Evidence that F-plasmid proteins TraV, TraK and TraB assemble into an envelope-spanning structure in Escherichia coli



We have examined the role of the F-plasmid TraV outer membrane lipoprotein in the assembly of F-pili. Yeast two-hybrid analysis with a traV bait repeatedly identified traK, which is predicted to encode a periplasmic protein, among positive prey plasmids. A traK bait in turn identified traV and traB, which is predicted to encode an inner membrane protein. A traB bait exclusively identified traK preys. Several additional observations support the hypothesis that TraV, TraK and TraB form a complex in Escherichia coli that spans the cell envelope from the outer membrane (TraV) through the periplasm (TraK) to the inner membrane (TraB). First, two-hybrid analyses indicated that TraV and TraB bind to different TraK segments, as required if TraK bridges a ternary complex. Secondly, all three proteins fractionated with the E. coli outer membrane in tra+ cells. In contrast, TraB fractionated with the inner membrane in traV or traK mutant cells, and TraK appeared in the osmotic shock fluid from the traV mutant. These results are consistent with a TraV–TraK–TraB complex anchored to the outer membrane via the TraV lipoprotein. Further, in traK mutant cells, TraV failed to accumulate to a detectable level, and the TraB level was significantly reduced, suggesting that TraV and TraB must interact with TraK for either protein to accumulate to its normal level. Both TraK and TraV accumulated in traB2[Am] cells; however, the TraB2 amber fragment could be detected by Western blot, and sequence analysis indicated that the fragment retained the TraK-binding domain suggested by yeast two-hybrid analysis. We propose that TraV is the outer membrane anchor for a trans-envelope, Tra protein structure required for the assembly of F-pili and possibly for other events of conjugal DNA transfer.


Conjugal or horizontal DNA transfer is ubiquitous among bacteria and even occurs from bacterial to eukaryotic cells (Heinemann and Sprague, 1989; Christie, 2001;Kunik et al., 2001). Among Gram-negative species, DNA donor activity is a specialized cell function requiring a set of DNA transfer genes usually contained within large, conjugative plasmids. The archetype of such conjugative plasmids is F, a 100 kbp plasmid whose 30 or so DNA transfer (tra) genes are contained within the 33 kbp tra segment, the entire nucleotide sequence of which has been determined (Frost et al., 1994). There is also information about the localization of individual Tra proteins in Escherichia coli and, broadly, the stage of conjugal DNA transfer at which each is required (Firth et al., 1996). Other models of conjugal DNA transfer that have been studied intensively include the broad-host-range IncP plasmid RP4 and the Ti plasmid of Agrobacterium spp. The transfer systems encoded by these plasmids are similar enough to each other to suggest a common evolutionary origin, perhaps among ancestral systems devoted to intercellular protein transfer (Kado, 1994; Lessl and Lanka, 1994; Christie, 2001). It is presently unclear to what degree, if at all, the Ti, RP4 and related systems resemble the DNA transfer systems encoded by F and the F-like R factors.

One feature evidently common to all Gram-negative DNA transfer systems is the requirement for conjugative pili (Paranchych and Frost, 1988; Ippen-Ihler and Maneewannakul, 1991; Haase et al., 1995; Fullner et al., 1996). These pili are extracellular filaments composed of a single subunit (F-pilin for F+ strains) and are the only structures so far unequivocally associated with conjugal DNA transfer. F-pili and, by inference, other conjugative pili as well function in the earliest cell–cell contact stages of transfer, just before the formation of specific and secure surface contacts between donor and recipient cells (Firth et al., 1996). F-pili are also the primary adsorption organelles for RNA and filamentous DNA coliphages (Valentine et al., 1969).

Functional F-pili are the end-point of a pathway involving about half the F-encoded Tra proteins (Firth et al., 1996). A mutation in any of a dozen tra genes abolishes F-pilus formation (assayed by electron microscopy), donor activity and donor-specific phage sensitivity. Mutations in yet other tra genes alter the number or length distribution of F-pili without entirely abolishing filament assembly.

