Innate and adaptive cellular immunity in flavivirus-naïve human recipients of a live-attenuated dengue serotype 3 vaccine produced in Vero cells (VDV3)

Vaccine 24 (2006) 4914–4926

Innate and adaptive cellular immunity in flavivirus-naıve humanrecipients of a live-attenuated dengue serotype 3 vaccine

produced in Vero cells (VDV3)

Violette Sanchez a, Sophie Gimenez a, Brian Tomlinson b, Paul K.S. Chan b,G. Neil Thomas b, Remi Forrat a, Laurent Chambonneau a,

Florence Deauvieau a, Jean Lang a, Bruno Guy a,∗a Sanofi Pasteur, Research and Development Department, Marcy l’Etoile, France

b Drug Development Center, Hong Kong, China

Received 21 December 2005; received in revised form 17 February 2006; accepted 20 March 2006Available online 4 April 2006

Abstract

eiaoetta©

K

1

mvatlla

6T

0d

VDV3, a clonal derivative of the Mahidol live-attenuated dengue 3 vaccine was prepared in Vero cells. Despite satisfactory preclinicalvaluation, VDV3 was reactogenic in humans. We explored whether immunological mechanisms contributed to this outcome by monitoringnnate and adaptive cellular immune responses for 28 days after vaccination. While no variations were seen in serum IL12 or TNF� levels,high IFN� secretion was detected from Day 8, concomitant to IFN�, followed by IL10. Specific Th1 and CD8 responses were detected

n Day 28, with high IFN�/TNF� ratios. Vaccinees exhibited very homogeneous class I HLA profiles, and a new HLA B60-restricted CD8pitope was identified in NS3. We propose that, among other factors, adaptive immunity may have contributed to reactogenicity, even afterhis primary vaccination. In addition, the unexpected discordance observed between preclinical results and clinical outcome in humans led uso reconsider some of our preclinical acceptance criteria. Lessons learned from these results will help us to pursue the development of safend immunogenic vaccines.

2006 Elsevier Ltd. All rights reserved.

eywords: Live dengue vaccine; Cellular immunity; Clinical trial; HLA; HLA B60

. Introduction

Dengue fever is a widespread viral illness affectingore than 50 million people each year [1]. An approved

accine against dengue does not yet exist, although severalpproaches are under evaluation [2]. In collaboration withhe Mahidol University of Thailand, sanofi pasteur has aongstanding commitment to the development of a tetravalentive-attenuated dengue vaccine for the protection of childrennd adults against dengue fever and its complications

∗ Corresponding author at: Building X, Sanofi Pasteur, Campus Merieux,9280 Marcy l’Etoile, France.el.: +33 4 37 37 38 75; fax: +33 4 37 37 36 39.

E-mail address: [email protected] (B. Guy).

[3–5]. A first generation of live-attenuated vaccine (LAV)candidates was obtained through passage in primary dogkidney cells and African green monkey kidney cell cultures.After a series of clinical trials with mono- and tetravalentLAV formulations, it was concluded that the LAV serotype 3was under-attenuated and too reactogenic. The LAV strainswere therefore plaque-purified and adapted to productionin Vero cells. The Vero-adapted dengue serotype 3 vaccine(VDV3) was thoroughly characterized in preclinical assaysand cleared for phase I clinical evaluation based on widelyaccepted criteria for satisfactory attenuation compared to itsparental strain (for a review see ref. [6]): small plaque pheno-type, temperature sensitivity, decreased growth and absenceof transmission in mosquitoes, and low viremia in monkeys(mean peak viremia <100 pfu/mL, mean duration <3 days)

264-410X/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.oi:10.1016/j.vaccine.2006.03.066

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4915

(V. Barban, in preparation). In addition, innate responseswere evaluated in vitro in human monocyte-derived dendriticcells. Upon infection with VDV3, the secretion of TNF�,IL6, and IL12 pro-inflammatory cytokines was undetectableor dramatically reduced in comparison to that triggered bythe wild-type parental strain. This reduction was associatedwith a higher type I IFN production. These results suggestedthat VDV3 would induce in vivo a safe and well-controlledearly innate response.

In the first clinical trial, monovalent VDV3 was evalu-ated in comparison with a Yellow Fever (YF) control vaccinein volunteers in Hong-Kong. The first 15 VDV3 recipientshad adverse reactions including fever, malaise, headache,macular rash, decreased neutrophil, lymphocyte, WBC, andplatelet counts, and increased levels of liver enzymes. Reac-tions occurred predominantly 6–12 days after vaccination andwere severe for 13/15 subjects (R. Forrat, in preparation). Theunexpected nature, number and severity of these reactions ledto the discontinuation of subject recruitment and injections,as well as to the discontinuation of the development of theVDV3 vaccine.

To better understand these reactions occurring after aprimary dengue contact, we monitored cellular innate andadaptive immune responses for 4 weeks after vaccination.We quantified serum levels of pro- and anti-inflammatorycytokines during the period of peak reactogenicity, andai

2

2

1isatjst2wafpfisi

2

p

dengue virus serotypes 1/16007 (batch 26/04/94), 2/16681(batch 27/04/94), 3/16562 (batch 07/94), and 4/1036(batch 03/03//94); and live-attenuated vaccines serotypes 1(02LSP020), 2 (03LSP14), 3 (LSP02), and 4 (02LSP021)[3,4].

2.3. Preclinical immunological analyses

Assays based on those previously developed with DEN2I6681 and LAV2 viruses were used to compare the immuno-logical changes induced inVDV3- or DEN3-infected humanmonocyte-derived DCs [7]. Whole blood from healthy adultvolunteers was obtained from Etablissement Francais duSang (ETS, Lyon). After isolation by density-gradient cen-trifugation, monocytes were purified by positive selectionusing CD14 microbeads and magnetic cell separator (MACS)according to manufacturer’s specifications (Miltenyi Biotec,Auburn, CA). CD14+ cells were cultured in RPMI 1640medium [Gibco, Paisley, Scotland], supplemented withhuman granulocyte-macrophage colony-stimulating factor(GM-CSF) and IL-4 (PeproTech, Rocky Hill, NJ) at 37 ◦C and5% CO2 for 6 days at a density of 1 × 106 cells/mL. GM-CSFand IL-4 were added every 2 days at 50 and 10 ng/mL, respec-tively. The appropriate phenotype of immature dendritic cells(CD14−, CD1a+, HLA-DR+, CD83−) was confirmed by flowcytometry before each experiment.

tca(Cmm(peslwi

alibA(aawB1pnu

ssessed specific cell-mediated immunity 28 days aftermmunization.

