A de Novo Designed Template for Generating Conformation-specific Antibodies That Recognize α-Helices in Proteins*

The generation of antibodies directed toward the surface-exposed regions of a protein using synthetic peptides as immunogens representing surface loops and turns has been widely successful. However, peptides representing α-helical regions are typically unstructured in solution and unable to produce antibodies that recognize α-helices in native proteins. We describe a de novo designed parallel two-stranded α-helical coiled-coil template for immunization to prepare antibodies that recognize α-helical protein sequences in the native protein. This template was designed for maximum stability through an Ile/Leu hydrophobic core and an interchain disulfide bridge. Surface-exposed helical residues are inserted into the template and used for immunization to generate polyclonal antibodies. To demonstrate the feasibility of this approach, 15 residues of the yeast transcription factor GCN4 were inserted into this template, and the resultant antibodies were screened for conformational specificity. Peptide antigens that contain the same surface-exposed residues but differ in structure were used as competitors in a competition assay. Direct competition between the capture peptide immobilized on a biosensor chip, the peptide antigens, and the antibodies generated by the template demonstrated that the antibodies were specific for helical structure in the native coiled-coil (synthetic GCN4 residues 250–280). These antibodies were unable to recognize the same inserted sequence in an unstructured analog. The helix-specific antibodies were also able to identify native GCN4 (31.3 kDa) from yeast whole cell extracts.

The immune system is a remarkable system of protein-protein interactions capable of recognizing an extraordinary number of antigens from a limited pool of genetic segments. While the immune system is of great importance for protecting an organism from foreign proteins, toxins, and bacterial infections, a great deal of work has also been devoted to research and diagnostic uses of antibodies. Development of antibodies as recognition molecules for bioassays and affinity purification is now commonplace.
Immunization with native proteins, domains, or segments is well within the scope of a typical laboratory using recombinant expression techniques. However, immunization with proteins typically yield epitopes directed toward surface loops (1), e.g.
nearly the entire surface of hen egg white lysozyme has been shown to be antigenic (2). Generation of monoclonal antibody libraries directed toward a protein target is also a common approach. The binding epitope of each clone is identified by mapping techniques (3) using combinatorial peptide or phage display libraries (Refs. 4 and 5 and references therein). Alternatively, the entire protein sequence of the immunogen can be screened for binding epitopes with the PEPSCAN approach (6). These epitope mapping techniques typically use short contiguous sequences and map linear epitopes. However, epitopes are often discontinuous. Nevertheless, these methods allow for the generation of protein-specific antibodies, although the epitope is not explicitly delineated during immunization.
As an alternative, synthetic peptide chemistry has also advanced such that peptides of moderate length can now be routinely made in the typical laboratory. Thus, it is possible to synthesize chemically selected protein segments of interest for immunization. Short peptides are poorly immunogenic on their own, but this drawback is circumvented by conjugation to a carrier protein (7). Polyclonal antibodies can be used in lieu of preparing monoclonal antibodies, because only a single epitope is typically used for immunization, and therefore only a single specificity is generated. Recent mutagenesis studies of antigenantibody interactions have shown that although many residues are often identified by structural means as comprising an epitope, relatively few residues contribute the majority of binding energy (8,9). Therefore, it is possible to minimize epitopes if key interactions are maintained.
The usefulness of synthetic peptides to represent protein domains is limited by conformational issues (10). The protein segment of interest is stabilized by secondary and tertiary interactions in the native protein, but the corresponding peptide on its own in solution will typically be a random coil. Without structural restraints, the flexibility of peptides will lead to a diverse set of conformations presented to the immune system, most of which are non-native (11). For example, a 19-residue peptide corresponding to a helical segment of myohemerythrin was shown to be bound to an antibody as a ␤-turn (12), whereas an overlapping peptide was later shown to be bound as a helix (13). Additional complications arrive from the induced fit of the antigen to the antibody recognition site for ligands such as peptides (14). Thus, a need is evident for structure-stabilized immunogens for the generation of antibodies that recognize a specific conformation, like ␣-helices in a native protein.
There are many means of stabilizing helical structure described in the literature. For instance, lactam bridges between Glu and Lys side chains in the i to iϩ4 positions can be incorporated to stabilize helical structure (15). Helical flanking sequences have been used to stabilize intervening sequences into helical structure (16) for use as both an epitope mapping library and helical-stabilized immunogens. De novo design of stabilized helical proteins has also been reported (17)(18)(19)(20). One particular motif, the coiled-coil stem loop, an anti-parallel coiled-coil connected by an intervening loop, was shown to be an effective and compact scaffold for presentation of epitopes as either loops or helices (21). Here, we focus on a method for the design of stabilized ␣-helical immunogens as a means of generating antibodies specific for helical protein segments using the two-stranded ␣-helical parallel coiled-coil as a template. The advantages of this template will be presented.
This laboratory has extensively studied the de novo design of coiled-coils, an excellent model of helical structure (22)(23)(24)(25). The coiled-coil motif is comprised of two right-handed ␣-helices that interact intricately with each other in a slight left-handed superhelical twist. This motif has been identified in many muscle proteins and leucine zipper dimerization domains (Ref. 26 and references therein). It is characterized by a heptad repeat (abcdefg) n in which the a and d positions are typically hydrophobic residues. This repeat pattern creates a hydrophobic face that provides a large thermodynamic driving force for dimerization. The remaining residues are solvent-exposed and often used in de novo design applications. One key advantage of using coiled-coils as a model of helical structure is that stability can be increased by relatively few changes at the a and d heptad positions. The effect of substitution of all 20 naturally occurring amino acids on stability within a model coiled-coil was studied at the a and d positions (27,28).
