INTRODUCTION

Top Abstract Introduction Methods Results and discussion References

Protein ion channels perform a wide variety of functions in cells. By virtue of their ability to respond to various inputs, including the concentration of specific analytes (e.g., neurotransmitters; Hille, 1992), channels might prove useful as sensors for water-soluble compounds. This could be accomplished by placing binding sites for the analyte of choice in the channel's pore. If the analyte bound reversibly to the site(s) and caused a well-defined conductance change, its concentration could be inferred from the mean conductance of a fixed number of channels or from the mean time a single channel spends in the analyte-bound conductance state.

In an effort to rationally design and construct sensors for specific chemicals in solution, we are genetically engineering various types of analyte binding sites into the channel formed by Staphylococcus aureus -hemolysin (HL) (Walker et al., 1994; Kasianowicz et al., 1994; Braha et al., 1997). -Hemolysin has several characteristics that make it particularly attractive for this purpose. Although the 293-residue monomer (Gray and Kehoe, 1984; Walker et al., 1992) is water-soluble, it binds spontaneously to a variety of lipid membranes, and self-assembles to form heptameric pores from identical monomers (Gouaux et al. 1994; Song et al., 1996). Since the experimentally determined narrowest segment of the channel's pore diameter is ~2 nm (Krasilnikov et al., 1992; Korchev et al., 1995; Kasianowicz et al., 1996; Bezrukov et al., 1996; Song et al., 1996; Bezrukov and Kasianowicz, 1997), the pore is large enough to accommodate engineered analyte binding sites. However, the pore is sufficiently small that the binding of an analyte to a receptor site inside the pore will cause marked changes in the channel's current. In addition, unlike many voltage-gated channels, the HL channel remains in the fully open state for long periods of time (Kasianowicz et al., 1996).

Wild-type HL (WT-HL) has been shown to sense different ions in solution and to determine their concentration. For example, the reversible binding of hydrogen or deuterium ions to the channel causes rapid current fluctuations (Bezrukov and Kasianowicz, 1993; Kasianowicz and Bezrukov, 1995). The pH dependence of the current spectral density allowed the determination of the rate constants and the equilibrium constants for these reactions as well. Those studies provided the analytical basis for the use of the HL channel in sensing applications.

We previously demonstrated that HL can be genetically engineered to detect transition metal divalent cations. The current carried by the mutant HL-H5, in which five consecutive amino acids in a glycine rich region of the protein (residues 130-134, inclusive) were replaced with histidines, was markedly blocked by the addition of 100 µM Zn(II) added to either side of the membrane (Walker et al., 1994; Kasianowicz et al., 1994). The work reported here extends those studies, identifies several residues that are near one mouth of the pore, and suggests that the pore's secondary structural motif is -sheet. These results are in good agreement with the recently determined crystal structure of WT-HL oligomerized in detergent (Song et al., 1996).

	METHODS

Top Abstract Introduction Methods Results and discussion References

Modified region of the primary sequence

Table 1 illustrates the segment of the WT-HL primary structure that was modified to include binding sites for divalent cations and illustrates the notation used herein to describe the mutations.

View this table: [in this window] [in a new window]   TABLE 1   A comparison of the primary sequences of wild type and mutant versions of HL used in this study

