OUP user menu

Comparison by flow cytometry of immune changes induced in human monocyte-derived dendritic cells upon infection with dengue 2 live-attenuated vaccine or 16681 parental strain

Violette Sanchez, Catherine Hessler, Aymeric DeMonfort, Jean Lang, Bruno Guy
DOI: http://dx.doi.org/10.1111/j.1574-695X.2005.00008.x 113-123 First published online: 1 February 2006


Dengue is an important threat for world-wide public health. Different vaccines are under development, which are currently assessed using a battery of in vitro and in vivo assays before moving on to humans. It is also important to assess vaccine characteristics on human primary cells; among them, dendritic cells, the most efficient antigen-presenting cells, are the first targets of dengue virus infection. In this study, we used flow cytometry to compare the consequences of such an infection by dengue serotype 2 live-attenuated vaccine (LAV2) or its parental strain DEN2 16681 (DEN2). Optimal conditions of infection have first been defined by a mathematical approach, and flow cytometry allowed studying modifications induced in both infected and noninfected dendritic cell populations after surface and intracellular labeling. Both DEN2 and LAV2 increased the expression of the phenotypic markers CD80, CD86, CD40, CD1a, HLA ABC and CD83, demonstrating cellular activation. Stimulated dendritic cells produced tumor necrosis factor-α in particular, and, to a lower extent, interleukin 6. Of importance, whereas DEN2 induced cytokine production both in the infected and noninfected populations, LAV2-induced cytokine production was restricted to the infected population. This limited activation triggered by LAV2 would be in agreement with its attenuation. In conclusion, these in vitro experiments using primary human dendritic cells may participate, in combination with other assays, to the evaluation of the immunogenicity and safety of dengue vaccine candidates.

  • dengue virus
  • dengue vaccine
  • dendritic cells
  • experimental design
  • flow cytometry


Dengue fever is a widespread viral illness found in about a hundred countries, and more than 50 million cases of dengue infection are recorded each year. It usually appears as a febrile syndrome (dengue fever) but it can lead to more serious symptoms such as hemorrhagic fever (or dengue hemorrhagic fever) or hypovolemic shock (or dengue shock syndrome), which is often fatal. Dengue hemorrhagic fever is distinguished from dengue fever by an increased vascular permeability, hemostatic anomalies and bleeding (Mairuhu et al., 2004).

Dengue disease is caused by a Flavivirus (serotypes 1–4), which carries a RNA sequence coding for three structural proteins and seven nonstructural proteins (Henchal & Putnak, 1990). Upon infection, the immune system develops both humoral and cellular responses with production of specific antibodies and activation of specific CD4+ and CD8+ T lymphocytes. A possible bias in this immune response could favor the induction of dengue hemorrhagic fever and dengue shock syndrome (Rothman & Ennis, 1999; Lei et al., 2001). It is suggested in particular that subneutralizing antibodies would facilitate penetration and dissemination of the virus by an antibody-dependent enhancement phenomenon (Watts et al., 1999; Chen et al., 2001; Libraty et al., 2002). Activation of lymphocytes is also responsible for the production of soluble mediators (tumor necrosis factor α (TNFα), interleukin 8 (IL8), γ-interferon (INFγ) and other proinflammatory and/or anti-inflammatory mediators), which could contribute to the severity of the disease (Chaturvedi et al., 2000).

