Intramembrane helix-helix interactions as the basis of inhibition of the colicin E1 ion channel by its immunity protein.

It had previously been hypothesized that the ability of a small number of immunity protein molecules in the cytoplasmic membrane to confer protection against the lethal effects of a channel-forming colicin involves a complex stabilized by electrostatic or polar interactions between immunity protein, the colicin channel, and specific sites on the cytoplasmic membrane surface defined by the presence of the tol gene translocation proteins. The hypothesis was tested (a) by constructing a hybrid colicin molecule, IaE1, containing the E1 channel domain, and the translocation and receptor domains of Ia, and (b) by altering charged residues in all peripheral regions of the immunity protein to neutral residues. It was concluded that the specificity of immunity protein requires neither specific translocation proteins, nor a specific arrangement of charged residues in the immunity protein. (c) In addition, by making 65 site-directed mutations, "immunity by-pass" mutants were found at five different loci, Ala474, Ser477, His440, Phe443, and Gly444, on two proposed membrane-spanning helices of the open colicin channel, one hydrophobic (A471-A488) and one amphiphilic (V441-W460). The mutants in the hydrophobic helix showed a larger bypass effect. The "bypass" phenotype could be assayed by (i) cytotoxicity and (ii) K+ efflux in imm+ cells caused by a bypass mutant but not wild-type colicin. It is concluded that the immunity protein exerts its specific effect through rapid lateral diffusion in the cytoplasmic membrane and helix-helix recognition and interaction with at least one hydrophobic and one amphiphilic trans-membrane helix of the colicin channel. Interaction with the amphiphilic helix implies that the immunity protein can react with the channel in the open state.

Intramembrane Helix-Helix Interactions as the Basis of Inhibition of the Colicin E l Ion Channel by Its Immunity Protein* (Received for publication, November 10, 1992, andin revised form, January 19, 1993) Yan-Liang Zhang and William A. Cramerz It had previously been hypothesized that the ability of a small number of immunity protein molecules in the cytoplasmic membrane to confer protection against the lethal effects of a channel-forming colicin involves a complex stabilized by electrostatic or polar interactions between immunity protein, the colicin channel, and specific sites on the cytoplasmic membrane surface defined by the presence of the to1 gene translocation proteins.
The hypothesis was tested (a) by constructing a hybrid colicin molecule, IaE 1, containing the E l channel domain, and the translocation and receptor domains of Ia, and ( b ) by altering charged residues in all peripheral regions of the immunity protein to neutral residues. It was concluded that the specificity of immunity protein requires neither specific translocation proteins, nor a specific arrangement of charged residues in the immunity protein. (c) In addition, by making 65 site-directed mutations, "immunity by-pass" mutants were found at five different loci, Ala4'", Ser4", His440, Phe443, and G1y444, on two proposed membrane-spanning helices of the open colicin channel, one hydrophobic (A471-A488) and one amphiphilic (V441-W460). The mutants in the hydrophobic helix showed a larger bypass effect. The "bypass" phenotype could be assayed by (i) cytotoxicity and (ii) K+ efflux in imm+ cells caused by a bypass mutant but not wild-type colicin. It is concluded that the immunity protein exerts its specific effect through rapid lateral diffusion in the cytoplasmic membrane and helix-helix recognition and interaction with at least one hydrophobic and one amphiphilic trans-membrane helix of the colicin channel. Interaction with the amphiphilic helix implies that the immunity protein can react with the channel in the open state.
The channel-forming colicins are toxin-like molecules that exert their cytotoxic effect on susceptible Escherichia coli cells by forming a highly conductive ion channel in the cytoplasmic membrane that depolarizes the membrane (Gould and Cramer, 1977). The depolarizing channel results in inhibition of active transport, depletion of intracellular ATP and K+ levels (Phillips and Cramer, 1973;Kopecky et al., 1975), and subsequent cell death. The channel-forming colicin molecule, like other colicins and toxins, is divided into functional domains. The COOH-terminal third of the colicin E l channel molecule * This research was supported by National Institutes of Health Grant GM-18457. 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 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence should be addressed. Tel.:  possesses channel activity whose in vitro properties resemble those of the whole colicin molecule (Dankert et al., 1982;Bullock et al., 1983). The structure of the membrane-embedded voltage-gated ion channel includes (i) a hydrophobic segment identified as a helical hairpin in the structure of the soluble colicin A peptide (Parker et al., 1992) that may  or may not (Parker et al., 1992) be an intrinsic part of the channel lumen; (ii) it also contains an amphiphilic helical hairpin whose voltage-dependent reversible insertion into the membrane may be responsible for channel gating (Merrill and Cramer, 1990). The COOH-terminal channel domain also contains the region of interaction with the protective immunity protein (Bishop et al., 1985), coded by the 4.2 megadalton ColEl plasmid.
