Agrobacterium exports at least five virulence proteins (VirE2, VirE3, VirF, VirD2, VirD5) into host cells and hijacks some host plant factors to facilitate its transformation process.
Random DNA binding selection assays (RDSAs), electrophoretic mobility shift assays (EMSAs) and yeast one-hybrid systems were used to identify protein-bound DNA elements. Bimolecular fluorescence complementation, glutathione S-transferase pull-down and yeast two-hybrid assays were used to detect protein interactions. Protoplast transformation, coprecipitation, competitive binding and cell-free degradation assays were used to analyze the relationships among proteins.
We found that Agrobacterium VirD5 exhibits transcriptional activation activity in yeast, is located in the plant cell nucleus, and forms homodimers. A specific VirD5-bound DNA element designated D5RE (VirD5 response element) was identified. VirD5 interacted directly with Arabidopsis VirE2 Interacting Protein 1 (AtVIP1). However, the ternary complex of VirD5–AtVIP1–VirE2 could be detected, whereas that of VirD5–AtVIP1–VBF (AtVIP1 Binding F-box protein) could not. We demonstrated that VirD5 competes with VBF for binding to AtVIP1 and stabilizes AtVIP1 and VirE2 in the cell-free degradation system.
Our results indicated that VirD5 may act as both a transcriptional activator-like effector to regulate host gene expression and a protector preventing the coat proteins of the T-complex from being quickly degraded by the host's ubiquitin proteasome system (UPS).
Agrobacterium (Agrobacterium tumefaciens) is a soilborne pathogenic bacterium that can induce crown gall on various plant species, including some agronomically important crops (Magori & Citovsky, 2012). During the Agrobacterium infection process, part of its genetic material, a single-stranded copy of the transferred DNA (T-DNA, a segment of its tumor-inducing (Ti) plasmid DNA), is transferred into host cells and subsequently integrated into the host genome (Gelvin, 2000, 2010; Tzfira & Citovsky, 2000, 2002; Pitzschke & Hirt, 2010; Magori & Citovsky, 2012). As a result, Agrobacterium has been widely used as a tool for the genetic modification of many plant species as well as some fungal species (Lacroix et al., 2006; Magori & Citovsky, 2012).
In addition to T-DNA, Agrobacterium also transfers at least five virulence effector proteins (VirE2, VirE3, VirF, VirD2 and VirD5) (Pitzschke & Hirt, 2010; Magori & Citovsky, 2011b) into the host cells through the type IV secretion system (T4SS), composed of the 11 VirB proteins and VirD4 (Fronzes et al., 2009), to facilitate the infection process. Vergunst et al. (2000, 2005) have established that a positively charged C-terminal transport signal on virulence proteins is essential for their translocation and that mutation negatively affects translocation. These proteins interact with numerous plant factors in plant cells during the infection process and the interactions are considered important for the Agrobacterium infection process. Through these interactions, Agrobacterium uses and abuses the host biological systems to meet its own needs (Djamei et al., 2007; Dafny-Yelin et al., 2008).
Within the host cells, before integration into the host genome, the T-DNA is believed to exist as a nucleoprotein complex (T-complex), in which one VirD2 molecule is covalently attached to the 5′ end of the T-DNA, and numerous VirE2 proteins coat the T-DNA (Citovsky et al., 1989; Zupan et al., 2000; Tzfira & Citovsky, 2002). The Arabidopsis (Arabidopsis thaliana) VirE2-interacting protein 1 (AtVIP1), which is a basic leucine zipper (bZIP)-family transcription factor, has been demonstrated to play key roles during Agrobacterium-mediated plant transformation (Tzfira et al., 2001, 2002; Ward et al., 2002; Li et al., 2005; Djamei et al., 2007). AtVIP1 is a direct target of the Agrobacterium-induced mitogen-activated protein kinase (MAPK) MPK3 and is delocalized from the cytoplasm to the nucleus upon phosphorylation by MPK3 (Djamei et al., 2007). Because it can form a ternary complex with VirE2 and ssDNA, AtVIP1 is believed to facilitate transportation of the T-complex into the cell nucleus (Tzfira et al., 2001). However, Agrobacterium exports into host cells another virulence protein, VirE3, which can also bind to VirE2 and mediate nuclear import of the T-complex (Lacroix et al., 2005; Garcia-Rodriguez et al., 2006). Furthermore, AtVIP1 can interact with all classes of core histones (H2A, H2B, H3, H4) and bind directly to the purified plant nucleosomes (Li et al., 2005; Loyter et al., 2005; Lacroix et al., 2008; Gelvin, 2010), indicating that AtVIP1 may act as a molecular adaptor to guide the T-complex to plant chromatin for subsequent integration. When acting as a transcription factor, AtVIP1 could bind to specific DNA elements (VIP1 response elements, VREs) and mediate MPK3-induced stress gene expression (Pitzschke et al., 2009).
It was reported that the AtVIP1 could be divided into two functional parts: the N-terminal portion (aa 1–164) interacts with VirE2, whereas the C-terminal portion (aa 165–341) is required for protein multimerization and for binding to the host H2A histone and specific DNA element (Li et al., 2005; Pitzschke et al., 2009).
