Surface-exposed Amino Acid Residues of HPV16 L1 Protein Mediating Interaction with Cell Surface Heparan Sulfate*

Efficient infection of cells by human papillomaviruses (HPVs) and pseudovirions requires primary interaction with cell surface proteoglycans with apparent preference for species carrying heparan sulfate (HS) side chains. To identify residues contributing to virus/cell interaction, we performed point mutational analysis of the HPV16 major capsid protein, L1, targeting surface-exposed amino acid residues. Replacement of lysine residues 278, 356, or 361 for alanine reduced cell binding and infectivity of pseudovirions. Various combinations of these amino acid exchanges further decreased cell attachment and infectivity with residual infectivity of less than 5% for the triple mutant, suggesting that these lysine residues cooperate in HS binding. Single, double, or triple exchanges for arginine did not impair infectivity, demonstrating that interaction is dependent on charge distribution rather than sequence-specific. The lysine residues are located within a pocket on the capsomere surface, which was previously proposed as the putative receptor binding site. Fab fragments of binding-neutralizing antibody H16.56E that recognize an epitope directly adjacent to lysine residues strongly reduced HS-mediated cell binding, further corroborating our findings. In contrast, mutation of basic surface residues located in the cleft between capsomeres outside this pocket did not significantly reduce interaction with HS or resulted in assembly-deficient proteins. Computer-simulated heparin docking suggested that all three lysine residues can form hydrogen bonds with 2-O-, 6-O-, and N-sulfate groups of a single HS molecule with a minimal saccharide domain length of eight monomer units. This prediction was experimentally confirmed in binding experiments using capsid protein, heparin molecules of defined length, and sulfate group modifications.

Human papillomaviruses (HPVs) 2 are non-enveloped epitheliotropic viruses mainly causing benign tumors of the skin and mucosa. Some of them, including HPV type 16 (HPV16), are the primary etiologic agent for anogenital tumors, especially cervical carcinoma (1). To initiate a successful life cycle, viruses need to attach to host cells. This is achieved by binding to cell surface receptor molecules, which can be proteins or sugar or lipid components. Specific attachment will then initiate internalization, which is followed by amplification of the viral genome and finally production of progeny virus. Certain HPVs of the genus of ␣-Papillomavirus have been shown to use cell surface proteoglycans as initial attachment receptors; the target glycosaminoglycan (GAG) chains have been tentatively identified as heparan sulfate (HS) (2)(3)(4).
Cell surface heparan sulfate proteoglycans (HSPGs), mainly syndecans and glypicans, are composed of a protein core with covalently attached HS chains. These unbranched polysaccharides are generated through a process initiated by polymerization of glucuronic acid and N-acetylglucosamine units in alternating sequence (5). This backbone undergoes a series of modifications, including N-deacetylation and N-sulfation of the glucosamine units, C-5 epimerization of glucuronic to iduronic acid residues, and finally O-sulfation at the 2-O-position of hexuronic acid and at the 3-O-and 6-O-positions of glucosamine units (5). The modification reactions are incomplete yet strictly regulated and give rise to a large number of distinct HS species that appear to be cell/tissue-specific. HS chains serve as ligands for a large number of proteins and thus influence a wealth of biological processes of importance to development and homeostasis. However, they have also been implicated as receptors for a growing number of viruses including dengue virus (6,7), herpes simplex virus type 1 (HSV-1) and HSV-2 (8,9), human immunodeficiency virus type 1 (10,11), adeno-associated virus type 2 (12)(13)(14) and adeno-associated virus type 3 (15), respiratory syncytial virus (16), pseudorabies virus (17), vaccinia virus (18,19), Sindbis virus (20), echovirus (21), coxsackievirus B3 (22), and human cytomegalovirus (23).
