice anthers for pollen fertility and microscopic analyses

To observe the dehiscence of anthers, mature panicles were collected during flowering and were examined under a dissection microscope (Zeiss Stemi 2000-C Stero Microscope, Carl Zeiss MicroImaging, Thornwood, NY, USA). For pollen fertility analysis, anthers from spikelets just before and 2 h after anthesis were collected. The pollen grains were then analyzed for fertility in iodine–potassium iodide (I2–KI) solution (Chen et al., 2007). For anther morphology analysis, fresh anthers from wild-type, siz1 and SIZ1-RNAi plants were collected 2 h after anthesis and loaded on a stub. The samples were frozen by liquid nitrogen slush, and then transferred to a sample preparation chamber at −160°C. After 5 min, the temperature was raised to −130°C, and the samples were fractured. The samples were then etched for 10 min at −85°C. After coating at −130°C, the samples were transferred to a scanning electron microscopy chamber and observed at −160°C using a cryo-scanning electron microscope (FEI Quanta 200 SEM/Quorum Cryo System PP2000TR FEI, Hillsboro, OR, USA). To analyze anther dehiscence, anthers before and after anthesis were collected, fixed and sectioned as suggested in the protocol (An Ellis, 2006). After staining with toluidine blue, the cross-sections (0.9 μm) were observed by light microscopy (BX51; Olympus, Tokyo, Japan) and images were recorded by cooling CCD (DP70; Olympus) through ×4 and ×10 (UPlanFI; Olympus) lenses.

Protein alignment and phylogenetic tree analysis

BlastP was used to search all SIZ1 orthologs in the National Center for Biotechnology Information and the TIGR version 6.1 rice genome annotation databases. Sequences of plant-specific SIZ proteins were then aligned using ClustalW2 (Larkin et al., 2007). A phylogenetic tree was constructed with the neighbour-joining algorithm (Saitou & Nei, 1987), as instructed in MEGA 4.0 (Kumar et al., 2008).


Isolation, identification and molecular analysis of siz1 mutant

Successful reproductive growth is important for crop productivity. To identify the genes essential for rice reproduction and grain yield, T-DNA activation mutants obtained from the Taiwan Rice Insertional Mutants (TRIM) Database (Hsing et al., 2007) were screened with the criteria of abnormal spikelet development and sterility. One of the mutants with a reduced plant height and lower seed production rate drew our attention (Fig. 1). One-quarter of the 19 segregated T2 plants showed reduced plant height, increased tiller number (Fig. 1a–c) and significantly reduced numbers of total spikelets and filled seeds (seed set percentage) (Fig. 1d,e). In addition, the panicles of the mutant did not completely emerge out of the leaf sheath (Fig. 1b). Many empty seeds (sterile spikelets) found in panicles, showing a white color without filling (Fig. 1d), may account for the lower seed production rate of the mutant. Flanking sequence analysis from the TRIM database indicated that the 15th of the 17 exons of SIZ1 encoding the rice SUMO E3 ligase was disrupted by T-DNA insertion (Figs 2a, S2). We then investigated the T2 plants for the co-segregation ratio between the T-DNA insertion and the sterile spikelet phenotype. Among 19 plants, five were siz1 homozygous lines with a reduced seed set phenotype, and the remaining 14 lines were either wild-type or hemizygous lines with normal seed development (Fig. 2b). The T3 population (= 72) also segregated well for reduced seed set as a single recessive trait (wild-type : mutant = 53 : 19; χ= 0.2 for 3 : 1) following Mendelian rules.

Figure 1.

 Detailed phenotypes associated with rice (Oryza sativa) siz1 mutant. Comparison of wild-type (Wt) and siz1 mutant plants at flowering (a) and mature (b) stages; siz1 plants showed semi-dwarfism, but increased tiller numbers (dark gray bars, plant height; light gray bars, number of tillers) (c). Wild-type panicles were normal and droopy when compared with the erect and incompletely emerged (arrows in b) panicles of siz1. The seed set percentage in siz1 mutants was substantially reduced to c. 60% compared with c. 93% in Wt plants (d, e). The right panels in (d) are magnified images of the rectangles in the left panel. Arrows and arrowheads indicate the filled and empty spikelets, respectively, found in Wt and siz1 panicles. Data are mean ± SD (= 15) in (c) and (e). Student’s t-test was used to analyze significant differences between Wt and siz1. **, < 0.01; ***, < 0.001. Bar, 5 cm.

