(Received for publication, February 5, 1997, and in revised form, April 8, 1997)
From the Laboratoire des Biomembranes, URA CNRS 1116, Université Paris-Sud, Bât 430, F-91405 Orsay Cedex, France
The Escherichia coli outer membrane protein FhuA catalyzes the transport of ferrichrome and is the receptor of bacteriophage T5. Purified FhuA was reconstituted into liposomes. The size of the proteoliposomes and the distribution of the proteins in the vesicles were determined by freeze fracture electron microscopy. Unilamellar vesicles with a diameter larger than 200 nm were observed frequently. FhuA was symetrically oriented in the proteoliposomes. Reconstituted FhuA was functional as binding of phage T5 induced the release of phage DNA and its transfer inside the vesicles.
FhuA is a 78.9-kDa Escherichia coli outer membrane
protein that catalyzes the high affinity transport of the ferric
siderophores ferrichrome and albomycin across this membrane. It is also
the receptor for bacteriophage T1, T5, and 80 and for colicin M (for review, see Ref. 1). Molecular modeling suggests that the protein contains about 30 transmembrane
-strands connected by loop regions. One of these external loops, extending from residue 316 to 356, is
involved in ligand binding (2-5). FhuA was recently purified in
OG1 using a combination of anion exchange
chromatography, chromatofocusing, and size exclusion gel
chromatography. Sedimentation data combined with size exclusion
chromatography indicated that the protein is monomeric. It is
characterized by a high (51%)
-sheet content and the absence of
-helical structure. The purified protein solubilized in OG is
functional as demonstrated by its capacity to bind phage T5 and to
trigger the release of the DNA in the external medium (6). Binding of
phage T5 to purified FhuA incorporated into planar lipid bilayers also
converts the transporter into an ion channel. It was proposed that the
function of this channel would be to trigger the diffusion of
ferrichrome and of phage DNA (7). Phage T5 DNA is a double-stranded
molecule of 121 kilobase pairs, 30 µm in length. Despite many
attempts (8-10), the mechanism by which the DNA of this and other
phage is transferred through the bacterial envelope remains an open
question. Since it is difficult to obtain information at the molecular
level in vivo, we have tried to analyze at least part of
this process in vitro. Indeed, if as proposed above, the
FhuA channel allows the passage of the phage DNA, and if no other
bacterial components are required, we expect the reconstitution of FhuA
into liposomes to trigger transfer of the phage genome inside the
vesicle. Early experiments suggested that some DNA might be transferred
into liposomes bearing the partially purified phage T5 receptor (11).
However, these liposomes were small (<100 nm) and appeared collapsed
and multilayered, which raised the question of how the DNA could pass
through this multiple membrane barrier. Roessner et al. (12)
also claimed that phage
DNA would be injected into small
unilamellar vesicles (75-100 nm in diameter) containing the LamB
receptor.
We report the reconstitution of FhuA into large unilamellar vesicles. The morphology of the proteoliposomes was analyzed by FFEM. We show that FhuA is symmetrically oriented in the vesicles and functional as its interaction with phage T5 triggers the release of DNA and its transfer into the proteoliposomes.
Lipids were from Avanti or Sigma and of the highest purified grade. OG was from Bachem. SM2 Bio-Beads were from Bio-Rad.
Preparation of Phage T5 and Purification of FhuAT5stamN5 (T5) phage was produced on E. coli Fsu+ (13). Phage stocks were prepared and
purified as described in Ref. 7. They were resuspended in 10 mM Tris-HCl, pH 7.2, 150 mM NaCl. The final
titer was 1 × 1013/ml. E. coli AB2847
(pHK232), a strain overproducing FhuA, was used for purification. The
different steps of purification of FhuA were as described in Refs. 14
and 6. The purified protein was stored at
20 °C in 20 mM Hepes, pH 7.2, 150 mM NaCl containing 33 mM OG.
