Toward Elucidating the Membrane Topology of Helix Two of the Colicin E1 Channel Domain*

The membrane-bound closed state of the colicin E1 channel domain was investigated by site-directed fluorescence labeling using a bimane fluorophore attached to each single cysteine residue within helix 2 of each mutant protein. The fluorescence properties of the bimane fluorophore were measured for the membrane-associated form of the closed channel and included fluorescence emission maximum, fluorescence anisotropy, apparent polarity, surface accessibility, and membrane bilayer penetration depth. The fluorescence data show that helix 2 is an amphipathic α-helix that is situated parallel to the membrane surface, but it is less deeply embedded within the bilayer interfacial region than is helix 1 in the closed channel. A least squares fit of the various data sets to a harmonic wave function indicated that the periodicity and angular frequency for helix 2 in the membrane-bound state are typical for an amphipathic α-helix (3.8 ± 0.1 residues per turn and 94 ± 4°, respectively) that is located at an interfacial region of a membrane bilayer. Dual quencher analysis also revealed that helix 2 is peripherally membrane associated, with one face of the helix dipping into the interfacial region of the lipid bilayer and the other face projecting outwardly into the aqueous solvent. Finally, our data show that helices 1 and 2 remain independent helices upon membrane association with a short connector link (Tyr363–Gly364) and that short amphipathic α-helices participate in the formation of a lipid-dependent, toroidal pore for this colicin.

Colicin E1 is a member of a large family of plasmidencoded antimicrobial protein toxins secreted by Escherichia coli. They are produced by the bacteria in response to a variety of assaults, including DNA damage, anaerobiosis, catabolite repression, and nutrient depletion (1), targeting susceptible E. coli and similar bacteria. Colicins can be grouped into the following three categories based on their mode of action: (i) membrane depolarization via formation of ion-conducting channels (2), (ii) inhibition of protein (3) or peptidoglycan synthesis (4); and (iii) DNA degradation (5). Because of their ability to cross the Gram-negative bac-terial membrane, colicins have become a model for studying bacterial protein import (6) as well as protein unfolding and folding (7,8), membrane insertion (9,10), and pore formation (11).
Colicin E1 belongs to the ion channel-forming group of colicins, which also includes colicins A, B, Ia, Ib, N, and K (12,13). The structural organization of these pore-forming colicins includes three domains as follows: receptor binding, translocation, and catalytic domains. The receptor binding domain initiates entry of the toxin into the target cells (14) by binding with the BtuB, or vitamin B 12 , receptor on the bacterial outer membrane (15,16). Next, the translocation domain associates with the trimeric ␤-barrel TolC outer membrane receptor, through which the unfolded translocation and catalytic domains of the colicin are transported into the periplasmic space (16). The catalytic domain subsequently refolds into an insertion-competent conformation, and the protein binds the inner membrane and, in a series of kinetically defined steps, the channel protein sequentially unfolds, binds, and spontaneously inserts into the membrane in a precursor state to the open channel (7,(17)(18)(19). The channel is opened upon imposition of a trans-negative membrane potential (20), and the newly created pore causes depolarization of the cytoplasmic membrane through the escape of cellular ions such as Na ϩ , K ϩ , and H ϩ . As the bacterial cell tries to equilibrate the ion concentration using the Na ϩ /K ϩ -ATPase, cellular ATP reserves are rapidly depleted and cannot be replenished sufficiently, and cell death ensues (21).
A crystal structure of the soluble channel domain of colicin E1 was determined to a resolution of 2.5 Å (21). The data revealed a globular protein composed of 10 amphipathic ␣-helices similar to the previously published structure of colicin A (22). Subsequently, the whole or catalytic domain structure of pore-forming colicins Ia (23), N (24), and B (25) have been solved, and predictably they are all structurally similar. These structures suggest that a key to their pore-forming ability are the two distinct hydrophobic helices that create a membranespanning hairpin upon bilayer association (17,26,27). Interestingly, this feature of the colicin structure is also seen in the mammalian intracellular apoptotic regulators Bcl-2 (28), Bcl-XL (29), Bax (30), and Bid (31), proteins known to create pores upon association with the mitochondrial or endoplasmic reticulum membranes.
A plethora of approaches have been employed over the years to derive information about the membrane-associated closed channel topology of the catalytic domain of colicin E1. Solid state NMR was used to identify the existence of trans-mem-*This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada (to A.R.M.) and by National Institutes of Health Grant GM 31986 (to E.L.). 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. 1 To whom correspondence should be addressed. Tel.: 519-824-4120 (ext. 53806); Fax: 519-837-1802; E-mail: rmerrill@uoguelph.ca.
