Interaction between the F plasmid TraA (F-pilin) and TraQ proteins


Philip M. Silverman. E-mail; Tel. (+1) 405 271 7663; Fax (+1) 405 271 3153.


Elaboration of conjugative (F) pili by F+ strains of Escherichia coli requires the activities of over a dozen F-encoded DNA transfer (Tra) proteins. The organization and functions of these proteins are largely unknown. Using the yeast two-hybrid assay, we have begun to analyse binary interactions among the Tra proteins required for F-pilus formation. We focus here on interactions involving F-pilin, the only known F-pilus subunit. Using a library of F tra DNA fragments that contained all the F genes required for F pilus formation in a yeast GAL4 activation domain vector (pACTII), we transformed yeast containing a plasmid (pAS1CYH2traA) encoding a GAL4 DNA-binding domain–F-pilin fusion. Doubly transformed cells were screened for GAL4-dependent gene expression. This screen repeatedly identified only a single Tra protein, TraQ, previously identified as a likely F-pilin chaperone. The F-pilin–TraQ interaction appeared to be specific, as no transcriptional activation was detected in yeast transformants containing pACTIItraQ plasmids and the Salmonella typhi pED208 traA gene cloned in pAS1CYH2. Two traQ segments isolated in the screen against F-pilin were tested for complementation of a traQ null allele in E. coli. One, lacking the first 11 (of 94) TraQ amino acids, restored DNA donor activity, donor-specific bacteriophage sensitivity and membrane F-pilin accumulation to wild-type levels. The second, lacking the first 21 amino acids, was much less effective in these assays. Both TraQ polypeptides accumulated in E. coli as transmembrane proteins. The longer, biologically active segment was fused to the GAL4 DNA-binding domain gene of pAS1CYH2 and used to screen the tra fragment library. The only positives from this screen identified traA segments. The fusion sites between the traA and GAL4 segments identified the hydrophobic, C-terminal domain IV of F-pilin as sufficient for the interaction. As TraQ is the only Tra protein required for the accumulation of inner membrane F-pilin, the interaction probably reflects a specific, chaperone-like function for TraQ in E. coli. Attempts to isolate an F-pilin–TraQ complex from E. coli were unsuccessful, suggesting that the interaction between the two is normally transient, as expected from previous studies of the kinetics of TraA membrane insertion and processing to F-pilin.


Gram-negative bacterial cells have evolved several mechanisms to secrete macromolecules across their protective surface layers (Hobbs and Mattick, 1993; Pugsley, 1993; Salmond and Reeves, 1993; Christie, 1997; Hueck, 1998; Russel, 1998). All such mechanisms have certain features in common, including multicomponent surface structures, chaperones to regulate protein–protein interactions, and metabolic energy to drive assembly and function (Finlay and Falkow, 1997). The conjugal transfer of DNA from one Gram-negative bacterial cell to another requires two such mechanisms, the first to elaborate conjugative pili (F-pili for F+ strains of Escherichia coli ), which are required to establish the stable and secure cell–cell contacts that ultimately lead to DNA transfer (reviewed by Firth et al., 1996), and the second to transport DNA and associated proteins.

F-pili, insofar as is known, are composed of a single subunit, F-pilin, which is itself derived in the inner membrane from the F plasmid traA gene product (Moore et al., 1981a). More than a dozen other Tra proteins are required for or affect pilus formation and activity. The primary structures of these Tra proteins have been inferred from the completed DNA nucleotide sequence of the F plasmid DNA transfer (tra) region (Frost et al., 1994). There is also information about the localization of individual Tra proteins in or at the E. coli cell envelope and, broadly, the stage of conjugal DNA transfer at which each functions (Firth et al., 1996). However, none of this information has shown how the Tra proteins are organized at the cell surface to mediate the assembly and function of F-pili.

As knowledge of how Tra proteins are organized with respect to one another would provide an invaluable guide to further structural and functional studies of conjugal DNA transfer and its regulation, we have begun to analyse interactions among the Tra proteins required for F-pilus assembly, using the yeast two-hybrid system (Fields, 1993). We focused on F-pilus assembly for several reasons. First, the F-pilus is the only visible structure unequivocally associated with conjugal DNA transfer from F+ cells. Second, mutational analyses indicate that the F-pilus is the end point of a pathway, involving about half the Tra proteins (Frost et al., 1994; Firth et al., 1996). Finally, the tra genes encoding the entire F-pilin secretion system, and with a few exceptions only those tra genes, are already contained in a single plasmid, pTG801 (Grossman and Silverman, 1989), making possible the construction of specific libraries for analysis.

Two general strategies have employed yeast two-hybrid screens to construct protein linkage maps defined by binary interactions among sets of functionally related proteins (Fields, 1993; Evangelista et al., 1996). Bartel et al., (1996) used a library of DNA fragments in both bait and prey plasmids to identify interactions among bacteriophage T7-encoded proteins. This approach, however, suffered from the problem of transcriptional activation by a relatively high fraction (several per cent) of bait library plasmids in the absence of any prey plasmid. Fromont-Racine et al., (1997), in an analysis of pre-mRNA splicing factors, chose to begin with a single gene as bait to screen a yeast DNA fragment prey library. Preys selected in this first screen were then used as baits to select additional preys, and so on. For several reasons, we have adopted this second, iterative approach. First, this approach allows us to build a Tra protein linkage map systematically, as all the interactors derived from a given bait are, of necessity, linked. Further, the iterative approach is self-verifying, in that if A as bait identifies B as prey then B as bait should identify A as prey. Finally, this approach is very well-suited to the relatively small number of tra genes involved in F-pilus formation.

