INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The engineering of ion channels and pores in order to mimic and improve upon the structures found in nature could have applications in many areas of biotechnology (Bayley, 1997, 1999), including cell permeabilization for cryopreservation (Eroglu et al., 2000), the development of cytotoxic agents (Pederzolli et al., 1995; Al-yahyaee and Ellar, 1996; Panchal et al., 1996), and the construction of biosensors (Ziegler and Göpel, 1998; Bayley, 1999; Bayley et al., 2000). -Hemolysin (HL), an exotoxin secreted by the bacterium Staphylococcus aureus (Gouaux, 1998), is a particularly attractive target for protein engineering. The HL pore is a heptamer made of identical subunits of 293 amino acids. Roughly globular molecules with molecular masses of up to ~2000 Da (Füssle et al., 1981), or larger elongated polymers such as single-stranded nucleic acids (Kasianowicz et al., 1996), can pass through a wide channel centered on the molecular sevenfold axis (Fig. 1 A).

View larger version (31K): [in this window] [in a new window]   FIGURE 1   Representations of -hemolysin (HL) and -cyclodextrin (CD). (A) Sagittal section through the WT-HL pore showing CD lodged in the lumen of the channel. The location is based on mutagenesis data (Gu et al., 1999; (B) structure of CD.

Earlier engineering studies focused on manipulating the assembly of the pore from monomers. It proved possible to control assembly with chemical, biochemical, and physical stimuli (e.g., divalent metal ions (Walker et al., 1995), proteases (Panchal et al., 1996), and light (Chang et al., 1995)). More recently, guided by the availability of a crystal structure of the heptamer (Song et al., 1996), greater attention has been given to manipulating the properties of the fully assembled pore. For example, binding sites formed by mutagenesis have been placed in the lumen of heteromeric pores to yield components for metal ion biosensors (Braha et al., 1997); ionic currents flowing through individual engineered pores are modulated by metal ion analytes, and the signal reveals both the concentration and identity of the ions. Single molecule detection of this kind is termed stochastic sensing (Braha et al., 1997; Bayley et al., 2000). Recently, targeted covalent modification has been used to place polymers in the lumen of the pore, increasing the variety of potential biosensor elements (Howorka et al., 2000).

We have also established a new stochastic sensing system by using HL equipped with noncovalent cyclodextrin adapters to mediate the sensing of organic molecules (Gu et al., 1999) (Fig. 1, A and B). Cyclodextrins are well-known to encapsulate organic molecules in aqueous solution and have been widely used in the food and pharmaceutical industries for this purpose (D'Souza and Lipkowitz, 1998). -Cyclodextrin (CD) is a cyclic molecule with a hydrophobic cavity, comprising seven D-glucose units. The primary 6-hydroxyl groups are on the narrower rim of the molecule, and the secondary 2- and 3-hydroxyl groups on the wider rim (Fig. 1 B). When CD or other cyclodextrins are lodged in the lumen of the HL pore, they alter the unitary conductance (Gu et al., 1999), change the ionic selectivity (Gu et al., 2000), and act as binding sites for channel blockers by contributing their host properties (Gu et al., 1999). The blocker sites provide for the detection of a wide variety of organic molecules.

Further studies of the interactions of cyclodextrins with the HL pore will advance the engineering of ion channels and aid the construction of biosensors. In particular, it is important to maximize the dwell time of cyclodextrins within the lumen of the pore. Here, we describe the sidedness, voltage-dependence, and pH-dependence of the interaction of CD with the wild-type HL pore. CD binds from the trans side of the lipid bilayer, with an apparent dissociation constant (K_d^app) that varies over a range of >100-fold.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Reagents

-Cyclodextrin (CD) was from Aldrich (Milwaukee, WI). Buffers for planar bilayer recording contained 1 M NaCl, 10 mM sodium phosphate, and 2 mM citric acid (Sigma, St. Louis, MO) in deionized water (Millipore Corp., Bedford, MA), and were titrated to pH 5.0 or 7.5 with 1 M HCl (EMScience, Gibbstown, NJ), and to pH 9.0 and pH 11.0 with 2 M NaOH (EMScience).

Protein

Heptameric WT-HL was formed by treating monomeric HL, purified from Staphylococcus aureus, with deoxycholate (Bhakdi et al., 1981; Walker et al., 1992) and isolated from SDS-polyacrylamide gels as described (Braha et al., 1997).

