Subunit Dimers of α-Hemolysin Expand the Engineering Toolbox for Protein Nanopores*

Staphylococcal α-hemolysin (αHL) forms a heptameric pore that features a 14-stranded transmembrane β-barrel. We attempted to force the αHL pore to adopt novel stoichiometries by oligomerizing subunit dimers generated by in vitro transcription and translation of a tandem gene. However, in vitro transcription and translation also produced truncated proteins, monomers, that were preferentially incorporated into oligomers. These oligomers were shown to be functional heptamers by single-channel recording and had a similar mobility to wild-type heptamers in SDS-polyacrylamide gels. Purified full-length subunit dimers were then prepared by using His-tagged protein. Again, single-channel recording showed that oligomers made from these dimers are functional heptamers, implying that one or more subunits are excluded from the central pore. Therefore, the αHL pore resists all structures except those that possess seven subunits immediately surrounding the central axis. Although we were not able to change the stoichiometry of the central pore of αHL by the concatenation of subunits, we extended our findings to prepare pores containing one subunit dimer and five monomers and purified them by SDS-PAGE. Two half-chelating ligands were then installed at adjacent sites, one on each subunit of the dimer. Single-channel recording showed that pores formed from this construct formed complexes with divalent metal ions in a similar fashion to pores containing two half-chelating ligands on the same subunit, confirming that the oligomers had assembled with seven subunits around the central lumen. The ability to incorporate subunit dimers into αHL pores increases the range of structures that can be obtained from engineered protein nanopores.

Protein pores have been devised for a variety of applications (1)(2)(3). For example, ␣-hemolysin (␣HL) 3 from Staphylococcus aureus has been developed for the controlled permeabilization of cells (4,5), stochastic sensing (6), nucleic acid detection and sequencing (7,8), the examination of single molecule chemistry (9), and the construction of "prototissues" based on droplets connected by bilayers (10,11). To approach these goals, engineered ␣HL pores are essential, and they have been prepared by site-directed mutagenesis with natural and unnatural amino acids and by both noncovalent and covalent chemical modification (12,13). The wild-type (WT) ␣HL pore is a homoheptamer (14,15), and engineered pores have been prepared in both homo-and heteroheptameric form (12). However, in the case of heteromers, the permutation of the subunits around the central axis of the pore has not been controlled. For example, where two of the seven subunits have been altered, there are three possible permutations ( Fig. 1A) (16). One goal of the present study was to demonstrate such control.
The staphylococcal ␣HL pore contains a 14-stranded ␤-barrel that spans the lipid bilayer (15). Two strands are contributed by each of the seven subunits. Under certain conditions, a fraction of the pores may be hexamers (17). By contrast, the structurally similar leukocidin pore is an octamer comprising four of each of two classes of subunits, both of which have a low extent of sequence identity with ␣HL (18,19). Again, the protective antigen of anthrax toxin, which is structurally similar to the ␣HL pore but lacks sequence similarity, exists in both heptameric and octameric forms (20,21). These results suggest that it might be possible to manipulate the subunit stoichiometry of the ␣HL pore, and this was attempted in this study.
One means to control subunit stoichiometry and permutation is to use concatenated subunits produced by genetic engineering, which has most often been used with ion channels (22). The structure and function of K ϩ channels (23), GABA A receptors (24), nicotinic acetylcholine receptors (25), and large conductance mechanosensitive channels (26) have been investigated in this way. Subunit dimers (S-F) of the F and S subunits of the leukocidin pore have been used to show that these subunits alternate around the central axis of the octamer (19). Because the N and C termini of the subunits in the ␣HL pore lie on the same side of the lipid bilayer, we felt that a similar approach might be used to create concatemers of ␣HL, with two goals in mind. First, we wanted to investigate whether subunit dimers of ␣HL would assemble to give oligomers with new stoichiometries (Fig. 1B). A second goal was to increase the scope for genetic and chemical modifications of the pore. By using subunit dimers, we aimed to form heptamers in which two modified subunits are adjacent to each other. Although we failed to accomplish the first goal, the latter was successful. (27). Two PCRs (supplemental Fig. 1) were set up with overlapping primers to generate the two halves of the dimer construct. Each reaction generated one copy of the ␣HL gene, the serine glycine linker, and half of the vector. The PCRs were carried out with the Expand Long Template PCR system (Roche Applied Science). Each reaction contained DNA template (pT7-SC1 containing a single WT ␣HL gene, 30 ng), dNTPs (1.75 l, 10 mM), primer 1 (1 l, 100 M, either SC46 or SC47), primer 2 (1 l, 100 M, either HLnSG/46 or HLnSG/47, where n is the number of residues in the linker), reaction buffer 1 (5 l), DMSO (3 l), and polymerase (1 l) made up to a final volume of 50 l with water. Thermal cycling was carried out as follows: 1 cycle at 94°C (5 min); 18 cycles at 94°C (50 s), 60°C (50 s), and 68°C (9 min); 1 cycle at 68°C (7 min).
