Preformed dimeric state of the sensor protein VirA is involved in plant ...

5 downloads 0 Views 1MB Size Report
Communicated by Earl W. Davie, June 30, 1993. ABSTRACT. Plant signal ..... vided by C. W. Jones, University of Leicester, Leicester,. U.K.) showed that ChvE ...
Proc. Natl. Acad. Sci. USA

Vol. 90, pp. 9939-9943, November 1993 Microbiology

Preformed dimeric state of the sensor protein VirA is involved in plant-Agrobacterium signal transduction SHEN Q. PAN, TREVOR CHARLES*, SHOUGUANG JIN, ZHI-LIANG WUt, AND EUGENE W. NESTER Department of Microbiology, University of Washington, Seattle, WA 98195

Communicated by Earl W. Davie, June 30, 1993

to Asp-52 of VirG (8). The phosphorylated VirG binds to upstream regions of each of the vir genes (vir box) to activate transcription by an unknown mechanism (9). Alteration of the VirA phosphorylation site from His-474 to Gln-474 results in a nonphosphorylatable VirA and abolishes vir gene induction by AS, suggesting that the phosphorylation activity of VirA is essential for its in vivo function (6). However, neither AS nor monosaccharide was necessary for the phosphorylation of the VirA derivative proteins in vitro. How VirA is involved in recognizing AS or how the autophosphorylation of VirA is triggered is still not known. To gain some insight into the molecular mechanism of the AS-VirA signal transduction pathway, we have initiated in vivo studies on any possible involvement of a VirA aggregate state in the process of signal recognition, signal transmission, or autophosphorylation. Here we present physical evidence for the existence of a VirA homodimer in intact A. tumefaciens cells, as well as genetic data suggesting that the homodimer is the functional state involved in AS-VirA signal transduction in vivo. To our knowledge, this is the first report that a bacterial membrane-bound sensor kinase is present as a homodimer and functions in a dimeric state in vivo.

Plant signal molecules such as acetosyringone ABSTRACT and certain monosaccharides induce the expression of Agrobaeterium tumefaciens virulence (vir) genes, which are required for the processing, transfer, and possibly integration of a piece of the bacterial plasmid DNA (T-DNA) into the plant genome. Two of the vir genes, virA and virG, belonging to the bacterial two-component regulatory system family, control the induction of vir genes by plant signals. virA encodes a membrane-bound sensor kinase protein and virG encodes a cytoplasmic regulator protein. Although it is well established from in vitro studies that the signal transduction process involves VirA autophosphorylation and subsequent phosphate transfer to VirG, the structural state of the VirA protein involved in signal transduction is not understood. In this communication, we describe an in vivo crosslinking approach which provides physical evidence that VirA exists as a homodimer in its native configuration. The dimerization of VirA neither requires nor is stimulated by the plant signal molecule acetosyringone. We also present genetic data which support the hypothesis that VirA exists as a homodinier which is the functional state transducing the plant signal in an intersubunit mechanism. To our knowledge, this report provides the first evidence that a bacterial membranebound sensor kinase exists and functions as a homodimer in vivo.

MATERIALS AND METHODS Strains and Plasmids. A. tumefaciens A348 is a derivative of A136, which harbors the octopine-catabolizing Ti plasmid pTiA6NC. A136 is a derivative of strain C58 lacking the nopaline Ti plasmid pTiC58. A1030 is a TnS insertional virA mutant of A723 (10). MX358 and MX368 are Tn3HoHol insertional mutants of A348 containing virE-lacZ and virBlacZ fusions, respectively (11). At11023 is a derivative of MX358 containing a TnS insertion at virA, which was constructed by electroporating A1016 (10) genomic DNA into MX358. The plasmid DNA of pSW191B containing the virA gene driven by the lac promoter (5) was cleaved with Pvu II and the fragment containing the virA gene was ligated into pSW213, a TcrIncP plasmid containing laclq (12), that had

