Intersubunit Complementation of Sugar Signal Transduction in VirA Heterodimers and Posttranslational Regulation of VirA Activity in Agrobacterium tumefaciens

Abstract

The VirA/VirG two-component regulatory system of Agrobacterium tumefaciens regulates expression of the virulence (vir) genes that control the infection process leading to crown gall tumor disease on susceptible plants. VirA, a membrane-bound homodimer, initiates vir gene induction by communicating the presence of molecular signals found at the site of a plant wound through phosphorylation of VirG. Inducing signals include phenols, monosaccharides, and acidic pH. While sugars are not essential for gene induction, their presence greatly increases vir gene expression when levels of the essential phenolic signal are low. Reception of the sugar signal depends on a direct interaction between ChvE, a sugar-binding protein, and VirA. Here we show that the sugar signal received in the periplasmic region of one subunit within a VirA heterodimer can enhance the kinase function of the second subunit. However, sugar enhancement of vir gene expression was vector dependent. virA alleles expressed from pSa-derived vectors inhibited signal transduction by endogenous VirA. Inhibition was conditional, depending on the induction medium and the virA allele tested. Moreover, constitutive expression of virG overcame the inhibitory effect of some but not all virA alleles, suggesting that there may be more than one inhibitory mechanism.

Agrobacterium tumefaciens is capable of transferring a defined fragment of its tumor-inducing (Ti) plasmid into a wide variety of dicotyledonous plants. The transferred DNA integrates into the plant genome, where expression of the genes leads to the accumulation of plant growth hormones and the formation of crown gall tumors. Expression of the vir genes required for this process is dependent on molecular signals that are present at the site of a plant wound (for a review, see reference 49). Phenolic molecules (e.g., acetosyringone) and acidic pH are essential for vir gene expression (34, 35, 44), whereas aldose monosaccharides (e.g., arabinose) enhance sensitivity to phenolic inducers but are not essential for vir gene induction (1, 33).

Expression of the virulence genes is regulated at the transcriptional level by the VirA/VirG two-component regulatory system (36). Induction signals are sensed by the VirA histidine kinase, which undergoes an autophosphorylation event on a conserved histidine (14, 26). Upon reception of the phenolic signal, the phosphate group is transferred from VirA to a conserved aspartate on VirG (16, 26). Phosphorylated VirG then stimulates transcription of the vir gene operons through binding conserved AT-rich sequences in their promoter regions (15).

VirA is arranged into distinct functional domains (Fig. 1) (9, 23, 43). Two transmembrane regions delineate a periplasmic domain at the N terminus of the protein. A direct interaction between the periplasmic domain and the sugar-binding ChvE protein is necessary for the sugar-enhanced response to low concentrations of phenolic inducers (32). Deletion of the periplasmic domain and point mutations in the region disrupt the effect of monosaccharides without eliminating response to phenolic inducers (4, 7, 9, 32, 33, 39).

FIG. 1.

Structural organization of VirA. The domains and functions of the VirA protein are represented as defined in reference 43. Two transmembrane regions represented by thick lines delimit the periplasmic domain. The sites of the E255L, and L332E mutations and the conserved histidine are indicated. G1/G2 indicates the region of the predicted ATP-binding site.

The cytoplasmic portion of VirA contains three additional regions referred to as the linker, kinase, and receiver domains. Helical configurations that are predicted to extend through the linker have been implicated as mechanistic structures that transmit signals from the site of reception into the kinase domain (39). As in other two-component systems, the VirA histidine kinase region contains an essential histidine (H474) that has been implicated as the site of autophosphorylation (14, 26). Downstream from H474 are additional conserved motifs, including G1 and G2, which are believed to form an ATP-binding site (31). VirA is somewhat atypical in having an extra “receiver” domain at its carboxyl end. The receiver domain includes some sequence similarity to the N-terminal region of VirG.

Cross-linking studies indicate that VirA forms membrane-bound homodimers (30). The VirA dimer appears to be a key structural element that is linked to at least two aspects of signal transduction. First, it has been demonstrated for several histidine kinases, including VirA, that autophosphorylation is an intersubunit reaction, whereby a phosphate group is transferred from the catalytic region on one dimer subunit to a conserved histidine on the second subunit (5, 6, 27, 37, 45). Second, the stabilizing influence of the leucine zipper has been used in fusions to VirA and Tar to demonstrate that a particular orientation at the linker interface between dimer subunits correlates with signal-transducing activity (11, 39).

Here we examined the question of intersubunit complementation of the sugar signal through coexpression of two mutant virA alleles. VirA_H474Q is a null mutant (14), but the unaltered state of its periplasmic region suggests that it should interact normally with ChvE. VirA_E255L is significantly impaired in its ability to augment vir gene expression in induction medium that contains sugar (4). However, the cytoplasmic portion of VirA_E255L is intact, and the protein is capable of responding to the phenolic signal, which requires only the linker and kinase domains of VirA (9). Phosphotransfer from VirA_E255L to VirG appears to occur normally at high concentrations of phenolic inducer. Thus, if the ChvE signal can be effectively received in the periplasmic region of VirA_H474Q, it might be integrated within the cytoplasmic portion of an H474Q-E255L heterodimer to enhance reception of the phenolic signal, thereby restoring the sugar enhancement lost with the E255L mutation.