Bilayer recordings

A 25-µm-thick Teflon film (Goodfellow, Malvern, MA) with a 100-150-µm diameter orifice was used as a partition between the two chambers (2 ml each) of a Teflon bilayer apparatus. The orifice was pretreated with 1:10 hexadecane (Aldrich)/pentane (HPLC-grade, Burdick and Jackson, Muskegon, MI). A solvent-free planar lipid bilayer membrane of 1,2-diphytanoyl-sn-glycero-phosphocholine (Avanti Polar Lipids, Inc., Alabaster, AL) was formed across the orifice (Montal and Mueller, 1972; Hanke and Schlue, 1993). The transmembrane potential was applied with Ag/AgCl electrodes with 1.5% agarose bridges (Ultra Pure DNA Grade, Bio-Rad Laboratories, Hercules, CA) containing 3 M KCl (Sigma). Protein was added to the cis chamber, which was at ground. A positive potential indicates a higher potential in the trans chamber, and a positive current is one in which cations flow from the trans to the cis side. Experiments were conducted at 22 ± 2°C.

Single-channel current recordings were obtained by using an Axopatch 200B patch-clamp amplifier (Axon Instruments, Inc., Foster City, CA), in the whole-cell ( = 1) mode, with a CV-203BU headstage, and filtered at 5 kHz with a built-in low-pass Bessel filter. The data were acquired by computer at a sampling rate of 20 kHz by using a Digidata 1200 A/D converter (Axon) and Clampex 7.0 software (Axon). The data were analyzed with the software pClamp 6.03 (Axon) and Origin (Microcal Software Inc., Northampton, MA). Conductance values were determined by fitting the peaks in amplitude histograms to Gaussian functions. The mean dwell time  at each conductance level was measured by fitting the dwell time distribution to an exponential function.

Experiments were initiated by the addition of heptameric HL to the cis compartment of the bilayer apparatus to a final concentration of 3-30 ng ml¹, with stirring until a single channel inserted into the membrane. CD was added to the trans chamber at 40 µM unless otherwise specified. For the determination of each set of kinetic constants, three or more separate experiments were performed and data acquired for at least 2 min were analyzed. Values for unitary conductance, k_on^app, k_off, and K_d^app are quoted as the mean ± SD.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Sidedness of current block by CD

As previously documented (Menestrina, 1986; Bezrukov and Kasianowicz, 1993; Korchev et al., 1995), except at extremes of pH and transmembrane potential, WT-HL pores exhibited uniform single-channel conductance states of long duration at both negative and positive transmembrane potentials (Table 1 and Fig. 2 A, left, 651 ± 4 pS, 40 mV, pH 7.5, 1 M NaCl; right, 721 ± 6 pS, +40 mV, pH 7.5, 1 M NaCl). The addition of 40 µM CD to the trans compartment produced reversible partial blockades of the ionic current (see below). By contrast, no current block was detected when 40 or 80 µM CD was added from the cis side (Fig. 2 B), even at the highest voltages (±120 mV) and lowest pH values (pH = 3.0) tested (data not shown).

View this table: [in this window] [in a new window]   TABLE 1   Conductance values and kinetic parameters for HL and HL · CD at ±40 mV

View larger version (19K): [in this window] [in a new window]   FIGURE 2   Representative current traces from single HL pores showing blockades by CD. All traces were recorded under symmetrical pH conditions in buffer containing 1 M NaCl; 40 µM CD was added where indicated. Left, traces recorded at 40 mV; right, traces recorded at +40 mV. The broken line indicates zero current. (A) HL alone, pH 7.5; (B) HL with CD, cis, pH 7.5; (C) CD, trans, pH 5.0; (D) CD, trans, pH 7.5; (E) CD, trans, pH 11.0.

pH- and voltage-dependence of the extent of current block by CD

The reversible partial blockades of the single-channel currents, obtained by the addition of 40 µM CD to the trans side of a bilayer, were examined further. At pH 5.0 and 40 mV, the conductance was reduced from 683 ± 3 pS (HL) to 266 ± 7 pS, while CD was bound (HL · CD) (Table 1 and Fig. 2 C, left). At pH 5.0 and +40 mV, the conductance was reduced from 746 ± 12 pS (HL) to 256 ± 7 pS (HL · CD) (Fig. 2 C, right). At pH 7.5, blockades of similar amplitude were observed: at 40 mV, from 651 ± 4 pS (HL) to 240 ± 3 pS (HL · CD) (Fig. 2 D, left), and at +40 mV, from 721 ± 6 pS (HL) to 253 ± 4 pS (HL · CD) (Fig. 2 D, right). Again, at pH 11, the extent of block was similar: at 40 mV, from 606 ± 7 pS (HL) to 212 ± 15 pS (HL · CD) (Fig. 2 E, left), and at +40 mV, from 654 ± 3 pS (HL) to 196 ± 2 pS (HL · CD) (Fig. 2 E, right).