Protein Expression by Coupled in Vitro Transcription and Translation-All proteins were prepared by coupled in vitro transcription and translation (IVTT) using an E. coli T7-S30 expression system for circular DNA (Promega) (27). The amino acid mixture minus methionine and the amino acid mixture minus cysteine (both 1 mM) were mixed in equal volumes to give a working mixture of all 20 amino acids. The T7-S30 extract was supplemented with rifampicin (1 l of 500 g/ml rifampicin per 150 l) to inhibit transcription by the endogenous E. coli RNA polymerase. For a 25-l reaction, premix solution (10 l) was combined with the complete amino acid mixture (2.5 l), plasmid DNA (4 l, 400 ng/l), T7 S30 extract (7.5 l), and L-[ 35 S]methionine (1 l, MP Biomedicals, 1174 Ci/mmol, 10 mCi/ml). The reaction was incubated at 37°C for 60 -90 min and then centrifuged for 8 min at 4°C at 25,000 ϫ g to pellet any insoluble proteins. The supernatant was removed and stored at Ϫ80°C. Protein samples were run on 10% BisTris polyacrylamide gels in MOPS running buffer (both from Bio-Rad).
Protein Purification by Ni-NTA Affinity Chromatography-The His-tagged dimers ␣HL-(10SG)-␣HL-H6 and ␣HL-(10SG)-␣HL-D8H6 were purified to remove truncated proteins. Ni-NTA magnetic beads (200 l, Promega) were washed three times with wash buffer (1 ml of 500 mM NaCl, 50 mM Tris⅐HCl, pH 8.0, 10 mM imidazole), resuspended in wash buffer (200 l), and mixed with IVTT protein (200 l). The beads were incubated for 2-4 h at 4°C on a rotator. The beads were then immobilized with a magnet and washed twice with wash buffer (200 l). The protein was removed from the beads with elution buffer (150 l of 500 mM NaCl, 50 mM Tris⅐HCl, pH 8.0, 500 mM imidazole). The protein solution was then passed through a Micro Bio-Spin 6 gel filtration column (Bio-Rad) equilibrated with 50 mM Tris⅐HCl, pH 8.0, to remove the imidazole.
Protein Oligomer Formation and Purification-IVTT proteins (20 -100 l) were mixed with rabbit red blood cell membranes (rRBCM, 5 l, ϳ4 mg of protein/ml) in MBSA buffer (50 l of 10 mM MOPS, titrated to pH 7.4 with NaOH, 150 mM NaCl, 1 mg/ml bovine serum albumin). Oligomer formation proceeded over 1 h at 37°C. The membranes containing the assembled oligomers were recovered by centrifugation at 25,000 ϫ g for 8 min. The pellets were resuspended in MBSA (50 l) and again recovered by centrifugation. Each membrane pellet was solubilized in 1ϫ sample buffer (50 l) without heating, loaded onto a 5% SDS-polyacrylamide gel, and subjected to electrophoresis at 50 V for 14 h with TGS running buffer (25 mM Tris⅐HCl, 192 mM glycine, 0.1% w/v SDS, pH 8.3). The gel was dried without heating onto paper (Whatman 3MM Chr) under a vacuum, and an autoradiograph was obtained. The protein oligomer bands were cut from the dried gel. After rehydration in buffer (150 -200 l of 25 mM Tris⅐HCl, pH 8.0), the paper was removed. The gel was then crushed using a pestle, and the slurry was filtered through a QIAshredder column (Qiagen) by centrifugation at 25,000 ϫ g for 10 min. The protein solution (ϳ5-20 ng/ml) was stored frozen at Ϫ80°C.
Hemolytic Activity Assay-Monomers and dimers of ␣HL (5 l, 100 -400 ng/l) were prepared by IVTT, diluted with MBSA (95 l), and then subjected to 2-fold serial dilutions across the 12-well row of a microtiter plate, such that the final volume in each well was 50 l. An equal volume of 1% washed rabbit erythrocytes in MBSA was quickly added to each well, starting with the most dilute sample. Hemolytic activity was recorded for 1 h by monitoring the decrease in light scattering at 595 nm with a Bio-Rad microplate spectrophotometer, using the Microplate Manager 5.2 software.