Here, we describe the results that we have obtained, using the F-pilin coding sequence as bait to screen a library of tra fragments. We have detected, and partly characterized, a two-hybrid interaction that reflects an early, perhaps the first, stage in the formation of F-pili, the accumulation of F-pilin monomers in the bacterial inner membrane. The interaction between the TraA (F-pilin) and TraQ proteins appears to be transient in E. coli, and difficult to detect using biochemical means. Hence, two-hybrid and similar approaches are likely to play an important role in the analysis of Tra protein interactions.


Library preparation from pTG801

As we wished to focus on the formation of F-pili, we used pTG801 as our source of tra DNA. pTG801 is a pUC replicon, containing the F tra genes required for cells to elaborate functional F-pili, and, with few exceptions (e.g. traN ), only those genes (Grossman and Silverman, 1989) (Fig. 1). Plasmid DNA was sheared to a mean fragment size of 1 kb. Blunt-ended DNA fragments were ligated to a linker, and introduced into the unique BamHI site of pACTII, a yeast GAL4 activation domain plasmid (see Experimental procedures) (Durfee et al., 1993). Ampicillin-resistant transformants of E. coli, strain DH5α, were pooled and stored at −80°. This pACTII prey library consisted of 120 000 transformants. A bait library in the GAL4 DNA-binding domain plasmid pAS1CYH2 (45 000 transformants) (Durfee et al., 1993) was constructed in the same way.

Figure 1.

. Plasmid pTG801. The plasmid (23 kb) includes the 19.5 kb of F-plasmid tra DNA, comprising EcoRI fragments f1, f15 and part of f6 (Grossman and Silverman, 1989). The tra genes of pTG801 (upper case characters) (Frost et al., 1994; Firth et al., 1996) are sufficient for E. coli cells with pTG801 to elaborate functional F-pili (Fpl+ phenotype) in the presence of the lac inducer IPTG (Grossman and Silverman 1989; Grossman et al., 1990). F trb genes are not indicated for the purposes of clarity (Firth et al. 1996). pTG801 cells are Tra, owing to the absence of other tra genes essential for DNA transfer. Cross-hatched and stippled segments indicate vector DNA.

Use of the BamHI site allowed fragments to be cloned in either orientation, an important consideration because not all tra genes are transcribed in the same direction (Frost et al., 1994). When this is taken into account, along with the facts that tra DNA constituted 85% of pTG801 and that 75% of the plasmids had inserts, we calculate that the pACTII library contains about one transformant per tra base pair in the correct orientation for expression. Statistically, there is a 95% probability of a fusion site within any given tra codon among the 120 000 pACTIItra plasmids constituting the library.

When the pACTIItra library was introduced into yeast strain Y190 in the absence of a binding domain plasmid, we found no His+ Lac+ colonies among 12 000 pACTIItra transformants. In contrast, and consistent with other reports (Bartel et al., 1996), we observed that 4% of the pAS1CYH2tra library plasmids activated HIS3 and lacZ transcription in the absence of any activation domain plasmid. To avoid this high background, we used individual tra genes as bait.

Construction and characterization of pAS1CYH2traA

To construct the traAGAL4 DNA-binding domain fusion, we used a 240 bp traA polymerase chain reaction (PCR) fragment from the traA+ plasmid pTG101 (Grossman and Silverman, 1989). The fragment, encoding the entire F-pilin polypeptide with five amino acids from the TraA leader peptide at the N-terminus, was cloned into pAS1CYH2. Plasmid DNA was isolated from E. coli transformants and sequenced to confirm that the fusions were in frame and that there were no mutations in the traA segment. The plasmid (pAS1CYH2traA) was used to transform yeast strain Y190; Trp+ transformants were His and LacZ, in the absence of any activation domain plasmid. Synthesis and accumulation of the fusion protein were confirmed by Western blot with antibodies against the flu virus haemaglutinin (HA) epitope of the GAL4 segment of the fusion protein (Durfee et al., 1993). HA antibody-reactive material from pAS1CYH2traA cells showed a decrease in electrophoretic mobility relative to that from pAS1CYH2 cells. The relative mobilities corresponded to the expected increase in molecular mass owing to the TraA segment (Fig. 2, lanes 1 and 2). Polypeptides with the same electrophoretic mobility from pAS1CYH2traA cells also reacted specifically, albeit weakly, with antibodies against F-pilin (not shown). We routinely observed multiple, closely spaced bands in these experiments, presumably the result of limited proteolysis during the preparation or storage of extracts.

Figure 2.

. Expression of GAL4–F-pilin fusion proteins in yeast. Two millilitres of an overnight culture were added to an equal volume of ice-cold 50 mM Tris-HCl (pH 7.5)/10 mM NaN3. The cells were collected by sedimentation, suspended in 0.03 ml of electrophoresis sample buffer and heated to 100°C for 3 min. Sufficient acid-washed glass beads (450–500 μm, Sigma Chemical) were added to reach the surface of the liquid (≈ 0.1 g). The mixture was agitated with a vortex mixer for 2 min. Seventy microlitres of electrophoresis sample buffer was added and the mixture was agitated briefly with a vortex mixer and heated for 1 min to 100°C. Ten microlitres of each sample was then analysed by Western blot using antibodies against the HA epitope, as described in Experimental procedures. Lane 1, Y190/pAS1CYH2 (GAL4 DNA-binding domain(HA) polypeptide); lane 2, Y190/pAS1CYH2traA (GAL4 DNA-binding domain(HA)–F-pilin polypeptide); and lane 3, Y190/pACTIItraA (GAL4 activation domain(HA)–F-pilin polypeptide). Molecular masses are those predicted for the respective polypeptides. Note the immunoreactive material with higher molecular mass in lane 3 (left bracket).

