7) 336–348www.elsevier.com/locate/yviro

Virology 359 (200

The neutralizing antibody response against West Nile virus innaturally infected horses

Melissa D. Sánchez a, Theodore C. Pierson c, Marciela M. DeGrace a, Lisa M. Mattei a,Sheri L. Hanna a, Fabio Del Piero b, Robert W. Doms a,⁎

a Department of Microbiology, University of Pennsylvania, 225 Johnson Pavilion, 3610 Hamilton Walk, Philadelphia, PA 19104, USAb Department of Pathobiology-School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA 19104, USAc Viral Pathogenesis Section, Laboratory of Viral Diseases, National Institutes of Health, Bethesda, MD 20892, USA

Received 31 May 2006; returned to author for revision 21 July 2006; accepted 29 August 2006Available online 20 October 2006

Abstract

A major neutralizing epitope (here referred to as the T332 epitope) located on the lateral surface of domain III (DIII) of the West Nile virus(WNV) envelope protein has been identified based on the analysis of murine monoclonal antibodies. However, little is known about the humoralimmune response against WNV in a natural host or whether DIII in general or the T332 epitope in particular are important targets of neutralizingantibodies in vivo. To characterize the types of antibodies produced during infection with WNV, we studied a group of naturally infected horses.Using immune adsorption assays coupled with the use of virus particles bearing mutations in the T332 epitope, we found that in some animalsneutralizing activity against DIII and the T332 epitope was below the limit of detection. In contrast, some animals generated a significant fractionof neutralizing activity to DIII and the T332 epitope. Thus, while antibodies to the T332 epitope did not represent a significant fraction of the totalantibody response in the infected animals studied, in some horses, they comprised a significant fraction of neutralizing activity, making this animportant but far from dominant neutralizing epitope. Rather, the neutralizing response to WNV generated in infected horses is both variable andpolyclonal in nature, with epitopes within and outside of DIII playing important roles.© 2006 Elsevier Inc. All rights reserved.

Keywords: West Nile virus; Neutralization; Flavivirus; Domain III; Envelope protein; Horses

Introduction

West Nile virus (WNV), the causative agent of West Nilefever and encephalitis, has rapidly spread throughout theWestern Hemisphere since its appearance in 1999. WNV isnot only an important human pathogen, but also a majorveterinary pathogen. Disease associated with WNV infectionhas become a significant clinical and economic burden for theequine industry in the United States, with 23,742 cases reportedthroughout the country since 1999 (as of March 2006) (http://www.aphis.usda.gov/vs/ceah/ncahs/nsu/surveillance/wnv/wnv.htm). As in humans, the majority of WNV infections in horsesare subclinical, with a subset of horses developing neurologicsigns including paralysis, ataxia, muscle fasciculations and

⁎ Corresponding author. Fax: +1 215 898 9557.E-mail address: doms@mail.med.upenn.edu (R.W. Doms).

0042-6822/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.virol.2006.08.047

tremors. Among clinically affected horses, mortality rates rangefrom 30 to 40% (Murgue et al., 2001; Ostlund et al., 2001;Porter et al., 2003; Salazar et al., 2004; Snook et al., 2001).Although no specific anti-viral treatments for WNV exist todate, several veterinary vaccines have been developed (Agrawaland Petersen, 2003; Minke et al., 2004; Siger et al., 2004). Inaddition, live, attenuated chimeric human vaccines are under-going clinical trials (Arroyo et al., 2004; Monath et al., 2006;Pletnev et al., 2006).

WNV virions contain two membrane proteins, pre-mem-brane/membrane (prM/M) and envelope (E). The E proteincovers the majority of the viral surface and is the major target ofantibodies generated against WNV during infection. The hu-moral immune response plays a critical role in protection fromand clearance of WNV infection in animal models (Ben-Nathanet al., 2003; Camenga et al., 1974; Diamond et al., 2003a, 2003b;Engle and Diamond, 2003; Gould et al., 2005; Tesh et al., 2002;

Fig. 1. Levels of WNV-specific IgG in naturally infected horses. WNV-specificIgG antibodies present in horse serum samples were measured using WNVsubviral particles (SVPs) (A) or domain III (DIII) (B) as antigen. Horses 4–7(black lines and symbols) represent uninfected controls. Horses 9–20 (gray linesand symbols) represent WNV-infected horses. Normalized optical density (OD)values are shown. Two WNV-infected horses had undetectable levels of WNV-specific IgG in this assay (9 and 12). Results represent the average of twoindependent experiments.

337M.D. Sánchez et al. / Virology 359 (2007) 336–348

Wang et al., 2001). The protective effects of WNV-specificantibodies may be attributed to several different functionalproperties including an ability to fix complement, mediateantibody-dependent cytotoxicity and to directly interfere withthe attachment or entry of virions. The latter process has beenstudied extensively in vitro (Beasley and Barrett, 2002; Gollinsand Porterfield, 1986; Gould et al., 2005; Oliphant et al., 2005;Peiris et al., 1982; Sanchez et al., 2005).

Flavivirus E proteins fold into three structural domains: I, IIand III (Modis et al., 2003, 2005; Rey et al., 1995). The largemajority of the potent murine neutralizing monoclonal anti-bodies (MAbs) described to date against WNV bind to domainIII (DIII), specifically to an epitope on the lateral surface of DIIIdescribed in the literature as a dominant neutralizing epitope(Beasley and Barrett, 2002; Oliphant et al., 2005; Sanchez et al.,2005). This neutralizing epitope includes residues clusteredaround amino acids 306, 307, 330, and 332 and is referred tohere as the T332 epitope. An additional neutralizing epitope inDIII has been described involving residues 310 and 394, wherea weakly neutralizing MAb mapped (Oliphant et al., 2005).Many neutralizing antibodies against related flaviviruses bind toDIII, although neutralizing antibodies against other domainshave also been described (Mandl et al., 1989; Roehrig, 2003;Roehrig et al., 1998).

While the most potent neutralizing MAbs to WNV describedto date bind to DIII, and in particular to the T332 epitope, theseantibodies have been derived by experimental vaccination andinfection of rodent animal models. The importance of DIIIneutralizing epitopes during natural WNV infection remains tobe determined. In this study, we investigated the importance andprevalence of antibodies against DIII, including the T332epitope, during natural WNV infection in horses. We found thathorses make antibodies to epitopes throughout the E protein,including the T332 epitope. However, removal of DIII-specificantibodies reduced the neutralizing activity of serum from onlya subset of horses. Likewise, WNV particles bearing multiplemutations in the T332 epitope were effectively neutralized byall horse sera examined, though sometimes less efficiently.Therefore, DIII is an important target for neutralizing antibodiesin some but not all infected animals. Furthermore, the T332epitope, while sometimes important, is not a predominantneutralizing determinant in WNV-infected horses. Rather, theneutralizing antibody response in infected horses is clearlypolyclonal in nature, with epitopes within and outside of DIIIbeing targeted.

Results

Antibody response against WNV in naturally infected horses

A group of WNV-infected horses was studied to examine thetypes of antibodies produced during natural infection and todetermine if epitopes in domain III (DIII) of the envelope (E)protein are important for neutralization of WNV in vivo. WNV-infected horses used in this study were examined at the Schoolof Veterinary Medicine at the University of Pennsylvania during2003. Serum from control uninfected horses was collected at the

School of Veterinary Medicine at the University of Pennsylva-nia prior to the introduction of WNV into the WesternHemisphere. Serum samples from uninfected (control) andWNV-infected horses were examined for antibody levels (IgGand IgM) and neutralization titers. IgG levels were measured byassessing the reactivity of horse serum to WNV subviralparticles (SVPs) and purified WNV DIII protein by ELISA(Sanchez et al., 2005). SVPs are non-infectious virus-likeparticles lacking the capsid protein and genomic RNA that aresecreted from cells expressing full-length prM and E proteinsand share many of the antigenic and functional properties ofnative mature virus particles (Allison et al., 1995; Hanna et al.,2005; Hunt et al., 2001; Konishi and Fujii, 2002; Konishi et al.,1992, 2001; Kroeger and McMinn, 2002; Mason et al., 1991;Schalich et al., 1996). SVPs were used to determine the overallimmune response against the envelope proteins of WNV (prM/M and E) while the purified DIII protein was used to assessspecific responses against the domain incorporating theconserved T332 epitope (Beasley et al., 2004; Beasley andBarrett, 2002; Oliphant et al., 2005; Sanchez et al., 2005).