brane and in-plane ␣-helices of membrane-bound colicin E1 (18), whereas circular dichroism, Fourier transform infrared spectroscopy, fluorescence resonance energy transfer, and differential scanning calorimetry were used to study membrane association and determine the ␣-helical and ␤-sheet content (32). Using site-directed mutagenesis to engineer conveniently located Trp residues, our group has estimated the depth of colicin E1 helical segments upon binding to the membrane using fluorescence quenching, fluorescence resonance energy transfer, and red-edge excitation shift analysis (33)(34)(35). These varied studies imply that the hydrophobic helices 8 and 9 insert into the bacterial inner membrane, whereas the remaining amphipathic helices lie on the membrane surface with considerable interaction with the aqueous medium. The exact orientation of the helices and the lipid and protein contacts made by each residue upon membrane binding are currently under investigation. Using electron paramagnetic resonance, the mobility and accessibility of nitroxide-labeled cysteine side chains generated one-at-a-time from residues 402-424 (which encompasses helix 4) showed a solvent-exposed and a lipidexposed face of the amphipathic helix, suggesting a possible transmembrane orientation of this helix upon bilayer binding (36). Recently we reported on the membrane topology of helix 1 of colicin E1 in the closed channel state and found that this helix retains its ␣-helical structure and helix length even when bound at the surface of large unilamellar vesicles (37).
Herein we continue our investigation of the membranebound topology of the helices within the closed state of the colicin E1 channel with a study of the membrane-bound disposition of helix 2 (Glu 365 -Ser 378 ). Using site-directed fluorescence labeling of a series of cysteine mutants coupled with rigorous analysis using various fluorescence techniques, we found that helix 2 is an amphipathic ␣-helix. However, it is less tightly appressed to the membrane surface than helix 1, and its membrane-bound length likely extends from Glu 365 -Ser 378 .

EXPERIMENTAL PROCEDURES
All chemicals, unless otherwise stated, were purchased from Sigma. All steady-state fluorescent measurements were collected using a PTI-Alphascan-2 spectrofluorimeter (Photon Technologies Inc, South Brunswick, NJ) equipped with a thermostated cell holder. All measurements are reported as the mean Ϯ S.D. and were performed at least in triplicate.
Mutagenesis, Protein Purification, and Monobromobimane Labeling-Cysteine-scanning mutagenesis, in which each amino acid residue from Glu 365 to Gly 380 of P190H 6 2 (except for K366C) was individually replaced with a cysteine, was performed using the Stratagene (La Jolla, CA) Quikchange TM mutagenesis kit. Plasmid DNA was purified using the High Pure Plasmid TM isolation kit from Roche Diagnostics. Wild type P190H 6 , P190H 6 /C505A (Cys-less wild type), and Cys mutant plasmids were prepared and purified from transformed lexA Ϫ E. coli IT3661 cells as described previously (37). The purity of each protein was assessed by SDS-PAGE. Protein concentration was determined by spectroscopy at A 280 , using the extinction coefficient (⑀) of 28,590 M Ϫ1 cm Ϫ1 (⑀ ϭ 27,310 M Ϫ1 cm Ϫ1 for the Tyr mutant protein, Y367C) (35). Purified mutant proteins were labeled with monobromobimane (mBBr) (Molecular Probes, Eugene, OR) at a 20:1 molar ratio (probe:protein), and labeling efficiency was determined as described previously (37).
Steady-state Intrinsic Trp Fluorescence-The folding properties of all proteins were examined using intrinsic Trp fluorescence as described previously (37). Briefly, WT P190H 6 , P190H 6 /C505A, and unlabeled and bimane-labeled mutant proteins were diluted to 4 M in PBS (50 mM NaH 2 PO 4 , 50 mM Na 2 HPO 4 , 100 mM NaCl, pH 7.0). Intrinsic Trp residues were excited at 295 nm (excitation slit width, 2 nm), and emission was detected from 305 to 450 nm (slit width, 4 nm). The resulting traces were corrected for the buffer and wavelength-dependent bias of the emission components of the spectrofluorimeter before calculation of the emission maximum ( em(max) ) from the first derivative of the smoothed spectra.
Bimane Fluorescence Emission Spectra-The steady-state bimane fluorescence emission spectra of all mutant proteins were measured as described previously (37). All measurements were made with a 4 M protein sample in DMG buffer (20 mM DMG, 130 mM NaCl, pH 4.0) in the presence or absence of excess LUVs (800 M, final concentration). The data were corrected for the buffer and wavelength-dependent bias of the equipment (34) before calculation of the em(max) from the first derivative of the smoothed spectra.
Solvent-accessible Surface Area (SASA)-To correlate the bimane fluorescence parameters with the local environment of each probed site, the solvent-accessible surface area of each helix 2 amino acid side chain was determined using a web-based program called GETAREA 1.1 (40) with a 1.4-Å water probe after input of the crystal structure coordinates of P190H 6 (21).
Steady-state Bimane Fluorescence Anisotropy-The steadystate fluorescence anisotropy (r) measurements were made using "T-format" detection by simultaneously comparing the intensities of the vertically (I VV ) and horizontally (I VH ) polarized emitted light when the sample was excited with a vertically 2 The abbreviations used are: P190H 6  The "G" instrumental factor, measured as I HV /I HH , was determined from the intensities of the vertically (I HV ) and horizontally (I HH ) polarized emitted light from horizontally polarized excitation light. For all measurements, the excitation was set at 381 nm (4 nm slit-width), and emission was collected at 470 nm (10 nm slit-width) with a signal integration time of 30 s. Each anisotropy value is the mean of three determinations. A solvent blank (buffer or LUVs in buffer) was subtracted from each intensity reading prior to the calculation of the anisotropy value as described previously (37).