Figure 2.

 Null mutation feature of siz1 was caused by a single recessive T-DNA insertion. (a) Schematic representation of the genomic organization of the SIZ1 gene, position of T-DNA insertion and primer sets used for genotyping and gene expression analyses. Solid boxes represent exons, and gray boxes indicate 5′- and 3′-untranslated regions. The T-DNA vector contains promoter-less GUS (GUS) in the right border and eight copies of 35S enhancers (8×35S) in the left border. The primer sets used for genotyping and semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR) are indicated as P1, P2 and P3, and 1F, 1R, 2F, 2R and GUS-R, respectively. (b) Genotyping of T2 seedlings with primer sets indicated to reveal the wild-type (W), heterozygous (H) and homozygous (M) progenies. T-DNA F and R are the forward and reverse primers designed within the T-DNA region, respectively. (c) Semi-quantitative RT-PCR analysis of gene expression revealed that T-DNA insertion truncated the expression of SIZ1 in siz1. 8×35S residing in T-DNA had no effect on the expression of the neighboring gene Gene1 (Os05g03440). The expression levels of SIZ2 (Os03g50980) were similar in both wild-type and siz1. (d) Northern blot with the 3′-probe revealed the absence of the expected SIZ1 transcript (3.3 kb) in the siz1 mutant. Elongation factor 1-α (EF1) and ribosomal RNAs were used as loading controls in semi-quantitative RT-PCR (c) and northern blot analysis (d), respectively. (e) Southern blot analysis of wild-type and siz1 genomic DNA digested with EcoRI (RI) and XbaI revealed single T-DNA insertion in the siz1 mutant with GUS used as a probe. (f) Southern blot analysis of retrotransposon Tos17 copy numbers in wild-type (Wt) and T2 segregants (W, wild-type; H, heterozygous mutant; M, homozygous mutant) genomic DNAs digested with XbaI, followed by probing with Tos17 probe. Asterisks indicate three copies of Tos17.

Because the T-DNA used to generate the mutant population contained 8×CaMV35S enhancers near the left border (Hsing et al., 2007), it might activate the expression of SIZ1 and/or any gene located within 15 kb around the T-DNA insertion site. From the annotated rice genomic sequences, only one additional gene, Os05g03440, was found in this ± 15-kb region, but showed no significant difference in expression profile between the wild-type and siz1 (Gene1 in Fig. 2c). The expression of SIZ2, encoding another rice SUMO E3 ligase, was similar between the wild-type and siz1 (Fig. 2c). Despite slight expression levels of SIZ1 and the SIZ1–GUS fused transcript behind the T-DNA insertion position, the SIZ1–GUS transcript was not in frame, and no GUS signal was detected in all tissues examined (data not shown). RNA blotting with a SIZ1 3′-probe located after the T-DNA insertion revealed the presence of a c. 3.3-kb SIZ1 transcript in the wild-type, but not siz1 seedlings (Fig. 2d), which suggests the null mutation feature of siz1. The TRIM activation T-DNA lines carry an average of 1.73 copies of T-DNA (Hsing et al., 2007), and occasional activation of retrotransposon Tos17 causes additional mutations in the rice genome during tissue culture and transgenic line generation (Hirochika, 1997). Southern blot results demonstrated a single T-DNA insertion (Fig. 2e) and no extra Tos17 copy (Fig. 2f) in siz1 seedlings. All these results indicate that the null mutation feature of siz1 is caused by a single T-DNA insertional mutation.