Sera from infected horses exhibited considerable variabilityin their reactivity to both SVPs and DIII, both with regard toantibody titer (IgG) as well as maximal reactivity seen at

338 M.D. Sánchez et al. / Virology 359 (2007) 336–348

saturating concentrations of antibody (Fig. 1). All uninfected(control) horses and two WNV-infected horses (Horses 9 and12) had little to no antibody reactivity to SVPs or to DIII. WNV-specific IgM antibodies were detected in the serum of horse 9which suggests that the sample was obtained during the acuteperiod of infection before the animal was able to mount a robustIgG response (data not shown). Horse 12, a 9-month-old foal,did not have any detectable IgM in its serum, which suggeststhat this young animal was unable to mount an immuneresponse to viral infection. Despite variability in antibody levels(IgG), sera from all animals exhibited similar reactivity to bothSVPs (Fig. 1A) and purified DIII (Fig. 1B).

We also tested the reactivity of horse sera against a panel of 56overlapping peptides (15–19 amino acids in length) spanningthe entire length of the E protein (Peptide Array West Nile VirusGene E, NIH Biodefense and Emerging Infections ResearchResources Repository, NIAID) to identify linear epitopes recog-nized during the course of infection. However, the majority ofthe naturally infected horses did not produce IgG antibodies thatreacted with any of the linear epitopes encoded by this peptidelibrary, while a few horses reacted with two peptides spanningsequences on domain II that, based on the structure of the denguevirus E protein, are likely to be surface exposed (Table 1). Nosingle peptide was recognized by all horses, and most horsesshowed very weak reactivity to the positive peptides. Therefore,naturally infected horses do not efficiently produce antibodiesagainst short, linear epitopes of the E protein.

Domain III epitopes targeted during natural infection

To determine the contribution of antibodies that bind residueson DIII to the total humoral immune response in horses duringWNV infection, we performed experiments to determine if thebinding of horse serum to soluble DIII could block subsequentbinding of monoclonal antibodies (MAbs) to this protein. Themajority of the potent murine neutralizing antibodies describedagainst WNV map to the T332 epitope, which is on the lateralsurface of DIII (Beasley and Barrett, 2002; Oliphant et al., 2005;Sanchez et al., 2005). Although many antibodies are producedagainst other regions of the E protein in mice, the regionsurrounding the T332 epitope seems to be a preferred target forthe production of potent neutralizing antibodies. Three MAbswere used in these experiments: 4E1, 8B10 and 11C2 (Sanchezet al., 2005).MAbs 8B10 and 11C2map to the T332 neutralizingepitope onDIII (Fig. 4A), while 4E1maps outside of this epitopeat an undetermined location in DIII. ELISA plates were coated

Table 1Reactivity of WNV positive horse serum against linear epitopes

Peptide Sequence

10 63LSTKAACPTMGEAHNDKR80

11 71TMGEAHNDKRADPAFVCR88

Peptides tested span the entire length of the WNV E protein. The number of the pepdetermined for each horse using an average of the absorbance values of the lower 10 pand divided into three groups: 4- to 6-fold, 6- to 8-fold and >8-fold. The total numbabsorbance of more than 4-fold the cut-off value are listed. A few peptides were consispositive horses, therefore reactivity with these peptides was deemed to be non-spec

with a purified WNV DIII protein after which serial dilutions ofhorse sera were added to the wells and allowed to bind for 1 h.The time of incubation and amount of horse serum added wereoptimized to ensure maximal occupation of antibody bindingsites on DIII. Non-saturating concentrations of the indicatedMAbs were subsequently added for 20 min, the least amount oftime needed to achieve adequateMAb binding. This was done tominimize the possibility of the MAbs competing off loweraffinity equine antibodies. A decrease in the amount of MAbbinding to DIII would suggest that, under the conditions tested,the specific horse serum has antibodies targeting the same regionwhere the MAb binds.

Competition ELISAs were performed with horses that hadintermediate (horses 10, 17 and 18) or high titers (horses 11, 14,15, 16 and 19) of anti-WNV antibodies (IgG) as measured byELISA (see Fig. 1). Horses with high WNV ELISA titersefficiently blocked binding of neutralizing MAbs to DIII, mostat serum dilutions ranging from 40-fold to 80-fold (Fig. 2A,MAb 8B10). Decreasing amounts of serum decreased the abilityto block MAb binding as expected (40-fold vs. 160-folddilution). These results indicated that the T332 neutralizingepitope is a target of the humoral immune response duringnatural infection of WNV in horses. Other sites on DIII are alsotargets since sera from these horses were able to block infectionof a different MAb (4E1) which binds outside of this dominantneutralizing epitope (Fig. 2B). However, sera from horses withintermediate titers by ELISA (10, 17 and 18) did not block orweakly blocked subsequent binding of both MAbs. Thissuggests that these horses either do not have significant levelsof antibodies against regions in DIII or that horses could havelow affinity antibodies that were unable to compete with highaffinity MAbs, and therefore not detected by the competitionassay. Lower affinity antibodies may also explain the lower IgGtiters obtained (Fig. 1) in those serum samples. No correlationwas evident between the antibody blocking ability of serumsamples and levels of WNV-specific IgM (data not shown).

Since the purified DIII fragment lacks other domains of the Eprotein that may be targeted during the immune response, werepeated the competition ELISAs with a subset of horse serumsamples using SVPs rather than soluble DIII as the antigen. Weperformed experiments to determine if the binding of horseserum to SVPs could block subsequent binding ofMAbs to the Eprotein in the context of a polyvalent virus particle. When SVPswere used as antigen, similar patterns of competition wereobserved with all DIII antibodies (4E1, 8B10 and 11C2) and a DIantibody (17D7) (Fig. 3). Interestingly, the degree to which

4 to 6-fold 6 to 8-fold >8-fold

0 1 (H 11) 1 (H 16)1 (H 17) 0 2 (H 16, 18)

tide, the sequence and magnitude of reactivity are shown. Cut-off values wereeptides. Absorbance values of more than 4-fold the cut-off values were calculateder of horses and the specific horse (H) which reacted with the peptides with antently positive (to the same level) with control, uninfected horses and with WNVific.

Fig. 2. Serum from WNV-infected horses blocks subsequent binding of mono-clonal antibodies (MAbs) to WNV DIII. Purified DIII proteins were incubatedwith different dilutions of horse serum fromWNV positive horses (10, 11, 14–19)or an uninfected control horse (C; used at 1:20 dilution). Dilutions of horse serumwere performed at a range of 1:20–1:640 (2–64 on x-axis). Each horse is shownwith the highest dilution first (i.e. 16, or 1:160) followed by lower dilutions and,therefore, increasing amounts of horse serum. Subsequent to horse serum binding,incubation with MAbs against DIII (8B10, 4E1) was performed. Binding ofMAbs was detected with an anti-mouse HRP conjugate. Percentage binding of theMAb represents the percent of signal obtained with incubation with serum fromthe uninfected control horse. Error bars represent the standard error of the mean.Increasing concentrations of most horse seras tested decreased subsequentbinding of the MAbs tested to DIII.

Fig. 3. Serum from WNV-infected horses blocks subsequent binding ofmonoclonal antibodies (MAbs) to subviral particles (SVP). SVPs were incubatedwith different dilutions of horse serum from a subset ofWNVpositive horses (10,15, 19) or an uninfected control horse (C; used at 1:80 dilution). Dilutions ofhorse serum were performed at a range of 1:80–1:4920 (8–492 on x-axis). Eachhorse is shown with the highest dilution first (i.e. 128 or 1:1280) followed bylower dilutions and, therefore, increasing amounts of horse serum. Subsequent tohorse serum binding, incubation with MAbs against DIII (11C2, 8B10, 4E1) orDI (17D7) was performed. Binding of MAbs was detected with an anti-mouseHRP conjugate. Percentage binding of the MAb represents the percent of signalobtained with incubation with serum from the uninfected control horse. Errorbars represent the standard error of the mean. Increasing concentrations of thehorse seras tested decreased subsequent binding of all MAbs to SVPs.

339M.D. Sánchez et al. / Virology 359 (2007) 336–348

MAb binding was blocked by horse serum was consistentlygreater when SVPs were used as the antigen. In addition, horsesera that did not block binding of MAbs to DIII (see Horse #10)did block binding of MAbs to SVPs, although the degree ofblocking was less than with other horse sera. These differencescould be due to a higher avidity of antibodies in horse seraagainst the E protein in the context of a virus particle.Alternatively, since E proteins exist as dimers on the surfaceof virus particles (Kuhn et al., 2002; Mukhopadhyay et al.,2003), binding of an antibody to one E protein could blocksubsequent binding of an antibody to a different site on anadjacent E protein. Nonetheless, these experiments show thathorses naturally infected with WNV produce antibodies againstseveral regions in DI and DIII, including antibodies that canblock subsequent binding of MAbs to the T332 neutralizingepitope.

