Harmonic Analysis of the Fluorescence Response of Bimane Adducts of Colicin E1 at Helices 6, 7, and 10

Background: Colicin E1 forms a voltage-dependent channel in the cytoplasmic membrane of target E. coli cells.

Results: Fluorescence analyses reveal that helices 6, 7, and 10 in the membrane-bound colicin E1 are separate amphipathic α-helices with different periodicities.

Conclusion: A new model of the closed channel of colicin E1 has been developed.

Significance: This study provides new data on the structure of the umbrella model for the channel-forming colicins.

Abstract

The pre-channel state of helices 6, 7, and 10 (Val⁴⁴⁷–Gly⁴⁷⁵ and Ile⁵⁰⁸–Ile⁵²²) of colicin E1 was investigated by a site-directed fluorescence labeling technique. A total of 44 cysteine variants were purified and covalently labeled with monobromobimane fluorescent probe. A variety of fluorescence properties of the bimane fluorophore were measured for both the soluble and membrane-bound states of the channel peptide, including the fluorescence emission maximum, fluorescence anisotropy, and membrane bilayer penetration depth. Using site-directed fluorescence labeling combined with our novel helical periodicity analysis method, the data revealed that helices 6, 7, and 10 are separate amphipathic α-helices with a calculated periodicity of T = 3.34 ± 0.08 for helix 6, T = 3.56 ± 0.03 for helix 7, and T = 2.99 ± 0.12 for helix 10 in the soluble state. In the membrane-bound state, the helical periodicity was determined to be T = 3.00 ± 0.15 for helix 6, T = 3.68 ± 0.03 for helix 7, and T = 3.47 ± 0.04 for helix 10. Dual fluorescence quencher analysis showed that both helices 6 and 7 adopt a tilted topology that correlates well with the analysis based on the fluorescence anisotropy profile. These data provide further support for the umbrella model of the colicin E1 channel domain.

Introduction

Colicins are toxic proteins produced by certain strains of Escherichia coli to provide a survival advantage in a “selfish gene” system (1, 2), and they are often used in response to metabolic challenges, including DNA damage, catabolite repression, and nutrient depletion (3). Colicins are a large bacteriocin family that targets susceptible E. coli and similar bacteria, which do not possess the protective immunity protein (4), by acting at a number of levels, including (i) membrane depolarization by ion-conducting channels (5), (ii) inhibition of protein (6) or peptidoglycan synthesis (7), and (iii) DNA degradation (8).

Colicins have become a model for study of bacterial protein import (9, 10), protein folding (11, 12), membrane insertion (13, 14), and pore formation (15, 16). The colicin polypeptide can be functionally divided as follows: receptor binding, translocation, and catalytic/channel domains (17). Colicin E1 has a catalytic/channel domain that forms a depolarizing ion channel causing cell death in a host-infected bacterial cell (18). In order for colicin E1 to enter a target bacterium, the receptor-binding domain must first bind to the BtuB outer membrane receptor (vitamin B₁₂ receptor) (19). The binding of the BtuB receptor induces unfolding of the translocation domain, which initiates migration of the entire protein through the TolC channel and facilitates entry into the periplasm. This translocation process is also mediated by both the TolA and TolQ inner membrane proteins. Finally, the channel domain adopts an insertion-competent state in which it spontaneously inserts into the inner membrane to form the closed channel (20). The channel then opens in the presence of a trans-negative membrane potential that allows the escape of various ions from the host cells, such as Na⁺, K⁺, and H⁺, and subsequently cell death ensues (21).

The crystal structure of the soluble channel domain (22, 23) is composed of 10 separate α-helices that form an extremely stable, water-soluble globular protein. The channel domain is a helical sandwich that is folded into three layers: layer A, the outer layer composed of H1,⁵ H2, and H10; layer B, the inner core layer, including H5, H8, and H9; and layer C, an outer layer composed of H3, H4, H6, and H7 (17). Interestingly, this protein also consists of a hydrophobic α-helical hairpin, H8 and H9, which acts as the nonpolar core of the protein. These two helices are critical to colicin pore formation because they create a membrane-spanning hairpin upon bilayer association (24). Previous fluorescence studies suggested that upon translocation across the host cell outer membrane, colicin adopts an insertion-competent state, which allows the hydrophobic core (H8 and H9) to penetrate the target membrane. Thus, the protein unfolds, binds, and spontaneously inserts into the membrane to form the closed channel in a series of kinetically defined steps (25). Subsequently, the channel opens in the presence of a trans-negative membrane potential, and the two channel states exist in rapid equilibrium (26).

Two well known structural models have been proposed for the closed channel state, which are the penknife and umbrella model. The penknife model was based on disulfide bond engineering experiments, which suggested that H1 and H2 move away from the body of the protein, with the remaining helices being deeply buried into the lipid bilayer (2). In contrast, the umbrella model suggests that only hydrophobic helices, H8 and H9, are inserted into the hydrophobic milieu of the membrane, whereas the remaining eight helices spread out onto the membrane surface to form an umbrella-like structure. In fact, the umbrella model was strongly supported by time-resolved fluorescence resonance energy transfer (FRET) studies on colicin E1 (27). However, the exact orientation of the helices, their depth of bilayer penetration, and the details of the lipid and protein contacts still remain unknown. Therefore, the objective of this study was to determine the three-dimensional orientation of each helix relative to the lipid membrane in the pre-channel state.

