Calorimetric Investigations of the Structural Stability and Interactions of Colicin B Domains in Aqueous Solution and in the Presence of Phospholipid Bilayers*

The effects of pH and temperature on the stability of interdomain interactions of colicin B have been studied by differential-scanning calorimetry, circular dichroism, and fluorescence spectroscopy. The calorimetric properties were compared with those of the isolated pore-forming fragment. The unfolding profile of the full-length toxin is consistent with two endothermic transitions. Whereas peak A (T m = 55 °C) most likely corresponds to the receptor/translocation domain, peak B (T m = 59 °C) is associated with the pore-forming domain. By lowering the pH from 7 to 3.5, the transition temperature of peaks A and B are reduced by 25 and 18 °C, respectively, due to proton exchange upon denaturation. The isolated pore-forming fragment unfolds at much higher temperatures (T m = 65 °C) and is stable throughout a wide pH range, indicating that intramolecular interactions between the different colicin B domains result in a less stable protein conformation. In aqueous solution circular dichroism spectra have been used to estimate the content of helical secondary structure of colicin B (≈40%) or its pore-forming fragment (≈80%). Upon heating, the ellipticities at 222 nm strongly decrease at the transition temperature. In the presence of lipid vesicles the differential-scanning calorimetry profiles of the pore-forming fragment exhibit a low heat of transition multicomponent structure. The heat of transition of membrane-associated colicin B (T m= 54 °C at pH 3.5) is reduced and its secondary structure is conserved even at intermediate temperatures indicating incomplete unfolding due to strong protein-lipid interactions.

Membrane proteins are characterized by water-exposed extramembranous as well as intramembranous regions, which are embedded in the interior of the lipid bilayer. As a conse-quence, the energetics of stabilization of membrane-inserted proteins is expected to include contributions from these two regions in different proportions. Whereas the extramembranous region is more likely to be organized by forces similar to those described for water-soluble proteins (1,2), the membrane-inserted regions are very stable because of considerable strengthening of hydrogen-bonding interactions within the lipidic milieu (3).
The transition of proteins from a soluble to a membraneassociated state offers a new challenge for describing the energetics involved in the process of membrane insertion, which are critical for the proper folding of membrane proteins (3,4). A key aspect in lipid-protein interactions is the partitioning of the protein from water into the membrane. The conformational changes, however, that are associated with protein insertion into, transport across, and maintenance within membranes remain poorly understood.
Pore-forming colicins are water-soluble proteins that insert into membranes. Merely by offering a different environment and without the need of chemical modifications, these proteins can be investigated as well in solution as in their membraneinserted state. Therefore, they provide excellent model systems to study the structural changes during membrane insertion. Pore-forming colicins (e.g. colicin A, B, N, E1, and Ia) invade Escherichia coli cells by parasitizing receptors of the outer membrane and use the Tol or Ton protein complexes for translocation. In a final step these proteins cause voltage-dependent pore formation. The three functions correlate with three distinct domains of the polypeptide chain, which are arranged along the primary sequence in the following order: translocation, receptor recognition, and channel formation.
Here we study colicin B (ColB), 1 which is a 54.6-kDa protein secreted by Escherichia coli as a monomeric unit. Primary overall sequence comparison and functional studies indicate that ColB is most closely related to colicins A, N, and Ia (5-7). It has been suggested that, after binding to the target cell but before insertion occurs, the colicin pore-forming domains undergo conformational changes to adopt an insertion-competent state. This transition is induced in vitro by acidic pH. Membrane-active colicins thereby exhibit close similarity to other proteins such as diphtheria, tetanus, and botulinum toxins (8). Previous calorimetric studies have contributed some understanding of the structural transitions of the colicin A and colicin E1 pore-forming domains (9,10). However, the conforma-tional changes of full-length colicin B or its thermolytic fragments, which are necessary for membrane binding and insertion, have so far not been characterized.