A typical amplitude histogram (Fig. 3, inset; recorded at 40 mV, pH 7.5, 1 M NaCl) shows only one current blockade level, which suggests that there is only one binding site for CD within the lumen of the HL pore. The kinetics of the interaction with CD are also in keeping with this interpretation (see below). The I-V (current versus voltage) curves for HL and HL · CD recorded in 1 M NaCl (Fig. 3) show that neither the conductance of HL nor that of HL · CD change dramatically over a wide range of pH values throughout the voltage range from 140 mV to +140 mV.

View larger version (23K): [in this window] [in a new window]   FIGURE 3   Reduction of the unitary conductance of the HL pore by CD. Single-channel I-V curves for HL with or without CD bound, under different symmetrical pH conditions, as indicated, in 1 M NaCl with 40 µM CD trans. Inset, Representative current amplitude histogram from a recording made in the presence of 40 µM CD trans, with 1 M NaCl at pH 7.5 in both chambers.

CD binds to HL at a single site in a simple bimolecular interaction

Histograms displaying the dwell time for the unoccupied state of HL (_on) and the dwell time for CD in the lumen of the pore (_off) could be fitted by single-exponential distributions for data obtained at 40 mV, pH 7.5, 1 M NaCl, and 40 µM CD (Fig. 4, A and B). Single time-constants for both _on and _off were also found at the other pH values examined (pH 5.0, pH 9.0 and pH 11.0) and throughout the voltage range 140 to +140 mV (data not shown). These data affirm that, under the wide range of conditions examined here, CD lodges at a single site in the lumen of the pore, which is accessible from the trans side of the bilayer. Therefore, the kinetics of the interaction between CD and HL should obey the simple kinetic scheme:

(1)

As expected from Scheme 1, the plot of 1/_on versus [CD] was a straight line, the slope of which yields the rate constant k_on^app (e.g., Fig. 4 C). We use k_on^app (rather than k_on) because certain models for the pH- and voltage-dependence of binding demand more than one state of HL. Again, as expected from Scheme 1, 1/_off is independent of [CD] (e.g., Fig. 4 D) and k_off was evaluated as the average of 1/_off at the different CD concentrations.

View larger version (39K): [in this window] [in a new window]   FIGURE 4   Kinetics of the interaction of CD with HL at 40 mV. Dwell time histograms of HL (A) and HL · CD (B) were fitted to single-exponential functions. Data were from a recording made with 40 µM CD trans, in 1 M NaCl at pH 7.5 in both chambers. (C) The rate constant konapp was obtained from the slope of the linear fit of 1/on versus [CD]. (D) The rate constant koff is independent of [CD] and was obtained from the average of the 1/off values.

pH- and voltage-dependency of block kinetics

The rate constants, k_on^app and k_off, were determined over a wide range of conditions (Fig. 5, A and B). For example, at pH 5.0, at negative transmembrane potentials, the dwell time (_off) of CD in the lumen of the pore (e.g., at 40 mV, _off = 3.1 ± 0.1 ms, k_off = 3.2 ± 0.1 × 10² s¹, Fig. 2 C, left) was longer than _off at positive potentials (e.g., at +40 mV, _off = 1.4 ± 0.2 ms, k_off = 7.0 ± 0.7 × 10² s¹, Fig. 2 C, right). By contrast, at pH 5.0, _on was shorter at negative potentials (at 40 mV and 40 µM CD, _on = 280 ± 20 ms, k_on^app = 9.0 ± 0.5 × 10⁴ M¹ s¹) than at positive potentials (at +40 mV and 40 µM CD, _on = 500 ± 20 ms, k_on^app = 5.0 ± 0.2 × 10⁴ M¹ s¹). At pH 11.0, the trends were the opposite of those observed at pH 5.0. _off was considerably shorter at negative transmembrane potentials (e.g., at 40 mV, _off = 0.37 ± 0.03 ms, k_off = 2.7 ± 0.2 × 10³ s¹, Fig. 2 E, left) than at positive potentials (e.g., at +40 mV, _off = 0.87 ± 0.04 ms, k_off = 1.2 ± 0.1 × 10³ s¹, Fig. 2 E, right). At pH 11.0, _on was longer at negative potentials (at 40 mV and 40 µM CD, _on = 190 ± 10 ms, k_on^app = 1.3 ± 0.1 × 10⁵ M¹ s¹) than at positive potentials (at +40 mV and 40 µM CD, _on = 57 ± 2 ms, k_on^app = 4.4 ± 0.2 × 10⁵ M¹ s¹).