Whereas the protein core of HSPG does not seem to contribute significantly to the interaction with HPV (24), modifica-* This work was supported by Deutsche Forschungsgemeinschaft Grant SFB490/E2 (to M. S. and R. E. S.), by the National Center for Research Resources, a component of the National Institutes of Health (Grant P20-RR018724, entitled "Center for Molecular and Tumor Virology"), and by Swedish Research Council Grant 15023, the Swedish Cancer Society, Swedish Foundation for Strategic Research Grant A303:156e, and Polysackaridforskning AB (Uppsala, Sweden). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. tions by sulfate groups have been shown to be essential for HPV11, -16, and -33 capsid interaction with cells (2,3), suggesting the involvement of electrostatic interactions of basic amino acid residues with the negatively charged sulfate groups. Not much is known about the viral components contributing to the interaction with HSPG. The papillomavirus shell is composed of 360 copies of the major capsid protein, L1, organized into 72 pentameric capsomeres and a small number of the minor capsid protein, L2. This protein shell encapsidates a circularized double-stranded DNA genome of ϳ8000 bp. Because HPVs are difficult to propagate in vitro due to their dependence on terminally differentiating keratinocytes for completion of their life cycle, DNA-free virus-like particles (VLPs) and DNA-containing pseudoviruses have been used to study HPV/cell interaction (25). L2 protein does not seem to contribute to the initial binding to HSPG as VLPs and pseudovirions composed of L1 protein alone can bind to cell surface-exposed proteoglycan as well as to heparin, a highly sulfated form of HS secreted by mast cells. Interaction of HPV with heparin requires an intact outer surface conformation (26,27) suggesting that the cluster of basic amino acids, as usually required for this type of interaction, is conformational rather than linear.
To identify regions of L1 protein involved in GAG binding, we performed site-directed mutagenesis of basic and polar residues located on the capsid surface of HPV16. We also determined the minimal saccharide length required for binding to papillomavirus capsids and evaluated the contribution of variously positioned sulfate groups. Moreover a molecular model is presented to illustrate how three amino acid residues, lysines 278, 356, and 361, may collectively mediate the interaction of HPV with HS-related GAG.

EXPERIMENTAL PROCEDURES
Plasmids and Mutagenesis-Expression plasmids carrying codon-modified HPV16 L1 and L2 structural genes have been reported previously (28). A marker plasmid coding for a dimeric green fluorescent protein (2) was co-transfected as a control for efficient transfection and as a reporter plasmid in infectivity assays. For each HPV16 L1 exchange mutant, two complementary PCR primers of 30 -40 nucleotides were used to introduce an alanine (GCC) or arginine (CGG) instead of a lysine, arginine, or threonine codon. During PCR the whole pUF3 plasmid was amplified, then the PCR products were digested with DpnI to remove methylated template DNA, and the remaining mutant plasmids were transformed. The mutations were confirmed by DNA sequencing.
Cell Culture-COS-7 cells were maintained at 37°C and 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics. The human embryonic kidney cell line 293TT was obtained from Buck et al. (29) and grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1% Glutamax I, 1% modified Eagle's medium nonessential amino acids, and antibiotics.
Preparation of HPV16 Virus-like Particles and Pseudovirions-Wild-type (wt) HPV16 VLPs were produced using the baculovirus expression as described previously (30). HPV16 pseudovirions were generated as described previously (29,31). Briefly expression plasmids carrying codon-optimized wt or mutant HPV16 L1 genes, respectively, were co-transfected with a codon-modified L2 expression plasmid and a marker plasmid coding for green fluorescent protein. After 48 h, the transfected 293TT cells were suspended in Dulbecco's phosphate-buffered saline (PBS) containing calcium and magnesium (DPBS, Invitrogen) supplemented with 9.5 mM MgCl 2 . After sonication for 20 s, 0.5% non-ionic detergent Brij 58 (Sigma) and 0.2% Benzonase (Sigma) were added, and the samples were incubated at 37°C for 16 -20 h. The nuclease-digested lysates were chilled on ice, adjusted to 0.8 M NaCl, incubated on ice for 10 min, and then clarified by centrifugation at 2,000 ϫ g for 10 min at 4°C. The capsids were purified from these clarified cell lysates as described previously (29) via ultracentrifugation through OptiPrep gradients. After centrifugation, fractions were collected and tested on SDS-PAGE, and gradient fractions containing appreciable amounts of capsid proteins were used for analysis. Pseudoviruses were quantitated by Western blot and conformation-dependent enzyme-linked immunosorbent assay (ELISA).