Fluorescence experiments were performed with an SLM 8000 spectrofluorometer in a 1 × 0.4-cm cuvette thermostatted at 37 °C. Excitation and emission wavelengths were set at 491 and 509 nm, respectively, and slits were 4 nm for excitation and emission. The final concentration of YO-PRO-1 (Molecular Probes) was either 2 or 4 µM, and the final volume was 1 ml.
Preparation of LiposomesLarge unilamellar vesicles were
prepared by reverse phase evaporation essentially as described (15)
using either a mixture of egg phosphatidylcholine and phosphatidic acid
(mol ratio 9:1) or L--phosphatidylcholine from soybean
(type IV-S). The lipid concentration was determined by adding a known
trace amount of [14C]phosphatidylcholine (4.1 × 108 Bq/mmol, specific activity 18.5 kBq/ml) to the initial
lipid solution. Liposomes were suspended in 20 mM Hepes, pH
7.2, 150 mM NaCl and sequentially extruded through 0.8-, 0.4-, and 0.2-µm polycarbonate membranes (Nucleopore). The final
phospholipid concentration was about 10 mg/ml (i.e. 12 mM). Liposomes were used within the 3 days following
preparation.
Proteoliposomes were prepared from the lipid vesicles (16, 17) as follows. Liposomes (330 µl, 5 mM) were incubated with OG (either 22 or 45 mM) for 30 min. Then FhuA (35 µg) was added and the detergent concentration adjusted to either 22 or 45 mM. For some of the fluorescent assays, YO-PRO-1 (4 µM final concentration) was also added. The mixture was incubated further for 1 h under gentle stirring. The detergent was then adsorbed onto SM2 Bio-Beads (18) and at a concentration of 80 mg of wet Bio-Beads/ml. The suspension was shaken gently. 3 h later a second portion of the same amount of Bio-Beads was added and the suspension shaken overnight at 4 °C. No significant binding of YO-PRO-1 to the Bio-Beads was observed. Proteoliposomes were removed gently after decanting the Bio-Beads and used within the same day. Solubilization and reconstitution experiments were performed at room temperature, except where otherwise indicated.
Freeze Fracture Preparations for Electron MicroscopyProteoliposomes were centrifuged at 100,000 × g in a TL100 ultracentrifuge (Beckman) for 30 min at
4 °C. The pellet was cryoprotected with glycerol (30% v/v). A small
drop of the sample was placed between a thin copper holder and a thin
copper plate and quenched in liquid propane, as described (19). The
frozen sample was fractured at 125 °C in vacuo of about
1.33 × 10
5 pascals by removing the upper plate with
a liquid nitrogen-cooled knife in a Balzers 301 freeze-etching unit.
The fractured sample was replicated with a 1-1.5-nm deposit of
platinum-carbon, cleaned in chromic acid, washed with distilled water,
and observed with a Philips 301 electron microscope.
98% of the liposomes prepared by reverse phase
evaporation and observed in FFEM were unilamellar. Their diameter
varied between 100 and 200 nm (data not shown). FhuA was reconstituted
into the liposomes following the procedures described in Ref. 20 and using OG since the protein was purified and functional in this detergent (6). The lipid:protein mol ratio (3800) was chosen so that
individual proteins (particles) could easily be observed by FFEM. It
should be kept in mind that because of steric hindrance no more than 20 phages are likely to bind on a proteoliposome having a diameter of 200 nm. To find the optimal conditions for reconstitution in terms of
orientation and distribution of the proteins in the liposomes, two
types of assays were performed. In the first the purified protein
solubilized in OG was added to a suspension containing preformed
liposomes saturated with detergent coexisting with mixed
lipid/detergent micelles. This stage (called stage II) is generally
obtained for an OG concentration of 22 mM (i.e.