It was thought that the coat proteins were removed before T-DNA integration into the host genome (Tzfira et al., 2004). Two F-box proteins, the Agrobacterium VirF (Tzfira et al., 2004) and plant AtVIP1-Binding F-box protein (VBF) (Zaltsman et al., 2010), function in this process. VirF, which is only present in the octopine Ti plasmid (Melchers et al., 1990; Regensburg-Tuink & Hooykaas, 1993) and not in the nopaline- and agropine-type Ti plasmids, is the first F-box protein identified in prokaryotes (Magori & Citovsky, 2011b). It interacts with plant Skp1 (S-phase Kinase-associated Protein 1) and functions as a subunit of the SCF (Skp1-Cull-F-box protein) ubiquitin E3 ligase complex (Schrammeijer et al., 2001) to degrade AtVIP1 and VirE2 by directly binding to them (Tzfira et al., 2004; Lacroix et al., 2008). The VBF is an F-box protein identified in plants. Its expression is induced by Agrobacterium (Zaltsman et al., 2010). Because VBF can form a ternary complex with AtVIP1 and VirE2 and can elicit their degradation via the SCFVBF pathway (Zaltsman et al., 2010, 2013), it is considered to be involved in the T-complex uncoating process and required for tumorigenicity. However, this produces two scenarios. First, if the T-complex is uncoated more rapidly before it targets host chromatin for integration, the T-DNA might not be integrated and might be expressed only transiently or degraded by intracellular enzymatic activity. Alternatively, the T-DNA might also not be integrated if VBF does not elicit the degradation of AtVIP1 and VirE2 on time after the T-complex reaches the host chromatin. For completing integration, how does Agrobacterium control when the coat proteins of the T-complex should be degraded and when they should not be?
Through nucleotide sequence analysis, multiple research groups found an open reading frame (ORF) downstream of the VirD4 locus of the VirD operon (Rogowsky et al., 1990; Lin & Kado, 1993; Kalogeraki et al., 2000; Schrammeijer et al., 2000). One group, using an analysis of this ORF in the nopaline-type Ti plasmid pTiC58, considered that this ORF is not within the virD operon (Rogowsky et al., 1990; Lin & Kado, 1993) because its expression is dependent on its own promoter and independent of acetosyringone (AS) induction. Another group, through an analysis of this ORF in octopine-type Ti plasmid, considered it to be the fifth ORF (ORF5) within the virD operon and designated it VirD5. This group found that the expression of VirD5 is strongly induced by AS (Kalogeraki et al., 2000).
Some early preliminary studies reported that VirD5 was not essential for tumorigenicity because its deletion mutant did not affect the tumor formation of A. tumefaciens (Stachel & Nester, 1986; Koukolikova-Nicola et al., 1993; Lin & Kado, 1993; Kalogeraki et al., 2000). However, later studies have shown that the VirD5 protein can be transported into the nucleus of the host plant (Vergunst et al., 2005; Magori & Citovsky, 2011a) and counteract host-induced degradation of VirF by directly interacting with VirF (Magori & Citovsky, 2011a). Lack of VirD5 would significantly reduce the efficiency of tumor formation by Agrobacterium, suggesting that VirD5 is required for efficient Agrobacterium infection (Magori & Citovsky, 2011a).
We studied VirD5 in the agropine-type Ti plasmid pTiBo542. Our results indicated that it has characteristics of eukaryotic transcription factors and is a putative transcriptional activator-like effector virulence protein. Furthermore, our results demonstrated that VirD5 could interact with AtVIP1 and form a ternary complex with AtVIP1 and VirE2, indicating that it may target the T-complex in the nucleus. However, it cannot form a ternary complex with AtVIP1 and VBF, but competes with VBF for binding to AtVIP1. We also showed that VirD5 could counteract VBF-induced degradation of AtVIP1 and VirE2. Our data suggested that VirD5 functions as both a transcriptional activator-like effector and a protector that prevents the coat proteins of the T-complex from being degraded more rapidly by the host's SCFVBF pathway.
Materials and Methods
Yeast two-hybrid assays
The cDNA of VirD5 (full-length and truncated) and AtVIP1 (coding for C-terminal amino acids 165-341) were cloned into pGBKT7 and/or pGADT7 vectors (Clontech, Mountain View, CA, USA). Pairs of the constructed plasmids were used to cotransform yeast strain AH109 following the user manual (Clontech). The transformed cells were first cultured on synthetic defined (SD) medium lacking Leu and Trp, and then transferred onto SD medium lacking Leu, Trp, His and Ade and containing 40 μg ml−1 X-α-Gal.
Random DNA binding selection assay (RDSA)
The general protocol is based on Pitzschke et al. (2009). Two random DNA fragments (RDSA1 and RDSA2) with 20 and 18 bp random regions flanked by two independent sets of primer annealing sites were used for the binding selection assays. The recombinant VirD5 protein immobilized by chitin beads (NEB) or glutathione sepharose beads (GE) was incubated with RDSA1 or RDSA2 for > 2 h at 4°C with gentle rolling. The RDSA buffer (Pitzschke et al., 2009) was used to wash the beads at least four times. Then, 500 μl sterile double-distilled water (ddH2O) was added and the beads were boiled for 5 min to release the bound DNA. The DNA was purified using 24 : 1 (chloroform : isoamyl alcohol), precipitated by isopropanol, and washed with 70% alcohol. The PCR reagents (74 μl ddH2O, 10 μl 10× PCR buffer, 10 μl 2.5 mM deoxyribonucleotide triphosphate, 2.5 μl of 20 mM RDSA1 fo/RDSA2 fo primer, 2.5 μl of`20 mM RDSA1 re/RDSA2 re primer, 1.0 μl rTaq DNA polymerase) were added to the tubes containing the DNA precipitates and PCR was performed using the following conditions: 5 min at 94°C; 30 cycles of 30 s at 94°C, 30 s at 50°C, and 30 s at 72°C; 10 min at 72°C. The PCR products were purified, precipitated, and washed as described earlier, and they were then used for the next RDSA cycle. In total, eight cycles were conducted. By increasing the amounts of the nonspecific competitor poly(deoxyinosinic-deoxycytidylic) acid sodium salt (poly(dIdC); 0, 50, 100, 150, 200, 250, 300 and 350 ng) and reducing the amounts of protein in the binding reactions, the stringency was increased in each subsequent purification cycle. The PCR products from the last cycle were cloned into pGEM-T Easy (Promega) and sequenced.
