1996, 70(10):6589. J. Virol. 

EstesL J White, J M Ball, M E Hardy, T N Tanaka, N Kitamoto and M K cell lines.virus capsids to cultured human and animal Attachment and entry of recombinant Norwalk

http://jvi.asm.org/content/70/10/6589Updated information and services can be found at:

These include:

CONTENT ALERTS more»cite this article),

Receive: RSS Feeds, eTOCs, free email alerts (when new articles

http://journals.asm.org/site/misc/reprints.xhtmlInformation about commercial reprint orders: http://journals.asm.org/site/subscriptions/To subscribe to to another ASM Journal go to:

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

on January 8, 2014 by guest

http://jvi.asm.org/

Dow

nloaded from

JOURNAL OF VIROLOGY, Oct. 1996, p. 6589–6597 Vol. 70, No. 100022-538X/96/$04.0010Copyright q 1996, American Society for Microbiology

Attachment and Entry of Recombinant Norwalk Virus Capsidsto Cultured Human and Animal Cell Lines

LAURA J. WHITE,1 JUDITH M. BALL,1 MICHELE E. HARDY,1 TOMOYUKI N. TANAKA,2

NORITOSHI KITAMOTO,3 AND MARY K. ESTES1*

Division of Molecular Virology, Baylor College of Medicine, Houston, Texas 77030,1

and Department of Laboratory Medicine, Kinan General Hospital, Wakayama,2

and Department of Food Science and Nutrition Himeji College of Hyogo,Himeji,3 Japan

Received 9 January 1996/Accepted 17 June 1996

Norwalk virus (NV) is the prototype strain of a group of noncultivable human caliciviruses responsible forepidemic outbreaks of acute gastroenteritis. While these viruses do not grow in tissue culture cells or animalmodels, expression of the capsid protein in insect cells results in the self-assembly of recombinant Norwalkvirus-like particles (rNV VLPs) that are morphologically and antigenically similar to native NV. We have usedthese rNV VLPs to examine virus-cell interactions. Binding and internalization of VLPs to cultured human andanimal cell lines were studied in an attempt to identify potentially susceptible cell lines for virus propagationin vitro and to determine if early events in the replication cycle were responsible for the narrow host range andrestriction of virus growth in cell culture. Radiolabeled VLPs specifically bound to a saturable number ofbinding molecules on the cell surface of 13 cell lines from different origins, including human intestine(differentiated and undifferentiated Caco-2) and insect (Spodoptera frugiperda 9) ovary. Differentiated Caco-2cells bound significantly more rNV VLPs than the other cell lines. Variations in the amount of bound VLPsamong the different cell lines did not correlate with the tissue or species of origin. VLP binding was specific,as determined by competition experiments with unlabeled rNV VLPs; however, only 1.4 to 6.8% of thespecifically prebound radiolabeled VLPs became internalized into cells. Blocking experiments using polyclonaland monoclonal anti-rNV sera and specific antipeptide sera were performed to map the domains on rNV VLPsinvolved in binding to cells. One monoclonal antibody (NV8812) blocked binding of rNV VLPs to human andanimal cell lines. The binding site of monoclonal antibody NV8812 was localized to the C-terminal 300 to 384residues of the capsid protein by immunoprecipitation with truncated and cleaved forms of the capsid protein.These data suggest that the C-terminal region of the capsid protein is involved in specific binding of rNV VLPsto cells.

Norwalk virus (NV) is the prototype strain of a group ofhuman enteric pathogens recently classified as caliciviruses(22, 29) that are responsible for outbreaks of acute gastroen-teritis in the USA and around the world (4, 9). Although NVwas visualized by immunoelectron microscopy (IEM) 23 yearsago (25), progress in understanding the molecular biology andreplication strategies of NV has been hampered by the lack ofa cell culture system or an animal model. The host rangespecificity for this group of viruses appears to be narrow, prob-ably limited to primates and humans. NV is unable to infectexperimental animals except for chimpanzees, which undergoan asymptomatic infection (24, 54). A narrow host range alsohas been suggested for other animal caliciviruses. For example,two closely related strains of rabbit caliciviruses (rabbit hem-orrhagic disease virus and European brown hare syndromevirus) that cause lethal liver disease exhibit narrow host tro-pisms for different species of rabbits (53).The low amounts of virus in stool samples from infected

individuals initially delayed understanding the molecular biol-ogy of NV. However, in 1990, the first cDNA clones of NVwere obtained (19, 32). Subsequently, the sequence andgenomic organization of NV and the closely related Southamp-ton virus were determined (22, 29). These breakthroughs re-vealed similarities between the sequence and genomic organi-

zation of NV and prototype animal caliciviruses andestablished these agents as members of the Caliciviridae family(22, 29, 32). NV and NV-like viruses contain a genome ofsingle-stranded positive-sense RNA of about 7.7 kb which ispredicted to encode three open reading frames (ORFs) (22).The second ORF (ORF2) encodes the capsid protein (appar-ent molecular weight, 58,000), which is the only structuralprotein (14, 21), a property characteristic of the caliciviruses(5).The capsid protein of NV spontaneously assembles into re-

combinant Norwalk virus-like particles (rNV VLPs) when ex-pressed in insect cells infected with a recombinant baculovirus(21). These empty rNV VLPs can be purified in high yieldsfrom the media of infected insect cell cultures. The resultingpreparations of rNV VLPs are morphologically and antigeni-cally similar to native virions (13, 21). The three-dimensionalstructure of the rNV VLPs has been determined to a resolu-tion of 2.2 nm by electron cryomicroscopy and computer imageprocessing techniques (40). The single capsid protein folds intothree domains to form a shell and protruding arches.In spite of the recent advances in understanding the genome

organization and structure of NV, the lack of a cell culturesystem continues to limit our ability to dissect the biology ofthese viruses. We have taken advantage of the ability to purifylarge amounts of rNV VLPs to study their binding and inter-nalization properties with different cells in culture. These earlyinteractions were studied directly, and after pretreatment with

* Corresponding author. Phone: (713) 798-3585. Fax: (713) 798-3586. Electronic mail address: mestes@bcm.tmc.edu.

6589

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

polyclonal and monoclonal antibodies to rNV, to help identifycell lines susceptible to infection and determine if early eventsin the replication cycle are responsible for the restriction ofvirus growth in cell cultures and the consequent failure tocultivate this virus in vitro. Among the cell lines chosen for thisstudy are cells from human colon adenocarcinomas, includingCaco-2 cells, which upon confluency show morphologic andphysiologic markers of differentiation that are characteristic ofthe mature small intestine enterocytes, also found in fetal hu-man colon. These cells represent the closest in vitro cell culturesystem available that mimics the in vivo characteristics of hu-man differentiated enterocytes (37). Human and animal celllines that have been successfully used to grow fastidious entericviruses were also tested (36, 38, 46, 48). Animal cell lines asdistantly related as human intestine and insect (Spodopterafrugiperda 9 [Sf9]) ovary were included in this study.

