The amino acid sequence of SPLCV-JS (GenBank accession No. FJ176701) C4 was aligned with selected geminivirus C4 protein sequences from GenBank using the ClustalX program. After multiple sequence alignments, phylogenetic analysis was performed with MEGA software version 4.0 using the neighbor-joining method and the bootstrap option (1000 replicates) (Tamura et al., 2007).

Arabidopsis thaliana ecotype Columbia (Col0) was transformed following standard protocols (Zhang et al., 2006). T1 seeds were collected and germinated on selection media and then plantlets were transplanted to soil in a growth chamber (22 ± 1°C, 16 h light/8 h dark photoperiod). The transgene copy number was confirmed using the Southern blot analysis (Sambrook et al., 2001).

The total genomic DNA was extracted from the transgenic Arabidopsis plants according to the method developed by Soni and Murray (1994). For the Southern blot analysis, an aliquot of 10 μg of the genomic DNA from each sample was digested with HindIII and analyzed using a standard protocol for Southern blot (Sambrook et al., 2001). Digested genomic DNA was hybridized with DIG-labeled probe specific to the SPLCV-JS C4 gene. Labeling, hybridization, and chemiluminescent detection were performed according to the instructions of the manufacturer (Roche Applied Science, Germany).

For SPLCV-JS C4 self-activation assay, yeast strain AH109 (Clontech) was transformed with pGBKT7-C4, pGBKT7-C4C31, pGBKT7-C4N55, and the pGBKT7 control. For the C4 and AtBIN2 interaction assay, plasmids pGBKT7-C4, pGBKT7-C4C31, and pGBKT7-C4N55 were co-transformed with plasmid pGADT7-AtBIN2 into the yeast strain AH109. A yeast colony co-transformed with pGBKT7-53 and pGADT7-T plasmids was used as a positive control in the experiment. A yeast strain co-transformed with pGBKT7-lam and pGADT7-T was used as a negative control. All transformants were grown on a dropout medium lacking tryptophan, leucine, histidine, and adenine but containing 5 mM 3-amino-1,2,4-triazole (3-AT) to inhibit residual HIS3 expression. The expression of LacZ reporter gene was initially monitored using X-α-Gal as the β-galactosidase substrate.

Nicotiana benthamiana plants were grown in pots at 23°C with a 16 h light/8 h dark photoperiod and 60% humidity. Before agroinfiltration, the solutions of Agrobacterium tumefaciens harboring binary C4-YFPN or AtBIN2-YFPC were mixed equally. N. benthamiana plants grown to the four to five leaf stages were agroinfiltrated with the bacterial suspension using a syringe. N. benthamiana plants agroinfiltrated with A. tumefaciens harboring C4-YFPN and YFPC or YFPN and AtBIN-YFPC were used as negative controls. The agroinfiltrated leaves were observed under a laser confocal microscope (Carl Zeiss, Jena, Germany) at 36 h post- infiltration (hpi).

N. benthamiana leaves were agroinfiltrated with the A. tumefaciens harboring the expression vectors AtBES1-eGFP and AtBZR1-eGFP, separately. The suspensions of A. tumefaciens harboring the expression vector AtBES1-eGFP/AtBZR1-eGFP and pCAMBIA1301-AtBIN2 were mixed equally and used for agroinfiltration to illustrate the effects of AtBIN2 and C4 on the subcellular localization of AtBES1/AtBZR1. In addition, the suspensions of A. tumefaciens harboring the expression vector AtBES1-eGFP/AtBZR1-eGFP, pCAMBIA1301-AtBIN2, and pCAMBIA1301-C4 were mixed equally for the agroinfiltration experiment. All agroinfiltrated leaves were observed under a laser confocal microscope (36 hpi).

Arabidopsis seeds were surface-sterilized and sown on Murashige and Skoog (MS) agar plates with 30 g/L sucrose. The plates were cold-treated for 2 days at 4°C. For root growth inhibition assays, MS plates were placed in a vertical orientation for 4 days in a growth chamber at 23°C with a 16 h light/8 h dark photoperiod and 60% humidity, and the seedlings were then transferred to MS plates amended with 1 μM epibrassinolide (BL) for 5 days. The root length was measured before and after BL treatment, and the hormone sensitivity was measured as root growth inhibition.

