DENV (dengue virus) induces UPR (unfolded protein response) in the host cell, which strikes a balance between pro-survival and pro-apoptotic signals. We previously showed that Salubrinal, a drug that targets the UPR, inhibits DENV replication. Here, we examine the impact on UPR after direct or ADE (antibody-dependent enhanced) infection of cells with DENV clinical isolates. THP-1 cells in the presence of subneutralizing concentration of humanized antibody 4G2 (cross-reactive with flavivirus envelope protein) or HEK-293 cells (human embryonic kidney 293 cells) were infected with DENV-1–4 serotypes. UPR gene expression was monitored under these infection conditions using real-time RT–PCR (reverse transcription–PCR) and Western blots to analyse serotype-dependent variations. Subsequently, in a blinded study, strain-specific differences were compared between DENV-2 clinical isolates obtained from a single epidemic. Results showed that THP-1 cells were infected efficiently and equally by DENV-1–4 in the ADE mode. At 48 hpi (h post infection), DENV-1 and -3 showed a higher replication rate and induced higher expression of several UPR genes such as BiP (immunoglobulin heavy-chain-binding protein), GADD34 (growth arrest DNA damage-inducible protein 34) and CHOP [C/EBP (CCAAT/enhancer-binding protein)-homologous protein]. The ADE infection of THP-1 cells with epidemic DENV-2 high-UPR-gene-expressing strains appears to correlate with severe disease; however, no such correlation could be made when the same viruses were used to infect HEK-293 cells. Our finding that UPR gene expression in THP-1 cells during ADE infection correlates with dengue disease severity is consistent with a previous study [Morens, Marchette, Chu and Halstead (1991) Am. J. Trop. Med. Hyg. 45, 644–651] that showed that the growth of DENV 2 isolates in human peripheral blood leucocytes correlated with severe and mild dengue diseases.
Dengue is the most common mosquito-borne viral diseases in the world [1,2]. There are four serotypes of the virus, all of which can produce an acute disease with a wide spectrum of clinical presentations. Although most patients present with a self-limiting, non-severe form of disease, referred to as DF (dengue fever), a small proportion can develop a more severe form of the disease with plasma leakage and/or haemorrhage [DHF (dengue haemorrhagic fever)] [3,4]. DHF is further classified into four severity grades, with grades III and IV being defined as DSS (dengue shock syndrome) . There are various factors that play a role in determining the severity of dengue infection. These include host factors [6–9], vector , environmental factors  and viral factors, such as genetic variability [11,12] and the level of circulating virus [13,14]. Although several recent publications have focused on host factor studies at the genomic level using siRNA (small interfering RNA) [15,16] and microarray [17–19] technologies, it is still unclear why some viral strains cause serious dengue disease while others do not.
When DENV (dengue virus) infects the host cell, it releases its positive-stranded RNA into the cytosol that is used as a template for translating the viral polypeptide. This takes place in convoluted membranes derived from the ER (endoplasmic reticulum). The viral proteins are then processed and folded in the ER . The increased translation of viral proteins in the ER leads to accumulation of unfolded and misfolded proteins in the compartment. This induces a pro-survival signal in the cell known as the UPR (unfolded protein response) . The pathway involves three proximal sensors: PERK [PKR (double-stranded-RNA-dependent protein kinase)-like ER kinase], ATF6 (activating transcription factor 6) and the ER transmembrane protein kinase/endoribonuclease IRE1, which are regulated by the ER chaperone BiP (immunoglobulin heavy-chain-binding protein)/GRP78 (glucose-regulated protein of 78 kDa) (Figure 1A). During homoeostasis in healthy cells, BiP is mostly bound to the UPR sensors in the ER membrane on the luminal side and maintains a repressed state. In cells undergoing stress, accumulation of misfolded or unfolded proteins causes the dissociation of BiP from the UPR sensors, which leads to their activation [22,23]. Activation of PERK induces phosphorylation of eIF2-α (eukaryotic initiation factor 2-α), which leads to global translational attenuation. It also activates ATF4 that transactivates GADD34 (growth arrest DNA damage-inducible protein 34), which acts in turn as a cofactor for the eIF2-α phosphatase, to relieve stress-induced translational inhibition. In the case of sustained ER stress, ATF4 activates CHOP [C/EBP (CCAAT/enhancer-binding protein)-homologous protein] that induces the pro-apoptotic pathway leading to cell death .