After establishing that horses infected with WNV produceantibodies to DIII, we sought to determine if the majority ofantibodies produced against DIII were against the T332neutralizing epitope. To do this, we determined if sera fromWNV-infected horses could detect a purified DIII protein where

340 M.D. Sánchez et al. / Virology 359 (2007) 336–348

four critical residues in the conserved neutralizing epitope weremutated (306, 307, 330 and 332; Δ67IA) (Fig. 4A). Wild-type(WT) and mutant Δ67IA DIII proteins were used in ELISA to

Fig. 4. Mutations in DIII of the WNV Envelope protein. (A) Diagram of theWNV Envelope protein with an enlarged schematic of domain III (in blue) withresidues mutated for either reporter virus particle (RVP) or purified DIIIproduction. In red are residues 306, 307, 330 and 332 (Δ67IA mutant) and ingreen are residues 310 and 394. (B) Neutralization assays with wild-type andmutant (Δ67IA and Δ310/394) RVP luciferase and monoclonal antibodies(MAbs). RVPs were incubated with no antibody (0, No Ab) or with twoconcentrations of MAbs (10 μg/ml or 0.1 μg/ml). Infections were normalized tothe amount of luciferase produced when RVPs were not incubated with antibody(No Ab). MAb 17D7 maps to DI and is non-neutralizing. Neutralizing MAbs11C2 and 8B10 map to residues 306/307/330/332 which have been mutated inΔ67IA RVPs, therefore these MAbs lose neutralizing activity with this mutant.Neutralizing MAb E1 maps to residues 310/394 which have been mutated inΔ310/394 RVPs, therefore E1 lost neutralizing activity with this mutant. Errorbars represent the standard error of the mean.

determine reactivity of sera from control and WNV-infectedhorses. All horse sera with titers high enough to reliably detectthe WT DIII protein also readily detected the mutant Δ67IADIII protein (data not shown), indicating that antibodiesdependent upon residues 306, 307, 330 and 332 compriseonly a minor fraction of all the antibodies directed against DIII.

Reporter virus particle neutralization assay

To determine the functional properties of antibodies presentin the sera of each horse, we utilized a recently describedreporter virus particle (RVP)-based neutralization assay tocharacterize the neutralizing activity in each serum sample(Pierson et al., 2006). The RVP neutralization assay employsparticles that are capable of only one round of infection andencapsidate a sub-genomic RNA that expresses a reporterprotein. We have recently completed a study examining theequine neutralizing antibody response using both the RVP-based neutralization assay and traditional assays that utilizeinfectious virus to confirm that this novel assay reliably detectsneutralization activity in serum samples from infected animals(unpublished data). All WNV-infected horses exhibited neu-tralizing activity except horse 12, which did not have detectableWNV-specific IgM or IgG antibodies by ELISA (Fig. 1 and datanot shown). One of the uninfected control horses (#4) had veryweak neutralizing activity against RVPs (Table 2). It is possiblethat this horse was infected with a related flavivirus (i.e. SaintLouis encephalitis virus) and had developed cross-reactiveantibodies against WNV. For this group of naturally infectedhorses, the levels of IgG antibodies (Fig. 1) did not correlatewith the strength of the neutralization titers (Table 2), whichmay imply that low antibody titers do not necessarily translateinto weak neutralizing activity or lack of protection fromdisease (Davis et al., 2001; Siger et al., 2004; Tesh et al., 2002).It is possible that making antibodies against a few specificresidues may provide a sufficient amount of protection againstthe virus if these antibodies target key residues in the protein.

Neutralization of mutant reporter virus particles

Our MAb competition studies indicated that WNV-infectedhorses produce antibodies that bind to or near the T332 neutra-lization epitope in DIII. To determine if antibodies to this epitopecomprise a significant fraction of neutralizing antibodies in vivo,we produced RVPs bearing mutations in DIII that preventbinding of antibodies to two major neutralizing epitopes. Weprepared RVPs in which residues 306, 307, 330 and 332(Δ67IA) were mutated, or in which residues 310 and 394 (Δ310/394) were mutated (Fig. 4A). Neutralization assays were thenperformed using WT and mutant RVPs to ensure that the muta-tions abrogated binding of MAbs targeting the mutated epitopes(Fig. 4B). Four different MAbs were used: 17D7 binds to do-main I and is not neutralizing: 8B10 and 11C2 are neutralizingMAbs that map to the epitope comprised of residues 306, 307,330 and 332; and E1 maps to the neutralization epitope com-prised of residues 310 and 394 (Oliphant et al., 2005; Sanchez etal., 2005). All neutralizing MAbs (8B10, 11C2 and E1)

Fig. 5. Mutations in residues 306/307/330/332 (Δ67IA) can affect neutralizingactivity of horse serum. Scatter plot shows results from at least 12 independentneutralization assays with reporter virus particles expressing luciferase as areporter protein. The EC50 neutralization titers were calculated using non-linearregression curve fits. The ratio of the EC50 neutralization titer for wild-type(WT) over mutant (Δ67IA) is plotted (WT/67IA). Each symbol represents theratio of an independent experiment. Horizontal bars represent the mean of theratios. A ratio of 1 indicates that there was no difference in neutralization ofWT or mutant reporter virus particles. The larger the ratio (>1), the more horseserum was required to neutralize the mutant versus the WT. Table shows resultsof statistical analysis of the difference in EC50 titer between WT and Δ67IA(two-tail paired t-test). P value<0.05 is significant. Mean EC50 ratios(WTEC50/Δ67IAEC50) and confidence intervals for the ratios obtained areshown.

Table 2Average neutralization titers of WNV-infected horses

Horse no. Average EC50 titer (reciprocal dilution)

4 (C) a 272.45 (C) <406 (C) <407 (C) <409 187,87510 403311 34,79412 <4014 26,15315 40,68916 42,03117 108,17818 104,30019 139,97920 20,534

The reciprocal of the dilution of horse serum required to block 50% of infection(EC50) on Vero cells is shown. At least two independent replicates wereaveraged for each horse serum sample.a C—control, uninfected horses.

341M.D. Sánchez et al. / Virology 359 (2007) 336–348

effectively blocked infection of WT RVPs, with E1 having theweakest activity as previously described (Oliphant et al., 2005).When the Δ67IA mutant RVPs were tested, the MAbs whichmap to the residues mutated (8B10 and 11C2) were not able toneutralize infection, while neutralization of these particles by E1was not affected. In contrast, when residues 310/394 weremutated, E1 lost neutralizing activity, while 8B10 and 11C2effectively neutralized this mutant particle. These results suggestthat the mutations engineered into the envelope protein do notsignificantly affect the overall structure or arrangement of the Eprotein on the virus particle since the resulting virus particleswere infectious and retained the ability to bind antibodies thatreact with distant epitopes.

To address the prevalence of antibodies against the T332neutralizing epitope in horses naturally infected with WNV, weperformed neutralization assays utilizing the mutant RVPs. Wehypothesized that if the majority of neutralizing activity presentin horse serum is against these neutralization epitopes (306/307/330/332 or 310/394), these horses would have decreased abilityto neutralize the mutant RVPs. In contrast, if the majority ofneutralizing activity elicited by natural WNV infection targetsother regions of the E protein, then the mutant RVPs should beneutralized as efficiently as wild-type (WT) RVPs. It isimportant to note that, in our assays, the concentration ofantigen present in RVP preparations was sufficiently low toallow measurements of neutralization at steady state andachieve results that are independent of the concentration ofRVPs used (Pierson et al., 2006). WT and mutant RVPs wereincubated with dilutions of horse serum and 50% neutralizationtiters (EC50) calculated. EC50 titers were similar for all horsestested with WT and Δ310/394 mutant particles (data notshown), suggesting that antibodies directed to the epitopedefined by residues 310/394 do not comprise a significantfraction of the neutralizing activity in serum from WNV-infected horses. However, when the Δ67IA mutant RVPs wereused, different results were obtained. Sera from some infected

horses neutralized Δ67IA RVPs as effectively as WT particles.However, for a subset of the horses, significantly higherconcentrations of serum were required to neutralize the mutantΔ67IA particles than WT. For these horses (horses 11, 16 and19), the EC50 neutralization titers differed 2- to 5-fold betweenthe mutant Δ67IA and WT particles (Fig. 5A and B). Thus, insome but not all WNV-infected animals, antibodies that bind tothe DIII epitope that includes residues 306, 307, 330 and 332comprise a significant fraction of total neutralizing activity.Levels of WNV-specific IgM did not correlate with theneutralization potency of serum samples against wild-typeversus mutant reporter virus particles (data not shown).