. Material and methods

.1. Clinical trial

In a monocenter, randomized, blind-observer phase I trial,5 healthy, flavivirus-naıve adults received an intramuscularnjection of a low dose of VDV3 vaccine (102 50% Tis-ue Culture Infective Dose [TCID50]) on Day (D) 0. Sevendults received a YF control vaccine (Stamaril®, sanofi pas-eur, France). Informed consent was obtained from all sub-ects. The VDV3 group was constituted of eight males andeven females, and the Stamaril group of five males andwo females. Mean ages in VDV3 and Stamaril groups were2.8 ± 1.5 years and 22.6 ± 1.1 years, respectively. Subjectsere followed for safety and immune response for 28 days

fter injection. Serum was sampled every 2 days from D0–16or viremia and serum cytokine analysis. On D0 and D28,eripheral blood lymphocytes (PBL) were collected, puri-ed, frozen, and transferred in liquid nitrogen containers toanofi pasteur in France for analysis. The trial was performedn accordance with good clinical practice.

.2. Vaccine and viruses

The parental strains and the VDV strains were pre-ared at sanofi pasteur, Marcy l’Etoile, France: parental

Immature dendritic cells were then infected with wild-ype DEN3 or VDV3 at a MOI of 0.5 in RPMI 1640 mediumontaining 2% of heat-inactivated FCS (2 × 106 cells/mL)nd supplemented with GM-CSF (50 ng/mL) and IL-410 ng/mL). Cells were incubated for 48 h at 32 ◦C, 5%O2 without washing. This temperature was chosen as opti-al based on previous experiments [7]. Controls includedock-infected cells and cells infected with heat-inactivated

56 ◦C for 30 min) viruses. Cells were collected after 48 h andlated in 96-well plates for immunostaining and flow cytom-try analysis. Cell-free supernatants were also collected andtored at −80 ◦C for cytokine quantifications. For intracellu-ar cytokines analysis, Brefeldin A (Sigma, St. Louis, MO)as added at 10 �g/mL after 48 h for an additional 18 h of

ncubation.After cells were fixed with 4% formaldehyde (Aldrich)

nd permeabilized with 0.1% saponin (Sigma), intracellu-ar staining for virus and cytokines was accomplished byncubation with Alexa 488-conjugated anti-flavivirus (4G2atch HB112-011012; Biotem, Le Rivier d’Apprieu, France;

488 kit Molecular Probes, Eugene, OR) or phycoerythrinPE)-conjugated monoclonal anti-human IL-12p40/70 andnti-human TNF� mAbs (BD Biosciences) diluted in perme-bilization buffer containing human IgG (Sigma). Analysisas done with FACScan or FACScalibur flow cytometer (BDiosciences) and the CellQuest software. In all experiments,00,000 to 200,000 events were gated on the FSC-SSC dotlot, according to the infection rate, to analyze a sufficientumber of cells among infected (4G2+) and surroundingninfected cells (4G2−).

4916 V. Sanchez et al. / Vaccine 24 (2006) 4914–4926

A cytometric bead array (CBA) commercial kit fromBD Bioscience was used to monitor IL6, IL12, TNF�, andIL10 secretions, and ELISA kits from PBL BiomedicalLaboratories (Piscataway, NJ), and Fujerebio Inc. (Tokyo,Japan; IFN�) were used to monitor respectively IFN� (multi-species) and IFN� cytokine levels in cell-free supernatantsafter 48 h infection.

2.4. Clinical analyses

Four types of analysis were performed: assessment ofserum cytokine levels, PBL lymphoproliferation, Th1 andTh2 cytokine quantification in supernatants, and intracellu-lar staining (ICS) evaluation of specific IFN� and TNF�cytokine production by CD4 and CD8 cells after stimula-tion with dengue virus/vaccine or NS3 peptide libraries. Thegroup assignment of each subject was masked for the primaryanalyses. Secondary analyses (ICS studies with individualHLA-restricted peptides) were performed after unmasking.

2.4.1. Serum cytokine levelLevels of serum pro/anti-inflammatory cytokines (IL6,

IL8, IL10, IL12p70, TNF�, IL1�) were quantified on aFACSCaliburTM (BD) using a cytometric bead array (CBA)kit and appropriate controls (BD Biosciences), according tothe manufacturer’s recommendations. Serum TGF�, IFN�aNi

22dft(w

2sudwwilrepstVrpli

D0 and D28 samples were analyzed on the sameday for each donor. Analyses were done in triplicate(2 × 106 cells/mL in RPMI + GSP + 5% FCS). Viral anti-gens were diluted in the same medium and used at a con-centration of 0.2 MOI. For inactivation, viral suspensionswere incubated 30 min at 56 ◦C. PHA was used as posi-tive control, while negative control was medium. Cultureplates were incubated for 4 days at 37 ◦C and 5% CO2.After 4 days, supernatants were collected from each well(≈100 �L) to measure the secretion of cytokines later on.Triplicates were pooled before freezing at −80 ◦C. Tritiatedthymidin (1 �Ci/mL) was added to each well and plates wereincubated for 16 more hours. The emitted �-radioactivitywas quantified with a Beta-counter (Top-Count-NXTTM-Packard). Index of stimulation (IS) was calculated for eachmean of triplicates using the following formula: IS = meanof assay triplicates CPM/mean of negative control triplicatesCPM.

2.4.2.3. Lymphoproliferation cytokine levels. Levels of Th1(IL2, IFN�, and TNF�) and Th2 (IL6, IL4 and IL10)cytokines were quantified as described above using a CBATh1/Th2 kit (BD Biosciences).

2.4.2.4. IFNγ and TNFα production by CD4 and CD8 cells

nd IFN� levels were assayed using ELISA kits (Piscataway,J; and Pharmingen CA), as these cytokines were not present

n the CBA kit.

.4.2. Adaptive responses in human PBMCs

.4.2.1. Cell culture. After thawing, cells collected asecribed in Section 2.1 were counted, numbered, checkedor viability by Trypan Blue exclusion and their concentra-ion was adjusted to 2 × 106 cells/mL in RPMI 1640 + GSPGibco) + 5% FCS (Hyclone). No significant cell mortalityas observed after thawing of frozen PBLs.

.4.2.2. Lymphoproliferation. Optimal conditions for celltimulation were determined in preliminary experimentssing PBMCs from dengue-positive donors. A stimulationuration of 4 days was determined, for which stimulationas maximal and background levels in mock-treated cellsere acceptable. The stimulation MOI was also determined

n the same assays, and ranged between 0.2 (lymphopro-iferation) and 0.5 (ICS assay). Purified virus was chosenather than cell lysates to stimulate cells, as it is a moreasily characterized and standardized reagent. In addition,urified inactivated viruses stimulate mostly CD4 cells, in aerotype-specific way, as they contain mostly structural pro-eins that are less cross-reactive than non-structural proteins.iruses and vaccines were used as monovalent purified prepa-

ations as follows: (i) Di1 to Di4, corresponding to inactivatedarental viruses serotypes 1–4; (ii) V1–V4, corresponding toive vaccines serotypes 1–4; (iii) Vi1–Vi4, corresponding tonactivated vaccines serotypes 1–4.

by ICS. The production of these cytokines in particular wasassessed as INF�/TNF� ratios in favour of IFN� have beenassociated with a favorable evolution of dengue infections[8–10]. Intracellular staining (ICS) was performed followinga technique similar to the one described by Mangada et al.[11] and is briefly described in the following paragraph. Cellswere adjusted to 4 × 106 cells/mL in AIM-V medium. TheD0 and D28 samples were analyzed on the same day andstimulated by whole viruses or peptides.

