Porins OmpC and PhoE of Escherichia coli as Specific Cell-surface Targets of Human Lactoferrin

The binding of lactoferrin, an iron-binding glycoprotein found in secretions and leukocytes, to the outer membrane of Gram-negative bacteria is a prerequisite to exert its bactericidal activity. It was proposed that porins, in addition to lipopolysaccharides, are responsible for this binding. We studied the interactions of human lactoferrin with the three major porins ofEscherichia coli OmpC, OmpF, and PhoE. Binding experiments were performed on both purified porins and porin-deficient E. coli K12 isogenic mutants. We determined that lactoferrin binds to the purified native OmpC or PhoE trimer with molar ratios of 1.9 ± 0.4 and 1.8 ± 0.3 and Kd values of 39 ± 18 and 103 ± 15 nm, respectively, but not to OmpF. Furthermore, preferential binding of lactoferrin was observed on strains that express either OmpC or PhoE. It was also demonstrated that residues 1–5, 28–34, and 39–42 of lactoferrin interact with porins. Based on sequence comparisons, the involvement of lactoferrin amino acid residues and porin loops in the interactions is discussed. The relationships between binding and antibacterial activity of the protein were studied using E. coli mutants and planar lipid bilayers. Electrophysiological studies revealed that lactoferrin can act as a blocking agent for OmpC but not for PhoE or OmpF. However, a total inhibition of the growth was only observed for the PhoE-expressing strain (minimal inhibitory concentration of lactoferrin was 2.4 mg/ml). These data support the proposal that the antibacterial activity of lactoferrin may depend, at least in part, on its ability to bind to porins, thus modifying the stability and/or the permeability of the bacterial outer membrane.

Lactoferrin (Lf) 1 is an 80-kDa iron-binding glycoprotein found in various biological secretions, mainly in milk (1), and in polymorphonuclear leukocytes (2). The biological roles of Lf include antibacterial activities through mechanisms not yet clearly elucidated. The ability of Lf to tightly chelate two ferric ions allows the protein to limit the iron availability to bacteria and ultimately causes bacteriostasis (3,4). Moreover, Lf and an Lf-derived peptide called lactoferricin (Lfcin) (5) (residues 1-47 and 18 -41 for human (hLf) and bovine (bLf) Lfs, respectively) were shown to bind to Gram-negative bacteria, including Escherichia coli and to release lipopolysaccharides (LPS) from the outer membrane (6 -8). Stable complexes are formed between free LPS and Lf (9 -12), but the mechanism of membrane destabilization is not clearly elucidated. During the last few years, Naidu and co-workers (13)(14)(15) reported potential interactions of either hLf or bLf with porins, the major pore-forming proteins of the outer membrane of various Gram-negative bacteria. In particular, the binding of Lf to porins OmpF and OmpC in E. coli was reported (13). However, the biochemical evidence of interactions between Lf and porins was mainly based upon Western blot analyses using SDS-extracted and heat-denatured porin monomers. Because porins are potential receptors for bacteriophages and colicins (16,17), it is likely that Lf may use porins as anchoring sites at the surface of the outer membrane of bacteria. One can also hypothesize that the binding of Lf to porins either facilitates the destabilization of the bacterial outer membrane or limits the permeability of the membrane.
In the present paper, we have investigated whether hLf specifically binds to porins and how it does. The binding characteristics of hLf to porins OmpC, OmpF, and PhoE in E. coli were studied. Purified native porins and porin-deficient isogenic mutants of E. coli K12 were used. The porin-binding site of hLf was located using various hLf-derived proteins. Finally, a possible relationship between hLf binding to porins and the antibacterial activity of the protein was investigated with E. coli porin mutants and through electrophysiological studies in a planar lipid bilayer system.