Agrobacterium tumefaciens induces crown gall tumors in a wide variety of plants by transferring a piece of DNA (T-DNA) of its tumor-inducing (Ti) plasmid into plant cells, where the T-DNA becomes integrated into the plant chromosome. A region on the Ti plasmid, termed the virulence (vir) region, codes for enzymes and structures required for processing, subsequent transfer, and possibly integration of the T-DNA into the plant genome. All of the vir operons are induced as a regulon by plant phenolic compounds, such as acetosyringone (AS), and specific monosaccharides. AS and monosaccharides synergistically induce vir gene expression. Mutational analysis of the vir region has demonstrated that virA and virG control the induction of the vir genes by plant signal compounds. Sequence analysis of virA and virG indicates that the VirA and VirG proteins are homologous to members of a family of bacterial two-component regulatory systems involved in sensing specific environmental stimuli. A chromosomal virulence gene, chvE, is required for vir gene induction by monosaccharides. ChvE is homologous to the periplasmic sugar-binding proteins in Escherichia coli and interacts with the periplasmic domain of VirA. (For reviews, see refs. 1-3.) VirA is a transmembrane protein (4, 5) that presumably interacts with the plant phenolic signal compounds. Purified VirA derivatives can be autophosphorylated in vitro with the y-phosphate group of ATP (6, 7). The site of phosphorylation was determined to be a highly conserved histine residue (presumably His-474) (6). This phosphate is then transferred

been cut with EcoRI and blunt-ended with the Klenow

fragment of DNA polymerase I. The resulting plasmid, pTC115, contains lacIq and the virA gene driven by the lac promoter. pTC158 containing the mutant virA(H/Q) was derived from pTC115 by replacing the virA internal Sst I fragment with the pRS0401 (6) virA internal Sst I fragment. pSQ7, containing virAA605/829, and pSQ8, containing virA(H/Q)A605/829, were generated from pTC115 and pTC158, respectively, by cleaving with HindIII, blunt-ending with the Klenow fragment of DNA polymerase I, and then ligating with the fl fragment containing the transcription and translation termination signals that was obtained from Abbreviations: AS, acetosyringone; BS3, bis(sulfosuccinimidyl) suberate; IPTG, isopropyl ,-D-thiogalactopyranoside. *Present address: Microbiology Group, Department of Natural Resource Sciences, McGill University, 21,111 Lakeshore, Ste-Annede-Bellevue, Quebec, Canada H9X 3V9. tPresent address: Department of Pharmacology, University of Washington, Seattle, WA 98195.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

9939

Microbiology: Pan et al.

9940

Proc. Natl. Acad. Sci. USA 90 (1993)

pHP45QC (13) by Sma I digestion and gel purification. pIB415 is a derivative of pIB410 (of IncP replicon) containing virAA63/240, a virA derivative lacking 178 amino acids in the periplasmic domain (14). The plasmids were introduced into A. tumefaciens strains by electroporation (15) or triparental mating (16). Induction of vir Genes. A. tumefaciens strains were cultured overnight in MG/L (17) supplemented with appropriate antibiotics. The cultures were centrifuged, resuspended in distilled water, and inoculated into induction medium supplemented with appropriate antibiotics and AS if required. Induction medium contained 50 mM morpholineethanesulfonic acid (Mes, pH 5.5), 1 x AB salts (17), 0.02% yeast extract, 0.5% (vol/vol) glycerol, and 10 mM arabinose. After incubation for 16 hr at 28°C, the ,B-galactosidase activity was assayed by the method described by Miller (18) as modified by Stachel et al. (19). Crosslinking. The A. tumefaciens cells grown in induction medium as described above were washed by centrifugation once in distilled water and then three times in phosphatebuffered saline. Unless stated otherwise, the washed cell pellet was suspended in phosphate-buffered saline with 1 mM bis(sulfosuccinimidyl) suberate (BS3) and incubated on ice for 30 min. After centrifugation for 5 min at 4°C in a microcentrifuge, the cell pellet was resuspended in quenching buffer (10 mM Tris HCl, pH 8.0/0.1 M glycerol) and then incubated on ice for 20 min. The cell suspension was centrifuged for 5 min at 4°C and the pellet was suspended in 5 volumes of 1 x Laemmli sample buffer (20). The cell suspension was heated for 5 min at 100°C, cooled briefly on ice, and then centrifuged for 5 min at room temperature. Proteins in the supernatant were separated by SDS/PAGE and analyzed by Western blot. Western Blot. SDS/PAGE was conducted in 7.5% polyacrylamide gels. The proteins were transferred onto Immobilon-P membranes (Millipore) and the VirA- or VirB-pgalactosidase fusion protein was visualized with the enhanced chemiluminescence (ECL) Western blot detection system according to the recommendations of the manufacturer (Amersham). The polyclonal antibody to VirA was generated in mice with a truncated VirA protein, VirA681 (6); the monoclonal antibody to ,3galactosidase was purchased from Sigma.