The formation of F-pili occurs in two easily distinguishable stages. The first is the accumulation of an F-pilin subunit pool in the bacterial inner membrane (Moore et al., 1981a). This stage requires traA, the structural gene for F-pilin, and traQ, which acts catalytically as a TraA-specific chaperone (Moore et al., 1982; Maneewannakul et al., 1993; Harris et al., 1999). The F-pilin secretion/assembly process is arrested at this stage unless other F-plasmid tra genes are expressed (Moore et al., 1981b; Sowa et al., 1983). The actions of these other tra genes constitute the second stage of filament assembly.

For the DNA transfer systems encoded by plasmids RP4 and Ti, there is accumulating evidence that proteins required for the early stages of DNA transfer form structures that span the bacterial cell envelope (Christie, 1997; Grahn et al., 2000). These proteins and, by inference, the structures into which they normally assemble are in each case also required for the formation of surface filaments, presumably conjugative pili (Haase et al., 1995;Fullner et al., 1996).

Fernandez et al. (1996) proposed a special role for the outer membrane lipoprotein VirB7, along with VirB9, in nucleating the assembly of other Vir proteins into a surface structure. Given the close relationship between the RP4 and Ti systems, the same is likely to be true for RP4. Here, we show that the TraV outer membrane lipoprotein encoded by F may play a role similar to that of VirB7 in nucleating the formation of a surface structure required for the assembly of F-pili and perhaps for subsequent stages of conjugal DNA transfer. Hence, the RP4/Ti and the F/F-like DNA transfer systems manifest some interesting similarities, whether or not they are related by a common ancestor.


Yeast two-hybrid analysis

We first used yeast two-hybrid screens to identify binary interactions among Tra proteins known from genetic studies to be required for the assembly of functional F-pili (Harris et al., 1999). As we were interested in possible envelope-spanning structures, we began with a traV bait plasmid. TraV is an outer membrane lipoprotein with a hydrophilic domain extending into the periplasm (Doran et al., 1994). It is the only outer membrane Tra protein required for F-pilus synthesis (Firth et al., 1996) and is therefore a prime candidate for the outer membrane anchor of a trans-envelope structure. The prey library was derived from pTG801, which contains all the F tra genes required for the elaboration of functional F-pili and, with one or two exceptions, only those genes (Grossman and Silverman, 1989; Harris et al. 1999).

A two-hybrid screen with a full-length traV bait (minus the segment encoding the TraV leader peptide) identified numerous prey plasmids; 1.3% of the Leu+ Trp+ transformants were His+ LacZ+ as well (Table 1). More than half (27/49) the prey plasmids sequenced contained traK. Of the 27 traK segments identified, all but one were out-of-frame with respect to the upstream GAL4 sequence. Both +1 and +2 shifts were observed at similar frequencies. Counterselection of in frame fusions has been observed in previous two-hybrid screens when an interacting fusion polypeptide is deleterious to growth (Fromont-Racine, 1997;Sourdive et al., 1997). The frameshift mutation evidently reduces the level of the protein to that which can be produced as the result of frameshift suppression. The same phenomenon would explain our results with TraK.

Table 1.  Interactions among TraV, TraK and TraB by yeast two-hybrid analysis.
tra baitCodons (full gene)Leu+ Trp+ transformantsHis+ LacZ+ positives tra sequences/ total sequences
traV21−17117145 000590 traK (27/49)
traB (3/49)
traV (3/49)
Other tra (5/49)
Miscellaneous (11/49)
traK 1–242 242131 000211traV (16/32)
traB (14/32)
Miscellaneous (2/32)
traB8−475475402 00016 traK (16/16)

Other prey plasmids hit more than twice included traV itself and traB. All these were in the correct reading frame. The remaining prey plasmids included tra sequences hit only once or twice, reverse tra sequences and unidentified sequences, perhaps from the pTG801 vector backbone.

We next used traK as the bait plasmid. This screen identified only two tra preys: traV and traB. A screen with traB bait identified only traK preys (Table 1). These three proteins, TraV in the outer membrane, TraK in the periplasm and TraB in the inner membrane, have the potential to span the E. coli cell envelope, perhaps accounting for the location of F-pili at inner membrane–outer membrane junctions (Beyer, 1979; Silverman, 1987).