To demonstrate the feasibility of using a de novo designed coiled-coil as a helically stabilized immunogen, residues 254 -273 of the transcription factor GCN4 (general control of amino acid biosynthesis protein N4) from yeast were designated as an epitope target, and we designed a peptide immunogen in which the surface residues were placed into our helical template (see Fig. 1). GCN4 is a well studied example of a coiled-coil (and hence helical) in which its crystal structure (29), oligomerization state (30), and the thermodynamic properties (31) of GCN4 and its variants have been reported. To evaluate the conformational specificity of the resultant helix-specific antibodies generated from the helical template, several peptides were also designed as antigens for competition experiments. Surface plasmon resonance (i.e. Biacore) was used as a convenient means of measuring antibody-antigen interaction (32). Direct competition with peptides of native sequence demonstrate that the generated antibodies specifically recognize helical epitopes derived from the inserted sequence.

EXPERIMENTAL PROCEDURES
Materials-N-␣-tButyloxycarbonyl-protected amino acids, coupling agents, and starting resin were purchased from NovoBiochem (San Diego, CA) or Advanced ChemTech (Louisville, KY). Trifluoroacetic acid was from Halocarbon Products Inc. (River Run, NJ). The remaining solvents (synthesis or HPLC 1 grade) were from Fisher. Other materials and buffer salts were from Sigma, Aldrich, or Fisher.
Peptide Synthesis, Cleavage, and Purification-The peptides were synthesized manually using standard solid phase N-␣-t-butyloxycarbonyl chemistry and 4-methyl-benzhydrylamine resin as described in Ref. 28. The ␣-amino N-␣-t-butyloxycarbonyl protecting group was removed with 50% (v/v) trifluoroacetic acid in dichloromethane and then neutralized with 20% (v/v) diisopropylethylamine in dichloromethane. Amino acids were activated with 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate, N-hydroxybenzotriazole, and N-methylmorphine in dimethylformamide. After each completed coupling, the remaining free amino groups were acetylated with a solution of acetic anhydride and diisopropylethylamine in dichloromethane. For peptides that require a N-terminal extension, the resin was split in half, and the additional residues were added. Butyloxycarbonyl-norleucine and 4-benzoylbenzoic acid (BB) were coupled in a similar manner.
Peptides were cleaved from the resin with hydrogen fluoride (10 ml/g of resin) containing 10% anisole (v/v) and 2% 1,2-ethanedithiol at Ϫ4°C for 1 h. For peptides that contain benzoylbenzoic acid, ethanedithiol was excluded. Following cleavage and removal of hydrogen fluoride, the crude peptide was washed several times with ethyl ether and then extracted with 50% acetonitrile/water (v/v) and lyophilized.
Crude peptides were purified using reversed-phase chromatography (RPC) on an Agilent Zorbax 300SB-C8 250 ϫ 9.4-mm-inner diameter column (particle size, 5 m; pore size, 300 Å; Palo Alto, CA). The separations were performed on a Beckman System Gold HPLC at a flow rate of 4 ml/min with a linear AB gradient rate of 0.25% B/min, where eluent A was aqueous 0.05% trifluoroacetic acid and eluent B was 0.05% trifluoroacetic acid in acetonitrile. The fractions were collected at 1-min intervals and analyzed on a 150 ϫ 4.6 mm-inner diameter Zorbax C8 column at 1 ml/min flow rate and 2.0% B/min gradient rate. The peptide identity was verified by matrix-assisted laser desorption ionization time-of-flight mass spectrometry and amino acid analysis on a Beckman 6300 amino acid analyzer (Beckman-Coulter, Fullerton, CA).
Formation of a Heterostranded Immunogen-To form specific twostranded heteropeptides, the procedure described in Ref. 33 was used. Briefly, the peptide strand that does not contain the BB moiety was chosen to be derivatized with dithionitropyridine. Excess dithionitropyridine was first dissolved in acetic acid:water (3:1) and then added to dry peptide. This reaction is quantitative and easily monitored by analytical RPC because the peptide derivative absorbs strongly at 210 and 320 nm. The solution was then diluted 10-fold with glacial acetic acid, lyophilized, resuspended in water, and then extracted three times with ethyl ether. The aqueous layer was retained and lyophilized to give the peptide-thionitropyridine derivative. Aliquots of a solution of the free thiol peptide containing the BB moiety were added into a solution of the peptide-thionitropyridine derivative, and formation of the twostranded heteropeptide was monitored by analytical RPC. After the reaction was deemed complete, the disulfide-bridged peptide was purified by RPC and lyophilized.
Two-stranded disulfide-bridged homopeptides were formed by air oxidation in 100 mM ammonium bicarbonate (pH ϳ9). This solution was stirred overnight in an open container, and completion was verified by analytical RPC. The peptide was then purified and lyophilized.
Preparation of Peptide-Carrier Protein Conjugates-Peptide immunogens were conjugated to keyhole limpet hemocyanin for immunization and to bovine serum albumin (BSA) for capture of specific antibodies. 4-Benzoylbenzoic acid (added during synthesis) serves as a photoactivated linker. Keyhole limpet hemocyanin was dissolved in 6 M guanidine hydrochloride, whereas BSA was dissolved in 100 mM ammonium bicarbonate (pH ϳ8) at 20 mg/ml. An aliquot of the carrier protein solution was used to dissolve dry peptide (ϳ3-5 mg placed in a quartz tube) such that the molar ratio was ϳ8:1 (peptide:carrier). The quartz tubes were then placed in a Rayonet photoreaction chamber (Southern New England Ultraviolet Company, Branford, CT) and irradiated with UV light (350 nm) for 2 h. After photolysis, the conjugation mixture was diluted to 2 ml and dialyzed against phosphate-buffered saline pH 7.4 (PBS) overnight using a Slide-a-lyzer TM 10,000 molecular weight cutoff cassette (Pierce). The number of peptides/carrier molecule was determined by amino acid analysis using the norleucine (peptide) to phenylalanine (carrier) molar ratio.
The GCN4 single-stranded peptide was conjugated to BSA through its cysteine sulfhydryl side chain with N-succinimidyl-(4-vinylsulfonyl)benzoate using protocols provided by Pierce (available online at www.piercenet.com).