Impact of DIII neutralizing antibodies

Our data showed that, while some WNV-infected horsesproduce significant neutralizing activity against the T332neutralization epitope in DIII, others do not. To determine ifin these animals neutralizing antibodies might bind to other asyet unidentified epitopes in DIII, we performed serum depletionstudies. To do this, horse serum was incubated with beadscoated with purified DIII protein, after which the beads wereremoved. In addition, horse serum was incubated with beadswithout DIII to control for non-specific removal of antibodies

342 M.D. Sánchez et al. / Virology 359 (2007) 336–348

by beads. We confirmed that this procedure effectively removedDIII antibodies from the sera as assessed by ELISA (Fig. 6A).Neutralization assays were then performed. If a large fraction ofthe neutralizing antibodies in horse serum are against DIII, wewould expect to see a decrease in neutralizing activity afterremoval of DIII-specific antibodies. We found that removal ofDIII antibodies from sera obtained from horses 10 and 15 had

Fig. 6. Removal of DIII-specific antibodies affects neutralization activity in a subsetantibodies by ELISA before binding to beads (pre-beads, white bars) or after binding bbars). A control, uninfected horse serum is shown for background (C). For all WNV pantibodies, while incubation with beads without DIII did not non-specifically remodensity. (B) Neutralization assays with serum pre-beads or after treatment with beads frtreated with ‘No DIII’ beads (▲, triangles) or serum treated with ‘DIII’ beads (●, cirexpressing GFP measured by flow cytometry 44 h after infection. For some horses (sethe serum. However, for other horses (see 11, 16 and 19), removal of DIII antibodies deinfection. One representative graph from each horse is shown.

no effect on neutralization titers, suggesting that in these horsesmost neutralizing antibodies are directed against epitopesoutside of DIII or to epitopes that are not present in soluble,monomeric DIII. These results are consistent with the fact thatsera from horses 10 and 15 neutralized the Δ67IA DIII mutantas effectively as wild-type virus particles. In contrast, removingDIII antibodies from sera obtained from horses 11, 16 and 19

of horses. (A) Horse serum samples were tested for the presence of DIII-specificeads with or without DIII protein (Beads-No DIII, black bars or Beads-DIII, grayositive horses, incubation with DIII beads removed the majority of DIII specificve DIII antibodies. Error bars represent the standard deviation. O.D.—Opticalom part A. RVPswere incubated for 1 hwith serum pre-beads (■, squares), serumcles). Vero cells were infected after the incubation time and the percent of cellse 10 and 15), removal of DIII antibodies did not affect the neutralizing activity ofcreased the neutralizing activity of the serum. The control horse did not neutralize

343M.D. Sánchez et al. / Virology 359 (2007) 336–348

decreased neutralizing activity by an average of 3-fold (Fig.6B). For horses 16 and 19, the effect of removing DIII-specificantibodies was more prominent than for horse 11. This suggeststhat a significant fraction of neutralizing activity in these horsesis produced against DIII since removal of these antibodiesnoticeably decreased the neutralizing activity of the serum. Thisconclusion is also supported by the fact that sera from theseanimals neutralized the Δ67IA DIII mutant at least several foldless effectively than wild-type virus. Combined, these dataindicate that, in some naturally infected horses, a significantfraction of the neutralizing activity targets not only DIII, but theT332 neutralizing epitope on the lateral surface of DIII. Indeed,removal of DIII antibodies from serum of these three horses hadan almost identical effect on neutralizing activity as the use ofthe Δ67IA DIII mutant RVPs, which confirmed that in thissubset of horses other epitopes in DIII were not targeted byneutralizing antibodies (data not shown). However, otherinfected animals mounted highly effective neutralizingresponses that most likely involve determinants outside of DIII.

Discussion

Previous studies have shown the importance of an intacthumoral immune response in controllingWest Nile virus (WNV)infection in animal models (Ben-Nathan et al., 2003; Camenga etal., 1974; Diamond et al., 2003a, 2003b; Engle and Diamond,2003; Gould et al., 2005; Halevy et al., 1994; Tesh et al., 2002;Wang et al., 2001). Case reports of recovery after treatment withWNV-specific immunoglobulin preparations in human patients(Agrawal and Petersen, 2003; Haley et al., 2003; Hamdan et al.,2002; Shimoni et al., 2001) have provided additional incentivesto understand the humoral immune response generated duringinfection and to create treatments based on antibodies. Inaddition, recent studies with monoclonal antibodies haveprovided insight about neutralization epitopes on the WNVenvelope (E) protein and mechanisms of viral neutralization byantibodies (Beasley and Barrett, 2002; Gould et al., 2005;Oliphant et al., 2005; Sanchez et al., 2005). However, little isknown about the humoral immune response in a natural host ofinfection. In this study, we investigated the antibody responseagainst WNV during natural infection of horses to improve ourunderstanding of the epitopes targeted during infection in anatural host for the virus.

The majority of the potent neutralizing murine MAbs thathave been described to date bind to DIII of the WNV E protein(Beasley and Barrett, 2002; Oliphant et al., 2005; Sanchez et al.,2005), with many of these binding to the T332 epitope on thelateral surface of DIII. Studies with murine MAbs have alsoshown that a group of neutralizing MAbs againstWNValso mapoutside of DIII, although these tend to be less potent thanantibodies against DIII (Oliphant et al., 2005; Razumov et al.,2005). Humans infected with WNV produce antibodies that canblock binding of MAbs to the T332 epitope, but whether thisregion is an important neutralization target in vivo is not known(Oliphant et al., 2005). Human phage display libraries have beenused to identify neutralizing single-chain variable regionantibody fragments that bind to epitopes in DI and/or DII

(Gould et al., 2005). Similarly, primate or human-derivedantibodies isolated from Fab libraries for other relatedflaviviruses have mapped outside of DIII (Daffis et al., 2005;Goncalvez et al., 2004b). Therefore, the WNV E proteinpresents a variety of neutralizing epitopes of uncertain in vivosignificance.

We sought to determine if epitopes in DIII, and the so calleddominant neutralizing T332 epitope in particular, are importanttargets for neutralizing antibodies during natural WNV infectionin a group of horses. Horses serve as a good model to studyWNV infection in a natural host. As in humans, the majority ofinfections in horses are subclinical, with a subset of horsesdeveloping devastating neurologic disease. In addition, bothhorses and humans develop low levels of viremia after infectionand are considered dead-end hosts for WNV transmission.Previous serologic surveys have shown that naturally infectedhorses develop strong neutralizing antibody responses againstthe virus (Davidson et al., 2005; Schmidt and Elmansoury,1963). However, it is not known which epitopes are targeted bythe antibodies generated after natural infection of equines in thefield.

The majority of WNV-infected horses developed a robusthumoral immune response (IgG) against SVPs and a purifiedDIII protein. Interestingly, the infected horses preferentiallyproduced conformational antibodies against the E protein sincethe serum samples showed poor reactivity against a library ofoverlapping peptides spanning the entire length of the WNV Eprotein. An earlier study has shown that immunization withdenatured E protein fragments failed to induce neutralizingantibodies in rabbits (Wengler, 1989), though studies withdengue virus have shown that humans infected with the virusproduce antibodies against a variety of linear epitopes on the Eprotein (Aaskov et al., 1989; Innis et al., 1989). It is unclear if thedifference in our results is due to the differences betweenflaviviruses or a species difference.

We found that sera from WNV-infected horses readilyrecognized both a purified wild-type DIII protein and a DIIIprotein with four mutations in the T332 neutralizing epitope(data not shown). Thus, antibodies that bind to the T332 epitopewere not prevalent in the sera of the naturally infected horsesstudied here. However, a more important question is whetherDIII and the T332 epitope are significant targets for neutralizingantibodies. Using two independent approaches, we found thathorses showed variability on the epitopes targeted by theneutralizing antibody response. Sera from some infected animalsneutralized WNV particles bearing multiple mutations in theT332 epitope as efficiently as wild-type particles. Likewise,when antibodies to soluble DIII were removed from these sera,neutralization titers were not affected. Thus, for some animals,neutralizing activity against both DIII and the T332 epitope werebelow the detection limit of our high-throughput neutralizationassay. In contrast, some infected animals did appear to generate asignificant fraction of neutralizing activity to DIII. When DIIIantibodies were adsorbed from these sera, neutralizing titerswere consistently reduced several fold. Likewise, while serafrom these animals could neutralize particles bearing multiplemutations in the T332 epitope, titers were again reduced several

344 M.D. Sánchez et al. / Virology 359 (2007) 336–348

fold. This suggests that, when neutralizing antibodies aregenerated against DIII, the T332 neutralizing epitope on thelateral surface of DIII is a functionally important major target. Inaddition, while T332 specific antibodies may not represent asignificant fraction of the total antibody response after infection,in a subset of horses, they do comprise a significant fraction ofneutralization activity.