2.4.2.5. NS3 peptides stimulation. One hundred and eigh-teen peptides consisting of 15 amino acid residues, that over-lapped each other by 11 amino acids (aa), were designedto span the most cross-reactive DEN3 NS3 sequence(access number 74487, ref. [12]) (region corresponding toaa 189–619, and three additional overlapping peptides inthe N terminus region). Peptides were manufactured inNeoMPS (Strasbourg), solubilized in DMSO and adjustedto 30 mg/mL. Peptides were then pooled into six pools of∼18 successive peptides: pools B (covering aa 62–85, and189–273), C (aa 274–345), D (aa 346–406), E (aa 407–478),F (aa 479–546) and G (aa 547–619). The final concentrationof each peptide in the different stock mixes was 2 mg/mL inRPMI1640 + 10%FCS + GSP. Each mix was aliquoted andstored at −80 ◦C until use. Additional potential HLA B60-,A11-, A2- and 24-restricted peptides were used: CTRL-F1(plnkdedha); CTRL-F2 (epereksaai); A11-1 (gtsgspiinre);B60-F-1 (gesrktfvel); B-60-F2 (velmrrgdl); A2-F-1 (kmlld-nint); A24-G-1 (kytdrkwcf); A2-G-2 (kvaseqiky); A2-G-3(qileenmdv).

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4917

2.4.2.6. Cell stimulation. Individual peptides or peptidepools were added to cells at a final concentration of 2 �g/mLfor each peptide. Cells were incubated at 37 ◦C, 5% CO2for 1 h. Brefeldin A secretion inhibitor was then addedfor four more hours. Cells were kept at +4 ◦C overnightuntil staining. Staphylococcus Enterotoxin B (SEB; Sigma)was used as a positive control. Cells were fixed (4%formaldehyde) and permeabilized (0.1% saponin). Anti-CD3 and anti-CD8 (antigens of surface); anti-IFN� andanti-TNF� (intracellular antigens) antibodies were used forstaining.

The four-color flow cytometric analysis was performed ona FACSCalibur flow cytometer (Beckton Dickinson). Datawere acquired using the Cell-questPro Software (BD). A R1gate was done around lymphocytes on FSC/SSC dot plot. AG2 (R1 × R2) and a G3 gate (R1 × R3) were done aroundCD8+ and CD8− lymphocytes, respectively, on CD3/CD8dot plots. Acquisition of events was done on R1 with a maxi-mum of 50,000 events acquired in G2. The number of IFN�+or TNF�+ cells per million T cells was determined for thefollowing cell populations: CD3+ CD8+ (analysis on G2);CD3+ CD8− (CD4+) (analysis on G3); according to the for-mulas:

IFNγ + CD4 + T cells

1 × 106 T cells= {%[IFNγ + CD8−]cells)

2itiu(nd5ia

It is important to note that cells were not stimulated withYF-specific reagents. Cells from the YF control group weretherefore viewed as negative controls.

2.4.2.8. HLA typing. HLA typing (ABC) was performedusing an aliquot of frozen PBL (for three subjects no aliquotwas available). Analyses were performed by Cellis Pharma(St Malo, France) using a Sequence-Specific-Primer-PCRmethod.

2.5. Statistical analyses

Analyses were mostly descriptive. The severity of reac-tions was summarized per subject using a symptom index thattakes into account the severity and duration of each systemicreaction, as previously described [3]. Briefly, the durationof each systemic reaction (fever, chills, malaise, headache,myalgia, arthralgia, rash, pruritis, eye pain, photophobia, orconjunctivitis) was multiplied by its severity score. Then foreach subject, the maximum value between fever versus chills,myalgia versus arthralgia, rash versus pruritus, and eye painversus photophobia versus conjunctivitis were retained. TheSI is equal to the sum of these maximum values plus thosefor malaise and headache.

Area under curve (AUC) and peaks were calculated ineach group to evaluate cytokines in serum and were com-pltfres

3

3Dm

isaDsd(pittlnp

× (%[CD8 − CD3+]cells)} × 100

TNFα + CD4 + T cells

1 × 106 T cells= {%[TNFα + CD8−]cells)

× (%[CD8 − CD3+]cells)} × 100

IFNγ + CD8 + T cells

1 × 106 T cells= {%[IFNγ + CD8+]cells)

× (%[CD8 + CD3+]cells)} × 100

TNFα + CD8 + T cells

1 × 106 T cells= {%[TNFα + CD8+]cells)

× (%[CD8 + CD3+]cells)} × 100

.4.2.7. Viral stimulation. PBLs (4 × 106/mL) werenfected by viral suspensions used at a final viral concen-ration of 0.5 MOI or corresponding concentrations fornactivated viruses/vaccines. Viruses and vaccines weresed as monovalent purified preparations as defined aboveSection 2.4.2.2). DS22 vaccine stabilizer was used as aegative control, at the same concentration than in vaccineilutions. SEB (Sigma) was used as a positive control. Afterh of incubation, Brefeldin A was added and cells were

ncubated overnight, then kept on ice until staining andnalysis as described above.

ared using Wilcoxon test. For quantification of cytokines inymphoproliferation supernatants, differences with the nega-ive controls 28 days after injection were plotted versus dif-erences observed before injection. Values <20 pg/mL wereeplaced by 10 and negative differences were represented asqual to 0. Statistical analyses were performed using SASoftware V8.2.

. Results

.1. Preclinical comparison of VDV3-and parentalEN3-induced innate responses in humanonocyte-derived dendritic cells

Human dendritic cells were first used to compare thennate response induced by VDV3 and its parental DEN3train. Infection levels as measured after 48 h by ICS with 4G2nti-flavivirus antibody were similar for VDV3 or parentalEN3 (5–10% of positive cells according to the donor; not

hown). Under these conditions, VDV3 induced in indepen-ent experiments no detectable IL12 secretion as seen by ICSFig. 1a, six donors) and CBA (Fig. 1b, five donors), whilearental DEN3 consistently induced such expression bynfected or bystander cells. TNF� production was restrictedo the minority of infected cells after VDV3 exposure, in con-rast to parental DEN3 exposure, and differences in cytokineevels were also confirmed by CBA (Fig. 1b). In addition,o/low expression of IL6 or of the potentially immunosup-ressive IL10 cytokine was detected (Fig. 1b). In contrast, the

4918 V. Sanchez et al. / Vaccine 24 (2006) 4914–4926

Fig. 1. (a) IL12 (left panel) and TNF� (right panel) production in human mDCs after infection with VDV3 or parental DEN3 virus, as measured by ICS.Values correspond to the % of infected or bystander cells positive for TNF or IL12 expression. Bystander cells correspond to cells negative for dengue antigenexpression after 4G2 staining. (b) IL12, TNF�, IL6 and IL10 secretion in mDCs supernatants after infection with VDV3 or parental DEN3 virus as measuredby CBA. (c) IFN� secretion in mDCs supernatants after infection with VDV3 or parental DEN3 virus as measured by ELISA.