Purification of Porins-Native OmpC, OmpF, and PhoE porin trimers were extracted and purified to homogeneity from EC1234, EC1233, and EC1230 E. coli K12 strains, respectively, as described previously * This work was supported in part by the Ministère de l'Enseignement et de la Recherche Scientifique (ACC SV5, Interface Chimie-Biologie) and the Centre National de la Recherche Scientifique. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 Spik et al. (21). The iron-saturated and apo forms of hLf were prepared as described previously (22,23). Mild tryptic digestion of hLf gave the N-terminally deleted proteins hLf Ϫ3N (residues 4 -692) and hLf Ϫ4N (residues 5-692) (24), the 30-kDa N-terminal (residues 4 -283) and 50-kDa C-terminal (residues 284 -692) fragments, as well as the 18-kDa N-terminal domain 2 (residues 91-255) of hLf (25). Human serotransferrin (hTf) and bLf were purchased from Sigma and Biopole (Brussels, Belgium), respectively. All chemicals used were of the highest analytical grade.
Expression and Purification of Recombinant hLfs-Nonmodified recombinant hLf (rhLf), rhLf Ϫ5N , a mutated rhLf, in which sequence 1 GRRRR 5 was deleted, and rhLf-EGS, a rhLf whose sequence 28 RKVRGPP 34 was replaced by EGS (the 365-367 C-terminal counterpart of sequence 28 -34) were produced in a baculovirus expression system as previously reported (12,24,26). The rhLf-SAST corresponds to rhLf in which 39 KRDS 42 was substituted by residues SAST, the 372-375 C-terminal counterpart of sequence 39 -42. For this purpose, a mutagenic oligonucleotide 5Ј-GTCAGCTGCATATCAGCATCAACCCC-CATCCAGTG-3Ј was synthesized by Eurogentec (Seraing, Belgium). The template for the mutagenesis was the phage M13-mp11, containing a 346-base pair EcoRI-AccI fragment of the coding sequence cloned into a pBluescript SK plasmid (27). The mutation was confirmed by DNA sequence analysis, and the mutated EcoRI-AccI was ligated back into pBluescript SK with the complementary part of the full-length cDNA of hLf as formerly described (27). Finally, the mutated cDNA was subcloned into pVL1392 (PharMingen), and the rhLf-SAST was produced in the baculovirus expression system and purified on a SP-Sepharose fast flow column, as described previously (26). The purity of the mutant protein was verified by 7.5% SDS-polyacrylamide gel electrophoresis followed by Coomassie Blue staining. The N-terminal amino acid sequence analysis was performed with the Edman degradation procedure, using an Applied BioSystem 477 protein sequencer.
Radiolabeling of hLf-Human Lf was labeled either with 125 I or with 59 Fe. Radioiodination of hLf (100 -200 g) was carried out with 0.3 mCi of 125 I in the presence of two Iodo-Beads (Pierce), according to the manufacturer's instructions. After 10 min of incubation on ice, free iodine was removed by gel filtration on a Sephadex G-25 PD-10 column (Amersham Pharmacia Biotech). Specific radioactivity was typically between 700,000 and 1,000,000 cpm/g of protein. For 59 Fe labeling, iron was first removed from hLf as described previously (22). Human Lf was then incubated 30 min at room temperature in a 0.25 mM sodium nitrilotriacetate, 0.1 M Tris/HCl, 0.1 M sodium bicarbonate, pH 8.2, solution containing an appropriate amount of 59 Fe (33 g/ml of carrierfree 59 FeCl 3 in 0.1 M of HCl, 0.1 mCi/ml, Amersham Pharmacia Biotech). Finally, free iron was eliminated by passing the solution through a Dowex 1 ϫ 8 column (Bio-Rad) and desalting on a PD-10 column (Amersham Pharmacia Biotech). A typical specific radioactivity of about 5,000 cpm/g of 59 Fe-hLf was obtained. To avoid radiation damage, the radiolabeled proteins were used immediately for the binding experiments.
