Synthesis and characterization of supramolecular protein aggregates ...

Protein Engineering vol.10 no.11 pp.1289–1294, 1997

Synthesis and characterization of supramolecular protein aggregates: self-assembled, molecularly-ordered, tubes from electrostatic complementation of glutamine synthetase dodecamers

Jing-Ping Chen, Michael J.Dabrowski and William M.Atkins1 The Department of Medicinal Chemistry, Box 357610, University of Washington, Seattle, WA 98195-7610, USA 1To whom correspondence should be addressed

Dodecameric Escherichia coli glutamine synthetase (GS) is formed from identical subunits arranged in face-to-face hexameric rings. In the presence of Zn21 and other transition metal ions the individual dodecamers ‘stack’ to form protein tubes. Previous results have suggested that six binuclear intermolecular metal binding sites are generated at each dodecamer–dodecamer interface by juxtaposition of the N-terminal helices of each subunit adjacent to an analogous helix from a docked dodecamer. In principle, replacement of one of the metal binding sites within each pair of helices with charged amino acids could generate electrostatic interactions that would provide the basis for heterospecific protein–protein interactions. In turn, this would allow for ordered assembly of protein tubes with alternating, chemically distinguishable, components. This hypothesis was tested by replacement of one of the metalligating histidines (His12) with aspartic acid, arginine or cysteine. The H12C mutant was further elaborated by selective thiol modification, with either of the charged reagents 2-iodo-acetic acid or 2-chloro-acetamidine, which yield glutamate (H12C-IA) or arginine (H12C-CA) mimics at position 12. Light scattering and electron microscopy were used to monitor the ‘stacking ability’ of these variants in the presence of Zn21. No, or few, GS ‘tubes’ were observed in solutions containing only H12D, H12R, H12CCA or H12C-IA, in the presence or absence of Zn21. In contrast, in mixtures containing H12C-CA and either H12D or H12C-IA, the complementary GS variants stack in the presence of 100 µM Zn21, with apparent second order rate constants that are comparable to the wild type dodecamers. Fluorescence energy transfer experiments with fluoresceinlabeled H12C-IA (donor) and rhodamine-labeled H12CCA (acceptor) were performed and compared with the energy transfer efficiency with mixtures containing variable ratios of acceptor-labeled and donor-labeled wild type GS; the wild type mixtures provide a benchmark for the extent of energy transfer expected in random linear arrangements of donor and acceptor. The efficiency of metal-dependent energy transfer in mixtures containing the acceptor-labeled H12C-CA and the donor-labeled H12C-IA was 3.2-fold greater than expected for a random distribution of charged variants. Together, the results indicate that the charged variants provide a mechanism for heterospecific interaction between chemically distinguishable dodecamers that align in an ordered one-dimensional array. Keywords: protein engineering/supramolecular self-assembly/ molecular complementation/nanofabrication © Oxford University Press