Recently, we reported the membrane topology of amphipathic α-helices, H1–H5, of colicin E1 in its pre-channel state, and it was found that all five N-terminal helices retain their α-helical structure, with H3 and H5 elongating upon membrane association (28–30). Herein we continue our investigation of the membrane-bound topology of the remaining helices within the prechannel state of colicin E1. Using site-directed fluorescence labeling combined with our novel helical periodicity analysis method, we found that H6 and H7 are two separate and distinct amphipathic α-helices in the membrane-associated, pre-channel state and that H6 becomes overwound as a 3₁₀ helix upon membrane binding. In contrast, H10 is a 3₁₀ helix in the soluble colicin protein but opens up into a standard α-helix upon membrane binding.

EXPERIMENTAL PROCEDURES

Materials

All chemicals, unless otherwise stated, were purchased from Sigma. All steady-state fluorescence measurements were conducted with a PTI-Alphascan-2 spectrofluorimeter (Photon Technologies Inc.) equipped with a thermostatted cell holder, and data were reported as the mean ± S.D. and were performed at least in triplicate.

Mutagenesis, Protein Purification, and Monobromobimane Labeling

Each residue from Val⁴⁴⁷ to Gly⁴⁷⁵ and from Ile⁵⁰⁸ to Ile⁵²² of P190H₆ was individually replaced with a cysteine using the Stratagene (La Jolla, CA) QuikChange^TM mutagenesis kit. Plasmid DNA was purified using the High Pure Plasmid isolation kit from Roche Applied Science. Both the wild-type (WT) P190H₆ and Cys variants were prepared from transformed lexA⁻ E. coli IT3661 cells as described previously (10). Protein purity was assessed by SDS-PAGE, and protein concentration was determined by spectroscopy at A₂₈₀, using an extinction coefficient (ϵ) of 29,910 m⁻¹ cm⁻¹ (31). Purified Cys variants were labeled with monobromobimane (mBBr) with a molecular weight of 271.11 g mol⁻¹ (Molecular Probes Inc.) at a 20:1 molar ratio (probe/protein), and the labeling efficiency was determined as described previously (28).

Preparation of Large Unilamellar Vesicles (LUVs)

LUVs were prepared from 1,2-dioleoyl-sn-glycero-3-phosphocholine and 1,2-dioleoyl-sn-glyerco-3-(phospho-rac-(1-glycerol)) vesicles at a 60:40 molar ratio (Avanti Polar Lipids). Lipids were prepared and quantified as described previously (28), except the buffer used to suspend vesicles consisted of 10 mm DMG and 100 mm NaCl (pH 4.0). Asolectin (Fluka) was purified according to the method of Schendel and Reid (32), and vesicles were prepared as described previously (33). Phospholipid concentration was determined using the micro-Bartlett assay (28).

6-Methoxy-N-(3-sulfopropyl)quinolinium (SPQ) Assay for in Vitro Channel Activity

This assay was performed using a Cary Eclipse spectrofluorimeter (Varian Instruments). Asolectin vesicles were loaded with 16 mm SPQ in 10 mm DMG, 100 mm KCl, 1 mm CaCl₂ buffer, pH 5, using a freeze-thaw method (33), and the vesicles were diluted to 0.3 mg/ml in 20 mm DMG, 100 mm NaNO₃ buffer, at the desired pH. The vesicle fluorescence was monitored continuously for 2 min at 20 °C with constant stirring, after which protein was added (2 ng/ml, final concentration). Fluorescence measurements were taken using excitation and emission wavelengths of 347 and 445 nm, respectively. The spectral bandwidth for both wavelengths was set to 5 nm. The extravesicular buffer was 100 mm NaNO₃, 10 mm DMG, pH 6.0, and the encapsulation buffer was 100 mm KCl, 10 mm DMG, 1 mm CaCl₂, at pH 5.0. The total amount of encapsulated Cl⁻ was released by the addition of Triton X-100 (0.1%, final concentration) to the sample. The fluorescence changes upon the addition of the protein were reported as the percentage of maximum SPQ fluorescence (%F_max): %F_max(t) = (F − F_b)/(F_T − F_b) × 100%, where F_b is the initial residual fluorescence of the dye-loaded vesicles and F_T is the maximal fluorescence intensity after detergent lysis of the vesicles.

Steady-state Intrinsic Trp Fluorescence

The folded integrity of all proteins was examined using intrinsic Trp fluorescence as described previously (28). Both the native and bimane-labeled variants as well as the WT P190H₆ were diluted to 4 μm in PBS (50 mm NaH₂PO₄, 50 mm Na₂HPO₄, and 100 mm NaCl, pH 7.0). Intrinsic fluorescence was generated by excitation of Trp residues at 295 nm (2-nm excitation slit width), and emission was detected from 305 to 450 nm (4-nm slit width). The resulting traces were corrected for the buffer and wavelength-dependent bias of the emission components of the spectrofluorimeter before calculation of the λ emission maximum (λ_em,max) from the first derivative of the smoothed spectra.

6-Methoxy-N-(3-sulfopropyl)quinolinium Assay for in Vitro Channel Activity

The 6-methoxy-N-(3-sulfopropyl) quinolinium assay was performed as described previously (33) using a Cary Eclipse spectrofluorimeter (Varian Instruments). Both the labeled and unlabeled proteins were diluted to 4 μg/ml in extravesicular buffer consisting of 100 mm KCl and 10 mm DMG, and all buffers were adjusted to pH 5.0.

