Preliminary Crystallographic Data, Primary Sequence, and Binding Data for an Anti-peptide Fab and Its Complex with a Synthetic Peptide from Influenza Virus Hemagglutinin*

X-ray qualityprystals which diffract to high resolution (51.9-2.1 A) have been grown of an anti-peptide Fab and its complex with a 9-residue peptide antigen. Both crystals arg mon0clinic~P2~, with uait cell dimensions a = 90.3 A, b = 82.9 A, co= 73.4 A, 9 = 122.5’ {or the native Fab and a = 63.9 A, b = 73.0 A, c = 49.1 A, /3 = 120.6’ for the complex. The peptide sequence corresponds to residues 100-108 of all influenza virus hemagglutinins (HA1) of the H3 subtype (1968-1987).

the three-dimensional structure of the complex will elucidate the details of how anti-peptide antibodies recognize a small peptide antigen and provide insights into the recognition of the same sequence in the intact protein antigen. As both native Fab and the peptide-Fab complex have been crystallized, we can also determine in addition whether changes in the structure of the antibody accompany antigen binding. The nucleotide sequence of the mRNA coding region of the antipeptide Fab has been determined to provide the amino acid sequence ultimately required for the high resolution three-dimensional structure determination.
Major advances in our understanding of the detailed interactions in antibody-antigen complexes have been made pos-* This work was supported by National Institutes of Health Grant AI-23498 (to I. A. W.) and a Medical Research Council (Canada) fellowship (to J. M. R.). This is Publication 5286-MB from the Research Institute of Scripps Clinic. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertkement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequencefs) reported in thispaper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) 504061. § To whom correspondence should be addressed.
sible recently with the structure determinations of three Fabprotein complexes. The structure of two Fab-lysozyme complexes (Amit et al., 1985(Amit et al., , 1986Sheriff et al., 1987) and a neuraminidase-Fab complex (Colman et al., 1987a(Colman et al., , 1987b have shown the very close association of antibody and antigen. Although the structure of the native antigens are available for comparison with their structure in each of the solved complexes, in none of these cases could the structure of the free Fabs be determined as suitable native Fab crystals could not be obtained. Consequently, one cannot be certain of the extent to which antibody conformational changes accompany binding to these protein antigens. Further advances in our understanding of antigenic determinants have arisen from the enormous proliferation of studies using synthetic peptides as immunogens (Arnon and Sela, 1969;Lerner, 1982Lerner, , 1984Berzofsky, 1985). Antibodies raised against such peptides have been shown to react not only with determinants previously identified by anti-protein antibodies but perhaps more importantly with epitopes not normally reactive when the intact protein is the immunogen (Green et al., 1982, Lerner, 1982, 1984Berzofsky, 1985). These antipeptide antibodies are currently being used as research tools for a multitude of purposes which include basic detection of a given antigen, recombinant protein purification, tissue specificity of viruses, serological testing, detection of biologically important conformational changes and vaccine development.
In light of such studies, we wish to investigate in structural detail how an anti-peptide antibody can react with both a peptide immunogen and its cognate sequence in the parent antigen. It has been suggested for anti-peptide antibody recognition that flexibility of the peptide determinant in the intact antigen is important (Tainer et al., 1984(Tainer et al., , 1985. This correlation has also been made for some anti-protein antibodies (Westhof et al., 1984;Williams and Moore, 1985), although some controversy still exists in both systems as to what parameters correlate best with antigenicity (Moudallal et al., 1985;Thornton et al., 1986;Novotny et al., 1987;Geysen et al., 1987).
In this study, we investigate the interaction of a peptide, corresponding to a sequence of the influenza virus hemagglutinin (residues HA1' 100-108), with an anti-peptide antibody.
The original influenza peptide study was initiated by Green et al. (1982), who investigated the reactivity of 25 different anti-peptide (HA1) rabbit antisera with intact hemagglutinin or influenza virus. An analysis of the location of the reactive peptides based on the x-ray structure of the A/Aichi/68 influenza virus HA (Wilson et al., 1981), showed that the entire surface of the HA molecule could be considered antigenic (Green et al., 1982). A panel of 21 mouse monoclonal antibodies was then raised against one of these peptides (HA1 75-110) by Niman et al. (1983), and their fine specificity was determined (Wilson et al., 1984(Wilson et al., , 1986Wilson, 1985;Houghten, 1985;Houghten et al., 1986). The major immunodominant site lies in the trimer interface in the intact HA (Wilson et al., 1984) consistent with the fact that the antibodies react in solution at neutral pH only with monomeric HA1 "tops" (residues 27-328 (KO M)) (Wilson, 1985;Wilson et al., 1986).