The total RNA was extracted from Arabidopsis seedlings or floral buds containing approximately stages 7–9 anthers using the Tiangen Plant RNAprep Pure Kit (Tiangen, Beijing, China). The first-strand cDNA was synthesized using the Takara PrimeScript First-Strand cDNA Synthesis Kit (TaKaRa, Dalian, China) and then used for real-time quantitative reverse transcription-PCR (qRT-PCR) analysis. Primer sequences for each gene are listed in Supplementary Table S2. PCR reactions were performed in a 20 μL volume containing 2× SYBR Green Master Mix (TOYOBO Co., Ltd.), 50 ng cDNA, 400 nM forward primer, and 400 nM reverse primer in a Bio-Rad CFX96 thermocycler. Actin or a U-BOX gene At5g15400 (Ye et al., 2010) was used as a reference control for seedlings or anthers, respectively. All samples were assayed in triplicate and repeated at least twice. The relative expression levels were calculated using the 2^-ΔΔCt method (Livak and Schmittgen, 2001).

Anthers at the anthesis stage were removed and used to brush pollens into the fluorescein diacetate (FDA) staining buffer on the slide. After staining for 20 min, the slide was observed under an optical microscope in blue light (wavelength = 495 nm), and viable pollen grains emitted fluorescence. FDA staining was performed as described by Li (2011). Nikon NIS-Elements BR 3.0 software was used to calculate the pollen viability.

Arabidopsis flowers that were just open were picked and the pollen grains were dusted on the solid pollen germination medium (Li et al., 1999) and allowed to germinate for 3–5 h or overnight at room temperature. The pollen germination was observed under a microscope Nikon TE2000-S, and the germination ratio was calculated using Nikon NIS-Elements BR 3.0 software.

Arabidopsis inflorescences were incubated for at least 12 h in formaldehyde–acetic acid–ethanol (FAA) fixing solution (10% formaldehyde, 5% acetic acid, and 47.5% ethanol in water), dehydrated in a graded ethanol series (2× 50, 60, 70, 85, and 95%, and 3× 100%), embedded in epoxy resin and sectioned. Anther transverse sections were stained with 1% toluidine blue at 42°C for 1–2 h (Sanders et al., 1999), bright-field photographs were taken using a Nikon TE2000-S microscope.

For scanning electron microscopic (SEM) analysis, wild-type and mutant inflorescences were fixed overnight in FAA, dehydrated in a graded ethanol series as described earlier, and samples were critical-point dried in liquid CO₂. Individual anthers and pollen from flowers that corresponded to a specific stage of wild-type anther development were mounted on SEM stubs. Mounted samples were coated with palladium–gold and then examined under an Autoscan SEM JMS-6360 LV (JEOL, Tokyo, Japan).

For transmission electron microscopic (TEM) analysis, the buds approximately containing anthers at stage 9 were collected and fixed in 2.5% glutaraldehyde in phosphate buffer (pH 7.2) for 4 h at 4°C. After fixation, the tissue was washed and post-fixed with 1% OsO₄ overnight at 4°C. The samples were then dehydrated in a graded ethanol series, infiltrated with a graded series of epoxy resin in epoxy propane, and embedded in Epon 812 resin. Ultrathin sections were stained with 1% uranyl acetate, followed by lead citrate solution and viewed under a TEM (HITACHI H-7650; Hitachi, Tokyo, Japan).

All data were presented as the mean ± standard deviation (SD) of at least three independent experiments. Statistical analysis was performed using the SPSS statistical package, version 15.0 (SPSS, IL, United States). Student’s t-test was used for analyzing the significance of the differences between the wild-type and transgenic Arabidopsis lines. A value of P ≤ 0.05 or P ≤ 0.01 was indicated by an asterisk (^∗) or double asterisks (^∗∗), respectively.

The 258-bp long open-reading frame of SPLCV-JS C4 encoded a 9.2-kDa protein consisting of 85 amino acids. The amino acids BLAST analysis showed that C4/AC4 proteins were diverse among different strains/isolates. The most similar ones were chosen to perform phylogenetic analysis. Most branches of the phylogenetic tree were less than 70 bootstrap values and the SPLCV-JS C4 protein was grouped together with other sweepoviruses (e.g., SPLCV, SPLCJV, and SPLCBV) (Supplementary Figure S1).