Bimane Fluorescence Emission Spectra

The steady-state bimane fluorescence emission spectra of all Cys variants were measured as described previously (28). All bimane labeled variants were diluted to 4 μm in DMG buffer (20 mm DMG and 130 mm NaCl (pH 4.0)) in the presence or absence of excess LUVs (800 μm, final concentration). The data were corrected for the buffer- and wavelength-dependent bias of the equipment before calculation of the λ_em,max from the first derivative of the smoothed spectra.

Steady-state Bimane Fluorescence Anisotropy

The steady-state 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 as described previously (28). Using the I_VV and I_VH fluorescence intensities, the anisotropy (r) was calculated as follows.

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. All bimane-labeled variants were diluted to 8 μm in DMG buffer (20 mm DMG and 130 mm NaCl (pH 4.0)) in the presence or absence of excess LUVs (800 μm, final concentration). The excitation wavelength 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. A solvent blank (DMG buffer or LUVs in DMG buffer) was subtracted from each intensity reading prior to the calculation of the anisotropy value. Probe mobility values were determined based on the inverse of the measured anisotropy values.

Dual Quenching Analysis

The depth measurement of each bimane-labeled residue relative to the bilayer was determined by measuring the quenching of bimane using both iodide (KI) and 10-doxylnonadecane (10-DN) as described previously (34). To measure iodide quenching (F_KI), the sample fluorescence was measured in ratio mode using semi-micro quartz cuvettes (0.5 × 0.5 cm) containing 100 μm LUVs and 7.5 μg of protein or LUVs only (background). F_KI was determined after the addition of a 50-μl aliquot of an aqueous solution from 1.7 m KI in 0.85 mm Na₂S₂O₃ stock solution. To measure 10-doxylnonadecane (10-DN) quenching, LUVs were prepared as above except that the LUVs were doped with 10 mol % 10-DN. All of the samples were incubated at 24 °C for 30 min before the measurement of initial fluorescence. For all measurements, the excitation wavelength was set at 375 nm with the emission intensity observed at 467 nm (2.5 and 5 nm for both excitation and emission bandpasses, respectively).

Calculation of the Iodide to 10-DN Quenching Ratio (Q-ratio)

The ratio of quenching by 10-DN relative to that by KI (Q-ratio) was used to determine the penetration depth of bimane in lipid bilayers as described previously (30). The Q-ratio was calculated from the formula,

where F₀ is the fluorescence of the sample lacking quencher. F_KI and F_10-DN are the fluorescence intensities in the presence of KI and 10-DN, respectively.

Prediction of Secondary Structure from Fluorescence Parameters

The secondary structure elements were predicted from the observed fluorescence parameters using a method adapted from Cornette et al. (35). The electronic center of a probe attached to the protein fluctuates with a trajectory confined in a cone of defined angle and will probably be experiencing an average dielectric constant, refractive index, electric field, dipole moment, etc., according to the average local probe environment during the lifetime of its excited state. Consequently, for a sequentially labeled helix, any fluorescent property exhibited by the probe will follow an amplitude and angular frequency described by a harmonic wave function. Now, considering both radial and longitudinal heterogeneity in the microenvironment along the length of the helix, we propose the following empirical function,

which corresponds to an amplitude-modulated, A(n), harmonic wave form of a generic fluorescent property Y, with period T (in residues per turn (r.p.t.)) and phase ϕ. This, in terms of the position of the residue n, for a given initial residue n_o (horizontal offset) gives the Y(n_o) value (vertical offset). In addition, the function includes a linear trend with slope m.

The amplitude of the sinusoidal function is damped within the envelope determined by the exponential profile,

with maximum amplitude, A_m, and time constant, τ. The fluorescent property considered for this analysis was the spectral centroid (SC) of the emission spectrum in wavelength mode, defined by the equation,

which reflects the center of mass of the whole emission band. The relevant parameters in Equations 3 and 4 (T, A_m, and m) were obtained by weighted nonlinear least squares regression of the experimental data from Equation 5, using the graphical and statistical package OriginPro 8.0 (OriginLab Corp.).

Data Fitting

The setting of the initial parameters for the nonlinear fitting of Equations 3 and 4, based on the calculated data with Equation 5, was performed as follows. (i) The initial residue n_o to consider (horizontal offset) was based on the secondary element reported in the crystal/soluble structure. (ii) The parameter SC_o (vertical offset) corresponded to the spectral center of mass at n_o, SC(n_o). It is an experimental value and was considered fixed for the most of the fitting. (iii) The setting of the slope m, either initialized for free-floating or fixed constant, was obtained according to the linear trend observed for the envelope of the harmonic function. (iv) The maximum amplitude A_m was usually a fixed parameter, set to half of the maximum Δ value observed. (v) The parameters ϕ and τ are correlated with each other; thus, in general, the phase was set fixed to either ϕ = 0 or π/2, depending on whether the SC for the second residue (n_o + 1) is higher or lower than the first residue (n_o), respectively.

For the data fitting, the weight of each value was kept uniform except for the last residue, which was set to a value of $\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$100$}\right.$ of the regular weighting. The rationale was to intentionally reduce the contribution of the last residue in the fitted result but then to evaluate the magnitude of the residuals for the basis of the assessment of the helical boundaries.