Spatiotemporal expression, domain analysis and subcellular localization of SIZ1 protein

To examine the spatiotemporal expression profiles of SIZ1, semi-quantitative RT-PCR analyses were conducted with the primer sets specific to SIZ1 (Table S1). In rice, SIZ1 was universally present in all examined tissues and all developmental stages (data not shown). Similar to most of the characterized and annotated SIZ/PIAS (SAP and MIZ/Protein Inhibitor of Activated STAT) SUMO E3 ligases from all organisms, rice SIZ1 also contains SAP, PINIT, SP-RING, a SUMO binding motif (hhhSXSaaa), a C-terminal nuclear localization signal (NLS) and the plant-specific PHD domain (plant homeodomain with C4HC3-type Zn-finger) (Fig. S2a,b). PINIT (Pro-Ile-Asn-Ile-Thr) and SP-RING, containing a zinc-finger (C2HC3), are essential for SUMO E3 ligase activity, and SAP (scaffold attachment factors SAF-A/B, Acinus, PIAS, a helix–extended loop–helix) forms a helix–extended loop–helix structure probably involved in DNA binding (Aravind & Koonin, 2000). Amino acid comparison revealed that all plant SIZs possess all consensus domains (Fig. S2b). A BlastP search analysis with SIZ1 amino acid sequences used as a query was employed to generate a rooted phylogenetic tree to illustrate the relationship of rice SIZs to their plant orthologs. As shown in Fig. S2c, rice SIZ1 and SIZ2 are positioned in different clades, and AtSIZ1 is closer to rice SIZ1 than to SIZ2. Sequence similarity analysis also confirmed that rice SIZ1 is closely related to AtSIZ1 (Fig. S2d). The amino acid sequences of SIZ1 and SIZ2 showed high similarity to sorghum loci Sb09g002225 and Sb08g000380, with 65.0% and 57.1% identity, respectively (data not shown). This is probably because both rice and sorghum originate from the grass family and have high synteny for sharing similar panicle architecture.

In Arabidopsis, AtSIZ1 protein is primarily localized in the nucleus, and PHR1, a MYB transcriptional activator, is its direct target in response to phosphate deficiency (Miura et al., 2005). Therefore, the subcellular localization is important to reveal the potential functions of rice SIZs. Recently, Park et al. (2010) showed the nuclear localization of OsSIZ1 and OsSIZ2 in rice protoplasts. To verify whether NLS in the SIZ1 C-terminus is responsible for its nuclear targeting, the NLS-containing C-terminal SIZ1 was fused to the N-terminus of eGFP for particle bombardment-mediated transient assay in onion epidermis (Fig. 3a). The mRFP-tagged Ethylene Response Factor 4 (P35S::ERF4-mRFP:T35S), a known transcription factor (Ohta et al., 2000), was used as a positive nuclear localization marker. Cells expressing the SIZ1–CT–eGFP fusion construct showed co-localization of GFP signals with ERF4 specifically in the nuclei (Fig. 3b). Our subcellular localization study suggested that the NLS-containing C-terminus of SIZ1 was sufficient for its nuclear targeting.

Figure 3.

 Nucleus-localized SIZ1 in bombarded onion epidermal cells. Various cauliflower mosaic virus 35S promoter (CaMV35S)-driven constructs listed in (a) were delivered into onion epidermal cells by particle bombardment. Images obtained from the central section of bombarded cells were recorded by fluorescent confocal laser scanning microscopy (b). P35S, CaMV35S promoter; eGFP, enhanced green fluorescent protein, T35S, CaMV35S terminator; SIZ1-CT, C-terminal open reading frame of SIZ1; ERF4, Ethylene Response Factor 4; mRFP, monomeric red fluorescent protein. Bars, 50 μm.

Spikelet sterility in siz1 and SIZ1-RNAi independent lines was caused by defective anther dehiscence, but not impaired pollen fertility