Electrophoretic mobility shift assay (EMSA)
The oligonucleotides for D5RE and D5REm, along with their complementary oligonucleotides, were labeled with biotin using the Biotin 3′ End DNA labeling Kit (Thermo Scientific, Waltham, MA, USA). The EMSAs were performed using the LightShift Chemiluminescent EMSA Kit (Thermo Scientific).
Yeast one-hybrid system
Three pairs of complementary oligonucleotides, D5RE-S and D5RE-AS, D5REm-S and D5REm-AS, and D5REm1-S and D5REm1-AS (Supporting Information, Table S1), were synthesized. Each pair of oligonucleotides was annealed and ligated into a yeast reporter vector, pAbAi (Clontech). The constructed vector plasmids were linearized with BstI and transferred into yeast strain Y1HGold for constructing the yeast bait strains. The empty pGADT7 (AD) and the constructed AD-VirD5 vector plasmids were used to transform each yeast bait strain. The transformed yeast cells were first spotted on SD minus uracil and leucine (SD/-Ura-Leu) medium, and then the cultured yeast colonies were resuspended in sterile ddH2O, adjusted to an OD600 of 0.8, and transferred onto SD/-Leu-Ura minimal medium and SD/-Leu-Ura minimal medium with 1000 ng ml−1 Aureobasidin A for culture.
Protoplast isolation and transformation
The methods for rice protoplast isolation and transformation were based on Bart et al. (2006). The methods for Arabidopsis protoplast isolation and transformation were as described by Yoo et al. (2007). Approx. 30 μg endotoxin-free plasmid DNA was used to transform rice and Arabidopsis protoplast cells.
Subcellular location of fusion proteins in protoplast
Three transient expression vectors were constructed: 35S::VirD5-YFP, 35S::VirD5C-YFP and 35S::VirD5N-YFP. The empty yellow fluorescent protein (YFP, as control) and the construct YFP vector plasmid DNA were used to cotransform rice protoplasts with 35S::Ghd7-CFP. The 35S::Ghd7-CFP was used as a nuclear localization marker. The subcellular location analyses were performed with a confocal laser scanning microscope (Zeiss LSM data server). Fluorescence was excited at 458 nm (cyan fluorescent protein, CFP) and 514 nm (YFP), and emissions were detected from the following wavelength ranges: 505–530 nm (YFP) and 465–480 nm (CFP). For each subcellular location analysis, at least three repeats were performed and at least 10 cells were examined in each repeat.
Bimolecular fluorescence complementation (BIFC), bridge-BIFC and multicolor BIFC
For BIFC analysis, the full-length coding sequences of VirD5 and Os12g0562400 were cloned into both pSPYCE(M) and pSPYNE173 vectors (Waadt et al., 2008). The coding sequences of IMPA-1, IMPA-4 and AtVIP1 (full-length, N-terminal and C-terminal sequences) were cloned into the pSPYNE173 vector (Waadt et al., 2008). Pairs of the constructed vector plasmids were used to cotransform protoplast cells. Fluorescent signals were detected with a confocal laser scanning microscope (Zeiss LSM data server).
For bridge-BIFC and multicolor BIFC analysis, the full-length coding sequences of VirE2 and VBF were cloned into both pSAT6-cCFP-N and pSAT6-nCerulean-N vectors (Lee et al., 2008). The full-length coding sequences of AtVIP1 and VirD5 were cloned into pSAT6-nCerulean-N, pSAT6-cCFP-N and pSAT6-nVenus-N vectors (Lee et al., 2008). For expression of free AtVIP1, we used the EcoRI-BamHI fragment of AtVIP1 to replace the YFP fragment in the PM999-YFP vector, yielding the 35S::AtVIP1 transient transformation vector. The transient transformation vector plasmid DNA was used to cotransform rice and/or Arabidopsis protoplast cells. Fluorescent signals were detected by confocal laser scanning microscope. The detection conditions were as follows: for cCFP/nVenus signal, line active 488 nm, BP 505–530, detector gain 550–750; for cCFP/nCerulean signal, line active 458 nm, LP 475, detector gain 650–850. All experiments were repeated at least three times.
Glutathione S-transferase (GST) pull-down, coprecipitation and Western blot analysis
For GST pull-down analysis, the full-length coding sequence of VirD5 was cloned into PET-28a(+) and pGEX-6p-1 vectors, yielding His-VirD5 and GST-VirD5 fusion expression vectors, respectively. The full-length coding sequence of AtVIP1 was cloned into the pGEX-6p-1 vector, and Os12 g0562400 was cloned into PET-28a(+) as a negative control. The constructed vectors were transferred into Escherichia coli BL21 (DE3) cells for heterologous expression of fusion proteins. The bacterial lysates (supernatant) in PBS buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, 5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, pH 7.3) were incubated with 30 μl glutathione sepharose beads for 1–2 h at 4°C. After four washes with PBS buffer, 20 μl of each sample was loaded onto an 8% sodium dodecyl sulfate polyacrylamide gel electrophoresis gel. Proteins were detected on a polyvinylidene difluoride membrane (Roche, Mannheim, Germany) by Western blot using His antibodies and GST antibodies.