Further sequence alignment of SPLCV-JS C4 was performed with three well-studied C4 proteins (also known as AC4 in bipartite begomoviruses), including C4 of BCTV, AC4 of TGMV, and AC4 of East African cassava mosaic Cameroon virus (EACMCV). The N-terminal G² site in a conserved consensus N-myristoylation motif was required for membrane binding of both the bipartite begomoviruses EACMCV AC4 (Fondong et al., 2007) and the monopartite geminivirus BCTV C4 (Piroux et al., 2007). Besides, the four residues P⁴⁴, A⁴⁵, S⁴⁹, and P⁵⁰ within the central domain of BCTV were reported to contribute to pathogenicity (Piroux et al., 2007). Mills-Lujan et al. (2015) showed that S⁴⁹ was phosphorylated, and this phosphorylation was required for BCTV C4 function. SPLCV C4 showed several conserved amino acid residues, including G², P⁴⁴, S⁴⁹, and P⁵⁰ (Figure 1).

Transgenic A. thaliana lines expressing the C4 protein were produced to better characterize the developmental defects induced by the C4 in planta in a system more amenable to genetic studies. The C4 gene was expressed under the control of the CaMV 35S promoter in order to have constitutive expression. After transformation, none of the transgenic Arabidopsis T1 generation showed a segregation ratio of 3:1, indicating that the transgene integration was not a single insertion. Indeed, Southern blot analysis showed that all the transgenic lines were multi-copy insertion (Supplementary Figure S2A). Three independent transgenic lines, C4-11, C4-36, and C4-37, were chosen for further study. Real-time qRT-PCR analyses showed that the expression level of C4 in C4-11 and C4-37 was three or two times higher than that of C4-36 (Supplementary Figure S2B). Compared with the wild type (Figure 2A), C4-transgenic seedlings showed elongated and twisted petioles and severely curled leaves (Figures 2B–D). During growth, the whole plantlet was distorted and deformed, as illustrated by the C4-11 plant line (Figure 2E). Different from Col0 in flowering status (Figures 2E–I), the flower buds of the C4 transgenic plants were not completely enclosed by the sepals and the early flowers were poorly organized (Figures 2J,K). The anthers were shrunk (compare Figure 2H with Figure 2L), and the siliques of the transgenic lines were distorted and had fewer seeds compared with the wild type after destaining (Figure 2M). The phenotype severity was in accordance with the C4 expression level in C4-11, C4-36, and C4-37 seedlings (Supplementary Figure S2B), in which C4-11 had the highest C4 expression level with the most severe phenotype.

SPLCV-JS C4 protein was divided into two segments, the N-terminal C4N55 and C-terminal C4C31, to study the functional domains of C4. Both truncated genes were expressed under the control of CaMV 35S promoter in Arabidopsis. The phenotype of the homozygote Arabidopsis is shown in Supplementary Figure S3. No obvious difference was observed between the C4N55 transgenic Arabidopsis and the wild-type control. However, the differences between the C4C31 transgenic Arabidopsis and the wild-type control were mild, including slightly curved leaves and plants, distorted flowers, and slightly distorted siliques (Supplementary Figure S3). These results demonstrated that the occurrence of severe developmental defects in the transgenic Arabidopsis was correlated with the expression of the N-terminal domain of SPLCV-JS C4. Attempts were also made to generate transgenic sweet potato expressing the C4 protein, but no transgenic plants could be produced using the routine protocol (Yang et al., 2011), suggesting that the expression of the SPLCV-JS C4 might be lethal.

The interaction between SPLCV-JS C4 protein and AtSK21 (AtBIN2) was confirmed by the yeast two-hybrid assay (Figure 3A). The yeast strain co-transformed with the C4 and AtBIN2 could grow well in the dropout medium (SD/-Trp/-His/-Ade) supplemented with 15 mM 3-AT and X-α-Gal, and was blue, in both combinations pGBKT7-C4 + pGADT7-AtBIN2 and pGADT7-C4 + pGBKT7-AtBIN2 (Figure 3A). When C4 was truncated into two domains, N-terminal C4N55 and C-terminal C4C31, the C4C31 could interact with AtBIN2 weakly, and the C4N55 did not interact with AtBIN2 (Figure 3A). Using the BiFC assay, it was further investigated whether the SPLCV-JS C4 protein interacted with the AtBIN2 in plant cells. The C4 alone was localized in both the plasma membrane and the nucleus of N. benthamiana protoplasts and epidermal cells as described in a previous study (Bi and Zhang, 2012). AtBIN2 was previously reported both in the cytoplasm and on the plasma membrane (Vert and Chory, 2006). After agroinfiltration of the constructs expressing C4 fused to YFP (C4-YFPN) and AtBIN2 fused with YFP (AtBIN2-YFPC) into intact N. benthamiana leaves, a strong fluorescent signal could be detected mainly in the plasma membrane of the co-infiltrated leaves 36 hpi (Figure 3B, left panel). No detectable signal was noted in the negative controls (Figure 3B, middle and right panels). The result indicated that SPLCV-JS C4 protein interacted with AtBIN2 in vivo, and the interaction mainly occurred at the periphery of the cell.