RESULTS Physical Evidence for a Dimeric State of VirA in Intact Cells. To determine the functional state of VirA, we have used an

Periplasmic region

TM1 cm

I,

r-

00 a

11 l

TM2

eI

C4

1--

in

--I

00W d"

-

~

~

~l ~ ~ ~

0%

IL

iivE

Tm1

a, 0s m 00 00O0%

XMe

in vivo approach in which the VirA protein in intact cells was crosslinked by a homobifunctional crosslinking reagent, BS3. BS3 is a water-soluble crosslinker which covalently links adjacent primary amine functions of e amine groups on lysine or available N-terminal amines. The crosslinking bond cannot be cleaved by any reducing reagent. Because of its high water solubility, BS3 cannot penetrate membranes and thus crosslinks only those lysine residues present outside the cytoplasmic membranes. Therefore, BS3 is an ideal reagent to study the native state ofthe VirA protein, which has a large periplasmic domain containing seven lysine residues (4, 5, 21) (Fig. 1). The use of BS3 provided evidence that VirA is present in a multimeric complex. When intact cells of A. tumefaciens strain A348 were crosslinked by 1 mM BS3, four protein bands with apparent molecular masses of 205, 210, 215, and 222 kDa appeared in addition to the 92-kDa VirA monomer (Fig. 2, lanes 6-8). These bands were visualized by Western blot using VirA antibody. When no BS3 was added, only the 92-kDa band was present (Fig. 2, lanes 2-4). All the bands were absent in a virA null mutant (Fig. 2, lanes 1 and 5). These data indicate that the VirA is a component of each of the four bands, and suggest that VirA is present as an aggregate in its native state. The fact that four bands were visualized may be explained as follows. There are seven lysine residues in the periplasmic domain that can be crosslinked, and the native conformation of VirA at the periplasmic domain may have rendered only four different crosslinking positions. Each of the products of crosslinking at the different positions has a different electrophoretic mobility in the gel. In support of this notion, other investigators have also observed that dimers of two identical molecules that are crosslinked at different positions have different mobilities in gels (22). If this explanation is correct, a deletion of the periplasmic region should reduce the number of bands. This was verified experimentally. After crosslinking with BS3, the mutant VirA protein VirAA63/240, which lacks much of the periplasmic domain (Fig. 1) but still responds to AS (14), yielded only two bands (Fig. 3, lane 1), suggesting two crosslinking sites. However, VirAA63/240 has only one lysine residue in the periplasmic domain (Fig. 1). Another crosslinking site probably comes from the N-terminal amine or either of the two lysine residues (Lys-13 and Lys-17) near the N terminus (Fig. 1). It was previously observed that cells with a TnphoA insertion at amino acid position 22 of VirA exhibited a low level of alkaline phosphatase activity (5), suggesting that the N terminus of VirA might also be present in the periplasm and therefore available for crosslinking. Also, a higher concen-

Kinase domain H474

WT

(pTC1 15)

VirA(H/Q)

(pTC158)

Q474

]-4

1

H474

VirAA605/829 c__

I

,

I I

-14

II

) -4

C eq

(pSQ7)

--

W eJ

I I%

.

Q474 _~~~~~~~~~~~~~... ...