Macromolecules are stabilized by the cooperative action of numerous weak forces. Because such highly cooperative structures undergo phase transition-like conformational changes upon exposure to heat, relevant thermodynamic, and structural information is obtained from differential-scanning calorimetry (DSC) and circular dichroism (CD) studies. Unlike the many calorimetric studies conducted with water-soluble proteins, the stability of only a few membrane proteins has been measured calorimetrically, and even less is known about the thermal stability of proteins that insert into membranes (9 -14). Investigations of full-length colicin B as well as its Cterminal fragment both in solution and the membrane-associated state allows for a direct comparison of the thermal characteristics in these different environments. In addition, by investigating the pH-dependent properties of full-length colicin and its pore-forming fragment, important information on the domain structure and inter-domain interactions is obtained (1). The observed modulations of thermal stability by interactions between domains and with the membrane have important implications also for the insertion and translocation competence of colicin B.

Preparation of ColB and ColB PFF-ColB was expressed in
E. coli. The full-length protein as well as its thermolytic fragment were purified following procedures described elsewhere (15,16). This latter domain of the homologous colicin A protein has been shown to exhibit pore-forming activities (e.g. Ref. 17) and the thermolytic fragment is, therefore, abbreviated PFF. SDS-polyacrylamide gel electrophoresis of ColB and its poreforming fragment shows one single band in Coomassie Bluestained gels.
Preparation of Lipid Vesicles-All phospholipids were from Avanti Polar Lipids (Birmingham, AL) and used without further purification. Small unilamellar vesicles of POPC:cardiolipin:POPE (5:1:2, w/w) were prepared from a homogeneous solution of the three lipids in chloroform/methanol (1:1). The solvent was completely removed with a stream of nitrogen and by exposure to high vacuum overnight. Thereafter, buffer was added and the suspension was subjected to several freeze-thaw cycles. A clear suspension of small unilamellar vesicles was obtained after repeated extrusion first through 100-nm (5 times) and thereafter 50 nm polycarbonate filters (Avestin, Milsch, Laudenbach, Germany). The lipid concentration for DSC, CD, and fluorescence experiments was 50 mg/ml. Differential-scanning Calorimetry-DSC was used to measure the transition temperatures (T m ) of full-length ColB and its pore-forming fragment. T m is defined as the temperature at which the excess heat capacity is maximal. Lyophilized protein was dissolved in buffer (0.1 M KCl, 0.02 M glycine at pH 3.5-4.5; or 0.1 M KCl, 0.02 M tris-malate at pH 6.0 -7.0), followed by dialysis overnight against the same buffer. Excess heat (C p ) versus temperature scans were obtained from 37 M protein solutions using a high sensitivity differential scanning calorimeter MicroCal VP-DSC. The sample and reference solutions were carefully degassed under vacuum for 5 min before loading the cells (0.56 ml). After equilibration of the system at 10°C, the temperature was increased to 100°C and decreased to 10°C at scan rates of 1°C/min. Thereafter another 10 -100°C scan was obtained. This rescan, which in all cases showed no evidence of reversibility, was subtracted from the initial scan, and the resulting baseline was corrected.
Circular Dichroism (CD) Spectropolarimetry-CD spectra (190 -250 nm) were recorded on a Jasco J-715 spectropolarim-eter using a quartz cuvette of 0.1-mm path length. The protein concentration was 3 M. For temperature scans at 222 nm, a heating rate of 1°C/min was used (20 -85°C). CD spectra were recorded every 5-10°C in the temperature range 20 -80°C as indicated in the figures. The helicity of the peptides was determined quantitatively only for spectra recorded at 25°C, because limited protein aggregation was observed above the transition temperatures. All measurements were corrected for buffer contributions. The lipid concentration was kept constant.
Intrinsic Emission Fluorescence Measurements-All emission fluorescence experiments were performed using a PerkinElmer LS-50 luminescence spectrometer with a waterthermostat cell holder. Excitation wavelengths of 290 nm were used, and the emission spectra were recorded within the range 300 -500 nm. The temperature dependence of the fluorescence intensity was recorded at a constant heating rate of 1°C/min from 25-85°C at 320 nm. The protein concentration of ColB or its pore-forming fragment was 3 M.

RESULTS
To test for the effect of pH on the thermal stability of ColB and its pore-forming fragment in solution as well as in the presence of lipids, the proteins were investigated in a temperature-dependent manner by DSC, fluorescence, and CD spectroscopy.