Owing to the high resolution of the prey library used in these experiments (Harris et al., 1999), we could infer interaction domains of the three proteins from the two-hybrid analyses. Among the prey plasmids isolated with the traV bait, the 27 traK fusion sites ranged from codon S12 to codon T124 (out of 242 in total). No two fusion sites were within the same codon. These data indicate that a TraV interaction site resides in the C-terminal half of TraK (Fig. 1). In contrast, all the 16 traK sequences obtained from a screen with traB bait began within the first 27 codons (Fig. 1). This limited target size suggests that the N-terminal segment of TraK binds TraB and could explain the low frequency of positives in this screen (1/25 000 Leu+ Trp+ transformants). Similar reasoning indicates that the TraK interaction site of TraV is within the C-terminal 44 amino acids, and the TraK interaction site of TraB is contained in a segment beginning at codon 138 (Fig. 1).

Figure 1.

Interaction domains of TraV, TraK and TraB inferred from yeast two-hybrid data. Each polypeptide is represented by a segment proportional to its length in amino acids. Light grey shading denotes the boundaries within which all GAL4–tra fusions were located among sequenced prey plasmids (e.g. V36–M138 for traB preys isolated with a traK bait). Dark grey shading denotes the boundaries within which 3′-termini of tra segments were located (e.g. D229–V316 for traB preys). Black shading denotes segments inferred to contain binding domains for the indicated partner(s). The dotted extension of the putative TraB interaction domain of TraK indicates uncertainty as to its C-terminus.

Some traB preys beginning near the 5′-terminus of the coding sequence did not extend through the entire gene, which at 475 codons is longer than the mean fragment length of the prey library, 1.0 kbp (Harris et al., 1999). From the position of 3′-termini among eight such sequences, we infer that the TraK interaction site is located within a TraB segment encompassing residues M138 to D229 (Fig. 1). This segment is distinctive in containing 21 proline residues (23%), 16 of which are between M138 and Q183 (Frost et al., 1994).

The traK preys isolated with the traB bait were all in frame, in contrast to those obtained from the traV screen (see above). When we transformed a traV bait plasmid with one of the in frame traK preys, colony size on agar plates was variable, but generally smaller than those from cells transformed with the same traK prey and the traB bait. It appears that the TraV–TraK interaction, as opposed to the TraK fusion per se, adversely affects yeast growth.

Taken together, the results of these two-hybrid analyses are consistent with the formation of a complex that includes TraV, TraK and TraB.

Detection of Tra proteins in E. coli

Using antibodies against TraV, TraB and TraK, we extended the results of the yeast two-hybrid analyses to tra+ strains of E. coli. Polyclonal antisera were raised in rabbits and affinity purified as described in Experimental procedures. The antibodies were tested by Western blot against proteins from XK1200/pOX38kan (tra+) and traV, traK or traB mutants of F′lac(Fig. 2). In all three cases, the antibodies detected polypeptides in the tra+ strain that were absent or altered in size in the corresponding mutant strain (Fig. 2). Furthermore, these polypeptides corresponded in molecular mass to those predicted for the corresponding Tra proteins, minus leader peptide segments where appropriate: 17 kDa for TraV; 24 kDa for TraK; and 52 kDa for TraB [TraB migrated more slowly than predicted, as noted also by Moore et al. (1987)].

Figure 2.

TraV, TraB and TraK accumulation in tra+ and tra mutant cells. Cells were grown to an OD600 of 0.6–0.8, collected by sedimentation and heated in electrophoresis sample buffer. Proteins were separated in 4–20% gradient gels, blotted onto PVDF paper and probed with anti-Tra protein antibodies. Plasmid genotypes were: lanes 1, 3 and 5, tra+ (pOX38kan); lane 2, F′lac traV569; lane 4, F′lac traB2; lane 6, F′lac traK105. Molecular mass standards are indicated on the left; masses are in kDa. Horizontal carets indicate the different Tra proteins discussed in the text. The standards nearest in mass to TraB and TraK are indicated to the left of lanes 3 and 5 respectively.