Protein Expression, Purification, and Assay

Histidine scanning mutagenesis

Cassette mutagenesis

For further mutagenesis of the DNA encoding the central region of the HL polypeptide, eight restriction endonuclease sites (HpaI, BsiWI, BstEII, SpeI, StuI, ApaI, AvrII, and AflII) were introduced into pT7-NPH8S (K8A). These sites do not occur elsewhere in the vector or insert. Four of these sites (HpaI, BsiWI, BstEII, and AflII), representing sequences between amino acid residues 114 and 148, were introduced by PCR of 5' and 3' segments of the HL gene using the following primers: HB203 (5' half; sense primer) CGGGATCCTAATACGACTCACTATAGGG, HB497 (5' half; antisense primer) CCAGTAAGGTTACCGTTGAATCCGTACGTTAACGTACTCATATACTC, HB498 (3' half; sense primer) GGTGCAAAGGTAACCCTAGGTCATACACTTAAGTATGTTCAACC, HB172 (3' half; antisense primer) AAACATCATTTCTGAAGCTTTCGGCTAAAG. The PCR product representing the 5' segment of the HL gene was digested with NdeI and BstEII and that representing the 3' segment with BstEII and HindIII. Subsequent three-way ligation with the vector PT7-SMC (digested with NdeI and HindIII (Cheley et al., 1997)) yielded an HL construct with a deletion in the central region, which was then digested with BstEII and AflII. Ligation of the digested vector to an oligonucleotide cassette flanked by the same sites resulted in a full-length re-engineered HL. The following oligonucleotides were used to generate the double-stranded synthetic cassette: HB499 (upper strand) GTAACCTTACTGGTGATGATACTAGTAAAATTGGAGGCCTTATTGGGGCCCAGGTTTCCCTAGGTCATACAC, HB500 (lower strand) TTAAGTGTATGACCTAGGGAAACCTGGGCCCCAATAAGGCCTCCAATTTTACTAGTATCATCACCAGTAAG. DNA sequencing of the entire re-engineered HL gene revealed two base deletions and one base substitution in the synthetic insert. These errors were repaired by replacing the DNA between BstEII and StuI sites with a synthetic oligonucleotide: HB521 (upper strand) GTAACCTTACTGGTGATGATACTAGTAAAATTGGAGG, HB522 (lower strand) CCTCCAATTTTACTAGTATCATCACCAGTAAG. The corrections were confirmed by DNA sequencing. Of the eight new restriction endonuclease sites, four (BstEII, SpeI, ApaI, and AvrII) resulted in conservative amino acid replacements (V124L, G130S, N139Q, I142L) and four (HpaI, BsiWI, StuI, and AflII) were silent. The partly synthetic gene was designated HL-RL.

The HL-RL plasmid DNA was then used to create mutant GNN/HQ/H. BstEII and StuI-digested plasmid was ligated to a cassette composed of two phosphorylated oligonucleotides: HB510 (upper strand) GTAACCTTACTGGTAATAATACTCATCAGATTGGAGG, HB511 (lower strand) CCTCCAATCTGATGAGTATTATTACCAGTAGA. Mutant GNN/HQ/H plasmid DNA was then digested with AvrII and AflII, and ligated with the following phosphorylated oligonucleotides to create the mutant GNN/HQ/N: HB523 (upper strand) CTAGGTAATACAC, HB524 (lower strand) TTAAGTGTATTAC. Three other mutants were generated using BstEII and StuI-digested GNN/HQ/N plasmid DNA. These mutants and the oligonucleotides that were used to generate them are 1) DNN/HQ/N-HB563 (upper strand) GTAACCTTACTGATAATAATACTCATCAGATTGGAGG, HB566 (lower strand) CCTCCAATCTGATGAGTATTATTATCAGTAAG; 2) GDN/HQ/N-HB564 (upper strand) GTAACC- TTACTGGTGATAATACTCATCAGATTGGAGG, HB567 (lower strand) CCTCCAATCTGATGAGTATTATCACCAGTAAG; and 3) GND/HQ/N-HB565 (upper strand) GTAACCTTACTGGTAATGATACTCATCAGATT- GGAGG, HB568 (lower strand) CCTCCAATCTGATGAGTATCATTACCAGTAAG. Mutations were confirmed by DNA sequencing. No sequence differences were observed in HL DNA other than those present in the mutagenic oligonucleotides.

Protein purification

WT-HL from S. aureus Wood strain 46 (American Type Culture Collection), recombinant HL from the HL negative S. aureus strain DU1090 (Fairweather et al., 1983) and recombinant HL expressed in Escherichia coli strain JM109(DE3) (Promega, Madison, WI) were purified as described elsewhere (Cheley et al., 1997; Palmer et al., 1993).