To functionally confirm that the sterile spikelet found in siz1 was truly caused by T-DNA insertion, we used an RNAi approach to knock down the expression of SIZ1. A 534-bp cDNA fragment of SIZ1 around the SP-RING domain was chosen to generate SIZ1-RNAi transgenic plants (Figs 4a, S1, S2a). Semi-quantitative RT-PCR of more than 10 SIZ1-RNAi independent transgenic lines revealed substantial down-regulation of SIZ1; examples are lines 44, 19 and 52 shown in Fig. 4b. Phenotyping results indicated that all SIZ1-down-regulated RNAi lines exhibited reduced plant height (Fig. 4c,d), increased tiller number (Fig. 4c,e) and reduced seed set percentages (Fig. 4f), which were comparable with that of the siz1 mutant during the tillering and maturity phases. The severity of the observed phenotypes in these SIZ1-RNAi independent lines was associated with the magnitude of the reduction in SIZ1 transcripts. In SIZ1-RNAi line 33, the expression of SIZ1 was not notably affected, and the line showed phenotypes similar to that of the wild-type. Although the RNAi target sequence is somehow unique to SIZ1, a BLAST search of the Rice Genome Annotation Project showed that the sequence still contains 59% homology to SIZ2. However, the SIZ2 expression profile revealed slight down-regulation of SIZ2 found in some transgenic plants, such as line 44 (Fig. S3), which showed phenotypes similar to those in other SIZ1-RNAi independent lines. These results demonstrate that the defective phenotypes observed in these SIZ1-RNAi independent lines, and in siz1, were strongly associated with the degree of down-regulated SIZ1 transcript levels.

Figure 4.

 Generation, characterization and phenotyping of several independent SIZ1-RNAi transgenic plants. (a) A 534-bp cDNA fragment around the SP-RING domain of SIZ1 was used to generate the SIZ1-RNAi construct in the pANDA vector and to produce corresponding RNAi transgenic plants. RB and LB, right and left borders; NptII, kanamycin-resistant gene; Hpt, hygromycin-resistant gene; Zm Ubi P, maize ubiquitin promoter; NosT, nopaline synthase terminator. (b) The expression profiles of SIZ1 in the wild-type (Wt) and SIZ1-RNAi transgenic seedlings revealed by semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR) analysis. Bottom panel shows the mean quantitative reduction level of SIZ1 expression in several SIZ1-RNAi transgenic seedlings from three independent experiments by use of the relative transformed value of SIZ1 to Elongation Factor 1 (EF1) in Wt plants set as 1.0. (c) Morphology of Wt and RNAi transgenic plants showing plant height and tillers during the flowering stage. (d–f) Quantitative analyses show significant differences in several agronomic traits, namely plant height (d), number of tillers (e) and seed set percentage of panicle (f) in SIZ1-RNAi transgenic plants when compared with Wt plants. Data are the mean ± SD (n ∼ 9–11) in (d–f). Student’s t-test was used to compare significant changes in RNAi lines relative to Wt. **, < 0.01; ***, < 0.001; ns; not significant.

After anther dehiscence, pollen grains released from a single anther are sufficient to fertilize the ovule for seed production in a self-pollinated crop such as rice. Normally, rice anthers reach the top of the spikelet before dehiscence and then release pollen through dehiscence, and pollen grains are shed over the stigma of the pistil just before spikelet opening to ensure successful self-pollination. Anthesis and filament elongation were normal in the wild-type, siz1 and SIZ1-RNAi mutant plants (Fig. 5a,b). In the wild-type, the anthers dehisced completely and shed pollen before the spikelet began to close; however, c. 70% of siz1 and c. 60% of SIZ1-RNAi spikelets showed various degrees of defects in anther dehiscence even after the spikelets had closed, and a variation in anther indehiscence was observed from one anther to six anthers of a single spikelet in siz1 and RNAi knock-down plants (Fig. 5b). To investigate the dehiscence events, scanning electron microscopy was used to compare the morphological variation of anthers among the wild-type, siz1 and SIZ1-RNAi plants just after anthesis. Wild-type anthers showed normal dehiscence to release fertile pollen grains, whereas > 50% of siz1 and SIZ1-RNAi anthers were completely indehiscent during or even after anthesis (Fig. 5c).

Figure 5.