Screen for Tra protein–F-pilin interactions

Y190/pAS1CYH2traA was transformed with the pACTIItra prey library. Trp+ Leu+ His+ colonies appearing by day 5 were tested for lacZ expression by colony lift. His+ LacZ+ positives were obtained at a frequency of 0.6 × 10−4 per Leu+ Trp+ transformant (Table 1). All 24 Leu+ Trp+ His+ lacZ+ colonies originally detected remained His+ and LacZ+ upon retesting. Single colonies were patched to YPDA plates, incubated at 30°C, and replica plated on to medium containing leucine and cycloheximide to select for cells that had lost the pAS1CYH2 bait plasmid, which confers cycloheximide sensitivity to the otherwise resistant Y190, but had retained the LEU2+ pACTII prey plasmid. Each pACTII plasmid was then isolated from Cyhr yeast derivatives and used to transform E. coli. Plasmid DNA isolated from E. coli was analysed by restriction endonuclease cleavage for the presence of an insert, using the linker NotI and HindIII sites. To screen out false positives, plasmids with inserts were used to transform Y190/pAS1CYH2traA, Y190/pAS1CYH2 (no insert) or pAS1CYH2 with each of four control inserts (encoding, respectively, CDK2, p53, SNF1 and lamin (Bartel et al., 1993).

Table 1. . Analysis of TraA–TraQ interactions by yeast two-hybrid assay.Thumbnail image of

All 24 His+ lacZ+ isolates became His LacZ upon loss of pAS1CYH2traA. Of these, 11 had pACTII derivatives with inserts detectable by gel electrophoresis. These plasmids rendered Y190/pAS1CYH2traA His+ and LacZ+, but not Y190 with pAS1CYH2 or any of the pAS1CYH2 plasmids with control inserts. All 11 were sequenced from GAL4 to determine whether the insert was a tra fragment and, if so, which tra gene was represented. All 11 isolates identified traQ fusions, which occurred at five sites within a 10 codon segment near the 5′ end of the gene (Fig. 3). At the nucleotide level, the fusion sites were not all separated by an integral number of codons, owing to multiple or frameshifted linker segments in some of the isolates. All but one of the isolates were in frame with respect to the GAL4 sequence. The exception was one of two otherwise identical fusions containing different numbers of linker repeats.

Figure 3.

. Fusion sites for 11 GAL4(HA)–TraQ fusion proteins. The abscissa shows the amino acid sequence of TraQ (Wu and Ippen-Ihler, 1989), and the ordinate shows the number of fusions at the indicated amino acids. The data were derived from DNA sequence analysis of pACTIItraQ plasmids isolated in a two-hybrid screen with a pAS1CYH2traA bait. Some isolates with the same fusion site are independent; the six plasmids in which traQ begins at codon 21 (serine; fusion site 4) represent at least three independent isolates, judging from the different insert sizes.

For several of the original His+ LacZ+ isolates, inserts were not detected by restriction analysis (see above). Three of these were analysed by PCR, using one primer complementary to the 3′ terminus of traQ and the other to the 3′ terminus of the GAL4 activation domain coding sequence. Two of the three yielded small PCR products, indicating that most of the remaining plasmids picked up in this screen also contained GAL4–traQ fusions.

Biological activity of N-terminal deletions of TraQ

All 11 traQ fusions that we sequenced mapped between traQ codons 11 and 21 (out of 94 total) (Fig. 3). To determine the effect of N-terminal deletions on biological activity, the traQ segments comprising codons 12–94, designated traQ1, and codons 21–94, designated traQ5, were amplified from the repective pACTIItraQ plasmids so as to include the HA epitope, and cloned into pUC18 in frame with the first 8 lacZ codons. The resultant plasmids, pUC18traQ1 and pUC18traQ5, were used to transform E. coli strain XK1200/F′traQ238::kan (Kathir and Ippen-Ihler, 1991). The traQ1 plasmid restored sensitivity to donor-specific bacteriophages R17 and M13 and donor activity to levels similar to that of XK1200/F′traQ238::kan containing the traQ+ plasmid pWP1001 (Table 2) (Paiva and Silverman, 1996). In contrast, the traQ5 allele failed to restore donor-specific phage sensitivity by spot test. Donor activity in pUC18traQ5 cells was significantly higher than in the absence of any traQ plasmid, but remained 100-fold less than in traQ+ cells (Table 2).

Table 2. . Complementation of the traQ238::kan allele by traQ1 and traQ5. a. The data are from two separate experiments. One hundred per cent activity corresponded to 0.42 and 0.16 transconjugants per donor cell in a 45 min mating respectively (see Experimental procedures).b. Assayed by spot test.Thumbnail image of

Membrane F-pilin pools correlated well with these biological assays. The TraQ1 polypeptide restored membrane F-pilin accumulation to wild-type levels, whereas no F-pilin was detected in membrane fractions from pUCtraQ5 cells (Fig. 4). The effect of the TraQ5 polypeptide on donor activity (Table 2) suggests that some F-pilin accumulated in these cells, but only to a low level not detectable by Western blot or insufficient to establish a membrane pool.

Figure 4.