Limited Proteolysis Assays-Proteinase K (Sigma) was dissolved in water (5 mg/ml) and used immediately. Limited proteolysis was performed on gel-purified oligomers. The enzyme (2 l, or water for the control reactions) was mixed with the protein (18 l in 20 mM Tris⅐HCl, pH 8.0) and incubated at 21°C for 6 min. The reaction was quenched by the addition of phenylmethylsulfonyl fluoride (2 l, 100 mM in isopropyl alcohol). The oligomers were mixed with Laemmli sample buffer (1ϫ final concentration), and a portion was heated to 95°C for 10 min before electrophoresis in a 10% BisTris polyacrylamide gel (Bio-Rad) in MOPS running buffer.
Single Channel Recordings-Single channel recordings were carried out by using folded planar lipid bilayers, as described previously (16,31). Both chambers contained 1 ml of recording buffer (2.0 M KCl, 2 mM succinic acid, pH 4.0, or 2.0 M KCl, 10 mM MOPS, pH 7.0). The protein was added to the grounded cis chamber. A potential difference of Ϫ50 mV was applied through a pair of Ag/AgCl electrodes, which were set in 2% agarose containing 3.0 M KCl. ␤-Cyclodextrin was added from stock solutions to the trans chamber. The single channel current was amplified by using a patch clamp amplifier (Axopatch 200B, Axon Instruments), filtered with a low pass Bessel filter (80 dB/decade) with a corner frequency of 1 kHz, and then digitized with a Digidata 1320 A/D converter (Axon Instruments) at a sampling frequency of 5 kHz, giving a time resolution of about 300 s. The acquisition software was Clampex 10.2 (Molecular Devices).
Single Channel Recording Data Analysis-Single channel conductance values were determined by fitting the peaks in amplitude histograms to Gaussian functions. Current-voltage (I-V) relationships for single channels were determined by recording the current obtained after stepwise changes in applied potential. For the ␤-cyclodextrin binding experiments, the dwell times () of the blocked and unoccupied states were determined by fitting dwell time histograms to single exponential functions. The Zn 2ϩ binding experiments were carried out and analyzed as described previously (30).

RESULTS AND DISCUSSION
Design and Expression of Subunit Dimers of ␣HL-The termini of the individual subunits in a concatemer are usually linked by short peptide sequences. Such linkers are kept as short as possible to prevent the formation of secondary structure in the linker and to prevent the formation of undesired oligomeric structures. Undesired oligomers include those with interspersed subunits ( Fig. 2A, structure 2) and those with concatenated subunits that are incorporated into separate oligomers (structure 3). As yet, no interspersed subunits have been recognized in channels or pores formed from concatenated subunits, but a subunit dimer can offer one subunit to each of two acetylcholine receptors if the linker between the subunits is sufficiently long (25). However, if the linker between the two  (2 4 1 0 7). B, the structure of the ␣HL heptamer (PDB code 7AHL) was used to determine the length of the linker required to connect the C terminus of the first subunit (red) in a subunit dimer to the N terminus of the second subunit (yellow). In a PyMOL model, we estimated the distance across the protein surface from the C terminus (blue) via Phe-42 (cyan) and Thr-12 (purple) (all on the red first subunit) to the N terminus (green) on the yellow second subunit to be 37 Å. This represents the minimum length of the linker if the N terminus of the second (yellow) subunit is not displaced.
subunits is short, then one subunit is excluded from the central pentameric pore of a single receptor.
The same principle could apply to concatemers of ␣HL; for example, a very short linker might force the exclusion of subunits from the central pore ( Fig. 2A, structure 4). In a PyMOL model, we used the crystal structure of the ␣HL heptamer ( Fig.  2B) to estimate the length of a desirable linker between two adjacent subunits. The distance from the C terminus (Fig. 2B, blue) of a specified subunit (red) to the N terminus (green) of the next anticlockwise subunit (yellow), via Phe-42 (cyan) and Thr-12 (purple) on the protein surface, is 37 Å. A suitable linker should be longer than this to cover the uneven protein surface and to prevent strain of the linker. This model for the linker assumes that the N terminus of one subunit latches tightly onto a second, as seen in the crystal structure of the WT heptamer. However, work in our laboratory has revealed that the N terminus of ␣HL is not critical for the formation of oligomers, and up to 17 residues can be deleted while retaining the ability to form pores (32). Furthermore, the related Vibrio cholerae cytolysin has no such N-terminal domain (33). We therefore had to consider the possibility that the N terminus of the second subunit (yellow, Fig. 2A) of an ␣HL subunit dimer is flexible, thereby shortening the required length of the linker. For this reason, the optimal length of the linker was explored experimentally. A tandem gene encoding a subunit dimer in a pT7 expression vector (Fig. 3A) was prepared by PCR followed by homologous recombination in E. coli. The two subunit genes were separated by a sequence encoding a serine/glycine linker. Such linkers have been employed to link subunits of the leukocidin pore (19) and are thought to be conformationally flexible (34). We designed four different linkers from 5 to 15 amino acids in length (Fig. 3B). In PyMOL models of the elongated sequences, the lengths between the terminal C␣ atoms are 16 to 51 Å (Fig. 3C).