 Anther indehiscence, but not impaired pollen fertility, causes spikelet sterility found in siz1 and SIZ1-RNAi independent plants. (a) Appearance of dehiscent anthers in wild-type (Wt), siz1 and SIZ1-RNAi plants during flowering. Spikelets of the panicles with anthesis are shown. Except for anther dehiscence, panicle morphology was normal in all plants. (b) Magnified images of corresponding plants marked by a square in (a) showing various degrees of anther dehiscence. Arrowheads and arrows in (a,b) mark dehiscent and indehiscent anthers, respectively. (c) Scanning electron micrographs of anthers collected after anthesis show anther indehiscence in T-DNA (siz1) and SIZ1-RNAi transgenic plants. Arrowheads indicate anther dehiscence during anthesis. Pollen grains from anthers of both Wt and the siz1 mutant examined by I2–KI staining before (d) and after (e) anthesis are fertile. (f) Transverse sections of Wt and siz1 anthers after anthesis. Wt anther shows complete dehiscence and sheds all pollen grains (left panel), whereas siz1 anthers show partial dehiscence (middle panel) or complete indehiscence (right panel) in the top and central theca, respectively. Arrowheads indicate the intact endothecium tissue in the siz1 theca. Inset shows the details around the cavity marked by a square line in siz1 and the arrowhead indicates the failure of endothecium to promote dehiscence before stomium tissue opening. C, cavity; EP, epidermis; EN, endothecium; L, locule; P, pollen grain; SP, septum; St, stomium. Bars: (a) c. 0.8 cm; (b) c. 1.2 mm; (c) 500 μm; (f) 50 μm.

Our phenotypic results suggest that the spikelet sterility found in siz1 and SIZ1-RNAi plants is mainly caused by anther indehiscence. However, we could not rule out the possibility that spikelet sterility was caused by pollen sterility. Therefore, we examined pollen fertility by I2–KI staining before spikelet opening and after spikelet closing. Wild-type and siz1 pollen grains were fertile and remained inside the anthers before spikelet opening (Fig. 5d). Nevertheless, when compared with wild-type anthers, siz1 anthers retained fertile pollen grains after anthesis (Fig. 5e) because of anther indehiscence. To further investigate the tissues showing failed anther dehiscence, we investigated transverse sections of anthers collected immediately before and after anthesis from wild-type and siz1 plants. Because the opening of rice theca initiates from the stomium of the apical part and near the anther wall of the basal part of the large locule towards the central theca (Matsui et al., 1999), we examined the top portions of the opening anthers. When compared with the wild-type anther, which showed complete dehiscence at the apical part of the theca (left panel in Fig. 5f), the siz1 anther showed only partial dehiscence (middle panel in Fig. 5f) or complete indehiscence (right panel in Fig. 5f) for the same regions. In completely indehiscent siz1 anthers, the endothecium and septum tissues near the stomium were still intact and left a small cavity and tight association of the epidermal cells of two adjacent locules after anthesis (inset in right panel in Fig. 5f). These defects may be unable to support stomium tissue opening during the final step and thus cause incomplete anther dehiscence. All these results suggest that SIZ1plays a significant role in regulating the development of endothecium tissues during anther dehiscence.

Sumoylation assay and expression of genes involved in anther dehiscence

To explore the possible biological function of SIZ1/sumoylation in anther dehiscence, we examined the sumoylation patterns of protein profiles from spikelets of wild-type, siz1 and two RNAi lines: RNAi-33 and RNAi-52. For the wild-type and RNAi-33 line with slightly reduced SIZ1 expression (Fig. 4b), intense protein sumoylation profiles were observed. However, siz1 and the RNAi-52 line showing a reduced SIZ expression level revealed a significant reduction of sumoylation profiles in higher molecular weight proteins (Fig. 6a). This result shows the convincing correlation between the expression levels of SIZ1 and sumoylation profiles in the spikelets, and implies that SIZ1 probably has an important function in anther dehiscence.

Figure 6.