Soluble Form

For helix 6, the beginning of the helix was set at n_o = 447. Inspecting the data, the phase was set to ϕ = π/2, and the slope was set to m = 0. Thus, minimizing the χ² of the fitting, the period converged at T = 3.34 ± 0.08 (R² = 0.76) and A_m = 8.26, with residuals distributed uniformly with relative low dispersion (±0.5%). For helix 7, the initial residue was set at n_o = 461 and phase ϕ = 0. The data for bim463 were excluded from the analysis because it appeared as a clear outlier of the regular harmonic function. Thus, fixing the slope at m = −0.41 according to the trend observed, the period converged to T = 3.56 ± 0.03 (R² = 0.93) and A_m = 9.96, regardless of the initial values of the free-floating parameters T, τ, and A_m. The n-2 first residues presented a residual deviation lower than ±1.0%. Helix 10 was considered to initiate at n_o = 514, and, according to the plot, the phase was set to ϕ = 0. Again, with m < 0 and with free-floating T, τ, and A_m, the period converged to T = 2.95 ± 0.03 (R² = 0.81), regardless of the initial values of the floating parameter, although the maximum amplitude A_m reported was significantly elevated (A_m = 21.4). As an alternative fitting strategy, considering the data at residue 519 as an outlier (actually, by assigning it a relative weight of $\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$100$}\right.$), with free-floating m and A_m, the fitting yielded T = 2.99 ± 0.12 (R² = 0.85) and m = 0 ± 0.19 (a flat slope), but now with A_m = 11.45 ± 1.95. Nevertheless, in both cases, the estimated periods were closer to each other.

Membrane-associated Form

In the membrane-bound state, the harmonics behaved in a simpler fashion because the lack of the damped amplitude forced Equation 4 to reduce to A(n) = A_m. For helix 6, from an initial inspection of the output plot, the parameters were set to m = 0.15 and A_m = 10. Thus, starting with the residue n_o = 448, the fitting converged to T = 3.00 ± 0.15 (R² = 0.90) and ϕ = 0, with a low residual distribution of ±1.4%. Helix 7 exhibited a slope of m = 0. Thus, setting the initial parameters to A_m = 10 and ϕ = π/4, both either fixed or free-floating, and starting at n_o = 461, the fitting converged to T = 3.68 ± 0.03 (R² = 0.93) and A_m = 10.6 (for the free-floating case). For helix 10, the beginning of the helix was set at n_o = 513, the phase was set at ϕ = π/2, and the slope was set at m = 1.67. The fitting converged to T = 3.47 ± 0.04 (R² = 0.95) and A_m = 11.04 ± 0.9, with residuals lower than ±1.0%.

Accordingly, for all three helices in both the soluble and membrane states, the maximum amplitudes were in the range of A_m ∼10 ± 2.0 nm. This number may represent a constant value that reflects the maximum variation of the observed variable (center of mass) between the total buried and the fully exposed probe, under the experimental conditions of the measurements.

In Silico Analysis of the Soluble Structure

The modeler and simulator suite, Molecular Operative System, MOE 2011.10 (Computing Chemical Group, Inc.) was used for all of the in silico analyses, and the x-ray crystal structure of the soluble colicin E1 channel domain (Protein Data Bank entry 2I88) was used as the template structure. This structure was protonated by the MOE module, Protonated 3D (36), at pH 6.0, 300 K, and 0.1 m ionic strength and was energy-minimized using the OPLS-AA (optimized potential for liquid simulation for all atom) force field in a generalized Born/integral volume implicit solvent model approach (37) (with columbic electrostatic of ϵ_int = 2 and ϵ_ext = 80), in order to eliminate low energy clashes.

The SASA (in Å) of each residue in the ColE1 channel peptide was calculated after rolling a spherical probe with 1.4-Å radii over the residue surface. This value was standardized by the total exposed SASA, calculated in the context of the tripeptide Gly-X-Gly (38), with X the residue in consideration. The local period, T_i, to evaluate the helical content of the backbone was assessed by measuring the angle of rotation, θ_i, defined by the angle N_{i − 1}-N_i-N_{i + 1} because, for a perfect helix, nitrogen lies on a cylinder with θ ≅ 100°, resulting in a constant period of T ≅ 360/100 ≅ 3.6 r.p.t. (35).

RESULTS

Mutagenesis, Protein Purification, and mBBr Labeling

The colicin E1 structure consists of 10 α-helices, where H8 and H9 (colored black) are the hydrophobic helices that form the core of the soluble protein (Fig. 1A). The spatial relationship of H6, H7, and H10 (shaded in gray) relative to the entire channel peptide is shown. There is only one natural Cys residue (Cys⁵⁰⁵) deeply buried within the hydrophobic core of the protein. Normally, it does not react with any thiol-specific reagents under non-denaturing conditions. The sequence under investigation in this study was composed of H6, H7, and H10 and includes Val⁴⁴⁷–Gly⁴⁷⁵ and Ile⁵⁰⁸–Ile⁵²² (Fig. 1B). The sequences underlined in black were subjected to cysteine-scanning mutagenesis. A total of 43 Cys variants were prepared, and mutation sites were confirmed by DNA sequencing.

FIGURE 1.