Binding of hLf to Purified Porins-Purified porins OmpC, OmpF, and PhoE were diluted in a 75 mM Tris/HCl, pH 6.8, buffer at a concentration of 46 nM. Denaturation of porins was achieved by boiling aliquots of the porin solutions at 95°C for 10 min in the presence of 1% SDS (w/v). Fifty l of the porin solutions (0.9 pmol of porin trimer) were then slot blotted to a 0.45-m nitrocellulose membrane (Schleicher & Schuell) using a Bio-Dot SF Microfiltration Apparatus (Bio-Rad). After drying, excess sites on the membrane were blocked with a phosphatebuffered saline/2% Tween 20 (w/v) solution for 45 min, and the membrane was washed with phosphate-buffered saline/0.05% Tween 20 (w/v) (washing solution). It was then incubated for 1 h with washing solution containing 125 nM unlabeled or radiolabeled hLf. For the optimal binding pH experiments, incubations were performed using 125 nM unlabeled hLf in the washing solution at pH values ranging from 5.0 to 8.0 with 0.5 pH unit increments. For Scatchard analysis (28), nitrocellulose strips were incubated with various hLf concentrations ranging from 0 to 250 nM in phosphate-buffered saline, pH 7.3. When experiments were performed with radiolabeled proteins, membranes were washed three times with the washing solution and dried, and the slots were cut for further counting on a Compugamma LKB-Wallac ␥-radiation counter. When immunochemical staining was required, washed membranes were first incubated with rabbit polyclonal anti-hLf antibodies (1/1000 diluted whole antiserum produced in our laboratory) in a washing solution for 40 min, washed again, and incubated for 40 min with goat peroxidase-labeled anti-rabbit IgG antibodies (dilution 1/2500 in washing solution) (Biosys). Staining was achieved by using the Diaminobenzidine Peroxidase Substrate Tablet Set (Sigma) according to the manufacturer's instructions. All incubations were carried out at room temperature. Samples containing 6.25, 3.12, 1.56, 0.75, 0.37, 0.19, and 0.09 pmols of either unlabeled or 125 I-or 59 Fe-labeled hLf dissolved in 75 mM Tris/HCl, pH 7.3, were also applied to nitrocellulose paper as standards for further quantification. Quantification of hLf bound to porins was performed by densitometry analysis using a Hewlett-Packard 4C scanner and Biosoft Quantiscan software. Scatchard plot analysis was performed using Biosoft Enzfitter software.
Study of the hLf Porin-binding Site-Purified native OmpC and PhoE trimers (0.9 pmol) and 0.09 -3.12 pmols samples of hLf-derived proteins were immobilized on nitrocellulose paper that was blocked and washed as described above. Blots were then incubated with 125 nM of each of the hLf-derived proteins for 1 h at room temperature and immunostained as previously mentioned. Quantitation of porin-bound proteins was performed with Quantiscan software in comparison with the blotted protein standards.
Binding of hLf to E. coli Porin-deficient Mutants-Radioiodinated proteins are generally used for cell binding studies. However, because we have demonstrated that 125 I-hLf was unable to bind to purified porins, two different methods allowed us to compare the binding parameters of hLf to mutant strains EC1230, EC1231, EC1233, and EC1234. First, the affinity constants were estimated by flow fluorocytometry. In this method, bacterial strains were grown 3-5 h in LB medium at 37°C to an A 600 of 0.5 (5 ϫ 10 8 cells/ml), washed in M9 medium supplemented with essential amino acids, pH 7.3 (29), harvested, and resuspended in M9 medium containing 0.2% hTf (w/v) to minimize nonspecific binding to plastic vials and to bacteria. Aliquots of a 200-l cell suspension containing 10 6 cells were incubated with hLf concentrations ranging from 0 to 250 nM for 45 min. After washing, bacterial cells were incubated for 45 min in M9 medium containing 0.2% (w/v) bovine serum albumin and purified rabbit fluorescein isothiocyanate-labeled anti-hLf antibodies prepared in our laboratory. Green fluorescence bound to bacteria was then measured using a Becton-Dickinson FACScan Flow Cytometer. Bacteria were gated for forward and side angle light scatters, and 10,000 events of the gated population were analyzed. The 1024 fluorescence channels were set on a logarithmic scale, and the mean fluorescence intensity was determined. The affinity constants of hLf to E. coli mutant strains were calculated by Scatchard plot analysis using the Biosoft Enzfitter software.