Generalized strategies for the synthesis of supramolecular aggregates with defined non-covalent architectures are required for fabrication of nano- and meso-scale advanced materials (Whitesides et al., 1991, 1995). For example, two-dimensional monolayers and molecular tubes represent molecular ‘arrangements’ that are widely sought for their potential applications in sensors, electronics, catalysis and drug delivery. Some strategies for nano-fabrication may be learned from biomolecules that self-assemble into non-covalent aggregates of useful size, shape and symmetry (Drexler, 1994; Hartgerink et al., 1996). Alternatively, the self-assembly properties of ‘scaffold’ biomolecules may be exploited to arrange desired functional groups, chromophores or catalysts by specific derivatization of the scaffold molecule (Meldrum et al., 1991; Stayton et al., 1992; Yitzchaik and Marks, 1996). Many proteins naturally form molecular ‘objects’ with potential utility in advanced materials, and thus provide a ‘pool’ of molecular scaffolds (Perham, 1994; Antson et al., 1996). However, the range of objects which could be constructed from protein scaffolds would be increased if the supramolecular connectivity between the components was not limited to the biologically assembled aggregate. Protein engineering strategies, therefore, provide a basis for expanding the pool of biomolecular reagents with utility in advanced materials. Here, we report progress towards in vitro generation of heterospecific protein–protein interactions, which yield molecular tubes with greater supramolecular complexity than provided by the natural protein. The results demonstrate the potential utility of proteins as synthetic reagents in the generation of ordered, novel, one-dimensional aggregates. Dodecameric bacterial glutamine synthetases (GSs) are ‘double donut’ assemblies derived from two face-to-face hexameric rings of identical subunits (Almassey et al., 1986). Each dodecamer possesses sixfold, and lower, symmetries. Some GSs self-assemble into molecular tubes in the presence of transition metal ions (Valentine et al., 1968; Maurizi et al., 1986), by aligning along their sixfold axis of symmetry. The available X-ray structures of the S.typhimurium dodecamer, and a molecular model of a stacked dodecamer complex (Dabrowski et al., 1994), provide a guide for reengineering the protein–protein interface in GS tubes. Site directed mutagenesis with the E.coli GS (Dabrowski et al., 1994) and mass spectrometry with model peptides (Schurke et al., 1996) indicate that binuclear metal binding sites between each of six pairs of adjacent helices from separate subunits in a stacked dodecamer–dodecamer complex are provided by His4 and His12 on each helix (Figure 1, top). Thus, twelve intermolecular metal binding sites are created at each dodecamer–dodecamer interface. Replacement of His4 or His12 with aliphatic or charged side chains abolishes the stacking ability (Dabrowski et al., 1994). In previous work, we have demonstrated that GS tubes also may be formed in a salting out process (Dabrowski et al., 1996). For that process, we observed that homogeneous solutions 1289

J.-P.Chen et al.

provide heterospecific protein–protein interactions leading to metal-dependent formation of tubes with alternately arranged, chemically distinguishable, components (Figure 1, bottom). This hypothesis was tested by preparation of H12D and H12R mutant GSs, and by thiol derivatization of the H12C mutant with 2-chloro-acetamidine (CA) or 2-iodo-acetic acid (IA) which yield arginine (H12C-CA) and glutamate (H12CIA) mimics, respectively, at position 12. The results indicate that a combination of genetic and chemical modification can be used to obtain supramolecular aggregates with ‘higher order’ than the naturally occurring assembly.

Fig. 1. (Top) Schematic model of intermolecular metal binding sites between the N-terminal helices at the interface of two stacked GS dodecamers. Helices from one dodecamer are black and from the other are gray. Two metal ions (black spheres) bind between each helical pair, of which there are six, at the dodecamer interface. Only one binuclear site is occupied in the figure. Metal ligands are provided by His4 and His12. Incorporation of charged functional groups at His12, either genetically or by chemical elaboration of engineered cysteines, provides six pairs of oppositely charged side chains at the interface. The metal site at His4 is retained. (Bottom) Idealized scheme for electrostatic complementation. Black and white disks represent oppositely charged dodecamers. GS variants with similar charge at residue 12 will not stack, but a mixture of the components yields molecularly ordered tubes.

containing some single mutants do no exhibit salt-dependent tube formation. However, when mixed with wild type GS dodecamers, these mutants co-precipitate with wild type GS, presumably through heterospecific interactions (Dabrowski et al., 1996). By analogy, genetically- or chemically-modified GS variants in which one of the inter-helical metal-binding sites is replaced by complementary charged groups would 1290