Immunization Protocol-All of the animal work was carried out at the University of Alberta Human Sciences Laboratory Animal Services in accordance with established protocols on file. Briefly, for each immunogen, five rabbits were immunized at two intramuscular sites. Primary immunization contained 50 g of the keyhole limpet hemocyanin-peptide conjugate (in PBS, pH 7.4) mixed 1:1 with Freund's complete adjuvant. Secondary, tertiary, and booster immunizations (at days 7, 28, and 50) also contained 50 g of conjugate but were mixed with Freund's incomplete adjuvant. Test bleeds (10 ml) were collected on day 35 and evaluated for positive immune response by ELISA. After exsangination on day 58, serum was collected and stored at Ϫ20°C.
Purification of the IgG Pool from Crude Sera-Polyclonal antibodies were precipitated from sera using a modified procedure from Ref. 34. Sera were diluted 5-fold with 50 mM sodium acetate, pH 4.5. Caprylic (octanoic) acid was added dropwise to 2.5% (v/v) with constant stirring at room temperature. This simple precipitation step rapidly and conveniently removes the vast majority of serum proteins by centrifugation at 10,000 ϫ g. PBS stock buffer (10ϫ) was added to the supernatant, and the pH was adjusted to 7.4. Crystalline ammonium sulfate was added to 45% saturation (0.277 g/ml) while stirring in an ice bath to precipitate the immunoglobulins. Centrifugation at 7,000 ϫ g was used to collect the precipitated antibodies. This pellet was resuspended in 2.5 ml of PBS and loaded into a Slide-a-lyzer TM 30,000 molecular weight cutoff cassette for dialysis overnight against PBS, pH 7.4. The antibody solution was effectively concentrated Ͼ50-fold by these precipitations and is generally pure enough to be used directly in ELISA or Biacore applications.
However, as a precautionary step, the antibody was further purified on a HIPAC TM protein G affinity column (ChromatoChem, Missoula, MT). The silica-based packing allows convenient use with conventional HPLC equipment, requiring only isocratic elution, and protein G is highly specific for rabbit IgG. After equilibrating the column in PBS, pH 7.4, the antibody solution was loaded, and unbound proteins were flushed through. Immunoglobulin elution with 0.5 M ammonium acetate, pH 3, was monitored by absorbance at 280 nm. The eluted antibody solution was immediately adjusted to pH 7-8 with ammonium hydroxide and dialyzed against PBS overnight. Subsequently, the antibody solution was concentrated to Ͻ5 ml in an Amicon concentration unit using YM30 ultrafiltration discs (Millipore Corp., Bedford, MA). The approximate concentrations were determined by amino acid analysis, assuming a molecular mass of 150,000. This combination of caprylic acid/ammonium sulfate precipitations with the protein G affinity purification resulted in the purest antibody with minimal nonspecific binding in ELISA and Biacore experiments. This method was preferred over the protein G affinity chromatographic step alone.
ELISA Protocol-High binding polystyrene 96-well ELISA plates (Costar 3590 from Fisher) were used. BSA-peptide conjugate (0.2 g) was adsorbed to the bottom of each well in 50 mM sodium carbonate, pH 9.6, overnight at 4°C. After washing with PBS (pH 7.4 with 0.05% Tween 20 added), each well was blocked with a 2% BSA solution (37°C, 1 h). After washing, crude serum or purified antibody (typically diluted 1:1000) was added to each well and incubated at 37°C for 1 h. After washing away unbound primary (rabbit) antibody, goat anti-rabbithorseradish peroxidase secondary antibody (Jackson Immunolaboratory, West Grove, PA) (diluted 1:5000) was incubated in each well at 37°C for 1 h. After washing, 2,2Ј-azino-bis-3-ethylbenzthiazoline-6sulfonic acid in 10 mM sodium citrate, pH 4.2, with 0.03% H 2 O 2 was incubated for 30 min. The plates were read on a SpectraMax 386 Plus plate reader (Molecular Devices, Sunnyvale, CA) at 405 nm with replicates of 6 wells averaged.
Antibody Recognition of Native GCN4 -YJJ662 yeast cells were grown overnight in minimal medium (50 ml of culture). The cells were collected by centrifugation, washed several times, and resuspended in 300 l of lysis buffer (0.2 M Tris, pH 7.9, 0.39 M (NH 4 ) 2 SO 4 , 10 mM MgSO 4 , 20% glycerol, 1 mM EDTA). 500 l of glass beads (0.2 m), 1 unit of DNase (Promega, Madison, WI), and Complete TM protease inhibitors (Roche Molecular Biochemicals) were added. The cells were then vortexed for 1 min and then cooled on ice for 1 min, repeated five times. Cell debris and glass beads were pelleted, and the supernatant was collected as the whole cell extract.
To remove nonspecific binding, the whole cell extract was precleared with protein A-Sepharose beads for 1 h at 4°C (shaken gently). After removal of the Sepharose beads, antibody (ϳ20 g) and BSA (final concentration, 1% w/v) was added to the whole cell extract and incubated for 1 h at 4°C. Fresh protein A-Sepharose beads were added and allowed to incubate for 1 h at 4°C to immunoprecipitate the antibody-GCN4 complex. After decanting away unbound proteins, GCN4 was eluted from the protein A beads with 0.5 M ammonium acetate, pH 3, and further concentrated by precipitation in 15% trichloroacetic acid. The precipitated protein was pelleted by centrifugation and washed several times with cold acetone. The pellet was finally dissolved in 40 l of SDS-PAGE loading buffer.
The protein samples were loaded onto a 12% SDS-acrylamide gel, separated using a Bio-Rad Mini-Protean II electrophoresis system, and transferred electrophoretically to a nitrocellulose membrane (using Bio-Rad protocols). After several washes with PBS (with 0.05% Tween 20), nonspecific protein-binding sites were blocked with 5% (w/v) BSA for 1 h at 25°C with gentle agitation. After washing, helix-specific antibody (diluted 1:2000) was allowed to bind for 1 h. Goat anti-rabbit horseradish peroxidase secondary antibody (Jackson Immunolaboratory) (diluted 1:2500) was then added after washing away the primary rabbit antibody. The protein bands were developed using a chemiluminescent substrate (horseradish peroxidase LumiBlot kit; Novagen) according to the included instructions.