The experiments presented here show that coexpression of virA_H474Q and virA_E255L restored sugar-enhanced sensitivity to vir gene expression. This contrasts with an earlier report (37) concluding that reception of the sugar signal requires that both subunits within a VirA dimer have wild-type periplasmic regions. However, the success of our heterodimer experiment depended on the plasmid used to express the virA alleles. When the virA alleles were carried on a pSa-derived vector, little or no sugar enhancement was observed. Yet vir gene expression was significantly increased by the presence of sugar if the same virA alleles were expressed from a pRi-derived vector. We examined the basis for the vector dependence of successful heterodimer function and found that cells that carried virA on a pSa-derived vector produced considerably more VirA protein than cells expressing the gene from the pRi-derived vector. We found that all virA alleles tested, including wild-type virA, expressed from our pSa-derived vectors inhibited vir gene expression in cells that carried wild-type virA on pTi. Thus, when present in excess, even wild-type VirA hindered the stimulation of vir gene expression that normally takes place under inducing conditions. The inhibitory effect of different virA alleles was influenced by the composition of the induction medium and, in some but not all cases, could be overcome with a constitutively expressed version of virG. Our findings suggest that overexpressed VirA inhibits vir gene expression by more than one mechanism.

MATERIALS AND METHODS

Bacterial strains and plasmids.

Table 1 lists the bacterial strains and plasmids used in this study. Escherichia coli DH5α was used for construction and amplification of the majority of plasmids. Escherichia coli XL1-Blue was used for plasmids that contained the P_N25-virG construct.

TABLE 1.

Strains and plasmids

Strain or plasmid	Characteristics	Reference or source
E. coli
XL1-Blue	recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac[F′ proAB lacIqZΔM15 Tn10(Tetr)]	Stratagene
DH5α	recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 ΔlacU169(φ80 lacZΔM15)	Laboratory stock
A. tumefaciens
A348	C58 background carrying pTiA6	13
A348-3	A348 with Kanr cartridge replacing a 4.6 kb KpnI fragment that contained virA on pTiA6	19
UIA143	autC58 recA::ery140 (pATC58)	12
A348-4	A348-3 recA::ery140	This study
Plasmids
pUC19	E. coli cloning vector	New England Biolabs
pAW17	pUC19 carrying virAH474Q at KpnI	This study
pAW18	pUC19 carrying virA at KpnI	This study
pAW19	pUC19 carrying Flag-tagged virA at KpnI	This study
pAW20	pUC19 carrying virAE255L at KpnI	This study
pAW21	pUC19 carrying virAL332E	This study
pYW10	IncW E. coli/Agrobacterium shuttle vector pUCD2-derived (pSa origin of replication)	(4, 10, 40)
pAW10	pYW10 with pUC19 cloning site at NcoI	This study
pAW13	pAW10 carrying virAE255L	This study
pAW14	pAW10 carrying virAH474Q	This study
pAW15	pAW10 carrying virAH474Q and virAE255L	This study
pAW16	pAW10 carrying wild type virA at KpnI	This study
pAW16a	pAW10 with Flag-tagged wild-type virA	This study
pAW50	E. coli/Agrobacterium shuttle vector with pRi-derived origin of replication	This study
pAW52	pAW50 carrying wild-type virA at KpnI	This study
pAW52a	pAW50 containing Flag-tagged virA at KpnI	This study
pAW53	pAW50 containing virAE255L at KpnI	This study
pAW54	pAW50 containing virAE255L and virAH474Q	This study
pAW55	pAW50 containing virAH474Q at KpnI	This study
pAW56	pAW50 containing virAL332E at KpnI	This study
pAW59	pAW10 containing virAL332E at KpnI	This study
pLV2	pAW10 containing virAE255L+H474Q at KpnI	This study
pYW47	PN25-virG at Ncol site of pYW10	40
pYW48	pYW47 with wild-type virA at KpnI	40
pAW58	pYW47 with virAH474Q at KpnI	This study
pAW60	pAW47 with virAL332E at KpnI	This study
pAW61	pYW47 with virAE255L at KpnI
pSW209	IncP plasmid carrying PvirB-lacZ fusion, Kanr	S. C. Winans
pSW209Ω	Derivative of pSW209 that is also Specr	40

Media and growth conditions.

E. coli strains were grown on Luria-Bertani (LB) medium. Agrobacterium spp. were grown on MG/L (7). A modified AB medium (44) was used for vir gene induction. AB medium (pH 5.5) included 20 mM 2-[N-morpholino]ethanesulfonic acid, 0.05% Casamino Acids, 1.2 mM phosphate, 0.001% thiamine, and either glycerol or arabinose at 0.2%. Acetosyringone was included as a vir gene inducer at the indicated amounts. Carbenicillin was added to liquid medium at 30 μg/ml to maintain plasmids containing a virA or virG allele. For maintenance of pSW209 and pSW209Ω, kanamycin or streptomycin was added to liquid medium at 10 and 25 μg/ml, respectively. Agrobacterium cells were grown at 25°C.

DNA methods.

Standard methods were used for plasmid isolation, restriction analysis, agarose gel electrophoresis, DNA ligation, and PCR amplification of DNA (3). Transformation of E. coli and Agrobacterium strains was done by electroporation with a Bio-Rad gene pulser as recommended by the manufacturer, but with resistance set to 400 Ω for Agrobacterium spp.

Plasmid construction.