MATERIALS AND METHODS

Cells. The origins and properties of the cell lines used in this study are listedin Table 1. The human intestinal cell lines Caco-2 and Int 407 were grown inEarle’s minimum essential medium (MEM), supplemented with 10% fetal bo-vine serum (FBS), L-glutamine, MEM nonessential amino acids, N-2-hydroxy-ethylpiperazine-N9-2-ethanesulfonic acid (HEPES) buffer, penicillin, and strep-tomycin in a 5% CO2 incubator. Caco-2 cells were considered undifferentiated(U-Caco-2) at confluency or 1 day postconfluency, when they showed no detect-able sucrase activity and domes (structures characteristic of polarized epithelialmonolayers undergoing vectorial transport) were lacking. Seven to 14 days post-confluency, Caco-2 cell cultures showed biochemical and morphologic markersof differentiation (sucrase activity of 2.3 to 2.8 mmol/mg/h and the presence ofdomes) and were considered differentiated cultures (D-Caco-2). Porcine kidneyLLC-PK cells were grown in MEM with 10% FBS. AGN (primary rhesus kidney)cells were grown in Lebovitz L-15 medium supplemented with 10% FBS, glu-tamine, penicillin, and streptomycin in a non-CO2 incubator. Human cell linesHT-29, HCT-8, HL, and HEC were grown in RPMI 1640 supplemented with10% FBS, penicillin, and streptomycin. Human embryo kidney 293, monkeykidney Vero, and mouse 3T3 E cells were grown in Dulbecco’s modified Eaglemedium supplemented with 10% FBS, penicillin, and streptomycin. Human T-84cells were grown in Dulbecco’s modified Eagle medium-Ham’s F12 medium (1:1[vol/vol]) supplemented with 10% FBS, penicillin, and streptomycin. Sf9 insectcells were grown in TNM-FH medium containing 10% FBS, penicillin, andstreptomycin at 278C.Preparation of rNV VLPs. rNV VLPs were prepared by infecting Sf9 insect

cells (3 3 106 cells per ml) at a multiplicity of infection of 10 PFU per cell withthe baculovirus recombinant Bac-rNV C8, which expresses the NV capsid pro-tein (21). Metabolically radiolabeled VLPs were prepared by placing the cells inmethionine-free Grace’s medium at 28 h postinfection (hpi) for 30 min and thenadding 25 to 30 mCi of [35S]methionine (Trans-35S-label; ICN, Irvine, Calif.) per

ml. After 4 to 6 h of labeling, 50 mg of unlabeled methionine per ml was added.The cultures were harvested at day 7 postinfection, the cells were pelleted, andrNV VLPs released into the medium were concentrated by centrifugation for 2 hin an SW28 rotor (Beckman) at 26,000 rpm. The pellets were suspended in asolution of CsCl (1.362 g/cm3) and centrifuged for 24 h at 35,000 rpm in aBeckman SW50.1 rotor. Peak fractions of the rNV VLPs were pooled andpelleted for 2 h at 26,000 rpm in an SW28 rotor to remove the CsCl, and thepellets were suspended in 0.01 M phosphate-buffered saline (PBS) and stored at48C. Pure VLP preparations consisted primarily of 38-nm VLPs (virion-like size).A small proportion of 19-nm VLPs also formed by the NV capsid protein werepresent in these pure preparations (50). The 19-nm VLPs were separated fromthe 38-nm VLPs by sedimentation through 5 to 25% sucrose gradients centri-fuged for 3.5 h at 26,000 rpm in a Beckman SW28 rotor. The separated particleswere concentrated by pelleting in the same rotor for 3 h at 26,000 rpm. VLPpreparations were examined by negative-stain electron microscopy and sodiumdodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE). Protein con-centration was determined by using the Bio-Rad protein assay (Life ScienceResearch Products, Hercules, Calif.) with bovine gamma globulin as the proteinstandard. The specific activity of the preparation was determined by calculatingthe number of counts per minute per microgram of protein.Antibodies. Polyclonal hyperimmune antisera to rNV VLPs raised in mice,

guinea pigs, and rabbits have been described elsewhere (21). Human pre- andpostinfection antisera were obtained from two adult volunteers who serocon-verted and excreted NV-specific antigen after being orally inoculated with NV8FIIa (patients 518 and 520) (12). Human preinfection antisera were obtainedfrom eight adult volunteers who had NV-specific antibody enzyme-linked immu-nosorbent assay (ELISA) titers of 5 to 40 and did not seroconvert or excreteantigen after NV inoculation (patients 508, 517, 526, 531, 532, 536, 537, and 553)(12). Ten immunoglobulin G1 (IgG1) monoclonal antibodies (MAbs) to rNVthat have been characterized elsewhere (16) were used (NV834, NV142, NV101,NV813, NV3901, NV3912, NV2461, NV7411, NV8812, and NV8301). TheseNV-specific MAbs are known to react with rNV by ELISA, IEM, Westernblotting (immunoblotting) (WB), and immunoprecipitation (IP). Rotavirus IgG1MAb 2G4 (45), which maps to VP5, was used as a control. IgG was purified fromascites fluids by treatment with caprylic acid followed by ammonium sulfateprecipitation (33). The IgG fraction was dialyzed against ammonium bicarbon-ate, lyophilized, and stored at 48C. The purity of this fraction was confirmed bySDS-PAGE. Working solutions of pure IgG in PBS were made at 2 mg/ml for theblocking experiments. Mouse antipeptide antisera to six NV-specific syntheticpeptides, N1, V1, V3, V5, C1, and C2, that localize to amino acid residues 65 to83, 285 to 305, 336 to 355, 377 to 400, 465 to 484, and 510 to 530 respectively, willbe described elsewhere (2).Binding assays. Each cell line was grown in monolayers to confluency (or 7 to

14 days postconfluency to obtain D-Caco-2 cells) in 24- or 96-well plates (Costar,Cambridge, Mass.). The cell monolayers were washed three times with cold PBS(pH 7.2) or serum-free Eagle’s minimum essential medium (pH 7.2) containing1% bovine serum albumin, fraction V (BSA; Calbiochem, La Jolla, Calif.) andchilled to 48C. Purified radiolabeled rNV VLPs were added to duplicate wells inserum-free medium–1% BSA at final volumes of 200 ml per well in 24-well platesor 30 ml per well in 96-well plates. The plates were incubated for 1 h, since thebinding plateau was shown at this time, with gentle agitation on a rocker platformat 48C to inhibit internalization. The binding reaction was terminated by gentlywashing the cells three times with cold PBS containing 0.1% BSA to remove theunbound VLPs, and then the cells were lysed with radioimmunoprecipitationassay (RIPA) buffer (0.15 M NaCl, 1% sodium deoxycholate, 1% Triton X-100,0.1% SDS, 1% trasylol, 10 mM Tris-HCl [pH 7.2]). The total radioactivity in thesample was determined by liquid scintillation spectrometry. The number of cellsper well was determined for each assay by counting the trypsinized cells fromtriplicate wells. Assays were done in duplicate and repeated at least twice.(i) Saturation binding. Dose-response experiments were performed to deter-

mine the saturability of binding of rNV VLPs to different cell lines. Increasingamounts of radiolabeled particles (from 0.5 to 40 mg/105 cells) were added to theconfluent monolayers in 24-well plates and incubated for 1 h at 48C in a finalvolume of 200 ml. Each rNV VLP concentration was tested in the presence andabsence of a 10-fold excess of unlabeled rNV VLPs to determine the nonspecificand total (specific plus nonspecific) binding, respectively. The specific binding foreach rNV VLP concentration was calculated by subtracting the nonspecificbinding from the total binding.(ii) Calculation of cell binding parameters. To calculate cell binding param-

eters, the number of free rNV VLPs in each well was calculated by subtractionof the number of counts representing the bound VLPs from the VLPs added tothe reaction. A Scatchard plot was derived by plotting the bound/free ratioagainst bound particles, and the best straight-line fit was determined for theresults by using regression analyses as previously reported (44). The intercept ofthe x axis of the Scatchard plot was multiplied by the amount of particles presentin 1 mg of purified particles (5.783 1010) and then divided by the number of cellsin the assay to determine the number of receptor sites per cell. To calculate thenumber of particles present in 1 mg of purified VLPs, the molecular mass of eachcapsid protein (58 kDa) was multiplied by 180 units forming a particle with anicosahedral symmetry of T53, times the mass of 1 Da (1.663 10224 g). The slopeof the line was used to calculate the dissociation constant (Kd), which is equiv-alent to the inverse of the slope. The value of the slope represented the Kd for

TABLE 1. Cell lines used in binding and internalization assays

Cell line Origin Tissue

Caco-2Differentiateda Human Colon adenocarcinomaUndifferentiateda Human Colon adenocarcinoma

HT-29a Human Colon adenocarcinomaHCT-8a Human Colon carcinomaT-84a Human Ileocecal carcinomaInt 407a Human Embryonic jejunum-ileum293a Human Embryonic kidneyHECb Human Endometrial carcinomaHLc Human LungAGNc,d Rhesus Kidney3T3 E Mouse FibroblastSf9e Insect OvaryLLC-PKf Porcine KidneyVeroa Monkey Kidney

a Obtained from the American Type Culture Collection.b Kindly provided by A. Menge.c Kindly provided by T. Metcalf.d Primary cell culture.e Kindly provided by M. Summers.f Kindly provided by L. Saif.