The anti-TraV antibodies also detected a component with an apparent mass of about 40 kDa that was absent in the traV569 mutant (Fig. 2). In order to assess the significance of this observation, we examined several other strains by Western blot (Fig. 3). The 40 kDa component was present in XK1200/pOX38kan(Fig. 3, lane 1) and HfrH (Fig. 3, lane 2), but not in JC3272/JCFL0 (Fig. 3, lane 3), notwithstanding the fact that all three plasmids have a full complement of tra genes. Moreover, in DH1/pTG801/pIQ2 cells, in which the pTG801 tra genes are inducible (Grossman and Silverman 1989), TraV levels increased significantly after IPTG addition, whereas the level of the 40 kDa component was unaltered (Fig. 3, lanes 5 and 6). Anti-TraV antibodies to the F-encoded TraV protein also detected the R100 TraV protein, which is essentially identical to that encoded by F (Anthony et al., 1999). The protein from derepressed R100-1 cells was readily detectable by Western blot but, as expected, could not be detected in cells with the repressed R100 wild-type plasmid; the 40 kDa component could not be detected in either strain (Fig. 3, lanes 7 and 8). Finally, as shown below, the 40 kDa component fractionated with osmotic shock fluid, whereas TraV is an outer membrane component. Thus, we find no correlation between the 40 kDa component and donor activity, tra gene expression or TraV localization and have not characterized this material further.

Figure 3.

Expression of the 40 kDa component reactive with anti-TraV antibodies. The data were obtained exactly as in Fig. 2. Strain or plasmid content: lane 1, pOX38 (tra+); lane 2, HfrH; lane 3, JCFL0 (F′lac tra+); lane 4, F′lac traV569; lane 5, pTG801 (grown without induction); lane 6, pTG801 (grown with 1 mM IPTG); lane 7, R100; lane 8, R100 drd1.

The anti-TraB antibodies detected a single component in the traB2 strain with an apparent mass of 27 kDa (Fig. 2; see also Fig. 6), compared with the larger apparent mass of the wild-type TraB polypeptide. As the traB2 allele is a nonsense mutation (JCFL2 in Achtman et al., 1972), this component is likely to be the TraB2 amber fragment. By sequence analysis, we identified the traB2 amber codon as the result of a C to T transition at tra bp 4861 (Frost et al., 1994), yielding the Q207Am mutation. The size of the TraB2 polypeptide should therefore be about 23 kDa but, given that the wild-type polypeptide has a lower than expected mobility (see above), the data are in reasonable agreement with prediction.

Figure 6.

Effect of tra mutations on the accumulation of TraV, TraK and TraB polypeptides. Whole-cell extracts were assayed by Western blot. Plasmid content: lanes 1, pOX38kan (tra+); lanes 2, F′lac traV569; lanes 3, F′lac traK105; lanes 4, F′lac traB2.

Localization of Tra proteins

Each of the F Tra proteins has been assigned at least a provisional location within the cell, based largely on fractionation experiments with inviable (UV-irradiated) cells expressing one or a subset of such proteins or on sequence-based predictions (Firth et al., 1996). TraV, TraK and TraB are predicted to reside in the outer membrane, the periplasm and the inner membrane, respectively, but there are no data reporting the actual localization of these proteins in normal tra+ strains. In fact, all three proteins fractionated quantitatively with the outer membrane of tra+ cells prepared by sucrose density gradient sedimentation (Fig. 4). In preparations containing material intermediate in density between inner and outer membrane fractions (M band material; Osborne et al., 1972), the TraV contents of the outer membrane fractions and intermediate density fractions, normalized to total protein, were indistinguishable by Western blot at excess antibody (data not shown). Note also that the 40 kDa component that reacted with anti-TraV antibodies and was absent from traV569 cells fractionated with shock fluid, suggesting a periplasmic location (Fig. 4A); this was also true in traK105 cells, in which TraV itself did not accumulate (see below).

Figure 4.

Localization of TraV, TraK and TraB in a tra+ strain. RD17/pOX38kan was fractionated as described in Experimental procedures. A portion of each fraction scaled to 2 × 107 cells was electrophoresed through 4–20% polyacrylamide gels. Western blots were performed as described in Experimental procedures with antibodies to TraV (A), TraK (B) or TraB (C). WC, whole-cell extract; SF, osmotic shock fluid; C, cytoplasm; M, crude membranes; IM, purified inner membrane; OM, purified outer membrane. The horizontal caret in (A) indicates the 40 kDa component. In (B), owing to cross-reacting material, whole-cell extract from the F′lac traK105 strain was run in parallel to identify TraK unequivocally. Molecular mass standards (in kDa) are indicated.