As of today, there is neither specific treatment nor vaccine to fight the disease, and the only — partially effective — method of curbing the propagation of the disease is to eliminate the mosquito vector, Aedes aegygti (World Health Organization, 1997). Production and preclinical testing of different dengue vaccines candidates are currently in progress (Kanesa-thasan et al., 2001, 2003; Halstead & Deen, 2002; Gwinn et al., 2003; Innis & Eckels, 2003; Sun et al., 2003). The aim of Sanofi Pasteur is to develop a safe dengue vaccine capable of inducing antibodies neutralizing the four serotypes at similar levels, in order to abolish the potential danger linked to the antibody-dependent enhancement phenomenon. The development of these candidate vaccines has, initially, been based on the attenuation of the live virus through passage in primary dog kidney cells or in African green monkey kidney cell cultures (LAV DEN vaccines) (Eckels et al., 2003). For each vaccine candidate, attenuation is addressed by different means. Genetic stability is assessed by virus sequencing (genotyping), and a set of phenotypic tests is conducted in parallel: growth, thermosensitivity and plaque size in different cell lines, transmission in mosquitoes, and neurovirulence and viscerotropism in mice and monkeys. In monkeys, vaccine candidates must induce upon subcutaneous administration the lowest possible viremia and significant seroconversion, which would upon challenge confer protection against viremia induced by parental virus (Guirakhoo et al., 2004). This set of tests is mandatory to support vaccine development, but not always fully predictive. It is therefore important to design other in vitro tests that use primary human cells in order to identify and understand better the attenuation criteria of the vaccine candidates in comparison with their parental viruses.

Many studies have shown that the dengue virus infects circulating or tissue-homing cells (Avirutnan et al., 1998; Huang et al., 2000) such as monocytes (Bosch et al., 2002; Espina et al., 2003), B lymphocytes (Lin et al., 2002) and dendritic cells (DCs) (Wu et al., 2000; Ho et al., 2001; Libraty et al., 2001; Palmer et al., 2005). DCs are antigen-presenting cells of myeloid and lymphoid/plasmacytoid origin, present in a large number of peripheral tissues (Banchereau et al., 2000; Steinman, 2003). The DCs, and particularly the Langerhans cells and dermal/interstitial DCs, would be the primary targets of natural dengue infection (Wu et al., 2000), which would promote maturation and migration from the peripheral tissues to the lymph nodes. They would then in a second step activate naive CD4+ and CD8+ cells. DCs stimulated by the dengue virus secrete cytokines such as TNFα and α-interferon (IFNα) (Libraty et al., 2001; Palmer et al., 2005), the latter cytokine being also produced by the plasmacytoid DCs, even though not significantly susceptible to infection (Izaguirre et al., 2003). Stimulated DCs secrete little or no interleukin 12 (IL12) (Libraty et al., 2001), whilst the secretion of this cytokine increases in the presence of IFNγ. Production by DCs of interleukin 10 (IL10) or IL12 during maturation may bias subsequent Th2 or Th1 immune responses respectively (Corinti et al., 2001). Thus, during the early innate response, DCs mature and modulate adaptive immune responses. Their positive or negative role in dengue disease evolution may depend in part on this orientation.

It is possible to generate immature dendritic cells in vitro from CD14+ monocytes or from CD34+ progenitor cells, and then to follow the phenotypic changes induced in immature DCs upon antigen stimulation. These changes can be analyzed in particular by monitoring the expression of maturation markers such as antigen-presenting molecules (MHC I, MHC II and CD1a), costimulation molecules (CD40, CD80 and CD86) and of molecules involved in DC migration (CD11b) (Banchereau et al., 2000). Expression of CD83 (potential immunoregulator) (Lechmann et al., 2002) can be analyzed as well. Expression of cytokines can be monitored in parallel using intracellular staining. This type of test allows comparing different vaccine preparations and serotypes vs. their respective parental virus in their capacity to stimulate DCs.

In the present study, we have first used an experimental design to determine optimal conditions for infecting DCs by dengue virus. This was performed on DEN2 16681 serotype (referred as DEN2 in this study), for which a live-attenuated-derived vaccine (referred as LAV2) has been tested in humans as a monovalent, bivalent, trivalent and tetravalent vaccine (Dharakul et al., 1994). LAV2 induced good seroconversion in humans in the absence of clinical signs and reactogenicity. It was therefore an appropriate reference for evaluating this assay, and has been compared to its parental field virus. Intracellular cytokine staining and phenotyping by flow cytometry were used to measure the responses to viral stimulation. These techniques enabled us to look simultaneously at the phenotype, cytokines expression and infection levels of the DCs, and thus to compare infected vs. noninfected cells in the total population for the same donors. This provided useful data for characterizing live-attenuated vaccine candidates with respect to potential immunogenicity and safety.