Differential-scanning Calorimetry-The temperature dependencies of the excess molar heat capacity of the full-length ColB between 10 and 100°C are shown in Fig. 1 as a function of pH. The scans were obtained at heating rates of 1°C/min. Denaturation was irreversible, because the transitions are absent during cooling or reheating of the samples. In this study the "heat of transition," ⌬H, therefore, relates to the heat absorbed during the process of denaturation and cannot be fully interpreted as the enthalpy of a reversible process. 2 Two superimposed endothermic transitions give the best theoretical fit of the scans obtained at pH 7.0: peak A with T m ϭ 54.9 Ϯ 0.5°C, 3 and peak B with T m ϭ 58.7 Ϯ 0.8°C (n ϭ 3) (⌬T m(B-A) ϭ3.9°C). At pH 3.5 both transitions occur at lower temperatures and with better resolution. The transition temperature of peak A is 29.8°C and that of peak B 40.9°C (⌬T (B-A) ϭ 11.1°C). The highest T m was observed at pH values of Ͼ4.5. When the pH is decreased in a stepwise manner from 7.0 to 3.5, peaks A and B decrease by 25.0°C and 17.8°C, respectively. Therefore, although both components are affected by pH, peak B is less sensitive to changes in proton activity and more resistant to thermal denaturation than peak A. The transition temperatures obtained at pH 2.5 and pH 9.0 are very similar to those of pH 3.5 and pH 7.5, respectively (not shown).
The heat absorbed during unfolding is obtained by integrating the area under the fitted curves of each transition. Table I lists the unfolding parameters determined using linear and quadratic baselines in a multiple-curve fitting routine of the DSC data. The calculated ⌬H, for peaks A and B at different pH are included in Table I. The pH dependence of ⌬H for peak A and B is depicted in Fig. 2. Although ⌬H A increases, ⌬H B decreases at the same rate. The observation of two well-distinguishable transitions could be due to different protein conformations or due to separate conformational unfolding steps of discrete domains within the same protein. To further discriminate between these possibilities, ColB was heated to 35°C, slightly above the T m of peak A at pH 3.5, and was kept at this temperature for 10 min. A DSC experiment with partially unfolded ColB was then performed at a heating rate of 1°C/min between 25 and 85°C (Fig. 3). A single transition with ⌬H ϭ 98 kcal/mol and T m ϭ 62.1 Ϯ 0.1°C (S.D., n ϭ 3) was observed, which is 21°C increased from T mB of the continuous scan. This result indicates that the transition of peak B is subject to interactions with components contributing to peak A.
To study the origin of the two transitions of ColB, a 203amino acid C-terminal fragment of ColB that contains the pore-forming domain was also investigated. Fig. 4 shows the temperature dependence of the excess molar heat capacity of the pore-forming fragment at different pH values at a heating rate of 1°C/min. Irreversible denaturation was observed after heating to 100°C. The denaturation profile of the pore-forming fragment in solution shows one main transition corresponding to 90% of the total ⌬H. The transition temperature for the main transition is 65.7 Ϯ 1.5°C at pH 7.0. A minor transition with T m ϭ 55.5 Ϯ 1.5°C is consistently present and corresponds to the remaining 10 Ϯ 4% of the total heat of transition. The transition temperature of the pore-forming fragment is pHindependent between pH 3.5 and 9.0 ( Fig. 4 and Table II). The pH-dependent changes in ⌬H of the PFF and peak B of ColB exhibit comparable characteristics (Tables I and II). By changing the pH from 3.5 to 7.0 the relative decrements in ⌬H are 63 and 59% for ColB peak B and the pore-forming fragment, respectively. A minor increase in heat of transition when the pH was lowered from 7 to 5 was observed for the related ColA PFF (Fig. 2 in Ref. 9). The effect of pH on ColB reveals more than one protonation state, suggesting the existence of several conformational states (Fig. 5A). In contrast, between pH 3.5 and 7.0 both transitions of the PFF occur in a pH-independent manner (Fig. 5B).