Immunity protein produced in association with the plasmidencoded colicins protects the cell against the lethal action of the colicin with which it is produced and against exogenous homologous colicin produced and extruded into the medium by other cells (Bazaral et al., 1968). This protective mechanism is a unique aspect of the interaction of colicins on membranes, compared with other toxin and toxin-like molecules. Furthermore, because the immunity proteins of the channel-formingcolicins E l (Bishop etal., 1985), Ia, Ib (Mankovich et al., 1986), and A (Geli et al., 1988) are hydrophobic proteins and are known to be localized in the E. coli cytoplasmic membrane, the protective interaction between the proteins occurs within the two-dimensional space of this membrane. The orientation and topography of the immunity proteins of colicins A (Geli et al., 1989a) and E l  have been determined. The colicin E l immunity protein spans the membrane bilayer three times, with NH2 and COOH termini on its cytoplasmic and periplasmic sides, respectively .
Protection by the immunity proteins for colicins E2 (Schaller andNomura, 1976) andE3 (Sidikaro andNomura, 1974;Jakes, 1982) against the degradative effects of these colicins could be demonstrated in vitro in aqueous solution by the ability of the respective immunity proteins to react directly with the COOH-terminal domains of the respective colicins to prevent the inhibition of DNA and protein synthesis caused by colicins E2 and E3. In the case of the channelforming colicins A and El, it has thus far not been possible to demonstrate i n vitro inhibition by immunity protein of colicin function (Geli et al., 198913;Shirabe et al., 1993).
The problem of the mechanism of action of immunity protein on the channel-forming colicins is interesting because of: (i) the uniqueness of this mechanism; (ii) the ability of a small number of immunity protein molecules (-102-103) 10cated in the cytoplasmic membrane to prevent the formation by colicin of a functional and lethal channel in that membrane; (iii) the high specificity of the immunity protein produced in conjunction with each colicin for that colicin, in spite of the high degree of sequence identity between some of  (Dankert et al. 1982); *, residues of Sites of immunity bypass in colicin E l channel domain. Numbers refer to sequence of colicin E l .

__~.__
colicin E l t h a t are identical or quasi-identical to those of colicin Ia or Ib.
the channel domains (58% between colicin E l with the closely related colicins Ia or Ib). It was previously proposed that the ability of a small number of immunity protein molecules in the cytoplasmic membrane to exert its specific inhibitory reaction could result from a ternary protein reaction involving (i) the immunity protein, (ii) the channel domain of the colicin, and (iii) the tol or ton translocation proteins at unique sites in the membrane associated with the attachment of the latter . This hypothesis has been tested in the present study. It was found to be incorrect, but evidence has been obtained for direct interaction between the hydrophobic truns-membrane helices of the colicin E l immunity protein and the trans-membrane helices of the channel domain of this colicin.

MATERIALS A N D M E T H O D S
Strains and Plasmids-E. coli JMlOl was used as the indicator strain for assay of cytotoxicity. JMlOl was also transformed with pSKHY(-)amp' imm" to test immunity bypass function and with pSK(+)amp' for comparison of cytotoxicity. The colicin-tolerant mutants TolA592 and TPS13 (tolQ) were kindly provided by R. E. Webster. TolA592 and TPSl3 were transformed with pSKEl imm' for assay of the dependence of hybrid colicin activity on immunity protein. E. coli BW8983 obtained from Dr. B. Wanner is tonB-tolA' and was used to determine cytotoxicity and translocation-dependence of the hybrid colicin IaE1. E. coli MV1190 (dut', ung') and CJ236 (dut-, ung-) (Bio-Rad) were the strains used for mutagenesis. Helper phages M13K07 or R408 were obtained from Promega (Madison, WI) for preparation of single strand DNA. The pBluescript I1 SK(+) phagemid (Stratagene, La Jolla, CA) was used for cloning, mutagenesis, and overexpression of the colicin IaEl hybrid colicin. pSKEl(-) was used for mutagenesis of colicin E l mutants . E. coli JK341/pJK1 obtained from J. Konisky was the source of colicin Ia and its immunity protein (Mankovich et al., 1986).