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4919

secretion of Type I interferons (as shown for IFN�, Fig. 1c)was greater upon VDV3 exposure. IFN� secretion presenteda similar profile (not shown). Similar analyses were alsoperformed directly ex-vivo on total PBMCs from dengue-negative donors, and double CD56-IFN� staining showedno cytokine production by NK cells after VDV3 infection(not shown). Compared to its parental virus, VDV3 wasthus shown to induce an absent/restricted pro-inflammatoryresponse, and it was assumed that its replication would be effi-ciently controlled by Type I interferons. This innate immuneprofile, together with the phenotypic and genotypic charac-teristics of VDV3, was considered to support the hypothesisthat VDV3 would be safe in humans, which led to its clinicalevaluation.

3.2. Clinical safety results summary

After vaccination, all 15 VDV3 recipients experiencedfever (oral temperature ≥37.5 ◦C), malaise, headache, andmacular rash. These were severe in 13/15 subjects. Mostsystemic reactions occurred between D6 and D12; a fewoccurred on D0 and D1. Clinical symptoms were associ-ated with biological abnormalities: decreased levels of neu-trophils, lymphocytes, WBC, and platelets, and increased lev-els of liver enzymes. The symptom index (min–max: 27–175)of each subject is summarized in Table 1. All subjects fullyrl

3

D(listI

tpo

sm

3

wwrc

m

tive responses were serotype 3-specific (inactivated parentalserotype 3 virus or live/inactivated VDV3; Fig. 3a andTable 1). Under these conditions, six VDV3 subjects hadan IS > 3 and a further four had a very low increase atD28 (IS > 1.5). Upon live VDV3 stimulation, D28/D0 ISratios were 3.13 ± 2.9 in the VDV3 group, compared to1.26 ± 0.42 in the YF group. After inactivated VDV3 stim-ulation, D28/D0 ratios were 2.93 ± 2.05 in the VDV3 groupcompared to 1.1 ± 0.23 in the YF group.

3.5. Cytokines in lymphoproliferation supernatants

None of the stimuli induced detectable changes in Th2(IL4, IL6 and IL10) cytokine levels between D0 and D28(data not shown).

Th1 responses were mostly serotype 3-specific and,with one exception, no specific stimulation was inducedafter serotypes 1, 2 and 4 stimulations (Table 1). Stimula-tion with live or inactivated VDV3 induced higher IFN�concentrations at D28 than at D0 in samples from most(11/15) VDV3 recipients (Table 1). In the YF group,IFN� concentrations did not differ between D0 and D28(except slightly in two cases). In VDV3 vaccinees afterlive VDV3 stimulation, the GMT (95% CI) of IFN�levels increased from 840 pg/mL (450–1540) on D0 to2816 pg/mL (1679–4722) on D28. Similar increases inIG1bo(Dtt(

aoVtcvcDi(

(swVts

t

ecovered from these dengue-like syndromes, without seque-ae.

.3. Early serum cytokine responses

In VDV3 recipients, IFN� and IFN� levels increased at8 and peaked at D10 following the viremia peak at D8

Fig. 2a and b; Table 1). IL10 levels rose later on D10. To aesser extent, IL6 levels also increased at the time of viremian half of the recipients (Fig. 2b). At D16, IFN� levels weretill higher than D0 baseline levels. Compared to YF con-rols, these changes (AUC) were significant: IFN� p = 0.0031;FN� p = 0.0052; IL6 p = 0.0214; IL10 p = 0.0003.

In YF controls, only IFN� showed changes comparedo baseline. Increases were lower than in VDV3 recipients,eaked earlier (D6 versus D10), and were seen in only fourf the seven subjects.

Serum levels of IL1, IL8, IL12, TGF� or TNF� remainedtable in all VDV3 or YF vaccine subjects. IL18 was noteasured in vaccinees.

.4. Lymphoproliferation

Regarding all adaptive responses described in the presentork, it is important to note that no YF-specific reagentsere used to stimulate cells (17D vaccine or peptides). In this

espect, cells coming from the Stamaril group of volunteersan be mostly viewed as negative controls.

In VDV3 recipients, lymphoproliferation responses wereoderate to low/absent, with IS values below 10. Posi-

FN� levels were seen after inactivated VDV3 stimulation:MTs increased from 651 pg/mL (299–1420) on D0, to702 pg/mL (994–2917) on D28. The GMT remained sta-le in the YF group after both live (775, 400–1504 pg/mLn D0; 738, 371–1468 pg/mL on D28) and inactivated605, 364–1214 pg/mL on D0; 781, 391–1558 pg/mL on28; Fig. 3b) VDV stimulation, with values equivalent

o those of control cells. Positive IFN� responses tendedo correlate with positive lymphoproliferation responsesTable 1).

Variations in TNF� levels were lower than those for IFN�,nd changes in Den-3 specific TNF� levels were detectedn Day 28 for only three subjects (Table 1). After liveDV3 stimulation, D0 and D28 TNF� GMTs were, respec-

ively, 15 pg/mL (9–25) and 12 pg/mL (7–20) in VDV3 vac-inees, and 17 pg/mL (10–75) and 27 pg/mL (10–75) in YFaccinees, and were similar to the values in mock-treatedells. Similarly, after inactivated VDV3 stimulation, D0 and28 GMTs were 18 pg/mL (12–28) and 14 pg/mL (8–23)

n VDV3 recipients, and 31 pg/mL (11–87) and 30 pg/mL10–80) in YF recipients.

The IFN�/TNF� ratios therefore increased from 5830–111) on D0 to 236 (130–429) on D28, after live VDV3timulation of samples from VDV3 vaccinees. These ratiosere 45 (12–165) on D0, and 28 (12–62) on D28 after liveDV3 stimulation of samples from YF vaccinees. Similar

rends were observed after inactivated VDV3 stimulation (nothown).

No specific variations were observed for IL2 whicheverhe stimulus.