Dual Quenching Analysis-Depth-dependent quenching of membrane-bound bimane-labeled mutant proteins was performed as described previously (41). To measure iodide quenching (F KI ), the fluorescence of samples was measured on a Spex Tau-2 Fluorolog spectrofluorimeter (Jobin Yvon Inc., Edison, NJ) in ratio mode using semi-micro quartz cuvettes (excitation path length 10 mm, emission path length 4 mm) containing 100 M LUVs and 7.5 g of protein or LUVs only (background). F KI was determined 5 min after the addition of a 50-l aliquot of an aqueous solution from a 1.7 M KI and 0.85 mM Na 2 S 2 O 3 stock solution. The fluorescence values after KI addition were corrected for dilution before quenching was calculated. The excitation wavelength was set at 375 nm while observing the emission intensity at 467 nm. The excitation and emission slit-widths were 2.5 and 5.0 nm, respectively. To measure the efficiency of 10-doxylnonadecane (10-DN) quenching, membrane-bound protein or vesicles lacking protein were prepared as described above except that the 10-DN quencher-containing LUVs contained 10 mol % of 10-DN. After preparation, all the samples were allowed to equilibrate for 30 min at 24°C before measurement of initial fluorescence.
Calculation of the Iodide to 10-DN Quenching Ratio (Q-Ratio)-The ratio of quenching by 10-DN to that by KI (Q-ratio) was used to determine bimane depth in lipid bilayers. The Q-ratio was calculated from Equation 2, where F o is the fluorescence of a sample lacking quencher, and F KI and F 10-DN are the fluorescence intensities in the presence of KI or 10-DN, respectively. Sensitivity of the Fluorescence Parameters of Bimane to Solvent Polarity-The sensitivity of the bimane fluorescence to solvent polarity was assessed using N-acetylcysteine conjugated with bimane (bimane-Cys) as a probe. Bimane-Cys was produced by reacting mBBr with 10-fold molar excess of N-acetylcysteine in 100 mM NH 4 HCO 3 buffer, pH 8.1, for 1 h. The reaction mixture was lyophilized overnight and subsequently resuspended in dioxane:water mixtures of 0 -100% (v/v) dioxane. The fluorescence emission and lifetime of bimane-Cys (2 M bimane-Cys) samples in dioxane:water mixtures of different dielectric constants (⑀ ϭ 2.3-80) were determined.
Predicting Secondary Structure from Fluorescence Parameters-The secondary structure elements were predicted from the observed fluorescence parameters using a method adopted from Cornette et al. (42). In brief, the periodicity and the angular frequency of the observed fluorescence parameters were obtained through a least squares fitting approach using the harmonic wave function shown in Equation 3, where a is the amplitude; b is the phase; p is the period, and c is an offset value. All calculations were made using an Excel 2002 template spreadsheet developed for us by Dr. Uwe Oehler.

RESULTS
Structural and Functional Analysis-The primary amino acid sequence and ribbon diagram of the crystal structure (21) of the channel peptide of colicin E1 (P190) can be seen in Fig. 1. The designation "P190H 6 " refers to the N-terminal His 6tagged, 190-amino acid peptide of colicin E1 that encodes the active channel domain. In Fig. 1A, the sequence of the protein corresponding to helix 2 that was subjected to Cys-scanning mutagenesis is indicated by the downward facing brace. The soluble structure of the domain (Fig. 1B) is arranged in three layers as follows: (a) helices 1, 2, and 10; (b) helices 5, 8, and 9; and (c) helices 3, 4, 6, and 7. The locations of the Cys-substituted residues are also highlighted in Fig. 1B by dark spheres that correspond to the C-␣ carbon of each residue.
The labeling efficiencies of most of the Cys mutants ranged from 60 to 110% indicating the newly incorporated cysteine residues were labeled as expected. However, problems were encountered with A371C (9% labeling efficiency) and K377C (200% labeling efficiency).
The WT P190H 6 protein contains a single Cys residue at position 505 within helix 9 of the structure. This cysteine residue is extremely buried and does not normally react with electrophilic reagents used to covalently tether a fluorophore to an engineered Cys residue at the surface of a mutant protein. However, Cys 505 labeling would be a problem for the A371C and K377C mutants. In the case of A371C, even a small amount of Cys 505 labeling would contribute to the bimane signal, and in the case of K377C, the extra labeling suggests Cys 505 probably gets labeled because of some conformational perturbation.
The most likely explanation for the behavior of the K377C mutant is that the mutation of the long lysine residue to a more compact cysteine possibly created a channel within the tertiary structure of the colicin E1 peptide that allowed the thiol-reactive probe (mBBr) access to the normally inert and buried Cys 505 . Supporting this theory, the Trp em(max) of K337C was 328 nm (data not shown) indicating potential structural perturbation with this mutant.