Construction ofpSKZaEl-The hybrid IaEl colicin used a fusion site at the start of the colicin E l channel domain defined by the tryptic cleavage site between Lys335 and Ala336 (Table I). pJKl was the source vector from which the cia and iia genes were isolated. pJK1 is a pBR322 vector encoding cia and iia genes that carries an EcoRI fragment at its multicloning site (Fig. 1, A and B ) . The EcoRI fragment was subcloned into pSK(+) (pBluescript SK, Stratagene product), generating pSKIa to allow both mutagenesis and overexpression. pSKHY, the expression vector for the colicin E l channel domain, contains SacII sites at the 5'-end of this domain and at the The abbreviations used are: imm, immunity protein genes, cea, colicin E l structural protein gene; cia, iia, colicin Ia structural and immunity protein genes; PAGE, polyacrylamide gel electrophoresis. multicloning site downstream of the 3'-end of the imm gene . The presence of the imm gene in the plasmid carrying the hybrid colicin IaE1, in which coding regions of the translocation and receptor-binding domains were contributed by colicin Ia, and the channel domain by colicin E l , allows the immunity protein to protect the expression host from the cytotoxic effect of hybrid IaE1. A Sac11 site was created in pSKIa at the anticipated position between the coding regions for the colicin Ia receptor-binding and channel-forming domains, based on sequence alignment of colicin Ia and El. As in pSKHY, another SacII site in pSKIa is in the multicloning site downstream from the 3'-end of the iia gene. pSKIa was then submitted to SacII digestion, deleting the region encoding the colicin Ia channel domain and iia gene. pSKHY was also submitted to SacII digestion, generating the fragment encoding the colicin E l channel domain and the imm gene. The fragment from pSKHY was inserted into the pSKIaA vector, generating pSKIaE1. The correct orientation of the fragment was determined from the pattern of restriction enzyme digestion.
Ouerexpression and Purification of Colicin IaEZ-Because it is likely that the colicin Ia gene is also under control of the SOS system, E. coli strain IT3661 which is lexA-, was chosen for constitutive overexpression of hybrid colicin IaE1, and the region containing the regulatory elements upstream of the colicin Ia gene was left intact and included in the fragment for cloning. IT3661 was then transformed with pSKIaE1. This cell strain was incubated for 16 h in 2 X YT media before testing the cell lysate. A protein with M, = 66,000, approximately the calculated molecular mass of the hybrid colicin, was shown by SDS-PAGE to be overexpressed.
Site-directed Mutagenesis-The method of Kunkel (1985) was followed and utilized strains MV1190 (dut+, ung') and CJ236 (dut-, ung-), (Bio-Rad 170-3576; Bio-Rad mutagenesis manual, 1989). pSKEl(-) was transformed into CJ236, and single strand DNA was isolated from the phage pellet by infecting the cells with the helper phage M13K07. After annealing the mutant oligonucleotide with the template, native T 7 DNA polymerase and T4 ligase were incubated (30 min, 37 "C) to synthesize the mutant strand in the presence of dNTP, the closed heteroduplex transformed into MV1190, several colonies picked, and plasmid DNA isolated from each. Mutants were identified by double strand DNA sequencing of the pSKE(-) plasmid.
Oligonucleotide Synthesis-All oligonucleotides used for the mutagenesis were constructed in the Laboratory for Macromolecular Structure at Purdue University. Mutants were made by using a separate oligonucleotide for each desired mutation. 19-and 30-mers were used for introduction of single base mismatches at one or two sites, and four sites, respectively. Oligonucleotides were directly used for mutagenesis.
DNA Sequencing-DNA sequencing for the screening of mutants was done with a Bethesda Research Laboratory model S2 apparatus. Sequenase (U. S. Biochemicals, Cleveland, OH), a modified T7 DNA polymerase, was used for both single and double strand DNA sequencing, following the protocol in the Sequenase kit (70770). For double strand DNA sequencing, template DNA was denatured by alkaline pH (Chen and Seeburg, 1985).
Cytotoxicity Assay-The cell strains, JMlOl and JMlOl/pSK, used as the colicin E l sensitive indicator, and JMlOl/pSKHY(-), used as imm+ bypass indicator, were grown overnight to early log phase and 100 pl was spread on 2 X YT or 2 X YT (+75 pg/ml ampicillin) plates. From a 10 pg/ml stock solution of the colicin, 5-fold serial dilutions to a minimum colicin concentration of 3.2 or 16 ng/ml were assayed. The range between 10 pg/ml and 16 ng/ml was chosen because it was found that all mutants showed observable activities within this range. 20 ~1 from each dilution was spotted directly on a plate that had been overlaid with the indicator cells. Plates were incubated at 37 "C overnight (8-12 h), and the lowest dilution of mutant colicin that gave clearing of a spotted site was used to calculate specific activity.
Purification of Wild Type and Mutant Colicin El-The DM1187 lexA-strain, which like the IT3661 strain produced colicin E l constitutively, was used for overexpression of the cea gene . DM1187/pSKEl(-) cells were grown overnight, harvested, resuspended in 50 mM sodium phosphate buffer, pH 7.0, with 2 mM EDTA, and broken in a French Press. The lysates was sedimented at 10,000 X g for 15 min, the supernatants loaded on a Mono-S fast protein liquid chromatography column at 6 "C and eluted with an NaCl gradient. The purity of the resultant colicin E l was routinely examined on an SDS-PAGE Phast gel system (Pharmacia LKB Biotechnology Inc.).