Materials and methods Protein construction, expression and modification Site directed mutagenesis of E.coli GS was accomplished as described previously (Dabrowski et al., 1994). Protein was expressed and purified as described previously (Dabrowski et al., 1994), using the modifications described for mutants that do not effectively zinc precipitate. Chemical modification with CA or IA were accomplished as described previously (Dhalla et al., 1994; Greenhalgh et al., 1992). Briefly, 1 ml of a solution containing 1.0 mg protein/ml was incubated with 200 mM IA or CA in 50 mM HEPES, pH 7.2, 100 mM KCl, 1 mM MnCl2 at room temperature for 3–4 h. Excess reagent was removed by gel filtration chromatography with Sephadex G-25. Extent of thiol modification was determined with DTNB as described previously (Dhalla et al., 1994). For several preparations of H12C, the extent of modification with CA or IA was 10 6 2 mol CA/mol dodecamer and 9 6 2 mol IA/ mol dodecamer, respectively. Modification of wild type protein was , 2 mol IA or CA/mol dodecamer under the same conditions. Labeling with FRET acceptor and donor Lysine-specific labeling of H12C-CA with TMRA was accomplished with 0.66 mg H12C-CA, 0.1 mg/ml 5- (and 6-)carboxytetramethyl rhodamine, succinimidyl ester (mixture of isomers) in a total volume of 0.5 ml bicarbonate buffer, pH 8.3, 1 h at 25°C, in the presence of 100 µM Zn21. Labeling under conditions that promote stacking was done in order to limit access of TMRA to any lysines on the GS dodecamers that were not on the exterior, peripheral surface. It was observed that when the labeling reaction was performed in the absence of Zn21, then the labeled dodecamers did not, subsequently, stack upon addition of Zn21. Free label was removed by gel filtration with Sephadex G-25. Labeling of H12C-IA with SFX was achieved under the same conditions, with 5-carboxyfluorescein, succinimidyl ester. The average extent of labeling was determined from known extinction coefficients at 585 and 494 nm for TMRA and SFX respectively, to be 0.7 mol TMRA/H12C-CA dodecamer and 1.2 mol SFX/H12C-IA dodecamer. The degree of chemical heterogeneity of the labeled samples was not determined. Based on the experiments described above, it is unlikely that label was incorporated other than on the external periphery of the dodecameric ring structures. SFX and TMRA were purchased from Molecular Probes (Eugene, OR). Wild type GS was labeled under identical conditions. Light scattering measurements Stacking kinetics were determined in 20 mM HEPES, pH 7.2, 1 mM MnCl2, 100 mM KCl at 25°C, unless noted otherwise. The formation of tubes from GS dodecamers is conveniently

Synthesis and characterization of supramolecular protein aggregates

Table 1. Relative rate constants for GS stacking Protein

Concentration (µM)

Relative rate constant

wild type H12C H12D H12R H12C-CA H12C-IA H12D/ H12R H12D/H12C-CA H12C-IA/H12C-CA

0.07 0.07 0.07 0.07 0.07 0.07 0.035/0.035 0.035/0.035 0.035/0.035

8.75 6 0.35 7.0 6 1.7 1.0 0.06 6 0.03 0.83 6 0.58 2.3 6 1.2 0.25 6 0.27 5.15 6 1.6 6.25 6 0.35

Values are normalized to H12D. The units of normalized bimolecular rate constants are sec–1 M total dodecamers–1. Reactions in 20 mM HEPES, pH 7.5, 100 mM KCl, 1 mM MnCl2, 25°C, initiated with 100 µM Zn(SO4)2. Light scattering intensity monitored at excitation 340 nm, emission 340 nm.

monitored by 90° light scattering, where the second order rate constant for assembly at low extent of reaction and at high dilution is obtained as previously described (Yanchunas et al., 1994) from It 5 (I0 1 I`[GS]0kt) / ([GS]0kt 1 1) Here I`, It and I0 are the intensity’s of scattered light at reaction completion, at time t after and before addition of metal ion, respectively, and [GS]0 is the total concentration of GS dodecamers. Under the conditions used here, the reaction kinetics fit well to a second order scheme (Yanchunas et al., 1994). Fluorescence energy transfer Fluorescence spectra were obtained on an SLM-Aminco 8100 spectrofluorimeter. Spectra were corrected for ‘buffer blanks.’ All slit widths were 5 mm. Relevant excitation and emission wavelengths are provided in text and tables. Electron microscopy Eectron microscopy was performed on a Philips 200 electron microscope at 80 keV. Samples were prepared by adsorption of samples on Formvar-coated grids, and subsequently blotted and air dried. Adsorbed samples were negatively stained with 0.2% uranyl acetate, blotted and air dried. Results Several combinations of GS variants were examined, by light scattering and electron microscopy (EM), for their ability to self-assemble into tubes, in the presence of Zn21. At low extent of reaction (,5%) and high dilution, the reaction approximates kinetically a bimolecular mechanism, and relative rate constants for stacking were determined (Table 1; wt k 5 1.53104 M–1 sec–1) as described previously (Yanchunas et al., 1994). Separately, the individual charged variants H12D, H12C-CA and H12C-IA stack poorly, if at all. However, in some mixtures containing two of these components, the rate and extent of stacking are increased significantly (Figure 2). That is, some mixtures of charged variants ‘complemented’ one another in the stacking reaction that yields GS tubes. The relative rate constants reported in Table 1 were determined at 0.07 µM total protein, for the individual components H12R, H12D, H12C-IA and H12C-CA, as well as for mixtures, where each component was 0.035 µM. At 0.035 µM, the individual components stack more slowly than at 0.07 µM, which is the concentration used for determination of the relative rate