In certain cases, as indicated in the text, mutant HL polypeptides were synthesized in vitro by coupled transcription and translation (IVTT), as described previously (Walker et al., 1992). Briefly, supercoiled plasmid DNA was used as the template in an E. coli S30 extract (Promega, No. L464) supplemented with rifampicin (20 µg/ml) and T7 RNA polymerase (2000 U/ml; New England Biolabs, Beverly, MA). To facilitate the monitoring of the purification, an IVTT reaction (25 µl), carried out in the presence of [³⁵S]methionine (400 µCi/ml, 1200 Ci/mmol; New England Nuclear Life Science Products, Boston, MA) was mixed with an unlabeled IVTT reaction (100 µl). Polyethylenimine (PEI, 10% aqueous solution), prepared as described elsewhere (Gegenheimer, 1990), was added to a final concentration of 0.2% and the mixture was then placed on ice for 10 min. The PEI-treated extract was centrifuged in a microfuge tube at 16,000 × g for 10 min at 4°C. A 50% slurry of SP Sephadex C50 cation exchanger (50 µl) that had been equilibrated in 10 mM Tris-HCl (pH 8) was added to the supernatant. After mixing, the suspension was applied to a 2-ml capacity 0.2 µm cellulose acetate microfilterfuge tube (Rainin) and centrifuged for 2 min at 16,000 × g. The filtrate was diluted 10-fold in 10 mM sodium acetate (pH 5.2). A 50% slurry of S-Sepharose FF cation exchanger (60 µl) that had been equilibrated with 10 mM sodium acetate (pH 5.2) was then added. The mixture was rotated gently for 1 h at 4°C, placed in another microfilterfuge tube, and centrifuged to remove unbound protein from the resin. The spin filter containing the ion exchange resin was then inserted into a fresh microfilterfuge tube. Bound HL was eluted from the resin by mixing it with 4 bed volumes of elution buffer (120 µl of 300 mM NaCl in 10 mM sodium acetate, pH 5.2) followed by centrifugation for 2 min. Recovery of the mutant HL protein was monitored by hemolytic assay (see below) as well as by autoradiography after electrophoresis in a 12% SDS-polyacrylamide gel.

Hemolytic assays in microtiter wells

After translation, the IVTT mix was centrifuged at 16,000 × g for 10 min and the supernatant was used immediately. For hemolysis assays in the presence of EDTA or metal ions, 10 µl of IVTT mix was diluted with Na MOPS-BSA (10 mM Na MOPS, 150 mM NaCl, pH 7.4 containing 1 mg/ml bovine serum albumin) buffer (70 µl) and subjected to twofold serial dilution with Na MOPS-BSA buffer (40 µl in each of 12 wells of a 96 well microtiter dish). EDTA (10 µl of 5 mM) or 1 mM metal sulfate (10 µl) was then added to the wells and lysis was measured for 1 h at 20°C after the addition of 1% washed rabbit erythrocytes (50 µl in Na MOPS-BSA) by monitoring the increase in light transmitted at 595 nm using a Bio-Rad microplate reader (Model 3550-UV).

Single channels in planar lipid bilayers

Solvent-free planar lipid bilayer membranes were formed from diphytanoyl phosphatidylcholine (DiPhyPC; Avanti Polar Lipids, Inc., Alabaster, AL), in high purity pentane (Burdick and Jackson, Muskegon, MI) on a 100-150 µm diameter orifice in a 17- or 25-µm-thick Teflon partition that separated two chambers (Montal and Mueller, 1972; Kasianowicz et al., 1996). The orifice was pretreated with a 10% (v/v) solution of purum grade hexadecane (Fluka, Buchs, Switzerland) in pentane.

The aqueous solutions bathing the membrane were prepared with 1 M KCl (Mallinckrodt, St. Louis, MO), 5 mM ultrol grade HEPES (Calbiochem, San Diego, CA), or 5 mM BioChemika micro select grade MOPS (Fluka) in deionized H₂O (Barnstead NANOpure II, Dubuque, Iowa) and titrated to pH 7.5 using 1 N NaOH (Fisher Scientific, Pittsburgh, PA). Trace heavy metal contaminants were subsequently removed by passing the solution through a 1.5-cm diameter glass column containing 10 ml Chelex-100 ion exchange resin (Bio-Rad, Hercules, CA). The bulk conductivity of aqueous solution samples was measured using a Radiometer Model 82 conductivity meter (Copenhagen, Denmark).

Single channels were formed by adding <1 µg of either WT-HL or mutant-HL to 1.8 ml of solution in the cis chamber while stirring. When a single channel appeared, the chamber was flushed extensively with fresh buffer to eliminate further channel incorporation.

Divalent cations were added to the cis or trans chambers by adding <20 µl total of concentrated stock solutions of the following reagents, which were used without further purification: zinc sulfate (Matheson, Coleman and Bell, Norwood, OH), cupric sulfate and nickel sulfate (Mallinckrodt, St. Louis, MO), calcium sulfate (Aldrich, Milwaukee, WI) and cobalt chloride (Fisher Scientific, Fairlawn, NJ). Although some of the mutant HLs had submicromolar binding constants for metal ions, as judged by the effect of Zn(II) on the single channel current, we restricted our study to the effects caused by divalent cations at concentrations between 1 µM and 1 mM. Where noted, excess EDTA (tri-sodium salt, Sigma, St. Louis, MO) titrated to pH 7.5 was added to the chambers to completely chelate the divalent cations. The pH of the solutions in each chamber at the end of an experiment varied by <0.15 from the initial reading, and the temperature was T = 23.0 ± 1.5°C.