6590 WHITE ET AL. J. VIROL.

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

the number of particles per well in the volume used for the assay. The finalresults were expressed in molar units that were calculated by dividing the inverseof the slope by the volume used for the assay and by Avogadro’s number (6.02331023).(iii) Competition of binding. The ability of homologous and heterologous

particles to compete with radiolabeled rNV VLPs for binding to the cell surfacewas determined by incubating confluent monolayers with a mixture of a fixedamount of radiolabeled VLPs (10 mg/105 cells) and 5-, 10-, or 30-fold-excessunlabeled competitor. The following competitors were tested: homologous rNVVLPs; heterologous rotavirus VP2/6 VLPs made as previously described (8);19-nm rNV VLPs (see above) (50); soluble NV capsid protein obtained byalkaline dissociation of rNV VLPs (17); and eight high-pressure liquid chroma-tography (HPLC)-purified synthetic peptides (N1, V1, V2, V3, V4, V5, C1, andC2) that map to different domains on the capsid protein and localize to aminoacid residues 65 to 83, 285 to 305, 305 to 333, 336 to 355, 349 to 375, 377 to 400,465 to 484, and 510 to 530, respectively. The peptides were dissolved in PBS andused in 20- and 200-fold molar excess.(iv) Blocking of binding. The ability of specific anti-rNV VLP sera (polyclonal,

monospecific, and monoclonal) to block the binding of rNV VLPs to Caco-2 andSf9 cells was examined by preincubating radiolabeled rNV particles (15 mg/105

cells) with serial dilutions of the antiserum or purified IgG in 0.01 M PBS (finalvolume, 20 ml) for 1 h at 378C. After this mixture was chilled on ice, 10 ml of 3%BSA in serum-free MEM was added to each reaction mixture, and the finalmixture was added to confluent monolayers in 96-well plates that had beenprewashed with cold serum-free MEM–1% BSA. Binding was assayed as de-scribed above.Internalization assays. Internalization assays were carried out in triplicate as

described by Bass et al. (3), with slight modifications. Briefly, radiolabeled rNVVLPs (70 mg/105 cells) were bound to confluent monolayers in 96-well plates for1 h at 48C, washed twice with serum-free MEM, and transferred to 378C to allowinternalization to occur. Following a 1- to 4-h incubation, the cells were washedtwice in cold PBS, treated with proteinase K (500 mg/ml; Boehringer Mannheim,Mannheim, Germany) for 30 min at 48C, washed three times with ice-coldPBS–0.1% BSA, and solubilized in RIPA buffer, and the cell-associated radio-activity was quantitated by scintillation counting. The efficiency of the proteinaseK treatment to remove surface bound VLPs was determined in control plateskept at 48C, instead of 378C, to prevent internalization of the VLPs. Thesecontrol experiments showed that 94 to 98% of the VLPs bound at 48C weredigested under these conditions. Specificity was assayed by the addition of a10-fold excess of unlabeled rNV VLPs.Preparation of N-terminal and C-terminal truncated forms of the capsid

protein. Baculovirus transfer plasmids containing only the full-length gene of theNV capsid protein and the four truncated mutants were generated by insertingPCR-amplified fragments flanked by restriction enzyme sites BglII and BamHIinto those sites in the transfer vector pVL 1393 (Invitrogen, San Diego, Calif.).The template used for PCR in all cases was the plasmid pGNV, which containsthe NV ORF2, as described previously (21). The PCR for the full-length capsidprotein gene amplified a segment corresponding to nucleotides (nt) 5349 to 6976by using primers NV-110a and NV-111a (59-GCGGCGAGATCTAATTCGTAAATGATGATGGCG-39 and 59-GCGGCGAGATCTAATTGCACCAATTATGGCTT-39). The N-terminal truncated form, NT35, included nt 5461 to 6976 andwas amplified by PCR by using primers NV-102 (59-CCGGGGATCCATGGCTGTAGCAGGTTCTTCGACA-39) and NV-109 (59-CCCGGGGGATCCAATTGCACCAATTATGGCTT-39). The expressed capsid protein from this clonelacks the N-terminal 35 amino acids, and the dipeptide methionine-alanine hasbeen added to the N terminus. The N-terminal truncated gene NT98 spans fromnt 5456 to 6976 and was amplified by using primers NV-106 (59-CCCGGGGGATCCCTAAATGATGATGGCGTCTAA-39) and oligonucleotide NV-109. Thefinal clone starts at a BamHI site internal to the site on the primer, and theexpressed protein from this clone starts at the methionine in position 99. TheC-terminal truncated capsid gene CT230, which expressed a protein with theC-terminal 230 amino acid residues deleted, was amplified by using primerNV-110a and primer NV-114a (59-GCGGCGAGATCTTCAGTTGATTACAGTGCCATT-39). A double deletion mutant, NCT98/146, spanning from nt 5456 to6510, expressed a protein containing amino acids 99 to 384. The primers used toamplify this fragment were NV-106 and NV-108a (59-AAAAAAGGATCCTCAGCCAGACGGGTGTGATGGGG-39). The PCR products were digested withBglII and ligated to the BglII-linearized, calf intestine alkaline phosphatase-treated pVL 1393 transfer vector (Invitrogen). The insertion site junctions wereconfirmed to be correct by sequencing. Baculovirus recombinants containing thecomplete or truncated forms of NV ORF2 were prepared by cotransfecting Sf9insect cells with the pVL 1393 recombinant plasmids along with Baculogold(Invitrogen) as described previously (8). The recombinant viruses were identifiedby slot blot hybridization followed by two rounds of plaque purification.35S-labeled full-length or truncated NV capsid protein was expressed in Sf9

cells infected with the respective baculovirus recombinant. The cells were labeledwith 25 mCi/ml at 28 hpi and harvested at 36 hpi. Cell lysates were prepared forIP as described previously (16). The expression of the full-length and truncatedproteins was confirmed by WB and IP, using rabbit polyclonal antisera to rNV.Mapping the MAb binding site by IP of truncated forms of the NV capsid

protein. Sf9 cell lysates containing 35S-labeled full-length and truncated forms ofthe capsid protein (NT35, NT98, CT230, and NCT98/146) corresponding to 105 cpm

were immunoprecipitated in nondenaturing buffer with MAb NV8812 and arabbit polyclonal antisera to rNV (diluted 1:20 and 1:50) as described previously(16). In some cases, the 32-kDa (32K) C-terminal trypsin cleavage product of therNV protein was prepared as described before (16) and used in the IP assays.Briefly, 35S-labeled rNV VLPs were dissociated by incubation in 10 mM Tris-HCl(pH 9.0) and then digested with 250 mg of trypsin (Worthington Biochemical,Freehold, N.J.) per ml for 1 h at 378C.

RESULTS

Two major criteria are commonly used to determine thespecificity of virus-receptor interactions. Specific bindingshould be (i) saturable and (ii) competitively inhibited by asecond ligand (unlabeled virus) known to bind to the samereceptor.Binding of rNV VLPs to cells is saturable and specific.

