Ligand-Specific Opening of a Gated-Porin Channel in the Outer Membrane of Living Bacteria

See allHide authors and affiliations

Science  23 May 1997:
Vol. 276, Issue 5316, pp. 1261-1264
DOI: 10.1126/science.276.5316.1261


Ligand-gated membrane channels selectively facilitate the entry of iron into prokaryotic cells. The essential role of iron in metabolism makes its acquisition a determinant of bacterial pathogenesis and a target for therapeutic strategies. In Gram-negative bacteria, TonB-dependent outer membrane proteins form energized, gated pores that bind iron chelates (siderophores) and internalize them. The time-resolved operation of the Escherichia coli ferric enterobactin receptor FepA was observed in vivo with electron spin resonance spectroscopy by monitoring the mobility of covalently bound nitroxide spin labels. A ligand-binding surface loop of FepA, which normally closes its transmembrane channel, exhibited energy-dependent structural changes during iron and toxin (colicin) transport. These changes were not merely associated with ligand binding, but occurred during ligand uptake through the outer membrane bilayer. The results demonstrate by a physical method that gated-porin channels open and close during membrane transport in vivo.

Transport through biological membranes is selective and directional, providing a regulated influx of solutes required for metabolism and an efflux of molecules that function outside the cell. Ligand-gated membrane channels—which exist in a variety of organisms and tissues, from bacteria to the human brain—form specific uptake pathways by binding solutes and moving them through transmembrane pores (1). In the outer membrane [OM (2)] of Gram-negative bacteria, TonB-dependent gated porins transport iron in the form of ferric siderophores (3). These metal chelates are too big (750 daltons) to traverse the OM through the open channels of general porins (4), and channels large enough to accommodate them pose a threat to bacteria, because they would also allow the entry of detergents, antibiotics, and other noxious molecules. Gated porins have evolved that solve this nutritional dilemma by elaborating ligand-binding surface loops that close their large channels. The acquisition of iron through ligand-gated porins is a determinant of bacterial pathogenesis (5), and the pores themselves are routes for the entry of toxins (6) and antibiotics (7). A prototypical gated porin, Escherichia coli FepA, binds the siderophore ferric enterobactin in closed surface loops (8) and transports it through an underlying transmembrane channel (9) into the cell. The latter internalization reaction requires the input of energy and the participation of another protein, TonB (10). Because the TonB NH2-terminus resides in the inner membrane (11) and TonB promotes uptake through channels in the outer membrane, TonB-mediated transport reactions involve energy transduction between two distinct bilayers.

The outer leaflet of the OM contacts the external environment, and various nutrients, noxious agents, and small molecules interact with OM proteins during the initial stages of transport (12). Because ferric enterobactin binds to the outside of a closed channel, through which it subsequently passes, conformational changes in FepA surface loops are a fundamental part of its suspected transport mechanism. To investigate this possibility, we reacted nitroxide spin labels with a genetically engineered cysteine residue in a ligand-binding surface loop [PL5 (13)] and analyzed the live bacteria by electron spin resonance (ESR) spectroscopy during transport. ESR detects the presence of unpaired electrons, and the introduction of a stable nitroxide free radical creates a molecular probe that conveys information about the structure and motional dynamics at its site of attachment. Site-directed spin labeling has generated extensive information about membrane proteins in vitro (14), including determinations of secondary and tertiary structure and characterization of conformational change. Previous ESR results with the purified mutant protein Glu280 → Cys280 (E280C) (15), labeled with either S-(-oxyl-2,2,5,5- tetramethylpyrroline-3-methyl) methane thiosulfonate (MTSL) or 4-maleimido-tempo [MAL6 (16)] and studied in detergent or liposomes (17), suggested that FepA could be specifically modified and analyzed in vivo (18). Two findings confirmed this expectation: Bacteria harboring pfepAE280C were labeled 30- to 50-fold more than the fepA host strain KDF541 or KDF541 expressing the fepA + allele that lacks the E280C mutation (Fig. 1). Furthermore, in Western blots (protein immunoblots) of cell lysates with anti-MAL6 sera, the engineered Cys residue at FepA residue 280 was the major spin-labeled site in the bacteria (Fig. 1).