just below the critical micellar concentration). In the second assay,
the protein was added to a suspension of mixed detergent/lipid
micelles. This stage (stage III) is observed when detergent is added to
the liposomes at a concentration well above the critical micellar
concentration (i.e. 45 mM). Upon lowering the
detergent concentration by adsorption onto Bio-Beads the protein either
reincorporates directly into the liposomes saturated in detergent
(stage II) or associates with the lipids upon removal of the detergent
to form proteoliposomes (stage III). These two procedures generally
lead to different orientations of the protein (20). Proteoliposomes
were then observed by FFEM. Figs. 1 and 2
represent, respectively, the FFEM and the size and particle distribution histograms of the different preparations. The histograms were obtained over a population of 100-200 vesicles. Calculations were
done according to Heegaard et al. (21). Fig. 1A
shows the result of a reconstitution performed at stage II. The bulk of the population comprised vesicles with smooth convex and concave fracture surfaces and corresponding to liposomes. Some multilamellar vesicles were also observed. Only a few liposomes contained proteins, which were vizualized as globular intramembranous particles. The particles were distributed equally between the concave and convex fracture surfaces. Large and numerous aggregates of particles were
observed; these probably correspond to protein aggregates. The
histograms Fig. 2, A and B, illustrate these
observations: 70% of the vesicles had a mean diameter of 50-100 nm
and contained no more than 10-20 particles/proteoliposome. Stage III
reconstitution led to two distinct populations of vesicles (Fig.
1B): small vesicles without particles and larger ones
containing particles that were distributed equally between the convex
and concave fracture surface. Neither protein aggregates nor
multilamellar vesicles were observed. 50% of the proteoliposomes had
an average diameter of 100-200 nm and contained 40 particles. The
diameter of 40% of the proteoliposomes was larger than 200 nm, and
these vesicles contained approximately 150 particles (Fig. 2,
A and B). Some of the vesicles had even a
diameter varying between 500 and 800 nm (not shown on Fig. 1).
An aliquot of stage III preparation was treated with phage T5 so that the final phage:protein ratio was 1:100 (about one phage/proteoliposome). The suspension was incubated for 30 min at 30 °C and then treated with DNase (20 µg/ml) and MgSO4 (2.5 mM) to degrade ejected DNA. Fig. 1C shows that the vesicle morphology and size were not modified by the presence of phage. Phage structures were occasionally observed: they appeared as clusters of particles, the morphology of which was different from that of the protein aggregates shown in Fig. 1A. Their size corresponds to that expected for the phage capsid (90 nm in diameter).
Proteoliposomes reconstituted at stage III were subjected to centrifugation on a flotation sucrose gradient essentially as described in Ref. 17. Fractions were collected, and the proteins and lipids were estimated from the trytophan fluorescence (FhuA contains 9 Trp) (4) and radioactivity, respectively. No fluorescence was recovered at the bottom of the gradient, indicating that all of the protein had been incorporated into the liposomes. Protein and lipids were found in a unique peak centered at 18% sucrose (w/v). Radioactivity was also found, albeit at a low level, throughout the gradient (data not shown). These results are in line with the ultrastructural data. Further functional analysis was limited to the proteoliposomes prepared at stage III.
Functionality and Orientation of FhuA in the ProteoliposomesPrevious experiments have shown that the addition of purified FhuA solubilized in OG to phage T5 resulted in the release of DNA from the phage capsid. Release of the DNA was measured from the increase of the fluorescence of YO-PRO-1, a dye that intercalates between the free DNA base pairs. The intensity of fluorescence of the dye is directly proportional to the number of DNA base pairs, and, provided that the fluorescence had been calibrated with DNA solutions of known number of base pairs, it is possible to design a quantitative assay of protein functionality (6).
To improve the effect of successive treatment of FhuA with detergent
and Bio-Beads, the fluorescence assay was applied first to
proteoliposomes solubilized by the addition of 25 mM OG
(Fig. 3). The successive addition of proteoliposomes,
phage T5, and OG to the cuvette containing YO-PRO-1 resulted in an
increase of the fluorescence which started after a lag of about 30 s corresponding to the solubilization of the proteoliposomes and
reached a plateau 5 min later. Quantification of the DNA ejected
indicated that 90 ± 8% of the DNA was released at the plateau.