Subunit dimers of ␣HL were prepared by coupled in vitro transcription and translation. SDS-PAGE analysis of the radiolabeled proteins (Fig. 4A) revealed a polypeptide that migrated slightly faster than the 69-kDa marker, consistent with a dimer of 66 kDa. The expression levels of all the dimer constructs were two to five times lower than the WT monomer, as seen in the diminished intensities of the bands (Fig. 4A). Upon incubation with rRBCMs, all of the dimers formed oligomers, which appeared as two closely spaced diffuse bands when nondenatured samples were run in SDS-polyacrylamide gels (Fig. 4B). The oligomers have faster electrophoretic mobilities than the WT heptamer, suggesting that they may have a different subunit stoichiometry or arrangement. A hemolytic activity assay (Fig. 4C) showed that all of the subunit dimers form active pores. Monomeric ␣HL has an HC 50 value (the concentration required for 50% hemolysis) of ϳ25 ng/ml at 20°C (18). By comparison, the subunit dimers of ␣HL have HC 50 values of ϳ50 -100 ng/ml. This activity may in part arise from truncated species that act as monomers, as discussed below.
Does the Linker-Internal N terminus Sequence of the Subunit Dimer Bind Tightly within the Cap Domain?-Because all the dimers formed oligomers with equal efficiency, the linker-internal N terminus sequence between the two subunits presumably does not latch tightly onto the first of the two subunits. To further probe the conformation of the linker-internal N terminus sequence, we carried out limited proteolysis experiments. Proteinase K is a serine protease with broad specificity, which cleaves the N terminus of ␣HL after Ile-7 and Ile-14 when these sites are accessible (35). In the WT heptamer, the N termini of the seven subunits are buried deeply in the cap domain of the protein and are therefore resistant to proteolytic cleavage (

35
). However, truncation mutants of ␣HL lacking part of the N terminus (but not the cleavage sites) are sensitive to protease cleavage, presumably because the shortened N termini are no longer occluded within the cap domain (32). Limited proteolysis of oligomers made from subunit dimers should therefore provide insight into the conformation of the linker-internal N terminus sequence. Proteinase K treatment rendered gel-purified ␣HL oligomers made from subunit dimers (Fig. 4B) more electrophoretically mobile in SDS gels (Fig. 5), but this was not the case for the WT heptamer (Fig. 5). The proteolyzed oligomers made from subunit dimers were also heated in SDS sample buffer at 90°C to disrupt the intersubunit interactions. Electrophoresis revealed the formation of a protein band with a slightly faster mobility than the WT monomer, implying that cleavage had taken place in the linker-internal N terminus sequence to yield truncated monomers. At lower concentrations of proteinase K, the same experiment yielded several protein bands that correspond to incomplete digestion (supplemental Fig. 2). Because even the shortest linker between the subunit dimers in oligomers is accessible to proteinase K, we conclude that the linker-internal N terminus sequences of all four types of subunit dimer do not bind tightly within the cap domain. Therefore, a shorter linker would be preferable to a longer linker, which could lead to the undesired oligomeric structures mentioned earlier (Fig. 2). However, we chose to proceed with a 10SG linker rather than a 5SG linker, as it offers more possibilities for the introduction of unique restriction sites in the DNA that encodes it. This will ultimately be important for preparing extended concatemers of ␣HL (subunit trimers, tetramers, etc.), as it will allow us to "cut and paste" mutant subunit genes into pre-made concatemer genes. Stoichiometries of Oligomers Made from ␣HL Subunit Dimers-The limited proteolysis experiments described above included a control condition, in which the oligomers were heated but not treated with proteinase K. We expected to recover only dimers of ␣HL; however, a mixture of dimers (major component) and monomers (minor component) was found (Fig. 5). Closer inspection of the IVTT products (Fig. 4A) revealed weak bands that migrate more slowly than the monomer but faster than the dimer. We suspected that these protein bands represent truncated dimers, generated by either incomplete transcription of the subunit dimer gene or incomplete translation of the mRNA. Mixtures of full-length concatenated protein subunits and truncated protein subunits have been found in oligomers built from concatemers of the P2X receptor and the K v -1.2/1.1 potassium channel, but their precise origin remains obscure (36,37). In our experiments, the ratio of truncated dimer to full-length dimer is greater in the heated oligomers than in the original IVTT reaction, suggesting that truncated dimers (which would be structurally similar to monomers) are preferentially incorporated into oligomers. Because subunit dimers cannot form a heptameric central pore, without the exclusion of one or more subunits from the ring, this result provided the first hint that ␣HL resists all but a heptameric stoichiometry. For example, such a stoichiometry could be achieved by combining three subunit dimers with one truncated dimer (effectively a monomer). At this point, however, it was not clear why the oligomers made from the subunit dimers showed a different electrophoretic mobility compared with the WT heptamer, especially if they actually share the same stoichiometry with respect to the number of subunits surrounding the central pore. However, it is known that oligomers made from certain mutants of ␣HL can run anomalously fast. For example, the heptamer made from ␣HL-D8, a mutant that carries an oligoaspartate extension at the C terminus (18), has similar mobility to denatured phosphorylase B (97 kDa) in a 5% Laemmli gel, although its actual molecular mass is 240 kDa.