The denaturation profile of ColB in the presence of small unilamellar lipid vesicles is shown in Fig. 6A. At pH 7.0 ColB exhibits the same two transitions observed during its unfolding in aqueous solution in agreement with weak membrane-association of the protein at neutral pH. The T m values of peaks A and B are 55.4 Ϯ 0.4°C and 57.9 Ϯ 0.4°C, respectively. At pH 3.5, however, peak A exhibits a T m of 53.3 Ϯ 0.9°C and peak B is absent (Fig. 6B). In the presence of lipids the total ⌬H of ColB at pH 7.0 is 172 kcal/mol. At pH 3.5 ⌬H equals 140 kcal/mol in the presence of lipids, a value similar to the ⌬H observed for peak A at pH 3.5 in aqueous buffer alone ( Fig. 1D and Table I). However, in the presence of lipids this latter transition has shifted to higher temperatures by at least 20°C (Fig. 6B). Fig. 6C shows the DSC profile of the pore-forming fragment in the presence of lipids. The transition occurs over a wide temperature range and ⌬H is significantly decreased (Table  III). After rescan subtraction and baseline correction line-fitting analysis reveals a pattern of multiple transitions at pH 7.0 or pH 3.5 (Figs. 6, C and D). The calculated values of T m and ⌬H total are listed in Table III. CD Spectropolarimetry-Far UV-CD spectra of ColB and the pore-forming fragment in solution and in the presence of lipids at pH 7.0 display a maximum at 191 nm as well as minima at 208 and 222 nm (Figs. 7C, 7D, 8B, and 8C). The CD line shapes are, therefore, characteristic of a large proportion of ␣-helical secondary structures. When the mean residue ellipticity at 222 nm is taken as an indicator, helix contents of 39 and 82% are obtained for ColB and its PFF, respectively, in aqueous solution before starting the heating scan (Figs. 7C and 8B). In the presence of liposomes optical distortions by absorption flattening and light scattering would result in an underestimation of the helical content (18). These effects are minimized when small unilamellar vesicles prepared by extrusion through 50-nm filters are used (18). Nevertheless, no attempt was made to calculate the helix content quantitatively from the CD spectral line shape in the presence of lipids or after thermal denaturation.
The relative intensity changes of the ellipticity at 222 nm, however, were recorded as a function of temperature and compared with the DSC thermal transitions (Fig. 7, A and B).  However, the maximal level of unfolding is reached at a slower rate when compared with ColB at neutral pH (Fig. 7, A and B). The inset shows the derivatives of the experimental curves. Fig.  7 (E and F) corresponds to the full-wavelength scans for ColB in solution and in the presence of lipids, respectively, at different temperatures and at pH 3.5.
The temperature dependence of the CD spectra of the poreforming fragment in solution at pH 7.0 exhibits single transitions at 64.4°C and 62.3°C in the absence or in the presence of lipids, respectively (Fig. 8A). This former value is in good agreement with the transition temperature of 65.7°C observed in DSC experiments.
Although in the absence of lipids much of the helix contributions are lost at temperatures above T m (Fig. 8B), in the presence of lipids a high helix content of the pore-forming fragment is conserved even at elevated temperatures (Fig. 8C). The inset in Fig. 8A shows the derivatives of the melting curves.
Intrinsic Emission Fluorescence-The temperature dependence of the intrinsic emission fluorescence of ColB and its pore-forming fragment in solution was also tested at scanning rates of 1°C/min from 25 to 85°C. ColB contains eight trypto-phan residues, five of which are located in the R/T domain. The temperature dependence of the tryptophan fluorescence of ColB at pH 7.0 reveals a nearly linear decrease in fluorescence intensity as a function of increasing temperature (not shown). However, the temperature-induced quenching also exhibits a sharp increase in intensity at around 50 -55°C, thereby matching T m of ColB peak A. The curves obtained during cooling to 25°C or during a second heating period did not show the stepwise increase in intensity. This is in agreement with the irreversible unfolding observed by DSC. In the presence of lipids, the same transition is observed. The tryptophans of the pore-forming fragments experience a temperature-induced quenching in fluorophore intensity in solution or in the presence of lipids. However, a stepwise change in fluorescence intensity is absent during the melting process (not shown).