 Down-regulation of spikelet sumoylation profiles and expression of several orthologous genes involved in rice (Oryza sativa) anther dehiscence in siz1 plants. (a) Total proteins extracted from maturing spikelets of wild-type (Wt), siz1, RNAi-33 and RNAi-52 lines were analyzed by immunoblotting with anti-SUMO1 polyclonal antibody to show the profiles of SUMO–protein conjugates. The large subunit (LS) of Rubisco (55 kDa) and free SUMO, revealed by Coomassie brilliant blue (CBB) staining and immunoblotting, respectively, were used as loading controls. Each lane contains 30 μg of total protein. (b) Twelve rice orthologs of known genes shown to be involved in anther dehiscence in different plant species were identified, and their expression profiles in Wt and siz1 plants were examined by semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR) analysis. The locus numbers of individual genes were obtained from the TIGR version 6.1 rice genome annotation database (http://rice.plantbiology.msu.edu/). Elongation Factor 1 (EF1) was selected as the loading control. The number of cycles in RT-PCR for individual genes in the experiment differed. The relative expression ratio of individual genes in siz1 was calculated by comparison with the expression in Wt. Arrows indicate the genes moderately down-regulated and the asterisk indicates a nonspecific PCR product. (c) Real-time quantitative PCR analysis (= 3) was used to evaluate the expression of four genes marked by arrows in (b), and endogenous EF1 was used as a reference. Wt, light gray bars; siz1, dark gray bars. Error bars, ± SD.

Next, to explore the possible molecular mechanism of anther indehiscence in siz1, we examined the expression profiles of the genes potentially participating in anther dehiscence in both wild-type and siz1 plants. About two dozen genes are involved in anther dehiscence, and 12 of 23 genes (orthologs of OsAID1, AtIRX1, AtNST2, ZmMADS2, AtCOI1, AtAOS and PhERS1) were found in rice after a search of the rice genome sequences. Typical results of three biological repeats of semiquantitative RT-PCR and real-time PCR revealed the expression of Os2g08420 (ortholog of AtIRX1) and Os2g36924 (ortholog of ZmMADS2) to be down-regulated by c. 50% in siz1 (Fig. 6b,c). This partial reduction of their expression may contribute to the partial spikelet sterility in siz1. These results suggest that SIZ1-mediated sumoylation of some regulatory players may regulate the expression of the downstream genes participating in rice anther dehiscence.


Successful reproductive growth is important for crop productivity, and various defects in reproductive organ development, such as male sterility, cause adverse crop yields (Ma, 2005). The identification and functional characterization of diverse mutants and their corresponding genes with defective anther development will reveal the regulatory mechanisms of male gametophyte development. In this study, an activation-tagged rice mutant siz1 was functionally characterized for its unique role in regulating spikelet fertility. SIZ1 is a universally expressed gene and regulates both vegetative and reproductive growth and development, but has adverse effects on spikelet fertility (Figs 1, 4). Results from tight phenotypic segregation with T-DNA insertion, as well as northern and Southern blot analyses (Fig. 2), clearly revealed siz1 as a recessive mutant, with SIZ1 disrupted by a single T-DNA insertion. SIZ1 contains all the conserved domains found in most plant SIZs (Fig. S2), localizes in the nucleus (Fig. 3; Park et al., 2010) and participates in the sumoylation process (Fig. 6a). Importantly, SIZ1 plays a unique and critical role in regulating anther dehiscence during rice anthesis and fertilization processes (Fig. 5). The function of SIZ1 in anther dehiscence was supported by spikelet sterility in several SIZ1-RNAi transgenic lines: the down-regulation of SIZ1 transcripts was closely associated with moderately reduced seed set percentage in several SIZ1-RNAi lines (Fig. 4). All these results show that SIZ1 is involved in regulating rice anther dehiscence.