4920V.Sanchez

etal./Vaccine24

(2006)4914–4926

Table 1Lymphoproliferation, serum and cellular cytokines, CD4/CD8 responses, and HLA type in VDV3 and Yellow Fever vaccinees

Subject HLA Safetya Serum cytokinesb Lymphoproliferationc Secreted cytokinesd ICSe peptide pools ICS individual peptides

A B C IFN� IFN� IL10 IFN� TNF� CD8 CD4 CD8

IFN� TNF� IFN� TNF� IFN� TNF�

VDV3 vaccine group1 A2/A11 B22/B60 Cw1/Cw7 + + +++ + + V3 + V3i + V3 + V3i − ++ F + F − − ND ND2 A24/A11 B75/B60 Cw9/Cw8 + ++ + ++ + D3i + V3 + V3i ++ D3i + V3 + V3i + D3i + V3 − + B, F − − ND ND3 A33/A11 B46/B58 Cw9/Cw10 ++ + +++ + + V3 + V3i ++ D3i + V3 + V3i − − − − − + A11 −4 A2/A203 B60/B60 Cw4/Cw15 + ++ +++ + − − ++ D3i + V3i − ++ F + F − − ++ B60-F1 ++ B60-F17 A2/A11 B75/B51 Cw8/Cw14 ++ ++ +++ ++ − − − − − − − − + A11 −8 A26/A11 B13/B58 Cw10/Cw10 ++ ++ ++ + ± V3 − − − − − − − −9 A2/A24 B46/B27 Cw1/Cw15 +++ + +++ + ± D3i + V3 + V3i ++ D3i + V3 + V3i − ++ G + G − − − −

10 A2/A11 B46/B60 Cw1/Cw7 ++ ++ +++ + + D3i + V3 + V3i ++ D3i + V3 + V3i + D3i + V3 + V3i +++ F ++ F − − ++ B60-F1 ++ B60-F113 A2/A30 B46/B13 Cw1/Cw6 + − +++ + + V3 + V3i ++ V3 + V3i − − − − − − −14 A2/A31 B46/B54 Cw1/Cw1 ++ − ++ + ± D3 + V3 + V3i ++ V3 + V3i − + G + G − −17 NDf ND ND ++ + + + − − − +++ F ++ F − − ND ND18 A2/A11 B46/B60 Cw1/Cw7 +++ + ++ + − + V3 + V3i + V1i − ++ F + F − − ++ B60-F1 ++ B60-F119 A2/A30 B46/B13 Cw1/Cw6 +++ + ++ + ± V3 + V3i + V4i ++ V3 + V3i + V4i − − − − −20 A11/A11 B62/B55 Cw1/Cw4 +++ + ++ +++ + D3i + V3 + V3i + D3i + V3 + V3i + V3 ++ B, +G ++ B − − − −21 A31/A11 B13/B61 Cw10/Cw8 ++ ++ +++ + − − − − − − − − − −

Yellow Fever vaccine group5 A2/A26 B51/B60 Cw15/Cw16 − − − − − + V1i − − − − − − −6 A2/A1 B46/B37 Cw1/Cw6 − − − − − − − − − − − − −

11 ND ND ND − − ++ − − + V3i − − − − −12 A2/A11 B46/B60 Cw1/Cw7 − − ++ − − − − − − − − −15 ND ND ND − − ++ − − − − − − − − ND ND16 A33/A11 B75/B58 Cw10/Cw8 ± − ++ − − − − − − − − − −22 A2/A24 B46/B71 Cw1/Cw7 − − ++ ± V1i to V4i − − − − − − −

a Safety symptom index: 0–5: −; 5–10: +/−; 10–40: +; 40–80: ++; >80: +++. This index was calculated as defined in Section 2 [3], based on the duration multiplied by the severity score of each systemicreaction.

b Serum cytokines (baseline <10): 10–100 pg/mL: +; 100–1000 pg/mL: ++; >1000: +++.c Lymphoproliferation stimulation index: 1.5–3: ±; 3–10: +. Antigens triggering positive responses are indicated in the corresponding columns: V, live vaccines; Vi, inactivated vaccines; Di, inactivated viruses;

serotypes 1–4.d Secreted cytokines in supernatants: IFN� levels (medium alone <100); 200–1000 pg/mL: +; 1000–10000: ++. TNF� levels (medium alone <10); 20–100 pg/mL: +.e Intracellular staining: (medium <500 and peptide/medium signal ratio ≥2): 500–2500 cells/million T cells: +; 2500–10000: ++; >10000: +++. Pools or peptides triggering positive responses are indicated in

the corresponding columns.f ND: not done.

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4921

Fig. 2. (a) Serum cytokine levels in Stamaril or VDV3 vaccinees from Day 0–16 post vaccination. Individual values are represented. Curves represent geometricmeans in Yellow Fever (triangles) and VDV3 (squares) groups. Minimal values correspond to 5 pg/mL (half of the lower limit of detection). (b) Kinetics ofserum cytokine responses and viremia in VDV3 recipients. Serum cytokine levels (geometric means; pg/mL) and viremia (geometric mean; pfu/mL). Viremiawas measured by plaque assay.

3.6. Quantification of IFNγ+ and TNFα+ CD4 andCD8 cells

With the exception of one VDV3 subject, in whomIFN� levels were higher than TNF� levels, vaccine/virusstimulation failed to induce detectable CD4 or CD8responses (not shown). However, some peptide poolsfrom the immunodominant NS3 antigen induced specific

CD8 responses (IFN� > TNF�; Table 1) in 9/15 VDV3-recipients, and these peptides were mainly those locatedin the C terminus. Some subjects responded strongly;e.g., Subject 10 (Fig. 4a), in whom more than 1% ofCD8 cells were positive for one pool (pool F). No CD4responses were detected using these NS3 pools. Nei-ther IFN� nor TNF� were detected with YF recipients’cells.

4922 V. Sanchez et al. / Vaccine 24 (2006) 4914–4926

Fig. 3. (a) Lymphoproliferation assay. Stimulation indexes (SI, scatter plotand median) after PBMCs stimulation with live VDV3. Day 0 and Day 28cells were stimulated for 4 days with live VDV3 before incorporation of3H thymidine. SI were calculated for VDV3 and Stamaril recipients. (b)IFN� (upper panel) and TNF � (lower panel) levels in cellular supernatants(pg/mL, scatter plot and median) after stimulation with live VDV3. Day 4supernatants were collected and assessed for IFN� and TNF� content byCBA.

3.7. HLA typing and HLA-binding peptides

HLA typing revealed the homogeneity of the study popula-tion, with dominant haplotypes (Table 1). In the 14/15 VDV3recipients for whom HLA-typing was performed, the follow-ing were most frequent: A11 (64%), A2 (57%), B46 (42%),B60 (35%), A2/B46 (35%), A2/B60 (29%), A11/B60 (29%)and A11/B46 (21%). A11 and B46 haplotypes are common in

Asian populations, but all typed VDV3 recipients were eitherA11 and/or A2, and, all but one (93%) was either A11 and/orA2/B46. A2/B46/Cw1 represented 47% of the VDV3 or YFrecipients.