Figure 1

Nitroxide labeling of FepAE280C in vivo. (A) ESR analyses of MTSL-labeled bacteria expressing FepAE280C. KDF541 (9) containing pITS449 or pFepAE280C was grown in LB medium and subcultured into T medium plus appropriate nutritional supplements at 1%. After 16 hours, bacteria were harvested and suspended in labeling buffer [50 mM Mops and 60 mM NaCl (pH 7.5)]. Spin labels were added (to a final concentration of 20 μM MTSL or 10 μM MAL6) and incubated 2 hours at room temperature with shaking, and the cells were washed twice with reaction buffer containing 0.05% Tween-20 and once with reaction buffer alone. After the labeling reaction, the viability of the bacteria was confirmed by plating on LB agar. For ESR analysis, 1010 cells were suspended in 100 μl of 0.01 M NaHPO4 (pH 6.9) containing 0.4% glucose. X-band spectra of MTSL-labeled KDF541/pFepAE280C (top), KDF541/pITS449 (fepA +; center), or KDF541 (fepA; bottom) were recorded on a Bruker EMX spectrometer, operating at 9.8 GHz with a total scan range of 150 G and 100-kHz modulation frequency, at 4°C in a quartz flat cell. Peaks a and b show the position of strongly and weakly immobilized spin labels, respectively. (B) [125I]-labeled protein A Western blots with anti-MAL6 sera. MAL6 was coupled to bovine serum albumin (BSA) and ovalbumin (OVA). New Zealand White rabbits were immunized with BSA-MAL6 conjugates [100 μg/week for 1 month (29)], and polyclonal serum was collected in week 5 and assayed for antigen specificity by enzyme-linked immunosorbent assay (30) and Western blots against BSA, BSA-MAL6, OVA, and OVA-MAL6 (lanes 1 through 4, respectively). Lysates (5 × 107 cells) of KDF541, KDF541/pITS449, and KDF541/pFepAE280C, previously labeled with MAL6, as described above, were reacted with the anti-MAL6 polyclonal serum (lanes 5 through 7, respectively). Arrows indicate the positions of (top to bottom) FepAE280C-MAL6, BSA-MAL6, BSA, and OVA-MAL6.

Although the in vivo ESR spectra of MTSL attached to residue 280 closely resembled its in vitro spectra (17), marked changes appeared during FepA-mediated transport in live bacteria. MTSL-labeled FepA showed a two-component spectrum in vitro, with the majority of its spins strongly immobilized and a small percentage (<1%) weakly immobilized (17, 19). FepAE280C-MTSL showed similar spectra in vivo at 4°C [3% weakly immobilized (peak b in Fig.2A)], but an increase in sample temperature to 37°C in the presence of the siderophore shifted a significant proportion of the spin-label population [16% in Fig. 2E; (20)] to the weakly immobilized state (Fig. 2). The effects of ferric enterobactin on FepA in vivo were distinct from those previously observed in vitro, and the binding of the siderophore caused a slight further immobilization of MTSL at E280C (17). The mobilization of spins in vivo was so conspicuous that we initially suspected chemical release of MTSL from FepA by ferric enterobactin. Other experiments refuted this possibility though. Recooling of the sample to 4°C reversed the effects, regenerating the properties and proportion of strongly immobilized spins attached to E280C (Fig. 2). Secondly, MAL6, which forms a chemically stable thioether bond with Cys, produced the same results with FepAE280C in vivo (21). Double integration of the X-band spectra, furthermore, showed that spin labels were neither destroyed nor lost during the experiment (Fig. 2). The conversion of spins from strong to weak immobilization was comparable to changes in MTSL motion that accompany FepA denaturation (22), strongly suggesting that the siderophore stimulates relocation of PL5 away from globular protein structure into the aqueous milieu. These data show that two distinct conformations of PL5 exist in vivo: a form that binds the siderophore, in which spin-label motion is restricted, and a form that arises after siderophore binding, in which spin-label motion is comparatively unrestricted. The appearance of the mobilized form coincided with the initiation of siderophore transport (23), suggesting that ferric enterobactin triggered the conformational changes in FepA, either before or during its uptake.