All pUC19, pAW10, and pAW50 derivatives constructed for this study carry virA alleles on a 4.6-kb fragment originally derived from pTiA6 (19). The DNA fragments include the natural virA promoter and all known regulatory elements.

The Flag epitope tag was added to the carboxyl end of wild-type virA with a sequential PCR method. In two initial PCRs that used pAW18 as the template, primer EcoRV 3811-3834 (5′-GAAGCCGATATCGTCCAGAACTAT-3′) was used with primer Flag 3M (5′-AGGAATTGGTTTTAGTTCTGCCTGATGTTCCTGCTGCTACTGTTC-3′), while primer Flag 3P (5′-GACTACAAGGACGACGATGACAAGTAGAGTTGCGACGTGTCAGG-3′) was used with primer EcoRV 4397-4373 (5′-AAGCGGATATCAACTGCGTCGCAAT-3′) to amplify 80- and 550-bp fragments, respectively. These first PCR products were used as templates in a self-priming reaction to produce a 610-bp fragment that was used to replace a natural EcoRV fragment at the end of virA. The replacement added the Flag epitope sequence immediately before the virA stop codon.

The construction of pLV2 was done by replacing the N-terminal region (NcoI to BstEII) of DNA fragments carrying virA_H474Q in pAW17 with the N-terminal region of the fragment carrying virA_E255L (4) in pAW20. To complete pLV2, the KpnI fragment containing the double mutant was cloned into pAW10.

For construction of pAW50, a 4.6-kb HindIII-BamHI fragment containing a portion of the Agrobacterium rhizogenes root-inducing plasmid (pRi) that allows stable replication (28) was ligated to a 1.8-kb portion of pUC19 containing the bla gene and the ColE1 origin of replication. A 400-bp region of pUC19 that includes the lac promoter, the multicloning site, and the alpha portion of lacZ was added to complete pAW50.

pAW21 was constructed by sequential PCR. The first two PCRs used primers VirA 1705 (5′-CTCAACTGCTACGCCAGC-3′) with primer L332E (5′-CCACTAGAGCTTCCGCGCACG-3′) and primer L332E (5′-CGTGCGCGGAAGCTCTAGTGG-3′) with primer VirA 2850 (5′-TGTGCTAATTCTGCGTGCC-3′). The third PCR included the products of the first two reactions in a self-priming PCR that reconstructed the linker domain of virA but changed a leucine to glutamate at codon 332. This fragment was used to replace the wild-type linker domain of the virA gene in pAW19.

Construction of A348-4.

Genomic DNA obtained from UIA143 (recA::erm) (12) with the mini-genomic DNA method (3) was used to transform A348-3/pSW209Ω with selection for erythromycin resistance. The recA::Erm^r mutation in A348-4 was confirmed with PCR.

vir gene assays.

vir gene induction was analyzed by triplicate measurements of β-galactosidase activity produced from a virB-lacZ fusion contained on pSW209 (S. C. Winans, Cornell University) or pSW209Ω (40). Overnight MG/L cultures were used to inoculate AB induction medium containing acetosyringone. Induced cultures grew for 22 h. Samples were prepared and assayed by the method of Miller (24).

Southern blot analysis.

Genomic DNA was isolated from A348/pAW52 and A348/pAW16 by the mini-genomic DNA method (3). DNA was digested to completion with either EcoRV (A348/pAW52) or HincII (A348/pAW16) and electrophoresed through agarose gels before transfer to nylon membranes (3). A 1.1-kb DNA fragment internal to virA (nucleotides 1705 to 2850) was labeled with dioxigenin-11-UTP (Roche Diagnostics) and used as the probe in a stringent hybridization reaction. The CDP-Star system (Roche Diagnostics) and exposure to Hyperfilm (Amersham Biosciences) were used for signal detection. Labeled bands on portions of the film where signal was determined to be linear by DNA concentration as analyzed by ImageQuant software (Amersham Biosciences) and Storm 860 scanner (Molecular Dynamics) were used to determine plasmid copy number.

Immunoblot analysis.

Membrane proteins were prepared from A348-3, A348-3/pAW52a, and A348-3/pAW16a for visualization of Flag-tagged VirA through immunoblotting. Each strain was grown for 22 h in 300 ml of AB induction medium containing arabinose and 150 μM acetosyringone. Cells were precipitated and washed with 50 mM NaH₂PO₄ (pH 7.5). Cell pellets were resuspended in NaH₂PO₄ containing 0.2 μg of DNase I per ml, 0.2 μg of RNase per ml, and 0.01% phenylmethylsulfonyl fluoride and subjected to three passes through a French press at 14,000 lb/in². Lysozyme was added to 4 mg/ml, and cells were left on ice for 30 min. Cellular debris was precipitated, and 0.2 N KCl was added to the supernatant. Ultracentrifugation of the supernatant was at 270,000 x g for 20 min. Membrane pellets were resuspended in NaH₂PO₄ with 0.8% sodium dodecyl sulfate. The membrane protein and protein concentration was determined with the bicinchoninic acid assay (38). For immunoblotting, 0.5 μg of membrane protein per lane was separated on a Nupage 10% Bis-Tris denaturing gel (Invitrogen Life Technologies) and transferred to a polyvinylidene difluoride membrane. VirA was visualized through Western blot analysis with anti-Flag (Sigma) primary antibody, horseradish peroxidase-conjugated anti-rabbit immunoglobulin secondary antibody (Amersham Biosciences), and ECL Plus (Amersham Biosciences).