Classic binding kinetics were used to study the binding of rNVVLPs to different cells in culture and determine if the bindingoccurred through a finite number of specific molecules on thecell surface. Dose-response curves of rNV VLPs binding toD-Caco-2, U-Caco-2, and Sf9 cells are shown in Fig. 1A to C.Nonspecific binding measured in the presence of 10-fold-ex-cess unlabeled particles was a small fraction of the total bind-ing and showed a linear relationship between the amount ofradiolabeled VLPs added and the amount bound for the threecell lines. The specific binding for each concentration of radio-labeled particles was calculated by subtracting the nonspecificfrom the total binding. Scatchard plot analysis of the specificsaturable binding of rNV VLPs for each cell line (Fig. 1D)showed that the Kd calculated from the slope of the curve waswithin the same range for the three cell cultures (1028 to 1029

M). The numbers of receptor sites per cell were estimated tobe 1.76 3 106 per cell for D-Caco-2 cells, 0.77 3 106 forU-Caco-2 cells, and 0.3 3 106 for Sf9 cells by multiplying theintercept of the x axis of the Scatchard plot (micrograms ofVLPs bound per 105 cells) by the amount of particles presentin 1 mg of pure particles (5.78 3 1010) and then dividing by thenumber of cells in the assay. These values represent a largenumber of binding sites per cell but are within the rangedescribed for other well-characterized virus-receptor interac-tions (51).One approach to address if the apparent narrow host range

of NV infection results from the presence of specific receptorsthat are restricted to cell populations in the human intestinewas to compare the binding of rNV VLPs to human intestinalcell lines with binding to other human and animal cell lines.Cells from different origins (Table 1), grown to confluentmonolayers in 24-well plates, were incubated with variedamounts (5 to 10 mg/200 ml/105 cells) of 35S-Trans-label-radio-labeled VLPs for 1 h at 48C. rNV VLPs bound to every cell linetested; the percentage of input virus that remained associatedwith the cells ranged from 1 to 19% (Fig. 2). A fraction of thisbinding was not competed for by unlabeled rNV VLPs, repre-senting nonspecific binding. Dose-response curves of the bind-ing of rNV VLPs to these cell lines also showed saturation(data not shown), indicating that VLPs bind to a finite numberof specific binding sites on the surfaces of all of the differentcell lines tested. Five of the cell lines tested originated fromhuman colon carcinomas. Among them, D-Caco-2 bound moreVLPs than the other cell lines. Binding did not seem to cor-relate with a particular tissue or species of origin of the celllines. The binding and internalization of rNV VLPs to D-Caco-2 cells were not affected significantly when the cells werepretreated with trypsin (0.25 to 8 mg/ml; Worthington) or pan-creatin (50 to 100 mg/ml; Gibco BRL) for 1 h at 48C (data notshown).

VOL. 70, 1996 BINDING OF NV CAPSIDS TO HUMAN AND ANIMAL CELLS 6591

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

The binding of rNV VLPs is competed for by homologousparticles but not by heterologous particles. To further examinethe specificity of the rNV binding to the different cultured cells,we compared the abilities of unlabeled homologous rNV VLPsand heterologous rotavirus VP2/6 VLPs to compete with ra-diolabeled rNV VLPs for binding to specific cell surface mol-ecules. A 10-fold excess of unlabeled rNV VLPs abolished 83and 76% of the binding of radiolabeled rNV VLPs to D- andU-Caco-2 cells, respectively, while rotavirus VP2/6 VLPs abol-ished only 7 and 8%, respectively, of the binding (Fig. 3A). Thecompetition of binding to Sf9 cells by unlabeled rNV VLPs wassimilar to that of Caco-2 cells (77%). Unexpectedly, unlabeledrotavirus VP2/6 VLPs appeared to compete for some bindingsites on Sf9 cells (34%), suggesting that the binding of rNVVLPs to Sf9 cells may have properties different from those ofthe binding to Caco-2 cells. Figure 3B shows the binding toD-Caco-2 cells in the presence of other competitors. A 10-foldexcess of unlabeled purified 19-nm rNV VLPs was able tocompete with the binding of radiolabeled 38-nm rNV VLPs toD-Caco-2 cells, suggesting that the virus attachment site isconserved in VLP structures of both sizes. The 19-nm VLPsare assembled in Sf9 cell cultures along with the 38-nm VLPswhen the NV capsid protein is expressed in the baculovirusexpression system. SDS-PAGE analysis has indicated that19-nm VLPs, like the 38-nm VLPs, are formed by the NVcapsid protein (50). Smaller empty particles have been foundin preparations of VLPs from a Mexican strain of humancalicivirus, MX virus (20), and in human stool samples con-taining typical human calicivirus particles (49). Soluble NV

FIG. 1. Dose responses of total, nonspecific, and specific binding of rNV VLPs to D-Caco-2 (A), U-Caco-2 (B), and Sf9 (C) cells. Confluent monolayers grown in24-well plates containing 4.5 3 105 U-Caco-2 cells per well, 7 3 105 D-Caco-2 cells per well, or 5.5 3 105 Sf9 cells per well were incubated in the presence of increasingconcentrations of 35S-labeled rNV VLPs (from 0.5 to 50 mg/105 cells) in a final volume of 200 ml for 1 h at 48C. Each concentration of radiolabeled VLPs was incubatedin the presence and absence of a 10-fold excess of unlabeled VLPs to determine nonspecific and total (specific plus nonspecific) binding, respectively. The specificbinding for each dose was calculated by subtracting nonspecific binding from total binding. The cell-associated radioactivity was determined after the unbound viruswas washed off and the cells were solubilized with RIPA buffer. The counts per minute bound was divided by the specific activity of the radiolabeled rNV VLPspreparation (30,770 cpm/mg) to obtain micrograms bound. (D) Scatchard plot analysis of data from specific binding of rNV VLPs to D-Caco-2, U-Caco-2, and Sf9 cells.Experiments were done in duplicate and reproduced at least twice. Each datum point represents the mean 6 range.

FIG. 2. Binding of rNV VLPs to different cell lines. Five to 10 mg of 35S-labeled rNV VLPs (specific activity, 3 3 to 20 3 104 cpm/mg) per 105 cells wasincubated with different cell lines in 24-well plates for 1 h at 48C. All reactionswere performed in a final volume of 200 ml of serum-free medium containing 1%BSA, in the absence (black bars) or presence (white bars) of a 10-fold excess ofunlabeled rNV VLPs. For cell lines HEC, HL, and AGN, the assay was done onlyin the absence of unlabeled competitor. The cells were washed, and cell-associ-ated radioactivity was determined by liquid scintillation spectrometry. Resultsare expressed as the percentage of total input counts that bound to each cell type.Experiments were done in duplicate and reproduced at least twice. Data plottedrepresent the means 6 ranges. The characteristics of the cell lines tested arelisted in Table 1.

6592 WHITE ET AL. J. VIROL.

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

capsid protein obtained by alkaline dissociation of purifiedrNV VLPs (17) also competed with rNV VLPs for binding siteson D-Caco-2 cells. These results indicate that the correct con-formation of the virus attachment site on the NV capsid pro-tein is present in the dissociated soluble forms.NV-specific synthetic peptides do not block binding. Eight

HPLC-purified NV synthetic peptides that map to differentdomains on the NV capsid protein were tested for the ability tocompete with radiolabeled rNV VLPs for specific binding siteson the surface of Caco-2 cells (see Fig. 6A). The sequences offive of these peptides cover the region on the capsid proteinbetween amino acids 285 and 400, which is predicted to containthe binding site for MAb NV8812 (see below). Binding ofradiolabeled rNV VLPs was not altered when a 20- or 200-foldmolar excess of the purified peptides was used in the compe-tition assays (data not shown), suggesting that the virus attach-ment site on the capsid protein is conformational. In agree-ment with this finding MAb NV8812 did not react with any ofthe NV peptides when tested by ELISA (2). These resultssuggest that MAb NV8812 binds to a conformational epitope,as shown earlier by Hardy et al. (16).Internalization of radiolabeled rNV VLPs by human and

insect cell lines. We determined whether the rNV VLPs were

internalized as a consequence of their specific binding to thesurface of D-Caco-2, U-Caco-2, T-84, and Sf9 cells. The inter-nalization assay was designed to determine how much of thebound VLPs entered the cells by allowing prebound radiola-beled VLPs to enter the cells by warming them to 378C. After1 h, surface-bound particles were removed by proteinase Kdigestion. The treatment of cells with proteinase K to removesurface-bound particles was effective in reproducibly removing94 to 98% of the counts attached to the cell surface in controlcells kept at 48C before the proteinase K treatment. The av-erage counts from internalization controls kept at 48C prior toproteinase K treatment were subtracted from the internalizedcounts. Between 1.4 and 6.8% of the counts bound to the fourcell cultures remained associated with the cells after the pro-teinase K treatment (Fig. 4). To exclude the possibility that thelow levels of entry measured were a consequence of nonspe-cific binding, the internalization assays were done in the pres-ence of a 10-fold excess unlabeled rNV VLPs. Under theseconditions, no counts were detected inside the cells by ourassay, indicating that the entry was a consequence of specificbinding (data not shown). Our assay was validated by showingthat it was able to detect 15% of 35S-labeled rotavirus SA11 4Finternalized into D-Caco-2 cells after 30% of the input viruswas bound to the cell surface. Internalization of rotavirus SA114F was blocked (56% inhibition) in the presence of a 10-foldexcess of unlabeled virus (data not shown).NV-specific MAb blocks binding of rNV VLPs to D-Caco-2

and Sf9 cells. Ten MAbs developed against rNV VLPs andknown to bind NV virions in stool samples by ELISA and IEMand to bind rNV VLPs by ELISA, IP, WB, and IEM (16) weretested for the ability to block binding to human intestinalD-Caco-2 cells and insect Sf9 cells. Ascites fluid from oneMAb, NV8812, was able to block the binding of radiolabeledrNV VLPs to D-Caco-2 cells at dilutions of 1:10, 1:20, and 1:50(data not shown). The specificity of this activity was confirmedwhen purified IgG from MAb NV8812 blocked binding at