The low labeling efficiency of the A371C mutant could be ascribed to the fact that the relatively hydrophobic Ala 371 residue is located in the middle of helix 2 facing toward the core hydrophobic hairpin helices of the channel peptide and is therefore inaccessible to the labeling reagent. However, bimane labeling experiments of A371C/C505A even in the presence of 8 M urea did not improve the labeling efficiency (data not shown).
Whatever, the origin of these problems, to circumvent them, a Cys-less P190H 6 (P190H 6 /C505A) was used as a background template. As expected, the use of the K377C/C505A double mutant returned the bimane labeling efficiency to normal levels (83%) for this mutant.
To determine the folding and functional properties of all mutants, the intrinsic Trp fluorescence and in vitro channel activities of each protein, with or without the bimane label, were determined. The intrinsic Trp fluorescence measurement provides an assessment of the folded integrity of the protein under investigation; a Trp em(max) value near 324 nm for colicin E1 indicates properly folded protein, whereas more redshifted values indicate perturbation of the folded structure (at a Trp em(max) value near 350 nm, the protein is considered fully denatured (43)). As shown in Table 1, the structural and functional integrities of most mutant proteins were comparable with the WT (P190H 6 ) or Cys-less template (C505A) as per their Trp em(max) values and their relative in vitro channel activities. One mutant, M370C, had a slightly red-shifted Trp em(max) (329 nm), indicating a degree of perturbation; however, upon bimane labeling, it seems to have reacquired the native folded structure (Trp em(max) , 325 nm). Two other mutants, Y367C and L374C, displayed a more significant Trp em(max) red shift to 333 and 332 nm, respectively, suggesting that these mutant proteins were somewhat destabilized by Cys substitution. Not surprisingly, these two mutants did not express high quantities of protein, and therefore protein purification was less successful (data not shown). In the case of Y367C, analysis of the crystal structure of the channel peptide revealed that the mutation likely caused the loss of a critical hydrogen bond with Thr 501 of helix 9. This most likely "loosened" the tertiary structure at this point causing local unfolding. The underlying cause of the red-shifted Trp em(max) seen for L374C was not immediately obvious; however, a general loosening of the structure induced by Cys substitution is likely the explanation. However, both of these mutant channel proteins were active and exhibited good in vitro channel activities (Table 1). Fig. 2A shows the bimane em(max) values for the Cys-labeled sites in helix 2 of the soluble channel domain. In the soluble form, most of the mutants showed bimane em(max) values Ն470 nm. These mutant proteins correspond to mostly polar and charged residue substitutions, E365C, Y367C, S368C, K369C, Q372C, E373C, D376C, K377C/C505A, S378C, K379C, and G380C. The corresponding SASA values of the substituted sites also showed nearly consistent accessible surface area values (SASA Ͼ50 Å 2 ; Fig. 3, solid line), a value that is generally accepted for surface-exposed sites within a soluble protein (44). In contrast to the surface-exposed mutations, the remaining mutant proteins, M370C, A371C/ A, the primary sequence and the secondary structure of the channel-forming domain of colicin E1 (P190H 6 ). Highlighted as a downward facing brace are the residues corresponding to the Cys substitutions. The tryptic and thermolytic digestion sites of the channel-forming domain are shown as Tr and Th, respectively. B, the ribbon topology diagram of the 2.5 Å crystal structure of the P190 peptide (21). The overall architecture of the structure consists of 10 helices, a hydrophobic helical hairpin (8 -9, black) surrounded by eight amphipathic helices that are arranged in a three-layer sandwich helical bundle as follows: layer A, helices 1, 2, and 10; layer B, helices 8, 9, and 5; layer C, helices 3, 4, 6, and 7 ) (21). The location of the cysteine mutation sites are highlighted with numbers and black spheres that represent the ␣-carbon of each substituted residue. The dark-colored ␣-helices are helices 8 and 9, which serve as a membrane-anchoring helical hairpin in the membrane-associated closed channel state of the protein. C505A, L374C, and A375C, showed relatively blue-shifted bimane em(max) values (455-469 nm) ( Fig. 2A), and the SASA values of the Cys sites of these mutants showed correspondingly lower surface-accessible areas (0 -5.23 Å 2 ) indicative of buried sites within the soluble channel domain. An exception to this correlation of SASA with the em(max) was observed for Y367C, S368C, S378C, and G380C mutants. According to the predicted SASA values, the Cys-labeled sites of these mutants have rather low accessible surface areas (0 -2.99 Å 2 ) and hence would be expected to be buried sites within the protein. However, the observed em(max) values of these mutants (470 -479 nm) are consistent with more surface-exposed sites. This discrepancy can be explained if one takes a closer look at the location of the substitution sites in the crystal structure of the colicin E1 channel domain (Fig. 1B). Tyr 367 and Ser 368 are situated at an extremely tight junction because of the small, tight turn between helix 