RESULTS

Coexpression of virA_H474Q and virA_E255L restored sugar enhancement to vir gene induction.

Intersubunit complementation of the ChvE-dependent sugar signal was examined by cloning two mutant versions of virA (virA_E255L and virA_H474Q) together into the pRi-derived vector pAW50 and introducing the resulting plasmid (pAW54) into the ΔvirA recA strain A348-4/pSW209Ω (virB-lacZ). For comparison with this heterodimer strain, the individual effects of the E255L and the H474Q mutations are shown in Fig. 2A. In agreement with previous findings (4), the E255L periplasmic mutation significantly reduced sugar enhancement of vir gene expression at 1 or 10 μM of the phenolic inducer acetosyringone. At 100 μM acetosyringone (AS), the signal-transducing activity of VirA_E255L was similar to that of wild-type VirA regardless of the presence of sugar. As expected, no vir gene expression was observed in cells containing virA_H474Q. Immunoblot analysis of an epitope-tagged version of virAH474Q confirmed its expression in cells that also carried a functional virA allele (data not shown.)

FIG. 2.

(A) The effect of sugar on vir gene expression in A348-3 (ΔvirA)/pS209Ω as mediated by wild-type VirA is compared with that determined by VirA_E255L. Cells that carry wild-type virA (pAW52) are indicated by squares; cells carrying virA_E255L (pAW53) are indicated by circles; and cells that carry virA_H474Q (pAW55) are indicated by an X. (B) Expression of vir genes is enhanced by arabinose when cells (A348-4/pSW209Ω) contain both virA_E255L and virA_H474Q. Cells carrying virA_E255L (pAW53) are represented as in A. Cells carrying both virA_E255L and virA_H474Q (pAW54) are represented by triangles. In both A and B solid symbols indicate the presence of arabinose in the induction medium, while open symbols indicate the presence of glycerol. Data shown are the averages of triplicate values. Standard deviations were calculated for all data points, although they are not always apparent on the graph.

In Fig. 2B, vir gene expression in A348-4/pSW209Ω, mediated by VirA_E255L is compared with that in cells that contain both VirA_E255L and VirA_H474Q expressed from pAW54. At low levels of phenol, vir gene expression was clearly elevated in cells that contained pAW54 if sugar was also present in the induction medium. Thus, sugar enhancement of vir gene expression was partially restored in the strain that carried both virA_H474Q and virA_E255L. Reconstitution of the sugar effect most likely depends on the formation of functional E255L-H474Q heterodimers.

Expression of virA_H474Q and virA_E255L from a pSa derivative limits vir gene expression.

Restoration of the ChvE-dependent sugar signal through intersubunit complementation was examined previously (37). The heterodimer experiments of Toyoda-Yamamoto et al. coexpressed the virA_E210V and virA_H474R mutant alleles from plasmids containing the pSa origin of replication. Sugar enhancement was not restored, and it was concluded that reception of the sugar signal requires VirA dimers with two wild-type periplasmic domains. We therefore assayed for restoration of the sugar signal through heterodimer formation with pSa-derived plasmids. pAW10 is maintained in Agrobacterium spp. through an origin of replication obtained from pSa (10). pAW15 is a derivative of pAW10 that carries both the virA_E255L and virA_H474Q alleles. A. tumefaciens carrying pAW15 exhibited significantly lower vir gene expression than cells that contained the same alleles on pAW54. Reduced vir gene expression was seen in medium containing arabinose or glycerol (Fig. 3). Negligible sugar enhancement was observed with pAW15.

FIG. 3.

Comparison of vir gene expression as mediated by pSa and pRi derivatives that each carry two mutant virA alleles in A348-4/pSW209Ω. Both virA_E255L and virA_H474Q are present together on pAW54 (indicated by circles) or pAW15 (indicated by squares). vir gene expression in induction medium with arabinose (solid symbols) is compared with that in induction medium containing glycerol (open symbols). Data shown are the averages of triplicate values. Standard deviations are indicated.

virA alleles carried on the pSa-derived vector alter vir gene expression mediated by pTi-encoded VirA.

Because intersubunit complementation of the sugar-dependent signal through coexpression of virA_E255L and virA_H474Q was shown to be vector dependent, wild-type and mutant virA alleles were individually cloned into the pSa-derived vector to be tested for their effect on vir gene expression in the wild-type background (Fig. 4). A348/pSW209 carries wild-type virA on pTi and a virB-lacZ reporter fusion. Surprisingly, the inclusion of wild-type virA (pAW16) reduced vir gene expression relative to that in cells containing the empty vector (pAW10). In induction medium with glycerol (Fig. 4A), virA_E255L (pAW13) or virA_H474Q (pAW14) further suppressed gene expression relative to that of cells containing wild-type virA at 30 μM acetosyringone. When the concentration of acetosyringone was increased to 150 μM, vir gene expression in cells containing virA_E255L (pAW13) was similar to that produced by cells carrying wild-type virA (pAW16). The increase in vir gene expression seen with pAW13 at the higher concentrations of phenolic inducer is noteworthy, as the signal-transducing activity of VirA_E255L relative to wild-type VirA is not defective under those conditions (Fig. 2A). Wild-type VirA expressed from pAW16 was noticeably more inhibitory in medium containing arabinose (Fig. 4B). However, virA_H474Q (pAW14) had the unexpected effect of allowing more vir gene expression in this medium than cells carrying pAW16. Apparently, the inhibition of native wild-type VirA by the null H474Q mutant was partly relieved by the presence of arabinose in the induction medium.