VirA(H/Q)A605/829 (pSQ8)

H474

VirAA63/240

(pIB415)

FIG. 1. Schematic presentation of wild-type (WT) and mutant VirA proteins. Solid boxes represent the two transmembrane regions (TM1 and TM2). The stippled bar indicates the protein kinase domain. The vertical lines represent lysine residues that can possibly be crosslinked by BS3; the numbers indicate the positions of lysine residues. The wild-type virA and mutant virA(H/Q), virAA605/829, and virA(H/Q)A605/829 genes were driven by the lac promoter; virAA63/340 was driven by the native promoter.

s~ ~ ~-92

Microbiology: Pan et al.

kDa

^,_ ,222

*C=

-215

a _ 210 205

1

3

2

4

5

6

7

8

FIG. 2. Protein blot analysis of wild-type VirA protein. A. tumefaciens strains A1030 (lanes 1 and 5), A348 (lanes 2 and 6), and A136(pTC115) (lanes 3, 4, 7, and 8) were grown in induction medium supplemented with 1 mM isopropyl ,B-D-thiogalactopyranoside (IPTG) (lanes 3, 4, 7, and 8) in the presence of 100 ,uM AS (lanes 1, 2, 4, 5, 6, and 8) or absence of AS (lanes 3 and 7) for 16 hr at 28°C. The bacterial cells were collected, washed, and divided into two parts. One part was crosslinked with 1 mM BS3 (lanes 5-8); another part was not (lanes 1-4). The cell lysates were electrophoresed in SDS/7.5% polyacrylamide gels. The proteins were transferred onto Immobilon-P membranes and the VirA protein was visualized by Western blot with VirA antibody. The apparent molecular masses were estimated based on protein standards of 221, 106, 75, and 46 kDa (GIBCO/BRL).

tration of BS3 (10 mM) was required to visualize these two bands, probably due to a much lower efficiency of crosslinking for the small periplasmic domain of VirAA63/240. The crosslinking experiments with the mutant VirA also provided evidence that VirA exists as a homodimer. The

apparent molecular

masses

of the four bands of crosslinked

wild-type VirA were lower than the expected molecular mass of a VirA trimer but slightly higher than that of a VirA homodimer (Fig. 2, lanes 6-8). One possibility is that a VirA molecule was dimerizing with other unidentified protein(s) of at least 113 kDa (heterodimer or heteromultimer). However, crosslinking of the 71-kDa mutant protein VirAA63/240 revealed two bands at 147 and 156 kDa (Fig. 3, lane 1), lower than the expected molecular mass of a heterodimer or hetkDa

222

_ 2~~215 __210 205 kDa

156147-

-92

71-_~

1

2

_ 3

4

5

FIG. 3. Protein blot analysis of wild-type and mutant VirA protein crosslinked by several concentrations of BS3. A. tumefaciens strains A1030(pIB415) (lane 1), A1030 (lane 2), and A348 (lanes 3, 4, and 5) were grown in induction medium in the presence of 100 '"M AS for 16 hr at 28°C. The washed cells were crosslinked by 0.1 mM (lane 3), 1 mM (lane 4), or 10 mM BS3 (lanes 1, 2, and 5). The cell lysates were electrophoresed in SDS/7.5% polyacrylamide gels. The proteins were transferred onto Immobilon-P membranes and the VirA protein was visualized by Western blot with VirA antibody. The apparent molecular masses were estimated based on protein standards of 221, 106, 75, and 46 kDa (GIBCO/BRL).

Proc. Natl. Acad. Sci. USA 90 (1993)