Circular Dichroism Spectroscopy and Chemical Denaturation Measurements-Circular dichroism spectroscopy was carried out at 20°C on a Jasco J-820 spectrapolarimeter with constant N 2 flushing (Jasco Inc., Easton, MD). Cylindrical cells of varying path lengths were used. Instrument parameters used were a resolution of 0.1 nm, a bandwidth of 1 nm, a scan rate of 100 nm/s, sensitivity of 100 mdeg, and an average of eight scans reported. The results were subsequently expressed as mean residue molar ellipticity [⌰] (deg⅐cm 2 ⅐dmol Ϫ1 ) as calculated in the following manner.
where MRW is the mean residue weight (molecular mass of the peptide divided by the number of residues).
All of the peptides were prepared as stock solutions in benign buffer (50mM phosphate, pH7, 100mM KCl) at ϳ5mg/ml. The exact peptide concentrations were determined by amino acid analysis in triplicate. For the spectra scans, the peptides were diluted to 0.2-1.0 mg/ml (dependent on signal intensity) in either benign buffer or trifluoroethanol (TFE) (50%v/v). The ellipticity was measured from 190 -260 nm in a 0. 5-mm cylindrical cell. For concentration dependence experiments, the peptides were diluted to ϳ1mg/ml in benign buffer, and a 2-fold dilution series was prepared. As the peptide concentration decreased, cells with longer path lengths were needed to maintain reliable signals.
Coiled-coil stability was determined by chemical denaturation. The peptide stock solutions were mixed with 8M guanidine hydrochloride and benign buffer in a dilution series such that peptide concentration was constant but denaturant concentration varied from 0 to 7M. These samples were allowed to equilibrate overnight, and ⌰ 222 nm for each was determined. Ellipticity was then normalized to the fraction folded (and hence contributing to the observed signal) in the following manner.
where ⌰ folded and ⌰ unfolded represent the observed ellipticity of the fully folded and fully unfolded peptides, respectively. Biacore-based Peptide Inhibition Experiments-Biomolecular interaction analysis was performed on a Biacore3000 system (Biacore, Piscastaway, NJ). BSA-peptide conjugates were immobilized on CM5 sensor chips (ϳ600 response units) using the amine coupling immobilization wizard provided in the control software and described in the BiaApplications Handbook. BSA (without any peptide) was immobilized in a similar manner on the control surfaces. Sensorgrams were run with background subtraction at a flow rate of 75 l/min of HEPESbuffered saline, pH 7.4 (HBS-EP buffer from Biacore). Injections of antibody and peptide (250 l) were made with KINJECT using 300 s of dissociation time. The interaction surface was regenerated with a short pulse (10 l) of 6 M guanidine hydrochloride in 10 mM glycine (pH 2).
Relative binding affinities were determined in a direct competition assay. A large volume of diluted antibody (ϳ8 M) was prepared in HBS-N buffer (from Biacore). An aliquot of peptide was mixed with a portion of the antibody solution. From this solution (ϳ80 M peptide), a 2-fold serial peptide dilution series was prepared using the remaining antibody solution. In this way, the antibody concentration remained constant. In addition to these peptide/antibody dilutions, an aliquot of the antibody solution without peptide was also passed over the sensor chip. The binding response was measured over the same interval for all dilutions. The data were then normalized to the uninhibited sensorgram and plotted on a semi-log scale. The data were fitted to the following dose-response equation.
from which the apparent binding affinity midpoint IC 50 was calculated. Lim1 and Lim2 were constrained such that curves ranged from 0 to 100, c is the peptide inhibitor concentration, n is the Hill coefficient, and K d (app) is the apparent dissociation constant of the antibody-capture peptide complex.

RESULTS
Helical Epitope Template Design-The sequence of the helical epitope template is shown in Fig. 1 (Template). Isoleucine and leucine were chosen as the residues to be incorporated at the a and d positions, respectively. Leucine was found to be the most stabilizing residue at the d position for coiled-coil formation (28). Isoleucine was shown to be the most stabilizing res-idue at the a position (35) and minimized higher order oligomers (27). Coiled-coils with an Ile/Leu a/d position hydrophobic core were shown to be two-stranded in at least three sequences (25,30,35). In this template (Fig. 1), the "cassette" into which a target sequence can be inserted can accommodate a sequence 24 residues in length, of which 18 are surface-exposed residues. These 18 residues are placed in the (-bc-efg) heptad positions, and all of the a and d positions will be occupied by Ile and Leu, respectively. A minimum of three heptads was shown to be required for a stable coiled-coil (36). An extension at the N terminus (CAA) incorporates a cysteine for disulfide bond formation, which not only stabilizes the template (37) but also maintains the parallel orientation of the two strands. Alanine was chosen as a spacer residue because of its high helical propensity (38). Several arginine residues are added at the C terminus to aid in solubility. The N terminus was acetylated, and the C terminus was amidated. Conjugation to carrier proteins requires a functional linker, which in this template is BB. This moiety was shown to be an efficient photo-labile linker of synthetic peptides (39) and native protein fragments (40). Norleucine was incorporated for quantitation of the peptide/carrier ratio after conjugation by photolysis. A general outline of this template preparation procedure is shown in Fig. 2.