These data, like the yeast two-hybrid analyses, are consistent with the formation of a stable complex of TraV, TraK and TraB (and possibly other proteins) whose fractionation properties are determined by the strong association of TraV with the outer membrane. If this is so, and if the predictions that TraB is an inner membrane protein and TraK is periplasmic are correct, then the fractionation properties of TraB and TraK should depend on whether or not TraV is present. This is in fact the case (Fig. 5). In traV569 cells, TraK was detected primarily in the osmotic shock fluid, indicating a periplasmic location, instead of the outer membrane (Fig. 5A). Some TraK was still detected in the outer membrane fraction, suggesting a weak association between TraK and an outer membrane protein other than TraV, perhaps a host protein. No TraK was detected in the inner membrane fraction of traV569 cells, suggesting that any TraK–TraB interaction is also relatively weak, at least in the absence of TraV. Similarly, TraB was detected primarily among inner membrane proteins of traV569 cells (Fig. 5B); none was found in the outer membrane fraction. Furthermore, full-length TraB also fractionated with the inner membrane of traK105 cells (Fig. 5C). In this experiment, several additional components were detected by anti-TraB antibodies in the 25–45 kDa region of the gel (Fig. 5C). These were not detected in other experiments (e.g. Fig. 5B). These additional components could be TraB degradation products, or they could reflect the fact that more E. coli protein and more anti-TraB antibody were used in this experiment owing to the low level of TraB in traK105 cells (see below), or a combination of these two possibilities. In any event, TraB appears to be an inner membrane protein, as predicted, and its fractionation properties from tra+ cells can be attributed to its association with a TraK–TraV complex anchored to the outer membrane. Similarly, several Tra proteins encoded by the RP4 Tra2 region fractionated with the outer membrane in Tra2+ cells, notwithstanding their localization to the inner membrane or shock fluid when expressed alone (Grahn et al., 2000)

Figure 5.

Fractionation of TraK and TraB from mutant cells. The procedure was as described for Fig. 4, except that, for (C), the antibody dilution was 1:250, and the sample size run in each lane corresponded to 108 cells.

Tra protein accumulation in mutant strains

For the Agrobacterium Ti plasmid conjugal DNA transfer system, the absence of certain VirB proteins reduced or effectively abolished the accumulation of others (Fernandez et al., 1996; Hapflmeier et al., 2000). Of special interest in this respect is the VirB7 outer membrane lipoprotein, in whose absence four out of six VirB proteins examined failed to accumulate (Fernandez et al., 1996). This was interpreted to mean that VirB7 plays an essential role in nucleating the assembly of a VirB protein complex at the cell surface. We therefore tested the effects of mutations in traV, traK and traB on the accumulation of the other proteins in the TraV/K/B linkage group.

Both TraK and TraB accumulated to normal or nearly normal levels in the traV569 mutant (Fig. 6). However, TraV itself failed to accumulate, and the TraB level was significantly reduced in the traK105 mutant (Fig. 6). The traB2 mutation had much less drastic effects. The TraV and TraK levels in the traB2 mutant were only somewhat lower than those in the wild-type strain (Fig. 6); note that the TraB2 amber fragment is again clearly visible. The level of the 40 kDa component was not reduced by the traK105 or the traB2 mutation; if anything, this material appeared to accumulate to a somewhat higher level in these mutant strains (Fig. 6). Thus, although TraK appears to be essential for the accumulation of TraV and, to a lesser degree, TraB, the absence of the TraV lipoprotein does not substantially affect TraK or TraB levels.


A long-standing hypothesis is that F-encoded Tra proteins act together at the cell surface to mediate the different stages of conjugal DNA transfer (Silverman, 1987;Ippen-Ihler and Maneewannakul, 1991). The data reported here provide the first evidence directly supporting this hypothesis.

By yeast two-hybrid analyses, TraV, TraK and TraB constitute a protein interaction group. The two-hybrid data are consistent with simultaneous binding of TraK to TraV and to TraB, thereby possibly forming a complex in the E. coli cell envelope.