Cell Binding Assay-COS-7 cells were grown to confluence as adherent monolayers in 6-well plates and incubated with mutant or wild-type HPV16 pseudovirions in Dulbecco's modified Eagle's medium for 2 h at 4°C (34). Binding of pseudovirions was determined in the presence or absence of antibody H16.56E or its Fab fragments. To generate these Fab fragments, antibody-containing cell culture supernatants (100 l) were digested for 90 min with 3 l of papain (P3125, Sigma; 27 g/ml) in the presence of 10 mM cysteine basically following the procedure described by Johnstone and Thorpe (35). Full antibody controls were mock-treated for 90 min in the absence of papain. Subsequently both Fab fragments and undigested H16.56E antibodies were used in cell binding assays. After binding, cells were extensively washed with PBS to remove unbound pseudovirions, lysed in Laemmli sample buffer, incubated at 100°C for 5 min, and subjected to SDS-PAGE and immunoblotting for the detection of cell surface-bound HPV16 L1 protein.
Pseudovirus Infection Assay-293TT cells were grown in 24-well plates and infected with mutant and wt pseudovirions from OptiPrep gradients (5000 -8000 infectious units/well for wt and corresponding amounts of assembled L1 protein for mutants) in a total volume of 1 ml of Dulbecco's modified Eagle's medium containing supplements. The cells were grown at 37°C for 72 h before manually counting cells with nuclear green fluorescence to determine infectious events.
ELISA-ELISA microtiter plates (BD TM Heparin Binding Plate, BD Biosciences) were coated with 25 g/ml heparin (H4784, Sigma) or heparan sulfate (H9902, Sigma) overnight at room temperature. Subsequently the plates were washed with PBS, 0.2% Tween 20 (PBST). Free binding sites were blocked with 2% milk in PBST for 1 h at 37°C. Mutant and wt pseudovirions were diluted in PBS (0.5 g/ml) in the absence or presence of 250 g/ml heparin (H4784, Sigma). After preincubation for 30 min at room temperature, pseudovirions were added to the plates in replicates and bound for 1 h at 37°C. Subsequently wells were incubated with primary antibody solution (polyclonal antiserum K75) for 1 h at 37°C. Bound pseudovirions were detected by the addition of horseradish peroxidase-coupled secondary antibody for 30 min at 37°C. The assays were developed with trimethylbenzidine (Moss, Inc.) and stopped with 1 N HCl. Absorbance was measured at 450 nm using a Multiscan RC plate reader (Thermo Fisher Scientific). Heparin and heparan sulfate binding was normalized with an ELISA detecting directly coupled pseudovirions that was always run in parallel as described previously (26) using identical amounts of pseudovirions and the same antibody dilutions.
Electron Microscopy-The wt and mutant pseudovirions were purified from OptiPrep gradient fractions using Sephadex G-50 Quick Spin columns (Roche Applied Science). The purified pseudovirions were directly spotted onto carbon-coated copper grids, washed with water three times, and negatively stained with 2% phosphotungstic acid. Grids were observed, and photomicrographs were obtained with a Zeiss EM900 transmission electron microscope or a Philips Tecnai 12 transmission electron microscope at an instrumental magnification of ϫ50,000.
Graphics and Docking Simulation-The crystal structure of HPV16 L1 particle (36) was used for docking a heparin 14-mer onto the L1 pentamer surface using the program O. After docking, an energy minimization was performed to optimize the interactions in the model.
Radiolabeling of Heparin and Heparin Fragments-Bovine lung heparin (The Upjohn Co.) was purified as described previously (37) and N-[ 3 H]acetylated with [ 3 H]acetic anhydride (500 mCi/mmol; Amersham Biosciences) at free amino groups as described before (38) to a specific activity of 90,000 dpm 3 H/nmol of heparin. Radiolabeled, size-defined heparin oligosaccharides were produced by partial deamination with HNO 2 , pH 1.5, reduced with NaB 3 H 4 (20 -25 Ci/mmol; Amersham Biosciences) to a specific activity of 0.5 ϫ 10 6 dpm/nmol oligosaccharide, and separated according to size on a Bio-Gel P-10 (Bio-Rad) column (1 ϫ 146 cm) as described before (39). Polysaccharide concentration was assessed by colorimetric determination of hexuronic acid using the meta-hydroxydiphenyl method (40) with glucuronic acid as a standard. A factor of 3 was arbitrarily used to convert values to saccharide mass.