Structural and Functional Characteristics of Colicin B and
Related Colicins-The high resolution x-ray structures of several channel-forming domains of colicins (A, N, E1, Ia) exhibit a 10-helix arrangement with 2 hydrophobic helices being surrounded by two layers of hydrophilic and amphipathic helices (19 -22). These conformations, therefore, represent water-soluble forms of the proteins. The primary sequences of the Cterminal parts of colicins A and N exhibit Ն54% identity and 71 or 62% homology with ColB PFF, respectively. In addition, both the high resolution x-ray structure of ColA PFF (19) and the ColB PFF CD spectra (Fig. 8B) exhibit Ն75% ␣-helix contributions. It is, therefore, reasonable to assume that the colicin A and B pore-forming domains adopt closely related structures.  In contrast, the length and amino acid composition of the pore-forming regions of colicins A, B, and N exhibit considerable differences when compared with those of colicins Ia, Ib, and E1 (e.g. Ref. 16). Therefore, these C-terminal fragments are assigned to different families. However, when the receptor and translocation mechanisms of colicins are compared with each other, the molecules group in a different manner. Although group A colicins (including colicin A and ColN) use the TolA system for translocation, group B colicins (including ColB and ColIa) parasitize the Ton B machinery (6). Despite this functional similarity between ColB and ColIa, the primary sequence homology is modest and the helix content of ColB (Fig. 7B) is considerably lower than that of colicin Ia, in particular when considering the R/T domains alone (22).
Structural details, such as the location of the Ton recognition sequence in ColIa (22), or the distribution of hydrophilic and hydrophobic residues within the pore-forming domains of colicins indicate that upon receptor-binding, translocation and membrane-insertion conformational changes occur. For example, a multitude of investigations show that the helical conformation of ColA PFF is preserved during pH-dependent membrane insertion when at the same time refolding of the tertiary structure occurs (23).
Thermal Transitions of Colicin B-The thermal melting characteristics of colicin B and its PFF provide valuable information about the conformational stability and the interactions between the domains also when high resolution structures have not been obtained. The heats of transition of colicin B and its thermolytic fragments in solution are Յ3.6 and Յ1.3 cal/g, respectively, and, therefore, depending on the experimental conditions, are smaller or within the range previously observed for globular proteins in solution at comparable transition temperatures (11,24).
The DSC traces of colicin B exhibit two well-separated transitions ( Fig. 1) suggesting denaturation of two discrete cooperative units. The R/T domain is known to account for 61% of the total length of colicin B, which corresponds closely to the 68% value obtained for the heat of transition A when averaged between pH 7 and 4. Comparison with the DSC traces obtained from the isolated PFF provides further support for this assignment. Although the transition temperatures of the isolated PFF are considerably higher when compared with peak B of full-length ColB, this difference is largely annihilated after transition A has been uncoupled from transition B by incubation at intermediate temperatures (Fig. 3). In addition, the size and pH dependence of the heat of transition (⌬H) of peak B of ColB parallel those of the main transition observed for the isolated pore-forming fragment (Tables I and II). The large reduction in heat of transition of both peak B and the ColB pore-forming fragment during membrane association of the corresponding proteins further demonstrate that transition B is closely associated with the pore-forming domain of fulllength ColB (Fig. 6, B-D).
The CD spectra of colicin B and its pore-forming fragments indicate a high content of helical secondary structures of the C-terminal domain at room temperature (Figs. 7 and 8). This is in excellent agreement with previous structural findings for the colicin A pore-forming domain (19,25) as well as the high degree of primary sequence homology between the pore-forming fragments of ColA and ColB (16). The calorimetric, fluorescence, and circular dichroism studies presented in this paper demonstrate that the two reported transitions of ColB are due  to denaturation of two discrete domains of the protein. Far UV-CD spectra indicate that during this process the ColB-poreforming fragment in solution loses most of its secondary structure (Fig. 8B). This is in contrast to the colicin A pore-forming domain, which at 90°C and at pH 5 retains a considerably higher proportion of secondary structure in CD spectra (9). Three out of eight tryptophans of colicin B are located within the pore-forming fragment of colicin B at positions analogous to those in colicin A. Structural comparison, therefore, strongly suggests that at ambient temperature the tryptophan side chains at positions 395, 439, and 449 are buried within the hydrophobic interior of the protein (19,25). A continuous decrease in fluorescence intensity is observed upon heating, which partly persist when the sample is equilibrated back to room temperature (not shown). This indicates a pronounced change in environment. The remaining five Trp residues in ColB are located in the R/T domain. At the transition temperature a stepwise increase in the fluorescence intensity is observed thereby paralleling the DSC thermal transition (not shown).