Recently, these antibodies have provided evidence for a pHinduced conformational change in the HA trimeric structure as might well occur in the biologically important HA-mediated membrane fusion event between the virus and the host cell endosomal membrane (White and Wilson, 1987). In addition, the free peptide (HA1 98-106) has been shown to have a surprisingly high percentage of type I1 @-turn in water (Dyson et al., 1985) which has led to the proposal of a possible role for the folding of peptide in solution in producing proteinreactive anti-peptide antibodies (Dyson et al., 1988).
In this paper, we report the crystallization of the Fab fragment of one of our panel of anti-peptide antibodies (17/ 9) with and without bound peptide. Preliminary crystallographic data and binding studies indicate promise for the successful high resolution structure determination of both the native Fab and the peptide-Fab complex, aided by the deduced amino acid sequence data presented here.

EXPERIMENTAL PROCEDURES
Antibody-The monoclonal antibody 17/9 (IgG2a;~) is a subclone of monoclonal antibody H17D09, which was raised against a 36amino acid residue peptide corresponding to influenza virus hemagglutinin HA1 (residues 75-110), coupled to keyhole limpet hemocyanin as described by Niman et al., 1983. Papain Cleavage of Monoclonal Antibody 17/9-Mercuripapain (Sigma) was incubated in PBS, containing 10 mM cysteine and 2 mM EDTA, at a concentration of 1 mg/ml for 15 min at 37 "C. The activated enzyme was desalted over a Sephadex SG-25 column (1.5 X 5 cm) equilibrated in PBS and the protein concentration determined by measuring the absorbance at 280 nm (Arnon, 1970). The papain was added to the ammonium sulfate precipitated IgG, which had been dialyzed against PBS at pH 7, in a 1:50 (w/w) ratio (enzyme-toantibody). The cleavage reaction was allowed to proceed for 90 min at 37 "C and was then stopped by the addition of iodoacetamide to a final concentration of 20 mM.
Purification of the Fab Fragment-The crude papain digest was fractionated by size-exclusion chromatography on an AcA 44 column (LKB, Bromma, Sweden) (100 X 2.5 cm) with 0.1 M sodium acetate buffer, pH 5.5, at a flow rate of 0.5 ml/min. The fractional eluate was analyzed both by absorbance measurements at 280 nm and by nonreducing sodium dodecyl sulfate-gel electrophoresis. In nonreducing gels, the Fab fragment appears as a band at 43 kDa whereas the Fc fragment shows up as a 20-22 kDa band. The Fab fragment containing fractions were combined, concentrated, and dialyzed against 10 mM Tris/acetate buffer, pH 7.8, which was the starting buffer for the following anion exchange chromatography.
The 17/9 Fab was purified by ion exchange chromatography on DEAE-cellulose (DE52, Whatman, Springfield Mill, Maidstone, Kent, United Kingdom). A salt gradient from 0-0.25 M sodium acetate in 10 mM Tris/acetate, pH 7.8, was used for developing a 20-ml column with a flow rate of 1.5 ml/min. The total gradient volume was 200 ml. Under these conditions the Fab fragment elutes as one major peak with a shoulder very early in the gradient. Contaminating Fc fragments are eluted at much higher salt concentrations. The Fab fragment, eluting in the major front peak, was concentrated and dialyzed against 0.1 M sodium acetate buffer, pH 5.5, 0.02% sodium azide, for crystallization.
The purification procedures were monitored by analytical scale size exclusion HPLC as well as sodium dodecyl sulfate-gel electrophoresis. Final analysis of the purified Fab by sodium dodecyl sulfategel electrophoresis and isoelectric focusing gels revealed only one band. The Fab binding activity for peptide remained unchanged from that of the intact IgG as judged from peptide binding studies in ELISA.