Chemical Modification of Heparin-Selective chemical desulfation of heparin was performed as described previously (39,41,42). In short, the pyridinium salt of heparin was N-desulfated by solvolysis in Me 2 SO/H 2 O (43) and N-acetylated (44) to yield N-desulfated heparin. A 6-O-desulfated sample was obtained by treatment of the pyridinium salt with N-methyl-N-(trimethylsilyl)trifluoroacetamide at 110°C (42). For 2-O-desulfation the sodium salt of heparin was adjusted to pH 12.5 and lyophilized (45) followed by re-N-sulfation (2-O-desulfated heparin) or complete N-desulfation and N-acetylation (N-/2-O-desulfated heparin) as described before (41). All samples were analyzed as described previously (39) with regard to molecular size by gel chromatography or disaccharide composition by anion exchange high pressure liquid chromatography following deaminative cleavage.
Interaction of HPV16L1L2 with Heparin and Size-defined Heparin Oligosaccharides-3 H-Radiolabeled heparin (ϳ0.5 g) or size-defined oligosaccharides (4 -26-mer; ϳ3 pmol of fragments each) were incubated with 5 g of HPV16L1L2 VLPs in 200 l of PBS for 1 h at room temperature as described previously (46). Protein along with any bound oligosaccharides was trapped on a nitrocellulose filter (diameter, 25 mm; Sartorius). The filter was washed with the incubation buffer, proteinbound oligosaccharides were dissociated from the membrane in 2 M NaCl, and the radioactivity was measured by scintillation counting. For competition assays selectively desulfated heparin preparations were added at the indicated concentrations together with 3 H-labeled heparin (ϳ0.5 g), and the assay was performed as described above. VLPs instead of pseudovirions were used because of easier availability of large quantities. Key experiments were repeated with pseudovirions as stated in the text.

Surface-exposed Basic Amino Acid Residues of HPV16 L1-
The attachment of viruses to cell surface proteoglycans is mainly governed by electrostatic interactions of basic amino acids with the negatively charged sulfate and carboxyl groups of the GAG side chains. A preliminary analysis of the electrostatic potential of the HPV16 capsomere surface using the Insight2000 software package revealed the presence of 15 highly positive charged amino acid residues present on the tip of the capsomere. These are lysine residues at HPV16 L1 positions 278, 356, and 361 and are contributed by the five L1 molecules found in a capsomere (see Fig. 4A). Additional surfaceexposed basic amino acid residues are found in the BC loop (Lys-54, Lys-55, and Lys-59) located at the upper outer rim of the capsomere and deep in the cleft between capsomeres (Arg-74, Lys-82, Arg-97, and Lys-443).
Virion Assembly of L1 Mutants-To investigate the role of positively charged amino acid residues in papillomavirus cell interaction, a point mutational analysis was performed. Using codon-optimized HPV16 L1, we substituted the residues for alanine or arginine as single, double, and triple mutants. These mutants were co-transfected with wt HPV16 L2 and a green fluorescent protein-based reporter plasmid into 293TT cells. Under these conditions, L1 and L2 self-assemble into pseudovirions and encapsidate the reporter plasmid, allowing easy scoring of infection by counting green fluorescent proteinexpressing cells (29). Pseudoviruses were extracted and purified by OptiPrep gradient centrifugation, and fractions were analyzed for the presence of L1 and L2 by Western blot analysis. Mutants K54A, R74A, and K82A were assembly-deficient as evidenced by the lack of L1 and L2 in pseudovirus-harboring fractions, although all were expressed at levels comparable to wt L1 after transfection of cells. In addition, viral particles could not be detected by electron microscopy (data not shown). Therefore, these mutants had to be excluded from further analysis. Analysis of the additional lysines in the BC loop (Lys-55 and Lys-59) was not pursued because antibody binding to this loop does not block cell attachment and binding to heparincoated ELISA plates (47). As shown in Fig. 1A, all other mutants yielded pseudoviruses. L2 protein was found co-migrating with L1 proteins in these gradients, indicative of particle formation with incorporated L2 protein. The level of L2 incorporation was similar for all wt and mutant capsids, although the yield of pseudoviruses differed somewhat between preparations due to differences in transfection efficiencies. The formation of intact particles was further confirmed by probing particles with conformation-dependent antibodies (not shown) and by electron microscopy (Fig. 1B). At 50,000-fold magnification no significant differences were observed between wt and mutant pseudovirions. All HPV particles, containing either single, double, or triple lysine or arginine to alanine mutations, were of the same size as wt particles and displayed the characteristic features of the icosahedral HPV capsids.