Materials and methods

Viruses, culture medium and reagents

The cell culture medium comprised RPMI 1640 (Gibco, Paisley, UK), glutamine 2 mM (Gibco), 10% heat-inactivated fetal calf serum (HyClone, Logan, USA), penicillin 100 U mL−1 and streptomycin 100 U mL−1 (Gibco). Recombinant GMCSF and IL4 were purchased from Novartis/Sandoz (Basel, Switzerland) and PeproTech (London, UK) respectively. The staining buffer comprised 0.1% BSA (Sigma, St Louis, MO) and 0.01% NAN3 (Sigma) in PBS. The fixing buffer comprised 4% formaldehyde (Aldrich, St Louis, MO) in staining buffer.

Of the phycoerythrin-labeled monoclonal antibodies mAbs against cell surface molecules, anti-HLA ABC, anti-CD83, anti-CD86, anti-CD40 and antimannose receptor were from Pharmingen (San José, CA); anti-HLA DR and anti-CD80 were from Becton Dickinson (San José, CA); CD1a was from DAKO; and CD11b phycoerythrin was from IOTest Immunotech (Marseille, France). Monoclonal antidengue 2 antibody (3H5-1-12) was produced by Biotem (Rivier d'Apprieu, France). The antibody was Alexa 488-conjugated using A 488 kit (Molecular Probes, Eugene, OR). The phycoerythrin-conjugated anticytokine antibodies anti-IL6, anti-IL10, anti-IL12p70, anti-TNFα were purchased from Pharmingen.

Parental DEN2 16681 (DEN2, Batch 13/06/1994) and PDK53 LAV2 vaccine strain (LAV2, Batch 16/11/01) were produced in Sanofi Pasteur.

Preparation of dendritic cells

Whole blood was collected in sodium citrate bags from healthy donors by EFS (Etablissement Français du Sang) and dendritic cells prepared from monocytes according to standard procedures (Libraty et al., 2001; Palmer et al., 2005).

Experimental design and statistics

Flow cytometry was used to compare immune responses obtained after viral stimulation (by LAV2 or DEN2). This comparison required significant infection of DCs, and cells had to remain viable to address accurately phenotypic changes after infection. Therefore, it was of interest to optimize the process with the aim of increasing the infection (Y1: infection rate, first experimental response studied) while maintaining the viability of the DCs (Y2: percentage of nonviable cells, second response studied). Because many factors can modulate these responses, we first had to select some of them, and fix some others (for instance, the percentage of serum was fixed at 2%). Five parameters were retained as they were considered to have an influence on cell viability and percentage of infection, and each parameter was tested at two distinct levels, coded −1 and +1 in the experimental matrix. These parameters were: multiplicity of infection (X1), tested at 0.25 and 1; duration of infection (X2), tested at 2 and 6 days; status of differentiation (X3), tested at 7 and 10 days; temperature (X4), tested at 32 and 37 °C; pH (X5), tested at 6.6 and 7.8. The full factorial matrix allowed 25 distinct combinations to calculate all main effects of five variables and interactions (Box & Hunter, 2000). Based on previous knowledge, a half fraction of the complete design (25–1) was considered sufficient. A total of 16 distinct experiments were thus performed twice and independently in a random order from a more statistical point of view (not shown): this allowed us both to check the reliability of the experiment and to estimate the experimental variability properly. Finally, negative and positive experiments were added for control and quality of the experimental design. The global validity of the model was established from the variance, the R2 coefficient and the residues distribution. The statistical significance of the individual regression coefficients was then calculated from a Student's t-test (Table 1). All calculations were carried out using nemrod software (Mathieu & Phan-Tan-Luu, 1981). Statistical analysis was carried out to evaluate the effects, and those that were biologically relevant were then taken into account when planning future experiments.