. Accumulation of membrane F-pilin in the presence of N-terminal TraQ deletions. Cultures of E. coli strain XK1200/F′lac traQ238::kan, or the same strain with pWP1001 (traQ+), pUC18traQ1 or pUC18traQ5, were grown to an OD600 = 0.7–0.9 in LB medium supplemented with antibiotics as appropriate. A volume corresponding to four OD units was subjected to centrifugation, the supernatant fluid was removed by aspiration, and the cells fractionated into membrane, shock fluid and cytoplasmic fractions, as described in Experimental procedures. Each fraction was then analysed for F-pilin by Western blot. Lane 1, whole cell extract, XK1200/F′lac traQ238::kan; lanes 2–5, 6–9 and 10–13, the same strain with pWP1001, pUC18traQ1 and pUC18traQ5 respectively; lanes 2, 6 and 10, whole cell extracts; lanes 3, 7 and 11, shock fluid; lanes 4, 8 and 12, membranes; lanes 5, 9 and 13, cytoplasm. Amounts analysed were normalized to a constant culture volume.

The HA epitope of TraQ1 was used to show that the protein fractionates with cell membranes, as reported for TraQ itself (Wu and Ippen-Ihler, 1989) (Fig. 5, lanes 1–4). Extraction of membrane-containing fractions indicated quantitative solubilization with Triton X-100 but not with 0.1 N NaOH, indicating that TraQ1 has at least one membrane-spanning segment (Fig. 5, lanes 5–7) (Steck and Yu, 1973; Jennings, 1989). Most of the TraQ5 polypeptide also fractionated with membranes, but a significant quantity (about half) of this material remained insoluble after Triton X-100 or NaOH extraction (data not shown). Thus, some TraQ5 molecules appear not to localize properly, but, in contrast to TraQ1, tend to form detergent-insoluble aggregates in vivo. Nevertheless, given that some TraQ5 does appear to localize correctly, as judged by insolubility in 0.1 N NaOH, weak complementation by pUC18traQ5 suggests that TraQ5, notwithstanding its ability to interact with F-pilin in the yeast two-hybrid assay, is functionally defective.

Figure 5.

. Fractionation properties of the TraQ1 polypeptide in E. coli. E. coli strain XK1200/F′lac traQ238::kan/pUC18traQ1was grown to an OD600 = 0.8 in LB, containing kanamycin, ampicillin and 1 mM IPTG. Cells were collected by sedimentation and suspended in 1 ml of spheroplast buffer (100 mM Tris-HCl, pH 8.0/0.5 mM EDTA (Na+)/0.5 mM sucrose). Shock fluid, membrane and cytoplasmic fractions were prepared as described (Manoil and Beckwith, 1986). Three 5 μl portions of the membrane fraction were diluted in 195 μL of Tris-HCl (pH 8.0)/1 mM EDTA (Na+). To one was added 2 μl of 10 N NaOH (final concentration = 0.1 N), to another 4 μl of Triton X-100 (final concentration = 2% by volume), and the third served as a control. After 30 min at ambient temperature, insoluble material was collected by sedimentation at 100 000 g for 30 min The pellets were dissolved in electrophoresis sample buffer and TraQ1 in each fraction determined using Western blot with anti-HA epitope antibodies. Volumes were scaled to equivalent culture volumes. Lane 1, whole cell extract; lane 2, shock fluid; lane 3, cytoplasmic fraction; lane 4, membrane fraction; lane 5, membrane fraction, diluted and sedimented; lane 6, membrane fraction extracted with 0.1 N NaOH; lane 7, membrane fraction extracted with Triton X-100. The figure indicates the mobility of an 18 kDa marker protein.

We also used strain XK1200/F′traQ238::kan, containing pUC18traQ1, to assay for stable TraA–TraQ1 complexes in E. coli by immunoaffinity chromatography. Detergent-solubilized membrane proteins incubated with or without anti-HA antibodies were passed over a Sepharose-protein A column. Material retained by the column was then assayed for TraQ1 and F-pilin using Western blot. The protein A-Sepharose quantitatively captured the TraQ1 polypeptide from membrane extracts incubated with anti-HA antibodies, but all the F-pilin was contained in the run-through fractions (data not shown). This indicates that wild-type F-pilin and biologically active TraQ1 do not form stable complexes in E. coli or, if they do, that these complexes were disrupted by the conditions of the assay.

Screen for interactions involving TraQ

To test for reciprocal interactions and to determine whether any other Tra proteins interact with TraQ, we transferred the traQ1 segment from pACTII to the pAS1CYH2 bait plasmid and used pAS1CYH2traQ1 to screen the pACTIItra library. Positives were detected at a frequency of 2.8 × 10−4 per Leu+ Trp+ transformant (Table 1). Of 42 positives, 35 were sequenced, and of these, 26 created in frame fusions with traA sequences. These were the only in frame tra fusions identified. Strikingly, all 26 traA isolates encoded domain IV of F-pilin. Some also included domain III and even part of domain II, but none included domain I or the N-terminal half of domain II (Fig. 6). The result of this screen indicates that F-pilin domain IV is sufficient for interaction with TraQ1 in the yeast two-hybrid assay.

Figure 6.

. Fusion sites for 26 GAL4(HA)–TraA fusion proteins. The data were acquired and are presented as in the legend to Fig. 3.