Cell Binding of Mutant Pseudovirions-To analyze the attachment of wt and mutant HPV pseudovirions to cell surface GAGs, binding studies were performed. Monolayer COS-7 cells were incubated with similar amounts of pseudovirions at subsaturating levels, and bound particles were detected by L1-specific Western blot analysis performed essentially as described before (34). Because it is important for these assays to utilize similar amounts of pseudovirions, pseudovirus preparations were normalized by L1-specific Western blot and ELISA using HPV16 VLP antisera that detect conformationally intact L1 protein. A representative binding assay is depicted in Fig. 2A. Strongest binding was observed with wt pseudovirions. Mutant R97A displayed binding comparable to wt pseudovirions, whereas the lysine to alanine single mutants K278A, K356A, and K361A showed a reduced binding to cells. Among these, K356A binding was most prominent in all assays, whereas K361A displayed the lowest binding. Replacement of two (K278A/K361A and K356A/K361A) and three lysine residues (K278A/K356A/ K361A) resulted in an even further reduction in cell binding, suggesting that these lysine residues, which are located in close proximity on the capsomere tip, may cooperatively contribute to interaction with the cell surface. With mutant K443A only a slight reduction was observed. Heparin drastically decreased cell binding of wt and mutant pseudovirions as expected for interaction with cell surface proteoglycan. Nevertheless a low background binding was observed even in the presence of heparin (Fig. 2B). Because heparin blocks infection completely under the chosen conditions (48), residual binding is probably unspecific and does not lead to infection. Quantification of several binding assays is presented in Fig. 2C. Using heparin-and heparan sulfate-coated ELISA plates, we confirmed that the mutations resulting in impaired binding of pseudovirions to cells do indeed preclude interaction with the type of cell surface polysaccharides implicated (Fig. 2, D and E).
Infectivity of Mutant Pseudovirions-To measure infectivity of mutant pseudoviruses, monolayer 293TT cells were incubated with OptiPrep fractions of wt and mutant pseudovirions, and infection was scored 72 h later by counting cells with green nuclear fluorescence. Infectious events obtained with wt and mutant pseudovirions were normalized to the amount of capsid protein L1 in these particles. Infectivity of wt virions was set to 100% (Fig. 3A). Significant difference between the groups was determined using a one-way analysis of variance (p Ͻ 0.001). As a "posthoc" multiple comparison, a Student-Newman-Keuls procedure was used to compare each group mean with each other group mean. It yielded a significant difference of all mutants compared with wt except for particles carrying the arginine to alanine mutation at position 97 (p Ͻ 0.001). R97A particles, which attached normally to cells, only showed a slightly reduced infectivity compared with wt particles. In contrast, particles with single point mutations in lysine residues 278, 356, and 361 displayed reduced infectivity, which correlated well with their impaired cell binding ability; K361A displayed the lowest infectivity (21%), whereas K278A and K356A retained 36% infectivity. In line with the cell binding analysis, both double mutants analyzed, K278A/K361A and K356A/ K361A, showed an even further reduction of infectivity to 13 and 18%, respectively. Infectivity of the triple mutant K278A/ K356A/K361A was reduced to 5%, again suggesting that the lysine residues 278, 356, and 361 cooperatively contribute to interaction with GAG. Interestingly the exchange of lysine 443 led to decreased infectivity down to merely 10%, although cell binding of K443A virions was only slightly reduced. The lack of a close correlation between cell binding and infectivity of this mutant suggests that lysine residue 443 may not contribute to the primary attachment to HSPGs but may be important for subsequent steps of infection. Based on the atomic model proposed by Modis et al. (49), Lys-443 lies at the vertex of the capsomere halfway down to the floor of the capsomere and thus is not in close vicinity to Lys-278, Lys-356, and Lys-361. Exchange of single, double, or triple lysine residues 278, 356, and 361 for arginine did not reduce infectivity of pseudovirions, suggesting that overall charge distribution rather than sequence specificity determines binding to GAGs (Fig. 3B).