Figure 2

Effect of ferric enterobactin on the mobility of MTSL attached to FepAE280C in live bacteria, observed in signal-averaged X-band spectra. KDF541/pFepAE280C was grown, spin-labeled, and prepared as described in Fig. 1. Signal-averaged X-band spectra (six field sweeps collected over an 8-min period) of spin-labeled bacteria were recorded at 4°C (A) and again after warming the flat cell to 37°C [B(31)]. A fresh aliquot of cells from the same labeling reaction was suspended in saturating ferric enterobactin (300 μM) at 4°C (C), the sample was warmed to 37°C, and spectra were recorded after 5 (D) and 15 (E) min (midpoints of the signal-average measurements). The bacteria were then cooled again to 4°C for 30 min, and spectra were recorded (F). Each of the spectra was double-integrated (G) from 3405 to 3555 G to determine the total number of MTSL spins (○) and from 3443 to 3477 G to determine the spins derived from the strongly immobilized (a) and weakly immobilized (b) peaks (•).

Another TonB-dependent ligand that penetrates via FepA, colicin B, also altered MTSL mobility during its transport, but the effects were opposite to those engendered by the siderophore. The addition of saturating colicin B at 37°C progressively eliminated weakly immobilized spins, enhanced the strongly immobilized peak, and increased the separation between low- and high-field extrema (Figs. 3 and 4). These changes did not result from colicin binding, because they did not take place during continual incubation of the cells at 4°C, even over the course of 1 hour. However, when cells with bound colicin were warmed to 37°C, MTSL at residue 280 showed a progressive immobilization that became more severe over 30 min, ultimately converting virtually all weakly immobilized spins to the strongly immobilized state (Figs. 3 and 4). As observed for the siderophore, these changes correlated with the transport phase of the colicin through the OM (24). Colicins presumably traverse the OM bilayer through porin channels (24), and the progressive immobilization of MTSL by colicin B may reflect steric restriction of spin-label motion as the colicin polypeptide passes through the FepA pore. The irreversibility of the immobilization during recooling supports the idea that the toxin polypeptide remains bound within FepA subsequent to OM transport. Both the kinetics and the nature of colicin-induced effects differed from the actions of ferric enterobactin, which probably reflects the different steric constraints placed on the spin-labeled site during uptake of the two structurally diverse ligands.

Figure 3

Effect of colicin B on the mobility of MTSL attached to FepAE280C in live bacteria, observed in signal-averaged X-band spectra. KDF541/pFepAE280C was grown, spin-labeled, and prepared as described in Fig. 1. Signal-averaged X-band spectra (six field sweeps collected over an 8-min period) of spin-labeled bacteria were recorded in the presence of saturating levels of colicin B (10 μM) at 4°C (A). The sample was warmed to 37°C and analyzed after 5 (B), 15 (C), and 75 (D) min (midpoints of signal average measurement) before recooling to 4°C (E). The same strain was grown in minimal media and starved for glucose for 2 hours to deplete its energy stores, spin-labeled, and exposed to colicin B (10 μM) at 37°C for 30 min (F). The spectra were double-integrated (G) as described in Fig.2.

Figure 4

Statistical analysis of ferric enterobactin– and colicin B–induced variations in FepAE280CMTSL motion. KDF541/pFepAE280C was grown, spin-labeled, and prepared as described in Fig. 1. Multiple signal-averaged spectra were collected as in Figs. 2 and 3 from several independent preparations of bacteria, and the mean and standard deviation of the ratio b/a was calculated at the indicated points after the addition of ferric enterobactin (A) (n = 4) and colicin B (C) (n = 6). The most sensitive measure of changes in spin-label mobility was the ratio of the amplitudes of b/a, shown and analyzed here. However, the effects of the siderophore and colicin were also analyzed by several other methods, including measurements of the absolute amplitudes, areas, separation of low- and high-field extrema (2T′) and half-width at half-height (32) for peaks a and b. All these methods showed the same tendency of the siderophore to mobilize spin labels and the colicin to immobilize them. (A) Bars 1 through 6 correspond to samples A through F, respectively, of Fig. 2. (C) Bars 1, 2, 3, 5, and 6 correspond to panels A through E, respectively, of Fig. 3; bar 4 shows the effects of colicin after 30 min at 37°C. Superimposed spectra from single experiments are also shown in the region of the strongly and weakly immobilized peaks. (B) Spectra were collected in the absence of ferric enterobactin at 37°C (purple), in the presence of ferric enterobactin at 4°C (blue), after increasing the temperature to 37°C for 5 (green) and 15 (red) min, and after recooling the sample to 4°C (black). (D) Spectra were collected in the presence of colicin B at 4°C (purple), after increasing the temperature to 37°C for 5 (blue), 15 (green), and 75 (red) min, and after recooling the sample to 4°C (black).

