An Aqueous Channel for Filamentous Phage Export

See allHide authors and affiliations

Science  28 May 1999:
Vol. 284, Issue 5419, pp. 1516-1519
DOI: 10.1126/science.284.5419.1516


Filamentous phage f1 exits its Escherichia coli host without killing the bacterial cell. It has been proposed that f1 is secreted through the outer membrane via a phage-encoded channel protein, pIV. A functional pIV mutant was isolated that allowedE. coli to grow on large maltodextrins and rendered E. coli sensitive to large hydrophilic antibiotics that normally do not penetrate the outer membrane. In planar lipid bilayers, both mutant and wild-type pIV formed highly conductive channels with similar permeability characteristics but different gating properties: the probability of the wild-type channel being open was much less than that of the mutant channel. The high conductivity of pIV channels suggests a large-diameter pore, thus implicating pIV as the outer membrane phage-conducting channel.

The pIV protein is one of three filamentous phage proteins that are not part of the f1 virion but are required for phage export from the host bacterium. Interest in pIV has been stimulated by its sequence similarity to proteins in the type IV pilus assembly and in transport pathways, including type II and type III secretion systems (1). Both of these complex secretion systems mediate the export of proteins in Gram-negative bacteria. In type II secretion, toxins or degradative enzymes are secreted into the extracellular milieu; in type III secretion, proteins are secreted and injected directly into the cytosol of eukaryotic host cells, causing cytotoxicity. Bacteria with type II or type III secretion systems include such notorious animal and plant pathogens asYersinia, Salmonella, Shigella, andErwinia, all of which express a pIV homolog necessary for secretion or virulence. Although it has been postulated that pIV and its homologs function as outer membrane channels, there has been no direct evidence to support this hypothesis.

The pIV protein exists as a large homomultimer in the outer membrane ofE. coli. Purified multimers are large cylindrical structures, as viewed by scanning transmission electron microscopy (STEM) (2). The filamentous phage is approximately 1 μm long with a diameter of 6 to 7 nm. A simple diffusion pore 6 to 7 nm in diameter would cause E. coli to be very sensitive to external stresses. However, phage-infected E. coli maintain long-term viability. Thus, if pIV were to form a channel, it would most likely be opened only during phage export by a gating mechanism.

We used two assays to test whether pIV increased the permeability of the E. coli outer membrane: sensitivity to large antibiotics and growth on large carbohydrates. Vancomycin is an antibiotic that cannot cross the bacterial outer membrane because of its hydrophilicity and large size [molecular weight (MW) = 1449]. Wild-type pIV (pIV+) and pIV with a point mutation at Ser324→ Gly324 (pIVS324G), which still functioned for phage export, were synthesized from plasmids at slightly lower levels than in phage-infected cells (3). Expression of pIVS324G substantially increased the sensitivity ofE. coli to vancomycin, whereas pIV+ did not (Fig. 1A) (4). Even without vancomycin, bacteria with pIVS324G did not grow well. However, their growth defect was fully rescued by the addition of 20% sucrose, an osmoprotectant that does not cross the inner membrane, further suggesting that pIVS324G makes the outer membrane permeable (3). The pIVS324G made bacteria sensitive to concentrations of vancomycin 100-fold lower than those affecting envA and tolQ mutants, which are known to have leaky outer membranes (5). Experiments with bacitracin (MW = 1411) gave similar results as the vancomycin experiments (3).

Figure 1

Vancomycin sensitivity and growth on maltodextrins. (A) Serial dilutions of E. colistrain K1712 containing the indicated plasmid were plated on tryptone broth plates ± 1 mM IPTG and increasing concentrations of vancomycin. Colony-forming units (CFU) were determined after 24 hours at 37°C. (B) Cultures of MC4100 pPMR131 (lamB +, empty vector), MCR106 pPMR131 (ΔlamB106, empty vector), MCR106 pPMR132 (ΔlamB106, pIV+), and MCR106 pPMR132S324GlamB106, pIVS324G) were washed with M63 salts, and plated on minimal media containing M63 salts, 10 μM IPTG, 50 μg/ml chloramphenicol, and 0.2% (w/v) of the indicated sugar. Growth was assessed after 36 hours at 37°C. Shown is one of three similar experiments. (C) The same strains as in (B) (minus MC4100ΔlamB106, pIVS324G) were washed and plated on minimal media containing M63 salts, 1 mM IPTG, 50 μg/ml chloramphenicol, and the indicated sugar. Growth was assessed after 6 days at 37°C. Shown is one of three similar experiments.