FIG. 3. Binding of radiolabeled rNV VLPs to different cell lines in thepresence of competitors. (A) Competition of binding of 35S-labeled rNV VLPs tothree cell lines by unlabeled rNV VLPs and rotavirus VP2/6 VLPs. Confluentmonolayers of D-Caco-2, U-Caco-2, and Sf9 cells grown in 24-well plates wereincubated with a mixture of 35S-labeled rNV VLPs with a specific activity of 6 3to 20 3 104 cpm/mg (3 to 5 mg/105 cells) and a 10-fold excess of unlabeled rNVVLPs or a 10-fold excess of unlabeled rotavirus VP2/6 VLPs. Cell-associatedradioactivity was determined after a 1-h incubation at 48C. Results are expressedas percentage of input counts. Data plotted represent the means 6 2 standarddeviations of three replicates. (B) Competition of binding to D-Caco-2 cells by38-nm rNV VLPs, 19-nm rNV VLPs, and alkaline-dissociated 38-nm rNV VLPs.Competition assays were performed as indicated for panel A.

FIG. 4. Internalization of 35S-rNV VLPs into different cell lines. The indi-cated cell lines grown as monolayers in 96-well plates were prechilled andincubated with 70 mg of 35S-rNV VLPs per 105 cells for 1 h at 48C. After twowashes with serum-free MEM, the cells were incubated at 378C for 1 h, rinsedtwice with cold PBS, and treated with proteinase K (500 mg/ml) for 30 min at 48C.Noninternalized counts were removed by three washes with cold PBS–1% BSA,and cell-associated radioactivity was measured by scintillation counting of cellssolubilized in RIPA buffer. The average counts from internalization controlskept at 48C for 1 h after binding and prior to proteinase K treatment weresubtracted from the internalized counts. Data are expressed as percentage ofbound counts. Results represent the means 6 2 standard deviations of triplicatewells.

VOL. 70, 1996 BINDING OF NV CAPSIDS TO HUMAN AND ANIMAL CELLS 6593

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

concentrations of 0.15 to 1.5 mg/ml, while IgG from rotavirusMAb 2G4 (45) and IgG from MAbs NV3901 and NV3912 hadno blocking activity (Fig. 5). MAb NV8812 IgG also blockedbinding to Sf9 cells (data not shown). These MAbs to rNVVLPs have been assigned to three groups based on their dif-ferential reactivities (16). MAb NV8812 is one of three groupIII MAbs that recognize discontinuous epitopes located in theC-terminal 230 residues of the capsid protein. Another groupIII MAb (NV8301) exhibited intermediate levels of blockingactivity, suggesting that it may map to residues that partiallyoverlap or are close to the site recognized by MAb NV8812.We hypothesize that the site recognized by MAb NV8812 isinvolved in the attachment of rNV VLPs to cells.Polyclonal antisera to rNV VLPs do not block binding.

Three polyclonal hyperimmune sera produced in our labora-tory against purified rNV VLPs were tested for the ability toblock the specific binding of rNV VLPs to D-Caco-2 cells.These hyperimmune antisera from rabbits, mice, and guineapigs immunized with three intramuscular injections of purifiedrNV have a high (.106) titer to rNV VLPs by ELISA and candetect NV in stools from infected persons by an antigen cap-ture ELISA (12, 21). These antisera also bind to rNV VLPsand to native NV following analysis by IEM, IP, and WB (21).We tested the ability of serial dilutions (1:20 to 1:4,000) of theantisera to block virus binding to cells. None of the antiseratested significantly inhibited or enhanced binding of rNV VLPsto D-Caco-2 cells (data not shown). Binding also was notblocked by human pre- and postinfection sera from two adultvolunteers who seroconverted and excreted antigen in stoolsafter being orally inoculated with NV 8FIIa (12). rNV VLPbinding to cells also was not blocked by preinoculation serafrom an additional eight patients who were resistant to infec-tion (no seroconversion or antigen excretion) with the 8FIIaNV inoculum.Monospecific polyclonal antisera directed against synthetic

peptides that map to known residues on the capsid protein donot block rNV VLP binding to cells. Hyperimmune antisera tosix NV capsid protein synthetic peptides (amino acid residues65 to 83, 285 to 305, 336 to 355, 377 to 400, 465 to 484, and 510to 530) produced in mice and guinea pigs (2) were not able to

block the binding of rNV VLPs to D-Caco-2 cells when testedin blocking experiments at dilutions of 1:50, 1:100, and 1:400(data not shown).Mapping the binding site of MAb NV8812. Since MAb

NV8812 blocked attachment presumably by binding to thevirus attachment site on the rNV VLP capsid protein, wepursued localizing the binding site of this MAb. The reactivityof MAb NV8812 was tested by IP with different truncatedforms of the rNV capsid protein and the 32K C-terminal tryp-sin cleavage fragment of the capsid protein (Fig. 6). Underconditions that would keep the protein properly folded duringthe assay (16), the hyperimmune polyclonal rabbit sera to rNVVLPs recognized all of the different forms of the capsid pro-tein. MAb NV8812 reacted with the 32K trypsin cleavage prod-uct, with the N-terminal truncated forms NT35 and NT98, andwith the double N- and C-terminal truncated fragment NCT98/146. MAb NV8812 did not recognize CT230 in this assay (Fig.6). These results suggest that MAb NV8812 binds to a confor-mational epitope on the capsid protein located between aminoacid residues 300 and 384. We hypothesize that the virus at-tachment site also maps in this region.

DISCUSSION

The initial step in a virus infectious cycle is the attachmentof the virion to the host cell membrane. For many viruses suchas human immunodeficiency virus and rhinovirus, virus attach-ment occurs through a specific receptor molecule present in afinite number of sites that mediates a high-affinity interaction(7, 15, 28). At the other end of the spectrum, other viruses likeSemliki Forest virus (18) have low-affinity interactions with cellsurface molecules present in many copies on the cell surface,and a less specific interaction leads to a productive infection. Ithas been demonstrated for several viral systems that the ad-sorption step is a major determinant of the host range andtissue tropism of the virus (6, 28, 43), and usually a narrow hostrange has been associated with high-affinity interactions with avery specific receptor (31). Because of the apparent narrowhost range of NV and the inability to grow this virus in tissueculture, we hypothesized that early events in its replicationcycle could determine the host range specificity and the failureof virus to grow in cell culture.For viruses that lack a cell culture system, the use of recom-

binant VLPs to study early events in the replication cycle is analternative approach to examine binding and internalizationfor a variety of cell lines. Data from such studies can unravelthe potential role of these interactions in the inability to growthe virus in cell culture and in host specificity. For example, thisapproach was used by Muller et al. (34) to study the initialinteractions of epitheliotropic papillomaviruses (PVs), whichare unable to complete their entire life cycle in cell culture(47). These investigators reported that bovine PV (BPV) syn-thetic capsids and BPV virions bound to the same receptormolecules in cultured cells from different origins, suggestingthat empty VLPs behaved like authentic virions in their inter-action with the cell surface of a wide range of cells. Theirresults suggested that the PV receptor is a conserved cellsurface molecule used by human PV and BPV VLPs and thatthe tropism of infection by different PVs is controlled by eventsdownstream from the initial steps of binding and uptake (41).Our initial approach to study human calicivirus-host cell

interactions is based on the assumption that the virus attach-ment site on the rNV is functionally equivalent to that on theinfectious virions. The following observations support this as-sumption. (i) VLPs are antigenically similar to authentic viri-ons, as determined by cross-reactivity in ELISAs and IPs using