Fig. 2. Kinetics of GS stacking. The intensity of scattered light after addition of Zn21 is monitored for the H12D, H12C-CA and a mixture of H12D 1 H12C-CA. The mixture stacks at a faster rate rate than either individual component. Recovered rate constants for various combinations of GS are summarized in Table 1.

constants in Table 1. Interestingly, the H12D/H12R mixture does not exhibit complementation, whereas the H12D/H12CCA and H12C-IA/H12C-CA mixtures do. Studies with additional mutants, chemical modifications and metal ions are required to define further the structural requirements for the observed complementation. These results suggest that H12D or H12C-IA form heterospecific complexes with H12C-CA. Because of the symmetry of the individual dodecamers, which places identical charged functional groups on both flat sides of the GS dodecamers, one-dimensional assembly of tubes would be expected with alternating, complementary variants with opposite charge (Figure 1, bottom). Because the light scattering experiments do not provide directly structural information for the aggregates formed, EM was used to determine their gross structure. In the absence of Zn21, or other transition metal ions, GS dodecamers routinely appear as homogeneously dispersed ‘ring’ structures in negatively stained EM micrographs, with some dodecamers oriented with their flat surface on the EM grid, and some dodecamers on edge (Figure 3). Each of the variants described here appeared as intact ‘double donut’ ring structures in EM micrographs in the absence of Zn21; the dodecameric structure was not grossly altered by mutagenesis or chemical modification. As suggested by the light scattering experiments, EM micrographs demonstrated that tubes were abundant only after addition of Zn21, in the presence of both H12C-CA and either H12D (Figure 3) or H12C-IA (not shown). In addition, the EM results confirm that the complex formation monitored by light scattering is due to specific formation of tubes, and not a non-specific aggregation. Direct elucidation of the arrangement of components in the target aggregate is a major challenge of non-covalent synthetic approaches, and ‘descriptive chemistry’ provides important reference points (Whitesides et al., 1995). Due to their size and tendency to form paracrystalline aggregates, the GS tubes are likely to be difficult to analyze by X-ray diffraction. As a qualitative, alternative probe of the distribution of complementary GS dodecamers in the stacked tubes, fluorescence energy transfer (FRET) was used with 5-carboxy-fluorescein-labeled H12C-CA (H12C-CA-SFX) and 5-carboxy-tetramethyl-rhodamine-labeled H12C-IA(H12C-IA-TMRA), at 0.9–1.2 mol label/mol dodecamer. The Forster distance for this donor– 1291

J.-P.Chen et al.

Fig. 3. Electron micrographs of a mixture of H12C-CA (0.035 µM) and H12C-IA (0.035 µM) before (Top) and after (Bottom) addition of 100 µM Zn21. In the Top micrograph, a homogeneous distribution of light colored particles is observed. After addition of Zn21, these particles have selfassembled into tubes, seen here lying down on the grid surface (Bottom). Efficient tube assembly requires both protein components and metal ion (Table 1). Magnification is ~240 0003.