View this table:
Table 1

Statistical analysis of the fractional matrix 25–1: influence of the five parameters on DC infectivity and viability

log Y1log Y2
ParametersCoefficientSignificance (P)CoefficientSignificance (P)
Multiplicity of infectionb10.020.146−0.010.2
Time of infectionb2−0.030.2340.26<0.01
Time of differentiationb3−0.23<0.010.19<0.01
  • Y 1, percentage infection; Y2, percentage nonviable cells.


Immature myeloid DCs were infected with DEN2 and LAV2 at a multiplicity of infection of 0.5 in RPMI 1640 medium containing 2% of heat-inactivated serum. For the experimental plan, 106 DCs were stimulated with different multiplicity of infection of LAV2 as defined above. DCs were then cultured (106 DCs mL−1), without washing cells, at 32 °C in 2% heat-inactivated FCS-IL4-GMCSF. Heat-inactivated virus (56°C for 30 min) was used as a negative control for virus infection (not shown). At 48 h, half of the cells were collected and dispensed into tubes for immunostaining of phenotypic markers. Brefeldin A (Sigma) was added to the other cells at 10 µg mL−1, and after 18 h incubation at 32°C, 5% CO2, cells were collected and labelled for intracellular dengue 2 antigen and cytokine expression (intracellular staining).

Flow cytometry

Analysis was performed according to standard procedures (Libraty et al., 2001; Palmer et al., 2005). All samples were analyzed on a FACScalibur flow cytometer (Becton Dickinson) within 24 h of staining.

Measurement of cytokines in supernatants

Cytokines were measured by a cytokine bead array (Becton Dickinson; proinflammatory kit, IL12, TNFα, IL10, IL1, IL6 and IL8) and data was analyzed by flow cytometry according to the instructions of the manufacturer.

Statistical methods

In order to take into account the increase of the type I risk α owing to multiple comparisons, this analysis was done by a two-step procedure: (1) comparisons were first made between all populations (infected or not) for each stimulation (LAV2 or DEN2) and control; and (2) the most relevant contrasts were then compared with the preceding comparisons. The test used in all analyses is a permutation test for paired data (nonparametric) (Siegel & Castellan, 1988). This allows calculating, on paired data, the exact probability that the differences observed between the two compared groups occurs by chance. Analyses were done with statxact 4 software (Witzenhausen, Germany) with a one-sided risk of 5% for comparisons with control (in order to check comparison higher than control), and a two-sided risk of 5% for comparisons between stimulations (in order to check differences). All graphics were done with sas® v8.2 software (SAS, Heidelberg, Germany). For more legibility, the values were log10-transformed, except for the values of %TNFα and %IL6.


Determination of optimal conditions for DC infection: use of an experimental design

In order to determine the optimal condition for DEN infection in DCs, a qualitative approach allowed us to establish the respective weight of the factors initially selected. These factors were multiplicity of infection (X1), duration of infection (X2, in days), duration of DC differentiation (X3), temperature (X4) and pH (X5). The influence of the five parameters on DC infection and viability was evaluated by means of the fractional matrix with 25–1 entries, a method that allowed us calculating the coefficients more accurately and at less cost (see the Materials and methods section). Experiments were performed in a random manner and all assays were done in duplicates. Calculations were carried out using nemrod software. For each parameter, values of coefficients and their significance, regardless of potential interactions, are shown in Table 1. The variance obtained with the duplicate experiments, the responses related to infection [(log Y1) SD=0.07283] and cell mortality [(log Y2) SD=0.04930] (where SD is the standard deviation) demonstrated the good reproducibility of the assay.

The statistical analysis of the coefficients of the Y1 response (percentage of infection) highlighted the significant effect of the temperature [regression coefficient −0.51 (b4), P<0.01], compared with the other factors. In addition, the duration of cellular differentiation (b3=−0.23, P<0.01) also had an influence on the levels of infection. Both these effects were negative (with the b coefficients negative): their increase (from 32 to 37°C and from 7 to 10 days, respectively) decreased the infection levels.