1 and helix 2. It is possible that a mutation and subsequent labeling at Tyr 367 and Ser 368 may elicit localized structural perturbation, rendering the tethered bimane fluoro-phore more exposed than the native residues (Tyr or Ser). In fact, the Trp em(max) of Y367C (333 nm) indicates some structural perturbation of this mutant in comparison with the WT (Trp em(max) ϭ 324 nm) as a consequence of the mutation and labeling procedures, and this mutant has the lowest channel activity of all those studied in helix 2 (60% of WT activity, see Table 1). For G380C, closer inspection of the crystal structure indeed confirms that the Gly 380 residue is surface-exposed, and the predicted low accessible surface area is attributed to the limitation of the algorithm used in determining the SASA of small side chain moieties such as Gly (H atom) (40). Despite these minor limitations, the observed bimane em(max) for most mutant proteins in aqueous solution appear to correlate well with the solvent accessibility of the probed sites for the solution structure of the colicin E1 channel domain. Most of residues with high SASA (Ͼ50 Å 2 ) showed consistently higher em(max) values (Ͼ470 nm) and vice versa. Such observations are con-   2I88). B, the solvent accessibility of each residue was calculated as the ratio of the side chain surface area as estimated from the crystal structure and the expected theoretical value of each residue in the random coil tripeptide, Gly-X-Gly. Residues with solvent accessibility ratios above 50% (top horizontal line) are considered to be surface-exposed, whereas residues with ratios Յ20% (bottom horizontal line) are buried. The calculated SASA data were fitted through nonlinear least squares analysis to a harmonic wave function (r 2 ϭ 0.91 and 0.92, respectively). The data points are joined by a dashed line, and the harmonic function fit to the data is shown as a solid line (periodic function). OCTOBER 27, 2006 • VOLUME 281 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 32379
sistent with our earlier characterization of helix 1 of the channel peptide (37). Fig. 2B shows the bimane em(max) of membrane-bound Cys mutant proteins. In comparison to Fig.  2A, the overall shape and profile of the em(max) values of this helix do not differ from that of the soluble state, indicating that the amphipathic ␣-helical nature of this segment of the channel protein was not significantly affected upon binding to the membrane surface. Overall, the observed em(max) values of the membrane-bound protein showed slightly blueshifted em(max) values in comparison to the solution state of the corresponding mutants. As shown previously for helix 1, this shift is consistent with overall changes in the polarity of the local environment of the bound peptide upon membrane association (37). On the basis of the general em(max) distributions, the membrane-bound mutants within helix 2 can be grouped into three categories as follows: 1) mutant proteins with redshifted em(max) (Ն470 nm), E365C, S368C, K369C, Q372C, E373C, D376C, K379C, and G380C; 2) mutant proteins with intermediate em(max) values (468 -469 nm), Y367C, L374C, and K377C/C505A; and 3) mutant proteins with blue-shifted em(max) (Ͻ468 nm), M370C, A371C/C505A, A375C, and S378C. Group 1 mutants, corresponding to largely polar and charged residues substituted with Cys, form the surface-exposed face of helix 1 in the soluble structure (Fig. 1B). Group 3 mutants, consisting of relatively nonpolar or polar uncharged residues replaced with Cys, make up the more buried face of this helix. Some of the residues in the intermediate group, such as Leu 374 , will likely be part of the hydrophobic face of the helix, whereas Tyr 367 and Lys 377 would be expected to interact with the aqueous solvent given the length of their side chain moieties, but a Cys substitution of the latter residues may lead to an interfacial location for the bimane chromophore.

Membrane-bound Bimane Fluorescence Emission and the Topology of the Cys Sites-
To further characterize the nature of the local environment surrounding the Cys-scanned sites within helix 2 of the membrane-bound channel domain, the apparent polarity corresponding to observed em(max) values were calculated using the equation em(max) ϭ 0.33 (nm/⑀) ϫ D ϩ 452.8 nm, where D is the dielectric constant of the bulk solvent (⑀). This relationship was determined in our previous study (37) with a bimane-Nacetyl-Cys standard and a dioxane-water solvent system; this polarity-calibration approach was found to be a reliable indicator of the apparent polarity of the local environment for the bimane chromophore. The calculated apparent polarities for group 1 mutants ranged between ⑀ ϭ 69 and 53; these values are consistent with highly aqueous, solvent-accessible environments. As for groups 2 and 3, the calculated apparent polarities ranged from ⑀ ϭ 48 -50 and ⑀ ϭ 13-41, respectively. Although the latter two groups reside in a relatively more nonpolar environment than the group 1 mutants, the calculated apparent polarities of these two groups are nonetheless outside the range generally accepted for even the interfacial layer of the membrane (⑀ ϭ 10 -25). Only two mutants, A371C/C505A and S378C, showed apparent polarity values (⑀ ϭ 13 and 22, respectively) that are within the range generally ascribed to the interfacial layer of the membrane.