TraQ specificity for F-pilin

We assayed the specificity of the TraA–TraQ interaction by comparing the abilities of TraQ1 and TraQ5 to interact with F-pilin and a related conjugative pilin in the yeast two-hybrid assay (see Experimental procedures). For the comparison, we chose pED208 pilin, which is 44% identical and 58% similar to F-pilin (Finlay et al., 1986). The two pilins are similarly organized into four structural domains, two of which, domains II and IV, are hydrophobic (Fig. 7A) and, at least for F-pilin, are membrane-spanning segments in E. coli (Paiva et al., 1992; Paiva and Silverman, 1996). Helical wheel projections of domains II and IV of the two pilins identified conserved surface patches that we have suggested might be involved in intersubunit interactions (Silverman, 1997). It seemed possible that these patches might also define sites for interaction with TraQ. However, we found no evidence for interaction between either TraQ5 or TraQ1 and pED208 pilin in the yeast two-hybrid assay. Yeast strain MAV203, transformed with pAS1CYH2traApED208 and either pACTIItraQ1 or pACTIItraQ5, failed to grow on plates lacking histidine and containing 3-amino triazole at a concentration allowing growth of pAS1CYH2traAF/pACTIItraQ transformants (Fig. 7B). The pAS1CYH2traAF/pACTIItraQ transformants, but not the traApED208 transformants, were positive by colony lift assay for β-galactosidase (data not shown). Quantitative analysis of β-galactosidase activities confirmed the specificity of the TraAF–TraQ interaction (Table 3). These measurements further suggested that the TraQ1 interaction with F-pilin might be stronger than the TraQ5 interaction, as judged by the difference in the level of β-galactosidase activity; different interaction strengths were also suggested by the relative colony sizes on His drop-out plates (Fig. 7B). Such a difference might explain, at least in part, the relative activities of the TraQ1 and TraQ5 polypeptides in E. coli (Table 2). In any case, these experiments show that the F-pilin–TraQ interaction cannot be attributed to the surface patches, or to any other feature common to the pED208 and F-pilins, including the generally hydrophobic character of their respective domains IV.

Figure 7.

. Specificity of the TraQ1 and TraQ5 two-hybrid interactions for F-pilin. A. Hydropathy profiles (Kyte and Doolitle, 1982) for F and pED208 pilins. Window setting = 7. Note the similar distribution of hydrophobic (index < 0) and hydrophilic (index > 0) regions (Paiva et al., 1992; Silverman, 1997). B. His phenotype of yeast strain MAV203 with either pACTIItraQ1 or pACTIItraQ5, each paired with pAS1CYH2traAF and pAS1CYH2traApED208. Plates contained synthetic complete medium with 40 mM 3-amino triazole and were lacking Trp, Leu and His.

Table 3. . Specificity of the TraQ interaction for F-pilin. a. Cultures were grown in selective medium to an OD600 of 0.66–0.8. Enzyme activity was measured as described in Experimental procedures.Thumbnail image of

F-pilin–F-pilin interactions by yeast two-hybrid assay

In view of the fact that F-pilin manifestly interacts with itself in the formation of filaments, we were surprised that TraQ was the only interacting protein identified in the two-hybrid screen with pAS1CYH2traA. To examine F-pilin–F-pilin interactions directly, we constructed a pACTIItraA fusion by cloning a restriction fragment, comprising all the traA coding sequence of pAS1CYH2traA, from pAS1CYH2traA into pACTII. Western blot analysis confirmed that pACTIItraA transformants of Y190 synthesized and accumulated the fusion protein, though a significant amount appeared as SDS-resistant aggregates not seen with Y190/pAS1CYH2traA (Fig. 2, lanes 2 and 3). Y190 transformed with both traA plasmids remained His and LacZ. Further, Y190 containing pAS1CYH2traQ1 and pACTIItraA, also remained His and LacZ, even though the reciprocal pair yielded a strong His+ LacZ+ phenotype (Table 1; Fig. 7B[link]). We conclude that in our constructs, full-length F-pilin fused to the GAL4 activation domain, but not the DNA-binding domain, misfolds in yeast in such a way that intermolecular interactions necessary for transcriptional activation cannot occur. This effect can be attributed to domains I and II, which were conspicuously absent from pACTIItraA prey plasmids isolated in the screen with pAS1CYH2traQ1 (Fig. 6).

In contrast to observations with full-length F-pilin fusions, the shorter fusions isolated in the two-hybrid screen with pAS1CYH2traQ1 were active symmetrically. We transferred three of these traA fusions from pACTII to pAS1CYH2; all three fusions interacted with the pACTIItraQ1, as judged by the His+ LacZ+ phenotype of transformants (not shown). We also tested several of the pACTIItraA plasmids isolated in the screen against pAS1CYH2traQ1 with pAS1CYH2traA, all with negative results.


These experiments provide the first evidence for direct, specific interactions between the F plasmid-encoded TraA (F-pilin) and TraQ polypeptides. Although our data were acquired with the yeast two-hybrid assay, they are consistent with data regarding the role of TraQ in the accumulation of membrane F-pilin in E. coli (Moore et al., 1981b; Maneewannakul et al., 1993) and the importance of specific F-pilin domains for such accumulation (Paiva and Silverman, 1996). For example, the TraQ requirement for accumulation of membrane F-pilin in E. coli involves domains of F-pilin itself, as opposed to the 51 amino acid TraA leader peptide (Majdalani et al., 1996). Our data are completely consistent with this evidence, insofar as the GAL4–TraA fusion protein used in our experiments lacked all but five of these 51 amino acids, and we found that F-pilin domain IV is sufficient for the two-hybrid interaction with the GAL4–TraQ fusion polypeptide. The evidence that TraQ interacts with domain IV of F-pilin is also consistent with the observation that membrane insertion of TraA–PhoA fusion polypeptides in E. coli was facilitated by the presence of other tra genes in the same cells only when the TraA segment included domain IV (Paiva and Silverman, 1996). Finally, the observation that pED208 traA failed to interact with the F TraQ polypeptide in the yeast two-hybrid assay could explain the failure of pED208 traA to complement an F traA mutant in E. coli (Finlay et al., 1986).