Sumoylation is emerging as one of the major post-transcriptional modification processes in plants by the PIAS (SIZs) group SUMO E3 ligase (Miura et al., 2007a). Arabidopsis AtSIZ1 is the best-characterized plant SUMO E3 ligase and plays critical roles under stress conditions (Miura et al., 2005, 2007b; Yoo et al., 2006), regulating plant growth (Catala et al., 2007), cell proliferation (Catala et al., 2007; Huang et al., 2009; Miura et al., 2010) and FLC-mediated flowering (Jin et al., 2008). Most eukaryotic SUMO E3 ligases share multiple domains SAP, PINIT and SP-RING for SUMO binding, together with a putative NLS for nuclear localization, but plant SUMO E3 ligases contain an extra PHD domain, a plant-specific zinc-finger homeodomain, which is probably involved in chromatin remodeling (Bienz, 2006) and phosphoinositide receptor activity (Gozani et al., 2003). A study of yeast PIAS-type ULL1/Siz1 revealed that N-terminal PINIT and SP-RING domains were required for SUMO ligase activity, whereas the SAP domain was involved in nuclear localization (Takahashi & Kikuchi, 2005). In addition, the N-terminal ULL1–GFP fusion protein was localized in the nucleus and the cytoplasm, which implies the necessary roles of ULL1 in both locations. In budding yeast, cytoplasmic proteins Ddc48, Vps72, Arc35 and Arc40, which localize in the cytoplasm, were confirmed to be sumoylated in vivo (Wohlschlegel et al., 2004). As shown in Fig. 3, the NLS residing in its C-terminus of OsSIZ1 is sufficient for its nuclear localization, although the functions of other domains need to be studied. Recently, structural studies of yeast and rice SIZs revealed the presence of four helix bundles within SAP domains exhibiting DNA binding ability (Suzuki et al., 2009). Its nuclear localization raises the possibility that it may function as a transcriptional coregulator for a transcriptional regulatory complex, as found in mammalian cells (Schmidt & Muller, 2003). Further detailed study of the individual domains of SIZ1 would generate more information about their specific functions in rice. The partial spikelet sterility found in siz1 and SIZ1-RNAi independent lines indicates that SIZ1 is important, but not definitely essential, for anther dehiscence. The partial effect may be the result of the presence of an additional functional SUMO E3 ligase, SIZ2, which can moderately complement the lost function of SIZ1 in the siz1 mutant. Although the identities of the potential target proteins in rice are lacking, our data suggest that SIZ1 may participate in anther dehiscence through post-translational modification of several key transcription factors (to be described further).

Anther dehiscence is a critical step in releasing mature pollen grains through the functions of three novel anther tissues: endothecium, septum and stomium. Anther dehiscence is initiated after the formation of tetrads, when a cavity gradually forms because of the dissociation of the stomium from the broken-down septum (Zhu et al., 2004). In the wild-type rice anther, the theca opens on the stomium from the apical part and near the anther wall of the basal part of the large locule (Matsui et al., 1999). During anther dehiscence, the degeneration of the septum forms a bilocular anther, which is followed by the enlargement of endothecial and connective cells and strengthening of the endothecium wall for correct dehiscence. Finally, with the driving force from the turgor pressure of swollen pollen grains and the inward bending of the locules, the stomium, a single layer of specialized cells defining the breakdown site for anther opening, splits to release pollen grains (Keijzer, 1987; Dawson et al., 1999; Matsui et al., 1999; Sanders et al., 1999; Zhu et al., 2004). However, in siz1, with its completely indehiscent anther, the endothecium and septum near the stomium were still intact after anthesis and the stomium was still linked to the epidermal cells of two adjacent locules; the cavity was slightly formed after the degeneration of the septum in siz1 (Fig. 5f). These defects may be unable to support the stomium tissue opening during the final step of anther dehiscence. In rice, rapid swelling of pollen grains just before dehiscence generates a significant mechanical force to rupture the septum (Matsui et al., 1999). For example, the rice aid1 mutant shows insufficient pressure, one of the reasons for anther indehiscence, because of the reduced size of pollen grains (Zhu et al., 2004). Nevertheless, this insufficient pressure from pollen grains is probably not the case in siz1, because most of the anthers from both wild-type and siz1 plants were plump and the pollen grains were fertile (Fig. 5). A detailed series of anatomical investigations of anther development in both wild-type and siz1 plants will provide realistic explanations for anther indehiscence in siz1.

