Modeling ligand-gated receptor activity. FhuA-mediated ferrichrome efflux from lipid vesicles triggered by phage T5.

An in vitro assay of iron-ferrichrome translocation across the FhuA protein of outer membranes from Escherichia coli has been devised. Upon reconstitution into large lipid vesicles, bacteriophage T5 binds to this polyvalent receptor, triggering a conformational change that resulted in channel opening. This facilitates the translocation of an iron(III)-siderophore, without the complexities involved in the in vivo process. Efflux of 55Fe(III)-ferrichrome across FhuA channels was determined quantitatively by monitoring the release of trapped radioactivity. The assay is rapid, reliable, and specific, because other bacteriophages, such as Φ80, fail to trigger channel opening of the FhuA receptor.

Understanding the function of ligand-gated channel proteins on the basis of their conformational changes has attracted much interest in recent years (1). Although the number of sequences of proteins belonging to this family has increased rapidly (2), structures are available for a mere few of them and at limited resolution (3). This is caused on the one hand by the difficulties in overexpressing these proteins, particularly those from eucaryotic sources (4), and on the other by the problems besetting the preparation of three-dimensional crystals of membrane proteins, which of course are prerequisite for the determination of high resolution structures by x-ray analysis. We have therefore chosen the iron(III)-ferrichrome translocating FhuA protein from Escherichia coli outer membranes (5)(6)(7)(8)(9) as a model of ligand-gated proteins. Proteins from this source have proven very useful in the determination of structures at atomic resolution (10 -13). The molecular genetics of E. coli, moreover, are well suited for the investigation of the structural basis of their functional properties (14 -16).
At limited availability of iron in the medium, its transport into the cytoplasm requires a complex cascade of events (17). First, ferric ions are chelated by one of several siderophores (18). Second, the iron-siderophore complex binds to its specific surface-exposed receptor. Binding is a necessary but not sufficient condition to trigger channel opening in a third step. This process requires energy, which is transduced from the plasma membrane to the channel protein by a protein complex consisting of at least three components, the products of the genes tonB, exbB, and exbD (19,20). The TonB protein appears to play a key role in this energy transduction by an as yet poorly understood allosteric transition, which results in the interaction of the TonB protein with a specific motif, the so-called TonB box, which is found in several channel proteins (21)(22)(23). Fourth, the substrate is scavenged by a ligand-specific binding protein in the periplasmic space (24). In a fifth step, the complex is delivered to a specific, active transport system in the plasma membrane. In the final, sixth step, the oxidized form of iron is reduced in the cytoplasm to the ferrous state (25,26). The rate-limiting step of this cascade appears to be the energy requirement of the translocation process across the outer membrane. The function of the TonB complex is needed also for infection of the cell with the several viruses and bacterial toxins that use FhuA as a receptor (phages T1, ⌽80, and colicin M), with the only exception being bacteriophage T5. This virus has been shown to trigger channel opening and ion flux through the FhuA protein in the absence of energy transduction (27). It thus circumvents this requirement, whereas the other ligands may serve as controls.
We are presently crystallizing the FhuA protein. This requires not only that the protein be available in large quantities but also that its native state can be assayed by a routine procedure throughout the lengthy process of crystallization. We here report an assay in which the FhuA protein and 55 Fe(III)ferrichrome were co-incorporated into large lipid vesicles: the former into the membrane and the latter trapped in the internal compartment. Interaction of T5 phage with the receptor protein, monitored by a fluorescent dye that interacts with the ejected phage DNA, caused the release of radiolabeled Fe(III)ferrichrome through the FhuA channel. Using radioactive iron, efflux from vesicles could readily be quantitated.

EXPERIMENTAL PROCEDURES
Protein Purification-FhuA protein was overexpressed in an E. coli BL21(DE3) strain not expressing OmpF, OmpC, PhoE, LamB, and OmpA proteins (gift of Dr. A. Prilipov), using the plasmid pHK763 carrying the fhuA gene (kindly provided by Dr. H. Killmann). The protein was solubilized from outer membranes using octyl-POE 1 (Alexis, Lä ufelfingen, Switzerland) as the detergent and purified in a procedure analogous to that used for porins from E. coli (28,29). After a final size exclusion chromatography step, SDS-polyacrylamide gel electrophoresis of the heat-treated protein revealed a single band corresponding to a polypeptide with a mass of 80 kDa.
Ferrichrome Labeling-Ferrichrome, kindly provided by Dr. J. B. Neilands, was labeled with 55 Fe(III) according to Weaver and Konisky (30). The final solution was 2.03⅐10 Ϫ3 M, as determined by its absorbance at 426 nm. The specific activity was 55 Ci⅐mol Ϫ1 .