Wu and Ippen-Ihler (1989) suggested that TraQ is an inner membrane protein whose C-terminal segment is likely to interact with TraA. In agreement with that suggestion, the solvent extraction data presented above indicate that TraQ has at least one membrane-spanning segment. However, assuming that the TraQ–TraA interaction occurs in the E. coli cytoplasm, as suggested by Wu and Ippen-Ihler (1989), the polar C-terminal segment of TraQ, with 60% charged or polar amino acids (Wu and Ippen-Ihler, 1989), would be an unlikely binding partner for the hydrophobic domain IV of F-pilin. Moreover, the ‘positive inside’ rule (Von Heijne, 1992) favours a TraQ orientation with a cytoplasmic N-terminus, and all 11 TraQ fusions, representing five different break points, isolated in the screen against pAS1CYH2traA mapped to a 10-codon segment, beginning at traQ codon 12. Were the C terminus of TraQ responsible for interaction with F-pilin, we might have expected shorter traQ segments from the screen. Finally, TraQ amino acids 12–20, though not required in the yeast two-hybrid assay, evidently include residues that are important for TraQ function in E. coli. For all these reasons, we suggest that TraQ is probably a bitopic membrane protein (Blobel, 1980) oriented with its N-terminus in the cytoplasm and its polar C-terminus in the periplasmic space. A monotopic topology, with both termini cytoplasmic, remains as another possibility, though such proteins are rare (Jennings 1989).

The data regarding TraQ and accumulation of membrane F-pilin can be summarized as a TraQ cycle (Fig. 8). We propose that TraQ binds directly and specifically to TraA domain IV while the latter is synthesized. We speculate that this binding prevents an interaction involving F-pilin that blocks further processing and leads to TraA degradation (Maneewannakul et al., 1993; Paiva and Silverman, 1996). We further propose that the TraA–TraQ interaction is transient, with domain IV dissociating from TraQ and partitioning into the membrane, leaving TraQ free to participate in another cycle. The data presented in this communication provide evidence for a direct and specific interaction between TraA domain IV and TraQ, and show that TraQ is itself an intrinsic membrane protein. The results of Maneewannakul et al., (1993) strongly suggest that TraQ acts catalytically (Paiva and Silverman, 1996), as the cycle requires. If so, TraQ–TraA complexes must dissociate to maintain a pool of free TraQ. This dissociation must be rapid to account for the kinetics of TraA processing to F-pilin catalysed by host leader peptidase B (Majdalani and Ippen-Ihler, 1996). The transient character of the interaction in E. coli might explain our failure to isolate TraQ–TraA complexes from cells. Finally, preliminary experiments suggest that F-pilin domain IV, at the C-terminus of an otherwise soluble polypeptide, inserts into the cell membrane in the absence of any other tra genes, including traQ (R. L. Harris and P. M. Silverman, unpublished observations). This result indicates that domain IV by itself partitions into the bacterial inner membrane, and suggests that interaction with TraQ might serve to delay this reaction until other parts of the TraA molecule have properly localized. In any event, each turn of the TraQ cycle produces an F-pilin monomer stably incorporated into the cell's inner membrane (Silverman, 1997). The F-pilin secretion/assembly process is arrested at this stage unless the cells are expressing other tra genes (Sowa et al., 1983).

Figure 8.

. The TraQ cycle for accumulation of inner membrane F-pilin. TraQ (shaded) is depicted as a bitopic inner membrane protein with its N-terminus (horizontal segment) cytoplasmic. The depiction of TraQ in the figure is not meant to imply a particular domain organization. Interaction with TraA domain IV (panel A) is required for proper membrane insertion of preceding TraA segments. This interaction is transient, with domain IV itself partitioning into the membrane (panel B), thereby releasing TraQ for another cycle. Cleavage between TraA leader peptide and domain I (panel B, downward arrowhead) completes the formation of membrane F-pilin (panel C). See the Discussion for further details.

As proposed (Fig. 8), the TraQ cycle does not involve any tra gene products other than TraA and TraQ. This is based partly on the results of Ippen-Ihler and co-workers, showing that TraQ was necessary and sufficient for cells to accumulate membrane F-pilin (Moore et al., 1981b; Maneewannakul et al., 1993), and partly on the yeast two-hybrid results presented above. Quantitatively, however, traQ alone was not as effective as F in restoring membrane F-pilin levels, suggesting that efficient operation of the TraQ cycle requires participation of other Tra proteins (Paiva and Silverman, 1996). This is still an open question, as such other proteins have not been identified, nor is it clear at what stage(s) of the cycle they might participate.

Although the TraQ cycle assigns a chaperone function to TraQ, this function differs from those of other secretion chaperones, such as the cytoplasmic Syc proteins of Yersinia spp. (Hueck, 1998), and the periplasmic chaperones active in chaperone/usher pilus assembly pathways (Hung and Hultgren, 1998). Those chaperones form stable, soluble complexes with their cognate target proteins, holding them until appropriate inner or outer membrane acceptor sites are found (Hultgren et al., 1991; Wattiau and Cornelis, 1993; Frithz-Lindsten et al., 1995). In contrast, TraQ, itself a membrane protein, need contact TraA only very briefly as TraA inserts into the inner membrane. Inner membrane phospholipids might hold F-pilin in a suitable conformation until other Tra proteins become available to mediate the later stages of pilus assembly. This could explain why we found no tra prey plasmids, other than traQ, with a traA bait; perhaps other protein–protein interactions involving F-pilin are specific for membrane F-pilin. Another possibility, not mutually exclusive of the first, is that such interactions require more than one other Tra protein. Using different tra bait plasmids, we have defined two other Tra linkage groups in addition to the TraA–TraQ group described above (Harris and Silverman, unpublished data). One or the other of these might be organized into a multiprotein complex in the E. coli cell envelope, and function to escort membrane F-pilin further along the pathway to filament assembly.