The lack of internal hLf standard in cytometry experiments led to 59 Fe-hLf binding studies to estimate the total number of hLf-binding sites on bacteria. For this, 10 8 bacterial cells in 1 ml of M9 medium were incubated with 250 nM 59 Fe-hLf for 45 min. Then, cells were washed three times with M9 medium, and pellets were counted on a Compugamma LKB-Wallac ␥-radiation counter. Nonspecific binding of hLf to bacteria was determined by adding a 100-fold excess of cold ironsaturated hLf to the incubation mixture. All incubations were performed at room temperature.
Planar Lipid Bilayer Assays-All experiments were carried out using double quartz-distilled water, reagent grade chemicals, and a septum pressed between two Teflon cups. Schindler (30) technique bilayers were formed across a 200-m hole in a 12-m-thick Teflon septum pretreated with a solution of 2% n-hexadecane in n-hexane. Conductance measurements and the criteria for bilayer formation were as described (19). The trans compartment was held to virtual earth. The sign of the membrane potential referred to that on the cis side of the membrane, and the values quoted, therefore, refer to V cis -V trans . 50 -85 pmols of OmpC, OmpF, or PhoE native purified porins were then added to the subphase on the cis side of the preformed bilayers with the aqueous solution stirred by a magnetic bar. The membrane current was amplified with a Burr Brown current-voltage converter with an operational amplifier and feedback resistors ranging from 10 8 to 10 9 ohms. Solutions of hLf or bLf at final concentrations of 56 -560 nM in a 10 mM Tris acetate, 5 mM CaCl2, 0.1 M NaCl, pH 7.0, were incubated in the cis compartment at a 50 mV potential to favor interactions of hLf with porins. Recordings were filtered at 1 kHz with a EG and G low-pass filter and recorded on a Racal FM tape recorder. All experiments were performed at room temperature in a Faraday chamber.
Minimal Inhibitory Concentration Assays-Bacterial strains EC1230, EC1231, EC1233, and EC1234 were grown 3-5 h in LB medium at 37°C to an A 600 of 0.5 (5 ϫ 10 8 cells/ml) and then diluted in 1% Bacto-tryptone (w/v) (Difco) to obtain 2 ϫ 10 5 bacteria/ml. Five l of this cell suspension were added to 250 l of 1% Bacto-tryptone (w/v) buff-ered at pH 7.3 with 20 mM sodium phosphate and containing ironsaturated hLf at concentrations of 0.12, 0.5, 1.2, 2.4, 3.6, and 16 mg/ml that was previously sterilized by filtration on 0.1 m sterile filters (Millipore). Incubations were performed in 5-ml sterile glass tubes at 37°C for 18 h under gentle shaking. Bacteria were then diluted and grown for 24 h at 37°C on LB agar plates for counting. The sample without hLf served as the negative control.

RESULTS
Binding Parameters of Lf to Purified Porins-We tested whether hLf was able to bind to nitrocellulose-immobilized porins, either in the SDS-and heat-denatured monomeric form or in the native trimeric form. Binding studies were performed with 125 nM unlabeled and iron-saturated hLf followed by immunostaining. As shown in Table I, hLf bound to native trimeric OmpC and, in a lesser extent, to denatured monomeric OmpC with hLf/porin molar ratios of 1.9 and 0.8, respectively. Human Lf bound to native trimeric PhoE in a way similar to that of OmpC trimers with a molar ratio of 1.8, but no binding was observed when PhoE porins were denatured (monomeric form). Concerning OmpF, a very low binding of hLf to either denatured or native forms of the porin was noted. The molar ratio was about 0.1. Similar results were obtained when either nonsaturated (apo) or 59 Fe-saturated hLf was used as a probe (data not shown). However, the binding of 125 I-labeled hLf was only observed on denatured porins OmpC and PhoE with molar ratios as low as 0.2 and 0.1, respectively (data not shown).