Design of Immunogens/Capture Molecules-As an initial test of this methodology, we inserted the GCN4 sequence (residues 254 -273) into our template (Fig. 1, GCN4 disulfide-bridged coiled-coil template) and immunized rabbits with this template. This involved inserting 15 surface-exposed residues of the target helical epitope into our template and the inclusion of three alanine residues to occupy the remaining open positions. Although the role of the central Asn 264 residue was shown to be important in maintaining GCN4 in a two-stranded form in solution, we substituted this residue with leucine to further FIG. 1. Sequences of peptides used for immunization, antibody capture and as competitive antigens. A template for helical epitopes is shown at the top. Hydrophobic core residues are indicated by the shaded boxes, whereas asterisks represent residues derived from the helical epitope of the target protein. The template allows for the insertion of 18 surface-exposed residues indicated by the asterisks. In this particular example, 15 surface-exposed residues from the coiled-coil region of GCN4 (residues 254 -273) were inserted into the template, and the remaining three open surface positions were filled with alanine residues (underlined). The immunogens and capture molecules are shown in the middle section. The peptide antigens used in the competition experiments are shown in the bottom section. GCN4(p1-md) is the native antigen, where p1 denotes the native sequence residues 250 -280 (29), and md denotes that this peptide forms a two-stranded coiled-coil that is concentration-dependent (i.e. monomer-dimer equilibrium). The antigen GCN4(p1-db) denotes the disulfide-bridged (db) coiled-coil of the native p1 sequence. The double underline denotes the target region for template antibodies prepared from the immunogens defined above. The helical heptad positions are indicated in italics. stabilize our template design. A GCN4 analog with a Ile/Leu a/d hydrophobic core was also previously shown to be twostranded (30). This immunogen is a well stabilized coiled-coil, and thus the peptides are expected to remain presented as helices during immunization.
An analog of our helical template was also designed to serve as a negative control. In the random coil analog of GCN4 (Fig.  1, GCN4 disulfide-bridged random coil analog), all of the hydrophobic residues in the core a/d positions were substituted with glycine residues. Glycine residues have low helical propensity and hydrophobicity (38,41), and their locations in the a and d positions in addition to their contribution to peptide backbone flexibility are expected to be extremely destabilizing to coiled-coil formation. Thus, we expect that this immunogen would be a random coil. However, the high degree of flexibility of this peptide chain will allow multiple conformations to be presented to the immune system. These two templates both contain the same GCN4 surface residues but differ in their structure.
The same peptides used as immunogens (Fig. 1, Immunogens/Capture Molecules) were also used as antibody capture molecules. The BB moiety allows us to conjugate these peptides to multiple carrier proteins, e.g. keyhole limpet hemocyanin for immunization and bovine serum albumin for antibody capture. Because polyclonal serum is used, the immunogen/capture peptide is used to select the antibody population that is specific for the peptide and not the carrier protein. BSA also serves as a convenient adaptor molecule, because it can easily be adsorbed to ELISA wells or immobilized to biosensor chips. An additional GCN4 single-stranded BSA conjugate was also produced by linking the free sulfhydryl side chain to the Lys residues of BSA using the N-succinimidyl-(4-vinylsulfonyl)benzoate method (see "Experimental Procedures"). This single-stranded analog of GCN4 serves as an additional control because both hydrophobic and hydrophilic faces are exposed in this peptide.
Design of Competitive Peptide Antigens-The native GCN4 sequence was synthesized (31-residue GCN4p1 form; residues 250 -280) as a single-stranded and disulfide-bridged twostranded molecule in addition to several analogs (Fig. 1, Peptide Antigens). The GCN4(p1-db) peptide is a classic coiled-coil stabilized by a disulfide bond. The GCN4(p1-md) peptide exists in a monomer-dimer equilibrium, and thus its structure is expected to be concentration-dependent (see below). An Ala core analog (in which all a/d hydrophobic core positions are substituted by alanine) was synthesized as a monomeric helical form of GCN4. Similarly, the Gly core analog represents a random coil GCN4 analog. These analogs serve as a series of different conformations of GCN4p1 but with the same surface residues at positions -bc-efg. This series of peptide antigens was used to compete with helix-specific antibodies for binding to the capture peptide immobilized on the biosensor chip surface. Using this direct competition, we are able to demonstrate the conformational specificity of our helix-specific antibodies.
Structure of the Peptide Antigens-The structures of the four peptides used as competitive inhibitors (Fig. 1) of the helixspecific antibodies were assessed by CD spectroscopy. The CD spectra of helical peptides show characteristic double minima at 208 and 222 nm that are distinct from random coil peptides (Fig. 3). The CD spectra of GCN4(Gly core) shows negligible helicity (Fig. 3A), even in the presence of the helix inducing co-solvent TFE, indicating that this peptide is a random coil and cannot be significantly induced into ␣-helical structure. GCN4(Ala core) was expected to be monomeric and helical because of the high helical propensity of alanine residues. However, this peptide shows poor helical structure in benign buffer (Fig. 3B) but is highly helical in 50% TFE. Therefore, this peptide can be regarded as an inducible helix. GCN4(p1-md) and GCN4(p1-db) peptides both show classical coiled-coil structure (Fig. 3, C and D) at the concentrations noted. The ratio of peak minima signals (⌰ 222 nm /⌰ 208 nm ) shifts from slightly greater than unity in benign buffer to slightly less than unity in 50% TFE, as expected for coiled-coils ( Fig. 3D and Table I).
The GCN4(p1-db) peptide is disulfide-bridged, and thus helicity should be concentration-independent. The Ala core and Gly core peptides lack sufficient hydrophobic surface area for effective dimerization and therefore exist as monomers. The helicity of these peptides is also concentration-independent. This is shown in Fig. 4 where [⌰], the mean residue molar ellipticity, is independent of concentration for these three peptides. The remaining GCN4(p1-md) peptide is known to exist in a monomer to noncovalent dimer equilibrium (42). As the concentration of GCN4(p1-md) decreases, the majority of the peptide population will be unfolded monomer, assuming a twostate unfolding model as suggested by thermal denaturation studies (31). At the lowest concentration studied (ϳ1 M), the helicity is significantly decreased (ϳ50%) (Fig. 4). This will lower the concentration of the coiled-coil in the peptide competition experiments because the competing peptide will be a mixture of unfolded monomer and folded dimer.