Protein Assay-Modified Lowry (Pierce Chemical Co. 23240x1 and BCA assays (Pierce 232356) were used to determine protein concentration. Purified colicin was extensively dialyzed against distilled water and lyophilized to make a 2.0 mg/ml solution as an internal standard.

Measurement of in Vivo K' Efflux-Cells transformed with
pSKHY (imm') or with the empty vector pSK (imm-) as a control were inoculated into 2 X Y T (+50 pg/ml ampicillin) medium. At early log phase, the cells were harvested, washed once with 0.4% glycerol, 0.5 mM KC1, 50 mM sodium phosphate, pH 7.0, resuspended in the same buffer at a cell density of 1.0 X 109/ml, and incubated with stirring a t 37 "C for about 20 min, to allow the cells to reaccumulate potassium. K+ efflux caused by colicin E l was measured with a potassium selective electrode (Orion model 93-14) immersed in 10 ml of the stirred cell suspension (Phillips and Cramer, 1973). Cytotoxicity of colicin added to the cell suspension was assayed from the same sample used for measurement of K+ efflux. The colicin multiplicity (m) was determined from the survival (S/So) level, m = ln(So/ S), of cells used in the K+ efflux assay.

RESULTS
In order to explain the specificity of the colicin E l immunity protein, and the ability of a small number (-lo2 -lo3) of immunity protein molecules to neutralize an inserted colicin channel molecule in all regions of the cell surface, it was proposed ( a ) that the immunity protein interacts in a ternary complex with the colicin channel domain and with the tolA and/or tolQ translocation proteins of the cell envelope that are localized in the cytoplasmic membrane . ( b ) It was also proposed that the extrinsic charged or polar segments of the immunity protein ( Fig. W ) would interact with extrinsic complementary regions of the channel domains and translocation proteins ( Fig. 2B) in order to provide the specificity of the interaction. Given the membrane topography and orientation of the immunity protein  and the orientation of the tolA-tolQ proteins (Levengood and Webster, 1989;, it was hypothesized that (i) the periplasmic L1 and COOHterminal T2 regions of the immunity protein might interact with the large COOH-terminal domain of tolA and/or the periplasmic NH2 terminus and exposed loops of tolQ, or (ii) A FIG. 2. Topographical models. A , the colicin E l immunity protein, with trans-membrane helices labeled Hl-H3, and the peripheral regions NH2-terminal T1, loop L1, loop L2, and COOH-terminal region T2; charged residues are circled, and residues mutagenized in the present work marked by an arrow. B, model of the channel domain of colicin E l in the "open" state; region upstream of is accessible to several proteases, but probably also contains a membrane-bound segment (Zhang and Cramer, 1992). B the highly charged cytoplasmic segment L2, as well as the NH, terminus TI of the immunity protein ( Fig. 2.4) might interact with the NH2-terminal region of the tolA protein and/or the cytoplasmic loop and COOH terminus of the tolQ protein.
The requirements of the to1 proteins and of extrinsic charges in the peripheral regions of the immunity protein for its function were tested as follows.
Construction and Characterization of Hybrid Colicin IaEl-The requirement of a particular array of to1 translocation proteins, or any individual to1 protein, for expression of the function of colicin E l immunity protein, was tested by making a hybrid colicin molecule, IaE1, whose translocation requires the tonB (Postle, 1990) instead of the to1 gene products.
A protein with approximately the same electrophoretic mobility, M , 66,000, of the hybrid colicin IaE1, containing the translocation and receptor domains of colicin Ia and the channel domain of colicin E l (predicted molecular weight, 67,861), was overexpressed (cf., "Materials and Methods") in E. coli strain IT3661 lexA-and purified (Fig. 3, lane 3 ) . It can be seen that colicin E l , with a molecular weight of 57,279 (Yamada et al., 1982), runs on the gel (Fig. 3, lane 5 ) with a substantially smaller Mr value compared to the hybrid colicin in lane 3 (Fig. 3). The hybrid protein was purified by the same procedure used for wild type colicin E l , using a fast protein liquid chromatography Mono-S cation-exchange column, indicating that the protein has a similar chromatographic behavior. Western blotting showed that the protein could be recognized by the antibody against the colicin El COOH-terminal channel-forming domain (data not shown). The IaEl hybrid exhibited cytotoxicity at a concentration of 80 ng/ml toward the tolA-deficient strain tolA592, whereas colicin E l showed no activity a t a concentration of 10 pg/ml (Table IIA), showing the absence of the colicin E l translocation domain in the hybrid colicin.