The conditions corresponding to the highest infection levels and obtained from the 16-experiment matrix were associated with a too high percentage of nonviable cells (not shown). We then defined in a second step conditions giving an infection yield of 30% for a cellular loss of lower than 10%. These conditions correspond to 7 days of DC differentiation, and 2 days of infection with a multiplicity of infection of 1, temperature of 32 °C and pH of 7.8.


Different tests were then performed in our laboratory, which confirmed that levels of infection were higher and, most importantly, more reproducible at 32 than at 37°C. This was observed not only for DEN2 and LAV2, but also for all four serotypes tested (parental and live-attenuated viruses, not shown). However, the observed immunological tendencies (phenotypic and cytokine secretions) were similar at 32 and 37°C (data not shown). Because the differences were more distinct at 32°C, this temperature has been retained for all our tests in order to obtain data that is more reproducible and easier to interpret. Conversely, the quantity of virus has been fixed at a multiplicity of infection of 0.5 rather than 1 for reasons of availability of viruses and vaccines. The same infection conditions were always used for DEN2 and LAV2. Even though it enabled us to fix certain criteria, the experimental plan has not, however, prevented significant interdonor variation in the obtained level of infection (from 2% to 13%, according to the donor). Similar infection variability has already been observed by other authors in human DCs (Cologna et al., 2005). Phenotypic studies were thus performed on eight donors in order to show significant trends; infection levels observed with LAV2 were consistently two to three times higher than those induced by DEN2 (not shown).

Application to live-vaccine evaluation: comparison of the phenotypic changes induced by LAV2 or parental D2 virus

Analysis of the infection by flow cytometry allowed simultaneous analysis of the phenotypic changes in the infected vs. noninfected populations. Immature myeloid DCs were infected by LAV2 or DEN2 in the conditions previously described. At 48 h, cells were analyzed for the expression of the following surface markers: HLA-ABC, HLA-DR, CD1a, CD11b, CD40, CD80, CD83, CD86 and mannose receptor. Results obtained on the dot-plots were expressed as mean fluorescent intensity of the specified marker on infected DCs or noninfected DCs present in the same sample. Infection with inactivated virus did not trigger detectable changes in DCs in the conditions of our assays (not shown).

In a first step, analysis was performed by comparing all populations (whether infected or not) to their controls, upon each stimulation (LAV2 or DEN2). Compared with the control DCs, the two markers CD80 and CD83 presented an increased expression after infection, whichever the stimulus (parental or attenuated virus), and whichever the analyzed population, infected or noninfected (Fig. 1). Similar trends were observed for the CD86 marker.

Figure 1

Dengue virus/vaccine infection of DCs induce changes in CD80, CD86 and CD83 phenotypic markers expression. Cells were incubated with antigen LAV2 or DEN2 for 48 h. After fixation and permeabilization, cells were stained for maturation marker (s) and intracellular dengue 2 expressions and analyzed by flow cytometry. The results of eight independent experiments are shown. Each line represents one individual. Analyses were done with statxact 4 software with a one-sided risk of 5% for comparisons with control, and a two-sided risk of 5% for comparisons between stimulations (d,e). All graphics were done with sas® v8.2 software. *increased mean fluorescence intensities (MFI), P≤0.05, **increased MFI, P≤0.01. Control: cells treated with medium only. ninf: cells treated with LAV2 or DEN2 and negative after staining with antidengue2 Mab; inf: cells treated with LAV2 or DEN2 and positive after staining with antidengue2 Mab.

In the case of HLA DR and HLA ABC expressions, differences were stronger upon DEN2 stimulation and did not reach statistical significance upon LAV2 stimulation. In addition, a slight increase was observed for CD40 and CD1a markers expression (not shown). Finally, the cells did not present any up-regulation in the expression of the CD11b and mannose receptor markers.

Besides comparisons to controls, other comparisons were made in a second step: (1) infected vs. noninfected population for each stimulus; (2) LAV2-noninfected vs. DEN2-noninfected population; and (3) LAV2- or DEN2-infected populations vs. each other. In general, no differences between the different populations have been observed. One can notice, however, for CD86 and HLA ABC markers, a significant difference between LAV2- vs. DEN2-noninfected populations: DEN2 induced higher expression levels of these markers than LAV2 (not shown).