pH Effects on the Thermal Stability of Colicin B-At pH values above the pI of full-length ColB (pI ϭ 4.65), the protein conserves its thermal stability within a wide pH range. However, the temperatures of both ColB transitions are lowered when the pH of the samples is decreased even before the pH-induced molten globule state of the ColA PFF is reached (26). The pH dependence of the R/T domain (peak A) is more pronounced when compared with peak B, therefore, upon acidification an increased resolution of both transitions is observed. At the same time the heat of transition of peak B increases and the one associated with peak A decreases (Fig. 2). The decrease in transition temperatures at constant ⌬H total indicates that entropic changes during the thermal transition increase when the pH is lowered (11). Models for changes in entropy during thermal transitions include alterations of the side-chain and backbone conformational freedom as well as changes in the hydration shell of macromolecules (11,24). Augmentation in ⌬H, as observed for peak B, has been taken as an indication of increased exposure of apolar residues during the thermal transition (11). Correspondingly, the decrease in ⌬H A suggests an increasing amount of solvent-exposed hydrophilic areas and/or augmentation of protonation reactions at low pH when compared with neutral conditions (11). The observed proton exchange during the unfolding of ColB below its pI, indeed, indicates that residues buried in the native structure are protonated during the ColB thermal transition (27). Similarly, pH-dependent changes in the thermodynamic parameters have been observed with other soluble proteins (1,28,29).
The decreased thermal stability of ColB at lowered pH correlates with the previously observed requirement of acidic conditions for optimal channel and molten globule formation of ColB (30), increased membrane association of ColA (31,32), augmented membrane-insertion rate of ColIa (33,34), and pH-sensitive conformational changes of ColE1 in the presence of membranes (35). The unfolding profile of diphtheria toxin has been published previously (36). This protein shares similarities with membrane-active colicin B such as acidic pHtriggered membrane insertion or the secondary and tertiary structure of the membrane-inserting domain (37). Diphtheria toxin also exhibits two transitions, which have been assigned to the A and B fragment (at pH 8, T m ϭ 54.5 and 58.4°C, respectively) (36). In a similar manner the transition temperatures and transition enthalpies exhibit a strong pH dependence, and the isolated A fragment exhibits a 6.5-10°C reduction in T m when compared with the full-length protein.
Thermal Transitions of the Isolated Pore-forming Domain of Colicin B-Investigation of the isolated ColB PFF indicates that only the heat of transition is pH-dependent, whereas the transition temperature remains unaffected between pH 3.5 and 7. In a similar manner, between pH 5 and 7 the heat of transition of ColA PFF exhibits a small tendency to decrease with increasing pH, at transition temperatures similar to those of ColB PFF (9). At pH Ͻ 5, however, both parameters of ColA PFF strongly decrease with pH. The ColB PFF-melting characteristics, therefore, do not provide direct indications of conformational instabilities that might be important during membrane insertion. It should be kept in mind, however, that the slow time scale of the DSC experiment is not sufficiently sensitive to detect the increased membrane insertion kinetics at acidic pH of ColA PFF or ColB PFF (26,38). Nevertheless, reduced numbers of negative charges at the surface of the protein might be important for low pH interactions with the membrane surface also of this pore-forming domain, in particular when these contain acidic phospholipids (39,40). Localized electrostatic interactions might help to increase membrane association and to orient the protein with respect to the lipid bilayer. Such interactions thereby open directed avenues for molten globule formation and membrane insertion (26,38,41).