TOM Translocation Protein Is Not Required for Action of Colicin E l Immunity Protein-The tolA-tolQ, and tonB proteins (Postle, 1990) are required for the translocation and in vivo activity of colicin E l or Ia, respectively, and are anchored in the cytoplasmic membrane.
The cell strain BW8983 (tonB-) in which the tolA and tonB translocation systems are functional and absent, respectively, is sensitive to colicin El, but not to the hybrid IaEl that contains the translocation and receptor binding domains for colicin Ia (Table IIB), confirming that the colicins IaEl and E l require the tonB and tolA systems, respectively, for translocation and activity. TolA592, in which the tonB and tolA systems are functional and non-functional, respectively, is sensitive to IaE1, but not to El, again showing that colicins IaEl and E l require the tonB and tolA systems for activity. A slightly reduced cytotoxicity of the IaEl hybrid toward the tolA-tonB+ strain compared to the activity of colicin E l toward the tolA+ tonBstrain (Table IIA versus IIB) may be due to a slightly lower translocation efficiency of the colicin E l channel domain by the Ia receptor and translocation domains.
The strain TolA592/pSKHY, in which colicin E l immunity protein is present (imm'), is sensitive to neither colicin E l nor colicin IaEl (Table IIC). Because colicin IaEl is translocated by the tonB system, but its action was inhibited by E l immunity protein in tolA592/pSKHY, it was concluded that structural interaction between colicin E l and its immunity protein does not require the tolA translocation protein.
Role of tolQ-Some evidence was also obtained for the lack of involvement of the tolQ protein in the action of immunity. Colicin E l was completely inactive in a tolQ-strain (Table  IID; Sun et al., 1986;Webster, 1991). The IaEl hybrid possessed a weak activity, observed a t concentrations 22.0 pg/ ml (Table IID). However, this activity was not expressed in the presence of immunity protein (Table IIE), indicating that the tolQ protein was not involved in the expression of immunity.
Modification of Peripheral Domains of Immunity Protein Does Not Affect Activity-The colicin E l immunity protein has been shown to span the membrane bilayer three times, with NHz and COOH termini on its cytoplasmic and periplasmic sides ( Fig. 2A). It was hypothesized that the specificity of the immunity protein might reside in its peripheral segments containing charged or polar residues ). This hypothesis was tested by site-directed mutagenesis of charges in all of the peripheral segments of the immunity protein. Charged residues in all peripheral segments including the terminal (T) and intermediate loop ( L ) regions, T1, L1, L2, and T2 (Fig. 2 A ) were changed to neutral residues as follows (Table 111): Arg4 + Gln4, in peripheral segment, T1, of the immunity protein ( Fig. 2A),

A~p~~-L y s~'
--$ A~n~~-G l n~' (in segment Ll), Arg73-Lys74 + Met73-Asn74 (in L2), Lys"' + Met"' (in T2), and A~p~~-Lys=-Lys"' + A~n~~-G l n~-M e t " ' (in segments Ll,T2). The configuration of charged or polar residues in COOH-terminal segment T2 was of particular interest. This region had been implicated from mutagenesis studies to be important for the activity of the colicin A immunity protein (Geli et al., 1986(Geli et al., , 1989, although it was previously noted that the structural requirements of the 113 residue E l immunity protein having three membrane spanning a-helices could be very different from those of the 178 residue colicin A imm protein . Changing the only positively charged residue, Lys"', in the colicin E l immunity protein COOH terminus (segment T2) to a neutral Met had no effect on the function of the imm protein, both in a single point mutant and in a triple mutant that also included changing 2 basic residues, Asp37 and Lys3", in the L1 region to neutral residues Asn and Gln (Table 111). The 2 positively charged residues Arg73 and in the highly charged cytoplasmic loop L2 were mutagenized to neutral MeC3 and but this double mutant also showed full activity. Removal of the single charged residue, Arg4, at the cytoplasmically located NH2 terminus was also without effect. Thus, it was concluded that a particular arrangement of charged or polar residues in the peripheral segments of the colicin E l immunity protein was not critical for activity.
Immunity "Bypass" Mutants in the Colicin E l Channel Effect on function of immunity protein of mutagenesis of the charged residues to neutral residues in the peripheral segments of the immunity protein ( Fig. 2A T2) a All mutant and wild type immunity protein carrying plasmids, and pSK (amp') cloning vector (as immcontrol to monitor colicin activity) were introduced into strain JM101. Colicin E l activity was assayed on amp' 2xYT plates. Assays were performed twice. + indicates clear zone and lack of activity of immunity protein;sign indicates no clearing associated with active imm protein; nt, not tested.