ELZSA-The relative affinity of the 17/9 IgG for various overlapping peptides was determined by inhibition-ELISA analysis (Table   I). For this purpose peptide 1 (Table I), dissolved in 0.17 M borate, 0.13 M NaC1, pH 8.4 (5 pglml), was adsorbed onto microtiter plates (Immunlon 2, Dynatech Lab. Inc., Alexandria, Virginia) at 4 "C overnight. The plates were washed with deionized water and blocked with 150 pl of 1% bovine serum albumin in PBS/well for 1 h at 37 "C. The plates were again washed with deionized water before the addition of 50 p1 serial 2-fold dilutions of antibody in 1% bovine serum albumin in PBS/well. The antibody dilutions were incubated for 2 h at 37 "C. After washing with deionized water the bound antibody was detected by adding 50 pl of peroxidase conjugated goat anti-mouse IgG (Organon Technika, Cappel Division, Westchester, PA) at a 1:500 dilution in 1% bovine serum albumin in PBS for 1 h at 37 "C. Unbound conjugated antibody was removed by washing with deionized water. The amount of conjugated antibody bound/well was quantitated spectrophotometrically after 100 p1 of a solution containing 4 mg of 2,2'-azino-di-[3-ethyl-benzthiazoline sulfonate] (Boehringer Mannheim) and 3 pl of 30% hydrogen peroxide in 10 ml of 0.05 M citrate buffer, pH 4, was added to each well. The developing color was quantitated after 15 min by absorbance measurements at 415 nm.
To measure the relative affinity of the antibody for various peptides, 50 pl of a solution containing both the antibody, at a concentration corresponding to 50% binding in the ELISA assay, and different inhibitor peptides, at a range of concentrations, were added to wells which had been coated with peptide 1. Otherwise the assay procedure remained the same.
Nucleotide Sequencing-The amino acid sequence of the antibody Fab was determined by sequencing the mRNA from the hybridoma. The mRNA was isolated using standard methods (Chirgwin et al., 1979). RNA was sequenced using the dideoxy chain termination methodology (Sanger et al., 1980), using oligodeoxynucleotide primers end-labeled with 32P and reverse transcriptase (Hamlyn et al., 1978;Gellebter, 1987). In areas of ambiguity, cDNA was synthesized from mRNA using specific primers followed by sequencing using specific primers end-labeled with 3zP, DNA polymerase Klenow fragment, and Sequenase (USB, Cleveland, OH). The problems of the endogenous SP2/0-Ag14 light chain, which is essentially identical in sequence to the mouse light chain PKAPPA (11)24, (Rabbit et al., 1980),' were overcome by the synthesis of two oligonucleotides complementary to the initially ambiguous sequence in a region containing only one ambiguity. These two primers were used in combination, labeled and unlabeled, in order to increase the specificity of the sequencing, by specific competition for these homologous sites. Compressions in the sequencing were resolved using 7 M urea and 40% formamide gels. The oligodeoxynucleotide primers were synthesized on an Applied Biosystems Synthesizer and used without further purification.
Crystallization Procedures-The native 17/9 Fab and the Fabpeptide complex were crystallized using the hanging drop vapordiffusion method (for a general review of crystallization methods, see McPherson, 1982). For the native Fab, equal volumes (2 pl) of protein solution (6-12 mg/ml), and reservoir buffer, 0.2 M imidazole/malate, pH 5.6-6.5, 0.1 M NaC1, 30-39% (w/v) PEG 600, were mixed in the drop and equilibrated against 1 ml of the reservoir buffer. For the cocrystallization experiments, peptide 4 (HA1 100-108, Table I) was added in 5-fold molar excess of the Fab solution (10 mg/ml in 0.1 M sodium acetate buffer, pH 5.5). If the peptide concentration was raised to 10-15-fold molar excess, crystallization was inhibited. Aliquots of the Fab-peptide solution (2 pl) were combined with reservoir solution, 7-12% (w/v) PEG 600,0.02% sodium azide, at a ratio of 3:l or 4:l (v/v) and equilibrated against 1 ml of the reservoir buffer as before. All crystallizations were then allowed to equilibrate at 22.4 'C in a constant temperature incubator.