All in all, DC activation was induced after LAV2 or DEN2 stimulation, as far as the expression of the antigen-presenting markers (CMH I, CMH II and CD1a), the costimulation markers (CD80 and CD86) and the CD83 marker were concerned. In this situation, apart from a few cases, the infected population was more activated than the noninfected population.

Comparison of the cytokines produced in DCs by intracellular staining

Results obtained were expressed as ratios in order to remove differences in infection between the DEN2 and LAV2. This ratio corresponds to the percentage of infected DCs (inf) secreting cytokine (CK+), expressed as follows. [R inf+]=inf CK+/(inf CK+inf CK+) × 100. Similar percentages were specified for the noninfected DCs (ninf) producing cytokines and present in the same sample: [R ninf+]=ninf CK+/(ninf CK+ninf CK+) × 100.

Analysis of cytokines induced by the different stimuli showed differences between DEN2 and LAV2, particularly for TNFα (for an example of dot-plots, see Fig. 2). Indeed, when DCs were infected with LAV2, only infected cells (inf) produced significant amounts of TNFα (Fig. 3). On the other hand, when DCs were infected by DEN2, the two populations, infected and noninfected, produced significant amounts of TNFα in comparison with their controls. In agreement, significant differences were observed between infected and noninfected populations when the cells were stimulated with LAV2, but not with DEN2. Regarding IL6 (Fig. 3), production levels were lower than for TNFα but the same trends were observed. In that case, DEN2 induced the highest production of cytokine in the two populations, infected and noninfected, in comparison with their controls. LAV2 induced a slight IL6 production by infected cells, but this effect was not significant compared with controls. The noninfected population stimulated by DEN2 always produced higher levels of cytokines than the corresponding population stimulated by LAV2. Finally, in the conditions of our intracellular staining assays, we did not observe intracellular production of IL10 and little production of IL 12p70 by the DCs, the later cytokine being induced only by parental virus in one donor (not shown).

Figure 2

Example of intracellular staining results. Cells were incubated with LAV2 or DEN2 for 48 h. Brefeldin A was added for another 18 h. After fixation and permeabilization, cells were stained for intracellular cytokine(s) and dengue 2 expressions, and analyzed by flow cytometry. In this experiment, 5% of the DCs were infected by DEN2% and 12.5% by LAV2. Controls: cells treated with medium only.

Figure 3

Dengue virus/vaccine infection stimulate TNFα and IL6 production. Cells were stimulated by LAV2 or DEN2 2 for 48 h. After fixation and permeabilization, the cells were stained for intracellular cytokine and dengue 2 expressions and analyzed by flow cytometry. The results of eight independent experiments are shown. Each line represents one individual. Analyses were done with statxact 4 software with a one-sided risk of 5% for comparisons with control, and a two-sided risk of 5% for comparisons between stimulations. All graphics were done with sas® v8.2 software. Control: cells treated with medium only. ninf: cells treated with LAV2 or DEN2 and negative after staining with antidengue2 Mab; inf: cells treated with LAV2 or DEN2 and positive after staining with antidengue2 Mab.

Measurement of cytokines in supernatants

Cytokines released in supernatants were measured in half of the donors using cytokine bead array and a proinflammatory kit (IL12, TNFα, IL10, IL1, IL6 and IL8; results not shown). Although we could not discriminate in that case between infected and noninfected cells production, results showed that DEN2 induced higher levels of IL6 and TNFα than LAV2, as seen by intracellular staining. According to the donor, five- to 10-fold higher levels of IL6 or TNFα were detected upon DEN2 compared with LAV2 stimulation. Interestingly, and in contrast to intracellular staining results where IL10 could not be detected, this difference was also observed for IL10 secretion, although overall levels were 10 times lower for that cytokine than those obtained with IL6 or TNFα. DEN2 then triggered expression of both proinflammatory (IL6 and TNFα) and anti-inflammatory (IL10) cytokines at higher levels than LAV2. No significant variations in IL12, IL1 or IL8 levels were detected compared with control medium conditions. Finally, although we did not quantify IFNα levels in these experiments, we performed such ELISA analyses in other experiments upon serotype 3 infection and detected, as observed by other authors (Libraty et al., 2001; Palmer et al., 2005) secretion of this cytokine (we also detected IFNβ production); no significant differences were observed between parental and live vaccine in this respect.