Brackets denote peripheral regions of imm protein shown in Fig.  2 A; this mutant and all those in this column have the imm+ genotype. JMlOl/pSK was used to monitor cytotoxicity of mutants and JMlOl/pSKHY to test the bypass effect of each mutant; each mutant was tested twice.
Ratio of minimum concentration of wild-type or mutant colicin E l that shows cytotoxic activity in immcompared to imm+ cells.
Domains-The approximate topography of the hydrophobic helical hairpin of the colicin E l channel domain (helices H3- Fig. 2B) was previously mapped by the relative cytotoxicity pattern within a large set of site-directed substitutions of charged for non-polar residues at 26 different positions in the hydrophobic region of the channel domain . These mutants showed diminished cytotoxic activity toward sensitive imm-cells compared to wild-type colicin El. This set of mutants in the H3-H4 region of the channel domain, supplemented with additional mutants in the Hl-HZ region (Fig. 2B), was examined for the existence of colicin E l molecules able to bypass the protective function of the immunity protein.

H4,
Among 65 mutants (Tables  IV and V) screened in the colicin E l channel domain (helices HI-H4, Fig. 2B), the following six, His440 "-f Arg, Phe443 -+ Lys, G1y444 -+ Lys, Ala474 * Glu, Ser477 -+ Lys, and Ser477 -+ Arg, were able to significantly bypass the protective function of the colicin E l immunity protein (Table IV). Whereas wild-type colicin added at a concentration of 200 wg/ml did not bypass or over-ride the immunity function, the Ser477 + Arg, + LYS, Ala474 + = Glu, Phe443 -+ Lys, -+ Arg, and G~Y~~ -+ LYS mutants bypass the imm function when added at concentrations of 0.4, 10, 10, 50, 50, and 100 d m 1 (Table  IV). An efficacy quotient index for the ability of a mutant allele to bypass the inhibitory effect of immunity protein is described by the ratio of the concentrations of mutant colicin that exerts a cytotoxic effect (clear zone on Petri plate) in the absence compared to the presence (imm') of immunity protein. By this criterion, the most pronounced bypass effect is displayed at residues Ser477 and Ala474 (Table IV, right-hand column) in helix H3 (Fig. 2B). Fifty-nine other mutants in the channel domain between residues 424 and 512 were tested and found to be unable to bypass the immunity protection function (Table V). Among these are the mutants Ser -+ Gly or Ala at position 477. Mutagenesis of the only cysteine in the channel domain of the colicin molecules, Cys505 -+ Ala, had no effect, ruling out the possibility that an intermolecular disulfide is involved in the interaction between the colicin channel and immunity protein, despite the occurrence of one cysteine in each of the three membrane-spanning helices of the immunity protein. The mutants Lys362 -+ Ile, Ala371 -+ Lys, -+ Lys, and Lys403 -+ Ile upstream of the four putative trans-membrane helices were also tested and found to be bypass-negative. The mutants Val4'' -+ Arg and Thr4" -+ Lys  were not active enough for the bypass assay.
Immunity Bypass Phenotype Detected by in Vivo Channel Activity-The rate of K+ efflux from sensitive cells treated with colicin E l has been shown to be correlated with the multiplicity of lethal colicin molecules adsorbed to the cells (Wendt, 1970;Phillips and Cramer, 1973). K+ efflux caused by addition of wild-type colicin E l in the absence of imm protein is characterized by a single channel activity of 6.5 x lo5 ions/channel/s ( n = 3) (Fig. 4a). The average channel conductance measured for in vivo K+ efflux in many ( n = 33) trials was 1.3 -+ 0.8 X lo6 K+/channel/s, which is comparable to the value of the single channel conductance, 3 x lo6 ions/ channel/s in 0.1 M NaCl of colicin E l measured in planar bilayers (Bullock et al., 1983). The specific rate of K' efflux of the colicin A channel was found to be 3 X lo5 K+/channel/ s (B6nkdetti et al., 1992). The presence of immunity protein is shown to prevent K+ efflux caused by wild-type colicin (Fig.   4b). The channel-forming ability of the immunity bypass mutant, Ser477 -+ Lys, in the presence of immunity protein is directly demonstrated by the similar rates of K' efflux caused by this mutant, added at a concentration approximately 100 times that of the wild-type colicin, to the sensitive imm-(6.0 x lo5 K+/channel/s, n = 2) and imm+ (1.5 x lo6 K'/channel/ s, n = 3) cells (Fig. 4, c and d).