acceptor pair is 45 Å (Selvin, 1995), and with an axial distance of 108 Å/dodecamer (Almassy et al., 1986), FRET will be observed only for donor–acceptor pairs that are axially aligned on directly adjacent dodecamers within a tube. For these experiments, the fluorescence emission spectrum of a mixture containing either H12C-CA-SFX, H12C-IA-TMRA, a mixture of both, or the analogous wild type proteins labeled with SFX or TMRA were recorded. Zn21 was then added, and light scattering intensity was monitored until no time-dependent change in scattering intensity was observed. This was considered as the completion of the stacking reaction. The emission spectra were then recorded again, and FRET was quantitated from the increase in emission intensity of the acceptor (TMRAlabeled protein). In fact, FRET was evident from both a decrease in intensity of H12C-CA-SFX (corrected for metaldependent decrease in the absence of acceptor), and from an increase in intensity of H12C-IA-TMRA, which was metal and donor dependent. It was demonstrated experimentally that stacking kinetics were not altered upon labeling with donor or acceptor; H12C-CA-SFX and H12C-IA-TMRA individually stack with the same apparent rate constants as H12C-CA and H12C-IA, respectively (not shown). The emission spectra of an acceptor–donor mixture (7:1 molar ratio) before and after 1292

Fig. 4. (Top) Fluorescence energy transfer between H12C-IA-SFX and H12C-CA-TMRA occurs only after addition of Zn21. Excitation was at 494 nm. The extracted ‘acceptor only’ spectra are also shown. The acceptor emission intensity increases after stacking. (Bottom) Calibration of stacking dependent increase in acceptor emission for various ratios of wild type donor–acceptor pairs. The increase in efficiency, ∆E, is the ratio of acceptor intensity after stacking (A 1 D) to the intensity before stacking (A). The wild type calibration yields expected FRET efficiencies for random distribution of donors and acceptors at these ratios. The H12C-CATMRA:H12C-IA-SFX mixture at 7:1 is nearly as ordered as the 1:1 ratio of wild type. For this mixture, the experimental error is within the size of the data point. The solid line is a visual aid only; it does not represent a specific mathematical model.

stacking are shown in Figure 4, along with the extracted ‘acceptor only’ spectra (Selvin, 1995). The efficiency of energy transfer (E) was determined from the change in quantum yield of the acceptor in the absence (IA) and presence of donor (IAD), as E 5 [εA/εD][IAD/IA – 1], where IAD and IA are emission intensities of the acceptor before and after stacking, and εA and εD are molar extinction coefficients of acceptor and donor at the relevant wavelengths (Selvin, 1995). In

Synthesis and characterization of supramolecular protein aggregates

addition, the relative ‘order’ of the stacked tubes can be assessed from the relative values of E in the complementary mixture (H12C-IA-SFX 1 H12C-CA-TMRA) compared with stacked mixtures of SFX-labeled and TMRA-labeled wild type GS with variable acceptor:donor ratios (Figure 4, bottom). The wild type mixtures provide a qualitative benchmark of E in a completely random distribution of different proportions of acceptor:donor. The analysis assumes one label per dodecamer, and an experimentally determined average length of 13 6 5 dodecamers/complex. For a linear array of 14 total dodecamers, containing two donor labeled, the number of possible arrangements is 14!/(12!2!) 5 91, if only statistical factors control the distribution. The total number of acceptor labeled dodecamers in this ensemble is 12391 5 1092. For this random distribution, the fraction of acceptor labeled dodecamers that are directly adjacent to donor labeled GS molecules is 277/ 1092 5 25.3%, by inspection. In contrast, a completely ordered array with alternating components places 100% of acceptor labeled dodecamers directly adjacent to donor labeled dodecamers. The fraction of acceptors for which FRET is possible is reduced further by a factor of 2 or 3, assuming that degenerate forms of the stacked complex, resulting from rotation about the sixfold axis of symmetry, will allow donor–acceptor pairs on adjacent dodecamers to lie within 45 Å for three out of six or two out of six rotational isomers. However, the same isomers will exist for a completely ordered array, and therefore, it is not necessary to include this factor in the analysis of the theoretical ratio, Emix/EWT, between wild type and complementary mixtures. Thus, the theoretical value is expected to be 100/25.3 5 3.9, when comparing a completely ordered versus a completely random array of acceptors:donors with a 12:2 ratio. This will be true as long as other terms which contribute to E (κ2, R, J) are identical for the wild type reference and for the complementary mixture. The experimentally observed Emix/EWT is 3.2 for the H12C-IA-SFX/H12C-CA-TMRA mixture compared with wild type at an acceptor:donor ratio of 7:1, suggesting that the array is nearly completely ordered (Figure 3, bottom). The analogous experiment, with H12CCA-SFX and H12C-IA-TMRA, was not performed due to the ‘background’ rate of stacking for H12C-IA. The ratio of E values for the complementary mixture versus the wild type benchmark, Emix/EWT, indicates that the former is nearly completely ordered (1 – 1 – 1 –), in a stacked complex generated from a 7:1 ratio of acceptor:donor. Even in the presence of a large stoichiometric excess of acceptors, nearly all acceptor labeled dodecamers are adjacent to donor labeled complements, as depicted in Figure 1. Discussion Taken together, the results described here indicate that heterospecific protein–protein interactions have been engineered in E.coli GS, by introduction of complementary functional groups at sites that are identical in the wild type complex. A possible source of the complementation is the favorable electrostatic interactions between the oppositely charged functional groups at position 12 that are expected to be in close proximity in the stacked complex formed between two variants. Static models do not suggest a structural basis for the lack of complementation in any mixtures containing H12R, although the H12C-CA side chain is ~0.8 Å shorter than the natural arginine side chain, and the H12C-IA side chain is ~0.8 Å longer than the natural aspartate. Clearly, additional experiments are required to provide a detailed structural model