Bacteriophages-Phage T5 (the gift of Dr. V. Braun) and ⌽80 (Biozentrum) were isolated and stored as described (27), using E. coli B E and DH5␣ (Biozentrum) for production of bacteriophages T5 and ⌽80, respectively. Using the same strains, phage titers of T5 and ⌽80 were determined to be 5⅐10 12 pfu/ml and 7⅐10 12 pfu/ml, respectively. protein and 55 Fe-ferrichrome were incorporated by resolubilization of the liposomes, addition of the desired constituents, and subsequent detergent removal (32,33). 14 C-Labeled vesicles were mixed with octylglucoside (0.6 M), 7 Ci of 55 Fe-ferrichrome, 20 g of purified FhuA protein, and HEPES buffer to yield final volumes of 330 l, containing 5 mM, and 40 mM of lipid and OG, respectively. The mixture was gently agitated for 30 min at room temperature. 27 mg of BioBeads SM2 (Bio-Rad) were enclosed in a small dialysis bag and added to the solution. The reaction vessel was agitated for 2 h at room temperature. A second batch of BioBeads SM2 (27 mg) was added, and the incubation was continued overnight at 4°C. Residual ferrichrome was removed from the vesicles by two sequential gel filtration chromatography steps on a 1 ϫ 9 cm Sephacryl S-200 HR (Pharmacia Biotech Inc.) column using the above mentioned HEPES buffer. Fractions were collected and aliquots counted for 14 C and 55 Fe during 20 min in a Packard 2200CA counter. Proteoliposomes containing 55 Fe-ferrichrome were stored at 4°C and used for assays within 24 h.
Fluorescence Studies-Ejection of phage DNA was measured at 37°C in a SLM 8000C fluorimeter (SLM-Aminco, Urbana, IL), using a fluorescent quaternary ammonium dye derivative (YO-PRO 1, Molecular Probes Inc.), which intercalated into double-stranded DNA (34). The excitation and emission wavelengths used were 490 and 509 nm, respectively. Calibration of the fluorescent signal caused by the release of phage DNA was performed by incubating an aliquot of phage T5 stock solution with an equal volume of 9 M LiCl at 50°C for 10 min and subsequent dilution of known amounts of phages in HEPES buffer (35). Reconstituted FhuA protein (0.6 g) was mixed with 5⅐10 9 phage T5 pfu in 1.5 ml of HEPES buffer containing 1 mM of CaCl 2 , 1 mM MgSO 4 , and 2 M YO-PRO 1. The increase of the fluorescence signal was recorded as a function of time. After reaching a plateau, 20 units of DNase I (Boehringer Mannheim) was added. To solubilize the vesicles, octyl-POE was added to a final concentration of 1%. In one control experiment, liposomes devoid of FhuA protein were used. In another control, 0.6 g of detergent-solubilized FhuA protein was used in 1% octyl-POE.
Translocation Assays-Ferrichrome efflux was monitored by separating free from entrapped 55 Fe-ferrichrome on a Sephacryl S-200 column following incubation of the vesicles with different ligands. A ratio of one pfu/liposome was used for incubations with phages. DNase I treatment was used to reduce the viscosity of the sample, because ejected phage T5 DNA interfered with chromatography. Translocation assays were performed by adding the following components to a total volume of 250 l: (a) 60 l of proteoliposomes from the peak fraction of the excluded volume (see Fig. 1A

RESULTS
Proteoliposomes reconstituted with FhuA protein were freed from extraneous 55 Fe-ferrichrome by gel filtration chromatography in two steps, of which the second reveals a negligible peak of free ferrichrome (Fig. 1A). When stored at 4°C, the vesicles remained stable over a period of 24 h. Upon the addition of phage T5 (1 per 30 proteoliposomes), ejection of DNA was revealed by the increase of the fluorescence intensity of the dye YO-PRO 1(see Fig. 2), which intercalated into doublestranded DNA (36). An instantaneous rise, seen in all curves, could be attributed to the interaction of the dye with free DNA originating from the small fraction of lysed T5 phages present in the added phage stock solution. The subsequent gradual rise (Fig. 2, curve A) was attributed to two components. Phage DNA ejected in the medium interacts with the fluorescent dye; the contribution of phage DNA injected into the vesicle interior is more difficult to quantitate. When the curve reached a plateau, all T5 phages had bound to a receptor protein molecule, as  supported by the observation that the addition of proteoliposomes after 1400 s had no further effect on the fluorescence signal (Fig. 2, arrow 1). Treatment of the solution with DNase I (Fig. 2, arrow 2) caused a rapid decrease of the fluorescence intensity of curve A by about 30%. This showed that the fraction of phage DNA existing in solution was immediately accessible to enzymatic hydrolysis. The liposome-entrapped DNA could also be degraded but only upon solubilization of the vesicles with a detergent (Fig. 2, arrow 3). This result represented strong evidence that phage DNA did enter the liposomes. Solubilization of the vesicles performed before addition of DNase I (Fig. 2, dotted line) caused a rapid increase in fluorescence intensity. The value reached corresponded to the level seen if detergent-solublized FhuA protein was present without liposomes (control experiment; Fig. 2, curve C). This result appeared to account for the quenching of the DNA-dye complex in the liposome interior and its release upon the addition of a detergent. The residual fluorescence, seen in curves A and C, was due to oligonucleotides resulting from DNase I activity (37).