For coprecipitation assays, in addition to those used in the pull-down analysis, two constructs (VBF-chitin binding domain (VBF-CBD) and VirE2-CBD) with the coding sequences of VBF and VirE2 in pTXB3 vector (NEB) were also used. Individual recombinant proteins (GST-AtVIP1, His-VirD5, VBF-CBD and VirE2-CBD) were expressed in the BL21 (DE3) strain of E. coli. E. coli cells were sonicated in PBS buffer. The lysate was centrifuged at 10 000 g for 10 min at 4°C. Trios of the supernatants (300–600 μl) GST-AtVIP1, His-VirD5 and VBF-CBD/VirE2-CBD were mixed and incubated for 2 h at 4°C. Then, 60 μl of PBS-washed chitin beads or glutathione sepharose beads were added to the mixture and incubated for 4 h at 4°C with gentle rolling. The CBD and GST proteins were used as negative controls. After four washes with PBS, the captured proteins were released by mixing with 60 μl of 2 × SDS sample buffer (4% SDS, 0.2% bromophenol blue, 100 mM Tris-Cl (pH 6.8), 20% glycerol, 200 mM dithiothreitol) and boiling for 6 min and were analyzed by Western blotting using anti-CBD (diluted 1 : 3000), anti-GST (diluted 1 : 3000), and anti-His (diluted 1 : 3000) antibodies.
Competitive binding assay
Aliquots of the GST-AtVIP1 bacterial lysates (supernatants, 50 μl) were incubated with incrementally increasing amounts of the His-VirD5 bacterial lysates (supernatants, 0–1000 μl). The mixtures were then added to the tubes containing the immobilized VBF-CBD with chitin beads and further incubated for at least 4 h at 4°C with gentle rolling. After at least four washes with PBS, the captured proteins were released by mixing 60 μl of 2× SB buffer and boiling for 6 min, and they were analyzed by Western blotting using anti-CBD (diluted 1 : 3000) and anti-GST (diluted 1 : 3000). The original GST-AtVIP1 and His-VirD5 mixtures were also analyzed by Western blot using anti-GST (diluted 1 : 3000) and anti-His (diluted 1 : 3000) antibodies, respectively, to indicate that the same amounts of GST-AtVIP1 and increasing His-VirD5 were used in different tubes. Three independent experiments were performed for this assay.
Cell-free degradation assay
Aliquots of the GST-AtVIP1 (20 μl) and VirE2-CBD (10 μl) bacterial lysates (supernatants) were incubated with PBS or His-VirD5 (500 μl) for at least 2 h at 4°C. The cell-free degradation assay was performed as previously described (Magori & Citovsky, 2011a), with some modifications. Nicotiana tabacum leaves were ground to a fine powder in liquid nitrogen. The total protein was extracted by resuspending the powder in degradation buffer (Magori & Citovsky, 2011a). The total protein (300 μl) and bacterial lysates containing CBD or VBF-CBD fusion protein (150 μl) were added to the tubes and incubated at room temperature. Samples were collected at the indicated times and analyzed by Western blot using anti-GST and anti-CBD antibodies. A major band at c. 50 kDa (putative ribulose-1,5-bisphosphate carboxylase oxygenase (RuBisCO) large subunit) on Coomassie-stained gels was used to indicate that the same amount of N. tabacum total protein was added. Three independent experiments were performed for each assay. All immunoblots were quantified with GelScan software (Frankfurt, Germany). The graphs were made in GraphPad Prism 5 software (La Jolla, CA, USA).
VirD5 has transcriptional activation activity in yeast
Previous researchers speculated that VirD5 may function as a transcription factor within host cells (Schrammeijer et al., 2000). To confirm this speculation, we conducted a transcriptional activation test in yeast cells. In this experiment, VirD5 was fused to GAL4-DNA-binding domain (BD) in pGBKT7 vector (Clontech). When the construct was alone or cotransferred with pGADT7-Rec-T into the yeast strain AH109, it could activate the expression of the reporter genes (HIS3, ADE2, MEL1) driven by GAL promoters, allowing the yeast to grow and show a blue color on minus medium supplemented with X-α-Gal (Fig. 1a,b). This result indicated that VirD5 has transcriptional activation activity.
To identify the potential activation domain region of VirD5, a series of N-terminal or C-terminal orderly truncated VirD5 fragments were fused to the BD domain to repeat the experiment described. The result (Fig. 1c) showed that all the C-terminal truncated fragments, except for D5▵13 (amino acids 1-583), presented transcriptional activation activity in yeast, whereas all the N-terminal truncated fragments did not. This result suggested that the N-terminal region of 102 amino acids contains the domain necessary for the transcriptional activation activity. However, D5▵13 did not show any activation activity. The Western blot result (Fig. S1) showed that the D5▵13 construct could be expressed in the transformed yeast cells. We hypothesized that this deletion might have changed the protein-folding patterns and thereby affected its activation activity (see below (Fig. 3c) for more evidence).