Addition of DNase in the presence of Mg2+ to hydrolyze the
released DNA caused an extinction of the fluorescence signal which
returned to its initial level in the absence of DNA. The same
fluorescence changes were observed with the FhuA preparation not
submitted to the reconstitution procedure. This indicates that the
reconstitution protocol has had no deleterious effects on the
protein.
We showed previously that the rate of fluorescence increase does not
correspond to the rate of DNA ejection, which occurs in few seconds,
but to the rate of binding of the phage to FhuA: the higher the FhuA to
phage ratio was, the shorter the time it took for the phage to bind.
Typically, increasing the concentration of FhuA from 0.015 to 100 nM reduced the half-time of fluorescence increase from 30 min to 30 s (6). Therefore, measuring the initial rate of
fluorescence increase should allow the number of active proteins to be
determined. We took advantage of this observation to determine the
orientation of FhuA in the proteoliposomes. The initial rate of
fluorescence increase was first measured for variable concentrations of
FhuA solubilized in OG ranging from 0.7 to 2.6 µg/ml. As expected,
this rate increased linearly with protein concentration (Fig.
4). This rate was then determined for the same
concentration of protein but reconstituted in the liposomes. The linear
dependence was still observed, but the slope was half that measured for
the solubilized protein. Solubilization of the proteoliposomes by the
addition of OG before adding the phage resulted in the same linear
dependence of the rate as that determined for the solubilized protein.
These results confirm that all of the protein molecules have been
inserted in the vesicles but that only about half of them are
accessible to the phage. The protein is therefore symmetrically
oriented in the vesicles.
Binding of Phage T5 to Reconstituted FhuA Triggers the Transfer of DNA into Proteoliposomes
A prerequisite to the determination of
entrapped DNA with the fluorescent probe YO-PRO-1 is that the dye
should be present both inside and outside the vesicles. Indeed, the
channel that is unmasked in FhuA upon binding of T5 allows the
diffusion of ferrichrome (7, 22). Since the molecular mass of YO-PRO-1 (629 Da) is close to that of ferrichrome (800 Da), it is likely that
the dye diffuses through FhuA upon binding of the phage. Therefore
proteoliposomes were prepared in the presence of 4 µM YO-PRO-1. Phage T5 was first added to a cuvette also containing 4 µM YO-PRO-1 (Fig. 5). The fluorescence
increased slowly and reached a plateau of 450 a.u. in about 15 min. This increase was DNase-insensitive and corresponded to diffusion
of some dye in the phage capsid. Proteoliposomes were then added. The
fluorescence increased after a lag of 2-3 min and reached a new steady
state corresponding to 800 a.u. 15 min later. Proteoliposomes were
then solubilized by OG. Ferrichrome, which competes with phage for binding to solubilized FhuA (6), was first added to prevent phage
attachment to any new receptor and especially to those whose binding
site was oriented inside the vesicles and which had become accessible
upon solubilization. Addition of DNase and Mg2+ decreased
the fluorescence to 150 a.u. immediately (data not shown). The
difference between the plateau value (800 a.u.) and the final value
(150 a.u.) therefore represents the total DNA whether released inside
or outside the proteoliposomes. When DNase and Mg2+ were
added to the proteoliposomes, but in the absence of OG, the
fluorescence was only partially decreased (450 a.u. at steady state).
Since DNase cannot diffuse through the liposomes, this indicated that
only part (54%) of the DNA was ejected in the external medium. DNase
was then inhibited by the addition of EDTA. Ferrichrome and OG were
successively added followed by an excess of Mg2+ to
reactivate DNase. This resulted in a new fluorescence decrease to a
plateau value of 180 a.u., indicating that some new DNA had become
accessible to DNase. It is likely that this DNA was entrapped in the
liposomes prior to solubilization.