Consistent with this hypothesis, all three proteins fractionated with the outer membrane of tra+E. coli, notwithstanding predicted locations in the periplasm for TraK and the inner membrane for TraB (Firth et al., 1996). In traK105 or traV569 cells, full-length TraB did fractionate with the inner membrane, as predicted, and in contrast to its fractionation with the outer membrane in tra+ cells. Similarly, most of the TraK in traV569 cells fractionated with soluble periplasmic proteins. These data suggest the formation of a TraV/B/K complex that spans the E. coli cell envelope and is firmly anchored in the outer membrane via TraV. Such a complex would also be consistent with electron microscopy showing that F-pili extend from surface sites at which inner and outer membranes are joined (Beyer, 1979; Silverman, 1987). At this stage, we cannot say whether or not this complex contains other proteins encoded by F or by E. coli itself.

Although there is little evidence from sequence comparisons that the F and F-like DNA transfer systems are ancestrally related to those encoded by RP4 and Ti, we can now point to some interesting parallels. In the RP4 and Ti systems, the relevant proteins also assemble at the cell envelope, where they appear to be anchored to the outer membrane (Thorstensen et al., 1993; Christie, 1997; Grahn et al., 2000). For the Ti system, this anchor includes the VirB7 lipoprotein (Fernandez et al., 1996); the RP4 system also includes the outer membrane lipoprotein TrbH (Grahn et al., 2000), which may well play a similar role to that proposed for VirB7 (Fernandez et al., 1996). We have shown here that F-encoded TraV, also an outer membrane lipoprotein (Doran et al., 1994), is an outer membrane anchor for at least two other Tra proteins, TraB and TraK. Our data on F-plasmid Tra proteins are therefore broadly similar to those obtained with the conjugal DNA transfer systems encoded by RP4 and Ti, notwithstanding the possibility that the F and RP4/Ti systems evolved independently.

VirB7 forms disulphide-linked homo- and heterodimers, the latter with VirB9 (Anderson et al., 1996; Baron et al., 1997; Spudich et al., 1996). Of four amino acid identities between mature VirB7 and TraV lipoproteins, two involve functionally important cysteines: the N-terminal cysteine common to all bacterial envelope lipoproteins and Cys-10 (Cys-24 of full-length VirB7; Cys-28 of full-length TraV), which is engaged in disulphide bond formation between VirB7 monomers and between VirB7 and VirB9. TraV contains one additional cysteine, Cys-18. Preliminary data suggest that both TraV cysteines are engaged in disulphide bond formation (R. L. Harris and P. M. Silverman, unpublished results), although the functional significance of this result, if any, is not yet known.

F-encoded TraB and Ti-encoded VirB10 have also been suggested to be similar (Kado, 1994). Although both are inner membrane proteins, a blast2 comparison returned no significant similarity at the amino acid sequence level. By yeast two-hybrid assay, VirB10 interacted strongly with VirB9 and VirB8 and weakly with itself (Das and Xie, 2000). It is unclear whether either of these last Vir proteins is analogous to TraK. In any event, all three Vir proteins fractionated with sedimentable components of the cell envelope, both inner and outer membranes. In our experiments, both TraK and TraB fractionated quantitatively with the outer membrane of tra+ cells. We interpret this result to mean that TraB is part of a trans-envelope structure whose fractionation properties reflect a high content of outer membrane.

The segment of TraB implicated in TraK binding by yeast two-hybrid data is remarkably proline rich. Proline accounts for 23% (21/92) of the residues in the segment M138–D229 and 35% in the segment M138–Q183. Proline-rich segments are commonly recognized in eukaryotic cell protein–protein interactions (Kay et al., 2000). Whether TraB uses a eukaryotic proline-rich ligand-binding motif to bind TraK remains to be determined but, in any case, the TraB2 amber fragment, which terminates at TraB amino acid Q207, includes most of the proline-rich region of TraB and may therefore contain the TraK binding site of the full-length protein. For this reason, no conclusion can be drawn from the present studies regarding the stability of TraK or TraV in the complete absence of TraB.

Although the formation of a TraV–TraK–TraB complex is the likeliest interpretation of our data, there is an alternative. The traK105 frameshift mutation reduced TraV accumulation to undetectable levels. Conceivably, TraK could act as a TraV chaperone. Lipoproteins are chaperoned from the inner membrane to their final sites in the outer membrane (Yokata et al., 1999). In this hypothesis, TraV and TraB interact directly. In the absence of TraV, owing to either a traV mutation or the absence of its TraK chaperone, TraB would fractionate with the inner membrane, as observed. We consider this hypothesis to be improbable because it does not account for the two-hybrid interaction between TraB and TraK or the fractionation of TraK with the outer membrane of tra+ cells.