FIG. 4.

Effect of virA alleles expressed from pSa-derived plasmids on vir gene expression in the wild-type background of A348 compared in induction medium with glycerol or arabinose. pAW10 is the empty vector; pAW16 carries wild-type virA; pAW13 carries virA_E255L; and pAW14 contains virA_H474Q. Data shown are the averages of triplicate values. Standard deviations are indicated.

We hypothesized that the interaction between ChvE and the periplasmic region of VirA_H474Q altered the protein in a way that reversed the inhibitory effect seen in medium with glycerol. An allele of virA that carries both the E255L and the H474Q mutations was created to ask if impaired sugar perception due to the first mutation would alter the inhibitory effect of the mutated histidine. vir gene expression in the wild-type background in cells that carried pLV2 (virA_E255L+H474Q), pAW13 (virA_E255L), or pAW14 (virA_H474Q) is compared in Fig. 5. In induction medium with glycerol, the inhibitory effect of the double mutant was similar to that of the virA_H474Q mutant (Fig. 5A). In medium containing arabinose (Fig. 5B), cells that contained the virA_E255L+H474Q double mutant demonstrated an intermediate level of vir gene expression, higher than that of cells containing pAW13 (virA_E255L), but less than that of cells carrying virA_H474Q. Thus, when sugar perception was damaged by the E255L mutation, the H474Q mutation was less effective in relieving inhibition of vir gene expression.

FIG. 5.

Effect of the H474Q mutation on vir gene expression in the wild-type background of A348. vir gene expression is compared in cells that carry wild-type virA on pAW16, virA_E255L on pAW13, virA_H474Q on pAW14, or the double mutant virA_E255L+H474Q on pLV2. Data shown are the averages of triplicate values. Standard deviations are indicated.

The behavior of the two proteins with mutations at the conserved histidine (VirA_H474Q and VirA_E255L+H474Q) was distinct from that of VirA_E255L which was more inhibitory under conditions in which the protein is known to function poorly, i.e., at low concentrations of the phenolic inducer, regardless of the presence of sugar. We examined the effect of an additional virA mutant which, like VirA_E255L, has a conditional mutant phenotype and retains the conserved histidine. We found that VirA_L332E, which carries a mutation in the linker domain, confers an absolute requirement for sugar in the inducing medium. The effect of the L332E mutation carried on pAW50 in A348-3 (ΔvirA)/pSW209Ω is shown in Fig. 6A and B. In the absence of sugar, VirA_L332E demonstrated no signal-transducing activity. In fact, at acetosyringone concentrations as high as 500 μM, VirA_L332E was inactive unless the medium also contained sugar (data not shown). When arabinose was present (Fig. 6B), VirA_L332E was able to communicate the presence of the phenolic inducer to VirG, although with less efficiency than wild-type VirA.

FIG. 6.

(A and B) vir gene expression in A348-3 (ΔvirA) as mediated by wild-type virA (pAW52) is compared with that determined by virA_L332E (pAW56). (C and D) vir gene expression in the wild-type background of A348 when cells carry the pAW10 vector or virA_L332E on a derivative of that vector (pAW59). Data shown are the averages of triplicate values. Standard deviations are indicated.

The virA_L332E allele was cloned into pAW10 and placed into A348/pSW209 to observe the effect of the protein on gene expression mediated by native wild-type VirA (Fig. 6C and D). The negative effect of VirA_L332E was more severe when the induction medium did not contain arabinose, decreasing the activity of the indigenous wild-type VirA by approximately 75% compared to the strain with the empty vector. Thus, like VirA_E255L, VirA_L332E was a stronger inhibitor of wild-type VirA activity under conditions in which the protein is inactive. This contrasts with VirA_H474Q which, while inactive under all conditions, demonstrated the conditional inhibitory effect described above.

Inhibitory effect of the virA alleles is vector dependent.

It was important to ascertain whether the reduced level of vir gene expression seen with pAW15 relative to pAW54 was due to some property of the pSa-derived vector (pAW10) which did not involve the virA alleles. In Table 2, vir gene expression in the wild-type background is compared for cells carrying pAW10 and the pRi-derived pAW50 vector and those vectors carrying wild-type virA (pAW16 and pAW52, respectively). Gene expression levels were not appreciably different between cells that carried either of the empty vectors and cells that expressed wild-type virA from the pRi derivative. However, wild-type virA expressed from the pSa derivative (pAW16) reduced vir gene expression to less that 30% of that in cells containing pAW10. Thus, the pSa-derived vector itself is not causing inhibition of vir gene expression but must contain the virA allele. A possible reason for the vector-dependent effects of the virA alleles on vir gene expression is copy number effects. If the pSa-derived vector exists at a significantly different copy number than the pRi-derived vector, it might strongly affect expression of the virA gene that it carries.

TABLE 2.

Inhibitory effect of wild-type virA is vector dependent^a

Acetosyringone (μM)	Vector		+ wild-type virA
	pAW10	pAW50	pAW16	pAW52
1	1,935 (63)	2,062 (16)	465 (16)	1,936 (89)
10	2,548 (350)	2,735 (52)	763 (88)	2,434 (154)

^aCells were grown in AB induction medium with arabinose. Values are the averaged β-galactosidase activities in Miller units from three independently inoculated cultures with standard deviations indicated in parentheses.

pSa-derived pAW16 exists at higher copy number than pRi-derived pAW52 and significantly increases the cellular concentration of VirA.