9941

eromultimer (at least 113 kDa + 71 kDa = 184 kDa) but roughly double that of the 71-kDa monomer, suggesting a homodimer. The apparent molecular sizes of crosslinked VirA dimer molecules were slightly higher than expected, presumably because crosslinking altered the electrophoretic mobilities. Since previous genetic data suggested that ChvE interacts with the periplasmic domain of VirA (14), we suspected that the crosslinked products might contain the ChvE protein. However, Western blot analysis with a ChvE antibody (provided by C. W. Jones, University of Leicester, Leicester, U.K.) showed that ChvE was not part of the crosslinked protein products (data not shown), presumably due to a lack of adjacent lysine residues present in the ChvE-VirA interaction site(s) and available for crosslinking. To estimate the proportion of VirA present in the dimeric state, the VirA protein in intact cells was crosslinked with several different concentrations of BS3. As the BS3 concentrations increased, higher levels of VirA protein present in the dimeric state were visualized (Fig. 3, lanes 3 and 4). When 10 mM BS3 was used, a smeared band appeared (Fig. 3, lane 5), suggesting nonspecific crosslinking. It appeared that 1 mM BS3 was the optimum concentration for crosslinking. Although it is difficult to precisely estimate how much VirA is present in the dimeric state because of the crosslinking efficiency involved, the data clearly show that a considerable proportion of VirA is present in the dimeric state in the intact cells. Role of AS in Dimerization of VirA. To determine whether AS plays a role in the formation of the VirA dimeric state, cells were grown in the presence or absence of AS. In this experiment virA was placed under the control of the lac promoter and induced by adding IPTG to the growing cells. As expected, VirA accumulated in the absence of AS (Fig. 2, lane 3). Crosslinking with BS3 indicated that VirA formed a homodimer in the absence of AS. Indeed, the crosslinking pattern of VirA was similar in the presence or absence of AS (Fig. 2, lanes 7 and 8). During the process of crosslinking, the cells were washed with phosphate-buffered saline. To eliminate the possibility that such washings might alter VirA conformation by depleting AS in the cells, 100 ,uM AS was added to the phosphate-buffered saline and crosslinking solutions. A similar crosslinking pattern was observed (data not shown). We conclude that AS plays no significant role in the dimerization of VirA. Genetic Evidence for Involvement of the VirA Dimeric State in Signal Transduction. The above crosslinking experiments clearly suggest that the sensor protein VirA exists as a homodimer in its native state. If this dimerization were required for signal transduction, aggregation of a defective VirA with a wild-type VirA protein would prevent the wild-type VirA from functioning. However, if the VirA monomer could transduce signals by itself, the presence of a defective VirA in a strain carrying a wild-type VirA would not significantly affect the functioning of the wild-type VirA. To test these possibilities, we coexpressed each of several defective virA genes together with the wild-type virA locus in the same A. tumefaciens cells and then assayed the level of virB or virE expression as an indicator of VirA signal transduction activity. Although these defective virA genes were driven by the lac promoter, in A. tumefaciens cells the level of virA expression under the control of the lac promoter was similar to that under the control of the native promoter (Fig. 2). pTC158, encoding a nonphosphorylatable protein, VirA(H/Q), whose phosphorylation site His-474 is replaced with Gln-474 (Fig. 1), could not complement the virA null mutant strain. pTC158 was introduced into MX368, which contains the wild-type virA and a virB-lacZ fusion, which was used to assay the level of virB expression. When VirA(H/Q) was induced by IPTG, the level of virB expression was reduced at least 10-fold (Fig. 4). The

Microbiology: Pan et al.

9942

Proc. Natl. Acad. Sci. USA 90 (1993)

strong inhibition of wild-type VirA function by VirA(H/Q) suggests that VirA(H/Q) dimerizes with wild-type VirA, resulting in a nonfunctional dimeric state. These genetic data strongly suggest that the VirA homodimer is the functional state, transducing the signal through an intersubunit mechanism (see Discussion). The VirA-VirG system is homologous to the EnvZ-OmpR system in Escherichia coli (for review see ref. 23). EnvZ both phosphorylates OmpR and dephosphorylates phospho-OmpR (24, 25). It is not clear whether VirA possesses phosphatase activity, but if it does, the point mutant VirA(H/Q) might retain some phosphatase activity which may quench signal transduction and hence contribute to the strong inhibition of wild-type VirA function by VirA(H/Q). In addition, VirA(H/Q) may also interact with VirG, thus competing with wild-type VirA for interaction with VirG. To test these possibilities we constructed two virA deletion mutants, virAA605/ 829 and virA(H/Q)A605/829, both of which lack one-third of the kinase domain and the entire C-terminal domain (Fig. 1). Presumably these deletions would have abolished the phosphatase activity (26) and the site of interaction with VirG. VirAA605/829 retains the highly conserved phosphorylation site His-474, while VirA(H/Q)A605/829 has an altered amino acid (Gln-474) at that position (Fig. 1). Neither pSQ7 nor pSQ8, containing virAA605/829 or virA(H/Q)A605/829, respectively, could complement the VirA- phenotype of Atl1023 (data not shown). However, both VirAA605/829 and VirA(H/Q)A605/829 reduced the level of virB expression at least 10-fold when coexpressed with wild-type VirA in MX368 (Fig. 4). The inhibitory activity of the two deletion mutants 'I