FIG. 5. Effects of MAbs to NV and rotavirus MAb 2G4 (45) on the bindingof rNV VLPs to D-Caco-2 cells. 35S-labeled NV VLPs (15 mg) were preincubatedwith increasing concentrations of IgG purified form the ascites fluids of thedifferent MAbs tested, or with 0.01 M PBS in the control, in a final volume of 20ml of PBS. After 1 h at 378C, the particles in the presence of antibody or PBSwere adsorbed to D-Caco-2 cells grown in 96-well plates (105 cells per well) for1 h at 48C. 35S-labeled VLPs bound were determined by liquid scintillationspectrometry. Binding was expressed as percentage of the control. Results shownare from a representative experiment. Data plotted represent the means ofduplicate wells. The apparent dose-related enhancement effect of MAb 2G4 onbinding was not reproduced in other experiments.

6594 WHITE ET AL. J. VIROL.

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

NV-specific hyperimmune antisera, MAbs, and antipeptide an-tisera (13, 16, 21). (ii) Comparison of computer-reconstructedimages at 2.2-nm resolution of rNV VLPs and a primate cali-civirus virion containing genomic RNA (39, 40) indicated thatthe absence of nucleic acid in the rNV VLPs does not alter thegeneral domain organization of the particle. (iii) The ability ofthe rNV capsid protein to self-assemble into empty capsidssuggests that protein-protein interactions, more than RNA-protein interactions, are responsible for assembling the correctNV capsid structure.Our results show that radiolabeled rNV VLPs bind to a

saturable number of specific molecules on the surfaces of thehuman and animal cells tested. Variations in the amount ofVLPs bound among the cell lines did not correlate with tissueor species of origin. Scatchard plot analyses revealed similarbinding affinities for human and animal cells, with values in therange previously described for virus-receptor interactions (108

to 109 M21) (51). The number of binding molecules per cellranged from 105 to 106. Similar binding characteristics arefound for viruses such as Semliki Forest virus, vesicular sto-matitis virus, and Autographa californica nuclear polyhedrosisvirus, which have a broad host range (51). D-Caco-2 cellsbound rNV VLPs with a higher efficiency than the other cell

lines. This enhanced binding could result from the binding ofrNV VLPs to a functional NV receptor molecule expressedonly in D-Caco-2 cells or to an increased number of bindingmolecules as a result of the increased microvilli surface areacharacteristic of the D-Caco-2 cells. Our approach does notaddress directly whether the binding molecules on the differentcell lines are the same, but the ability of MAb NV8812 to blockbinding of rNV VLPs to Caco-2 cells as well as to Sf9 cellssuggests the same virus attachment site for both cell lines. Ifthis binding molecule is the biologically relevant receptor, thenthe receptor for NV might be an evolutionarily conservedprotein that is expressed in many cell types. If this is the case,NV infection is restricted not by the availability of specificreceptor molecules but by other events that occur later in thereplication cycle.Specific saturable binding of virions or VLPs to the cell

surface does not necessarily result in viral uptake. Some virusesrequire interaction(s) with more than one cellular surface mol-ecule to initiate a productive infection (30, 52). In our study,the specific binding of rNV VLPs to human and animal celllines led to low levels of internalization (1.4 to 6.8%). Thepercentages of particles internalized by the different cell lineswere not significantly different, but internalization was inhib-

FIG. 6. Constructs used to map the virus attachment site. (A) Schematic diagram of the NV capsid protein, indicating the locations of NV-specific syntheticpeptides, truncated forms, and trypsin-cleaved 32K fragment of the NV capsid protein. Results of IPs with polyclonal rabbit antibodies and MAb NV8812 aresummarized on the right. The predicted region that contains the virus attachment site (VAS) is indicated in the bottom. (B) IP of full-length (FL) and truncated (CT230and 32K) forms of the NV capsid protein by polyclonal rabbit (R) antiserum to rNV VLPs (dilution, 1:100) and MAb NV8812 (dilution, 1:50). Full-length and CT230radiolabeled lysates were prepared from Sf9 cells infected with recombinant baculovirus rNV ORF2 and rNV CT230. The cells were labeled for 8 h with 35S-Trans-labelat 28 hpi. IPs were performed in nondenaturing conditions, and proteins were resolved by SDS-PAGE (10% gel).

VOL. 70, 1996 BINDING OF NV CAPSIDS TO HUMAN AND ANIMAL CELLS 6595

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

ited by excess unlabeled rNV VLPs, indicating that the lowlevels of internalization observed were due to the entry of asmall fraction of specifically bound VLPs. The ability of ourassay to detect viral entry into Caco-2 cells was confirmed bymeasuring the internalization of radiolabeled purified rotavi-rus. Our assay showed that 15% of bound rotavirus was inter-nalized, validating the assay. However, because of the lowefficiencies of internalization of rNV VLPs, we cannot rule outthe possibility that nonspecific pinocytosis is the mechanismresponsible for the entry levels measured by our assay. Withouthaving a cultivation system, our present data cannot distinguishwhether the small amount of internalization is sufficient toinitiate replication, or if this represents the block to replicationin cell culture. If entry is the block, experimental manipula-tions to enhance entry might facilitate virus cultivation.Very little is known about the early interactions of other

caliciviruses with cells and host tropism. The binding and entrycharacteristics of only feline calicivirus (FCV) have been ex-amined (26, 27). FCV was shown to bind efficiently (40 to 80%)to permissive cells and less efficiently (20%) to nonpermissivecells (27). Infectious virus was recovered after transfection ofFCV RNA into nonpermissive cells, suggesting that FCV rep-lication in nonpermissive cells is primarily restricted by theabsence of appropriate receptors on the cell surface (27). Thisinterpretation is consistent with available data, although inter-nalization, penetration, and uncoating of particles were notexamined. Therefore, it is possible that one of these other earlyevents plays a role in the growth restriction of FCV in non-permissive cells. Entry of FCV to permissive cells appeared torequire a low-pH-dependent pathway (26).If these data with FCV are relevant to all caliciviruses, one

interpretation of our results is that the low efficiency of bindingand internalization of rNV to the cell lines tested is due to theabsence of the appropriate receptors on the surface of thosecells. If this interpretation is correct, other cells in the intes-tine, different from the mature differentiated enterocytes,could be the ones susceptible to NV infection. Permissive cellswould show higher binding and internalization efficiencies thanthose observed in this study. Alternatively, enterocyte functionmight be modified in the intestine in response to interactionswith diverse adjacent cells or signals from these cells. In thatcase, our results could be a consequence of the inability of thecell culture systems available to simulate the in vivo environ-ment and terminal stage of differentiation required for theproductive attachment and entry of NV into the host cell.Flynn and Saif (10) have reported the successful in vitro serialpropagation of a porcine calicivirus in primary porcine kidneycells. This occurred only in the presence of intestinal contentsfrom uninfected gnotobiotic piglets, suggesting that the lack ofspecific in vivo conditions in in vitro systems may limit thebinding, internalization and/or later events in the replicationcycle. When we pretreated D-Caco-2 cells with different con-centrations of trypsin and pancreatin, the binding or internal-ization to those cells was not affected significantly, suggestingthat other factors may be required to achieve successful culti-vation.Our data show that D-Caco-2 cells bind rNV VLPs most

efficiently and probably internalize enough particles to initiatea productive infection. These may be an appropriate cell cul-ture for virus cultivation. However, our data also indicate thatNV capsid soluble protein present in stools competes for par-ticle binding to cells. Successful virus cultivation may requireremoval of this soluble protein prior to cultivation attempts.The specific binding of a virus to its cellular receptor can