The potential difference was applied across the bilayer with Ag-AgCl electrodes in 3 M KCl, 1.5% agarose bridges (Bethesda Research Laboratory, Gaithersburg, MD). The polarity of the potential is defined as positive when it is greater at the side of protein addition (cis). The current was converted to voltage and amplified by either a Dagan 3900A (Minneapolis, MN) or an Axopatch 200A (Foster City, CA) patch-clamp amplifier in the whole cell mode with a modified 3902 or a CV-201AU headstage, respectively. The signal was digitized using a SONY/Dagan/Unitrade DAS 75 data recorder onto DAT tape and subsequently transferred to a personal computer through a Frequency Devices model 902 or 9002 8-pole Bessel filter (Haverhill, MA), the corner frequency of which was set to < of the sampling frequency of a National Instruments AT-MIO-16X A/D board (Austin, TX). The membrane chamber and headstage were isolated from external electrical and magnetic noise sources by a mu-metal box (Amuneal Corp., Philadelphia, PA).

Under certain conditions, the binding of protons and divalent cations to WT-HL causes it to gate to much lower conductance states (the ratio of the final to the initial conductance is ~0.1) over time scales that range from 100 ms to seconds, depending on the concentration of each analyte and the magnitude of the applied potential (Menestrina, 1986; Korchev et al., 1995, Kasianowicz and Bezrukov, 1995). To avoid these cation-induced gating effects, we used a higher electrolyte concentration (1 M KCl), a relatively low applied potential (40 mV), relatively low concentrations of Zn(II) ([ZnSO₄]  1 mM), and a relatively high pH (pH 7.5), compared to those used previously (e.g., Menestrina, 1986).

	RESULTS AND DISCUSSION

Top Abstract Introduction Methods Results and discussion References

Hemolytic assay

We first screened the scanning histidine mutants T129H, G130H, K131H, I132H, G133H, and G134H for pore-forming activity as a function of Zn(II) concentration using the hemolysis assay described above. WT-HL lyses rabbit red blood cells (rRBCs) by forming large pores in the erythrocyte membrane. As the cells rupture, the light transmitted by the sample increases. In the absence of divalent cations, WT-HL (solid circles) and the point His mutants had closely similar lytic activities (Fig. 1). For visual clarity, the only mutant HL results that are shown are T129H (solid diamonds), G130H (solid triangles), and G134H (solid squares). Adding 100 µM Zn(II) had no effect on WT-HL's hemolytic activity (open circles), a slight effect on the lytic activity by T129H (open diamonds), or K131H (not shown) and a marked inhibitory effect on the lytic activity of G130H (open triangles), G134H (open squares), I132H (not shown), or G133H (not shown). The results of this experiment suggest that the His-mutants G130H, I132H, G133H, and G134H might be useful as sensing elements for metal ions.

View larger version (21K): [in this window] [in a new window]   FIGURE 1   Wild-type and histidine point mutants of HL form lytic pores. Zn(II) inhibits the hemolytic activity of several mutant proteins. The filled and open symbols indicate the absence and presence of Zn(II), respectively. Without Zn(II), WT-HL (circles), T129H (diamonds), G130H (triangles), and G134H (squares) caused 50% lysis of rabbit RBCs within 15 min. Adding Zn(II) had virtually no effect on the hemolytic activity of WT-HL but inhibited hemolysis by the point histidine mutants to varying degrees. Ionizable side chains are underscored.

Determining the Zn(II) binding site and structural implications

The hemolytic assay did not allow us to distinguish whether Zn(II) inhibited the assembly of the pore or occluded a fully formed channel. To address that question, and to identify which residues constitute the metal ion binding site, we studied the effects of divalent metal cations on the single-channel currents of one of the point His mutants, G130H, and the last five mutants listed in Table 1 (see above). We show below that by altering the number and location of ionizable side chains in select locations between 126 and 144, we determined the residues that comprise the binding site and confirmed part of the channel's proposed structure (Song et al., 1996).

In the absence of Zn(II), the single channel currents for WT-HL and each of these mutants were relatively noise free (e.g., see Fig. 3). In the presence of 1 µM Zn(II) in the trans compartment, the current carried by potassium and chloride ions through WT-HL is quiescent (Fig. 2, leftmost current trace). However, Zn(II) induced pronounced current fluctuations in three of the five mutant HLs studied here (Fig. 2, traces 2-6).