Many genes have been found to be involved in anther development (Goldberg et al., 1993), but the regulation and mechanism of anther dehiscence in rice are unclear. Recent genetic studies from plants highlighted > 20 genes participating in anther dehiscence, and 12 of their orthologs were found in the rice genome (Fig. 6). When compared with the wild-type, orthologs of AtIRX (Os2g08420) and ZmMADS2 (Os2g36924) were down-regulated in siz1 (Fig. 6). Remarkably, maize ZmMADS2, also highly expressed in anther endothecium and connective tissues before dehiscence and in mature pollen after dehiscence, is essential for anther dehiscence and pollen maturation (Schreiber et al., 2004). In addition, large amounts of ZmMADS2 protein accumulate in degenerating endothecial and connective cells. Both AtIRX and AtNST1 are involved in secondary wall thickening in the anther and subsequent dehiscence (Mitsuda et al., 2005), and their expression is up-regulated by AtMYB26 during secondary wall thickening (Yang et al., 2007). Exploring the existence of putative MYB binding sites in their promoter regions is of importance. From cis-element prediction (Prestridge, 1991; Higo et al., 1999), the promoter regions of Os2g08420 and Os2g36924 contain putative MYB binding sites. All these results suggest that defects in the development of endothecium and connective tissues regulated by the above genes may cause anther indehiscence in the siz1 mutant, and imply that an unknown MYB factor may be the candidate of SIZ1. A few MYB-class transcription factors, such as OsAID1 (MYB) and AtMYB26, regulate anther dehiscence in rice and Arabidopsis. Phenotypes of aid1 anthers share high similarity with the siz1 dehiscent anther. From the above results, we hypothesize that the MYB transcription factors, such as AID1 or rice orthologs of AtMYB26, may be the targets of SIZ1 for post-translational modification, which is essential for their function in regulating the expression of the downstream genes involved in secondary wall thickening in the endothecium and connective tissues during rice anther dehiscence. As SIZ1 is a nuclear protein, SIZ1 may function as a SUMO E3 ligase to indirectly sumoylate the upstream putative transcriptional factor. Or, it may function as a transcriptional coregulator, as found in mammalian cells, to directly sumoylate other components of this transcriptional complex and up-regulate the expression of AtIRX (Os2g08420) and ZmMADS2 (Os2g36924) orthologs before anther dehiscence. Nevertheless, additional genetic and biochemical studies are necessary to illuminate the cellular and molecular mechanisms of these genes and their sumoylation sites recognized by SIZ1 for anther dehiscence. Interestingly, Arabidopsis atsiz1 also showed a reduced number of siliques in addition to slow growth and development (Catala et al., 2007); however, the anther development and dehiscence in the atsiz1 mutant were not fully characterized. It will be interesting to explore whether SIZ1 plays a universal function in the regulation of anther dehiscence in both monocot and dicot plants.

In conclusion, by using T-DNA activation mutants with defective spikelet development and sterility, we have identified a rice gene, SIZ1, encoding a nucleus-localized rice SUMO E3 ligase, and have provided the first preliminary evidence of its role in controlling anther dehiscence in rice, particularly the failure of the degeneration of the endothecium and septum near the stomium during the final step of anther dehiscence.


We greatly appreciate the contributions of Drs Yue-Ie Hsing, Chyr-Guan Chern, Ming-Jen Fan and Su-May Yu in generating the TRIM database. We also thank Dr Ko Shimamoto (Nara Institute of Science and Technology, Japan) for sharing the pANDA-RNAi vector, Dr Shu-Hsing Wu (Institute of Plant and Microbial Biology, Academia Sinica) for her valuable comments on the manuscript, and Drs Su-May Yu (Institute of Molecular Biology, Academia Sinica) and Wann-Neng Jane (Cell Biology Core Facility, Institute of Plant and Microbial Biology, Academia Sinica) for assistance in rice transgenic and scanning electron microscopy experiments, respectively. We also thank Ms. Lin-Yun Kuang (Transgenic Plant Laboratory, Institute of Plant and Microbial Biology, Academia Sinica) for assistance in particle bombardment. We appreciate Ms Laura Smales for her excellent editing work. This work was supported by research grants from Academia Sinica (Taiwan), the National Science and Technology Program for Agricultural Biotechnology (NSTP/AB, Taiwan), the National Science Council (98-2313-B-001-001-MY3, Taiwan) and the Li Foundation (USA) to G.Y. Jauh.