50

c

=

40

Z--

30

3:;

0

() 20 0

C)10 (5

C~

0

1 IPTG

+

2 -

+

3

4

+

FIG. 4. Inhibitory activities of defective VirA proteins against the wild-type VirA protein. A plasmid containing a defective virA, pSQ7 (lanes 1), pSQ8 (lanes 2) or pTC158 (lanes 4), or the wild-type virA, pTC115 (lanes 3), was introduced into MX368, which contains a virB-4acZ fusion generated with Tn3HoHol. Cells were grown in induction medium supplemented with 100 ,M AS and with (shaded bars) or without (open bars) 1 mM IPTG for 16 hr at 28°C. The f-galactosidase activity was determined as an indicator for virB expression (Upper). In addition, the remaining bacterial cells were collected, washed, and lysed; the lysates were electrophoresed in SDS/7.5% polyacrylamide gels; and the proteins were transferred onto Immobilon-P membranes and the f-galactosidase protein (the lower panel) was visualized by Western blot with the monoclonal antibody against f-galactosidase. As documented previously, Tn3HoHol generates both transcriptional and translational fusions (19). The arrowhead and arrow indicate the transcriptional and translational fusion products, respectively.

was similar to that of the full-length nonphosphorylatable VirA(H/Q). This suggests that the major reason for the strong inhibitory activity of a defective VirA against the wild-type VirA was due to formation of a nonfunctional dimer and not because of phosphatase activity or competition for VirG. One could also argue that the strong inhibitory activity observed with a defective VirA might be at least partially a result of a reduced availability of AS for the wild-type VirA, because a defective VirA may retain the ability to interact with AS. To test this possibility we determined the levels of virB expression in MX368(pTC158) grown in the presence of a series of concentrations of AS. The full-length nonphosphorylatable VirA(H/Q) most likely retains the ability to interact with AS. When VirA(H/Q) was expressed, the level of virB expression was strongly reduced and the increased AS concentrations (up to 400 ,uM) did not relieve any inhibition. This suggests that the availability of AS was not a factor involved in the strong inhibition shown by a defective VirA.

DISCUSSION It is becoming increasingly evident that two-component regulatory systems are widely distributed in various bacteria and are involved in recognizing and responding to diverse environmental stimuli (for review, see ref. 23). Typically, the two-component regulatory systems consist of a "sensor" protein that monitors some environmental parameter and a cytoplasmic "response regulator" protein that mediates changes in gene expression. Most sensor proteins are located in the cytoplasmic membranes. Although it is well established that autophosphorylation and phosphate transfer are involved in the process of signal transduction, the structural state of the membrane-bound sensor proteins which functions in signal transduction is still unknown. In E. coli, two defective osmosensor EnvZ proteins can complement each other, suggesting that the autophosphorylation of EnvZ is an intermolecular phosphorylation reaction (26). However, no physical evidence for a dimer or oligomer state of EnvZ was presented. It is also unclear whether two native sensor molecules interact with each other as a normal functional state in the native condition. In this paper we provide both physical and genetic evidence that the A. tumefaciens sensor protein VirA exists as a homodimer in the native state. Our genetic studies suggest that the VirA homodimer is the natural functional state involved in signal transduction. When a defective VirA was expressed in A. tumefaciens cells containing a wild-type VirA, virB gene expression was reduced at least 10-fold. Although the defective virA genes were driven by the lac promoter, in A. tumefaciens cells the level of virA expression under the control of the lac promoter was similar to that under the control of the native promoter (Fig. 2). If VirA functioned as a heterodimer or heteromultimer, the defective VirA would compete with the wild-type VirA only for the unidentified protein(s), and the defective VirA would be expected to inhibit about 50%. If the VirA monomer were active in transducing signals by itself, the coexistence of a defective VirA with the wild-type VirA would not significantly reduce the signal transduction activity of the wild type. If a VirA homodimer is the functional state transducing signals, the coexistence of a defective VirA and the wild-type VirA will inhibit the function of the wild-type by 4-fold. However, the inhibition of a defective VirA against a wild-type VirA was >10-fold. We presume that the inhibition was higher than the expected 4-fold probably because the defective VirA proteins expressed under the control of the lac promoter accumulated earlier than the wild-type VirA whose expression was under the control of the native promoter that is regulated in a virA-, virG-dependent fashion (27). An earlier presence of a defective VirA would be expected to slow down the accumulation of the wild-type