View larger version (11K): [in this window] [in a new window]   FIGURE 2   Single channel recordings of genetically engineered HL mutants define the Zn(II) binding site and part of the channel structure in planar lipid bilayers. The side chains of amino acids at sites 126, 127, 128, 130, 131, and 144 are listed above the recordings, and the Zn(II) concentration in the trans chamber was 1 µM. See text for details.

The first two recordings show that the single point mutation, Gly replaced by His at position 130 (G130H, i.e., GDD/HK/H), is sufficient to confer Zn(II) sensitivity to the channel. Crystallographic and other evidence suggest that the channel is composed of seven identical HL monomers (Gouaux et al., 1994; Song et al., 1996). Zn(II) can coordinate four ionizable residues. Therefore, it was conceivable that Zn(II) was coordinated by His-130 side chains on different monomers. The third current trace demonstrates that this conclusion was not correct. Specifically, a mutant HL with His-130 and with the four ionizable side chains immediately up and downstream of it replaced with nonionizable counterparts (GNN/HQ/N) showed no Zn(II)-induced current fluctuations. Therefore, the metal binding site consists of His-130 and one (or more) of the four ionizable amino acids that are proximal to that site (i.e., D127, D128, K131, and H144).

Zn(II)-induced current fluctuations reappeared when Asp-128 and His-130 were both present (GND/HQ/N, fourth trace). However, the fifth trace shows that the current fluctuations were virtually absent, if instead, the coordinating side chains were Asp-127 and His-130 only (GDN/HQ/N). These two results suggest that the binding site is primarily His-130 and Asp-128. We do not know whether the chelation occurs between these two residues within a monomer or between Asp-128 and His-130 on different monomers.

We determined the channel's structural motif in this segment of the polypeptide with an additional substitution. When the only ionizable side chain proximal to His-130 was Asp-126, the current fluctuations reappeared (DNN/HQ/N, sixth current trace). Interestingly, the current noise was even more pronounced than that observed with the other mutants reported here. Because His-130 only coordinates Zn(II) with ionizable side chains at residues 126 or 128, the minimum interval defining a binding site is two residues, not four. This suggests the secondary structure of the channel in this region is -sheet and not -helix (Arnold and Haymore, 1991). This conclusion is consistent with the crystal structure of WT-HL oligomerized in detergent (Song et al., 1996).

Note that the Zn(II)-induced current blockades for the point mutants G130H and GND/HN/N (Fig. 2, second and fourth traces, respectively) were qualitatively different, which suggests the metal binding site is not completely defined by His-130 and Asp-128. Thus, one or more of the ionizable side chains Asp-127, Lys-131, or His-144 might play a minor role in determining the geometry and/or electrostatics of the M(II) coordination, but probably do not directly participate in the metal ion binding activity reported here. This conclusion is justified experimentally for Asp-127 because it caused virtually no Zn(II)-induced current fluctuations (Fig. 2, fifth trace). His-144 is too far from His- 130 to coordinate Zn(II) (Song et al., 1996).

At first glance, it might seem remarkable that Zn(II) can be chelated by Asp-126 and His-130 (Fig. 2, rightmost trace) because of the large distance that would separate the two side chains if they were within the same continuous segment of a -sheet. This result suggests that these two side chains lie adjacent to each other on opposite sides of a reverse turn in the -sheet. That conclusion is consistent with the crystal structure of the heptamer, which identifies T129 as the locus of the -sheet turn at one entrance of the pore (Song et al. 1996).

The crystal structure of WT-HL also predicts that amino acid side chains at positions 126 and 130 would point outside of the water-filled pore (Song et al., 1996). However, that structure cannot predict whether these residues are located in the hydrophobic milieu of the membrane or in the aqueous phase outside the channel's lumen. Because the energy barrier for divalent cations to partition into a lipid bilayer membrane is formidable (Parsegian, 1969), the accessibility of the side chains Asp-126 and His-130 to Zn(II) suggests that they are both in the aqueous phase and not buried in the hydrophobic part of the membrane. Measurements of the spectral response of a polarity-sensitive fluorescent dye attached to point Cys HL mutants (Valeva et al., 1996) suggested that side chain 126 is located in the membrane when the protein is reconstituted into liposomes. This is contrary to what we conclude here for single HL channels formed in DiPhyPC planar bilayer membranes and may be due to a difference in how HL is situated in the two membrane systems or because the fluorophore is large and therefore capable of extending into the bilayer. It is also interesting that current blockades were induced by Zn(II) coordinated to side chains located on the outside of the channel. We return to this point later.