Bimane Fluorescence Anisotropy and the Mobility of the Helix 2 Residues-The mobility of the substituted sites within helix 2 were determined from the fluorescence anisotropy of the tethered bimane of the mutant proteins in the presence and absence of LUVs. Probe mobility is quantified as the inverse of the observed fluorescence anisotropy (Fig. 4). As observed for the bimane em(max) parameters, the observed fluorescence anisotropy parameters are influenced by the local environment of the probed sites. Surface-exposed sites are expected to have lower anisotropy given the higher rotational freedom of their side chains (and the attached fluorophore in the Cys mutants). In contrast, sites facing the interior of the protein or that are buried in the membrane would be expected to show lower probe mobility (higher anisotropy) given the extensive tertiary steric hindrances of the attached probe. In addition to tertiary structure-induced restricted mobility, the mobility of the bimane probe can also be affected by the local secondary structure of the labeled site. Sites with defined secondary structures (␣-helix and ␤-strand) will also have lower rotational freedom than sites within unstructured regions (random coil). The observed probe mobility for most of the mutant proteins (Fig. 4A) in solution correlated well with the available surface area for the side chains of the probed sites, as per the calculated SASA from the solution crystal structure (Fig. 3) as well as the em(max) (Fig.  2A). Mutant proteins corresponding to sites predicted to have large accessible surface areas and shown to possess high em(max) values also showed higher probe mobility relative to the buried sites. E365C, K369C, E373C, D376C, K377C/C505A, K379C, and G380C, all of which showed em(max) values greater than 470 nm in the soluble state and that correspond to sites with greater than 50 Å 2 accessible surface area, also exhibited higher probe mobility (7.9 -10.5).
In contrast, L374C, A375C, and S378C, which correspond to sites with less than 10 Å 2 accessible surface areas and are thus buried in the core of the protein, showed lower probe mobility (5.8 -6.4). It should also be noted that the observed probe mobility in the soluble state of mutant proteins of the N-terminal region of helix 2 do not agree well with the estimated accessible surface area of the substituted side chains (Fig. 3) and the observed em(max) of these proteins in solution. For instance, M370C and A371C/C505A show unusually high probe mobility (10.5 and 7.4, respectively) given the predicted accessible area of their substituted side chains (5.2 and 0.5 Å 2 , respectively) and the corresponding em(max) values (469 and 455 nm), both of which suggest the labeled sites to be facing the interior of the proteins. Similar discrepancies are also observed for Y367C, S368C, and Q372C, which also show lower than expected probe mobility (Fig. 4A). It is likely that these discrepancies may reflect structural alterations to the tight packing of this region of the protein (Fig. 1B) as a consequence of the tethered bimane and the introduced mutations. Interestingly, despite some localized discrepancies for the observed fluorescence parameters with the crystal structure data, the overall profiles of both em(max) and the probe mobility data show that the ␣-helical structure of helix 2 of the soluble channel domain remains intact after Cys mutation and also after bimane attachment.
The observed probe mobility for membrane-bound helix 2 mutant proteins (Fig. 4B) correlated well with their em(max) values (Fig. 2B). Labeled sites predicted to form the hydrophobic side of the helix that faces the membrane surface in the membrane-bound state, as per em(max) and estimated apparent polarities, showed equally low probe mobility values (4.7-5.4). Similarly, labeled sites shown to form the surface-exposed face of helix 2 also displayed relatively higher probe mobility values (5.9 -7.9). Comparison of the overall profiles of the probe mobility values of the membrane-bound state and the soluble state of the labeled sites, with the exception of the N-terminal region for the above described reasons, reveal a similar pattern and periodicity, consistent with an amphipathic ␣-helical secondary structure conformation for helix 2 in both the soluble and membrane-bound states.
Dual Quenching Analysis of the Membrane-bound Depth of P190H 6 Helix 2-To assess the relative bilayer depth of helix 2 residues, a dual fluorescence quenching method that measures the depth of fluorescent groups in membranes was used (37,41,45,46). This method exploits the quenching ((F o /F q ) Ϫ 1) of the fluorescent groups at each site by two quencher species (KI and 10-DN) that possess differential solubility with respect to the aqueous solvent and the membrane bilayer (F o and F q represent the fluorescent intensity of the bimane fluorophore in the absence and presence of quencher, respectively). In these experiments, membrane-inserted fluorescent groups are strongly quenched by the membrane-embedded quencher (10-DN), giving a high value of ((F o /F 0-DN ) Ϫ 1). In contrast, surface-exposed fluorescent groups, including bimane, are strongly quenched by the aqueous quencher (KI), giving a high value of ((F o /F KI ) Ϫ 1). Given that the accessibility to any single quencher could be affected by the local protein conformation, the quenching ratio (Q-ratio; ((F o /F 10-DN )) Ϫ 1/(F o /F KI ) Ϫ 1) of the nonpolar and polar quencher pair best describes the relative depth of the probed site. A low Q-ratio indicates that the bimane probe is quenched more by KI and less by 10-DN and would be expected to be localized outside the membrane environment.
Similarly, a high Q-ratio indicates significant membrane depth penetration of the probe into the bilayer and therefore shows less quenching by the water-soluble KI and more by 10-DN (41,45,46). Fig. 5 shows the dual quenching analysis data for both helix 1 and helix 2 residues (37). The Q-ratio data for helix 2 residues (Fig. 5, right-hand side) are in good agreement with the data obtained by bimane emission and probe mobility data, showing that this helix is lying close to the membrane-water interface with considerable exposure to the aqueous KI quencher and limited contact with the hydrophobic 10-DN quencher. The tethered bimanes of all of the mutant proteins of helix 2 are quenched to a much greater degree by the water-soluble quencher, KI (Fig. 5A, gray bars), than the membrane-embedded quencher, 10-DN, under the conditions chosen for the experiments (Fig. 5A, black bars). The only proteins quenched by 10-DN at an appreciable level are M370C, A371C/ C505A, L374C A375C, and S378C, all of which represent sites within helix 2 that would be expected to face the membrane surface based upon the em(max) and probe mobility data ( Fig.  2B and Fig. 4B).