Tests of Hypotheses for the Protective Immunity Protein Interaction
In a previous study of the topography and orientation of the colicin E l immunity protein in the cytoplasmic membrane, the following mechanism for specific interaction of immunity protein with colicin channel domain was proposed : (a) immunity protein, present in a low copy number (-10') per cell, diffuses laterally in the membrane until the specific sites of interaction including the colicin channel domain are found (b) the specific sites are the positions of entry into the membrane of the imported colicin channel, which would be defined by sites of apposition with the tolA and/or tolQ translocation proteins that would Location of mutations in colicin E l channel domain with no effect on immunity bypass JMlOl/pSK was used to monitor cytotoxicity of the mutants; JMlOl/pSKHY was used to test the bypass effect of each mutant; each mutant was tested twice.
The studies reported in the present study on retention of protective immunity function (i) in IaEl hybrid colicin acting on tonB'tolAor tonB'tol Qcells (Table 11), (ii) with sitedirected mutants of immunity protein in which charged residues were removed in each of the extramembrane loops T1, L1, L2, and T2 (Table 111), and (iii) through the existence of In the absence of a mechanism that would convey direction to the immunity protein produced in low copy number, the above hypothesis (a) of lateral diffusion of imm protein in the cytoplasmic membrane to the reaction sites is still regarded as correct. The measured half-time for K' loss from the cell population, after a delay attributed to binding and translocation, is -60 s, similar to a half-time for efflux of 40 s predicted from an intracellular K+ concentration of 0.25 M in 0.15 M osmolal medium (Richey et al., 1987), an intracellular volume of 5 X liter (Neidhardt, 1987), respectively, and a single channel conductance of -lo6 K+/s (Fig. 4). The mean-square distance, ? = 4Dt, over which a membranebound protein with a diffusion constant of -IO-' -10"' cm2 s-' (Cramer and Knaff, 1991) can diffuse during the lag time of 20-40 s that precedes channel formation (Fig. 4) is approximately 10-7-10-8 cm'. This area is greater than or equal to that of the cytoplasmic membrane surface, consistent with the concept that a single immunity protein molecule has a reasonable probability of finding the colicin channel molecule in the cytoplasmic membrane before the channel opens or before the open channel has allowed the efflux of a large amount of intracellular K'. The existence of an immunity bypass site on amphiphilic channel helix HZ, which has been shown to insert into the membrane upon imposition of a membrane potential (Merrill and Cramer, 1990), implies that immunity protein can react with the channel in the open state.

Mechanism of Specific Interaction of Immunity Protein with
Colicin E l Channel Domain Direct Interaction-The ability of the colicin E l immunity protein reaction to accommodate the tonB as well as the to1 translocation proteins implies that these macromolecule translocation systems do not have any specific role in the recognition mechanism of immunity protein with the channelforming colicin. As for colicins E2 and E3, immunity protein must directly interact with colicin El, although in the case of the latter, the interaction involves the channel domain in the membrane.
Based on (i) the importance of the extramembrane region between helices H3-H4 in colicin A immunity protein analogous to peripheral segment T2 in the colE1 immunity protein (Fig. 2 A ) for immunity protein function (Geli and Lazdunski, 1989), (ii) the high proportion (8 out of 18 residues) of charged amino acids in loop L2 of the ColEl imm protein (Fig. 2 A ) , and (iii) precedent for stabilization of an oligomeric integral membrane protein complexes by favorable electrostatic interactions (Szczepaniak et al., 1991), it was proposed that the specificity of immunity protein recognition could occur through polar or electrostatic interactions between the extramembrane loops of the immunity protein and the target channel domain . This hypothesis appears wrong.
Neutralization by mutagenesis of (i) the only charge in extramembrane segments T1 and T2, (ii) of the only charges in loop L1, or (iii) of two of the four consecutive charges a t positions 71-74 in loop L2 (Fig. 2 A ) , had no effect on immunity protein function ( Table 111). Substitution of a charged residue at position 39 in loop L2 or at positions 43 or 45 near the H2-L2 boundary was also without effect .
Although these data do not rule out the involvement of the polar extramembrane segments in immunity protein-colicin E l recognition, they suggest that the segments of these proteins that are intrinsic to the membrane, the trans-membrane a-helices, may carry information necessary for the recognition. A somewhat similar conclusion was also obtained in a fusion protein study of the immunity interaction with hybrid colicin A-colicin B (Geli and Lazdunski, 1992), in which the region of the channel domain that reacts with immunity protein was inferred to start a short distance on the NH2terminal side of the hydrophobic a-helices (i.e. corresponding to a start position some residues on the NH2-terminal side of L Y S~~' in the model shown in Fig. 2B) (Geli and Lazdunski, 1992).