Microbiology: Pan et al. VirA in the same cells so that the cumulative level of the wild-type VirA was much lower than that of the defective VirA, resulting in an even stronger inhibition of vir gene expression. In fact, the wild-type VirA was observed to accumulate more slowly when the defective VirA was coexpressed, and the defective VirA exhibited a weaker inhibition against the wild-type VirA after the cells were grown in the induction medium for a prolonged period (data not shown). The genetic data also suggest that the VirA dimer transduces signals through an intersubunit mechanism. The point mutant VirA(H/Q) lacks the phosphorylation site but probably retains kinase activity, as does its homolog EnvZ(H243V) (26); the deletion mutant VirAA605/829 retains the phosphorylation site but probably lacks kinase activity, as does its homolog EnvZdlA (26). However, the inhibitory activities of both of these mutant VirA proteins against the wild type were comparable to that of the mutant VirA(H/ Q)A605/829, which lacks both the phosphorylation site and (probably) the kinase activity. Because the three mutant proteins still retain the domain involved in recognizing AS (28), the data suggest that the signal transduction process requires an interaction of two subunits after recognizing AS. It appears that this intersubunit interaction differs from the intermolecular phosphorylation observed previously (26). If only an intermolecular phosphorylation is responsible for this intersubunit interaction, the heterodimer between a subunit of wild-type VirA and a subunit of VirA(H/Q) or VirAA605/829 should be functional and these mutant proteins would not have shown a significant inhibition. Unlike eukaryotic membrane-bound receptor tyrosine kinases, whose dimerization is often mediated by their signal molecules (29, 30), VirA forms a homodimer in the absence of the plant signal molecule AS. This resembles the Salmonella aspartate receptor (a methyl-accepting chemotaxis protein), which is dimeric in the absence of its signal aspartate (31). However, unlike the aspartate receptor, which transmits signaling through a single subunit (32), VirA transduces signals through an intersubunit mechanism as several eukaryotic receptor kinases (29, 30). Presumably AS is only involved in triggering the signal transduction process, probably by causing autophosphorylation of the VirA homodimer. Low basal levels of both VirA and VirG are present in A. tumefaciens in the absence of signal molecules; and plant signals induce a much higher level of both VirA and VirG (27). We assume that VirA is present in a homodimer state and is ready to function as soon as it detects plant signals. Such a preformed dimeric state ready for plant-Agrobacterium signal transduction probably was adopted by bacteria as a strategy for quickly adapting to a changing environment. We thank Colin W. Jones for kindly providing the antibody against ChvE and Kathy Stephens for critical reviewing the manuscript. This work was supported by Public Health Service Grant 5RO1GM32618-18 from the National Institutes of Health (to E.W.N.), Postdoctoral Fellowship GM15674 from the National Institutes of Health (to S.Q.P.), and a postdoctoral fellowship from the Natural Sciences and Engineering Research Council of Canada (to T.C.).