VirD5 is transported into the plant cell nucleus and interacts with importin-α
The VirD5 homologous proteins are highly conserved in nearly all types of Ti and Ri plasmids (Fig. S2a), and all of them contain potential eukaryotic nuclear location signals (NLSs) (Fig. S2b). For example, PSORT analysis (http://psort.hgc.jp/form.html) showed that the VirD5 from the agropine-type Ti plasmid pTiBo542 (Gene ID: 6382152) used in this research contained three NLSs: 173KRKR176, 769KKDLEAKSVGVRRKKKE785, and 820RRVYDPRDRAQDKAFKR836. One of them is located on the N-terminal, and the other two are located on the C-terminal. To detect whether this VirD5 protein can be transported into the plant cell nucleus, we performed a subcellular localization assay. The transient expression vectors, 35S::VirD5–YFP, 35S::VirD5N–YFP and 35S::VirD5C–YFP, were each cotransferred into rice protoplast cells with another transient expression vector, 35S::Ghd7-CFP. The Ghd7 was used as a cell nucleus localization marker (Xue et al., 2008; Yan et al., 2013). The results (Fig. 2a,b) showed that the full-length construct, the N-terminal fragment (VirD5N, amino acids 1–317), and the C-terminal fragment of VirD5 (VirD5C, amino acids 318–836) could be found in the plant cell nucleus. The nuclear localization was consistent with previous results for VirD5 from octopine-type and nopaline-type Ti plasmids (Vergunst et al., 2005; Magori & Citovsky, 2011a). This result also indicated that both the N-terminal and C-terminal NLSs have functional significance.
Importin-α can bind to NLS-containing proteins and import them into the cell nucleus (Chook & Blobel, 2001; Goldfarb et al., 2004). It was reported that two Arabidopsis importin-α proteins (IMPA-1, AT3G06720; IMPA-4, AT1G09270) interacted with another virulence protein VirE3, and mediated its nuclear import (Garcia-Rodriguez et al., 2006). We performed a BIFC assay to assess whether VirD5 could interact with these two importin-α proteins. Fig. 2(c) showed that VirD5 interacted with two Arabidopsis importin-α proteins in Arabidopsis and rice protoplast cells and that the interactions were specific because VirD5 did not interact with the unrelated nuclear protein (Os12g0562400, for which the subcellular localization result is shown in Fig. S3) in the BIFC system. The recombinant YFP fluorescence signal was colocalized with the Ghd7-CFP signal, indicating that the detected interaction location is the plant cell nucleus. This interaction with importin-α indicated that VirD5 could be translocated into the plant cell nucleus, which was consistent with our subcellular localization result.
VirD5 forms homodimers in vivo and in vitro
The bioinformatics analysis showed that VirD5 had no apparent homology to other proteins. This made it difficult to characterize in terms of its biological functions. It is well known that eukaryotic transcriptional factors often form homo- or heterodimers when performing their functions (Deppmann et al., 2006; Amoutzias et al., 2008). To determine whether VirD5 could form homodimers in host cells, we conducted a BIFC experiment. VirD5 was tagged with nYFP and cYFP in the BIFC vector system (Waadt et al., 2008). After they were cotransferred into the rice (monocotyledon) or Arabidopsis (dicotyledon) protoplast cells, reconstructed YFP singles were detected in the nucleus (Fig. 3a). This result indicated that VirD5 could form homodimers in the plant cell nucleus.
To verify this interaction, we performed a GST pull-down experiment. His-tagged VirD5 and GST-tagged VirD5 were expressed for the pull-down assay. In this experiment, the GST-tagged VirD5 was used as bait, and the His-tagged VirD5 was used as prey. GST and His-tagged Os12g0562400 constructs were used as negative controls in this assay. The glutathione sepharose bead pull-down and immunoblotting assays showed that VirD5 could form homodimers in vitro (Fig. 3b).
The BIFC and GST pull-down results were further confirmed by the yeast two-hybrid assay (Clontech). To rule out the influence of autoactivation, we used truncated VirD5 fragments for the yeast two-hybrid experiment. Fig. 3(c) showed that the BD-truncated VirD5 constructs could specifically interact with the AD-truncated VirD5 constructs in the yeast cells except for the truncated VirD5 fragment D5▵13, which did not show any interaction with other truncated VirD5 fragments, most likely because this deletion changed the protein-folding patterns. This was consistent with our earlier results from the transcriptional activation experiment. These interaction results show that VirD5 could form homodimers in vitro and in vivo in the cell nuclei of dicotyledons and monocotyledons.
VirD5 binds to a specific DNA element
The earlier studies (Schrammeijer et al., 2000) mentioned that VirD5 might contain eukaryotic DNA binding characteristics and might function as a transcription factor in plant cells. Our results reported earlier in this paper also showed that VirD5 has some characteristics similar to those of eukaryotic transcriptional factors. To identify the DNA elements bound by VirD5, we performed RDSAs (Pitzschke et al., 2009). We sequenced 20 independent candidate clones from the DNA fragments enriched by the GST-VirD5 fusion protein and found that nine (45%) contained the common element: CCGCNC (Fig. 4a). We then sequenced another nine independent candidate clones from the DNA fragments enriched by the VirD5-CBD fusion protein and found that seven (77.8%) contained the same common element (Fig. 4b).
We named this consensus element D5RE (VirD5 response element). To test whether VirD5 binds directly to D5RE, an EMSA was conducted using recombinant VirD5 protein and 3′-end biotin-labeled DNA fragments (Fig. 4c). The result (Fig. 4d) showed a clear band shift when the VirD5 was incubated with the biotin-labeled D5RE, and this shifted band was not eliminated by poly(dIdC) but was eliminated by a 200-fold molar excess of unlabeled D5RE. When CCGCNC was mutated to atGCNC (D5REm), no corresponding band shift could be detected. We observed that poly(dIdC) could decrease the binding to some extent in repeat experiments; this may be because our DNA element was GC-rich. These results indicated that VirD5 directly targeted the CCGCNC DNA element.