Effect of Zn(II) on single channel currents of WT-HL and GND/HQ/N

The effect of 1 µM to 1 mM Zn(II) (trans) on the single channel current of WT-HL and that of the mutant in which the metal binding site is defined by Asp-128 and His-130 (GND/HQ/N) is shown in Fig. 3 (the mean current was subtracted from each recording). Concentrations of Zn(II)  1 mM had no effect on the WT-HL single channel current (left column of current traces). In contrast, 1 µM Zn(II) caused well-defined current blockades in GND/HQ/N and the current fluctuations became less prevalent as the divalent cation concentration was increased beyond 10 µM (right column of current traces). This result can be qualitatively described by assuming that Zn(II) complexes reversibly with each metal-ion binding site on the channel formed by GND/HQ/N. Higher concentrations of Zn(II) saturate the binding sites, which causes the current fluctuations to occur less frequently. This effect is similar to the fluctuations in current caused by the reversible binding of hydrogen or deuterium ions to ionizable residues in the WT-HL channel (Bezrukov and Kasianowicz, 1993; Kasianowicz and Bezrukov, 1995). However, the analogy is not completely identical because the kinetics of the protonation reactions are more rapid compared to those of Zn(II) binding to sites on these mutant channels and because the binding of protons to WT-HL causes the current to increase.

View larger version (23K): [in this window] [in a new window]   FIGURE 3   [Zn(II)]  1 mM had no effect on WT-HL but induced concentration-dependent current fluctuations in GND/HQ/N. The recordings show that Zn(II) induced single channel current fluctuations by binding to a coordination site defined by Asp-128 and His-130. When the [Zn(II)] was increased, the fluctuations in the current through the mutant channel reached a maximum (at [Zn(II)] = 10 µM) and then decreased. Although the channel is formed by seven identical monomers, only three well-defined conductance states were observed with this mutant. The mean current was subtracted from all recordings. ZnSO4 was added to the trans compartment only and a 40 mV potential was applied across the membrane. The data were acquired at 100 Hz and filtered at 10 Hz.

Because the channel is a heptamer formed from identical monomers (Gouaux et al., 1994; Song et al., 1996) and at least two ionizable residues per monomer define the binding site, there could be seven or more distinct metal binding sites per channel. The current recordings in Fig. 3 show that in the presence of Zn(II), there were predominately three distinct conductance states in the channel formed by GND/HQ/N. Higher bandwidth recordings (f_c = 1 kHz) showed that in the presence of 1 to 10 µM Zn(II), there were at least four distinct conductance states in this channel or with that formed by G130H (data not shown). Thus, only three Zn(II) binding sites per channel were readily apparent. Because the sites are in close proximity, the probability that an individual metal binding site is occupied by Zn(II) could be altered by the state of Zn(II) occupancy of the neighboring sites. A similar conclusion was drawn for hydrogen (Bezrukov and Kasianowicz, 1993, Kasianowicz and Bezrakov, 1995) and deuterium ions (Kasianowicz and Bezrukov, 1995) binding to the WT-HL channel. However, in those studies, it was not possible to directly observe conductance changes caused by the binding of single protons to the channel. Instead, the number of these sites (N = 4) was deduced from the H⁺- or D⁺-induced current noise assuming that the sites had identical pKs and the binding of each proton contributes equally to the total difference in conductance at the extremes of pH.

Location of the Zn(II) binding site

To confirm at which end of the pore the binding site was located, we compared the single channel current blockades caused by Zn(II) added to the cis or trans compartments. Approximately 50- to 100-fold less Zn(II) trans elicited the same effect on the single channel current of GND/HQ/N for a given concentration of Zn(II) cis (data not shown). As we showed above, the turn in the -sheet is between side chains 126 and 130. We therefore conclude that the residues between 126-130 are located adjacent to the trans mouth of the HL channel. A similar conclusion was drawn for a Cys mutation of HL at position 130 (Krasilnikov et al., 1997) and is also consistent with the HL crystal structure (Song et al., 1996).

In addition, if all seven engineered heavy metal binding sites were inside the lumen, we would expect that Zn(II) would reduce the channel conductance to a greater extent than that shown in Fig. 3. The relatively subtle effect at high concentrations of Zn(II) is another indication that the sites located are external to the pore.