We also investigated a possible role of three threonine residues at positions 353, 354, and 358, which are located right in between the three lysine residues 278, 356, and 361. Heparin binding domains often contain polar amino acid residues that can stabilize heparin binding via hydrogen bond formation (40). In the case of HPV16, pseudovirions with threonine to alanine exchanges at these positions neither displayed reduced cell binding (data not shown) nor reduced infectivity (Fig. 3B), indicating that these residues do not significantly contribute to cell binding.
Lys-361 is in direct vicinity to the FG loop (amino acids 260 -282), which is located between Lys-361 and Lys-278 (Fig. 4A) (36). The FG loop is recognized by virus-neutralizing antibodies. One neutralizing monoclonal antibody with specificity for experiments. B, infectivity of mutant pseudovirions with threonine to alanine or lysine to arginine mutations. C, wt HPV16 virions were incubated with mouse monoclonal antibody H16.56E or its Fab fragment and subsequently bound to cells. Bound pseudovirions were detected as described in Fig. 2A. Ab, antibody. Error bars indicate standard deviations of at least six independent experiments. FG loop amino acids 260 -270, H16.56E, blocks binding of HPV16 particles to the cell surface (Fig. 3C) as well as to heparin-coated microtiter plates (47), supporting our finding that HS interacts with the tip of capsomeres. To obtain further evidence for the involvement of the identified lysine residues in cell attachment, we tested Fab fragments of H16.56E for their inhibitory effect on cell binding. We reasoned that removal of the bulky Fc portion of the antibody should reduce the likelihood of steric hindrance at distant sites. As shown in Fig. 3C, Fab fragments of H16.56E strongly reduced attachment of wt HPV16 pseudovirions to COS-7 cells albeit at a somewhat reduced level compared with untreated antibodies. The slightly reduced efficacy is most probably due to a reduction in binding avidity of monovalent Fab fragments. These data strongly suggest that amino acid residues involved in cell attachment are in close vicinity to the H16.56E epitope and are thus in line with our mutational analysis.
Computer Model of Heparin/HPV Interaction-A computersimulated heparin docking experiment, based on the structural data of the papillomavirus 16 capsomere surface (36,49) and the proposed three-dimensional structure of heparin (50), revealed a binding mode in which a heparin molecule can simultaneously interact with three positively charged L1 amino acid residues within the pocket on the capsomere surface, Lys-278, Lys-356, and Lys-361 (Fig. 4, A and B). Previous experiments had suggested the involvement of sulfate residues in the HPV33/HSPG interaction (2). The model of heparin docking to the HPV16 capsomere supported these data and suggested that a saccharide sequence of eight or more residues is required to span the three lysine residues within the putative receptor binding pocket (Fig. 4B). The model also suggests that within such a sequence of contiguous, fully sulfated disaccharide units all three types of major sulfate substituents (iduronic acid 2-O-sulfate, glucosamine N-sulfate, and 6-O-sulfate) (Fig. 4B) are amenable to interaction with the lysine units (see also "Discussion").
Length Requirement for Heparin Binding-To further corroborate these observations, 3 H-labeled heparin oligosaccharides with defined lengths were incubated with HPV16 L1L2 VLPs, and protein-bound saccharide was recovered by nitrocellulose trapping. The smallest heparin fragment retaining the ability to bind VLPs was an octamer (Fig. 5), consistent with the modeling result. The proportion of bound oligosaccharide gradually increased when longer heparin was used. However, a binding plateau was reached when the heparin length reached 14-mer or longer. The amount of heparin bound under these conditions corresponded to ϳ0.9 14-mer per L1 molecule, thus approaching saturation. Similar size requirements were observed using affinity chromatography of labeled oligomers on HPV16 L1L2 VLP-Sepharose columns (data not shown).