The RDSA and EMSA results were further confirmed using the yeast one-hybrid system. In this experiment, DNA fragments (D5RE, D5REm, D5REm1) (Fig. 4c) were cloned into the yeast reporter vector pAbAi to construct the Y1HGold yeast bait strains. The pGADT7 (AD) empty vector and the AD constructs of VirD5 were used to transform each yeast bait strain. Fig. 4(e) showed that only when the AD construct of VirD5 was transferred into the D5RE yeast bait strain (pD5RE-AbAi) could the yeast grow on the SD/-Ura-Leu plus aureobasidin A (AbA) medium, demonstrating that VirD5 could bind to the specific DNA element, D5RE, in yeast cells.
To identify the potential DNA binding domain of VirD5, we also detected the binding activity of the truncated VirD5 fragments. These results (Fig. 4f) showed that the VirD5∆12 (1–719), VirD5∆13 (1–583), VirD5∆14 (1–444), VirD5∆21 (103–836), VirD5∆22 (339–836) and VirD5∆23 (445–836) constructs could bind to D5RE. This indicated that VirD5 might contain two sequential DNA binding domains (from amino acid 339–444 and 445–583).
VirD5 interacts with AtVIP1
Based on these results, VirD5 is an eukaryotic transcription factor-like Agrobacterium virulence protein. It is well known that the plant transcriptional factor, AtVIP1, plays important roles during the Agrobacterium infection process (Tzfira et al., 2001; Li et al., 2005; Djamei et al., 2007). To determine whether these two factors could interact with each other, we performed BIFC experiments in rice and Arabidopsis protoplast cells. Fig. 5(a) showed that when cYFP-tagged VirD5 and nYFP-tagged AtVIP1 (full-length, N-terminal and C-terminal) were coexpressed, a strong fluorescence signal of the reconstructed YFP could be detected, indicating that VirD5 could interact with AtVIP1 in vivo and that both the N-terminal and C-terminal of AtVIP1 mediated this interaction. The YFP fluorescence signal could be completely merged with the CFP-tagged Ghd7 signal, indicating that the location of the interaction is the cell nucleus. The localization of this interaction was consistent with their subcellular location.
To verify this interaction, we performed GST pull-down assays. GST-tagged AtVIP1 and His-tagged VirD5 were expressed for the pull-down assay. In these experiments, the GST-tagged AtVIP1 was used as bait, and His-tagged VirD5 was used as prey. The results (Fig. 5b) confirmed that VirD5 could specifically interact with AtVIP1 in vitro. The yeast two-hybrid system was used to further confirm the interaction between VirD5 and AtVIP1. Because the full-length AtVIP1 protein has transcriptional activation activity (Tsugama et al., 2012), its C-terminal domain (aa 165-341) was used to conduct this experiment. The interaction between VirD5 and AtVIP1 could be detected in yeast two-hybrid assays (Fig. 5c). Together, these results demonstrated that VirD5 could specifically interact with AtVIP1.
VirD5 forms a ternary complex with AtVIP1 and VirE2 in the plant cell nucleus
Previous studies showed that AtVIP1 could interact with VirE2 (Tzfira et al., 2001), and that VirD5 could not interact with VirE2 (Magori & Citovsky, 2011a). We conducted a bridge-BIFC assay to detect whether VirD5, AtVIP1 and VirE2 could exist in the same complex. VirD5 was tagged with nVenus and VirE2 with cCFP (Lee et al., 2008). Fig. 6(a) shows that when free AtVIP1 was coexpressed with VirD5-nVenus and VirE2-cCFP, a strong YFP signal (nVenus and cCFP interaction signal) could be observed in the cell nucleus. If only VirD5-nVenus and VirE2-cCFP were used for coexpression, no YFP signal could be detected in rice protoplast cells, and only feeble YFP signal could be observed in some Arabidopsis protoplast cells (data not shown), probably as a result of the effect of the endogenous AtVIP1. This result indicates that VirD5, AtVIP1 and VirE2 could form a ternary complex in the nucleus.
Next, we performed a multicolor–bridge BIFC assay in which the nCerulean-tagged AtVIP1 was used to replace the free AtVIP1 and coexpressed with VirE2-cCFP and VirD5-nVenus in Arabidopsis and rice protoplast cells. This experiment detected both yellow fluorescence produced by the cCFP with nVenus and blue fluorescence produced by the cCFP with nCerulean in the same cell (Fig. 6b), indicating that VirD5, AtVIP1 and VirE2 could exist in the same complex in living cells.
To further confirm this result, we performed an in vitro coprecipitation experiment. His-tagged VirD5, GST-tagged AtVIP1 and CBD-tagged VirE2 were expressed for this assay. When the GST-tagged AtVIP1 was used as bait, glutathione sepharose bead precipitation and immunoblotting assays showed that both His-VirD5 and VirE2-CBD could be captured efficiently in vitro compared with when GST alone was used as bait (Fig. S4), indicating that AtVIP1 could interact with both VirD5 and VirE2 in the same system. When the CBD-tagged VirE2 was used as bait, chitin bead precipitation and immunoblotting assays showed that both GST-AtVIP1 and His-VirD5 could be captured efficiently in vitro compared with when CBD alone was used as bait (Fig. 6c), indicating that these three proteins exist in a ternary complex.