To avoid producing a mixture of truncated species and dimers by IVTT of the subunit dimer gene, we used a gene with a C-terminal His 6 affinity tag. Because only full-length protein carries the tag, Ni-NTA purification should give only subunit dimers of ␣HL. This was confirmed by SDS-PAGE (Fig. 6A). By comparison, the nonpurified dimer contained a truncated protein, in amounts that varied from batch to batch (compare Fig.  4A with 6A). The purified ␣HL subunit dimer was allowed to oligomerize in the presence of rRBCMs. The band corresponding to the oligomer in a 5% polyacrylamide gel (Fig. 6B) was sharp compared with the band derived from the oligomer made from nonpurified dimers. Faint bands could be detected above and below the main band generated from purified subunit dimers, indicating the presence of other types of oligomer. The major band again had a faster mobility than the WT heptamer. When the major oligomer made from purified subunit dimers was gel-purified and heated, only dimers were released (Fig.  6C). Therefore, the truncated proteins seen in Fig. 5 are not the result of proteolytic cleavage during or after oligomer assembly, rather they are IVTT products.
We tried to elucidate the stoichiometry of the oligomers made from purified His-tagged subunit dimers by gel-shift experiments, which had been used successfully to determine the heptameric stoichiometry of the ␣HL pore (14) and the octameric stoichiometry of the leukocidin pore (18,19). The most likely oligomers to be formed by subunit dimers of ␣HL are a hexamer or an octamer (supplemental Fig. 3A). We cooligomerized ␣HL-(10SG)-␣HL-H6 with ␣HL-(10SG)-␣HL-D8H6, expecting to see a stepladder of four bands for a hexamer or five bands for an octamer, based on the possible combinations of subunits. However, no clear pattern was found (supplemental Fig. 3B). The lack of a clear band pattern for oligomers made from ␣HL-(10SG)-␣HL-H6 and ␣HL-(10SG)-␣HL-D8H6 excludes a hexameric (2 3 1 0 6) or octameric (2 4 1 0 8) oligomer structure, 4 leaving a heptamer with excluded subunits as a possible structure. It is hard to predict how many subunits would be excluded from the central pore, but because both ␣HL-(10SG)-␣HL-H6 and ␣HL-(10SG)-␣HL-D8H6 alone give rise to minor oligomer bands as well as one major band (Fig. 4B and supplemental Fig. 3), we suspect that several different arrangements can exist (e.g. 2 4 1 0 7 and 2 5 1 0 7).
Oligomers Made from Subunit Dimers of ␣HL Form Functional Heptamers with Excluded Subunits-To settle the question of pore stoichiometry conclusively, we carried out single channel recordings of oligomers made from subunit dimers. The unitary conductance of ␤-barrel pores is extremely sensitive to the pore stoichiometry and mutation. For example, the conductance of the octameric leukocidin pore is more than three times that of the heptameric ␣HL pore under the same recording conditions (38). Furthermore, systematic mutation of only one residue (Met-113) in homoheptameric ␣HL pores produces pores with conductance values that differ by as much as 24% (39). We extracted the protein from the major bands of oligomers from an SDS-polyacrylamide gel (Fig. 4B) and allowed multiple pores to insert into planar lipid bilayers in 2.0 M KCl, 10 mM MOPS, pH 7.0, under an applied potential of Ϫ50 mV (Fig. 7). By assuming that each current step corresponds to the insertion of a single pore, current amplitude histograms were constructed. WT ␣HL gave uniform quiet channels with a unitary conductance of 1.71 Ϯ 0.26 nS (n ϭ 50, Fig.  7A). By comparison, the major oligomers made from either nonpurified subunit dimers or purified subunit dimers gave noisier channels with wide distributions of conductance values (Fig. 7, B and C). The histograms indicate that a fraction of the pores made from subunit dimers share the same single channel conductance values as the WT heptamer, suggesting that they might also possess heptameric central pores.