Sulfate Groups of Heparin Important for Binding HPV16 Capsids-The contribution of various sulfate groups to the interaction between VLPs and heparin was also verified. Different unlabeled, selectively desulfated heparin preparations were incubated with 3 H-labeled heparin and the papillomavirus particles. The protein-bound fraction was again recovered on a nitrocellulose filter, and bound radioactivity was quantified. As shown in Fig. 6, a 10-fold excess of unlabeled, unmodified hep-arin completely abrogated binding of labeled heparin. In contrast, no significant reduction of bound label was observed when up to a 200-fold excess of 6-O-desulfated heparin was added to the mixture. Similar results, i.e. no competition, were obtained with heparin lacking both N-and 2-O-sulfate groups. However, heparin lacking either N-or 2-O-sulfate groups remained somewhat competitive although much less so than the fully sulfated polysaccharide. A maximal reduction of ϳ70% of [ 3 H]heparin binding was attained in the presence of excess amounts of either 2-O-or N-desulfated heparins, respectively. Similar results were obtained with HPV16 L1L2 pseudovirions (data not shown). These results indicate that 2-O-sulfate groups (located primarily on iduronic acid residues in heparin and heparan sulfate), glucosamine N-sulfate, and in particular

DISCUSSION
Several lines of evidence suggest that the primary cell attachment site of HPV particles is situated on the tip of capsomeres. Using cryoelectron microscopy and image reconstruction, Booy et al. (51) demonstrated that the bovine papillomavirus type 1-specific cell binding-neutralizing antibody 9 binds to the tip of capsomeres of bovine papillomavirus type 1 virions, whereas the neutralizing antibody 5B6, which does not block cell attachment, binds to the vertex of hexavalent capsomeres. Although this finding does not exclude that attachment occurs via the vertex of pentavalent capsomeres, such interaction seems unlikely. First, neutralizing antibodies that recognize epitopes in the cleft between capsomeres, either located in the BC loop or the carboxyl-terminal arm, do not block binding (27,52). Second, all tested monoclonal antibodies that bind to the DE, FG, and/or HI loops, all of which are found on the tip of capsomeres, interfere with the primary cell attachment (27,52). Third, Fab fragments of the HPV16-specific antibody H16.56E also interfere with virus/cell attachment, suggesting that the cell attachment site is in close proximity to the FG loop. Our mutational analysis supports this notion. Exchange of lysine residues at position 278, 356, or 361 for alanine as single, double, or triple mutants did not impair the formation of intact HPV16 virus particles but affected binding of pseudoviruses to cells and HS-coated substrates to various degrees. Compared with single point mutations, double and triple mutations displayed a further reduction in HS binding and infectivity, indicating that these basic residues cooperate in interacting with HS. This is expected for the mainly charge-driven interaction of HS and its ligands. For these mutants, we observed a good correlation of reduction in HS binding (measured by interaction with both cells and HS-coated substrate) and infectivity, suggesting that the primary reason for reduced infectivity is impairment of attachment to the cell surface mediated by HS molecules.
Our data suggest that two basic residues (Arg-97 and Lys-443) located outside the putative binding pocket but in the cleft between capsomeres do not participate in initial binding events. The phenotype of K443A mutant pseudovirions is especially interesting. Lys-443 seems to play an important role in infection as evidenced by a 90% drop in infectivity of mutant K443A pseudovirions, although the primary binding to the cell surface was only marginally reduced. It is tempting to speculate that Lys-443 may play a role in secondary binding events, which could include additional interactions with cell surface HS or HS-independent receptors. Indeed our data showing that interaction of heparin with VLPs was only partly abolished by removal of either 2-O-or N-sulfate (Fig. 6) suggest the presence of two HS binding sites, which is in line with previously published observations (27). Candidates for interaction with non-HS binding sites include the extracellular matrix protein laminin 5 and ␣6 integrin, which have both been implicated in playing a role in the HPV infection process (53,54). Other surface-exposed basic amino acid residues like Lys-54, Arg-74, and Lys-82 had to be excluded from the current study because replacement by alanine resulted in assembly-deficient L1 protein. Their possible involvement in virus/cell interactions remains speculative. The three lysine residues Lys-278, Lys-356, and Lys-361, which were shown to be important for HPV binding and infection, are arranged in a cluster of basic amino acid residues and form a positively charged flat pocket on the surface of the capsomere. A Clustal analysis (data not shown) revealed that Lys-356 and Lys-361 as well as Arg-74, Lys-82, and Arg-97 are found in each member of the human papillomavirus family causing malignant mucosal lesions (species 9 (55)). Lys-278 is present in six of seven members, Lys-443 is present in only five of seven members, and Lys-54 is found in a total of only three species 9 HPV types. Within the genus ␣-Papillomavirus, Lys-278 and Lys-361 are also highly conserved, confirming the importance of these basic residues. Correlating with the observed weak effect of K356A on cell binding, the lysine at position 356 is significantly less conserved. Although the here identified lysine residues are not fully conserved among the family of papillomaviruses, they may be functionally replaced by other basic amino acid residues in close proximity. Bovine papillomavirus types 1 and 2, for example, have an asparagine residue at the position corresponding to HPV16 Lys-361 but have an additional lysine residue at a position corresponding to HPV16 amino acid 359, which is surface-exposed in the vicinity of Lys-361. This would be consistent with our observation that interaction of HPV with HS is not peptide sequence-specific but rather depends on charge distribution (Fig. 3B).