Preparation of colicin E1 Cys variants. A, ribbon topology diagram of the 2.5 Å crystal structure of the P190 peptide (Protein Data Bank entry 2I88). The overall structure consists of 10 α-helices, where H8 and H9 (shown in black) are the hydrophobic helices that serve as a membrane-anchoring helical hairpin in the membrane-associated state. Gray residues (shown in sticks) were subjected to Cys codon replacement. B, the primary sequence and secondary structure of the channel-forming domain of the colicin E1 (P190H₆). Residues underlined with a boldface line were subjected to Cys substitution.

All Cys variants exhibited WT expression levels except for F464C, which showed almost no expression. This indicates that Phe⁴⁶⁴ might play an important role in the folded structure of the ColE1 channel domain (CD). As a result, this variant was eliminated from this study. All Cys variants were purified as described previously (39), and purity was assessed by SDS-PAGE analysis with >95% homogeneity.

In order to report on the local environment of the residues, each Cys variant was subjected to mBBr labeling as described previously (40). Monobromobimane was used for labeling in this particular study because it is a well characterized, relatively small, non-perturbing fluorophore that is essentially non-fluorescent until conjugated to the protein of interest through disulfide linkage with a cysteine side chain (Fig. 2A). Therefore, it is an ideal fluorophore for the purpose of spectroscopic studies. Although there is one naturally occurring cysteine residue located at residue 505 in H9, this residue is deeply buried and does not react with any thiol-specific reagents under non-denaturing conditions.

FIGURE 2.

Labeling of Cys ColE1 variants with mBBr. A, left, energy-minimized structure of the colicin E1 adduct 467BIM, labeled in silico based on the x-ray crystal structure. The helices were colored according to the position in the domain: from blue (N-terminal) to red (C-terminal). Right, chemical structure of mBBr. B, labeling efficiency measurement of the 43 Cys ColE1 variants with and without the C505A mutation (hollow and dark bars, respectively). Every other residue (odd numbers) is shown on the abscissa. Missing data for F464C are due to low expression level of this variant.

The mBBr labeling efficiency, which represents the fraction of the variant being labeled (no label, 0%; complete labeling, 100%) is shown in Fig. 2B. Notably, only 8 of the 43 Cys variants exhibited labeling efficiency of 40% or less. In order to improve the labeling efficiency, one option is to slightly denature the variant with 4 m urea prior to labeling followed by removing the urea to refold the variant. However, the only problem with this method is that Cys-505 may react with the thiol-specific probe during this procedure. To resolve the issue, an additional eight variants were prepared using the C505A mutation to facilitate the preparation of single-Cys variants. It was previously reported that the C505A mutation does not perturb the folded structure and does not impair channel activity (30). Labeling efficiency was significantly restored with the incorporation of C505A mutation with the above approach except for the I454C variant (Fig. 2B). In general, labeling efficiency of most variants ranged between 60 and 100%.

Structural and Functional Analysis of Cys Variants

Prior to spectroscopic measurements of the variants, it is important to assess both the structural and functional integrity of the proteins. Herein, the Trp fluorescence emission maximum values (Trp λ_em,max) were measured to provide a measure of the average local environment of the Trp residues and, hence, the folded integrity of the proteins. A red shift in the Trp λ_em,max values would be expected for mutation-induced alteration of the folded integrity. As shown in Fig. 3A, the WT showed a Trp λ_em,max value near 318 nm. In fact, most Cys variants exhibited Trp λ_em,max values similar to the WT protein except for a few variants, including T457C, W460C, A472C, G475C, L513C, and L520C. According to the crystal structure, these residues are all located at the boundaries of H6, H7, and H10. It is likely that these residues provide critical non-covalent interactions between individual helices that help to maintain proper folding of the channel protein. Alternate hypotheses might be (i) the original residues are contributing to the hydrophobic environment (blue feature), or (ii) the replacement Cys residues are interacting with Trp emitters. Nonetheless, Cys substitution at these sites induces only a minor perturbation to the folding integrity of the protein.

FIGURE 3.

Structural and functional tests of Cys ColE1 variants. A, Trp λ_em,max (nm) values were measured and compared against the WT ColE1 to assess the folding integrity of the variants. B, the channel-forming activity of the variants with bimane modification (dark bars; representative variants shown) and without bimane (hollow bars) was assessed by measuring the flow rate of SPQ (fluorescent dye) through the channel in LUVs. Data are missing for F464C due to the low expression level of this variant. Error bars, S.D.

In order to assess the effect of Cys replacement on the ColE1 CD channel activity, an SPQ activity assay was performed to measure the in vitro channel activity of the protein as described previously (27). As shown in Fig. 3B, most Cys variants showed channel activity ranging from 80 to 120% of the WT level, with the lowest value (A472C) at 75% of the WT. Thus, correlating with the data shown in Fig. 3A, it would seem that Ala⁴⁷², Thr⁴⁵⁷, and Leu⁵¹³ may be important residues for maintaining CD folded integrity, and thus Cys replacement causes a minor reduction in channel activity. Consequently, Cys substitution of these residues did not significantly impair channel activity of the variants. Also, modification of the Cys side chain with bimane did not impair channel function (Fig. 3B).