Attempts to diffuse peptide 4 (Table I)   3.5

RESULTS AND DISCUSSION
Peptide Binding-In order to determine the minimum-sized peptide for the crystallization experiments, we determined the relative affinities of the 17/9 IgG to a set of overlapping peptides, from the major antigenic site (HA1 residues 98-110) of the immunizing 36-amino acid peptide (Wilson et al., 1984) ( Table I). The binding studies defined the minimum antigenic determinant to be part of peptide 4 (residues 100-108)) which is in agreement with affinity measurements for other monoclonal antibodies against the same 36-mer peptide (Wilson, 1985;Wilson et al., 1986). The precise delineation of the site to residues 101-106 was determined by exhaustive peptide substitution and deletion studies by Houghten (1985;1986). Consistent with the often deleterious effect of free amino and carboxyl-terminal peptide residues on binding by an antibody, peptide 4 (residue 100-108) was shown to be the minimumsized peptide for high affinity binding (Table I, Wilson et al., 1986). The concentration of this peptide necessary for 50% inhibition in an ELISA assay when competed with the standard 23-mer (peptide 1) as absorbed antigen, is 3 X 10" M. Affinity measurements in the presence of ammonium sulfate or PEG gave results similar to those in PBS buffer? In addition, the peptide binding results were quantitatively the same in both solution immune precipitation assays, and ELISA (Wilson, 1985;Wilson et al., 1986): Native Fab 17/9 Crystals-The native Fab 17/9 crystallizes in the monotlinic space group P21 wi !h unit cell dimensions of a = 90.3 A, b = 82.9 A, c = 73.4 A, and j 3 = 122.5". The crystals typically grow to dimensions of 0.6 X 0.2 X 0.2 mm in 2-4 days (Fig. 1). An assumption of two Fab molecules in the asymmetric unit results in a packing density, V,,,, of 2.28 A3/dalton or a solvent content of approximately 46% (Matthews, 1968). The hkO zone showing the systematic absences due to the 21 screw axis along the b axis is shown in FiG. 2. A striking feature of this zone is that at low resolution (6 A) the reflections h + k = 2n are very much stronger than those where h + k = 2 n + 1. A plot of the intensity of the h + k K. F. Bergmann and I. A. Wilson, unpublished data. even versus h + k od$ reflections shows this trend is significant to around 4.5 A resolution. These pseudoextinctions suggest that the second molecule in the asymmetric unit is centered in the a b projection with respect to the first molecule. Precession photos of the h01 zone and the Okl zone also show pseudoextinctions at low resolution such that h = 2n and k = 2n, respectively. These special conditions indicate that the spacegroup can be considered as C121 a t low resolution where the two molecules in the P21 asymmetric unit are located at (x, y, z ) and ( x + Y2, y + Y2, z).
The crystals are exceptionally well-ordered diffracting to 9 I FIG. 1. Photomicrograph of a native Fab 17/9 crystal. These high quality prismatic crystals were grown from PEG 600 imidazolemalate buffer, pH 5.6-6.5 as described unde: "Experimental Procedures." These crystals diffract to at least 1.9 A resolution. The crystal shown is 0.6 X 0.2 X 0.2 mm and the photomicrograph was taken a t X 60 magnification between crossed polarisers. better than 1.9 A resolution as determined from oscillation photographs taken at the Stanford synchrotron p-ray laboratory (Fig. 3). X-ray crystallographic data to 2.0 A resolution have been collected on a Nicolet-Xentronics area detector and reduced to give an unweighted absolute R-factor on intensities of 8.6%. We are currently attempting to solve this structure by molecular replacement, using the five available Fab coordinate sets in the Brookhaven data bank, as has been successfully reported for four Fab structures (Cygler et al., 1987(Cygler et al., , 1988Sheriff et al., 1987;Vitali et al., 1987;Prasad et al., 1988). However, several of the peptide antigens have been iodinated or mercurated and can be used as heavy atom derivatives if this should prove necessary. 5. 1 ' oscillation Dha " complex rystal taken in a general orientation at the Stanford synhrotron radiation laboratory. The exposure time was 274 s at 47 IA, 3 GeV, using ap.2-mm collimator, with a crystal-to-film distance f 100 mm at 1.08 A wavelength. The dimensions of the crystal were .5 y 0.04 X 0.01 mm. The reflections at the edge of the film are a t .1 A resolution. Photomicrograph of an Fab-peptide complex crystal. These crystals grow as long, thin plate-like rods in sodium acetate buffer, pH 5.5, from a much lower PEG 600 concentration than the native Fab crystals, as described in the text. The crystals are monoclinic but have differen$ unit cell parameters than the native. The crystals diffract to 2.1 A (see Fig. 5). The crystal shown has dimensions of 1.0 x 0.02 x 0.01 mm.