The data presented in this paper indicate that FhuA can be functionally reconstituted into liposomes. Two reconstitution protocols were chosen because they were likely to result in different orientations of the protein in the liposomes. Reconstitution of FhuA was less efficient when the protein was added to liposomes saturated with detergent coexisting with mixed lipid/detergent micelles (stage II); few proteins were inserted into the liposomes, and large protein aggregates that were not incorporated into the liposomes were found. It is likely that upon removal of the detergent, aggregation of the protein has preceded insertion in the liposomes and that the only proteoliposomes formed originate from mixed detergent/lipid micelles. In contrast, when reconstitution was performed at stage III, i.e. from totally solubilized liposomes, neither protein aggregates nor multilamellar vesicles were found, and large proteoliposomes (mean diameter ranging from 100 to 200 nm, some of them having a diameter of 500-800 nm) were observed which contained fairly large amounts of proteins. The fact that the proteoliposomes were larger than the initial liposome preparation suggests that the protein incorporates into the liposomes during their formation and that the open vesicles tend to fuse and form large vesicles. The difference in size of the vesicles formed upon reconstitution at stage II or III is probably due to the rate of removal of detergent, which is in turn dependent on the initial detergent/lipid ratio and on detergent concentration (23).
Functionality of reconstituted FhuA was assessed from a fluorescence
assay of the DNA released upon binding of phage T5. Interaction of T5
with proteoliposomes prepared at stage II resulted in some release of
DNA (data not shown). However, the preparation also contained protein
aggregates, making it difficult to discriminate between phage bound to
the aggregates or to proteoliposomes. For the different reasons given
above proteoliposomes prepared at stage III were more suitable for
functional assays. The protein was shown to be symmetrically oriented
in these liposomes. DNA released from the phage was found both outside
and inside the proteoliposomes. The DNA present outside is likely to
originate from aberrant phage. These phages eject their DNA outside the bacterial envelope and represent 10-30% of the T5 stocks (24). DNA
might also be mistargeted as a consequence of inappropriate interactions between the phage and the receptor. This could occur in
vesicles having a small curvature radius. We are rather confident that
our measurements have allowed us to determine entrapped DNA and not
that released upon solubilization of the liposomes since binding to any
newly accessible receptor was prevented by the addition of ferrichrome.
Such a precaution was not taken by Tosi et al. (11),
therefore rendering their data on T5 DNA transfer in liposomes
difficult to interpret. Similarly, Roessner et al. (12)
measured injection of phage DNA into small unilamellar vesicles
containing the LamB receptor after disruption of the liposomes with
chloroform or phospholipase followed by DNase treatment. We cannot
exclude that this treatment has unmasked inside-oriented receptors to
which phage
would have bound. Our experiments strongly suggest that
no bacterial component other than FhuA is required for DNA transport.
They however indicate that only 10-20% of the total phage DNA is
transferred into the liposomes. The simplest explanation is that
transfer takes place only in the largest (500-800 nm in diameter)
liposomes, which represent only a small part of the population. The
state of compaction of T5 DNA in the liposomes is not known. We cannot
exclude that polyamines present in the phage capsid would be
transferred along with the DNA into the liposomes. Phage
DNA (49 kilobase pairs) compacted with polyamines appears as a sphere of less
than 100 nm in diameter in electron microscopy (25). If T5 DNA is also
compacted by polyamines then it could fit into the proteoliposomes. The
route used by DNA to cross the liposomes also remains unknown. DNA
transfer is accompanied neither by significant perturbations of the
vesicles morphology nor by changes in permeability to solutes of high
molecular mass (i.e. cytochrome c) (26). The
simplest model therefore still supposes that the DNA diffuses through
the channel opened by T5 in FhuA.
We thank P. Boulanger and T. Pugsley for constant interest in this work. We are indebted to T. Gulik Krzywicki (Center de Génétique Moléculaire, Gif sur Yvette) in which Laboratory the FFEM experiments were done. We thank V. Braun for providing the E. coli strain overproducing FhuA.