Proc. Natl. Acad. Sci. USA 90 (1993)

9943

1. Kado, C. I. (1991) Crit. Rev. Plant Sci. 10, 1-30. 2. Winans, S. C. (1992) Microbiol. Rev. 56, 12-31. 3. Zambryski, P. C. (1992) Annu. Review Plant Physiol. Mol. Biol. 43, 465-490. 4. Melchers, L. S., Regensburg-Tuink, T. J. G., Bourret, R. B., Sedee, N. J. A., Schilperoort, R. A. & Hooykaas, P. J. J. (1989) EMBO J. 8, 1919-1925. 5. Winans, S. C., Kerstetter, R. A., Ward, J. E. & Nester, E. W. (1989) J. Bacteriol. 171, 1616-1622. 6. Jin, S., Roitsch, T., Ankenbauer, R. G., Gordon, M. P. & Nester, E. W. (1990) J. Bacteriol. 172, 525-530. 7. Huang, Y., Morel, P., Powell, B. & Kado, C. I. (1990) J. Bacteriol. 172, 1142-1144. 8. Jin, S., Prusti, R. K., Roitsch, T., Ankenbauer, R. G. & Nester, E. W. (1990) J. Bacteriol. 172, 4945-4950. 9. Jin, S., Roitsch, T., Christie, P. J. & Nester, E. W. (1990) J. Bacteriol. 172, 531-537. 10. Garfinkel, D. J. & Nester, E. W. (1980) J. Bacteriol. 144, 732-743. 11. Stachel, S. E. & Nester, E. W. (1986) EMBO J. 5, 1445-1454. 12. Chen, C.-Y. & Winans, S. C. (1991) J. Bacteriol. 173, 11391144. 13. Fellay, R., Frey, J. & Krisch, H. (1987) Gene 52, 147-154. 14. Cangelosi, G. A., Ankenbauer, R. G. & Nester, E. W. (1990) Proc. Natl. Acad. Sci. USA 87, 6708-6712. 15. Cangelosi, G. A., Best, E. A., Martinetti, G. & Nester, E. W.

(1991) Methods Enzymol. 204, 384-397.

16. Ditta, G., Stanfield, S., Corbin, D. & Helinski, D. R. (1980) Proc. Natl. Acad. Sci. USA 77, 7347-7351. 17. Chilton, M.-D., Currier, T., Farrand, S., Bendich, A., Gordon, M. P. & Nester, E. W. (1974) Proc. Natl. Acad. Sci. USA 71, 3672-3676. 18. Miller, J. H. (1972) Experiments in Molecular Genetics (Cold Spring Harbor Lab. Press, Plainview, NY). 19. Stachel, S. E., An, G., Flores, C. & Nester, E. W. (1985) EMBO J. 4, 891-898. 20. Laemmli, U. K. (1970) Nature (London) 227, 680-685. 21. Leroux, B., Yanofsky, M. F., Winans, S. C., Ward, J. E., Ziegler, S. F. & Nester, E. W. (1987) EMBO J. 6, 849-856. 22. Pakula, A. A. & Simon, M. I. (1992) Proc. Natl. Acad. Sci. USA 89, 4144-4148. 23. Parkinson, J. S. & Kofoid, E. C. (1992) Annu. Rev. Genet. 26, 71-112. 24. Aiba, H., Mizuno, T. & Mizushima, S. (1989) J. Biol. Chem. 264, 8563-8567. 25. Igo, M. M., Ninfa, A. J., Stock, J. B. & Silhavy, T. J. (1989) Genes Dev. 3, 1725-1734. 26. Yang, Y. & Inouye, M. (1991) Proc. Natl. Acad. Sci. USA 88, 11057-11061. 27. Winans, S. C., Kerstetter, R. A. & Nester, E. W. (1988) J. Bacteriol. 170, 4047-4054. 28. Chang, C.-H. & Winans, S. C. (1992) J. Bacteriol. 174, 70337039. 29. Ueno, H., Colbert, H., Escobedo, J. A. & Williams, L. T.

(1991) Science 252, 844-848.

30. Heldin, C. H., Ernlund, A., Rorsman, C. & Ronnstrand, L. (1989) J. Biol. Chem. 264, 8905-8912. 31. Milligan, D. L. & Koshland, D. E., Jr. (1988) J. Biol. Chem. 263, 6268-6275. 32. Milligan, D. L. & Koshland, D. E., Jr. (1991) Science 254, 1651-1654.