To test if pIV increased the permeability to carbohydrates, we expressed pIV in the strain MCR106, which has a 501–base pair internal deletion of the gene lamB encoding an outer membrane maltoporin (6), and grew the strain on plates containing sugars of increasing size (Fig. 1B). In the absence of LamB,E. coli are unable to grow on maltodextrins larger than maltotriose (7). Escherichia coliexpressing low levels of pIVS324G grew on 0.2% sugars up to maltohexaose (8). They also grew on maltoheptaose, albeit poorly, when sugar concentrations were increased to 0.4% (3). Bacteria expressing pIV+ grew poorly on maltotriose and did not grow on the larger maltose sugars under these conditions.

Escherichia coli expressing pIV+ grew on maltodextrins larger than maltotriose only when both the sugar concentration and growth time were increased (Fig. 1C). Under these conditions, pIV+ conferred a growth advantage over the empty vector control. As the sugar size increased, bacteria with pIV+ required higher sugar concentrations to grow; colonies could be seen on maltohexaose only when the concentration was increased to 0.6%. The modest growth advantage due to pIV+ was not the result of mutation, because all of the plated cells formed colonies and these bacteria grew just as slowly after restreaking (9).

To directly test for channel activity in electrophysiological assays, we purified His-tagged pIV+ and pIVS324G and reconstituted them into proteoliposomes (10). The pIV proteins were purified by nickel-chelate and size exclusion chromatography (Fig. 2). Both wild-type and mutant proteins eluted in similar fractions corresponding to ∼670 kD, indicating that they exist as multimers of similar size. In addition, both forms of pIV appeared similar by negative staining electron microscopy (3). Both His-tagged proteins were functional, as assessed by their ability to function in phage export (11). Proteoliposomes with pIVS324G were fused to planar lipid bilayers. Large, single channels were observed at positive and negative voltages (+V m and –V m) (Fig. 3, A and B). Initially, an additional smaller channel was observed that had characteristics similar to those previously reported for OmpC (12). Subsequently, pIV was purified from an E. coli ompR strain, which contains low amounts of the porins OmpC and OmpF. The ratio of porin to pIV from cell lysates of theompR strain was 50 times lower than in the original strain (3), and contaminating channels were very rarely observed after purification. A recording of pIVS324G channels with a contaminating channel illustrates the markedly greater current through pIV channels (Fig. 3A). Both pIVS324G and pIV+independently displayed the same channel behavior when purified from either strain (3).

Figure 2

Purification of mutant pIV. (A) Silver-stained SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gel showing major purification steps. His-tagged pIVS324G was expressed from plasmid pPMR132S324G in strain K1312 (MC4100ompR::Tn10). Lane 1 (1× load), total cell lysate; lane 2 (4×), solubilized membranes; lane 3 (40×), elution from the Ni2+-Sepharose column; and lane 4 (40×), pooled peak fractions from the BioGelA5M column. (B) Silver-stained SDS-PAGE gel showing the elution profile of pIV from the BioGel column. Gel filtration standards are indicated. Purification of wild-type pIV was similar (3). F23 through F44 indicate the fractions analyzed.

Figure 3

Current traces of pIVS324G and pIV+. (A) Channels of pIVS324G (117 pA) are ninefold larger than a porin channel (*) at +80 mV. (B) pIVS324G channels from ompRstrain K1312. (C) pIV+ channels at various voltages. In (B) and (C), the traces were vertically displaced for clarity and marked as closed (C), or with one (O1) or two (O2) open channels. The solution for (A) was 285 mM NaCl in the cisand 150 mM in the trans chamber. In (B) and (C), 150 mM KCl was in both chambers. All solutions contained 10 mM NaHepes (pH 7.4), 5 mM MgCl2, and 5 mM CaCl2.

Most attempts to detect channels with pIVS324G were successful (n > 40). At low voltages (–20 mV <V m < +40 mV) there were occasional channel openings (Figs. 3B and 4C). At intermediate voltages (V m < –20 mV orV m > +50 mV) channels opened to two different current levels, O1 and O2, where O2 was double the size of O1. At larger voltages (V m < –80 mV or > +120 mV) only the larger O2 current level was observed. When several pIV multimers were incorporated into the bilayer, the maximum current was a multiple of the O2 current. This suggests that a pIV multimer has two conductance states or that the multimers reconstitute in pairs.