All subjects responding to pool F were B60 (one homozy-gous) and those responding to pool G were either A2 orA24. No responses were detected in B46-positive vaccineesin absence of A2 or B60 HLAs. The National Institutesof Health’s BioInformatics and Molecular Analysis Section(BIMAS) WWW-accessible analyses were used to identifypotential 9- or 10-mer B60-binding peptides in pool F withvery high scores (two overlapping peptides, gesrktfvel andvelmrrgdl, scores of 640,000 and 320,000, respectively), anda A24-binding peptide in pool G (kytdrkwcf, score 240,000).Potential A2-binding peptides were also identified in poolsF and G, but with lower scores (kmlldnint, kvaseqikyt andqileenmdv; scores of 69,000, 28,000 and 147,000, respec-tively). An A11-binding epitope has previously been iden-tified in dengue NS3 and the level of reactivity against thispeptide was associated with disease severity [13]. This pep-tide, located in the N terminus part of NS3, was not in ourinitial peptide pools. Therefore, in a second set of assays,this A11 NS3 serotype 3 epitope (gtsgspiinre; 11 mer, almostidentical for serotypes 1 and 3), together with the proposedB60, A24 and A2-binding peptides, as well as two controlpeptides predicted to not bind to these HLAs, were synthe-sI

I1Isas

4

attvwacmtlciatat

ized (NeosMPS) and further tested as individual peptides byCS.

The predicted B60 epitope (gesrktfvel) induced strongFN� responses in all tested HLA B60 vaccinees (subjects 4,0 and 18; up to almost 1% of CD8 cells), (Fig. 4b, Table 1).n contrast, the A11 epitope (gtsgspiinre) induced modest butpecific IFN� production in only 2 HLA-A11 vaccinees (3nd 7). The predicted A24- and A2-binding peptides did nottimulate CD8 cells.

. Discussion

After vaccination with VDV3, all 15 subjects experienceddverse reactions, despite the favorable pre-clinical evalua-ion that showed that VDV3 was clearly attenuated comparedo its parental strain, including a moderate and controlled initro induction of innate immunity, as seen in a previous studyith a LAV2 vaccine candidate [7]. As previously proposed

fter the human testing of a reactogenic Dengue 1 45AZ5andidate, the relevance of some of the preclinical criteriaay need to be re-evaluated [14]. These criteria may not be

he same for each considered vaccine technology; in particu-ar, for this type of live dengue vaccine passaged on primateells, even low (below 100 pfu/mL) and short-lived viremian monkeys (<3 days mean duration) may be unacceptable,nd instead no/minimal viremia should be observed. Addi-ionally, a single cycle of infection of DCs in vitro may notdequately predict the final level of replication in vivo, andhus minimal re-infection of DCs should be observed in this

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4923

Fig. 4. (a) Intracellular staining assay: CD8 and CD4 IFN� and TNF� positive cells after NS3 peptide pools stimulation. Example of Subject 10 (VDV3recipient). Cells were stimulated with peptide pools and analyzed by flow cytometry as described in Section 2. M: control medium. +: SEB positive control. Bto G: NS3 peptide pools. Grey bars: Day 0; black bars: Day 28. (b) Intracellular staining assay: Example in Subject 10 (A2/A11-B60/B46-Cw1-Cw7) of CD8IFN� and TNF� positive cells after stimulation with potential individual B60-, A11-, A2- or A24-binding peptides.). AIMV: control medium. SEB: positivecontrol. CTRL1 and CTRL2: control peptides (non-significant scores after BIMAS prediction). Pool F: NS3 pool F.

model. It is also possible that the VDV3 parental strain is aparticularly aggressive serotype or a particular biotype [6].In this case, the vaccine derivative may have retained somevirulence, despite its marked attenuation. Thus, current pre-clinical endpoints filter out overly-virulent candidates, butmay inadequately predict the subsequent clinical safety ofretained candidates. Further preclinical analyses and a re-definition of the acceptable thresholds in the various assaysmay be needed for such second-generation vaccines.

Another hypothesis to explain the discordance betweenpreclinical and clinical results was that the virus that repli-

cated in vaccinees may have evolved and recovered somenon-attenuated characteristics. This reversion was refuted byphenotypic and genotypic assays performed on the VDV3vaccine and the subjects’ sera (V. Barban, in preparation).Although some heterogeneity in the VDV3 bulk and masterlots was found with a subpopulation of virus presenting a lesstemperature-sensitive phenotype, no change occurred in vivoand no specific mutants with wild-type like, non-attenuatedcharacteristics were identified.

We therefore investigated the vaccine-induced immuneresponses possibly contributing to the clinical outcomes. As

4924 V. Sanchez et al. / Vaccine 24 (2006) 4914–4926

observed with other live-attenuated dengue vaccines [15],volunteers developed 28 days after vaccination a serotype3-specific Th1 response (stimulated similarly by live or inac-tivated purified viruses) associated with NS3-specific CD8responses and a high IFN�/TNF� ratio. We did not observeserotype cross-reactive responses upon stimulation with puri-fied monovalent live/inactivated vaccines/viruses. It is pro-posed that under these conditions, Th responses would havebeen triggered mostly against structural proteins (presentinga low cross-reactivity), as it would have been the case in vivoafter VDV3 vaccination. In addition, the rather low overallTh responses at D28 after vaccine/virus stimulation, togetherwith the absence of IL2 secretion, suggests some persistenthyporesponsivness in some donors as observed after acutedengue [16].

Regarding anti-NS3 responses, it is well known that thisantigen is immunodominant for CD8 T cell responses [9], butthe majority of epitopes identified so far were in the first halfof the protein. In our study, reactive pools (F and G) were inthe C terminus, which may be linked to the particular HLAsof the recipients. In this respect, experiments identified a verypotent B60-specific epitope (gesrktfvel), almost identical toa known HLA.B07-restricted peptide (gesrktfve [17]).

Our analyses of early and late immune responses afterthis primary infection suggest that adaptive immunity mayhave contributed to vaccine reactogenicity. Firstly, althoughcashatIIsttagvlcpswsfitsecc

ntG

with an A24-restricted epitope recently identified in DEN2NS3 (inyadrrwcf [32]), but our assays with the correspond-ing DEN3 peptide were negative, nevertheless in agreementwith the results of the same authors with this serotype [32].Unexpectedly, only two A11 donors reacted to a previouslyidentified NS3 epitope [13]. It has been shown previously thatA11-reactive T cells were undergoing massive apoptosis dur-ing acute phase and that subsequently, their number rapidlydeclined [13]. This may explain why no/low responses wereinduced by this A11-peptide on Day 28 in our study.