Our data indicate that, in horses naturally infected withWNV,neutralizing activity directed against the T332 epitope, and moregenerally DIII, is not consistently detected even though the T332epitope is conserved among many WNV strains (Li et al., 2005;Nybakken et al., 2005). Given recent reports of neutralizingmonoclonal antibodies outside of DIII (Goncalvez et al., 2004a,2004b; Gould et al., 2005; Oliphant et al., 2005; Razumov et al.,2005), we speculate that DII may be an important target forneutralizing antibodies in these horses. While the neutralizationepitopes in DIII may be important and definitely favored in theisolation of potent neutralizing MAbs derived from mice, theyare clearly not the only neutralizing epitopes targeted duringnatural infection. It is evident that neutralizing antibodies againstother regions of the envelope protein can be effective inneutralizing virus infection in vitro and are produced in vivo.

Materials and methods

Cell lines

Human embryonic kidney cells (HEK) 293Twere maintainedin Dulbecco's Modified Essential Media (DMEM) (LifeTechnologies, Gaithersburg, MD) containing 10% fetal calfserum (FCS) and 1% penicillin/streptomycin (P/S) (Invitrogen,Carlsbad, CA). Vero cells weremaintained inMinimumEssentialMedia (MEM) supplemented with 10% FCS/1% P/S/1% L-glu-tamine. All cell lines were maintained at 37 °C in 5% CO2.

Horse serum samples

Samples were obtained from WNV-infected horses seen atthe School of Veterinary Medicine at the University ofPennsylvania-New Bolton Center during 2003. All bloodsamples were obtained during either acute infection or post-mortem analysis of patients. Horses were diagnosed with WNVinfection by virus detection (virus isolation, RT-PCR, immu-nohistochemistry) or WNV-specific antibody detection asdetailed by the Center for Disease Control and Preventionguidelines (http://www.cdc.gov/ncidod/dvbid/westnile/resources/wnv-guidelines-apr-2001.pdf). Detailed clinical information re-garding outcome of infection or time of blood sample relative toinfection was not available for most horses. Control, uninfectedserum samples were collected from control horses housed at NewBolton Center in 1992, prior to the introduction of WNV into theWestern Hemisphere.

Production of subviral particles

Wild-type SVPs were produced by lipid-mediated transfec-tion of HEK 293T cells with a DNA plasmid encoding wild-

type Lineage I pre-membrane/membrane and envelope protein(prM-E) from strain NY99-6480 (pCBWN) (Davis et al.,2001) following manufacturer's instructions (Lipofectamine2000, Invitrogen). Five to six hours after transfection, themedia was changed to a low-glucose formulation of DMEM.Supernatants were harvested 48 h after transfection, clarifiedby low-speed centrifugation, filtered through a 0.45 μm-poresize filter, and ultracentrifuged in Beckman SW41Ti rotor at4 °C as previously described (Sanchez et al., 2005). Forunpelleted SVPs, supernatants were harvested, filtered andstored at 4 °C.

WNV DIII protein production

Production and purification of the WNV DIII protein(WED3) in bacteria have been previously described (Sanchezet al., 2005). Briefly, a fragment containing amino acid residues296–415 was amplified by PCR and cloned into the pBAD/ThioTOPO expression vector (Invitrogen). TOP10 Escherichia colicells were transformed and 5-liter cultures grown and inducedwith arabinose, as specified by the manufacturer. The proteinwas purified under denaturing conditions, dialyzed against PBS,quantified and purity assessed by silver stain. Proper folding ofthe DIII fragment was assessed by reactivity with conforma-tional MAbs by ELISA.

The following amino acid residues in domain III weremutated using the QuikChange Site-Directed Mutagenesis kit(Stratagene): S306A, K307R, K310A, T330I, T332A andN394A. Double and quadruple mutants of these residueswere produced as follows: Δ310/394 and Δ306/307/330/332.Sequence analysis of the entire open-reading frame confirmedthat the only mutations present were those listed above.

ELISAs

Horse serum titers were evaluated by ELISA using bothpurified DIII protein and pelleted wild-type SVPs as antigen. Forexperiments using DIII as antigen, 0.5 μg/ml of DIII was ad-sorbed on 96-well plates (Immulon 2 HB, Thermo Labsystems)in ELISA coating buffer (0.15 M sodium carbonate, 0.35 Msodium bicarbonate in PBS pH 9.6) overnight at 4 °C. Plateswere washed using an automatic plate washer (Ultrawash PLUSMicroplate Washer, Dynex Technologies) with PBS/0.05%Tween 20 and blocked with PBS/0.05% Tween/1% BSA for 1h at room temperature (RT). Two-fold serial dilutions of horseserum were added to the plate and allowed to bind for 1 h. Afterwashing, a rabbit anti-horse IgG antibody (Sigma) was added ata 1:5000 dilution and allowed to incubate for 1 h. Following afinal wash, TMB Peroxidase Substrate and Peroxidase Substratesolution B (KPL) were mixed at 1:1 dilution, 100 μl added toeach well, the reaction stopped with 50 μl 2MH2SO4 per welland optical density at 450 nm (OD450) measured. For experi-ments using SVPs as antigen, 100 ng/well of the E-specificflavivirus MAb 4G2 was coated on plates in coating bufferovernight. After washing, a 1:60 dilution of pelleted SVPs wasadded to the plates and allowed to bind for 2 h. The rest of theELISA was performed as described above. OD values for all

345M.D. Sánchez et al. / Virology 359 (2007) 336–348

serum samples were normalized by setting the OD value of horse11 serum at the highest dilution (1:40) to 100%. Horse 11 waschosen as a normalization control due to its high titer ofantibodies to the WNV E protein. Serial dilutions of horse 11sera were included on all plates to control for variability betweenplates and between experiments performed on different days.Therefore, values are presented as a percent of the maximalvalue of horse 11 (normalized O.D. values).

Peptide library ELISAs were performed as follows. Peptideswere obtained from the NIH Biodefense and Emerging Infec-tions Research Resources Repository, NIAID, NIH (PeptideArrayWest Nile Virus Gene E). The 67-peptide array spans the Egene of the NY99-flamingo382–99 strain of West Nile Virus(GenbankAF196835). Peptides are 15- to 19-mers, with 10 or 11amino acid overlaps. ELISA plates were coated with 3 μg ofeach peptide overnight in ELISA coating buffer. After overnightincubation, the plates were washed and blocked with PBS/0.05% Tween/10% FCS for 1 h at RT. Horse serum was added tothe plates at a 1:100 dilution and incubated for 2 h at RT. Afterwashing, anti-horse IgG antibody (Sigma) was added at a 1:5000dilution and allowed to incubate for 1 h. Development of plateswas performed as described above. The cut-off value for thedetermination of positive peptides was established by averagingthe absorbance levels of the 10 lowest signal peptides for eachhorse. Any peptide with an absorbance level higher than 4-foldthe cut-off value for each horse was determined to be positive.Peptides that were positive with control, uninfected horses werenot counted as positive even if they also reacted with the WNVpositive horses.

Competition binding ELISAs

Competition ELISAs were performed using both purifiedDIII protein and wild-type SVPs as antigen. For DIIIcompetition ELISAs, 0.5 μg/ml of WNV DIII protein wasadsorbed in ELISA coating buffer overnight at 4 °C. Plates werewashed with PBS/0.05% Tween 20 and blocked with PBS/0.05% Tween/1% BSA for 1 h at RT. Horse serum samples fromWNV naturally infected horses were diluted from 1:20 to 1:640in PBS/0.05% Tween/1% BSA and incubated at RT for 1 h.These horse serum dilutions corresponded to the concentrationsufficient to saturate the DIII binding sites. After incubation withthe horse serum, the indicated “competitor” MAb was added.Different concentrations of MAbs were used keeping in mindtheir Kd values and the minimal amount of antibody needed toobtain a reliable signal in the assay (Sanchez et al., 2005). Thefollowing MAbs were added at a non-saturating concentration:4E1 0.4 ng/100 μl (Kd 0.06 nM), 8B10 6 ng/100 μl (Kd0.09 nM) and 11C2 18 ng/100 μl (Kd 0.24 nM) without washingand the plate was incubated for 20 min at RT, which wasdetermined to be the optimal incubation time for MAb binding.The purity of all MAbs was similar (about 60% pure) (data notshown). Plates were then washed, and a 1:5000 dilution of ECLanti-mouse IgG from sheep conjugated to horseradish peroxi-dase (HRP) (Amersham Biosciences) was added and incubatedfor 1 h at RT. Following a final wash, the wells were developedas described above.