AtBIN2 phosphorylated transcription factors BES1/BZR1 to target them for degradation in the proteasome (Yang et al., 2011). However, AtBIN2 activity is inhibited in the presence of BR. Subsequently, unphosphorylated BES1/BZR1 accumulated in the nucleus and recruited proteins such as BIM1 and Myb30 to form diverse transcriptional complexes to regulate the expression of BR-response genes (Yang et al., 2011). The pattern of AtBES1 and AtBZR1 in the presence or absence of the C4 and/or AtBIN2 was studied to better understand the working mode between SPLCV C4 and AtBIN2 interacting proteins.

Both AtBES1 and AtBZR1 tagged with eGFP protein were localized in the nucleus in the N. benthamiana epidermal cells (Figures 4A,B), different from the eGFP control where fluorescence was reported in the cytoplasm and nucleus (Figure 4C). When co-infiltrated with C4, the patterns of AtBES1-eGFP and AtBZR1-eGFP in the N. benthamiana epidermal cells were unchanged 36 hpi (Figures 4D,E) in comparison with the eGFP control (Figure 4F), indicating that C4 protein did not directly affect the subcellular localization of AtBES1/AtBZR1. When AtBES1-eGFP or AtBZR1-eGFP were co-infiltrated with AtBIN2, both AtBES1-eGFP and AtBZR1-eGFP were mainly localized at the cell periphery and in the cytoplasm (Figures 4G,H), suggesting that the proteins were phosphorylated and relocalized from the nucleus to the cytoplasm. This finding was in accordance with previous reports that AtBIN2 could phosphorylate AtBES1/AtBZR1 and target them for degradation in the proteasome (Gampala et al., 2007; Ryu et al., 2007). In contrast, when AtBES1-eGFP or AtBZR1-eGFP were co-infiltrated with AtBIN2 and SPLCV-JS C4, the fluorescent signal was detected only in the nucleus (Figures 4I,J), suggesting that the C4 protein reversed the effects of AtBIN2 on the subcellular localization of AtBES1/AtBZR1 and maintained the accumulation of these two transcription factors in the nucleus.

The effects of C4 on the phenotype and the expression of two BR biosynthetic genes CPD and DWF4, which were regulated by AtBES1/AtBZR1 (He et al., 2005; Yang et al., 2011), were further studied in the presence of BL. After treatment with 1 μM BL, the wild-type Arabidopsis showed severely curled leaves, elongated and twisted petioles, and shorter taproot with more secondary roots (Figure 5C), similar to the C4-expressing lines (Figures 2B–D). Real-time qRT-PCR indicated that before BL treatment, the expression of CPD and DWF4 was much lower in the C4-expressing lines than the wild type. After BL treatment, the gene expression of both CPD and DWF4 was repressed to a lower level in all the C4-expressing lines and the wild type, compared with mock-treated samples (Figures 5A,B). These data showed that the phenotypes with curled leaves, elongated and twisted petioles, and repressed expression of CPD and DWF4 of C4-expressing lines were similar to those of the wild-type plants treated with BL. These results indicated that, just like the case of BL-treated plants, the BR-signaling pathway was activated in the C4-expressing Arabidopsis lines.

The microsporogenesis and pollen morphology of C4 transgenic lines were studied to further characterize the impact of SPLCV-JS C4 on Arabidopsis development. With FDA staining, the pollen viability of C4-11, C4-36, and C4-37 was 15.9, 24.5, and 22.0%, respectively, which was significantly lower than that of wild type (64.9%, Supplementary Figure S4). The pollen germination ratio of C4-11, C4-36, and C4-37 was 26.4, 30.8, and 26.7%, respectively, which was also significantly lower than the 64.4% of wild type (Supplementary Figure S4). The anthers of the C4-expressing Arabidopsis lines usually contained fewer pollen grains than that of the wild type. Transverse section of anthers at stage 9 showed that the microspore numbers of C4-11, C4-36, and C4-37 were reduced compared with that of the wild type. Meanwhile, the shape of many microspores of the transgenic lines was abnormal and vacuolated (Figure 6).