consistent with these results. Although the experiments do not provide a structural basis for the differences in behavior of H12C-CA and H12R, the observed complementation in the formation of tubes with H12C-CA and either H12D or H12CIA dramatically reveals the possibility of synthesizing protein aggregates with more complex ‘order’ than is provided in the natural aggregate. These results provide important precedent for engineering supramolecular protein complexes with defined architecture. Interestingly, analogous results have been achieved with electrostatic interactions in heterospecific peptide–peptide interactions, but we are unaware of examples of reengineering supramolecular protein aggregates in this manner. The utility of proteins in nano-fabrication remains largely speculative. However, a brief comparison of our results with conceptually analogous experiments serves to highlight the potential applications of linear protein assemblies. Nanotubes constructed from inorganic zeolites (Martin, 1995), carbon (graphite, Tamai et al., 1997), lipids (Schnur, 1993) and organic ‘dyes’ with liquid crystalline properties (Adams et al., 1994) have attracted attention for their potential use (van Nostrum, 1996) as molecular ‘containers’, ‘channels’ or ‘wires’. In addition, cyclic peptides containing L- and D-amino acids have been used to develop design strategies for peptide nanotubes with ‘tunable’ channel dimensions and with specific inter-tube packing interactions (Hartgerink et al., 1996; Ghadiri, 1995). To date, no single class of molecules provides all of the features desired for particular nanotube applications. In the case of proteins or peptides, we expect the major application of self-assembled tubes to be as scaffolds or templates to direct spatial alignment of redox-active functional groups, chromophores or catalysts in linear arrays. In order to maximize this potential utility, additional control over self-assembly reactions is required. The experiments presented here provide a compelling demonstration of the feasibility of engineering proteins for self-assembly of useful supramolecular arrays. Notably, in the absence of Zn21, the electrostatically paired variants do not stack. Thus, the putative electrostatic pairing between functional groups at position 12 is insufficient to drive the stacking reaction, which still requires metal ions. In fact, this was studied experimentally by varying the concentration of KCl and monitoring the formation of GS tubes. Even at low ionic strength (20 mM HEPES, 1 mM MnCl2, no KCl) no stacking was evident in complementary mixtures in the absence of Zn21. Apparently, the charged variants provide kinetic selectivity for heterospecific recognition in the metaldependent self-assembly reaction, but they cannot provide sufficient driving force in the absence of metal ions. Presumably, the symmetry of the dodecamer–dodecamer interface is essential for cooperative interactions between weak, individual, electrostatic interactions at multiple, spatially separated, sites. Regardless of the molecular driving force, the results demonstrate that heterospecific interactions can be exploited to yield protein aggregates with more complex, or ‘heightened’, supramolecular order than is found in the naturally occurring assembly. The pool of biomolecular scaffolds with defined supramolecular order can be expanded by simple chemical strategies when symmetry provides cooperative interactions. Acknowledgements This work was supported by The National Science Foundation (NSF9305202) and The Whitaker Foundation (930272).