usually be blocked by polyclonal antisera raised against thevirion and/or by MAbs to the virus attachment site. Exceptions

to this blocking effect have been reported for picornavirusesthat have developed different strategies to avoid recognition ofthe virus attachment site by the host immune system, such thatneutralizing antibodies do not block virus binding. For entero-viruses and rhinoviruses, the virus attachment site is located ina cleft or canyon on the virus surface (42), where antibodiesare not able to interact because of steric hindrance. For foot-and-mouth disease virus, an RGD conserved sequence in the Cterminus of VP1, surrounded by hypervariable residues, hasbeen proposed as the virus attachment site (11). This site isassociated with a flexible structural component or disorderedloop that hinders the immune system from recognizing theepitope (1).Our experiments found that one MAb made to rNV VLPs

blocked binding of rNV VLPs to Caco-2 and Sf9 cells, whilepolyclonal hyperimmune antiserum to rNV VLPs and humanserum from infected individuals failed to block binding. Theinability of polyclonal antiserum to rNV to block specific bind-ing to human intestinal cell lines can be interpreted as a con-sequence of a poor accessibility of the virus attachment site tointeract with antibody molecules, either because of its locationin a structural cleft, similar to enteroviruses and rhinoviruses(42), or possibly because the virus attachment site is a disor-dered structure as reported for foot-and-mouth disease virus(1). Another factor that may explain the inability of the rNVpolyclonal antisera to block binding is its suspected monospeci-ficity. For example, the polyclonal hyperimmune antiserum torNV made in our laboratory is type specific, recognizing onlyNV or strains of the same serotype in ELISAs (21). Thisfinding suggests that the antibodies in the polyclonal sera rec-ognize a few immunodominant epitopes in the capsid proteinof NV. In light of these results, the reported lack of correlationbetween preexisting serum antibodies to NV and protectionagainst disease in human volunteers challenged with NV (23,35) could be explained by the absence of antibodies that blockattachment. Short-term clinical immunity has been found inadult volunteers after multiple challenges with NV, but it is notclear what the correlates of this short-term protection are. It ispossible that repeated exposures are required to induce anantibody response capable of blocking the binding of virions tothe target cells.The reactivity of MAb NV8812 with truncated mutants and

trypsin-cleaved forms of the capsid protein by IP indicates thatthe binding site for this MAb is between amino acid residues300 and 384 in the C terminus of NV capsid protein. Thisassignment is made with the caveat that the lack of binding ofMAb NV8812 to the CT230 construct is not influenced byconformational changes caused by removal of the C terminus.If it is, the binding site should be between amino acids 228 and384. On the basis of predictions derived from structural studiesof rNV VLPs (40), it is likely that these residues form part ofthe arch-like structures in the rNV capsids. We propose thatthe NV capsid protein amino acid residues from 300 to 384contain a domain involved in the specific binding to Caco-2cells as well as to other animal cell lines. This hypothesis andthe biological relevance of this site will be confirmed if thisMAb blocks NV replication in volunteers or blocks replicationin culture once NV is successfully cultivated.

ACKNOWLEDGMENTS

This work was supported by funding from Public Health Servicegrants AI-30448, AI-38036, and T32 DK-07664-03 from the NationalInstitutes of Health.We thank Susan Henning for critical comments and suggestions

throughout this work.

6596 WHITE ET AL. J. VIROL.

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from

REFERENCES

1. Acharya, R., E. Fry, D. Logan, D. Stuart, F. Brown, G. Fox, and D. Rowlands.1990. The three-dimensional structure of foot-and-mouth disease virus, p.211–217. In M. A. Brinton and F. X. Heinz (ed.), New aspects of positive-strand RNA viruses. American Society for Microbiology, Washington, D.C.

2. Ball, J. M., and M. K. Estes. 1996. Unpublished data.3. Bass, D. M., M. R. Baylor, C. Chen, E. M. Mackow, M. Bremont, and H. B.Greenberg. 1992. Liposome-mediated transfection of intact viral particlesreveals that plasma membrane penetration determines permissivity of tissueculture cells to rotavirus. J. Clin. Invest. 90:2313–2320.

4. Blacklow, N. R., and H. B. Greenberg. 1991. Viral gastroenteritis. N. Engl.J. Med. 325:252–264.

5. Carter, M. J., I. D. Milton, and C. R. Madeley. 1991. Caliciviruses. Rev. Med.Virol. 1:177–186.

6. Collins, A. R. 1993. HLA class I antigen serves as a receptor for humancoronavirus OC43. Immunol. Invest. 22:95–103.

7. Colonno, R. J., J. H. Condra, S. Mizutani, P. L. Callahan, M. E. Davies, andM. A. Murcko. 1988. Evidence for the direct involvement of the rhinoviruscanyon in receptor binding. Proc. Natl. Acad. Sci. USA 85:5449–5453.

8. Crawford, S. E., M. Labbe, J. Cohen, M. H. Burroughs, Y.-J. Zhou, andM. K. Estes. 1994. Characterization of virus-like particles produced by theexpression of rotavirus capsid proteins in insect cells. J. Virol. 68:5945–5952.

9. Estes, M. K., and M. E. Hardy. 1995. Norwalk virus and other entericcaliciviruses, p. 1009–1034. In M. Blaser, P. Smith, J. Ravdin, H. B. Green-berg, and R. Guerrant (ed.), Infections of the gastrointestinal tract. RavenPress, New York.

10. Flynn, W. T., and L. J. Saif. 1988. Serial propagation of porcine entericcalicivirus-like virus in primary porcine kidney cell cultures. J. Clin. Micro-biol. 26:206–212.

11. Fox, G., N. R. Parry, P. V. Barnett, B. McGinn, D. J. Rowlands, and F.Brown. 1989. The cell attachment site on foot-and-mouth disease virusincludes the amino acid sequence RGD (arginine-glycine-aspartic acid).J. Gen. Virol. 70:625–637.

12. Graham, D. Y., X. Jiang, T. Tanaka, A. R. Opekun, H. P. Madore, and M. K.Estes. 1994. Norwalk virus infection of volunteers: new insights based onimproved assays. J. Infect. Dis. 17:34–43.

13. Green, K. Y., J. F. Lew, X. Jiang, A. Z. Kapikian, and M. K. Estes. 1993.Comparison of the reactivities of baculovirus-expressed recombinant Nor-walk virus capsid antigen with those of the native Norwalk virus antigen inserologic assays and some epidemiologic observations. J. Clin. Microbiol.31:2185–2191.

14. Greenberg, H. B., J. R. Valdesuso, A. R. Kalica, R. G. Wyatt, V. J. McAuliffe,A. Z. Kapikian, and R. M. Chanock. 1981. Proteins of Norwalk virus. J. Vi-rol. 37:994–999.

15. Greve, J. M., G. Davis, A. M. Meyer, C. P. Forte, S. C. Yost, C. W. Marlor,M. E. Kamarck, and A. McClelland. 1989. The major human rhinovirusreceptor is ICAM-1. Cell 56:839–847.

16. Hardy, M. E., T. N. Tanaka, N. Kitamoto, L. J. White, J. M. Ball, X. Jiang,and M. K. Estes. 1996. Antigenic mapping of the recombinant Norwalk viruscapsid protein using monoclonal antibodies. Virology 217:252–261.

17. Hardy, M. E., L. J. White, J. M. Ball, and M. K. Estes. 1995. Specificproteolytic cleavage of recombinant Norwalk virus capsid protein. J. Virol.69:1693–1698.

18. Helenius, A., and E. Fries. 1979. Binding of Semliki Forest virus and its spikeglycoproteins to cells. Euro. J. Biochem. 97:213–220.

19. Jiang, X., D. Y. Graham, K. N. Wang, and M. K. Estes. 1990. Norwalk virusgenome cloning and characterization. Science 250:1580–1583.

20. Jiang, X., D. O. Matson, G. M. Ruiz-Palacios, J. Hu, J. Treanor, and L. K.Pickering. 1995. Expression, self-assembly, and antigenicity of a SnowMoun-tain Agent-like calicivirus capsid protein. J. Clin. Microbiol. 33:1452–1455.