Increased Stability of GCN4 Template-Our interest in generating antibodies to a helical epitope necessitates several changes to the native sequence to stabilize helical structure. The stability of the GCN4 disulfide-bridged coiled-coil template and the native GCN4 peptides (GCN4(p1-md) and GCN4(p1db)) were studied by CD using chemical denaturation. These stability profiles are shown in Fig. 5. The GCN4(p1-md) peptide at a concentration of 153 M shows a guanidine hydrochloride midpoint of 1.2 M, whereas the GCN4(p1-db) disulfide-bridged peptide shows a guanidine hydrochloride midpoint of 3.2 M. Therefore, incorporation of a disulfide bond not only eliminates the concentration dependence of ␣-helical coiled-coil formation but also increases stability by 2 M units. The GCN4 disulfidebridged coiled-coil template was even more stable and was not significantly unfolded at 7 M guanidine hydrochloride, showing an increased stability of at least 5 M units over GCN4(p1-db) and at least 7 M units over native GCN4, assuming a midpoint of Ͼ8 M. This increased stability, caused by the change in the hydrophobic residues at positions a and d, ensures that the coiled-coil template will remain folded in an ␣-helical conformation for antibody preparation regardless of the stability of the helical epitope inserted into the template.
Antibody Specificity by Direct Peptide Competition-Each lot of crude sera was screened against the corresponding BSApeptide conjugate in a qualitative ELISA. Positive responses were compared against preimmune sera, and the lot that showed the greatest difference (and hence the highest titer of specific antibodies) from each set of animals was purified as described. Purified antibody generated with the coiled-coil template (L298) and random coil template (L392) were then characterized for conformational specificity on a Biacore 3000 instrument. Three biosensor surfaces were prepared by immobilizing the following BSA-capture peptide conjugates: BSA-GCN4 disulfide-bridged coiled-coil template, BSA-GCN4 single-stranded, and BSA-GCN4 random coil analog. The first two surfaces were used to capture helix-specific antibody (L298), whereas the BSA-GCN4 (random coil analog) surface was used to capture the non-helix-specific antibody (L392). Helix-specific antibody (L298) did not recognize the random coil analog surface, and the non-helix-specific antibody (L392) did not recognize the coiled-coil template surface (data not shown). Nonspecific binding to the BSA control surfaces was negligible.
The antibody specificity was evaluated by a direct competition assay. Antibody solutions were incubated with varying concentrations of peptide before interaction with the chip surface. If the peptide is effectively recognized by the antibody, the free antibody concentration will be reduced, and thus the resultant binding to the template immobilized on the chip surface will be reduced. A 2-fold serial dilution series was prepared with each of the four peptide antigens, and each series was passed over the three surfaces (an example is shown in Fig. 6). The binding response was calculated, normalized to the unin-hibited response, and plotted as a function of peptide concentration for each series (Fig. 7). The relative binding specificity of a given antibody for each peptide antigen is readily compared in this format.
The competition of the helix-specific antibody with the panel of four peptide inhibitors is shown in Fig. 7A. In this case, the capture surface is the GCN4 disulfide-bridged coiled-coil template, the same construct used in immunization. The GCN4(p1db) disulfide-bridged peptide was the most effective competitor of antibody binding. This was expected, because this antibody  was raised against the coiled-coil template, and this peptide is fully helical. GCN4(Ala core), which is inducible into helical structure, is also recognized, albeit much weaker than the coiled-coil dimer. This suggests that this helix-specific antibody can induce helical structure in antigens that have helical propensity and the identical surface-exposed residues. The GCN4(Gly core) peptide is very weakly recognized by this helixspecific antibody. CD experiments showed that this peptide is unstructured in benign buffer and is not significantly induced into helical conformation in TFE. However, the very weak but measurable inhibition of antibody capture by the Gly core peptide antigen suggests that the helix-specific antibody can in fact induce this peptide into an ␣-helical conformation. The helix-disrupting effect of glycine residues in the a and d positions makes the affinity of this antigen extremely weak. Of particular interest is the GCN4(p1-md) peptide inhibition curve. The helical structure of this peptide is concentration-dependent. We expect that although the monomer is unstructured (31), it can be induced into a helical conformation upon binding to the antibody as was seen for the GCN4(Ala core) peptide. Alternatively, the helix-specific antibody may select the peptide that exists as dimer from the population. This "conformational selection" was also demonstrated in Ref. 43, which shows that single chain antibody specific for a random coil variant of GCN4 was able to select the monomeric peptide from a population of monomeric and dimeric peptides. In either case, as the concentration is increased, coiled-coil formation allows for easier recognition by the antibody than induction of the unstructured monomer. Therefore, the inhibition profile of the single-stranded GCN4(p1-md) peptide should resemble that of the GCN4(p1-db) peptide at higher concentrations but should become less effective at lower concentration, exactly as observed (Fig. 7A). The same experiment was repeated in Fig. 7B, but the capture peptide in this case is the single-stranded GCN4 (Fig. 1). This capture peptide has both the hydrophobic core residues and the display epitope accessible for antibody recognition. The peptide inhibition profile is nearly identical between the two chip surfaces. Interpolated IC 50 values are shown in Table II. This demonstrates that the surface-exposed residues that were inserted into our template and not the hydrophobic core residues are being recognized and that the antibody can induce the single-strand capture peptide into ␣-helical structure. In addition, helix-specific antibody (L298) was also successfully used to immunoprecipitate and identify by Western blotting a single band of ϳ31 kDa (data not shown). This is in good agreement with the reported molecular mass of 31,310 Da for native GCN4. Thus, antibodies generated with the GCN4 coiled-coil template are capable of recognizing the coiled-coil domain in the native protein as well as synthetic peptides.