Continuous monitoring of MTSL mobility (at 3472 G) over time showed that ferric enterobactin stimulated a burst of motion that peaked and receded over a 5-min period (Fig. 5). This surge in motion accrued from the sum of individual, weakly immobilized MTSL spins in the bacterial sample. The changing mobility of the spin label in PL5 very likely reflects the opening and closing of the gated FepA channel during siderophore uptake. Kinetic analysis of the spin-label motion tracing, which showed two peaks of mobility approximately 10 and 20 min after the temperature shift to 37°C, revealed in both cases a sigmoidal rise to a maximum followed by an approximately symmetrical decay to a minimum. The integrated areas beneath the two curves were the same (25), suggesting that the changes in motion arose from an initial, relatively synchronous transport cycle and a second, less synchronous cycle in the same number of cells. The first surge in motion peaked in 140 s and decayed in 140 s; because ferric enterobactin was supplied in excess (26), the decrease in motion did not result from its depletion. Thus, the time-resolved fluctuations in PL5 were consistent with an active opening and closing of the channel during the ferric enterobactin transport reaction (27).

Figure 5

TonB and energy dependence of conformational changes in FepAE280C. (Top)fepA,tonB + bacteria (KDF541) expressing FepAE280C were spin-labeled with MTSL and analyzed in the presence of glucose. The intensity of the weakly immobilized peak (b in Fig. 1) was monitored versus time by setting the spectrometer center magnetic field at the peak maximum (3472.3 G) and the sweep width to zero. Bacteria were incubated at 4°C without added ligands (blue tracing) and warmed to 37°C after 5 min (marked by an arrow). The experiment was repeated in the presence of ferric enterobactin (300 μM; yellow) or colicin B (10 μM; red), as described in Figs. 2 and 3. (Center)fepA,tonB + bacteria (KDF541) expressing FepAE280C were analyzed, but were energy-depleted by glucose starvation described in Fig. 3 and subjected to the conditions described above. (Bottom) Analysis of fepA,tonB bacteria [KDF570 (9)] expressing FepAE280C, subjected to the conditions described above.

The relocation of spin labels attached to E280C, by both ligands, showed the hallmarks of energy- and TonB-dependence: Glucose deprivation, low temperatures, and use of a tonB strain all prevented the effects (Fig. 5). During ligand internalization, FepA PL5 changed to either mobilize (by decreasing their contact with other components of FepA structure) or immobilize [by increasing their contact with the FepA polypeptide or the protein (colicin B) that interacted with it] attached spin labels. Hence, FepA fluctuates between at least two conformations during ferric enterobactin transport, and colicin entry creates a different kind of conformational dynamic. The TonB- and energy-dependence of the observed structural changes confirms their physiological relevance (28), and the ligand-specific oscillation of probes attached to FepA PL5 suggests how siderophores enter the normally closed pores: TonB triggers energy-dependent structural changes in the ligand-binding site that displace PL5 and release the bound siderophore into the FepA channel. These results constitute physical evidence for the in vivo operation of a ligand-gated channel. They demonstrate what has been widely hypothesized, that gated porins open and close during solute transport. The perception of OM porins as rigid molecules containing passive channels, derived from crystallographic studies in vitro (4), contrasts with this characterization of FepA as a dynamic entity in vivo.

  • * To whom correspondence should be addressed. E-mail: peklebba{at}


View Abstract

Navigate This Article