The functions of TraV, TraK and TraB in conjugal DNA donor activity are not precisely known. Individually, traV569, traK105 and traB2 mutants have broadly similar phenotypes: absence of F-pili visible by electron microscopy; low or no detectable DNA donor activity; and resistance to donor-specific, F-pilus-binding bacteriophage by plaque assay (Firth et al., 1996). All these phenotypes are consistent with a failure of the initiation or elongation of F-pilus filaments. Anthony et al. (1999) reported that traB2 and traV569 mutants could still be infected by M13K07, as measured by transduction of kanamycin resistance, albeit at reduced levels. This would suggest a failure of filament elongation, as M13 binds to the tips of F-pili. In contrast, the traK105 mutant was completely resistant to M13K07 by this very sensitive test (Anthony et al., 1999), suggesting a failure of initiation. Although these functional designations may be correct, our data suggest a simpler alternative, namely that the traK105 mutation has pleiotropic effects, reducing both TraB and TraV levels, whereas the traV and traB mutations do not. Thus, the effects of the traK mutation might be expected to be more severe than those of mutations in either of the other two genes, no matter what stage of filament formation was affected.

It is of some interest that the two-hybrid screens presented above defined a closed interaction group consisting of TraV, TraK and TraB. We have assumed that all or most of the Tra proteins encoded by F assemble into a single surface structure that mediates conjugal DNA transfer (Silverman, 1987). In this case, one might have expected to find a more extensive protein interaction group by yeast two-hybrid analysis. There are several reasons why this might not have been observed, for example the two-hybrid screens we carried out would not detect interactions requiring more than two Tra proteins or E. coli proteins. It is of course possible that the F-encoded Tra proteins assemble into more than one distinct structure, although this possibility is unappealing. In any case, we have defined a second interaction group by yeast two-hybrid assay, comprising six Tra proteins other than TraV, TraK and TraB (R. L. Harris, V. Hombs and P. M. Silverman, unpublished data). It will be of interest to determine whether or how this interaction group is related to the TraV/B/K group described here.

Experimental procedures

Strains, plasmids and growth conditions

All bacterial strains were derivatives of E. coli K-12. Unless otherwise specified, the F′lac plasmids used in these studies were obtained from Dr L. Frost. JCFL0 (F′lac tra+) (Achtman et al., 1972) was from our laboratory collection. F′traV569 contains an amber mutation in codon 76 (of 153 in the mature protein) (Doran et al., 1994). The mutant F′ was obtained from Dr N. Firth in the supD strain JC6255 (Achtman et al., 1972). It was then transferred by conjugation into a nal derivative of the sup° strain JC3272 (Achtman et al., 1972). The traV569 allele in this background was complemented for mating and bacteriophage sensitivity by a traV+ plasmid cloned in pUC19 (data not shown). F′traK105 contains a +1 frameshift at codon P170 (of 221 in the mature protein) (Frost et al., 1994). F′traB2 contains an amber mutation in codon 207 (of 475) (R. L. Harris and P. M. Silverman, unpublished result). pOX38kan is a tra+ derivative of F, lacking insertion sequences (Chandler and Galas, 1983). R100 and R100 drd1 were from our laboratory collection. JCFL0, F′traV569, R100 and R100 drd1 were in strain JC3272 (Achtman et al., 1972). F′traK105 and F′traB2 were in strain XK1200 (Moore et al., 1987). pOX38kan was in strain RD17 [Δ(lac-pro)X111recA56 rel-1 supE44 thi-1]. pWP201HPC4 was constructed from pWP201, a pUC18 derivative that contains the F-plasmid traA translation initiation sequences and traA codons 1–57, followed by a BamHI site (Paiva and Silverman, 1996). Using mutually primed synthesis, we constructed a small DNA segment consisting of a BamHI site, the HPC4 epitope coding sequence (Rezaie et al., 1992), restriction sites corresponding to those in the prey plasmid pACTII, translation termination codons in all three reading frames and an EcoRI site. This segment was cloned into pWP201 digested with BamHI and EcoRI to yield pWP201HPC4.