Bimane Fluorescence Emission Maxima of H6, H7, and H10 Cys Variants

In order to determine the relative local environment of the bimane fluorophore within the protein according to its water accessibility, the bimane λ_em,max was measured for each Cys variant. In theory, more deeply buried residues are expected to have a lower bimane λ_em,max (blue-shifted) than solvent-exposed residues (red-shifted) in the membrane-bound state. To calibrate the bimane λ_em,max values, a standard polarity curve was generated previously in our laboratory (28) using a bimane-N-acetyl-Cys model compound in a series of dioxane/water mixtures with known dielectric constant (ϵ) values. Previously determined parameters (λ_em,max ≤ 455 nm = buried; 455 nm < λ_em,max < 470 nm = moderately accessible; λ_em,max ≥ 470 nm = solvent-accessible) were used to calibrate the bimane local environment.

The bimane λ_em,max data suggest that H6, H7, and H10 are amphipathic in nature, and the polar faces of these helices contain residues that are highly solvent-accessible (Fig. 4A). Importantly, the data correlate well with the crystal structure; residues with higher bimane λ_em,max values are also the hydrophilic residues that are exposed to the solvent side and vice versa. By visual inspection, the data suggest that both helices 6 and 7 are three-cycle α-helices, whereas H10 is a two-cycle α-helix. Interestingly, there was no significant difference in the helical lengths between the soluble and lipid-bound states, which suggested that H6, H7, and H10 retain their helical pattern in the lipid-bound state (Fig. 4A).

FIGURE 4.

Fluorescence emission maximum and probe mobility of the bimane-labeled Cys variants of the colicin E1 channel domain. A, the bimane λ_em,max values of the Cys variants in the soluble (■) and membrane-associated state (□) were measured. B, probe mobility (1/r) was calculated as the inverse of the observed fluorescence anisotropy (r). Absolute probe mobility (1/r) of the bimane-labeled Cys variants in both the soluble (■) and membrane-associated state (□) were measured. Average and S.D. (error bars) values for at least triplicate measurements are shown.

Bimane Fluorescence Anisotropy and Probe Mobility of H6, H7, and H10 Cys Variants

In order to determine the probe mobility of the individual residue side chains within H6, H7, and H10, this can be calculated by measuring the inverse of the bimane fluorescence anisotropy. In theory, surface-exposed residues are expected to have lower anisotropy (higher probe mobility) than buried residues that are facing the interior of the protein or membrane bilayer due to differences in the viscosities of the membrane bilayer compared with the aqueous medium. As shown in Fig. 4B, most Cys variants showed lower probe mobility (1/r) values in the membrane-bound state, and this supports the idea that the side chain mobility is often reduced upon lipid association, probably due to steric hindrance due to the higher viscosity.

Based on the probe mobility data (Fig. 4B), it appears that the residues within H10 in the soluble protein have higher mobility than the residues in either H6 or H7, which is not surprising because it is the last helix of the channel protein. Notably, residues within H6 appear to have higher mobility than H7 in the soluble ColE1 CD. The bimane λ_em,max and probe mobility data are well correlated in terms of both the helical periodicity and pattern. Residues that are highly solvent-accessible tend to have higher mobility, and vice versa.

Dual Quenching Analysis of the Membrane-bound Depth of H6, H7, and H10 Cys Variants

In order to determine the relative membrane penetration depth of each residue within these three helices in their lipid-bound state, a dual fluorescence quenching method was used as described previously (41). Two types of quencher species were employed in this assay (KI, aqueous quencher; 10-DN, membrane-embedded quencher). In theory, surface-exposed residues in the lipid-bound state are expected to be more quenched by KI, resulting in higher values of (F₀/F_KI − 1), where F₀ and F_KI represent the fluorescence intensity of bimane in the absence and presence of KI, respectively. In contrast, residues that are buried in the lipid-bound state are expected to be more quenched by 10-DN, resulting in higher values of ((F₀/F_10-DN) − 1), as shown in Fig. 5. In order to calculate the relative membrane penetration depth, the quenching ratio (Q-ratio; ((F₀/F_10-DN) − 1/(F₀/F_KI) − 1)) was calculated. It should be noted that H8 and H9 form the hydrophobic hairpin within the membrane; therefore, it is not surprising that H7 adopts a tilted topology with a more buried C terminus because it is connected to the anchor domain.

FIGURE 5.

Plot of the quenching ratio. The Q-ratio represents the ratio of quenching by 10-DN to that by KI (see “Experimental Procedures” and Equation 2), and it correlates to the penetration depth of the residues within the bilayer. Average and S.D. (error bars) values from triplicate measurements are shown.

Harmonic Analysis

The initial fluorescence emission results support the amphipathic character of H6, H7, and H10. However, a more robust quantitative analysis is required for helical characterization. In this sense, a harmonic analysis was performed using Equations 3–5. The strategy employed for the fitting was to start with an initial value of T = 3.6 r.p.t. (regular α-helix) and to include a range of residues in the analysis according to the crystal structure (for details, see below).

Soluble Form (Fig. 6A)

FIGURE 6.

A, harmonic analysis of the SC for the soluble state of the colicin E1 CD. Connected symbols, calculated SC (Equation 5) for the emission spectrum of the bimane adduct of the CD as a function of residue position. Solid lines, the best fit according to the functions in Equations 3 and 4. The signal at position 519 (boxed symbol) was considered an outlier for the fitting. B, the backbone local period was the local period of the helical segments according to the measured angle of rotation, as defined under “Experimental Procedures,” for the soluble structure. The solid horizontal line was set at T = 3.6 for reference. C, correlation SASA-mobility for the soluble form. Solid symbols and solid line, calculated SASA for the original side chains, standardized according to the value for the residue totally exposed and then normalized by the highest value in the helix. Empty symbols and dashed line, normalized mobility of the bimane adduct, normalized by the highest value in the helix.