For competition ELISAs using SVPs, a 1:32 dilution ofSVPs (unpelleted and in media with no serum) was coated onplates with coating buffer overnight. Plates were blocked, andhorse serum dilutions added, from 1:80 to 1:4920 and incubatedfor 1 h at RT. After incubation with horse serum, the indicated“competitor” MAb was added. Different concentrations ofMAbs were used keeping in mind their Kd values (concentra-tions at least 30-fold greater than their Kd values were used) andthe minimal amount of antibody needed to obtain a reliablesignal in the assay. The following competitor antibodies wereused at the following concentrations: 4E1 at 40 ng/100 μl (Kd0.06 nM), 8B10 80 ng/100 μl (Kd 0.09 nM), 11C2 180 ng/100 μl (Kd 0.24 nM), 17D7 375 ng/100 μl (Kd 0.83 nM)(Sanchez et al., 2005). MAbs were added without washing andthe plate was incubated for 20 min at RT. A higher concentrationof MAbs was needed to obtain a reliable signal (OD) whenusing SVPs than when using purified DIII. Subsequently, plateswere washed and developed as described above.

Production of WNV reporter virus particles (RVPs)

WNV RVPs were produced by transfection of 293T cellswith a combination of three different plasmids (Pierson et al.,2006) using Lipofectamine 2000 according to the manufac-turer's instructions. Preparation of RVPs using this systemwas accomplished by transfection of pre-plated 293T cells ina 10-cm dish with plasmids encoding 13 μg capsid-prM, 4 μg E(wild-type or mutant) and 13 μg of WNV replicon expressing areporter gene (Renilla luciferase or GFP). Lipid–DNA com-plexes were washed off transfected cells 12–16 h post infectionand replaced with a low glucose formulation of DMEM/10%FCS/1% P/S/25 mM HEPES pH 8.0. RVP-containing super-natants were harvested 48 h post-transfection, filtered through a0.45 μm filter and frozen at −80 °C.

Reporter virus particle neutralization assay

For experiments with Renilla luciferase expressing particles,8000 Vero cells were plated in 100 μl per well of a 96-wellwhite opaque tissue culture treated plate (Microtest™ FlatBottom, Falcon, Becton Dickinson Labware) and allowed tosit for 1–6 h before infection. Infections were harvested at42–46 h later, and cells analyzed using a commerciallyavailable Renilla luciferase assay system (Promega) permanufacturer instructions. Briefly, after removing the super-natant, cells were lysed for 15 min at room temperature in45 μl lysis buffer, and 45 μl of assay buffer with luciferasesubstrate was added immediately before reading the reactionin a Luminoskan Ascent luminometer (Thermo ElectronCorporation). For experiments with GFP expressing particles,32,000 Vero cells in 100 μl per well were plated in a clear96-well plate and allowed to adhere. Cells were harvested36–40 h after infection, and GFP expression analyzed using aGuava EasyCyte cytometer (Guava Technologies).

For experiments involving neutralization of RVPs withmonoclonal antibodies (MAbs), luciferase expressing particleswere used. MAbs were incubated with a 1:1000 dilution of WT

346 M.D. Sánchez et al. / Virology 359 (2007) 336–348

RVP, 1:200 dilution of Δ67IA or 1:800 dilution of Δ310/394for 1 h before infecting Vero cells as above. Four differentMAbs were used: 17D7 (a DI non-neutralizing antibody),11C2 and 8B10 (DIII neutralizing antibodies) and E1 (DIIIneutralizing antibody) (Oliphant et al., 2005; Sanchez et al.,2005). Two different concentrations of MAbs were used,10 μg/ml or 0.1 μg/ml. Experiments were performed in triplicate.

For experiments involving neutralization of RVPs with horseserum samples, serial dilutions of the indicated horse serumwere incubated with the indicated dilution of RVP (eitherluciferase or GFP). For experiments using GFP expressingparticles, WTandΔ67IA RVPs were used at a dilution of 1:100.For experiments using luciferase expressing particles, WTparticles were used at 1:1000 and Δ67IA RVPs were used at adilution of 1:400. For all concentrations of RVP used, WT andmutant stocks, we empirically determined that antibody was inexcess by performing curves with different concentrations ofRVPs as detailed in Pierson et al. (2006). Horse serum dilutionswere filtered through a 0.2 μm-pore size filter before incubationwith RVPs. The initial dilution for the horse serum samples inneutralization assays with WT or mutant RVPs ranged from1:50 to 1:200 depending on the potency of the neutralizingactivity of the serum. Ten three-fold serial dilutions of horseserum were performed in MEM. Neutralization curves for bothluciferase and GFP assays were evaluated using non-linearregression analysis (GraphPad Prism 4, GraphPad Software,Inc.). The concentration of horse serum required to neutralize50% of infection (EC50 titers) was determined for all curveswith an R2 value of more than 0.7.

For experiments comparing neutralization of WT versusmutant RVPs, the ratio of the EC50 values (WT/mutant) wasdetermined for each independent experiment in order to accountfor the variability between assays and analyze the data in a pairwise fashion. Statistical analysis was performed on thecalculated ratios using a two-tail paired t-test analysis(GraphPad Prism 4, GraphPad Software, Inc.), with significancedetermined as P<0.05.

DIII adsorption experiments

Purified WNV DIII protein (described above) was used toremove DIII-specific antibodies from horse serum. 50 μl ofTalon Metal Affinity Resin (Clontech) was washed with PBSand resuspended in 100 μl of PBS (No DIII samples) or PBSwith 40 μg of purified DIII protein. After overnight incubationwith or without DIII at 4 °C, beads were washed with PBS andpelleted to remove unbound DIII protein. Subsequently, a 1:100dilution of control, uninfected horse serum was added to thebeads as a blocking agent and incubated for 30 min at RT. WNVpositive or control serum was then added to the tubes to a finaldilution of 1:500. Therefore, each horse serum sample wasadded to one tube with beads without DIII, and one tube withbeads with DIII. After 1 h incubation, beads were pelleted andserum removed and filtered through a 0.2 μm filter.

Successful adsorption of DIII antibodies was determined byELISA as follows. 96-well plates were coated with 0.5 μg/ml ofDIII in coating buffer and incubated overnight. Plates were

blocked with PBS/0.05% Tween/10% FCS for 1 h. Subse-quently, horse serum before incubation with beads (pre-beadssample) or after incubation with beads (with and without DIII)was added to the plate and incubated for 2 h at RT. Anti-horseIgG–HRP conjugate was added, and plates developed asdescribed above. Neutralization assays were performed withwild-type and Δ67IA RVP GFP stocks as described aboveexcept the initial dilution for the horse serum samples used inthese experiments was 1:2000. Five five-fold serial dilutions ofhorse serum were performed in MEM in order to obtainsigmoidal curves for non-linear regression analysis as describedabove for RVP GFP experiments.

Acknowledgments

We thank Dr. Pamela Wilkins and Ms. Terry Fyock from theNew Bolton Center of the School of Veterinary Medicine at theUniversity of Pennsylvania for providing us with horse serumsamples. We also thank Ms. Hao Wang from the Center forClinical Epidemiology and Biostatistics at the University ofPennsylvania for assistance with statistical analysis and Dr.Bridget Puffer for help in obtaining the flow cytometry data forGFP-based RVP neutralization assays. We thank members of theDoms laboratory, especially Dr. Carl Davis, for useful discus-sions. This work was supported by NIH U54 457173 to R.W.D.,by a grant from the University of Pennsylvania, by NIH F31RR05074 to M.D.S. and by NIH T32 AI-07324 which supportsS.L.H.

References

Aaskov, J.G., Geysen, H.M., Mason, T.J., 1989. Serologically defined linearepitopes in the envelope protein of dengue 2 (Jamaica strain 1409). Arch.Virol. 105 (3–4), 209–221.