Because hLf exhibited significant binding to the OmpC and PhoE porins in the native trimeric form, further studies were performed only with these two porins. The binding of ironsaturated hLf to native porins OmpC and PhoE was estimated at pH values ranging from 5.0 to 8.0. Maximal binding of hLf to OmpC and PhoE occurred around pH 6.6 and 7.5, respectively (data not shown). On the basis of these results, an intermediate pH value of 7.3 was used to determine the hLf binding parameters to porins according to the Scatchard method (28). As shown in Fig. 1, the binding of hLf to porins was both concentrationdependent and saturable at concentrations ranging from 10 to 250 nM. Calculated dissociation constants (K d ) were 39 Ϯ 18 and 103 Ϯ 15 nM, and hLf/porin trimer molar ratios were 1.9 Ϯ 0.1 and 1.8 Ϯ 0.2 for OmpC and PhoE, respectively.
Location of the hLf Porin-binding Site-To localize the porinbinding site of hLf, the binding of eight hLf-derived proteins to native OmpC and PhoE was assayed. Furthermore, hTf and bLf were used to evaluate the specificity of hLf binding. Table  II shows that none of the N-and C-terminal fragments or the N2 fragment of hLf significantly bound to OmpC or PhoE. Furthermore, the absence of residues 1-3, 1-4, and 1-5 in hLf strongly decreased the binding of the protein to PhoE. The binding of hLf Ϫ3N , hLf Ϫ4N , and rhLf Ϫ5N to PhoE was only 26, 24, and 9% of that of intact hLf, respectively. Similar results were obtained with OmpC, except that the deletion of residues 1-3 (hLf Ϫ3N ) from hLf did not affect the binding of the protein.
The binding of hLf Ϫ4N and rhLf Ϫ5N to OmpC was 42 and 7% of the native hLf control, respectively. Furthermore, Table II shows that rhLf-EGS and rhLf-SAST, which correspond to hLf modified at sequences 28 -34 and 39 -42, respectively, also exhibited low binding capacities to both OmpC and PhoE. Lastly, whereas the binding of hTf to the porins was not detectable, bLf bound to OmpC and PhoE in a way similar to that of hLf.
Effect of hLf on the Permeability of Porins-The effect of hLf on the channel conductance of OmpC, OmpF, and PhoE integrated in azolectin bilayers was studied. Porin trimers were incorporated into these bilayers by injecting detergent-solubilized protein into the bathing solution (cis compartment). The appearance of channels was shown by the stepwise increases in current when voltage clamped at 50 mV. When the potential across the bilayer was raised from 80 to 200 mV, downward current steps confirmed the closure of individual monomers. Depending on the experiment, conductance measurements revealed 1-4 porin trimers integrated into the bilayers. Raising the hLf concentrations to 560 nM significantly increased the frequency but not the amplitude of the current fluctuations (Fig. 2b). Fluctuation did not occur when the polarity of the membrane potential was reversed (Fig. 2b). When bLf was used instead of hLf, a blocking effect on OmpC was also observed but at a lower frequency (Fig. 2c). In contrast to OmpC, neither PhoE (Fig. 2d)  Binding of hLf to E. coli Porin-deficient Mutants-To assess the role of OmpC or PhoE in binding hLf to the bacterial cell surface, the binding parameters of hLf to the isogenic E. coli K12 strains EC1230 (PhoEϩ), EC1231 (no major porins), EC1233 (OmpFϩ), and EC1234 (OmpCϩ) were investigated. Because we observed that 59 Fe-hLf, but not 125 I-hLf, was able to bind to purified porins, the radioiron-labeled protein was used as a probe. However, because of its low specific radioactivity (about 5000 cpm/g), 59 Fe-hLf was unsuitable to perform Scatchard analysis at low ligand concentrations. Therefore, we carried out flow cytofluorometry studies using immunorevealed hLf to estimate the dissociation constants (K d ) of the ligand. Then, the number of binding sites/bacterium was compared through binding experiments with radioiron-labeled hLf. As reported in Table III, lower K d were calculated for the strains expressing OmpC and PhoE than for the strains expressing OmpF or no major porins at all. Moreover, 40 -50,000 more binding sites/bacterium were calculated for both OmpCϩ and PhoEϩ strains compared with the two other strains (Table III).