Antibodies Generated to GCN4 Disulfide-bridged Random Coil Analog-A parallel set of immunizations and peptide in-  (GCN4(p1-db)) was diluted and mixed with antibody (L298) such that antibody concentration is constant but peptide concentration varied. Immobilized on the chip surface is the BSA-GCN4 disulfide-bridged coiled-coil template. The binding response of each antibody/peptide dilution to the biosensor chip surface was measured between the time points indicated by the arrows. These data were then normalized to an injection containing no peptide antigen (topmost curve) and plotted as a function of peptide concentration (0.001 to 17 M).
hibition studies was carried out with the GCN4 disulfidebridged random coil analog (Fig. 1). This construct contains the same surface-exposed residues as the coiled-coil template but is unlikely to adopt any helical conformations. The highly flexible backbone of glycine residues will likely lead to multiple conformations of this peptide presented to the immune system and as a capture molecule. However, we expect that the majority of peptide conformations will be in the extended form rather than well defined helices because of entropic considerations. The analogous peptide inhibition profiles of non-helix-specific antibody (L392) binding to the random coil analog capture surface are shown in Fig. 7C. In this case, the best peptide antigen competitor is GCN4(Gly core), as would be expected because the capture peptide contains the same sequence as that used for immunization. However, this positive response is at least 100-fold weaker than in the case of the helix-specific antibody (Table II), demonstrating the benefit of structurally stabilized epitopes. The very weak but measurable inhibition by the structured peptides (e.g. GCN4(p1-db)) suggests that structured antigens can be recognized by this antibody. It is reasonable to assume that the immune response to a peptide that can adopt multiple conformations may generate a series of antibodies specific for various conformations. During the generation of antibodies, a low frequency but accessible conformation of the immunogen may have been selected and hence amplified. Therefore, in our polyclonal sera, there may exist a small population of antibodies that could recognize a helical conformation and thus are inhibited by the helical peptide antigens.

DISCUSSION
The specificity of antibodies for detecting specific regions within a protein can be invaluable in protein structure/function studies. A minimalist approach is often pursued in generating antibodies with desired specificity where a peptide corresponding to a segment of interest is used as an immunogen. The selection of potential surface-exposed linear epitopes on proteins from amino acid sequence information (44) and the presentation of polyclonal antibodies that recognize these regions in the native protein have been widely successful (45). This approach works well for small linear sequences that are surfaceexposed, unstructured loops, or ␤-turn containing regions. However, a method to generate antibodies that recognize helical segments in proteins when using short synthetic peptides is needed. Because short peptides of helical sequences are unstructured when removed from the protein, there is a need to constrain these sequences in an ␣-helical conformation. We designed de novo a parallel two-stranded ␣-helical coiled-coil as a stabilized template suitable for presentation of helical epitopes. The design of the peptide involves relatively small changes to the native protein segment (only two residues/heptad). By substituting residues on the hydrophobic face, a stable hydrophobic core can be formed, thus driving the association of the two helices into a coiled-coil. Although we have described a homodimer in this work, this technique could easily be applied to create a heterodimer for immunization with two different epitopes.
The approach described requires the identification of a helical protein segment of interest and its orientation. The typical helix found in a globular protein is short (less than 20 residues) and easily accommodated in our template. Alanine residues can be included to fill any open surface positions. Because ϳ50% of all ␣-helices in proteins are amphipathic, the solvent- FIG. 7. Antibody specificity as determined by competition experiments with peptide antigens shown in Fig. 1. BSApeptide conjugates were immobilized to the biosensor chip surface (as described under "Experimental Procedures") in the following manners. A, BSA-GCN4 disulfide-bridged coiled-coil template. B, BSA-GCN4 single-stranded. C, BSA-GCN4 random coil analog. The four peptide antigens GCN4(p1-db) disulfide-bridged (q), GCN4(p1-md) (ƒ), GCN4(Ala core) (f), and GCN4(Gly core) (छ) were used as competitive inhibitors of antibody binding to the biosensor surface. The conformational specificity of antibody L298 (prepared to GCN4 disulfide-bridged coiled-coil template) was assessed in A and B, whereas that of antibody L392 (prepared to the unstructured random coil analog) was assessed in C. a The sequences are shown in Fig. 1. b Antibody L298 was generated with the GCN4 disulfide-bridged coiled-coil template, whereas antibody L392 was generated with the GCN4 random coil template.
c These capture peptides were conjugated to BSA, and the resultant BSA-peptide conjugate was immobilized onto the biosensor chip surface.
d Extrapolated value.
exposed face of the helix is readily defined. The spacing of hydrophobic residues will follow a three-four or four-three residue hydrophobic repeat. In this case, the design of a helical epitope immunogen simply involves placing the helical sequence into the template in Fig. 1 in an appropriate register such that hydrophobic residues in the amphipathic helix are replaced by the hydrophobic residues in the template. The oxidized coiled-coil template will have the same solvent-exposed positions as that in the amphipathic helix from the protein of interest. If the helix of interest is not clearly amphipathic, there are seven possible helical orientations in which a given peptide sequence can be inserted into the template and, thus, seven sets of a/d core residues. The contribution of each a/d residue in each set could be simply summed using the data of Refs. 27 and 28, and the frame that yields the greatest stability is selected as the most probable orientation of surface-exposed and hydrophobic core residues. In cases where several orientations are equally plausible, multiple immunogens could be prepared, and the resultant antibodies could be used to probe the correct surface accessible residues in the helix of the native protein.
With our current design criteria, we maximized the stability of the template (Fig. 5) by incorporating an interchain disulfide bridge and maximizing the hydrophobicity of the a and d residues in the hydrophobic core. Although the incorporation of the disulfide bridge increases stability (compare GCN4(p1-md) and GCN4(p1-db)), the additional increase in stability by optimizing the hydrophobic core was substantially greater (compare GCN4(p1-db) and GCN4 coiled-coil template). Thus, the stability of our template design ensures that antibodies are generated to an ␣-helical structure regardless of the residues inserted into the template. The observed increase in stability of the template could be predicted from previous research on model coiled-coils showing that the major contributing factor to coiled-coil stability is the hydrophobic core positions a and d. Substitution of three Val residues by Ile in the a position was shown to increase the guanidine hydrochloride midpoint of a model coiled-coil by 1.8 M (35). An Asn to Ile substitution in the center of another model coiled-coil, also at the a position, was also found to increase the guanidine denaturation midpoint by 1.7 M (27). The disulfide bridge contribution to coiled-coil stability is also very significant, even when replacing a large hydrophobe like Leu in the hydrophobic core. The increase in the guanidine hydrochloride midpoint was ϳ2.5 M (37). The observed increase in stability of the GCN4 disulfide-bridged coiled-coil template over the GCN4(p1-db) peptide is predicted by the sum of these changes.