Serotype-specific DENV replication in HEK-293 and THP-1 cells
Umareddy et al.  showed that DENV-1 (MY 10245) and DENV-2 (TSV01) infection of A549 cells increased the phosphorylation of eIF2-α, indicative of ER stress. Treatment with Salubrinal, an inhibitor of GADD34, resulted in a decreased viral titre. A recent report suggested a role for BiP in dengue protein folding and production of infectious virus . Previous studies have also shown that BiP is induced in cells infected with paramyxoviruses , hantaviruses  as well as flaviviruses such as JEV (Japanese encephalitis flavivirus)  and other viruses .
Since increased viral protein translation can lead to higher ER stress, we hypothesized that the level of ER stress can be a determinant of the viral replication and a potential predictor of dengue severity in patients. We found that there are serotype-specific differences in the level of ER stress in HEK-293 cells (human embryonic kidney 293 cells) and human monocytic cells (THP-1) cells as determined by UPR gene-specific mRNA levels. When direct infection of HEK-293 cells with clinical isolates of DENV-2 from Indonesian outbreak was used to test the hypothesis, the results showed that the up-regulation in UPR genes is not predictive of disease severity. However, when DENV-2 clinical isolates were used to infect THP-1 monocytes in ADE (antibody-dependent enhancement) mode, we found that the strains isolated from patients with severe dengue disease produced higher ER stress compared with those strains isolated from patients with uncomplicated DF, indicating the need to use the ADE infection model for virulence and dengue severity prediction.
MATERIALS AND METHODS
Cell culture and viral infection
HEK-293 cells were grown in DMEM (Dulbecco's modified Eagle's medium) and THP-1 cells were grown in RPMI 1640 medium supplemented with 10% (v/v) FBS (fetal bovine serum) and 1% penicillin/streptomycin. All cells were cultured at 37°C in a humidified incubator with 5% CO2. For infection of HEK-293 cells, the cells were incubated with the virus at an MOI (multiplicity of infection) of 1 for 1 h in 2 ml of medium [with 2% (v/v) FBS], washed once with PBS and replaced with fresh growth medium. THP-1 cells were infected by incubating virus (MOI = 1) with cells in 1 ml of medium (2% FBS) for 1 h, washed once with PBS and replaced with 3 ml of growth medium. For ADE infection of THP-1 cells, virus (with MOI = 1) was initially incubated with subneutralizing concentration of humanized 4G2 mAb (monoclonal antibody) against DENV E protein (0.1 μg/ml), constructed using a previously published protocol  for 30 min on ice in 1 ml of medium containing 2% FBS and the mixture was used to infect THP-1 cells. Cells were washed once after 1 h with PBS and replaced with 3 ml of fresh growth medium. THP-1 cells were collected by centrifugation at 0, 6, 12, 24 and 48 hpi for further analysis as described below.
The four dengue serotypes DENV-1–4 were isolated from patients recruited into the EDEN (early dengue infection and outcome) study in Singapore . This study was approved by the Domain Specific Review Board of the National Healthcare Group (DSRB/B/05/013). The strains were DENV-1–2402DK1, DENV-2–3295, DENV-3–863DK1 and DENV-4–2270DK1. The Indonesian viruses were collected as part of a DHF surveillance programme in Jakarta, Indonesia, conducted by Duane J. Gubler in collaboration with Dr P. Sumarmo (Cipto Mangunkusumo Hospital) and Dr H. Wulur (Sumber Waras Hospital) from 1975 to 1978. The latter viruses were isolated, stored in Aedes albopictus mosquitoes [32,33] and subsequently expanded in C6/36 cells (less than three passages for each isolate) utilizing RPMI 1640 medium containing 10% FBS (Invitrogen). Virus titres [pfu (plaque-forming units)/ml] were determined by a plaque assay on BHK-21 cells (baby-hamster kidney 21 cells) as previously described .