Agrawal, A.G., Petersen, L.R., 2003. Human immunoglobulin as a treatment forWest Nile virus infection. J. Infect. Dis. 188 (1), 1–4.

Allison, S.L., Stadler, K., Mandl, C.W., Kunz, C., Heinz, F.X., 1995. Synthesisand secretion of recombinant tick-borne encephalitis virus protein E insoluble and particulate form. J. Virol. 69 (9), 5816–5820.

Arroyo, J., Miller, C., Catalan, J., Myers, G.A., Ratterree, M.S., Trent, D.W.,Monath, T.P., 2004. ChimeriVax-West Nile virus live-attenuated vaccine:preclinical evaluation of safety, immunogenicity, and efficacy. J. Virol. 78(22), 12497–12507.

Beasley, D.W., Barrett, A.D., 2002. Identification of neutralizing epitopes withinstructural domain III of the West Nile virus envelope protein. J. Virol. 76(24), 13097–13100.

Beasley, D.W., Holbrook, M.R., Travassos Da Rosa, A.P., Coffey, L., Carrara,A.S., Phillippi-Falkenstein, K., Bohm Jr., R.P., Ratterree, M.S., Lillibridge,K.M., Ludwig, G.V., Estrada-Franco, J., Weaver, S.C., Tesh, R.B., Shope,R.E., Barrett, A.D., 2004. Use of a recombinant envelope protein subunitantigen for specific serological diagnosis ofWest Nile virus infection. J. Clin.Microbiol. 42 (6), 2759–2765.

Ben-Nathan, D., Lustig, S., Tam, G., Robinzon, S., Segal, S., Rager-Zisman, B.,2003. Prophylactic and therapeutic efficacy of human intravenous immuno-globulin in treating West Nile virus infection in mice. J. Infect. Dis. 188 (1),5–12.

Camenga, D.L., Nathanson, N., Cole, G.A., 1974. Cyclophosphamide-potentiated West Nile viral encephalitis: relative influence of cellular andhumoral factors. J. Infect. Dis. 130 (6), 634–641.

Daffis, S., Kontermann, R.E., Korimbocus, J., Zeller, H., Klenk, H.D., TerMeulen, J., 2005. Antibody responses against wild-type yellow fever virus

347M.D. Sánchez et al. / Virology 359 (2007) 336–348

and the 17D vaccine strain: characterization with human monoclonalantibody fragments and neutralization escape variants. Virology 337 (2),262–272.

Davidson, A.H., Traub-Dargatz, J.L., Rodeheaver, R.M., Ostlund, E.N.,Pedersen, D.D., Moorhead, R.G., Stricklin, J.B., Dewell, R.D., Roach, S.D.,Long, R.E., Albers, S.J., Callan, R.J., Salman, M.D., 2005. Immunologicresponses to West Nile virus in vaccinated and clinically affected horses.J. Am. Vet. Med. Assoc. 226 (2), 240–245.

Davis, B.S., Chang, G.J., Cropp, B., Roehrig, J.T., Martin, D.A., Mitchell, C.J.,Bowen, R., Bunning, M.L., 2001. West Nile virus recombinant DNAvaccine protects mouse and horse from virus challenge and expresses in vitroa noninfectious recombinant antigen that can be used in enzyme-linkedimmunosorbent assays. J. Virol. 75 (9), 4040–4047.

Diamond, M.S., Shrestha, B., Marri, A., Mahan, D., Engle, M., 2003a. B cellsand antibody play critical roles in the immediate defense of disseminatedinfection by West Nile encephalitis virus. J. Virol. 77 (4), 2578–2586.

Diamond, M.S., Sitati, E.M., Friend, L.D., Higgs, S., Shrestha, B., Engle, M.,2003b. A critical role for induced IgM in the protection against West Nilevirus infection. J. Exp. Med. 198 (12), 1853–1862.

Engle, M.J., Diamond, M.S., 2003. Antibody prophylaxis and therapy againstWest Nile virus infection in wild-type and immunodeficient mice. J. Virol.77 (24), 12941–12949.

Gollins, S.W., Porterfield, J.S., 1986. A new mechanism for the neutralization ofenveloped viruses by antiviral antibody. Nature 321 (6067), 244–246.

Goncalvez, A.P., Men, R., Wernly, C., Purcell, R.H., Lai, C.J., 2004a.Chimpanzee Fab fragments and a derived humanized immunoglobulin G1antibody that efficiently cross-neutralize dengue type 1 and type 2 viruses.J. Virol. 78 (23), 12910–12918.

Goncalvez, A.P., Purcell, R.H., Lai, C.J., 2004b. Epitope determinants of achimpanzee Fab antibody that efficiently cross-neutralizes dengue type 1and type 2 viruses map to inside and in close proximity to fusion loopof the dengue type 2 virus envelope glycoprotein. J. Virol. 78 (23),12919–12928.

Gould, L.H., Sui, J., Foellmer, H., Oliphant, T., Wang, T., Ledizet, M.,Murakami, A., Noonan, K., Lambeth, C., Kar, K., Anderson, J.F., de Silva,A.M., Diamond, M.S., Koski, R.A., Marasco, W.A., Fikrig, E., 2005.Protective and therapeutic capacity of human single-chain Fv-Fc fusionproteins against West Nile virus. J. Virol. 79 (23), 14606–14613.

Halevy, M., Akov, Y., Ben-Nathan, D., Kobiler, D., Lachmi, B., Lustig, S., 1994.Loss of active neuroinvasiveness in attenuated strains of West Nile virus:pathogenicity in immunocompetent and SCID mice. Arch. Virol. 137 (3–4),355–370.

Haley, M., Retter, A.S., Fowler, D., Gea-Banacloche, J., O'Grady, N.P., 2003.The role for intravenous immunoglobulin in the treatment of West Nile virusencephalitis. Clin. Infect. Dis. 37 (6), e88–e90.

Hamdan, A., Green, P., Mendelson, E., Kramer, M.R., Pitlik, S., Weinberger, M.,2002. Possible benefit of intravenous immunoglobulin therapy in a lungtransplant recipient with West Nile virus encephalitis. Transplant. Infect.Dis. 4 (3), 160–162.

Hanna, S.L., Pierson, T.C., Sanchez, M.D., Ahmed, A.A., Murtadha, M.M.,Doms, R.W., 2005. N-linked glycosylation of West Nile virus envelopeproteins influences particle assembly and infectivity. J. Virol. 79 (21),13262–13274.

Hunt, A.R., Cropp, C.B., Chang, G.J., 2001. A recombinant particulate antigenof Japanese encephalitis virus produced in stably-transformed cells is aneffective noninfectious antigen and subunit immunogen. J. Virol. Methods97 (1–2), 133–149.

Innis, B.L., Thirawuth, V., Hemachudha, C., 1989. Identification of continuousepitopes of the envelope glycoprotein of dengue type 2 virus. Am. J. Trop.Med. Hyg. 40 (6), 676–687.

Konishi, E., Fujii, A., 2002. Dengue type 2 virus subviral extracellular particlesproduced by a stably transfected mammalian cell line and their evaluationfor a subunit vaccine. Vaccine 20 (7–8), 1058–1067.

Konishi, E., Pincus, S., Paoletti, E., Shope, R.E., Burrage, T., Mason, P.W.,1992. Mice immunized with a subviral particle containing the Japaneseencephalitis virus prM/M and E proteins are protected from lethal JEVinfection. Virology 188 (2), 714–720.

Konishi, E., Fujii, A., Mason, P.W., 2001. Generation and characterization of a

mammalian cell line continuously expressing Japanese encephalitis virussubviral particles. J. Virol. 75 (5), 2204–2212.

Kroeger, M.A., McMinn, P.C., 2002. Murray Valley encephalitis virusrecombinant subviral particles protect mice from lethal challenge withvirulent wild-type virus. Arch. Virol. 147 (6), 1155–1172.

Kuhn, R.J., Zhang, W., Rossmann, M.G., Pletnev, S.V., Corver, J., Lenches, E.,Jones, C.T., Mukhopadhyay, S., Chipman, P.R., Strauss, E.G., Baker, T.S.,Strauss, J.H., 2002. Structure of dengue virus: implications for flavivirusorganization, maturation, and fusion. Cell 108 (5), 717–725.

Li, L., Barrett, A.D., Beasley, D.W., 2005. Differential expression of domain IIIneutralizing epitopes on the envelope proteins of West Nile virus strains.Virology 335 (1), 99–105.