Southern blot analysis was used to compare the copy number of plasmids pAW16 and pAW52 relative to pTi in A. tumefaciens A348 (Fig. 7A). A probe internal to the virA gene revealed bands of the sizes predicted from restriction maps of pAW16, pAW52, and the DNA sequence upstream from virA on pTi. From cells carrying pAW52, EcoRV restriction fragments of 5.7 kb (from pTi) and 8.5 kb (from pAW52) were predicted and identified. Similarly, HincII restriction fragments of approximately 4.8 kb (from pTi) and 8.5 kb (from pAW16) were predicted and observed from cells carrying pAW16. Analysis with ImageQuant software revealed that pAW16 was present at between 15 and 20 copies per pTi, while pAW52 was present at between four and six copies per pTi.

FIG. 7.

(A) Southern blot analysis of virA copy number as present on pAW16 or pAW52 in the wild-type background of A348. Each gel was loaded with 0.5 μg of DNA (lane 1) and 1 μg of DNA (lane 2). virA on a plasmid and pTi are indicated. virA on pTi served as an internal control. (B) Immunoblot analysis of Flag-tagged VirA in A348-3 (ΔvirA) carrying either pAW16a (lane 1), pAW52a (lane 2), or no plasmid (lane 3). Each lane was loaded with 0.5 μg of membrane proteins. The position of VirA is indicated. A band corresponding in size to dimerized VirA is indicated with an arrow. A nonspecific band (X) served as an internal control.

The higher copy number of pSa-derived pAW16 relative to pRi-derived pAW52 suggested that cells carrying pAW16 might produce more VirA than cells carrying pAW52. Immunoblot analysis (Fig. 7B) was used to compare the amount of VirA produced in A348-3 (ΔvirA) cells that carried a Flag-tagged virA allele on the pSa-derived vector with VirA produced by cells carrying the same gene on the pRi derivative. (Our unpublished results indicate that a Flag tag at the carboxyl terminus of VirA does not significantly alter its capacity to support vir gene expression.) The cellular content of the plasmid-encoded protein increased dramatically when it was carried on pAW16a (pAW10 with virA_Flag) rather than pAW52a (pAW50 with virA_Flag). We also noted a band of the size predicted for dimerized VirA in the lane loaded with protein from A348-3/pAW16a. The fragment carrying the virA_Flag allele is the same on both plasmids. Thus, the increase in protein must be due to the higher copy number of the pSa-derived plasmid.

Constitutively expressed virG overcomes the inhibitory effect of overexpressed wild-type and L332E virA alleles on vir gene expression.

The inhibition of vir gene expression caused by overexpression of wild-type virA in the wild-type background (Table 2 and Fig. 7) is consistent with the concept that increasing the amount of VirA in the cell reduces the availability of some factor that is essential for vir gene expression. Although vir gene expression is affected by several proteins, VirA and VirG are the only regulatory proteins that are known to be essential for vir gene expression. Furthermore, VirA and VirG regulate their own and each other's expression (44). Thus, normal expression of the virulence genes might require the two regulatory proteins to be present in stoichiometric amounts. To study the problem, we analyzed vir gene expression in cells that carried a constitutively expressed virG allele on a pSa-derived plasmid together with a wild-type or mutant version of virA.

pYW47 carries a P_N25-virG fusion that is constitutively expressed in Agrobacterium spp. (40). The vector portion of the pYW47 is exactly the same as that of pAW10. Thus, the pAW10 derivatives and pYW47 derivatives can be reasonably expected to have the same copy number. In addition, P_N25 (derived from bacteriophage T5) decouples expression of virG from VirA activity. However, in the presence of P_N25-virG, expression of the virulence genes, including virA, still requires a phenolic inducer and the signal-transducing activity of VirA (26). Mutant and wild-type virA alleles were cloned into pYW47 and introduced into the wild-type background. We analyzed vir gene expression in A348/pSW209 cells that carried the constitutively expressed virG allele, with and without virA. The addition of P_N25-virG (pYW47) by itself resulted in a modest increase in vir gene expression, relative to that seen in cells containing the vector (pAW10), regardless of whether the induction medium contained glycerol or arabinose (Table 3 and 4).

TABLE 3.

Ability of constitutively expressed virG to relieve inhibition of vir gene expression by virA alleles in induction medium containing glycerol^a

Plasmidsb	virA on vector	Acetosyringone (μM)	β-Galactosidase (Miller units)		Ratio with/without PN25-virG
			Without PN25-virG	With PN25-virG
pAW10, pYW47	None	10	58	94	1.6
		30	601	814	1.4
		150	1,211	1,445	1.2
pAW14, pAW58	virAH474Q	10	31	59	1.9
		30	145	301	2.1
		150	268	507	1.9
pAW59, pAW60	virAL332E	10	14	54	3.9
		30	96	478	5.0
		150	192	1,126	5.9

^aValues are the averaged β-galactosidase activity in Miller units from three independently inoculated cultures.

^bWhen present, the virA allele is the same on both plasmids. The second plasmid listed also contains P_N25-virG.

TABLE 4.