The anthers at the anthesis stage were also observed by SEM analysis. At this stage, the dehiscent anthers of the wild type contained few pollen grains (Figure 7A). A considerable number of pollen grains remained in the inner wall of C4-11, C4-36, and C4-37 anthers (Figures 7B–D), indicating that these C4-expressing lines were deficient in pollen release after anther dehiscence, similar to the previously reported BR-related mutants that were defective in pollen release (Ye et al., 2010). Further observation showed that the inner surface of dehisced anthers from C4-11, C4-36, and C4-37 was similar to that of the wild type, but the exines of the pollen grains of the transgenic lines were abnormal (Figures 7E–H). Under higher magnification, the wild-type exine had a network-like structure with many lacunae and three narrow apertures (Figure 7I). In contrast, the lacunae of C4-11, C4-36, and C4-37 lacked the regularity exhibited by the wild type, and some of the apertures of the transgenic lines were sunken (Figures 7J–L). Further examination of pollen development using TEM analysis showed that the microspores of C4-11, C4-36, and C4-37 at stage 9 were poorly organized compared with that of the wild type (Figure 7M), and the exine was irregular (Figures 7N–P). The complete bacula/tectum structure in the wild type was deficient in the transgenic lines (Figures 7Q–T). Moreover, the bacula/tectum structure had almost disappeared in some transgenic plants (Figures 7R–T). The abnormal structure of the exine in the C4 transgenic lines could explain the deficiency of pollen grains release from the inner surface of the anther wall, much like the bri1-116 mutant of Arabidopsis (Ye et al., 2010).

To investigate the morphological changes that lead to pollen grain defects in the transgenic lines, the anther and microspore morphology of the C4-11, C4-36, and C4-37 lines was examined using TEM. At stage 5, the cell arrangements of tapetum and microspore mother cells of C4-11, C4-36, and C4-37 were irregular compared with those of the wild type (Figures 8A–D), and their tapetal cells were larger and more vacuolated also (Figures 8E–H). At stage 8, a number of vesicles containing fibrous materials could be found in the tapetum of the wild type (Figure 8I), while the tapetum of the transgenic lines was still significantly vacuolated (Figures 8J–L). Since tapetum is responsible for the synthesis and secretion of exine material, the abnormal development of tapetum may be an important reason for the reduced male fertility in the C4 transgenic plants.

The expression levels of key genes related to the genetic program controlling anther and pollen development in Arabidopsis (Ma, 2005) were analyzed in the buds containing anthers at approximately stages 7–9 of the C4 transgenic lines using qRT-PCR to further identify the effect of C4 on male fertility. The expression levels of SPROROCYTELESS/NOZZLE (SPL/NZZ), DYSFUNCTIONAL TAPETUM 1 (DYT1), DEFECTIVE IN TAPEL DEVELOPMENT AND FUNCTION 1 (TDF1), ABORTED MICROSPORES (AMS), MYB domain protein 103 (MYB103), MALE STERILITY 1 (MS1), and MALE STERILITY 2 (MS2) (Ye et al., 2010) were significantly lower than that of the wild type (Figure 9). Furthermore, the expressions of several MS1 downstream genes were measured (Ye et al., 2010). It was found that the expression levels of At4g28395, At3g42960, At3g51590, At1g07340, At3g23770, At1g61110, and At5g62320 in C4-11, C4-36, and C4-37 were lower than that in the wild type, while the expression level of At2g18550 was higher than the wild type (Figure 9). This was in accordance with the case of the BR mutant cpd reported previously (Ye et al., 2010). Since AtBES1 could directly bind to the promoter regions of genes SPL/NZZ, TDF1, AMS, MS1, and MS2 to regulate their expression (Ye et al., 2010), C4-activated AtBES1/AtBZR1 modified the expression pattern of downstream genes that were essential for anther and pollen development, consequently affecting the development of tapetum and microspore and leading to reduced male fertility and decreased seed number in the transgenic lines.