We have used flow cytometry and intracellular cellular staining assay in the dengue model with the aim of performing in vitro characterization of our different candidate vaccines. This technique enabled us to differentiate between the infected and noninfected cells in a total population.

Infection of DCs: use of experimental design

Optimal conditions of infection were determined by an experimental approach. This is not yet common in biology, although it is a powerful technique for screening variables and optimizing their effects, and also for considering their interrelations. In fact, its implementation requires fewer experiments than other approaches and the information obtained is more robust and reproducible. Our results showed that infection was more effective if it took place at 32 °C, being consistent with the thermosensitive nature of the virus (Vithanomsat et al., 1983). The thermosensitivity of vaccine candidates at 37 or 39°C represents an important attenuation criterion, but has been addressed accurately in other assays, not described here (Vithanomsat et al., 1983). Regarding our DC experiments, although 32°C is not a physiological temperature, no qualitative differences were observed when experiments were performed at 37°C (not shown), and results were more reproducible at 32°C, which incited us to keep this latter temperature in our assay.

Within our experimental design, other parameters carried less weight. However, DCs presented a higher permissivity to infection when the number of days of differentiation was lower (7 days) and, expectedly, when the amount of infecting virus increased. Nevertheless, this latter parameter had a negative influence on cell viability. Despite optimization, infection levels remained subject to significant interdonor variation. Separate analysis of infected and noninfected populations allowed this variation to be compensated for.

In vitro analysis of the cell response in the dengue model: the role of APCs

We have compared the capacity of DEN2 and LAV2 to induce DC maturation and cytokine production. LAV2-induced changes in phenotypic markers were similar to DEN2 ones, whereas for HLAs this effect was more pronounced for DEN (Table 2). The observed increase in HLA-ABC (Class I) and CD1a expression on DCs suggests an increase in their capacity to present viral antigens. In addition, the increased expression of costimulation molecules (CD40-CD80-CD86) and CD83 up-regulation suggests some improved stimulation and presentation to T cells. Thus, LAV2 seems to be as efficient as DEN2 in triggering an immune response in vitro (and may be supposed to be so in vivo).

View this table:
Table 2

Dengue virus or vaccine infection induces changes in dentritic cells

Phenotypic markersCytokines
HLA ABCHLA DRCD1aCD11bCD40CD80CD83CD86Mannose receptorTNFαIL6IL10IL12
  • For phenotypic markers, comparisons were made between mean fluorescence intensities (MFI), and for cytokines, between percentages of positive cells (%)

  • No statistically significant variation.

  • Increased MFI or %, P≤0.05.

  • ** Increased MFI or %, P≤0.01.

  • Control, cells treated with medium only. ninf, cells treated with LAV2 or DEN2 and negative after staining with antidengue2 Mab; inf, cells treated with LAV2 or DEN2 and positive after staining with antidengue2 Mab.

Infections of DCs by DEN2 and LAV2 are accompanied by the production of TNFα and IL6, with little IL12 p70 (one donor in eight) and no IL10 in these intracellular staining assays. A kinetic study (24, 48 and 72 h) on two donors was performed for each cytokine, giving similar results for the different time points (not shown). As we detected IL10 in supernatants by the cytokine bead array method, this would suggest a lack of recognition by anti-IL10 antibody in the intracellular staining assay, possibly as a result of partial cytokine denaturation during the permabilization/fixation steps.