Numerous viruses bind to glycosaminoglycans of HSPGs, and many of them feature binding domains consisting of clustered basic amino acid residues located in shallow depressions on the virus surface (56 -59). This seems to be characteristic for the interaction of viruses with sugar residues. Polyomaviruses, which are structurally related to papillomaviruses, display a similar pattern of receptor binding sites for sialic acids (60). In both viruses, neighboring L1/VP1 monomers cooperatively form the binding site in a shallow groove.
To investigate the putative heparin binding sites on the papillomavirus capsomere surface in more detail, a computer-simulated docking experiment was performed. Based on the severe effects of mutations at Lys-278, Lys-356, and Lys-361 on cell binding and infectivity, molecular modeling was used to examine the possible interactions of a heparin oligomer and these amino acids based on the x-ray crystal structure of HPV16 small virus-like particles (36). The model demonstrates that all three amino acid units can form hydrogen bonds simultaneously with the sulfate oxygens or hydroxyl groups of a heparin molecule. Such concerted interaction is achieved with a minimum of eight monosaccharide units. Heparin molecules longer than eight units can slide along the three lysines, increasing the chance of being bound in the pocket (Fig. 4, A and B). However, as the heparin molecule gets longer, the entropy of binding will increase by the dangling ends, which will eventually offset the increased docking probability by a longer heparin chain and even reduce the binding when heparin length becomes too long (16-mer or longer). This modeling result agrees well with our heparin binding assays in which eight monosaccharide units were sufficient for binding of HPV16 L1L2 VLPs, and the binding increased as the heparin chain increased in size from eight to 14 units (Fig. 5). With 16 or more units, the binding starts to decrease, possibly due to the extra entropy at both ends on a bound heparin molecule. This phenomenon occurs within the same range as previously observed for HSV-1 (9) and pseudorabies virus (17).
The requirement of a combination of sulfate residues at different positions in GAG chains was reported previously for the interaction with pseudorabies virus glycoprotein C (17). For HSV-1, 2-O-and 6-O-sulfation also appeared to be required for binding to cells (41), whereas HSV-1 internalization and infection was promoted by the presence of 3-O-sulfated glucosamine residues (8,61). Our findings support the notion (2,27) that sulfated domains within HS chains may provide primary receptors for HPV that are essential to cell binding and infection. The proposed lysine residues located within the pocket on the surface of HPV capsid protein L1 may form a binding domain for these primary receptor molecules. The model interaction with heparin implicates the three major types of sulfate substituents in binding to the critical lysine residues Lys-278, Lys-356, and Lys-361 (Fig. 4B) in accord with the competitory effects of variously desulfated heparins on native, radiolabeled heparin binding to VLPs (Fig. 6). This observation, however, should be interpreted with some caution. The HS "receptor" at the cell surface has a more variable and heterogeneous structure than heparin. Ensembles of Ն4 consecutive, fully N-and O-sulfated disaccharide units are rare in HSs, which may instead display composite interaction sites of short N-sulfated domains interspersed by N-acetylated sequences ("SAS domains," see Ref. 62). The three-dimensional structure of such sites is unknown and is not readily deduced by comparison with the helical shapes reported for heparin polymers (50) and adapted to modeling such as that shown in Fig. 4B. In fact, the interaction properties of other sulfated polysaccharides, for instance carrageenan (40) and chondroitin sulfate E, 3 argue against a strict requirement for defined saccharide sequence. Instead the role of cell surface HS as primary HPV attachment site would depend heavily on the organization of N-sulfated domains along the polysaccharide chain.