Predicting Secondary Structure from Fluorescence Parameters-To determine the secondary structure of helix 2 of the membrane-bound channel domain, we used the method of Cornette et al. (42) to analyze the periodic variation of the observed fluorescence parameters with the residue number within this helix as described earlier for helix 1 (37). Shown in Table 2 is the summary of the nonlinear least squares harmonic wave function analysis of the observed fluorescence parameters of helix 2 for the both the soluble and membrane-bound states of the channel domain. For the soluble state, the fitting of the observed fluorescence parameters and the SASA data obtained from the crystal structure confirmed the amphipathic ␣-helical structure of this segment of the protein. The calculated residue per turn periodicity ( p) of both observed parameters ( em(max) and probe mobility) and those obtained from the crystal structure show a consistent value of 3.7 Ϯ 0.1, which is in agreement with the periodicity of an amphipathic ␣-helix (40,44). The phase of the fitted sine wave function for the em(max) and SASA data show comparable values, indicating that the orientation of the two faces of helix 2 did not change significantly upon Cys replacement and bimane labeling. However, the reduced periodicity and phase values of the probe mobility data in comparison to the SASA data may be an indication of slight alteration in tertiary contacts without concomitant secondary structure changes, especially at the N-terminal region of helix 2, induced by the Cys scanning method. The analysis of the amphipathic ␣-helical property of this segment of protein in the membraneassociated state showed that helix 2 maintains its amphipathic character with an average p value of 3.8 Ϯ 0.1 residues per turn. The phase property of this periodicity (228 -233 o ) and residues mapped to form the helix (365-379) are in agreement with those obtained from the analysis of the SASA data for helix 2 in the soluble state.

DISCUSSION
Previously, we showed that helix 1 within the colicin E1 channel domain maintains its amphipathic ␣-helical structure upon binding to the membrane surface in the closed channel state (37) and that the soluble helix consists of residues 350 -362, whereas the membrane-bound helix consists of residues 347-362 with the Tyr 363 -Gly 364 sequence serving as a helix breaker between helices 1 and 2. The data pertaining to helix 2 herein demonstrate that this helix is also relatively insensitive to perturbation (Table 1). Furthermore, channel activity is also insensitive to inactivation by Cys replacement mutagenesis and subsequent bimane labeling (37) ( Table 1). The greatest reduction in the in vitro channel activity for the helix 2 mutant proteins was only 40% (G380C mutant, Table 1), which supports previous mutagenesis studies that demonstrated the robust nature of the colicin E1 channel domain and its insensitivity to single point mutations (47)(48)(49)(50)(51).
The site-directed fluorescence labeling data presented herein provide compelling evidence for the amphipathic nature of helix 2 within the channel domain when it is bound to the membrane bilayer in the closed channel state (Fig. 6). It is clear from the harmonic wave function analysis of both helices 1 and 2 that the periodicities and angular frequencies of these helices do not change significantly upon binding to the membrane surface. In a similar fashion observed for helix 1, helix 2 also is appressed to the membrane surface with its nonpolar face bathing the hydrocarbon portion of the bilayer. However, the em(max) and apparent polarity data indicate that helix 2 is less tightly appressed to the bilayer surface than helix 1 and is overall more exposed to the aqueous solvent. Importantly, the connection between helix 1 and 2, Tyr 363 -Gly 364 in the membranebound channel domain, is not helical and therefore points to the individual nature of helices 1 and 2 in the closed channel state, i.e. that helices 1 and 2 do not join to form a longer ␣-helix lying on the membrane surface (Fig. 6) and that likely this linker   (33); furthermore, a plethora of biochemical and biophysical studies have led to various models for colicin precursor and active channels, including the carpet model for the closed channel where both helices 1 and 2 lie on the surface of the membrane with their nonpolar and polar faces interacting with the bilayer and the aqueous medium, respectively (34,35,52,53). The dual quencher analysis data (Fig. 5) allow a scan of the bilayer depth of the bimane fluorophore and provides a comparison of the relative depths of both helices 1 and 2 for the closed channel state. The Q-ratio analysis also indicated that helix 2 is less deeply embedded within the membrane bilayer than helix 1 (Figs. 5 and 6). Importantly, the results from the Q-ratio analysis were in excellent agreement with the em(max) , apparent polarity, and relatively mobility (fluorescence anisotropy) data for helix 2 ( Fig. 2B and Fig. 4B; Table  2), providing convincing evidence that both helices 1 and 2 are bound to the surface of the membrane in a conformation parallel to the bilayer surface. Fig. 6 is a simple model of both helices in the closed channel state of colicin E1 that show the disposition of these amphipathic helices and the deeper insertion within the membrane bilayer of helix 1 compared with helix 2. The shallow nature of helix 2 corroborates previous proteolysis data of the membrane-bound colicin E1 channel domain that indicated the relatively accessibility of helix 2 to protease digestion (54).