Experimental procedures

Biological materials and media

Yeast strain Y190 (MATa gal4 gal80 his3 trp1-901 ade2-101 ura3-52 leu2-3,-112 cyhrGAL1::lacZ@URA3 GAL1::HIS3@ LYS2) and plasmids pACTII and pAS1CYH2 were obtained from Dr Steven Elledge, Baylor University College of Medicine. pACTII is a GAL4 activation domain plasmid similar to pACT (Durfee et al., 1993), except for more extensive cloning sites in pACTII. It includes the yeast LEU2 gene and ori and amp genes for replication and selection in E. coli. pAS1CYH2 is similar to pAS1 (Durfee et al., 1993), except for the addition of the yeast CYH2 gene, which confers a dominant cycloheximide-sensitive phenotype. pAS1 plasmids include the yeast TRP1 and E. coli ori and amp genes. Yeast strain MAV203 (relevant genotype: GAL1::LACZ GAL1::HIS3@URA3) is a MATα derivative of MAT103 (Vidal et al., 1996), and for the purposes of the present studies was used interchangeably with Y190. E. coli K-12 strain DH5α was from our laboratory collection, except that electroporation-competent cells were obtained from Life Technologies. E. coli plasmid pTG801 has been described (Grossman and Silverman, 1989; Grossman et al., 1990).

Yeast YPD and synthetic complete (SC) drop-out media were as described (Kaiser et al., 1994). Yeast strains were routinely grown at 30°C with aeration; growth was monitored by total cell count or by optical density at 600 nm. E. coli was routinely cultured in Luria–Bertani (LB) medium at 37°C with aeration. Growth was monitored by optical density at 600 nm. Antibiotics were added at concentrations of 100 μg ml−1 (ampicillin), 50 μg ml−1 (kanamycin), 25 μg ml−1 (chloramphenicol) or 10 μg ml−1 (tetracycline).

Preparation of a tra DNA library

pTG801 DNA (100 μg per 5 ml of TE buffer) was sheared to a mean fragment size (mass average) of 1 kb using two passes through a French pressure cell at 4°C and 3000 psi. After concentration by Centricon filtration and ethanol precipitation, the sheared DNA (42 μg in 50 μl of TE buffer) was treated with Klenow fragment (25 U) for 10 min at 37°C, and then for 60 min at 25°C after the addition of the four deoxyribonucleotide triphosphates (100 μM each). After purification by spun-column chromatography (Pharmacia S300 resin) and ethanol precipitation, the DNA (23 μg in 50 μl of TE buffer) was treated at 14°C for 30 min with T4 DNA polymerase (40 U) in the presence of all four deoxyribonucleotide triphosphates as above, and purified again by spun-column chromatography and ethanol precipitation. Purified, blunt-ended DNA (14 μg in 50 μl of TE buffer) was then ligated to linker oligonucleotide added at 100-fold molar excess over sheared pTG801 DNA, assuming a 1 kb fragment size. Incubation with T4 DNA ligase (40 U) was for 24 h at 10°C, 18 h at 15°C (after the addition of an additional 40 U of enzyme), and 2 h at 25°C. The linker (5′-GATCGCGGCCGCAAGTTCC/3′-CGCCGGCGTTCGAAGGp) included internal NotI and HindIII sites, one blunt terminus with a 5′ phosphoryl group and one 5′ overhang compatible with a BamHI site. One copy of the linker added 16 bp when cloned into vector BamHI sites. For reasons noted below, the primers were designed so as not to generate a BamHI site upon ligation into a BamHI-digested vector; thus, the 5′ terminal sequence was GATCG instead of GATCC. The final DNA preparation was purified by PEG precipitation (Fulton et al., 1995), spun-column chromatography and ethanol precipitation. The final yield was 6.3 μg.

Library DNA fragments were ligated into the unique BamHI site of the yeast GAL4 activation domain plasmid pACTII, in a reaction containing BamHI-digested pACTII (1.3 μg) and library DNA (0.5 μg). BamHI (30 U) was included in the ligation reaction to maintain pACTII in linear form during the course of the ligation. As the linker used with the library fragments contained a degenerate BamHI site (see above), cloned library fragments were immune to BamHI digestion. This increased transformants μg−1 vector DNA about threefold. After ligation, the DNA was used to transform E. coli strain DH5α (Life Technologies) by electroporation. Ampicillin-resistant transformants were pooled and stored at −80°C.

Total library plasmid DNA (3 μg) isolated from E. coli was used to transform yeast strain Y190. Leu+ transformants were tested for His+ LacZ+ phenotype, using replica plating to Leu His plates that contained 40 mM 3-aminotriazole [3AT; added to reduce HIS3 (imidizole glycerol phosphate dehydratase) activity in Y190 in the absence of GAL4-dependent activation] and by colony lift respectively (Durfee et al., 1993). Colonies appearing within 5 days at 30°C were considered to be His+, and those yielding blue colour after 18 h at 41°C were considered to be LacZ+.

Construction of pAS1CYH2traA and related plasmids

To construct the GAL4 DNA-binding domain fusion with F-encoded traA, we used a 240 bp traA PCR fragment from the traA+ plasmid, pTG101 (Grossman and Silverman, 1989). The forward primer (CTTCCCGCAGCTGGCGATGGC) overlapped a PvuII site 5 codons before the end of the traA+ leader peptide, and the reverse primer (CTCGTCTGTCGACATCGTTT TATTTCC) introduced a SalI site 15 bp after the traA termination codon. The fragment was digested with PvuII and SalI and cloned into pAS1CYH2 that had been digested with BamHI, filled-in with Klenow fragment and digested with SalI. This preserved the correct reading frame between the GAL4 and traA segments. The result was a GAL4–traA fusion, encoding the entire F-pilin polypeptide with five additional amino acids at the N-terminus. DNA sequencing confirmed that the fusion was in frame and that there were no mutations in the traA segment.