FIG.
1719 Fab-Peptide Complex Crystals-The crystals grown from the Fab-peptide solutions are also in tbe space g r p p P2, but yith unit cell dimensions of a = 63.9 A, b = 73.0 A, c = 49.1 A, and @ = 120.6 '. It should be noted that both the Fab-peptide complex and the native molecule crystallize in unit cells having one edge close to the 72.4 A dimension common to X light chain dimer crystals (Schiffer et al., 1985).
For these crystals the constant domains form an "infinite" psheet pattern resulting in the 72.4 A repeat. Conceivably,

M E T S e r T r p v a l A r g G l n T h r P r o A s p L y s A r g L e u G l u T r p V a l A~~T h r I~~~~~A S~G~Y~~Y~~Y 40 50
A T y r T h r T y r T y r P r o A s p S e r v a 1 L y s G l y A r g P h e T h r I l e S e r A r g A s P~s~~a L Y s~s~~~~~~~ TACACCTACTATCCAGACAGTGTGAAGGGGGCGATTCACCATCTCCAGAGAC~TGCCAAG~CACCCTG 60 similar interactions could be occurring with both our two Fab crystal forms. The Fab-peptide complex crystals grow as long thin rods with dimensions of 1.0 X 0.02 X 0.01 mm within 1-2 weeks (Fig. 4). One Fab molecule inothe asymmetric unit results in a packing density, Vm, of 1.97 A3/dalton, or solvent content of 38%. The crystals diffract to at least 2.1 A resolution as determined at the Stanford synchrotron radiation laboratory (SSRL) (Fig. 5). A 7" precession photo of the Okl zone is shown in Fig. 6. These results suggest that a high resolution crystallographic analysis of these cocrystals will be possible which will provide an independent solution of the Fab-peptide complex. Attempts to obtain larger crystals are in progress.

GACGAGCGGGTTTGCTTACTGGGGCCAAGGGGACTCTGGTCACGGTCTCTGCAGCC~C~CAGCC ~~p~l u A~n G l y P h e A l a T y r T r p G l y G l n G l y T h r L e u V a l T h r V~~S e r A~~~a L Y s T h~~h r A~~ loo
In order to confirm the presence of peptide in the cocrystals, various control experiments were performed. Crystallization experiments with either Fab or peptide alone under the crystallization conditions, characteristic for the assumed Fabpeptide complexes gave no crystals. Reverse-phase HPLC analysis of complex crystals, which were washed twice with the well solution and then dissolved in 0.1% trifluoroacetic acid, showed an approximate 1:l (mol/mol) ratio of peptide to Fab. It could of course be argued that the peptide is simply trapped in the solvent channels and not specifically bound in the antibody-binding site. However, the relatively high affinity of the peptide for the antigen even in high salt or PEG3 suggests that specific Fab-peptide complexes predominate in the crystal.
Analysis of Gene and Protein Sequence-The cDNA sequences determined for the light and heavy chains are shown in Table 11. The VH sequence demonstrated that this was part of the VH 7183 gene family, one of the two VH region families proximal to the J H D region (Alt et al., 1987). Those sequences found to show the highest similarity for the VH region were that of an anti-dinitrophenyl antibody (Riley et al., 1986) a VH 7183 gene member, (Yancopoulos et al., 1984) an anti-CEA antibody (Cabilly et al., 1984), an anti-SRBC antibody (0110 et al., 1984) and an anti-@-(1,6)galactan (Hartmann and Rudikoff, 1984) VH gene. The analysis of the V, region demonstrated only two nucleotide sequences with significant similarity. The anti-lysozyme loop (gloop) antibody (Darsley and Rees, 1985) and the anti-dinitrophenyl antibody (Riley et al., 1986) showed 87 and 69% similarity, respectively. Comparison of the amino acid sequence of Fab 17/9 with that of the sequences of the Fabs for which crystal structures are known indicate that McPC603 (Satow et al., 1986) to be the best model for molecular replacement for the light (86%) and heavy (58%) chain variable regions, VI, and VH, although 5539 does have a slightly better agreement in the VH region (69%), while HyHEL-5 (Sheriff et al., 1987) is the best model for the constant regions, (CL (100%) and CHI (85%)).
of an Anti-peptide Fab