1293

J.-P.Chen et al.

References Adams,D., Schumacker,P., Simmerer,J., Haussling,L., Slemensmeyer,K., Etzbach,K.H., Ringsdorf,H. and Haarer,D. (1994) Nature, 371, 141–143. Almassy,R.J., Janson,C.A., Hamlin,R., Xuong,N.H. and Eisenberg,D. (1986) Nature, 323, 304–309. Antson,A.A., Dodson,E.J. and Dodson,G.G. (1996) Curr. Opin. Struct. Biol., 6, 142–150. Chilkoti,A., Schwartz,B.L., Smith,R.D., Long,C.J. and Stayton,P.S. (1995) Biotechnology, 13, 1198–1204. Dabrowski,M.J., Yanchunas,J. Jr., Villafranca,B.C., Dietze,E.C., Schurke,P. and Atkins,W.M. (1994) Biochemistry, 33, 14957–14964. Dabrowski,M.J., Dietze,E.C. and Atkins,W.M. (1996) Prot. Engng, 9, 291–298. Dhalla,A.M., Li,B., Alibhai,M.F., Yost,K.J., Hemmingsen,J.M., Atkins,W.M., Schineller,J. and Villafranca,J.J. (1994) Prot. Sci., 3, 476–478. Drexler,K.E. (1994) Annu. Rev. Biophys. Biomol. Struct., 23, 377–405. Ghadiri,M.R. (1995) Adv. Mater., 7, 675–677. Greenhalgh,D.A., Subramaniam,S., Alexiev,U., Otto,H., Heyn,M.P. and Khorana,H.G . (1992) J. Biol. Chem., 267, 25734–25738. Hartgerink,J.D., Garanja,J.R., Milligan,R.A. and Ghadiri,M.R. (1996) J. Am. Chem. Soc., 118, 43–50. Martin,C.R. (1995) Acc. Chem. Res., 28, 61–68. Maurizi,M.R., Kasprzyk,P.G. and Ginsburg,A. (1986) Biochemistry, 25, 141–151. Meldrum,F.C., Wade,V.J., Nimmo,D.L., Heywood,B.R. and Mann,S. (1991) Nature, 349, 684–686. Perham,R.N. (1994) Biosensors Bioelectronics, 9, 753–760. Schnur,J.M. (1993) Science, 262, 1669–1676. Schurke,P., Jochheim,C.M., Dabrowski,M.J. and Atkins,W.M. (1996) In Ornstein,R. (ed.) Biomacromolecules: From 3D Structure to Application. Proceedings 34th Hanford Symposium on Health and the Environment, October 2, Pasco, WA. Selvin,P.R. (1995) Methods Enzymol Rel. Areas Mol. Biol., 246, 300–335. Stayton,P.S., Olinger,J.M., Jiang,M., Bohn,P.W. and Sligar,S.G. (1992) J. Am. Chem. Soc. 114, 9298–9299. Tamai,H., Ikeuchi,M., Kogimas,S. and Yasuda,H. (1997) Adv. Mater., 9, 55–58. Valentine,R.C., Shapiro,B.M. and Stadtman,E.R. (1968) Biochemistry, 7, 2143–2152. van Nodstrum,C.F. (1996) Adv. Mater., 8, 1027–1030. Whitesides,G.M., Simanek,E.E., Seto,J.P., Chin,D.N., Mammen,M. and Gordon,D.M. (1995) Acc. Chem. Res., 28, 37–44. Whitesides,G.M., Mathias,J.P. and Seto,C.T. (1991) Science, 254, 1312–1319. Yamashita,M.M., Almassy,R.J., Janson,L.A., Lasek,D. and Eisenberg,D. (1989) J. Biol. Chem. 264, 17681–17689. Yanchunas,J. Jr., Dabrowski,M.J., Schurke,P. and Atkins,W.M. (1994) Biochemistry, 33, 14949–14956. Yitzchaik,S. and Marks,J. (1996) J. Acc.Chem. Res., 29, 197–202. Received May 1, 1997; revised July 17, 1997; accepted July 21, 1997

1294