We suspected that either the linkers between the subunit dimers or subunits excluded from the central pore might be responsible for the noise observed in the recordings and that they might also partially block the channel, resulting in lower conductance values. We therefore subjected the oligomers made from purified subunit dimers to limited proteolysis with proteinase K, which we had shown previously would cleave the linkers and release any excluded subunits that might be attached to them. Following proteolysis, the oligomers made from purified subunit dimers gave quiet uniform channels with a mean unitary conductance similar to the WT heptamer, 1.59 Ϯ 0.22 nS (n ϭ 161, Fig. 7D). The unitary conductance values of these pores mirrored those of WT heptamers over potentials ranging from Ϫ100 to ϩ100 mV (Fig. 8A).
We also investigated the binding of the molecular adapter ␤-cyclodextrin (␤CD) to these pores (Fig. 8D). The mean dwell time ( off ) of ␤CD in the ␣HL pore can vary by up to 5 orders of magnitude between various mutants (39), so it is another sensitive indicator of the pore architecture. The dwell times of ␤CD in WT pores and pores made from purified subunit dimers and then subjected to limited proteolysis were the same within experimental error (0.33 Ϯ 0.03 and 0.39 Ϯ 0.03 ms, respectively). We therefore conclude that oligomers made from subunit dimers adopt a heptameric central pore architecture, with one or more excluded subunits.
A Single Subunit Dimer Can Be Incorporated into a Heptameric ␣HL Pore-Although we were not able to change the stoichiometry of the ␣HL pore by the concatenation of subunits, a second aim was to incorporate a single modified dimer into a heptameric pore. Because we had learned that oligomers made from subunit dimers readily incorporate monomeric subunits, we set out to make mixed oligomers. The assembly on rRBCMs of different ratios of subunit dimers and monomers that carried a C-terminal D8H6 extension resulted in a stepladder of oligomeric bands upon SDS-PAGE analysis (Fig. 9A,  bands B-F). As expected from previous experiments with (␣HL-D8) 7 (18), the (␣HL-D8H6) 7 oligomer showed the fastest electrophoretic mobility. With increasing amounts of subunit dimer, the oligomer bands shifted upwards, suggesting that untagged subunit dimers are incorporated into the oligomers in various ratios. To analyze the composition of the mixed oligomers, we separated them by electrophoresis in a prepara-tive 5% gel (Fig. 9B) and extracted the bands for further analysis. After heating the oligomers from bands 1 to 4 (Fig. 9B) in SDS to disrupt the intersubunit interactions, we separated the constituent polypeptides in a second gel and determined the ratio of monomer to dimer (Fig. 9C). Band 2 (Fig. 9, B and C) was expected to contain an oligomer that includes one subunit dimer and five monomers (2 1 1 5 7, Fig. 9D), and the intensities of the bands after heating support this view (expected 29% dimer, 71% monomer; observed 31, 69%). Similarly, band 3 (Fig. 9, B and C) includes two dimers and three monomers (2 2 1 3 7, Fig.  9D; expected 57, 43%; observed 55, 45%). Band 4 (Fig. 9, B and C) was initially thought to include three dimers and one monomer (2 3 1 1 7). However, after heating, it was found that this oligomer contains a higher percentage of monomers than expected (expected 86, 14%; observed 74, 26%). A second possibility is an oligomer with two monomers and three dimers, with one subunit excluded from the central pore (2 3 1 2 7, Fig. 9D; expected 75, 25%).