For H6, the fitting for residues 447–457 gave a helical periodicity of T = 3.34 ± 0.08 r.p.t. According to the residue dispersion, residue 457 was included as part of H6. For H7, the fitting over residues 461–475 converged to T = 3.56 ± 0.03 r.p.t., with residues 474 and 475 clearly not a part of the periodicity. H10, when fitted over residues 514–522, gave a smaller periodicity of T = 2.99 ± 0.12 r.p.t., and clearly residue 522 did not follow the harmonic function. Thus, for all three helices, the maximum amplitudes used/reported were A_m = 10 ± 1.5 nm, which is consistent with the same range of values either for exposed or buried probe locations.

On the other hand, the helical content of H6, H7, and H10 was quantified by the average angle of rotation and the derived periodicity (Fig. 6B) for all but the end residues belonging to the helices. Effectively, the backbone of H6 presented a shorter average period (〈T〉 = 3.48 r.p.t.) than for an ideal α-helix, with irregular dispersion around the medium value and a particularly low outlier seen for residue 448. On the contrary, for H7, the average value is nearly optimal (〈T〉 = 3.58 r.p.t.) and quite regular except for residues 473 and 474. Curiously, H10 shows typical helix properties for the backbone but shows extreme values for both terminal residues (514 and 521).

Anisotropy of the ColE1 CD Bimane Adducts

Protein adducts were analyzed according to the mobility (M = 1/r), using the same harmonic function as described above, with limited success either for the soluble or the membrane-bound form. Despite the amplitude signal differences between high and low values, there was no simple periodic waveform detected. However, certain trends were evident in the overall pattern. For instance, the maximum mobility for H6 in the soluble form followed an exponential waveform (Equation 4), whereas the minimum values follow a linear trend. In the case of the soluble H7, the behavior is opposite to that observed for H6 but with two layers of decreasing exponential mobility. Nevertheless, the mobility observed by the bimane moiety is in good agreement with the calculated solvent accessibility for the original residue depicted in the correlation shown in Fig. 6C for the standardized signals.

Membrane-associated Form (Fig. 7A)

FIGURE 7.

A, harmonic analysis of the SC for colicin E1 for the membrane-bound state. Connected symbols, calculated SC (Equation 5) for the emission spectrum of the bimane adduct of colicin E1 at the indicated residue. Solid lines, best fit according to the functions in Equations 3 and 4. B, schematic diagram illustrating the predicted penetration depth based on the Q-ratio data and the harmonic analysis. The penetration depth was calibrated using dielectric constants measured from the bimane λ_em,max data.

In the membrane-bound state, all cases did not exhibit damped behavior (τ → 0, A = A_m), as was observed for the soluble state of the ColE1 CD. For H6, the fitting clearly converged to T = 3.00 ± 0.15 r.p.t., and according to the distribution of residuals, neither residue 447 nor 457 should be part of the helix. For H7, the fitting yielded T = 3.70 ± 0.1 r.p.t., and the analysis of the residuals led to extension of H7 to residue 473. In the case of H10, working in the range from residue 513 to 522, the fitting converged to T = 3.47 ± 0.04 r.p.t., and the helix can be extended to residue 522.

DISCUSSION

In this study, we have successfully scanned the membrane-bound topology of H6, H7, and H10 and showed that they are also amphipathic α-helices that are located on the membrane surface which provide further support for the umbrella model of colicin E1 in the pre-channel membrane-bound state. The calculated helical periodicity in this study revealed that H6, H7, and H10 are three separate amphipathic α-helices. These data help to refute earlier models, which proposed that both H6 and H7 adopt a transmembrane orientation to assist with channel formation in the open channel state (22). In fact, H1–H5 of the colicin E1 channel domain were previously demonstrated to exist as amphipathic α-helices upon membrane association by a similar method (24–27).

In this study, all of the Cys ColE1 variants were highly purified, and the bimane labeling efficiency was close to 80% for most variants (Fig. 2). Both structural and functional tests revealed that the Cys substitution did not negatively impact either the folded integrity or channel activity of the ColE1 CD (Fig. 3). In this study, both the Q-ratio and probe mobility data strongly indicate that H7 adopts a tilted topology on the membrane surface with a more buried C terminus. This appears to accommodate both H8 and H9 to adopt an anchor structure in the pre-channel state, which lends further credence to the umbrella model of the pre-channel state for colicin E1 (Fig. 8). In addition, because H6 exhibits a more buried N terminus in this study, these results also correlate well with data from our previous study, which revealed that the H5 loop region is deeply embedded within the membrane bilayer in the pre-channel state (27).

FIGURE 8.

Models of colicin E1 in the membrane-bound state. The models were adapted from Ref. 39, where the tilt angles and depth for H6 and H7 were modified while maintaining the membrane context of the connecting coil regions. The dots represent the raw data for lipid carbonyl groups of the outer leaflet of the plasma membrane. A, upper view. B and C, side views with a 90° rotation difference between B and C.