VirD5, AtVIP1 and VBF cannot exist in the same complex
Previous studies demonstrated that VBF could interact with AtVIP1 (Zaltsman et al., 2010) but not with VirD5 (Magori & Citovsky, 2011a). VBF could mediate the degradation of AtVIP1 and its interactor VirE2, which could form a ternary complex with AtVIP1 and VBF (Zaltsman et al., 2010). Therefore, we examined whether VirD5, like VirE2, could form a ternary complex with AtVIP1 and VBF. The bridge-BIFC results (Fig. 7a) showed that VBF could not interact with VirD5, consistent with previous studies, and even when free AtVIP1 was coexpressed with VBF-cCFP and VirD5-nVenus, no yellow fluorescence signal could be detected. This result indicated that VirD5, AtVIP1 and VBF could not form a ternary complex in planta.
To verify this conclusion, multicolor-bridge BIFC was performed. These results (Fig. 7b) showed that when VBF-cCFP, AtVIP1-nVenus and VirD5-nCerulean were coexpressed in the Arabidopsis protoplast cells, yellow but not blue fluorescence signal could be detected. When we instead coexpressed VirD5-cCFP, AtVIP1-nVenus and VBF-nCerulean, we could again only detect yellow signal. However, when AtVIP1-cCFP, VirD5-nVenus and VBF-nCerulean were coexpressed, we could detect both yellow signal and blue signal. These results indicate that both VBF and VirD5 could interact with AtVIP1, but VBF, AtVIP1 and VirD5 could not form a ternary complex in living cells.
To further confirm this result, a coprecipitation experiment was conducted. In this experiment, VBF, AtVIP1 and VirD5 were tagged with CBD, GST and 6× His, respectively. Then anti-CBD, anti-GST and anti-His antibodies were used to detect the input and samples precipitated by chitin or glutathione sepharose beads. These results (Fig. S5, Fig. 7c) showed that AtVIP1 could interact with both VBF and VirD5 in the same system (Fig. S5), but that the three proteins could not exist in the same complex (Fig. 7c).
VirD5 competes with VBF for binding to and stabilizing AtVIP1
To assess whether VirD5 can compete with VBF for binding to AtVIP1, we performed a competitive binding assay. The same amount of GST-AtVIP1 fusion protein was incubated with a series of incrementally increasing amounts of His-VirD5 fusion protein. Then, the same amount of recombinant VBF-CBD was used as bait to capture AtVIP1. Pull-down with chitin beads and immunoblotting assays showed that the captured AtVIP1 rapidly decreased with the increase in His-VirD5. When an excess of His-VirD5 was incubated with GST-AtVIP1 before the addition of VBF-CBD, no captured GST-AtVIP1 could be detected in the pull-down samples (Fig. 8), indicating that VirD5 could compete with VBF for binding to AtVIP1. Because VBF induces the degradation of AtVIP1 and VirE2 via the SCFVBF pathway (Zaltsman et al., 2010, 2013), VirD5 may counteract this degradation. To determine whether VirD5 could stabilize AtVIP1 and VirE2 by preventing their VBF-mediated degradation, we analyzed the stability of GST-AtVIP1 and VirE2-CBD in the presence of VBF-CBD and in the presence or absence of His-VirD5 in a cell-free degradation system with total proteins extracted from N. tabacum. These results (Fig. 9a,b) showed that, in the presence of VBF-CBD protein, the amounts of GST-AtVIP1 and VirE2-CBD protein declined rapidly and nearly disappeared after 4 h of incubation. However, this degradation was significantly delayed when GST-AtVIP1 and VirE2-CBD were incubated with His-VirD5 fusion protein before adding VBF-CBD. These results demonstrated that VirD5 could compete with VBF for binding to AtVIP1 and protect AtVIP1 and VirE2 from degradation via the SCFVBF pathway.
Whether a virulence protein is induced by AS is very important for the characterization of its functions during the Agrobacterium infection process. In this study, we analyzed the VirD5 locus from the agropine-type Ti plasmid pTiBo542. Using reverse transcription polymerase chain reaction (RT-PCR), we found that VirD5 and VirD2 are present in one polycistronic mRNA (Fig. S6b) and that the expression of VirD5 could be induced by AS (Fig. S6c), indicating that it could be transcribed within the virD operon. This result was consistent with some previous studies of the VirD5 gene from an octopine-type Ti plasmid (Kalogeraki et al., 2000). However, it had also been reported that the expression of VirD5 from the nopaline-type Ti plasmid pTiC58 was independent of AS induction (Lin & Kado, 1993). Sequence analysis of the agropine-type Ti plasmid pTiBo542 showed that there is a 100 bp spacer region between the stop codon of VirD4 and the start codon of VirD5, a putative ribosome binding site (RBS) 8 bp upstream of the start codon of VirD5, and the -10 and -35 promoter sequences, which overlap the 3′ end of VirD4 (Fig. S6a). We cannot exclude the possibility that VirD5 is transcribed independently under some conditions.
One recent study revealed that Agrobacterium uses VirD5 to stabilize VirF in the host cells by directly interacting with VirF. However, VirF is only present in the octopine Ti plasmid (Melchers et al., 1990; Regensburg-Tuink & Hooykaas, 1993) and not in the nopaline and agropine-type Ti plasmids, whereas VirD5 is, to our knowledge, conserved in all strains of A. tumefaciens and Agrobacteriumrhizogenes (Fig. S2), suggesting that VirD5 may have other functions.