21. Jiang, X., M. Wang, D. Y. Graham, and M. K. Estes. 1992. Expression,self-assembly, and antigenicity of the Norwalk virus capsid protein. J. Virol.66:6527–6532.

22. Jiang, X., M. Wang, K. Wang, and M. K. Estes. 1993. Sequence and genomicorganization of Norwalk virus. Virology 195:51–61.

23. Johnson, P. C., J. J. Mathewson, H. L. DuPont, and H. B. Greenberg. 1990.Multiple-challenge study of host susceptibility to Norwalk gastroenteritis inUS adults. J. Infect. Dis. 161:18–21.

24. Kapikian, A. Z., M. K. Estes, and R. M. Chanock. 1996. Norwalk Group ofviruses, p. 783–810. In B. N. Fields, D. M. Knipe, and P. M. Howley (ed.),Virology. Raven Press, New York.

25. Kapikian, A. Z., R. G. Wyatt, R. Dolin, T. S. Thornhill, A. R. Kalica, andR. M. Chanock. 1972. Visualization by immune electron microscopy of a 27nm particle associated with acute infectious nonbacterial gastroenteritis.J. Virol. 10:1075–1081.

26. Kreutz, L. C., and B. S. Seal. 1995. The pathway of feline calicivirus entry.Virus Res. 35:63–70.

27. Kreutz, L. C., B. S. Seal, and W. L. Mengeling. 1994. Early interaction offeline calicivirus with cells in culture. Arch. Virol. 136:19–34.

28. Lamarre, D., A. Ashkenazi, S. Fleury, D. H. Smith, R.-P. Sekaly, and D. J.Capon. 1989. The MHC-binding and gp120-binding functions of CD4 are

separable. Science 245:743–746.29. Lambden, P. R., E. O. Caul, C. R. Ashley, and I. N. Clarke. 1993. Sequence

and genome organization of a human small round-structured (Norwalk-like)virus. Science 259:516–519.

30. Li, Y., S. vanDrunen Littel-van den Hurk, L. A. Babiuk, and X. Liang. 1995.Characterization of cell-binding properties of bovine herpesvirus 1 glycopro-teins B, C, and D: identification of a dual cell-binding function of gB. J. Vi-rol. 69:4758–4768.

31. Marsh, M., and A. Helenius. 1989. Virus entry into animal cells. Adv. VirusRes. 36:107–151.

32. Matsui, S. M., J. P. Kim, H. B. Greenberg, W. Su, Q. Sun, P. C. Johnson,H. L. DuPont, L. S. Oshiro, and G. R. Reyes. 1991. The isolation andcharacterization of a Norwalk virus-specific cDNA. J. Clin. Invest. 87:1456–1461.

33. McKinney, M. M., and A. Parkinson. 1987. A simple, non-chromatographicprocedure to purify immunoglobulins from serum and ascites fluid. J. Im-munol. Methods 96:271–278.

34. Muller, M., L. Gissmann, R. J. Cristiano, X.-Y. Sun, I. H. Frazer, A. B.Jenson, A. Alonso, H. Zentgraf, and J. Zhou. 1995. Papillomavirus capsidbinding and uptake by cells from different tissues and species. J. Virol.69:948–954.

35. Parrino, T. A., D. S. Schreiber, J. S. Trier, A. Z. Kapikian, and N. R.Blacklow. 1977. Clinical immunity in acute gastroenteritis caused by Norwalkagent. N. Engl. J. Med. 297:86–89.

36. Parwani, A. V., W. T. Flynn, K. L. Gadfield, and L. J. Saif. 1991. Serialpropagation of porcine enteric calicivirus in a continous cell line. Effect ofmedium suppliments with intestinal contents or enzymes. Arch. Virol. 120:115–122.

37. Pinto, M., S. Robine-Leon, M.-D. Appay, M. Kedinger, N. Triadou, E. Dus-saulx, B. Lacroix, P. Simon-Assmann, K. Haffen, J. Fogh, and A. Zweibaum.1983. Enterocyte-like differentiation and polarization of the human coloncarcinoma cell line Caco-2 in culture. Biol. Cell. 47:323–330.

38. Pinto, R. M., J. M. Diez, and A. Bosch. 1994. Use of the colonic carcinomacell line CaCo-2 for in vivo amplification and detection of enteric viruses.J. Med. Virol. 44:310–315.

39. Prasad, B. V. V., D. O. Matson, and A. J. Smith. 1994. Three-dimensionalstructure of calicivirus. J. Mol. Biol. 240:256–264.

40. Prasad, B. V. V., R. Rothnagel, X. Jiang, and M. K. Estes. 1994. Three-dimensional structure of baculovirus-expressed Norwalk virus capsids. J. Vi-rol. 68:5117–5125.

41. Roden, R. B. S., R. Kirnbauer, A. B. Jenson, D. R. Lowy, and J. T. Schiller.1994. Interaction of papillomaviruses with the cell surface. J. Virol. 68:7260–7266.

42. Rossmann, M. G. 1989. The canyon hypothesis. Hiding the host cell receptorattachment site on a viral surface from immune surveillance. J. Biol. Chem.263:14587.

43. Rossmann, M. G., A. E. Erickson, E. A. Frankenberger, J. P. Griffith, H.-J.Hecht, J. E. Johnson, G. Kamer, M. Luo, A. G. Mosser, R. R. Rueckert, B.Sherry, and G. Vriend. 1985. Structure of a human common cold virus andfunctional relationship to other picornaviruses. Nature (London) 317:145–153.

44. Scatchard, G. 1949. The attractions of proteins for small molecules and ions.Ann. N. Y. Acad. Sci. 51:660–672.

45. Shaw, R. D., P. T. Vo, P. A. Offit, B. S. Coulson, and H. B. Greenberg. 1986.Antigenic mapping of the surface proteins of rhesus rotavirus. Virology155:434–451.

46. Superti, F., A. Tinari, L. Baldassarri, and G. Donelli. 1991. HT-29 cells: anew substrate for rotavirus growth. Arch. Virol. 116:159–173.

47. Taichman, L. B., F. Breitburd, O. Croissant, and G. Orth. 1984. The searchfor a culture system for papillomavirus. J. Invest. Dermatol. 83:2s–6s.

48. Takiff, H. E., S. E. Straus, and C. F. Garon. 1981. Propagation and in vitrostudies of previously non-cultivable enteric adenoviruses in 293 cells. Lancetii:832–834.

49. Taniguchi, K., S. Urasawa, and T. Urasawa. 1981. Further studies of 35-40nm virus-like particles associated with outbreaks of acute gastroenteritis.J. Med. Microbiol. 14:107–118.

50. White, L. J., and M. K. Estes. 1996. Unpublished data.51. Wickham, T. J., R. R. Granados, H. A. Wood, D. A. Hammer, and M. L.

Shuler. 1990. General analysis of receptor-mediated viral attachment to cellsurfaces. Biophys. J. 58:1501–1516.

52. Wickham, T. J., P. Mathias, D. A. Cheresh, and G. R. Nemerow. 1993.Integrins alphavbeta3 and alphavbeta5 promote adenovirus internalizationbut not virus attachment. Cell 73:309–319.

53. Wirblich, C., G. Meyers, V. F. Ohlinger, L. Capucci, U. Eskens, B. Haas, andH.-J. Thiel. 1994. European brown hare syndrome virus: relationship torabbit hemorrhagic disease virus and other caliciviruses. J. Virol. 68:5164–5173.

54. Wyatt, R. G., H. B. Greenberg, D. W. Dalgard, W. P. Allen, D. L. Sly, T. S.Thornhill, R. M. Chanock, and A. Z. Kapikian. 1978. Experimental infectionof chimpanzees with the Norwalk agent of epidemic viral gastroenteritis.J. Med. Virol. 2:89–96.

VOL. 70, 1996 BINDING OF NV CAPSIDS TO HUMAN AND ANIMAL CELLS 6597

on January 8, 2014 by guesthttp://jvi.asm

.org/D

ownloaded from