Ability of constitutively expressed virG to relieve inhibition of vir gene expression by virA alleles in induction medium containing arabinose^a

Plasmids	virA on vector	Acetosyringone (μM)	β-Galactosidase (Miller units)		Ratio with/without PN25-virG
			Without PN25-virG	With PN25-virG
pAW10, pYW47	None	1	1,478	1,936	1.3
		10	2,739	2,955	1.1
		100	2,962	4,313	1.5
pAW16, pYW48	Wild-type virA	1	531	752	1.4
		10	458	1,246	2.7
		100	1,018	3,207	3.2
pAW13, pAW20	virAE255L	1	119	155	1.3
		10	388	778	2.0

^aSee Table 3, footnotes a and b.

As virA_H474Q and virA_L332E were most inhibitory in induction medium containing glycerol (Fig. 4 and 6), we used that medium to examine the effect of including the P_N25-virG construct on the plasmid for these two alleles (Table 3). For virA_H474Q, adding virG increased vir gene expression approximately twofold, but it was not restored to the level seen in cells that carried the empty vector (pAW10) at 30 and 150 μM acetosyringone. Inhibition caused by virA_L332E was even more severe than that seen with virA_H474Q. The addition of P_N25-virG increased vir gene expression between three and sixfold in cells that also contained virA_L332E on the plasmid, reaching levels that were 80 to 90% of those observed in cells with pAW10. In contrast to the virA_H474Q and virA_L332E alleles, the inhibitory effects observed from overexpression of wild-type virA or virA_E255L were strongest when the induction medium contained arabinose. Therefore, for these two alleles, we examined the effect of constitutive expression of virG in medium with arabinose (Table 4). The addition of P_N25-virG to the plasmid countered the negative effect of wild-type virA, resulting in vir gene expression levels similar to that in cells carrying pAW10 at 100 μM acetosyringone. Expression in cells carrying virA_E255L also increased but remained well below uninhibited levels.

DISCUSSION

VirA is unusual in that maximization of its signal-transducing activity depends on the simultaneous presence of several different signals: sugar, phenolic molecules, acidic pH, and low phosphate (7, 33, 34, 35, 44). The fact that the essential phenolic signal is received through a cytoplasmic portion of the protein while the auxiliary sugar signal is received in the periplasmic domain (9, 23, 32) allowed us to use the phenomenon of intersubunit transphosphorylation to observe the effect of a signal received in the periplasmic region of one subunit on signal reception in the cytoplasmic domain of the second subunit. Specifically, coexpression of virA_E255L and virA_H474Q in A. tumefaciens lacking wild-type virA provided for sugar-enhanced vir gene expression.

Our results support a model in which E255L-H474Q heterodimers form to mediate reception and transduction of the sugar signal. The interaction between the wild-type periplasmic region of the H474Q subunit and the sugar-binding ChvE protein probably results in distortion of the E255L-H474Q heterodimer. The distortion could then be transferred from the periplasm through the transmembrane region into the linker domain. The dimer interface would thereby be altered, affecting the orientation of the kinase region in a manner that facilitates phosphotransfer to VirG from the wild-type kinase region of the E255L subunit. Leucine zipper fusions to the cytoplasmic portion of VirA indicate that the subunit interface within a VirA dimer is correlated with the protein's signal-transducing activity (39). Our model is consistent with those experiments and current models describing the structure-function relationships of signal-transducing proteins (2, 42).

Our inability to see sugar-enhanced vir gene expression with the pSa derivative pAW15 was probably due to the accumulation of homodimeric forms of the mutant proteins to a level that somehow interferes with the sugar-enhanced activity of otherwise functional heterodimers. We found that the copy number of the pSa-derived plasmid was severalfold greater than that of pTi in the same cell. Moreover, the higher plasmid copy number resulted in huge overproduction of the VirA protein (Fig. 7). Pan et al. (30) found that, in an otherwise wild-type strain of A. tumefaciens, virA_H474Q transcribed from an isopropylthiogalactopyranoside (IPTG)-inducible promoter carried on an IncP vector resulted in faster accumulation of the mutant protein relative to wild-type VirA encoded on pTi. Similar to our results, the increase in mutant protein resulted in the inhibition of vir gene expression. Our results and those of Pan et al. (30) are in agreement with the concept that large amounts of inactive VirA homodimers prevent activation of VirG, thereby reducing the cell's capacity to express the vir genes. The same factor, mutant homodimers, may also have prevented sugar-enhanced vir gene expression mediated by the mutant alleles encoded on pRi-derived pAW54 from reaching the same levels seen with wild-type virA (Fig. 2B). Alternatively, or additionally, the interaction between ChvE and the E255L-H474Q heterodimers may be less efficient than with wild-type homodimers.

Our analysis of the negative effect of VirA and the VirA mutants reveals two additional points. First, there are signal-specific effects on the inhibitory activity of different versions of the VirA protein. Wild-type virA carried on the pAW10 vector is significantly more inhibitory compared to the vector alone when sugar is present in the medium (Fig. 4). The extent of inhibitory activity of mutant alleles also changed depending on the inducing conditions. VirA_E255L and VirA_L332E were each more inhibitory in the tests carried out under inducing conditions, with arabinose and glycerol, respectively (Tables 3 and 4), in which these alleles by themselves had little or no capacity to induce vir gene expression (Fig. 2A for E255L and Fig. 6 for L332E).