Introduction of Functionality across Two Subunits of the ␣HL Pore-To demonstrate the utility of a subunit dimer in a heptameric pore, we introduced the half-chelating ligand PIDA (Fig. 10A) on each of the two subunits within a dimer of a 2 1 1 5 7 FIGURE 7. Representative pore insertion events and unitary conductance histograms. ␣HL oligomers were extracted from SDS-polyacrylamide gels, before insertion into planar lipid bilayers. The oligomers were made from A, WT ␣HL. B, subunit dimers of ␣HL. C, Ni-NTA-purified subunit dimers of ␣HL. D, Ni-NTA-purified subunit dimers of ␣HL, and the oligomers were treated with proteinase K before the recording was made. The current traces were recorded in 2.0 M KCl, 10 mM MOPS, pH 7.0, at an applied potential of Ϫ50 mV. pore. We have shown previously that two PIDA ligands on the same subunit can come together to form a complex with a single divalent metal ion (30). This requires the ligands to be in close proximity, which can be achieved by placing them on two adjacent subunits in the heptamer but not on two nonadjacent subunits. We prepared two types of modified pore, containing one or two PIDA ligands, by adapting the method published previously. PPIDAЈ has a single PIDA ligand at position 117 on the second subunit of the subunit dimer and is similar to the original PPIDA pore, which has a single PIDA ligand at position 117 on one of the seven monomeric subunits (30). The P(PIDA) 2 Ј pore has one PIDA ligand at position 145 on the first subunit of the dimer (red) and a second PIDA ligand at position 117 on the second subunit of the dimer (yellow) (Fig. 10B). The distance between the two PIDA ligands in the P(PIDA) 2 Ј pore is similar to that in the original P(PIDA) 2 pore, which had one PIDA ligand at each of positions 117 and 143 on one of the seven monomeric subunits (the distances between the two sets of C␣ atoms are both 8.5 Ϯ 0.1 Å in the WT heptamer structure, Protein Data Bank code 7AHL).
The PPIDAЈ and P(PIDA) 2 Ј pores were characterized by single channel recording in planar lipid bilayers, and their properties were compared with the PPIDA and P(PIDA) 2 pores from the original study. The PPIDAЈ pore carried a single channel current of Ϫ86.1 Ϯ 1.6 pA at Ϫ50 mV in 2.0 M KCl, 2 mM succinic acid, pH 4.0, which is larger than the value of Ϫ75.7 Ϯ 1.6 pA reported for the PPIDA pore under similar conditions. The recordings were performed at low pH because Zn 2ϩ binds very tightly to P(PIDA) 2 at neutral pH values (30). The addition of Zn 2ϩ to the trans side of the PPIDAЈ pore resulted in the fluctuation of the ionic current between two discrete levels separated by ⌬I ϭ 2.7 Ϯ 0.1 pA, where ⌬I is the current difference between the level partially blocked by Zn 2ϩ and that of the unoccupied pore (Fig. 10C, upper panel). A kinetic analysis of the binding events gave k on-mono4 ϭ 1300 Ϯ 400 M Ϫ1 s Ϫ1 and k off-mono4 ϭ 24 Ϯ 6 s Ϫ1 to yield an overall formation constant of K f-mono4 ϭ 52 Ϯ 19 M Ϫ1 . The k on value is about half that of the original PPIDA pore (Table 1), although the k off values are nearly the same. The percentage block by Zn 2ϩ relative to the unblocked pore is 3.1% for PPIDAЈ and 2.1% for PPIDA (Table 2). Therefore, in this regard, pores that include subunit dimers differ only subtly from similar pores made from monomers alone.
The P(PIDA) 2 Ј pore carried a single channel current of Ϫ67.1 Ϯ 2.2 pA under the conditions described above, which is again larger than the Ϫ56.4 Ϯ 2.0 pA found for the P(PIDA) 2 pore. After the addition of Zn 2ϩ to the trans recording chamber, we observed both individual and complex current blockades (Fig. 10C, lower panel). The individual binding events to either level A (⌬I ϭ 1.3 pA) or level B (⌬I ϭ 3.1 pA) are consistent with Zn 2ϩ binding to one or the other of two different sites. The complex blockades of the P(PIDA) 2 Ј pore included three different current levels and always began as a step from the unoccupied pore to either level A or level B, followed by a step to a new current level, C (⌬I ϭ 2.0 pA) ( Table 2). As supported by subsequent findings, level C represents a Zn 2ϩ cation fully chelated by both PIDA ligands. Current steps to level A or B from level C were observed, but we never detected steps from C directly to the level of the unoccupied pore. The reaction rates for each step in the binding scheme for Zn 2ϩ by the P(PIDA) 2 Ј pore (Fig. 10D) were determined by fitting the mean lifetimes of the four conductance states (unoccupied, A, B, and C) to the scheme by using the QuB software package ( Table 3). Plots of the rates v A-on and v B-on at pH 4.0 versus the Zn 2ϩ concentration are of the form v ϭ k[Zn 2ϩ ], where k is k on-A4 or k on-B4 (M Ϫ1 s Ϫ1 ), which confirms that both proposed half-chelator⅐Zn 2ϩ complexes are formed in a bimolecular reaction (supplemental Fig. 4). As a further test of the kinetic scheme, we note that at equilibrium the product of the rate constants for clockwise movement around the diamond (Fig. 10D) must equal the product for anticlockwise movement (30,40). We found that k on-A4 k on-AC4 k off-BC4 k off-B4 ϭ 33 Ϯ 9 ϫ 10 6 M Ϫ1 s Ϫ4 and k on-B4 k on-BC4 k off-AC4 k off-A4 ϭ 31 Ϯ 11 ϫ 10 6 M Ϫ1 s Ϫ4 , which are the same within experimental error.