Bacteria were grown routinely with aeration in Luria–Bertani medium at 37°C. Antibiotics (all from Sigma) were added at concentrations noted previously (Paiva and Silverman, 1996). Cells were harvested for experiments at an optical density of 0.6–0.8 (600 nm).

Yeast strains used for two-hybrid analysis and the tra fragment library in the prey plasmid pACTII have already been described (Harris et al., 1999). Bait plasmids were constructed by polymerase chain reaction (PCR) amplification of desired tra genes, which were then cloned into the bait plasmid pAS1CYH2 (Harris et al., 1999). The source of DNA was purified pTG801 or F′lac JCFL0 (Achtman et al., 1972) obtained by DNA extraction of E. coli K-12 strain JC3272/JCFL0 or by adding a few such cells directly to the PCR reaction. PCR primers, in F tra bp co-ordinates, were: traV forward, 6745–6766; traV reverse, 7198–7221; traK forward, 3480–3502; traK reverse, 4235–4260; traB forward, 4250–4272; traB reverse, 5686–5711. Point mutations were introduced into non-coding segments to create restriction sites for cloning. The tra segments of all three bait plasmids were sequenced to confirm the correct reading frame. Yeast transformants were also tested for accumulation of the desired Tra protein by Western blot, using the flu virus haemagglutinin (HA) epitope encoded by the GAL4 segment of pAS1CYH2 (Harris et al., 1999).

Yeast transformants were tested for transcriptional activation in the absence of any prey plasmid.

Two-hybrid analyses

These studies were exactly as described by Harris et al. (1999) except for the bait plasmids.

Antibody production

Antibodies were raised in rabbits against Tra polypeptides produced in E. coli. tra segments were subcloned from plasmids obtained by two-hybrid analysis into pWP201HPC4 (traK codons 27–200; traB codons 147–475) or the maltose-binding protein fusion vector pMal-c2 (New England Biolabs) (traV codons 21–171). TraB and TraK polypeptides were purified by affinity chromatography over HPC4 antibody columns (Rezaie et al., 1992). MBPTraV polypeptide was prepared as inclusion bodies (Tear et al., 1996). Rabbits were boosted (100 µg of protein) on days 14, 21 and 49 after immunization. Sera were collected on day 56.

All antibodies were affinity purified. Tra polypeptides were separated by electrophoresis through 4–20% polyacrylamide gradient gels. After transfer to polyvinylidene difluoride (PVDF) paper and blocking as for Western blots, the strip containing the Tra polypeptide was cut out and incubated at 4°C for 3–18 h with 0.5 ml of immune serum diluted 1:1 with TBS. The strips were washed with TBS and antibodies eluted in 0.1 M Tris–glycine buffer, pH 2.85, followed immediately by neutralization with 1 M Tris-HCl, pH 8.8.

Biochemical fractionations

Cells were fractionated into shock fluid, cytoplasm and crude membranes as described by Manoil and Beckwith (1985). Inner and outer membranes were prepared as described by Osborn et al. (1972).

Western blots

These were carried out as described previously (Paiva and Silverman, 1996). Sample volumes were scaled to about 2 × 107 cells (≈ 1.5 µg of protein), except for the experiment in Fig. 5, in which samples were scaled to 108 cells. Primary antibodies were diluted 500-fold, except for the experiment in Fig. 5, in which anti-TraB antibodies were diluted 250-fold. Alkaline phosphatase-conjugated goat anti-rabbit IgG secondary antibody (Jackson ImmunoResearch Laboratories) was diluted 1:2000 or 1:2500.

DNA sequence analysis of traB2

The traB2 gene was amplified from cells containing F′lac traB2 with the same primers used to construct the pAS1CYH2traB bait plasmid. The segment expected to contain the traB2 mutation, inferred from the size of the TraB2 amber fragment (see Results), was amplified with primers corresponding to tra nucleotides 4791–4811 (forward primer) and 5181–5200 (reverse primer). The 0.4 kbp amplification product was purified (Wizard PCR Preps DNA purification system; Promega Life Sciences) and subjected to automated DNA sequencing.


We are grateful to Dr Laura Frost and Dr Neville Firth for providing biological materials used in these studies. This work was supported by the National Science Foundation grant MCB-9900533 and funds provided by the Oklahoma Medical Research Foundation. P.M.S. acknowledges support from the Marjorie Nichlos Chair in Medical Research.