A further demonstration of signal-dependent negative effects was observed in the case of the null virA_H474Q mutant. This allele was more inhibitory than overexpressed wild-type virA, but only if the induction medium did not contain sugar (Fig. 4). The sugar-dependent relief of inhibition suggests that VirA_H474Q changes from inhibitory to less so when it interacts with the ChvE sugar-binding protein. Analysis of cells carrying the VirA_E255L+H474Q double mutant supports this model. The double mutant remained more inhibitory than the H474Q allele at low acetosyringone levels, where the sugar effect, in part mediated by the E255L residue, is most necessary for vir gene expression (Fig. 5). When tested in arabinose, the effect of the virA_E225L+H474Q double mutant was intermediate between that of virA_E255L and virA_H474Q. Thus, in interfering with reception of the sugar signal, the E255L periplasmic mutation also interfered with the relief of inhibition that was observed with VirA_H474Q.

The second important observation regarding the negative effects of overexpressed VirA is that suppression of this phenotype through constitutive expression of VirG varies markedly as a function of the virA allele tested. While the inhibitory effect of excess wild-type VirA or VirA_L332E was substantially overcome by the addition of P_N25-virG to the virA-containing plasmid, vir gene expression was only moderately increased in cells that carried P_N25-virG along with virA_H474Q and virA_E255L. In particular, constitutive expression of virG was inefficient at overcoming the inhibition caused by the virA_E255L allele. The observation that inhibition by wild-type VirA and VirA_L332E is efficiently corrected while inhibition by VirA_E255L is not is in accord with our inference that the latter protein (at low concentrations of acetosyringone) inhibits vir gene expression by a mechanism distinct from that of wild-type VirA or VirA_L332E.

At least two models may be considered to explain the inhibition of vir gene expression that occurs when virA is overexpressed. An obvious consideration is phosphatase activity. Phosphatase activity has been demonstrated to be a possible negative factor in several two-component regulatory systems (18, 25, 46, 48). Our results do not argue against phosphatase activity as a source of the inhibitory effect seen when virA alleles are overexpressed. However, an investigation of phosphatase activity in the VirA/VirG system found that phosphorylated VirG is unusually stable in the presence of VirA (15). Furthermore, if phosphatase activity alone was responsible for the inhibitory effect of overexpressed virA alleles, it is difficult to understand why the P_N25-virG construct does not correct inhibition by VirA_H474Q and VirA_E255L as efficiently as for wild-type VirA or VirA_L332E. Moreover, VirG activity is fundamentally different from that of some response regulators because vir gene expression is tightly controlled and, even in the presence of constitutively expressed VirG, absolutely requires VirA and a phenolic inducer (14, 26). Thus, regulation of VirG activity may differ from that in systems that are regulated by transitions between high and low levels of phosphorylated response regulators.

In our experiments, inhibition of vir gene expression was clearly associated with a large increase in VirA protein. An interaction between a response regulator and overabundant inactive histidine kinase may have an inhibitory effect, particularly when accumulation of the response regulator is (like VirG) autoregulated. Complex formation would reduce the availability of phosphorylated response regulator in the cell by preventing the upregulation of its expression. If this is the case, one need not postulate an enzymatic cause, such as phosphatase activity, for the inhibitory effect seen when virA alleles are overexpressed.

A growing body of evidence supports the concept that a histidine kinase can form a stable complex with its cognate response regulator if it is unable to activate it through phosphorylation (8, 17, 20, 21, 22, 29, 46, 47). In the CitA/CitB two-component system of Klebsiella pneumoniae, an inactive version of the response regulator (CitA_D54E) can be efficiently copurified with a MalE fusion to the CitB histidine kinase if ATP is present (17). Complex formation between the wild-type version of CitA and MalE-CitB (which retains kinase activity) is much less stable. Furthermore, in the UhpA/UhpB two-component regulatory system of E. coli, inactive versions of the UhpB histidine kinase have been shown with in vivo and in vitro assays to inhibit UhpA-dependent transcription (47). However, transcription is not inhibited by constitutively active versions of UhpB (46, 47). Additionally, in the PhoP/PhoQ system, overexpression of the PhoQ histidine kinase is inhibitory under conditions that do not favor the phosphatase activity of that protein (8). Finally, the yeast two-hybrid system has been used to demonstrate a stable interaction between the phosphotransfer segment of the NtrB histidine kinase and its cognate response regulator, NtrC (21). This interaction does not occur when the NtrB fragment contains a mutation that normally renders the full-length protein constitutively active (22).

Our observation that overexpression of wild-type virA decreased vir gene expression below levels seen in A348 cells carrying the empty vector indicates that an excess of wild-type VirA interferes with VirG activity. As this manuscript was being prepared, a similar result was reported by Brencic et al. (5). The autoregulatory nature of VirA and VirG suggests that normal vir gene expression might require these two proteins to be present in set proportions. Furthermore, vir gene expression increased modestly when pYW47 (P_N25-virG) was present in A348 (Tables 3 and 4). In addition, constitutive expression of virG substantially corrected vir gene inhibition caused by overexpression of wild-type virA (Table 3). We conclude that VirG is a limiting factor for vir gene expression that becomes even less available if the cellular content of VirA is significantly increased. An unresolved question derived from our experimental results is why overexpressed wild-type VirA remains inactive in the presence of inducer. Perhaps there are yet other unidentified factors that can become limiting in the presence of excess VirA.