View larger version (29K): [in this window] [in a new window]   FIGURE 5   Dependence of the kinetic constants for the interaction of CD with HL on pH and voltage. (A) koff; (B) konapp; (C) Kdapp. The dissociation constant Kdapp was calculated as koff/konapp.

Equilibrium dissociation constants (K_d^app = k_off/k_on^app) were calculated for HL · CD and differ by over 100-fold under the conditions examined (Fig. 5 C). The most extreme values are seen at pH 5.0 and pH 11.0. For example, at 120 mV, CD binds 87 times more strongly at pH 5.0 (K_d^app = 1.6 ± 0.1 × 10³ M) than at pH 11.0 (K_d^app = 1.4 ± 0.1 × 10¹ M). By contrast, at +120 mV, CD binds 150 times more strongly at pH 11.0 (K_d^app = 7.7 ± 0.3 × 10⁴ M) than at pH 5.0 (K_d^app = 1.1 ± 0.1 × 10¹ M). Interestingly, when extrapolated to a transmembrane potential of 0 mV, the K_d^app values for CD show little variation over the range pH 5.0 to pH 11.0 (Fig. 5 C). However, it is important to note that changes in both k_on^app and k_off occur with pH (Fig. 5, A and B). For example, while the K_d^app value for CD at 0 mV at pH 5.0 is similar to the K_d^app value at pH 7.5, compensating changes occur in the on- and off-rate constants.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

CD binds at a specific site within the lumen of the HL pore

When CD enters the lumen of the HL pore, it reduces the single-channel conductance to 33-43% of the value in the absence of CD. The exact value of the reduced conductance varies with the transmembrane potential and the pH of the buffer in the chambers (Fig. 3). Several arguments support the notion that CD reduces the conductance by binding at a single site within the channel lumen in a defined orientation and conformation.

First, at each pH value and voltage condition, there is only one conductance state that can be assigned to HL · CD (Fig. 3). Second, there is no additional noise associated with the HL · CD state, suggesting that the rather rigid cyclodextrin (Corradini et al., 1996; Schneider et al., 1998; Saenger et al., 1998) is firmly held at the binding site (data not shown, but compare peak widths in Fig. 3 B, inset). Third, at all pH values and voltage conditions tested, we find that dwell time histograms for _off and _on can be analyzed as single exponentials (Fig. 4, A and B), consistent with simple bimolecular kinetics for the interaction of CD with HL (Scheme 1). Fourth, the additional channel block produced when guest molecules bind to CD lodged in the pore (Gu et al., 1999) suggests that CD is bound in an orientation in which its interior is exposed to solvent, rather than (for example) being occupied by an amino acid side chain. The residual current observed in the presence of a guest may represent ions flowing between the outer surface of the cyclodextrin ring and the HL wall (Gu et al., 1999). Finally, the idea of a discrete binding site for CD within the lumen of the HL pore is strengthened by the existence of mutants with enhanced affinity for CD and other cyclodextrins (Gu et al., 1999, 2000). For example, in the case of M113N, the K_d^app is decreased from 3.6 × 10³ M to 1.5 × 10⁷ M at pH 7.5, 40 mV. Thus, the situation with cyclodextrin adapters differs from that with flexible polymers such as PEG or polynucleotides, which sustain brief encounters with the lumen or rapidly pass through it without lodging at a defined site (Kasianowicz et al., 1996; Bezrukov, 2000).

The binding site for CD can be accessed from the trans but not the cis side of the bilayer

CD can reach its binding site in the lumen of the HL pore from the trans side of the membrane (Fig. 2, C-E), but it has no effect when applied from the cis side (Fig. 2 B), even at extremes of pH and transmembrane potential. The cis entrance to the lumen is wider than the trans entrance (Fig. 1), so binding from the cis side must be hindered by an internal barrier (Merzlyak et al., 1999), which we suggest is formed by the rings of Glu-111 and Lys-147 side chains (Song et al., 1996). Interestingly, cis binding events of CD can be detected with the mutant E111N/M113N/K147N, in which the internal barrier is removed, and the dwell times (_off) are very similar whether CD binds from the trans or the cis entrance, again suggesting a single site (L.-Q. Gu and S. Cheley, unpublished observations).