Figure 4

Current-voltage relationships of pIVS324G. (A) The average current for the closed, single open (O1), and double open (O2) pIVS324Gchannels in 150 mM salt (cis and trans) was plotted. (B) The O1 currents for pIVS324Gchannels were measured in salt gradients of 150/10, 150/150, and 150/300 mM KCl in the cis and trans chambers, respectively. Both chambers had 10 mM KHepes (pH 7.4), 5 mM MgCl2, and 5 mM CaCl2. In (A) and (B), each data point is the current average ± SD calculated from an all-points histogram from several traces of a single experiment. Shown is one of five similar experiments. (C) Probability of time in the closed state (P closed) for mutant and wild-type pIV. The voltage was increased in 20-mV steps with a return to 0 mV between each step. We calculated P closedfrom an all-points amplitude histogram of a 2-min recording for each voltage: P closed = Σt closedt closed+ t open. Experiments were done in 150 mM KCl, 10 mM NaHepes (pH 7.4), 5 mM MgCl2, and 5 mM CaCl2. Shown is one of two similar experiments.

There were three distinct effects of voltage on the activity of the channel. First, at positive potentials, the channels required a greater voltage to open than at negative potentials. Second, the single-channel conductance of pIVS324G was larger at +V m than it was at –V m. This can be seen both in the single-channel recordings (Fig. 3B) and in the current-voltage plot, where opening to the first conductance level (O1) was 1.22 ± 0.03 nS (±SD) at 80 mV and 0.90 ± 0.10 nS at –80 mV in 150 mM KCl (Fig. 4A). The asymmetric response to the polarity of the voltage suggests that pIVS324G channels reconstitute into the lipid bilayer with a common asymmetry. Third, the channel conductance increased with increasing voltage. The selectivity of pIVS324G channels was determined by two criteria: quantification of channel current and reversal potential in varying salt solutions (Fig. 4B). They were approximately four times more permeable to potassium than to chloride.

In contrast to pIVS324G, channels from pIV+(n = 7) were observed only at very high voltages (V m > 180 mV or V m< –120 mV), which made them more difficult to characterize. For a given voltage, the O1 conductance for pIV+ was less than that for pIVS324G. At +200 mV, the O1 conductance of pIV+ was 1.2 ± 0.2 nS (13) and the percentage of time open was 5% (Fig. 4C). At +80 mV, the pIVS324G channel conductance and probability of being open closely resembled that of pIV+ at +200 mV. Thus, the pIVS324G channel behaved as if the mutation shifted its voltage dependence, thereby increasing the likelihood of the channel being open at lower voltages.

The channels formed by pIVS324G and pIV+ have many features in common. (i) Both reconstitute into membranes with a common asymmetry, with a larger channel conductance at +V m than –V m. (ii) Both have similar cationic selectivity. (iii) Both have a greater probability of opening when at greater V m. (iv) Both are more likely to open when V m is negative (Fig. 4C). (v) Both channel conductances increase with increasingV m. (vi) Both are extremely large channels in comparison to known porin molecules such as OmpC, whose conductance is 110 pS at 150 mM KCl (12). The primary difference between pIVS324G and pIV+ channels is their probability of opening. This difference confirms that the channel activity is due to pIV and not a contaminant.

The pIV pore diameter is estimated to be 6 nm if it is assumed that a pIV multimer has two conductance states (14). This diameter is large enough to accommodate an extruding phage (6 to 7 nm) and is consistent with measurements of pIV pore diameter (7 to 8 nm) in the STEM (2). The pIVS324G is open much more frequently than pIV+at voltages likely to exist across the outer membrane (15). This is consistent with the growth and antibiotic sensitivity experiments, indicating that the electrophysiological recordings reflect the in situ behavior of the protein. It is also consistent with pIV+ being a tightly gated channel. Transmembrane aqueous channels have been shown to function in the transport of ions and metabolites and the translocation of DNA and unfolded proteins (16, 17). The sequence similarity between pIV and numerous proteins involved in pilus assembly or secretion of folded proteins (18) suggests that use of large, gated channels may be a general mechanism for supramolecular transport.

  • * To whom correspondence should be addressed. E-mail: russelm{at} (M.R) and simon{at} (S.M.S.)


View Abstract

Stay Connected to Science

Navigate This Article