The properties of the P(PIDA) 2 Ј pore are similar to those previously reported for the P(PIDA) 2 pore (30); however, there are some differences. Although level C of the P(PIDA) 2 Ј pore corresponds to the largest current block (⌬I), level C of the P(PIDA) 2 Ј pore lies between level A and B. As noted previously, small changes within the lumen of the pore can have unpredictable effects on pore currents (39,41), i.e. there is no reason to assume that fully complexed Zn 2ϩ should always produce the largest current block. Neither is it possible to unambiguously assign level A or level B to one of the two locations of the PIDA ligand in either the P(PIDA) 2 or the P(PIDA) 2 Ј pore, although, based on the observed sequence of transitions and the Zn 2ϩ binding kinetics, we can be sure that the two levels correspond to Zn 2ϩ binding events. In contrast to the two sites in the P(PIDA) 2 pore, the sites (145 and 117) in the P(PIDA) 2 Ј pore although similar are no longer almost kinetically equivalent. For example, for P(PIDA) 2 Ј, k on-A4 is 2200 Ϯ 100 M Ϫ1 s Ϫ1 , although k on-B4 is 1400 Ϯ 200 M Ϫ1 s Ϫ1 . We note that both k on-A4 and k on-B4 for formation of the half-chelator complexes, as well as k on-AC4 and k on-BC4 for formation of the fully chelated complex, are lower for the P(PIDA) 2 Ј pore by comparison with the corresponding rate constants of the P(PIDA) 2 pore (Table 3). These differences could arise because different buffers were used for the two sets of experiments (succinic acid, pH 4.0, for P(PIDA) 2 Ј; potassium acetate, pH 4.0, for P(PIDA) 2 ). Alternatively, there may be small structural changes in pores that contain subunit dimers, which affect the rate constants. In any event, the effects are small, and the fundamental behaviors of the P(PIDA) 2 and the P(PIDA) 2 Ј pores are similar.
General Implications and Future Prospects-Although the utility of the concatemer approach has been carefully documented, the data from this study reinforce the need for rigorous analysis of translation products to ensure that only full-length concatemers are produced. We were able to obtain full-length subunit dimers by Ni-NTA purification of a His 6 -tagged protein. By placing the affinity tag at the C terminus of the protein, we could ensure that only fully translated proteins would bind to the Ni-NTA column. This approach, however, cannot solve the problem of proteolytic cleavage of the concatemer after translation, which has been a problem in other studies (36).
Even though there are several examples of pore-forming proteins that can form oligomers with variable stoichiometries, we have shown that ␣HL oligomers made from subunit dimers resist all but a heptameric pore stoichiometry. This implies that one or more subunits must be excluded from the central pore. Subunit exclusion is not uncommon for ion channels made from concatenated subunits, when there are more subunits to The PPIDAЈ pore contains a single half-chelator at position 117, while the P(PIDA) 2 Ј pore contains two half-chelators, one at each of positions 145 and 117. The four current levels for the P(PIDA) 2 Ј pore correspond to the unoccupied pore, a single Zn 2ϩ ion bound to one or the other of the half-chelators (levels A and B), and Zn 2ϩ fully complexed with both half-chelators (level C). D, proposed kinetic scheme.

TABLE 1
The association and dissociation rate constants for Zn 2؉ with the PPIDA and PPIDA pores be accommodated in the oligomer than in the native protein (25,26). Based on our findings, we have developed a method to prepare ␣HL pores containing one subunit dimer and five monomers, allowing us to selectively introduce mutations and/or chemical modifications on adjacent subunits of the ␣HL pore. Our metal chelation experiments show that pores containing subunit dimers behave similarly to pores containing only monomeric subunits, showing that they assemble correctly. Subunit concatenation has therefore widened the scope for engineering of the ␣HL pore and might be expanded to extended concatemers (trimers, tetramers, etc.). Ultimately, the approach could allow us to control the nature and position of every subunit in the pore. For practical reasons, it would be useful to introduce unique restriction sites into each linker at the DNA level, to allow the cutting and pasting of specific subunit genes into the construct.