It has been proposed that haplotype and peptide recogni-tion might be associated with disease severity [13,17,33–35].In our study, haplotypes were homogeneous as 100% oftyped VDV3-recipients were either A11 and/or A2 (93% A11and/or A2/B46). Although the limited number of subjectsprevents us from making conclusions, we noticed that the 10subjects with the severest reactions (++ or +++; subjects 3,7–10, 14 and 18–21) were A11-homozygous, A11, and/orA2/B46, while only two were B60. Whereas, three out of thefour subjects with the mildest reactions (+; subjects 1, 2, 4 and18) were B60, and only one was A2/A46 positive; the lowestsafety index (27) was seen in a B60-homozygous patient, andthis haplotype could be associated with lower reactogenicity.All in all, the similarity of the clinical reactions of each sub-ject (more uniform than that usually observed after wild typedengue infection) may therefore be partly explained by thehpIhsa

tvmsTipttvtibthiT

t[VhV

linical symptoms were similar to dengue fever caused bywild type virus, the observed cytokine profile was incon-

istent with a typical primary infection, in which one wouldave in particular expected more dramatic changes in IL12nd TNF� levels. The cytokine profile observed in sera at theime of viremia and symptoms was more “adaptive” (IFN�,L10, IL6) than strictly “innate” (IL8, IL12, TNF�), and theFN�/IFN�/IL10 kinetics were similar to that observed inecondary infection [9,18–23]. IFN� secretion in the presentrial could thus have been caused and amplified, at least par-ially, by specific T cells, the early activation of which haslready been demonstrated after administration of a reacto-enic vaccine [24]. This particular cytokine profile in VDV3accinees suggests that, while the very high viremia is theikely primary cause of cellular activation and clinical out-omes, it is possible that innate NK cells were not the only orredominant source of IFN�. For instance, mouse data havehowed that after primary DEN2 infection, IFN� secretionas T cell-dependant and not associated to early NK cell

timulation [25]. In our study, this would have been ampli-ed after viremia, due to highly efficient antigen presentation

o T cells [26]. Subsequent IL10 secretion would have repre-ented a negative feedback, and IFN� production would haveventually controlled viremia [27–30]. IL10 may also haveontributed to the severity of symptoms, including thrombo-ytopenia [31].

Concerning NS3 peptides, although we did identify aew B-60 epitope, we did not confirm our in silico predic-ion of A2- or A24-specific epitopes. The predicted A24-1 epitope tested here (kytdrkwcf) corresponded partially

omogeneity of the class I HLA profiles among the studyopulation (class II influence has not been addressed here).t is also possible that other, non-HLA, genetic factors mayave contributed to the uniform clinical picture, as a recenttudy showed an unexpectedly high symptomatic outcomefter primary infection in Chinese workers in Singapore [36].

Finally, another factor possibly contributing to high reac-ogenicity was the low viral dose in the vaccine, enabling theirus to initially replicate without triggering innate immuneechanisms. Monkey experiments carried out after the trial

upport this. When VDV3 was injected at either 2 or 4 log10CID50, viremia was higher and longer in those monkeys

njected with 2 log10 TCID5 (not shown), in agreement withrevious findings [37–40]. Similarly, the low-dose adminis-ration of a DEN4 vaccine candidate was recently reportedo trigger more frequent rashes of longer duration and a lateriremia than higher doses [41]. The YF vaccine, adminis-ered at a dose 65–100 times higher than VDV3, may havenduced a more efficient initial innate response, as indicatedy the earlier IFN� peak in YF-vaccinees. It is also possiblehat the high proportion of defective/subviral material that weave observed to be released by infected cells after VDV3nfection (not shown) could bias the initial innate response.his is under investigation.

In conclusion, in this context of primary vaccination wherehe potential role of pre-existing antibodies can be excluded42], it is possible that the severity of clinical outcomes inDV3 recipients was the result of a succession of viral andost factors including the characteristics and infectivity ofDV3, but also the low dose administered, and the particular

V. Sanchez et al. / Vaccine 24 (2006) 4914–4926 4925

and uniform HLA profile of the recipients. Lessons learnedfrom this trial will help us in our continued efforts to developa safe and efficient tetravalent dengue vaccine.

Acknowledgements

The authors want to acknowledge the critical help of G.Marsh in the preparation of this manuscript, Evelyn Chauand Winnie Yeung for their help in the clinical study, N. Bur-din and E. Trannoy for their support and helpful discussions,P. Chaux for his contribution in identifying HLA-restrictedepitopes, V. Barban, C. Fournier, P. Riou and D. Crevat forproviding help and reagents, and Dr. S Yoksan and ProfesssorA. Sabchareon for their constant support, and all the volun-teers for agreeing to take part.

References

[1] Mairuhu AT, Wagenaar J, Brandjes DP, van Gorp EC. Dengue: anarthropod-borne disease of global importance. Eur J Clin MicrobiolInfect Dis 2004;23:425–33.

[2] Hombach J, Barrett AD, Cardosa MJ, Deubel V, Guzman M, KuraneI, et al. Review on flavivirus vaccine development. Proceedings of ameeting jointly organised by the World Health Organization and theThai Ministry of Public Health. Vaccine 2005;23:2689–95.

[3] Sabchareon A, Lang J, Chanthavanich P, et al. Safety and immuno-

[

[

[

[

[14] McKee Jr KT, Bancroft WH, Eckels KH, Redfield RR, SummersPL, Russell PK. Lack of attenuation of a candidate dengue 1 vaccine(45AZ5) in human volunteers. Am J Trop Med Hyg 1987;36:435–42.

[15] Gwinn W, Sun W, Innis BL, Caudill J, King AD. Serotype-specificT(H)1 responses in recipients of two doses of candidate live-attenuated dengue virus vaccines. Am J Trop Med Hyg 2003;69(6Suppl.):39–47.

[16] Mathew A, Kurane I, Green S, et al. Impaired T cell proliferationin acute dengue infection. J Immunol 1999;162:5609–15.

[17] Zivna I, Green S, Vaughn DW, et al. T cell responses to an HLA-B*07-restricted epitope on the dengue NS3 protein correlate withdisease severity. Immunology 2002;168:5959–65.

[18] Chaturvedi UC, Agarwal R, Elbishbishi EA, Mustafa AS. Cytokinecascade in dengue hemorrhagic fever: implications for pathogenesis.FEMS Immunol Med Microbiol 2000;28:183–8.

[19] Pacsa AS, Agarwal R, Elbishbishi EA, Chaturvedi UC, Nagar R,Mustafa AS. Role of interleukin-12 in patients with dengue hemor-rhagic fever. FEMS Immunol Med Microbiol 2000;28:151–5.

[20] Nguyen TH, Lei HY, Nguyen TL, et al. Dengue hemorrhagic feverin infants: a study of clinical and cytokine profiles. J Infect Dis2004;189:221–32.

[21] Gagnon SJ, Mori M, Kurane I, et al. Cytokine gene expression andprotein production in peripheral blood mononuclear cells of childrenwith acute dengue virus infections. J Med Virol 2002;67:41–6.

[22] Raghupathy R, Chaturvedi UC, Al-Sayer H, et al. Elevated levels ofIL-8 in dengue hemorrhagic fever. J Med Virol 1998;56:280–5.

[23] Libraty DH, Endy TP, Houng HS, et al. Differing influences ofvirus burden and immune activation on disease severity in secondarydengue-3 virus infections. J Infect Dis 2002;185:1213–21.