The structure of the open channel formed by colicins upon voltage imposition has been controversial. It is generally believed that the voltage-gated, open channel consists of a single polypeptide with an even number of transmembrane helices having the N and C termini lying on the cis side of the membrane and involves the movement of surprisingly large segments of the membrane-bound precursor across the membrane. Less controversial is the identity of the anchor domain of the channel, which is generally believed to consist of helices 8 and 9 of the soluble protein. However, the identity of the remaining transmembrane segments of the open channel is less clear, various models have been proposed, most of which involve the formation of at least two additional transmembrane helices (13, 34, 54 -59). A number of these models include an extended helix formed by the expansion of helices 1 and 2 of the soluble structure to provide one of the pillars of the open channel structure. Although helix rearrangements are an important aspect of the mechanism of colicin E1 membrane association, insertion, and pre-channel state formation, our current data preclude the melding of helices 1 and 2 into an extended helix as a feature of the closed channel state for colicin E1 (37) (Fig. 6).
Our results contained herein for helix 2 and our previous data for helix 1 (37) favor a toroidal model for colicin E1 as proposed earlier by Cramer and co-workers (13,58) that involves the formation of the open channel with relatively short amphipathic helices (15-16 residues) that employ phospholipids in an inverted micellar conformation in order to form a functional channel. Furthermore, it has been demonstrated recently that a much larger conductance channel (600 pS compared with 60 pS) may be formed in thicker bilayers and that colicin E1 may induce lipid "flip-flop" in such membranes (59). The toroidal lipid pore model has been used to describe the action of large toxins such as equinatoxin II (60) and sticholysin (61), antibiotic peptides (62,63), and apoptotic pore-forming proteins Bax (64) and tBid (65). An important requirement for the toroidal model is the sensitivity of the pore-forming activity to membrane lipid curvature, where lipids with positive and negative spontaneous curvatures stimulate and inhibit pore formation, respectively. Cramer and co-workers (58) clearly demonstrated the sensitivity of colicin E1 pore-forming activity on membrane curvature lending support to the toroidal model for this bacteriocin and also that the nature of the channel is influenced by the thickness of the membrane bilayer (59).
The harmonic wave function analysis of our fluorescence data for helices 1 and 2 shows a break in the periodicity of all profiles beyond Lys 362 , both in the soluble and membranebound states of the colicin E1 channel domain. In the x-ray structure of the colicin E1 P190, Lys 362 marks the C terminus of helix 1 and the start of a very tight turn (Tyr 363 to Gly 364 ) that separates helix 1 from helix 2 (37) (Fig. 1, A and B). Moreover, Gly 364 is conserved in all helical colicins and is located in this short turn region that separates helix 1 and 2 in all known structures of channel-forming colicins. Additionally, our wave function analysis suggests that the C terminus of helix 2 is marked by Ser 378 and is terminated by Lys 379 -Gly 380 . Accordingly, our data suggest that the helical boundaries of helices 1 and 2 remain unaffected upon binding to the membrane surface. Notably, our data refute earlier models that suggested the extension of parts of helix 1 and 2 into a single long amphipathic transmembrane helix as part of the open channel structure (53). These conclusions support the findings of Slatin et al. (57), who showed that segments of helix 1 from both colicin Ia and A are FIGURE 6. Schematic representation of membrane-bound topology of helices 1 and 2. Shown in the top panel is the ribbon diagram with residue side chains of helices 1 and 2 oriented with their nonpolar faces toward the hydrocarbon milieu of the bilayer and their polar faces toward the aqueous medium. The bottom panel shows the relative membrane-bound topology of both helices. Helix 1 lies on the surface of the membrane with its nonpolar face bathing the hydrocarbon portion of the bilayer, whereas helix 2 is not as tightly appressed to the membrane surface but is more solvent-exposed with its nonpolar face pressed against the head group region of the phospholipid bilayer with a few residues extending into the interfacial layer. Also shown in the bottom panel are the accepted dielectric constants (⑀) of the interfacial layer (IF; ⑀ ϭ 10 -24) and the hydrophobic core of the acyl chain (HC; ⑀ ϭ 2-4). OCTOBER 27, 2006 • VOLUME 281 • NUMBER 43 inserted into the membrane with the intervening loop forming a pivotal point between the transmembrane and the translocated segments of the open channel. In colicin A, this short loop is partially exposed to the trans side of the membrane, and the majority of helix 2 is translocated across the membrane. At present it is not clear as to how this translocation event contributes to the channel-gating mechanism. Obviously, there still exists considerable controversy as to the nature and structure of both the closed and open states of the pore-forming colicins. We will continue our pursuit of the membrane-bound topology of the closed state of colicin E1 through a residue-by-residue analysis using fluorescence-based methodology. It is anticipated that this approach, although tedious, will help to clarify some of the ambiguities and the enigma that currently define both the closed and open states of the pore-forming colicins.