Differences observed in the level and source of cytokine production would be in favor of viral attenuation as far as safety is concerned. Actually, in contrast to LAV2 stimulation, stimulation by DEN2 effectively induced the production of high levels of TNFα and IL6 by both the infected and noninfected populations (Table 2). This difference is not a result of higher initial infection by DEN2. On the contrary, LAV2 was shown to be more infectious in our assays. However, different viral outputs from DEN2- or LAV2-infected cells, as observed for different viruses (Cologna et al., 2005), may have contributed to different stimulation and eventually cytokine production by noninfected cells through contact with viral particles, even in absence of detectable productive infection. This requires more investigation. Regarding IL10, it seems that in addition to proinflammatory cytokines, DEN2 can also trigger anti-inflammatory signals, thus compensating partially the effects of the former cytokines.

Stimulation of the noninfected population of myeloid dendritic cells by DEN2 (as seen by up-regulation of CD83 and surface activation markers) is a phenomenon that has already been observed (Libraty et al., 2001). In this respect, although some differences exist between our experimental conditions and those used by other authors with DEN2 16803 (Palmer et al., 2005) or 16681 virus strains (Libraty et al., 2001), we obtained results qualitatively similar to those obtained by Libraty and coworkers with the same 16681 strain (Libraty et al., 2001). Compared with controls, significant up-regulation of phenotypic markers was observed by these authors in both infected and noninfected DC populations, a higher response being observed, however, in noninfected cells for some markers. These latter results and ours were nevertheless different from the findings obtained with the 16803 DEN2 virus strain by Palmer (2005), who observed no up-regulation of phenotypic markers in infected cells compared with controls. Therefore, it appears that differences in virus strains rather than differences in experimental conditions would be responsible for such discrepancies. Actually, the DC-based assay used in these studies appears suitable for discriminating between different field strains, as well as between parental and vaccine strains such as those in the present work. In any case, one has to be aware that even for one given virus strain, different preparations coming from different laboratories may correspond to some extent to virus possessing different characteristics.

The overall effect of DEN2 or LAV2 on the noninfected cells may result from action by the soluble factors such as the cytokines and/or by the fixing of viral particles on the surface of these cells, as stated above. In subsequent studies, it will be useful to follow through DNA arrays the induction of a larger number of immune-regulatory genes (in progress). One has to notice, however, that in these conditions (cytokines in supernatant or DNA array) it will not be possible to distinguish infected from noninfected cells.

Regarding TNFα in particular, its presence in serum has been observed in humans upon infection with dengue virus, and it has been proposed that disease severity would be related to TNFα plasma levels, higher amounts being detected in dengue hemorrhagic fever patients compared with dengue fever or asymptomatic patients (Azeredo et al., 2001). Based on the present results, vaccination with attenuated vaccine would induce a more targeted and better-controlled immune response than parental virus, in favor of better safety. It will be important in future experiments to measure the levels of viral secretion by LAV2- or DEN2-infected DCs, and to investigate the capacity of the viruses produced after a first round of replication to infect new DCs or other cell types, which could also play a critical role in the cascade of events following initial infection in vivo.

In conclusion, characterization of dengue vaccine candidates is currently done via a battery of assays, both in vitro and in animals. Flow-cytometry-based assays such as those described here allowed assessing in vitro human DCs responses potentially linked to the immunogenicity and safety of dengue vaccine candidates. We are aware that one single assay cannot fully predict on its own such vaccine characteristics, and that only a combination of different in vitro and in vivo assays can give a more accurate prediction. Nevertheless, such an approach may contribute to this global knowledge. The experimental conditions defined here with serotype 2 are currently applied to the evaluation of the three other vaccine serotypes, measuring similar endpoints (in progress).


The authors want to acknowledge Nicolas Burdin and Emanuelle Trannoy for constant support and helpful discussions, Beatrice Barrere, Pascal Champalle, Denis Crevat and Christophe Fournier for providing viruses and vaccines, and Sutee Yoksan and Arunee Sabchareon for their constant help.


  • Editor: Willem van Eden


View Abstract