Immunity Bypass Mutants-In order to localize the sites of interaction of the channel domain of colicin E l with immunity protein, a search was made for the existence of immunity bypass mutants in the collection of 29 site-directed non-polar + charged residue mutants that was used to map the membrane topography of the Ala471-Ile508 hydrophobic domain, and to infer that it is a hydrophobic helical hairpin . The existence of mutations in this region that can bypass immunity function strongly implies interaction within the bilayer. Most of these mutants possess low cytotoxicity . Mutants were studied that had a lowered but readily measurable activity in an imm-strain and were examined for increased activity in an imm+ strain. The 2 residues at which mutants with the largest bypass effect could be found were Ala"' " and Ser477 in hydrophobic helix H3 of the colicin E l channel domain (Fig. 2B). The Ala474 4 Glu, Ser477 * Arg, and Ser477 + Lys mutants had cytotoxic activities approximately 5-, 25-, and 3000-fold lower than wild-type when assayed on the imm-indicator (Table IV; cf., . These mutants were able to bypass the imm+ protective function a t concentrations 625, 5.0, and 1.0 times the minimum concentration needed to demonstrate cytotoxic activity on imm-cells, in contrast to the wild-type colicin that does not show cytotoxicity when added to imm+ cells at 200 pg/ml, a concentration 7 X lo4 greater than needed for activity with imm-cells (right-hand column, Table IV). Charge-substitution mutants at 27 other sites in the Ala471-Ile508 hydrophobic H3-H4 helices were found to be negative for the immunity bypass phenotype (Table V).
Because of the 3-residue spacing between the bypass mutants at positions 474 and 477, and the possibility of the bypass phenotype being associated with a-helical periodicity (see below), particular attention was paid to mutants offset by 3-4 residues from positions 474 and 477. This subset included mutants Ala471 + Lys, Val480 -.+ Asp, Ala"" + Asp and Ser485, + Arg in hydrophobic helix H3 (Table V). Fourteen of the 15-17 residues in helix H4, including 5 at the same approximate depth in the membrane as Ala474 and Ser477 in the model of Fig. 2B, have been tested and found to be bypass-negative (Table V). Thus, the potential for bypass of immunity function in the hydrophobic helical hairpin appears to be localized in residues 474 and 477 of helix H3. Substitution of a large and/or basic residue at positions 477 and 440 appears to be required for immunity bypass, as Ser477 + Ala or Gly and His""' -Cys or Glu mutants are unable to bypass immunity (Table V).
Three additional, although less pronounced, sites for bypass mutants were found in an additional set of site-directed substitutions, His440 -+ Arg, Phe443 -+ Lys, and G1y444 + Lys, that were constructed in the polar and somewhat amphiphilic helix H2. These mutants have a cytotoxic activity relative to wild-type that is, respectively, unchanged, 5-fold smaller, and 5-fold lower when assayed on the imm-indicator. They are able to bypass immunity function in imm+ cells at concentrations approximately 1.7 X lo", 3.1 x lo3, and 6.3 x lo3, respectively, of the minimum concentration needed for activity with inm-cells (reciprocals of latter numbers shown in right-hand column, Table IV). The existence of bypass mutants on helix H2 implies that the protective interaction with the immunity protein can occur after the amphiphilic helical hairpin has inserted into the membrane (Merrill and Cramer, 1990) and initiated the opening of the channel. The two sets of sites on helices H2 and H3 of the colicin channel domain are on opposite sides of the membrane according to the model of Fig. 2B. As can be seen from the aligned sequences of colicins El, la, and Ib (Table I), the residues His440, Phe443, G~Y""~, and Ala474 are at non-conserved sites. This suggests that the specificity of the colicin El-immunity protein interaction may be partly determined by these nonhomologous or non-conserved residues.
Protein-Protein Recognition via Trans-membrane Helices-From the above data and discussion, it was concluded that the mechanism of the specific inhibitory complex of the immunity protein with the colicin E l channel domain involves (i) lateral diffusion in two-dimensional space of the highly dilute immunity protein in the membrane, and (ii) recognition and formation of an interprotein helix complex by interaction at least one of the three helices of the immunity protein with two of the helices, H2 and H3, of the colicin E l channel domain (Fig. 5). The distribution of the bypass mutants a t positions 440, 443, and 444 on helix H2, and a t 474 and 477 on helix H3, suggests that one face of each helix is involved in helix-helix interaction with the immunity protein. Precedents for the existence of important helix-helix interactions in the assembly of polytopic membrane proteins such as bacteriorhodopsin and lactose permease, in the formation of a tightly associated dimer of glycophorin A, and in proposed mechanisms of receptor-mediated trans-membrane signaling, have been reviewed (Bormann and Engelman, 1992). The helix-helix interactions could be mediated by salt bridges, hydrogen bonds, or specific packing and interactions of nonpolar residues, as in the case of glycophorin A (Lemmon et al., 1992). Because the bypass mutants occur on both a hydro-