Divalent cation specificity of the G130H channel

Metal-induced current fluctuations for single channels formed by G130H depend on the type of divalent cation added to the trans chamber (Fig. 4). For example, 100 µM Zn(II) trans causes many rapid blocking events per unit time and a distinctive kinetic pattern unlike those caused by the other divalent cations (e.g., Cu(II), Ni(II), or Co(II)) at this concentration. The binding constants for these metal ions follow the sequence Zn(II) > Cu(II) > Ni(II) > Co(II) (data not shown). Ca(II) was virtually ineffective at causing channel blockades, even at 1 mM (data not shown). The current fluctuations induced by divalent cations are completely reversible and are abolished by subsequently adding excess EDTA to the trans chamber (data not shown).

View larger version (25K): [in this window] [in a new window]   FIGURE 4   The point mutant G130H formed pores with divalent cation specificity. Different divalent cations induced distinct current blockade kinetic patterns in a single G130H channel. The solution on the trans side contained 100 µM of either ZnSO4, CuSO4, NiCl2, or CoCl2. The recordings are arranged in order of decreasing blockade frequency (Zn(II), Cu(II), Ni(II), and Co(II)). The data were acquired at 3000 points/s and filtered at 1 kHz.

Relation of single channel current blockades to Zn(II)-induced inhibition of hemolytic activity

In the absence of Zn(II), the mean single channel current of the pore formed by G130H (GDD/HK/H) was ~33 pA at 40 mV (Fig. 2, second trace). Most of the relatively long-lived Zn(II)-induced current blockades at any Zn(II) concentration decreased the current by <15% or 5 pA (Fig. 4, G130H, 100 µM Zn(II) trans). Thus, the Zn(II)-induced inhibition of this mutant's hemolytic activity (Fig. 1, open triangles) might be caused by a block in channel assembly in rRBC membranes rather than by the occlusion of pre-formed pores. The latter mechanism of hemolysis inhibition was dominant for the mutant H5-HL in which the five amino acid residues between 130-134 (GKIGG) were each replaced by His (Walker et al., 1994).

On the use of scanning mutagenesis and conductance measurements to locate residues in a channel's pore

The three-dimensional crystallographic structures of most membrane-bound proteins are unknown. Thus, alternative methods to deduce their structure and function have been developed. For example, site-specific mutagenesis combined with nitroxide spin labeling of those sites allowed the assignment of secondary structure motifs to specific regions in bacteriorhodopsin (Altenbach et al., 1990), colicin E1 (Todd et al., 1989), and in voltage-gated channel-forming peptides (Barranger-Mathys and Cafiso, 1996).

Because ionic channels have the unique property of having their functional pathway available for interrogation by ionic conductance measurements, scanning mutagenesis studies are also performed to determine which amino acid side chains line the channel lumen. Side chains that bind or react with a chemical reagent are substituted into the channel's primary structure. If the mutant channel's conductance is altered by the reagent, it is assumed that the amino acid that contains the binding site lines the channel's lumen. Cysteine scanning mutagenesis was used to identify sites in the lumens of several channels (e.g., Backx et al., 1992; Akabas et al. 1992; Akabas and Karlin, 1992; Ramirez-Latorre et al., 1996). In this study we showed that 1 µM Zn(II) caused current fluctuations when the metal binding site was defined by either Asp-128 and His-130 (GND/HQ/N) or Asp-126 and His-130 (DNN/HQ/N) (Fig. 2). In the absence of other structural information, it would be tempting to conclude that these side chains are, in fact, inside the pore. However, the crystal structure of WT-HL locates Gly-126, Asp-128, and Gly-130 outside the channel lumen, near the trans chamber mouth (Song et al., 1996). Thus, the divalent cation-induced conductance fluctuations reported here might not be caused by blockades of the channel lumen. Instead, they may be the result of Zn(II)-induced alterations in either the channel's structure or the electrostatic potential that drives ions through the channel.

It is also conceivable that the channel's structure is sufficiently flexible in this region to allow the residues 126, 128, and 130 to occasionally lie in the pore. Although we cannot rule that out at this time, previous work with fluorescent probes suggest this is possible (Valeva et al., 1997). Nevertheless, we have demonstrated here that analyte-induced conductance fluctuations are not a sufficient condition to conclude that a coordinating or reactive side chain is always inside a channel's lumen.