Translocation of Fe(III)-ferrichrome was shown in Fig. 1  (B-D). Phage T5 triggered channel opening of the FhuA protein reconstituted in lipid vesicles, facilitating the efflux of approximately 85% of the labeled ferrichrome (Fig. 1B). The control experiments showed that with phage ⌽80, less than 15% of the 55 Fe label was released, an amount also observed if no phage was added under otherwise identical conditions. The addition of octylglucoside, causing the solubilization of the proteoliposomes, released Ͼ95% of the Fe(III)-ferrichrome into the medium (Fig. 1D). The remaining 55 Fe label was eluted in the excluded volume adsorbed to lipid/protein/detergent aggregates. DISCUSSION We have designed an in vitro assay system that allows the quantitative determination of translocation of an iron(III)-siderophore complex through FhuA channels. The energy requirement of the in vivo translocation process has been circumvented by exploiting the properties of phage T5, which upon binding to the membrane-incorporated FhuA protein, causes DNA ejection. This process is specific, because it was not observed with other ligands such as phage ⌽80. A significant difference in the time courses of T5 phage DNA ejection with FhuA protein, either reconstituted into liposomes or in its detergent-solubilized form, was observed, for which three explanations may apply: (a) The effective concentration of binding sites exposed to the solution is half that of the total protein if the inserted membrane protein is oriented randomly. (b) Accessibility of surface-exposed binding sites may be considerably reduced, due to a taut conformation of the protein in the lipid bilayer (38). (c) When phage T5 DNA is injected into vesicles, the fluorescence probe YO-PRO 1 may give rise to a reduced signal. This may be due to limited accessibility of dye to the vesicles, quenching of the signal, or both. The observation that the fluorescence intensity rose significantly upon solubilization of the vesicles by detergent revealed that phage DNA was indeed trapped in the liposomes. The results from these experiments provide strong evidence that the FhuA protein was reconstituted in the vesicles in an active form and that it allowed DNA to cross the lipid bilayer. This should now give the opportunity to address the question of the specificity of transported solutes and of the mechanism by which macromolecules cross biological membranes.
Our choice of gel filtration chromatography to quantitate ferrichrome efflux is based on the accuracy and reproducibility with which potential adsorption of ferrichrome to vesicles and leakage can be distinguished from actual efflux. Thus, we could determine that ferrichrome efflux following incubation with phage ⌽80 amounts to about 15% of the total release of 55 Fe label from the vesicles. This tallied with the value of another control experiment in which no phage was added at all and thus represented the background leakage from liposomes under the reaction conditions used. The value of 85% of ferrichrome released during incubation with phage T5 can be assigned to leakage on the one hand and to a fraction of about 80% of liposomes that encountered a phage particle on the other. Keeping the ratio of phage-to-liposome constant, efflux could be maximized by adjusting the ratio of incorporated FhuA protein per vesicle to a value between 5 and 10. Now that these values have been established, it will be possible to use filter assays for the routine determination of efflux with large sets of samples.
In conclusion, iron(III)-ferrichrome transport across the outer membrane of E. coli, the physiologically relevant in vivo function of the FhuA protein, could be assessed in an in vitro system using bacteriophage T5 as the triggering ligand. This stratagem avoids the problems involved in the energy transduction from the plasma membrane to the channel protein. The assay allows the specificity of the gating ligand (phage T5 versus ⌽80) to be assessed and should now provide a tool to determine the vectoriality and specificity of the transported solutes, as well as the question of DNA transfer. As to our structural studies of the receptor protein (29), its functionality can now be monitored at any stage during purification and crystallization. In addition, it opens the possibility for kinetic studies of the transport rate through the FhuA protein and sets the stage for reconstitution of the entire translocation cascade.