Voltage- and pH-dependence of the interaction of CD with the HL pore: ruling out familiar mechanisms

Woodhull's mechanism for a charged blocker

(2)

Voltage-dependent two-state conformational change of the protein

(3)

A viable model for the voltage- and pH-dependence of the interaction of CD with the HL pore

A third model involves a continuous change in the free energy of HL or HL · CD (or both) as a function of the membrane potential. For example, CD might bind to HL in a single step by induced fit (Koshland et al., 1966), rather than through one of two states required by the second model. In a simplified version of the third model, the free energy of HL · CD remains unaltered in an applied field, while the free energy of HL varies as G = z_p F  V (Fig. 6). The dependence of K_d on voltage is then given by a relation with the same form as Eq. 3:

(4)

Despite the similarity between Eqs. 3 and 4, it is clear that the free energy barrier to the transition state (G) varies in model 3 (Fig. 6), with a dependence on voltage that would account qualitatively for the variation in k_on and k_off (Fig. 5). The use of z_p and  imply that a localized charge moves as CD binds. The same result would obtain if, when CD binds, there were a change in the dipole moment of the protein with contributions from several regions of the molecule (Moczydlowski, 1986).

View larger version (13K): [in this window] [in a new window]   FIGURE 6   The pH- and voltage-dependent interaction of CD with HL depicted as conceptual free energy profiles. In this simplified rendition, the occupied form of the protein, HL · CD, is supposed to be unaffected by voltage. The free energy of the unoccupied form, HL, is voltage-dependent. For example, at low pH a positive charge on the pore might be located part way through the lipid bilayer in the unoccupied form of the protein. Thus, a negative potential would increase the free energy of the unoccupied protein relative to the occupied form and the binding of CD would be favored.

At pH 5.0 and pH 11.0, and at low potentials, plots of log K_d^app versus V (Fig. 5 C) can be fitted to straight lines as predicted by Eq. 4, yielding z_p values of +0.60 and 0.61, respectively. These data imply that at low pH, positive charge on the protein moves toward the trans side of the bilayer as CD binds; hence, binding is favored at negative potentials. By contrast, at high pH, negative charge must move toward the trans side when CD binds. At intermediate pH values and at extremes of pH, plots of log K_d^app versus V are not linear and a more complex model would be required to fit the data, which is not warranted given our present knowledge of the system.

The third model implies that the conformational change of HL that occurs when CD binds (Fig. 6) does not occur in the absence of CD, i.e., that the conformationally altered state of the protein is separated by an insurmountable free energy barrier in the absence of CD. Alternatively, the conformationally altered, but unoccupied, state and intermediate conformational states might be energetically accessible, in which case the structure of the protein would gradually change with membrane potential to forms with increased or decreased affinity for CD. This effect of voltage would also satisfy Eq. 4 and the experimental observations on k_off. Model 2 is an extreme version of this last case in which there is not a gradual change in structure, but two states separated by a surmountable barrier.

pH-dependence of the properties of the HL pore

Several properties of the -hemolysin pore are pH-dependent, including the unitary conductance, ion selectivity, and magnitude of single-channel noise (Menestrina, 1986; Bezrukov and Kasianowicz, 1993; Kasianowicz and Bezrukov, 1995; Krasilnikov et al., 1997). Most relevant to the present work is the finding that the interactions of neutral PEG molecules with the lumen of the transmembrane channel vary with pH; at lower pH values, the size of polymers that interact with the lumen is reduced (Bezrukov and Kasianowicz, 1997). There is not enough in common between these experiments and the present work to determine whether the same ionizable groups on the protein are involved in both cases. For example, the interactions of PEG with the channel lumen are weak compared with those of the cyclodextrins. Nevertheless, it is interesting that in both cases the kinetics of binding of a neutral molecule are affected.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

We have measured the interaction of the adapter molecule CD with WT-HL over a wide range of pH values and transmembrane potentials. When extrapolated to a transmembrane potential of 0 mV, the K_d^app values for CD show little variation over the range pH = 5.0-11.0. While this suggests that ionizable groups play no role in the interaction of HL with CD at 0 mV, the situation is in fact more complex, because compensating changes in k_on^app and k_off occur. K_d^app values for CD vary continuously with voltage. Because CD is a neutral molecule, this finding suggests that charge movement on the protein is associated with the binding event. Because k_off, the dissociation rate constant of CD from its binding site, also varies continuously with voltage, a simple two-state model, in which there are high- and low-affinity forms of HL, is ruled out. Instead, a model is suggested that features a single unoccupied state of the channel with a free energy relative to HL · CD that is voltage-dependent. For applications in sensors, it is important that the dwell time of an adapter within the HL pore be as prolonged as possible. The data presented here show that the dwell time can be manipulated with pH and voltage. These tactics might be combined with increases in dwell time obtained by mutagenesis (L.-Q. Gu and S. Cheley, unpublished data) to improve the characteristics of HL pores as sensor elements.