Through analysis of the deduced protein sequence, previous studies speculated that VirD5 might function as a transcription factor in the host cells (Schrammeijer et al., 2000; Vergunst et al., 2005). Here, our results demonstrated that VirD5 has the following characteristics: having transcriptional activation activity in yeast cells; being translocated into the host cell nucleus and interacting with importin-α; forming homodimers in living cells; and binding a specific DNA element (D5RE). Our results strongly suggested that VirD5 is an Agrobacterium transcriptional activator-like virulence protein. We have already found candidate host genes with at least two D5REs on their promoters (500 bp upstream of the translation initiation codon (ATG)) that may be regulated by VirD5 and involved in the Agrobacterium infection process. Their verification is in progress via further experiments in Arabidopsis, tobacco and rice.
We also demonstrated that VirD5 could interact with AtVIP1. The VirD5–AtVIP1–VirE2 ternary complex could be detected both in vivo and in vitro. Because VirE2 coats the T-DNA (Grange et al., 2008) to form a nucleoprotein complex (T-complex), and AtVIP1 interacts with VirE2 to facilitate nucleus-importing and host chromatin-targeting of the T-complex (Tzfira et al., 2001), we proposed that, in the plant cell nucleus, VirD5 could target the T-complex by interaction with AtVIP1. We found that, after VirD5 binds to AtVIP1, VBF cannot bind to AtVIP1. Because VBF is an E3 ubiquitin ligase and can elicit the degradation of target proteins, such as AtVIP1 and its interactor VirE2 (Zaltsman et al., 2010, 2013), we hypothesized that Agrobacterium evolved VirD5 to act as a protector of the T-complex to prevent the ‘coat proteins’ from being quickly degraded after importation into the cell nucleus but before attachment to the host chromatin. The results from a cell-free degradation assay supported this hypothesis.
How does Agrobacterium prevent the T-complex from degradation before it targets host chromatin for integration? Our results provided some insights into this process. Based on previous knowledge and our novel findings, we proposed a model for this process (Fig. S7). Many T-complexes could enter the host cell nucleus. However, most of the invading T-complexes are rapidly uncoated by the host SCFVBF pathway activated by Agrobacterium before they target host chromatin for integration. These T-DNA molecules may be expressed only transiently or subjected to degradation by the activity of intracellular enzymes. A small proportion of T-complexes may be bound by the translocated VirD5. VirD5 delays the uncoating process of the T-complex by competing with VBF for binding to AtVIP1. Once the T-complex has been attached to host chromatin, VirD5 is removed from the T-complex, VBF binds to AtVIP1 to disassemble the T-complex, and then the T-DNA is integrated into the host genome for stable expression. To some extent, this model could explain how a small proportion of the invading T-DNA molecules might be integrated and expressed stably, whereas most of them are expressed transiently (Zaltsman et al., 2013) and the expression of T-DNA is bimodal (Janssen & Gardner, 1990).
Based on analysis of an Agrobacterium VirD5 deletion mutant, some researchers reported that VirD5 was not absolutely essential for the infection process (Stachel & Nester, 1986; Porter et al., 1987; Koukolikova-Nicola et al., 1993; Lin & Kado, 1993; Kalogeraki et al., 2000). However, more recent studies revealed that lack of VirD5 significantly affected the tumorigenicity of Agrobacterium, suggesting that VirD5 is required for efficient Agrobacterium infection (Magori & Citovsky, 2011a). This contradiction may arise from the fact that VirD5, like VirE3 and VirF, is a host range factor, a type of protein that can play roles in optimizing the transformation process but is not absolutely essential for infection in some plants (Vergunst et al., 2005; Magori & Citovsky, 2011a). This speculation also suggests that Agrobacterium may hijack some host-cell factors that are functionally redundant with VirD5 or that some other unknown mechanisms are used by Agrobacterium to avoid more rapid uncoating of the T-complex. Such functional redundancy between virulence proteins and host factors has been demonstrated for VirE3 and AtVIP1 (Lacroix et al., 2005; Garcia-Rodriguez et al., 2006) and for VirF and VBF (Zaltsman et al., 2010, 2013). Therefore, further studies are required to elucidate the whole functions of VirD5, to discover the putative host cell factors that are functionally redundant with VirD5 and to reveal the detailed processes of chromatin-targeting and the integration of T-DNA.
The interactions between Agrobacterium and host cells are extensive. Agrobacterium-mediated genetic transformation is widely used in plant functional genomics studies. To date, however, this method is still impractical for many plants or inefficient for transformation with binary vectors containing large DNA fragments, such as binary bacterial artificial chromosome (BIBAC) clones. We have previously constructed a maize B73 BIBAC library and have been transforming rice plants using the BIBAC clones to generate transgenic rice populations containing large fragments of maize DNA (Shi et al., 2011; Wang et al., 2013). However, the transformation efficiency is much lower than that of the generally used binary vectors. Clearly, in-depth studies and molecular analysis of Agrobacterium–host cell interactions will not only increase our understanding of pathogen infection and host defense strategies, but will also enrich our ability to apply the Agrobacterium-mediated transformation method.
We thank Tingxiang Yan for providing 35S::Ghd7-CFP and 35S::Ghd7-RFP constructs and Professor Jian Xu for providing the PM999 vectors. We also thank Dr Stanton B Gelvin and ABRC for providing the pSAT series vectors, Dr Heribert Hirt for suggestions about RDSA, and Qin Hu (Central China Normal University) for help with experiments on Arabidopsis. This work was supported by the Natural Science Foundation of China (grant no. 30971748).