Interactions between R/T and C Domains-At low pH the transition of the isolated ColB PFF occurs at considerably higher temperatures than transition B of full-length ColB (Tables I and II). Preincubation at temperatures between the transition temperatures of peaks A and B, however, shifts T m of peak B from 40 to 62°C, thereby closely approaching the transition temperature of ColB PFF (Figs. 1D and 3). Similarly, interaction of the PFF with lipid bilayers at low pH results in an increase of T m of peak A by Ͼ20°C (Figs. 1D and 6B). These data indicate that interaction between domains results in a reduction of the stability of both cooperative units. The decrease in conformational entropy when two protein domains are tightly coupled to each other is expected to contribute to the destabilization of large protein domains, but more specific interactions between the proteins, such as van der Waals, hydrogen bonding, or electrostatic interactions, can also be important. In a related manner, functional differences in ColA PFF and colicin A have been attributed to interactions between the pore-forming and the receptor binding domains (32,42).
Although the isolated C-terminal fragment by itself is competent to insert into membranes (16), and pore-forming domains of other colicins exhibit pore-forming activities in vitro (17,(43)(44)(45), all three domains are required for recognition, translocation, and pore formation at intact cells in vivo. It appears that mutual interactions between different regions of colicin B help to maintain the translocation and the poreforming domains in less stable but functionally competent conformations. Intra-and intermolecular interactions thereby reduce the energies involved in conformational changes during protein translocation and insertion (cf. e.g. Ref. 22).
Thermal Stability of Colicins in the Presence of Lipids-At pH 3.5 the denaturation profile of ColB shows a decrease in total heat of transition, disappearance of transition B, and stabilization of transition A due to the presence of lipids. Similarly, the heat of transition of membrane-associated ColB PFF is decreased by two orders of magnitude. In addition, CD spectra indicate that a large fraction of its secondary structure is preserved at higher temperatures. These observations are similar to those reported for the pore-forming domains of ColA (9), ColE1 (10), as well as the pore-forming proteins equinatoxin II (46), staphylococcal ␣-toxin (12), diphtheria toxin (36), or other transmembrane proteins that appear to be resistant to complete thermal denaturation (11). The hydrogen bonds of helices within hydrophobic environments are considerably stronger than in water (4), therefore, the helical structures of residues buried inside the lipid bilayer are often remarkably stable. In contrast, the least stable regions of membrane proteins are generally those that are extramembranous. As a consequence, these regions seem to unfold in a similar manner than soluble proteins (11).
Proton-decoupled 15 N solid-state NMR spectra of oriented membrane samples indicate that the helices of the ColB and ColE1 C-terminal domains are predominantly orientated parallel to the bilayer surface (16). In addition, neutron scattering experiments show a location of large parts of the ColA PFF close to and beyond the membrane surface (47). We, therefore, suggest that the small but detectable decrease in ␣-helix content and the heat of transition observed by calorimetric measurements of the ColB pore-forming fragment in the presence of lipids (Figs. 6 and 7) is associated with unfolding of loop structures, hydrophilic ends of membrane-associated helices, and/or complete unfolding of those helices that are weakly associated with the membrane interface (Ref. 11 and references cited therein).
The residual change in heat of transition observed in denaturation profiles of lipid-associated C-terminal domains of ColB (this study), ColA (9), and ColE1 (10), therefore, strongly suggest that a fraction of the membrane-inserted protein is located in extramembranous or interfacial regions. The reproducible fine structure encompassing at least four peaks in the DSC traces of ColB PFF is to our knowledge observed for the first time for membrane-associated proteins and could derive from, for example, independent helical structures at the membrane interface (Fig. 6, C and D).
In summary, this study provides thermal denaturation profiles of proteins in solution and in their membrane-associated states. The same protein reveals pronounced differences in stability and thermodynamic properties when investigated in solution or in the membrane. The membrane-inserted parts of the proteins seem to a large extent resistant to thermal denaturation and exhibits an interesting fine structure in DSC experiments. In contrast, those regions exposed to the aqueous environment lose much of their native conformation. The domain structure of colicin B is also reflected in two well-separated thermal transitions in aqueous buffer. Nevertheless, at low pH the interaction between these domains results in their mutual destabilization and transition temperatures well below those of the domains when uncoupled from each other. Intramolecular interactions may, therefore, play an important role during conformational changes that are associated with translocation and insertion into the inner E. coli membrane.