Several coiled-coils have been shown to exist in higher order oligomers (28,30,46), which may be a problem in presenting the surface residues properly. However, it is unlikely that sequences in the -bc-efg positions of helices can override the oligomerization state of this extremely stable disulfide-bridged coiled-coil. Each template molecule after conjugation will present two helical epitopes per coiled-coil with the hydrophobic core residues buried and not presented.
Purified sera were qualitatively screened using ELISAs, but we opted to use the Biacore instrument for comparative analysis of specificity. The same interaction surface is used for the entire peptide dilution series and is amenable to automation. Real time data collection allows for estimation of association and dissociation rates and hence equilibrium binding constants. In addition to the peptide antigen competition experiments shown in Fig. 7, the binding affinity of the two antibodies for their respective capture peptides was determined by an antibody dilution series (data not shown). The affinity of the helix-specific antibody (L298) for the coiled-coil template cap-ture peptide was estimated to be ϳ1 ϫ 10 7 M Ϫ1 using the bivalent analyte model, whereas that of nonhelix-specific antibody (L392) for the random coil analog capture peptide was ϳ3 ϫ 10 5 M Ϫ1 . This is in good agreement with our peptide competition results (Table II). Helix-specific antibody (L298) recognizes not only the coiled-coil domain (peptide GCN4p1md) but also the entire full-length native 31.3-kDa protein through immunoprecipitation and Western blotting (data not shown). This result reiterates the utility of using a designed epitope specific for a small functional domain to generate antibodies that successfully recognize the native full-length protein. Additionally, only the surface-exposed residues (once identified) are required for this approach.
The direct competition between the capture peptide (on the biosensor chip surface) and peptide antigens (in solution) for antibody binding allows for a relative ranking of binding affinity. The helical specificity of antibodies generated with the coiled-coil template was assessed by direct competition experiments ( Fig. 7) with the series of peptide antigens shown in Fig.  1. Each peptide differs only by the residues in the a and d heptad positions and contains the same surface-exposed residues in the cassette region. The GCN4(p1-db) disulfide-bridged peptide is the most effective inhibitor of the helix-specific antibody (L298) as expected because it is a stabilized coiled-coil and closely resembles the immunogen. Because this antibody is effectively captured by the coiled-coil template immobilized on the chip surface and effectively competed by GCN4(p1-db) peptide, this demonstrates that the inserted sequence is specifically recognized. In addition, the extremely high stability of the coiled-coil template ensures that only the helical conformation is presented to the immune system, and therefore it would be expected that this polyclonal serum will have only helical specificity. The GCN4(Ala core) peptide is also recognized by our helix-specific antibody, albeit 100-fold weaker than the GCN4(p1-db) coiled-coil. This implies that the antigen-binding site is capable of inducing helical conformation in peptides with helical propensity. The GCN4(Gly core) peptide is poorly but measurably recognized by helix-specific antibody L298, suggesting weak induction of helical structure in this peptide antigen. Because these peptides contain the same surface-exposed residues in the same register as the other peptides in the series, specificity for helical structure is also demonstrated. The helicity of the GCN4(p1-md) peptide was shown to decrease in a concentration-dependent manner (Fig. 4). We expect that induction of helical structure as seen with GCN4(Ala core) will also occur with the GCN4(p1-md) monomer. At lower concentrations, the monomer-dimer equilibrium shifts toward predominately unfolded monomer, and therefore the GCN4(p1md) peptide in solution becomes less effective as an inhibitor.
The same experiment was performed over the GCN4 singlestranded capture surface (Fig. 7B) and yielded nearly identical results. In this single-stranded template strand, the hydrophobic residues are exposed, whereas they were buried in the coiled-coil template and presumably not accessible to the immune system. This suggests that the BSA-GCN4 singlestranded conjugate is easily induced to the helical conformation. The nearly identical peptide inhibition profile suggests that the surface-exposed residues that were inserted into the template are being specifically recognized and not the hydrophobic core residues.
The target sequence was also inserted into a random coil analog of the coiled-coil template (Fig. 1). This random coil analog was designed to represent an unstructured peptide that contains the same epitope sequence. Glycine residues have poor helical propensity and a high degree of backbone flexibility. As such, many conformations are expected to be presented to the immune system, although the most probable conformation is expected to be that of the extended peptide chain. The antibody (L392) generated against this random coil analog is specific for the GCN4(Gly core) peptide, but the binding affinity appears to be weaker than that of antibody (L298) that was generated against the coiled-coil template (Table II). This illustrates the benefit of using a structurally stabilized template for immunization. Interestingly, the remaining three peptide antigens are very weakly recognized by this antibody (L392). The coiled-coil structure of the GCN4(p1-db) disulfide-bridged peptide is unlikely to adopt any nonhelical conformations. This suggests that there is a small population of antibodies in our polyclonal serum that are helix-specific. This minor component is also likely to be able to induce helical structure in the GCN4(p1-md) single-stranded (at low concentrations) and the GCN4(Ala core) peptides because these peptide antigens have high helical propensity. The concept of conformation adaptation of antigens as opposed to induced fit by the antibody was recently discussed (47).
We have demonstrated that by using a de novo designed coiled-coil as a template for immunization, we are able to generate antibodies specific not only for the sequence of interest but also for the helical conformation. This will allow the facile generation of specific antibodies targeted against helical regions of proteins. In principle, once a helical segment of interest is identified (either by structural data or computer prediction), analogous peptides (inserted into this helical template) could be synthesized and used for immunization. In cases where structural data are not available, predicted helical segments could be used as epitopes as a means of mapping surface accessibility of different regions of the protein. Potential neutralizing antibodies could be generated in a similar fashion because many viral proteins are helical in nature (48,49). Pathological protein precipitates (such as the prion protein) are also of great interest. Although the pathological form is the ␤-sheet, the cellular form is believed to be helical (50). Recognition patterns by a panel of helix-specific antibodies could identify the regions that undergo structural change (51,52). We have demonstrated here the feasibility of using a de novo designed coiled-coil as a helical template for generation of helix-specific antibodies and will in the future apply this to a wide range of applications.