Mandl, C.W., Guirakhoo, F., Holzmann, H., Heinz, F.X., Kunz, C., 1989.Antigenic structure of the flavivirus envelope protein E at themolecular level, using tick-borne encephalitis virus as a model. J.Virol. 63, 564–571.

Mason, P.W., Pincus, S., Fournier, M.J., Mason, T.L., Shope, R.E., Paoletti, E.,1991. Japanese encephalitis virus-vaccinia recombinants produce particulateforms of the structural membrane proteins and induce high levels ofprotection against lethal JEV infection. Virology 180 (1), 294–305.

Minke, J.M., Siger, L., Karaca, K., Austgen, L., Gordy, P., Bowen, R., Renshaw,R.W., Loosmore, S., Audonnet, J.C., Nordgren, B., 2004. Recombinantcanarypoxvirus vaccine carrying the prM/E genes of West Nile virus protectshorses against a West Nile virus-mosquito challenge. Arch. Virol. (Suppl.18), 221–230.

Modis, Y., Ogata, S., Clements, D., Harrison, S.C., 2003. A ligand-bindingpocket in the dengue virus envelope glycoprotein. Proc. Natl. Acad. Sci.U.S.A. 100 (12), 6986–6991.

Modis, Y., Ogata, S., Clements, D., Harrison, S.C., 2005. Variable surfaceepitopes in the crystal structure of dengue virus type 3 envelope glycoprotein.J. Virol. 79 (2), 1223–1231.

Monath, T.P., Liu, J., Kanesa-Thasan, N., Myers, G.A., Nichols, R., Deary, A.,McCarthy, K., Johnson, C., Ermak, T., Shin, S., Arroyo, J., Guirakhoo, F.,Kennedy, J.S., Ennis, F.A., Green, S., Bedford, P., 2006. A live, attenuatedrecombinant West Nile virus vaccine. Proc. Natl. Acad Sci. U.S.A. 103 (17),6694–6699.

Mukhopadhyay, S., Kim, B.S., Chipman, P.R., Rossmann, M.G., Kuhn, R.J.,2003. Structure of West Nile virus. Science 302 (5643), 248.

Murgue, B., Murri, S., Zientara, S., Durand, B., Durand, J.P., Zeller, H., 2001.West Nile outbreak in horses in southern France, 2000: the return after 35years. Emerging Infect. Dis. 7 (4), 692–696.

Nybakken, G.E., Oliphant, T., Johnson, S., Burke, S., Diamond, M.S., Fremont,D.H., 2005. Structural basis of West Nile virus neutralization by atherapeutic antibody. Nature 437 (7059), 764–769.

Oliphant, T., Engle, M., Nybakken, G.E., Doane, C., Johnson, S., Huang, L.,Gorlatov, S., Mehlhop, E., Marri, A., Chung, K.M., Ebel, G.D., Kramer,L.D., Fremont, D.H., Diamond, M.S., 2005. Development of a humanizedmonoclonal antibody with therapeutic potential against West Nile virus.Nat. Med. 11 (5), 522–530.

Ostlund, E.N., Crom, R.L., Pedersen, D.D., Johnson, D.J., Williams, W.O.,Schmitt, B.J., 2001. Equine West Nile encephalitis, United States. EmergingInfect. Dis. 7 (4), 665–669.

Peiris, J.S., Porterfield, J.S., Roehrig, J.T., 1982. Monoclonal antibodies againstthe flavivirus West Nile. J. Gen. Virol. 58 (Pt. 2), 283–289.

Pierson, T.C., Sanchez, M.D., Puffer, B.A., Ahmed, A.A., Geiss, B.J., Valentine,L.E., Altamura, L.A., Diamond, M.S., Doms, R.W., 2006. A rapid andquantitative assay for measuring antibody-mediated neutralization of WestNile virus infection. Virology 346 (1), 53–65.

Pletnev, A.G., Swayne, D.E., Speicher, J., Rumyantsev, A.A., Murphy, B.R.,2006. Chimeric West Nile/dengue virus vaccine candidate: preclinicalevaluation in mice, geese and monkeys for safety and immunogenicity.Vaccine 24 (40–41), 6392–6404.

Porter, M.B., Long, M.T., Getman, L.M., Giguere, S., MacKay, R.J., Lester,G.D., Alleman, A.R., Wamsley, H.L., Franklin, R.P., Jacks, S., Buergelt,C.D., Detrisac, C.J., 2003. West Nile virus encephalomyelitis in horses: 46cases (2001). J. Am. Vet. Med. Assoc. 222 (9), 1241–1247.

Razumov, I.A., Kazachinskaia, E.I., Ternovoi, V.A., Protopopova, E.V.,Galkina, I.V., Gromashevskii, V.L., Prilipov, A.G., Kachko, A.V., Ivanova,

348 M.D. Sánchez et al. / Virology 359 (2007) 336–348

A.V., L'Vov D, K., Loktev, V.B., 2005. Neutralizing monoclonal antibodiesagainst Russian strain of the West Nile virus. Viral Immunol. 18 (3),558–568.

Rey, F.A., Heinz, F.X., Mandl, C., Kunz, C., Harrison, S.C., 1995. The envelopeglycoprotein from tick-borne encephalitis virus at 2 A resolution. Nature 375(6529), 291–298.

Roehrig, J.T., 2003. Antigenic structure of flavivirus proteins. Adv. Virus Res.59, 141–175.

Roehrig, J.T., Bolin, R.A., Kelly, R.G., 1998. Monoclonal antibody mapping ofthe envelope glycoprotein of the dengue 2 virus, Jamaica. Virology 246 (2),317–328.

Salazar, P., Traub-Dargatz, J.L., Morley, P.S., Wilmot, D.D., Steffen, D.J.,Cunningham, W.E., Salman, M.D., 2004. Outcome of equids with clinicalsigns of West Nile virus infection and factors associated with death. J. Am.Vet. Med. Assoc. 225 (2), 267–274.

Sanchez, M.D., Pierson, T.C., McAllister, D., Hanna, S.L., Puffer, B.A.,Valentine, L.E., Murtadha, M.M., Hoxie, J.A., Doms, R.W., 2005.Characterization of neutralizing antibodies to West Nile virus. Virology336 (1), 70–82.

Schalich, J., Allison, S.L., Stiasny, K., Mandl, C.W., Kunz, C., Heinz, F.X.,1996. Recombinant subviral particles from tick-borne encephalitis virus arefusogenic and provide a model system for studying flavivirus envelopeglycoprotein functions. J. Virol. 70 (7), 4549–4557.

Schmidt, J.R., Elmansoury, H.K., 1963. Natural and experimental infection of

Egyptian equines with West Nile virus. Ann. Trop. Med. Parasitol. 57,415–427.

Shimoni, Z., Niven, M.J., Pitlick, S., Bulvik, S., 2001. Treatment of West Nilevirus encephalitis with intravenous immunoglobulin. Emerging Infect. Dis.7 (4), 759.

Siger, L., Bowen, R.A., Karaca, K., Murray, M.J., Gordy, P.W., Loosmore, S.M.,Audonnet, J.C., Nordgren, R.M., Minke, J.M., 2004. Assessment of theefficacy of a single dose of a recombinant vaccine against West Nile virus inresponse to natural challenge with West Nile virus-infected mosquitoes inhorses. Am. J. Vet. Res. 65 (11), 1459–1462.

Snook, C.S., Hyman, S.S., Del Piero, F., Palmer, J.E., Ostlund, E.N., Barr, B.S.,Desrochers, A.M., Reilly, L.K., 2001. West Nile virus encephalomyelitis ineight horses. J. Am. Vet. Med. Assoc. 218 (10), 1576–1579.

Tesh, R.B., Arroyo, J., Travassos Da Rosa, A.P., Guzman, H., Xiao, S.Y.,Monath, T.P., 2002. Efficacy of killed virus vaccine, live attenuated chimericvirus vaccine, and passive immunization for prevention of West Nile virusencephalitis in hamster model. Emerging Infect. Dis. 8 (12), 1392–1397.

Wang, T., Anderson, J.F., Magnarelli, L.A., Wong, S.J., Koski, R.A., Fikrig, E.,2001. Immunization of mice against West Nile virus with recombinantenvelope protein. J. Immunol. 167 (9), 5273–5277.

Wengler, G., 1989. An analysis of the antibody response against West Nile virusE protein purified by SDS-PAGE indicates that this protein does not containsequential epitopes for efficient induction of neutralizing antibodies. J. Gen.Virol. 70 (Pt. 4), 987–992.