[24] Kurane I, Innis BL, Hoke Jr CH, Eckels KH, Meager A, Janus J, etal. T cell activation in vivo by dengue virus infection. J Clin Lab

[

[

[

[

[

[

[

[

[

[

[

[

genicity of tetravalent live-attenuated dengue vaccines in Thai adultvolunteers: role of serotype concentration, ratio, and multiple doses.Am J Trop Med Hyg 2002;66:264–72.

[4] Sabchareon A, Lang J, Chanthavanich P, et al. Safety and immuno-genicity of a three dose regimen of two tetravalent live-attenuateddengue vaccines in five- to twelve-year-old Thai children. PediatrInfect Dis J 2004;23:99–109.

[5] Rothman AL, Kanesa-thasan N, West K, Janus J, Saluzzo JF, EnnisFA. Induction of T lymphocyte responses to dengue virus by acandidate tetravalent live attenuated dengue virus vaccine. Vaccine2001;19:4694–9.

[6] Halstead SB, Marchette NJ. Biologic properties of dengue virusesfollowing serial passage in primary dog kidney cells: studies at theUniversity of Hawaii. Am J Trop Med Hyg 2003;69(6 Suppl.):5–11.

[7] Sanchez V, Hessler C, DeMonfort A, Lang J, Guy B. Comparisonby flow cytometry of immune changes induced in human monocyte-derived dendritic cells upon infection with Dengue 2 live attenuatedvaccine or 16681 parental strain. FEMS Immunol Med Microbiol2006;46(1):113–23.

[8] Mangada MM, Endy TP, Nisalak A, et al. Dengue-specific T cellresponses in peripheral blood mononuclear cells obtained prior tosecondary dengue virus infections in Thai schoolchildren. J InfectDis 2002;185:1697–703.

[9] Rothman AL. Dengue: defining protective versus pathologic immu-nity. J Clin Invest 2004;113:946–51.

10] Mangada MM, Rothman AL. Altered cytokine responses of dengue-specific CD4+ T cells to heterologous serotypes. J Immunol2005;175(4):2676–83.

11] Mangada MM, Ennis FA, Rothman AL. Quantitation of dengue virusspecific CD4+ T cells by intracellular cytokine staining. J ImmunolMethods 2004;284:89–97.

12] Osatomi K, Sumiyoshi H. Complete nucleotide sequence of denguetype 3 virus genome RNA. Virology 1990;176(2):643–7.

13] Mongkolsapaya J, Dejnirattisai W, Xu XN, et al. Original antigenicsin and apoptosis in the pathogenesis of dengue hemorrhagic fever.Nat Med 2003;9:921–7.

Immunol 1995;46:35–40.25] Shresta S, Kyle JL, Robert Beatty P, Harris E. Early activation of nat-

ural killer and B cells in response to primary dengue virus infectionin A/J mice. Virology 2004;319:262–73. Related Articles, Links.

26] Rothman AL, Ennis FA. Immunopathogenesis of Dengue hemor-rhagic fever. Virology 1999;257:1–6.

27] Kurane I, Ennis FA. Production of interferon alpha by dengue virus-infected human monocytes. J Gen Virol 1988;69:445–9.

28] Kurane I, Innis BL, Nimmannitya S, Nisalak A, Meager A, EnnisFA. High levels of interferon alpha in the sera of children withdengue virus infection. Am J Trop Med Hyg 1993;48:222–9.

29] Diamond MS, Roberts TG, Edgil D, Lu B, Ernst J, Harris E. Mod-ulation of Dengue virus infection in human cells by alpha, beta, andgamma interferons. J Virol 2000;74:4957–66.

30] Pichyangkul S, Endy TP, Kalayanarooj S, et al. A blunted bloodplasmacytoid dendritic cell response to an acute systemic viralinfection is associated with increased disease severity. J Immunol2003;171:5571–8.

31] Green S, Vaughn DW, Kalayanarooj S, et al. Elevated plasmainterleukin-10 levels in acute dengue correlate with disease sever-ity. J Med Virol 1999;59:329–34.

32] Simmons CP, Dong T, Chau NV, et al. Early T-cell responses todengue virus epitopes in Vietnamese adults with secondary denguevirus infections. J Virol 2005;79:5665–75.

33] Zivny J, DeFronzo M, Jarry W, Jameson J, Cruz J, Ennis FA, et al.Partial agonist effect influences the CTL response to a heterologousdengue virus serotype. J Immunol 1999;163:2754–60.

34] Loke H, Bethell DB, Phuong CX, et al. Strong HLA class I-restrictedT cell responses in dengue hemorrhagic fever: a double-edgedsword? J Infect Dis 2001;184:1369–73.

35] Stephens HA, Klaythong R, Sirikong M, et al. HLA-A and -B alleleassociations with secondary dengue virus infections correlate withdisease severity and the infecting viral serotype in ethnic Thais.Tissue Antigens 2002;60:309–18.

36] Seet RCS, Ooi EE, Wong HB, Paton NI. An outbreak of pri-mary dengue infection among migrant Chinese workers in Singapore

4926 V. Sanchez et al. / Vaccine 24 (2006) 4914–4926

characterized by prominent gastrointestinal symptoms and a highproportion of symptomatic cases. J Clin Virol 2005;33:336–40.

[37] Guirakhoo F, Pugachev K, Zhang Z, et al. Safety and efficacy ofchimeric Yellow Fever-dengue virus tetravalent vaccine formulationsin nonhuman primates. J Virol 2004;78:4761–75.

[38] Monath TP, Guirakhoo F, Nichols R, et al. Chimeric live, attenu-ated vaccine against Japanese encephalitis (ChimeriVax-JE): phase 2clinical trials for safety and immunogenicity, effect of vaccine doseand schedule, and memory response to challenge with inactivatedJapanese encephalitis antigen. J Infect Dis 2003;188:1213–30.

[39] Monath TP, McCarthy K, Bedford P, et al. Clinical proof of prin-ciple for ChimeriVax: recombinant live, attenuated vaccines againstflavivirus infections. Vaccine 2002;20:1004–18.

[40] Fox JP, Penna HA. Befavior of 17D Yellow Fever virus in rheusmonkeys. Relation to substrain, dose and neural or extraneural inoc-ulation. Am J Hyg 1943;38:152–72.

[41] Durbin AP, Whitehead SS, McArthur J, et al. rDEN4delta30, a liveattenuated dengue virus type 4 vaccine candidate, is safe, immuno-genic, and highly infectious in healthy adult volunteers. J Infect Dis2005;191:710–8.

[42] Guy B, Chanthavanich P, Gimenez S, et al. Evaluation byflow cytometry of antibody-dependent enhancement (ADE) ofdengue infection by sera from Thai children immunized with alive-attenuated tetravalent dengue vaccine. Vaccine 2004;22:3563–74.