As mentioned above, the use of the SC (Figs. 6A and 7A) offered more uniform and consistent results because this variable summarized the overall spectra, including the position of the emission maximum, the bandwidth at half-maximum, and any skewness, being a more robust reporter of the environment around the probe than any independent property by itself. Effectively, a plot of the wavelength at the maximum (Fig. 4A) reported for the bimane probe was noisier and less periodic than the SC. Furthermore, the calculation of SC revealed consistent differences between the soluble and membrane-associated states in regard to the bimane fluorescent properties.

In addition, the current harmonic function employed herein offered a more powerful fitting capability than previously employed, with parameters that hold physical meaning. For a mixed solvent system (i.e. membrane-water interface), a homogenous value for any physical parameter (i.e. the average polarity, P) in the direction parallel to the interface (xy-plane) or for a vertical gradient in either direction from z = 0 (z-gradient) is expected. Thus, for a helix embedded at the interface for this system, the angular distribution of the fluorescence should be independent of the position of the probe in the helix, unless the principal inertial helix axis presents a tilt angle with respect to the plane of the membrane. The parameter m in the harmonic function can account for this topology. Now, for a soluble protein, it is expected that the environment around a particular helix will be more heterogeneous because there is expected to be much more variability in the composition and distance according to the sequence and structure of the protein. Thus, we included an element of asymmetry by considering a longitudinal gradient given by the exponential amplitude A(n). That is to say, for a globular protein, the angular and longitudinal heterogeneity of P(n) can be described by the parameter τ and m. However, any specific effect by a particular interaction at any location will be reported as an outlier of the regular harmonic trend.

According to the solution structure of the ColE1 CD, H6 shows some variability, whereas H7 is nearly a perfect α-helix (Fig. 6, A and B). Incidentally, both helices were similarly predicted by the harmonic analysis. In contrast, H10 appears as an overwound helix, as predicted by the harmonic analysis with T ≅ 3.0 r.p.t. (i.e. a 3₁₀-helix- while its backbone still behaves as a short regular α-helix). The origin of this discrepancy might be addressed by studying the particular location and interaction of the probe using molecular modeling approaches. It is noteworthy that only H7 required both parameters (m and τ) to account for the scattering of the fitted data. This is consistent with its location in the folded structure. This helix goes from the outside to the inside of the protein core, consistent with the negative slope of the SC.

For the membrane-associated state of the ColE1 CD, there is no high resolution structural model to correlate with the output from the harmonic analysis. Nevertheless, the current thinking is that the surface location for H6, H7, and H10, and consequently the bimane probe will experience a more homogenous environment than in the globular soluble form in regard to the angular and longitudinal dimensions. This statement was consistent with the fact that all of the helices lacked the damping behavior (τ → 0, A(n) = A_m), and, contrary to the general trend for the soluble protein, the parameter m could account for longitudinal variation. Thus, according to this analysis, the inertial axis of H6 and H10 seem to be tilted with a positive angle (leaving the membrane), whereas H7 is horizontal with the plane of the membrane. Upon membrane binding, H6 exhibited 3₁₀-helix character (T = 3.0 r.p.t.), whereas H7 and H10 exhibited more relaxed structures than in the soluble form. The anomalies for H6 are also evident in its shorter length by two residues. On the contrary, H7 and H10 apparently are extended by one residue each.

The harmonic analysis for the probe mobility is complex and might require the combination of multiple sinusoidal components (one of them may be the periodicity of the helix, T) to account for the data scatter. The caveat here is the lack of physical meaning ascribed to such new fitting components. However, the overall change of this variable must be associated with the topology of the helix. Indeed, the increased mobility is consistent with the probe environment, for example when the probe approaches the H6-H7 loop (C terminus of H6) and, conversely, decreased mobility when the probe approaches the N terminus of H7, exhibiting in both cases a complementary funnel-like shape. The low mobility shown by the linear trends in both cases probably describes the restricted environment for the probe inside the core of the protein. In contrast, when the probe is facing the external surface or is found within the loop region, the mobility is much higher. Thus, the degree of side chain solvent exposure for the original residues correlates well with the mobility of the bimane probe (Fig. 7A). However, some discrepancies that do exist reflect site-dependent interactions that might limit the range of rotation of the probe. This is the case for Ser-449. On the contrary, the null calculated exposure of Val⁴⁷³ and Ala⁴⁷⁴ predict restricted mobility. The relatively high mobility for these residues provides indirect evidence of local structural changes reported by the probe in relation to the original side chains, and/or the probe is inside an internal pocket of ample size with little or no interactions.

In general, the fluorescence data in this study correlated well with the crystal structure of the soluble ColE1 peptide (16, 17), as deduced from the overall comparison between the harmonic analysis of the signal and the periodicity of the backbone. These data also prove invaluable toward the construction of an improved membrane topology model of the CD in the pre-channel state (Fig. 8). Our revised model further demonstrates the circular arrangement of the helices in the membrane-bound state of ColE1. However, one cautionary note is that the three-dimensional orientation of the channel is still unclear because a wide range of feasible three-dimensional configurations are possible on the membrane surface. To deduce the various possible three-dimensional orientation models, our laboratory will apply a rigorous and high density FRET analysis approach to map the interhelical distances within the ColE1 CD. This approach uses three Trp donors with the Cys-bimane as the acceptor fluorophore, and this method should provide new data for the construction of a low resolution three-dimensional model of the pre-channel state of the ColE1 CD (33). This method will also provide information toward the possible oligomeric state of the colicin E1 channel in the membrane because it was previously demonstrated that more than one molecule may participate in the channel structure of colicin Ia (34).