PACTIItraA was constructed by isolating traA from pAS1CYH2traA as a SalI/NcoI fragment. PAS1CYH2traA was first digested with SalI, the ends were filled in with Klenow fragment and the traA fragment was liberated by digestion with NcoI. The purified fragment was cloned into pACTII digested with SmaI and NcoI. Candidate pACTIItraA plasmids were isolated from E. coli transformants and characterized by restriction enzyme and DNA sequence analyses.

pAS1CYH2traApED208 was constructed using PCR amplification of a 238 bp fragment from the pED208 traA+ plasmid pBF152 (Finlay et al., 1986), which we obtained from Dr Laura Frost, University of Alberta. The forward primer (GTG GGACATATGGCCAGTGC) encoded seven amino acids of the traA leader peptide and included a G→T mutation to introduce an NdeI site. The reverse primer (GTCTAAATCGGA TCCCTCCATATC) introduced a BamHI site. The fragment was digested with NdeI and BamHI and cloned into pAS1CYH2 digested with the same two enzymes. Candidate plasmids were isolated from E. coli transformants and characterized using restriction enzyme and DNA sequence analyses.

Construction of pAS1CYH2traQ and pUC18traQ plasmids

traQ fragments were transferred from pACTII to pAS1CYH2 by digesting the pACTIItraQ plasmids with EcoRI, filling in with Klenow fragment and liberating the traQ fragments by digestion with NcoI. The purified fragments were then cloned into pAS1CYH2 digested with NcoI and SmaI. Candidate plasmids were isolated from E. coli transformants and characterized using restriction enzyme and DNA sequence analyses. Accumulation of the Gal4–TraQ polypeptides in yeast was confirmed by Western blot, using the HA epitope, as shown in Fig. 2 for the GAL4–F-pilin fusion polypeptide. To construct pUC18traQ plasmids for expression in E. coli, traQ fragments were amplified by PCR from pAS1CYH2traQ derivatives. The forward primer (CGCCGGAATTCGATG GCTTACTT) included a pAS1CYH2 EcoRI site just 5′ to the segment encoding the HA epitope (Durfee et al., 1993); the primer also contained a + 1 frameshift mutation to correspond to the pUC18 lacZ reading frame. The reverse primer (CTGCACTGCAGCGGCGCTC) corresponded to tra nucleotides 16 894–16 912 (Frost et al., 1994) and contained a T→A mutation to create a PstI site. The PCR fragment was digested with EcoRI and PstI and cloned into pUC18 digested with the same two enzymes. Candidate plasmids were isolated from E. coli transformants and characterized using restriction enzyme and DNA sequence analyses.

Other materials and methods

Restriction fragments and PCR products were purified from agarose gels by subjecting frozen (liquid N2) gel segments to centrifugation (10 min at 13 000 g) through siliconized glass wool.

Yeast transformations were carried out by the lithium acetate/PEG method, as has been described (Kaiser et al., 1994); transformation frequencies with the pACTtra library were generally 104–105 Leu+ Trp+ colonies per μg of DNA. Plasmid DNA was prepared from zymolase-treated yeast cells (Kaiser et al., 1994) and introduced into E. coli by electroporation. Otherwise, E. coli was transformed by the CaCl2 method.

Western blots were carried out as described by Paiva et al., (1992), using 12% or 15% polyacrylamide gels. Anti-HA antibodies (mouse monoclonal line 12CA5; Boehringer Mannheim) at 6.6 mg of IgG per ml was diluted 1:2000. Goat anti-mouse IgG (alkaline phosphatase conjugate; 1 mg ml−1) was obtained from Promega Life Sciences and used at a 1:2000 dilution.

Bacteriophage sensitivities were determined by spotting 5 μl of different bacteriophage dilutions, containing, respectively, 50 000, 500, 50 and 5 pfu, on an agar overlay plate seeded with the strain to be tested. Sensitive strains showed confluent lysis or individual plaques at all dilutions. Resistant strains showed no lysis, except at the lowest dilution in which turbid lysis was occasionally evident. The spherical RNA bacteriophage R17 was from our laboratory stock and M13mp18 was obtained from Invitrogen.

Conjugal DNA transfer was measured as described (Paiva and Silverman, 1996) with AE2280 (Silverman et al., 1991) as the recipient strain. Donor and recipient strains (OD600nm = 0.5) were mixed in a 1:4 volume ratio. KanR StrR transconjugants were determined after a 45 min transfer interval. Donor cells were measured at the same time as KanR CamR or KanR AmpR colony-forming units, depending on the plasmid content of the donor strain.

β-Galactosidase activities in yeast were measured using chlorophenyl-red-β-d-galactopyranoside (CPRG; Boehringer Mannheim), essentially as described by Durfee et al., (1993). Cells in 5 ml of each culture to be assayed at an OD600 = 0.6–0.8 were collected by sedimentation, suspended in 1 ml of H buffer (Durfee et al., 1993), permeabilized with SDS and CHCl3 and assayed. Chlorophenyl-red released by hydrolysis was measured by absorbance at 574 nm.


The authors acknowledge support from NSF grants MCB 95-07089 (M.E.D.) and MCB-9900533 (P.M.S.) and the Oklahoma Medical Research Foundation (P.M.S.). P.M.S. also acknowledges support from the Marjorie Nichlos Chair in Medical Research.