The need for a more cautious interpretation of the physical cause of analyte-induced current changes is made even more apparent by the effects of Zn(II) on the single channel current of mutant DNN/HQ/N. At low Zn(II) concentrations, transient current blockades were evident (Fig. 5, top). As the Zn(II) concentration was increased the current blockades became less apparent. However, Zn(II) altered the mean current in a surprising manner. Specifically, the Zn(II)-induced current blockades slightly decreased the mean current at 1 µM Zn(II) (bottom). However, the mean current actually increased with increasing Zn(II) concentration above 1 µM (both panels). Similar results were obtained at +40 mV (data not shown). While we do not know the physical causes underlying these two contrasting effects, the 12% increase in the mean current was not due to the negligible (< 0.2%) increase in the bulk solution conductivity caused by 1 mM ZnSO₄.

View larger version (14K): [in this window] [in a new window]   FIGURE 5   Zn(II) had two opposing effects on the single channel current of a channel with a chelation site defined by Asp-126 and His-130 (DNN/HQ/N). Top panel: Increasing the concentration of ZnSO4 (trans) caused marked transient increases and decreases in the current. As the Zn(II) concentration increased, the current fluctuations disappeared and the mean current increased compared to that in the absence of Zn(II). Bottom panel: The dependence of the mean current on the ZnSO4 concentration. The solid line through the points illustrates the result of a least squares fit to a single binding constant model (K = 18 µM) to the data. Upon transfer to the computer, the current was filtered at 10 Hz and oversampled at 100 Hz. The applied potential was 40 mV.

It is also interesting to compare the effects of other charged analytes (e.g., protons) on different channels. The currents through some Ca²⁺ channels are blocked by protons (Pietrobon et al., 1988; Prod'hom et al., 1987; Root and MacKinnon, 1994; Chen et al., 1996) whereas the binding of protons to WT-HL causes the current to increase (Bezrukov and Kasianowicz, 1993; Kasianowicz and Bezrukov, 1995). In all of these studies it was concluded that the proton binding sites for each of these channels were inside the pore. Our results suggest that these conclusions, and those based on the results of scanning mutagenesis, warrant further consideration.

Summary

The crystal structure of WT-HL locates residues 126, 128, and 130 on the outside surface of the pore's -barrel. A representation of the channel's structure, which highlights only one of the seven anti-parallel -sheets that form the pore, is shown in Fig. 6. The experimental evidence shown here corroborates part of the pore's -sheet motif and the location of the turn at the trans channel mouth. It also suggests that the trans membrane surface is located above side chain 126 and that channel blockades can occur even when the binding site is located outside a channel's lumen.

View larger version (67K): [in this window] [in a new window]   FIGURE 6   A representation of the -hemolysin ion channel's crystal structure. Only one of the seven antiparallel -sheet pairs, which comprise the pore, is shown. The results of this study corroborate the pore's structure adjacent to the trans entrance.

Future goals

In addition to corroborating structural information about the HL ion channel, we are using scanning histidine mutagenesis to assess HL's ability to function as a biosensor component. One of our long-range objectives is to design and produce ion channels with different sensitivities to, and specificities for, a variety of water-soluble analytes. To accomplish this goal, we are constructing stable ion channels with well-defined binding sites associated with the pore to which an analyte of choice can couple. The results in Figs. 1-5 show that simple modifications to HL induced Zn(II) sensitivity to this channel. Fig. 4 demonstrates that genetic engineering can also be used to confer some degree of analyte specificity to it as well. Although highly desirable, it is not an absolute requirement that a sensor possess perfect selectivity. Spectral methods (Stevens, 1977; DeFelice, 1981; Kasianowicz et al., 1994; Kasianowicz and Bezrukov, 1995) combined with an estimate of the single channel current should prove useful in distinguishing between two different ions in solution, because the rate constants for the reactions will be characteristic of each ion species that binds to the site. In principle, the relative contribution of several analytes competing for the same binding sites should be accessible to spectral decomposition as well.

Cornell et al. (1997) recently reconstituted an ion channel "switch" into rugged supported bilayer membranes. By using impedance spectroscopy they showed that the channel-forming activity of gramicidin was altered by the presence of the analyte, which either released inactive peptide from anchorage sites or sequestered it there.

We note here that genetically engineered pores might also prove useful in other applications, including the characterization and separation of charged and neutral polymers. For example, the WT-HL channel was used to determine some of the physical properties of polynucleotides. Specifically, an applied electric field forced RNA and DNA to traverse the WT-HL channel, giving rise to ionic current blockades with characteristic lifetimes that were proportional to the polynucleotide length (Kasianowicz et al., 1996).