Effect of hLf on the Growth of E. coli Mutant Strains-Because preferential binding of hLf occurred on OmpC and PhoE either in the isolated form or present on E. coli K12, we investigated whether this phenomenon would affect the growth of bacteria. To avoid iron deprivation of the medium by hLf and subsequent bacteriostasis as described previously (3,4) and because holo and apo hLf are able to bind to porins in the same manner, iron-saturated hLf was used for these experiments. Minimal inhibitory concentration assays were performed on  (Fig. 3). As shown in Fig. 3a, the growth of only one strain, EC1230 (PhoEϩ), was totally inhibited by hLf at concentrations above 2.4 mg/ml. No clear growth inhibition was evidenced for the other strains (Fig. 3, b-d).

DISCUSSION
Porins were pointed out as potential targets for the binding of Lf to the surface of Gram-negative bacteria (13). However, the target porins were not clearly identified. Furthermore, the mechanisms by which Lf binds to porins and, hence, exerts its antibacterial activity have not been thoroughly investigated. Here we show that hLf binds with high affinity to purified native OmpC and PhoE trimers but not to OmpF. This binding also occurs when OmpC and PhoE porins are present on the surface of E. coli K12 mutant strains. Indeed, whereas binding of hLf to OmpF-expressing cells was not significantly greater than the porin-deficient cells, 40,000 -50,000 extra binding sites were present on OmpC-and PhoE-expressing strains. Furthermore, 2 and 3.5 times higher affinities, respectively, were observed on these mutant strains. Because the E. coli strains differ only by the expression of the major porins (18), it can be concluded that OmpC and PhoE porins constitute functional hLf-binding sites at the surface of bacteria. However, hLf bound to approximately 70,000 sites on EC1231 (no major porin). Thus, other binding sites may exist on E. coli K12.
According to x-ray diffraction studies, porins are organized as trimeric complexes, and each monomer is folded as a ␤-barrel built by 16 anti-parallel ␤-strands with short turns aligning the channel at the periplasmic side and long loops extending into the medium (31). Most loops, i.e. L1, L2, and L4 -L8, are exposed at the cell surface and exhibit reactivities to various phages and colicins (32)(33)(34). A striking feature of porin loops is the presence of several acidic amino acid patterns, e.g. sequence 290 DDED 293 of OmpC in L7 (Fig. 4). Because of its basic charge (pI 8.5-9), hLf could interact in a nonspecific manner with negatively charged regions of porins. Nevertheless, the binding of hLf to purified OmpC and PhoE trimers cannot be attributed to simple electrostatic interactions because of the following: (i) hLf binding occurs with a high affinity and (ii) no significant hLf binding to OmpF was noted despite the presence of acidic patterns in its amino acid sequence. As shown in Fig. 4, all the surface-exposed loops of OmpC are either shorter or longer than the OmpF counterparts. When calculated using the Alignp program from the Fasta package (35), sequence identities between OmpC and OmpF range from 35.3% for L6 to less than 10% for L4 and L1. It can thus be assumed that hLf binds to structural motifs that are present on OmpC but not on OmpF. L1 of OmpC, which poorly matches with the OmpF sequence, is 57.1% identical to L1 of PhoE and contains a

TABLE II
Binding of hLf-derived proteins to purified OmpC, OmpF, and PhoE porin trimers Solid-phase ligand-binding assays were performed as described under "Experimental Procedures." Results are expressed as the amount of hLf-derived protein (pmol) bound to 1 pmol of purified native porin and are given as mean Ϯ S.E. for three independent experiments. 0, below limit of detection (Ͻ0.2 for the hLf N2 fragment and Ͻ0.1 for the other proteins).  (Fig. 4). Therefore, L1 may participate in the binding of hLf. The N-terminal sequence 2 RRRR 5 and a loop 28 RKVRGPP 34 of hLf interact with many target molecules of mammalian cells such as receptors (36 -39), DNA (11), and proteoglycans (11, 24, 40 -42). It was also hypothesized that another loop region 39 KRDS 42 , which inhibits platelet aggregation (43), is a part of the hLf receptor (36). These three regions are present in human Lfcin, an antibacterial peptide consisting of residues 1-47 of hLf (5). Moreover, 2 RRRR 5 and 28 RKVRGPP 34 were previously shown to bind with a high affinity to the lipid A moiety of LPS (9 -12). Our results demonstrate that residues 2-5, but also loops 28 -34 and 39 -42 of hLf, are involved in the interactions with porins OmpC and PhoE. First, as shown in Table II, the C-terminal fragment (residues 284 -692), the N-terminal fragment (residues 4 -283), thus lacking 1 GRR 3 , nor the N2 fragment (91-255) isolated from hLf significantly bound to purified porins. Furthermore, sequential removal of the N-terminal amino acid residues 1 GRRRR 5 of hLf gradually but strongly decreased the binding of hLf to porins. As it could be expected from the estimated binding parameters of hLf to purified porins, the participation of residues 2-5, 28 -34, and 39 -42 in binding to either OmpC or PhoE is not identical. Whereas the modification of loop 28 -34 inhibited binding to both porins to a similar extent, modification of loop 39 -42 inhibited more efficiently the binding of hLf to PhoE. It was also observed that removing residues 1-3 from hLf altered its binding to PhoE but not to OmpC. Fig. 5 shows that three residues of hLf within regions 1-5, 28 -34, and 39 -42 (Arg 4 , Lys 29 , and Arg 40 ) are identical to bLf, but different from hTf. These basic residues are potential candidates for porin binding. However, because iodination of hLf prevents the binding to porins but no tyrosine residue lies in the 1-50 region, other amino acids may interact with porins. This is supported by our finding that both hLf and bLf alter the permeability of OmpC and thus probably fit closely to several surface-exposed loops of the porin channel.
Porins have traditionally been described as permanently open pores (44). High voltages allow the closure of pores, but the physiological relevance of this effect is still unclear (45). Polyamines such as cadaverine, which is normally associated with the outer membrane of E. coli, were shown to promote closures of porins and therefore could be regulators of OmpC and OmpF activity in vivo (46). We show that Lf, in a similar way to polyamines, induces open-closed transitions of the FIG. 2. Effect of hLf and bLf on the conductance of porins OmpC and PhoE integrated in a planar lipid bilayer system. A, after activation of four OmpC porin trimers, hLf (56 nM final concentration) was injected in the cis compartment at 100 mV applied potential. After a short time lag, the current trace became noisy. The inset shows that this noise was because of single channel fluctuations. Upon reversal of the membrane potential to Ϫ100 mV, the noise disappeared. B, increasing ten times the hLf concentration increased the noise. C, same experiment as in b but with 560 nM bLf. Three OmpC porin trimers were activated. D, in this experiment, only one PhoE trimer was activated. Because PhoE is more sensitive to high voltage than OmpC, closing events of PhoE single channels occurred. However, the injection of hLf (56 or 560 nM) did not affect the conductance. The applied voltage was ϩ110 mV. OmpC channel and thereby decreases the total amount of ion flux across the planar membrane. The effect is more pronounced with hLf than with bLf and is observed only at positive voltage. This is in agreement with the basic nature of the hLf region predicted to interact with the porin. At positive voltage, the electric field in the pore mouth will promote plugging of the pore by positively charged ligands, whereas negative potentials will repel them. No evidence of interactions of hLf with PhoE could be obtained from conductance studies. This apparent contradiction with direct binding studies can be resolved considering that the PhoE channel is anion selective, whereas OmpC porin is cation selective. The channel lumen of PhoE contains an additional lysine residue (31) that may prevent direct interaction of hLf with the constriction of the pore as hLf may do with OmpC. This hypothesis supports the idea that there is more than one contact site between hLf and porins.
The relationships between bactericidal activity and binding of Lf in numerous Gram-negative bacteria have been previously reported (7,8), but the mechanism(s) by which hLf exerts its activity at the membrane level was not clearly elucidated. Both lactoferrin and its pepsin-derived peptide fragment, Lfcin (5), cause membrane disruption thus leading to the release of LPS (47). Affinity sites for Ca 2ϩ and Mg 2ϩ on the pyrophosphoryl-and/or phosphodiester groups of the ketodeoxyoctulonic acid-lipid A region and on the ketodeoxyoctulonic acid trisaccharide unit of LPS play an important role in the assembly or maintenance of the organization of the outer membrane (48). Therefore, it has been hypothesized that Lf or Lfcin, in a manner similar to EDTA, may chelate divalent cations or bind to the cation-binding sites of LPS (47,49). Evidence of direct interactions between either Lf or Lfcin and LPS strongly supports the second hypothesis (10,12,50). Lf was shown to interact with the lipid A moiety of LPS (9). Elsewhere, Lfcin and Lfcin-derived basic helicoidal peptides bear similarities with several classes of natural antimicrobial peptides, which also interact with the cation-binding sites of LPS (50 -52). However, the mode of action of these peptides is still much debated. It was proposed that cationic amphipathic ␣-helical peptides form ion channels through membrane bilayers (53). The basic amphipathic region of Lfcin (residues 21-31 in hLf) adopts an ␣-helical structure, which is critical for its antimicrobial potency (50,54). Furthermore, the homologous region in bovine Lfcin appears able to adopt a helical or sheet-like conformation similar to what has been proposed for the prion proteins and Alzheimer's peptides (52). It was hypothesized that membrane interactions alone are capable of inducing a similar ␣-to-␤-transition in the N-terminal region of intact Lf (52). Lf has a less potent bactericidal effect than Lfcin and its derived peptides (5,50,54). This may have been explained by the following: (i) the degree to which the peptides can interact with the bacterial membrane, as a result of their increased flexibility and (ii) their ability to enter the outer membrane and to reach the cytoplasmic membrane (50,54). Another explanation is that the carbohydrate moiety of LPS may prevent or limit the binding of Lf, a larger molecule than Lfcin, to the outer membrane. We detected about 10 5 hLf bound to one E. coli cell, whereas a 10-fold higher binding of Lfcin was reported (55). In the present work, we show that OmpC and PhoE provide anchoring sites on E. coli that may facilitate the accessibility of hLf to the outer membrane. As illustrated in Fig. 6, tight interactions between Lf and OmpC probably limit ion and nutrient fluxes through the outer membrane but do not lead to a marked growth inhibition. In contrast, Lf binds to PhoE in a way that does not limit the permeability of the porin but permits an antibacterial effect. We postulate that preliminary binding of Lf to PhoE with a slightly lower affinity than to OmpC allows further interactions of the protein with the outer membrane thus facilitating its destabilization (Fig. 6).
In conclusion, our findings show that hLf may use OmpC and PhoE porins of E. coli as high affinity anchoring sites. They open the way to investigating the nature and physiological relevance of hLf binding to porins and other outer membrane proteins of pathogenic bacteria.  (60). Brackets indicate the regions of hLf that were modified by limited proteolysis or mutagenesis.