RNA extraction and real-time RT–PCR (reverse transcription–PCR) analysis
Total RNA was isolated using Trizol® (Invitrogen) from 1×105 cells infected as described above in the Materials and methods section. Reverse transcription was performed on 500 ng of RNA using the ImProm II reverse transcription system (Promega), and real-time PCR was carried out in iQ5 iCycler (Bio-Rad) using the primers listed in Table 1. Normalization was done using primers for 18S rRNA. Results are presented as a fold increase in logarithmic scale over the control (uninfected) sample at the same time point, normalized to 18S rRNA level. Statistical analysis was performed for 48 hpi time-point data using the paired Student's t test and results were considered statistically significant if P<0.05.
|Primer .||Sequence (5′→3′) .|
|Primer .||Sequence (5′→3′) .|
Cells (1×105; HEK-293 and THP-1 cells) were mock-infected with medium or infected with DENV for 48 h as previously above in the Materials and methods section. Total cell lysate was collected in lysis buffer containing 0.1% Triton X-100 with a protease inhibitor cocktail (Roche). After 30 min on ice, lysates were centrifuged at 10000 g for 10 min and the supernatant was used to quantify the amount of total protein by the BCA (bicinchoninic acid) assay (Pierce). Equal amounts of protein (20 μg per lane) were subjected to SDS/PAGE (10% gel), and Western blotting was performed using anti-BiP (Abcam), anti-phospho-eIF2-α and anti-eIF2-α (Cell Signaling Technology) antibodies. Chemiluminescence was observed using ImageQuant (Bio-Rad).
THP-1 cells (1×105) were infected with DENV for 24 h in the absence or presence of hAb (humanized antibody) 4G2 as described above in the Materials and methods section. Since THP-1 is a suspension cell line, in order to collect good images, the cells were subsequently plated on to poly-D-lysine-coated coverslips and allowed to adhere for a further 24 h. The cells were fixed with methanol and incubated with 4G2 antibody (recognizing DENV E protein) followed by secondary antibody (goat anti-mouse) labelled with Alexa Flour® 488. The coverslips were visualized under a fluorescence microscope in the presence of ProLong Gold antifade reagent with DAPI (4′,6-diamidino-2-phenylindole; Invitrogen). For quantitative analysis, relative fluorescence was measured from six separate fields using ImageJ software (National Institutes of Health, Bethesda, MD, U.S.A.) and the ratio of cells positive for dengue fluorescence to DAPI fluorescence was calculated.
Replication of DENV serotypes in HEK-293 and THP-1 cells
Previously, direct infection of A549 human lung carcinoma cells has been used to examine the UPR pathway induction . Here, we selected HEK-293 epithelial lineage cells  to test UPR induction during direct infection of DENV-1–4 serotypes from the EDEN prospective study in Singapore. In order to model the well-documented phenomenon of ADE infection, we used hAb 4G2 to facilitate the infection of THP-1, a human acute monocytic leukaemia cell line. Real-time RT–PCR results showed that all four dengue serotypes directly infected HEK-293 cells (Figure 1B) as well as THP-1 (Figure 1C) cells (in the ADE mode) equally as observed at 6 hpi. There were serotype-specific differences in replication at 48 hpi, with DENV-1 and -3 replicating to significantly higher levels (P<0.05) compared with DENV-2 and -4. As shown in previous studies [5,35,36], virus infection (MOI = 1) of THP-1 cells without antibody leads to a very low level of infection compared with infection in the presence of subneutralizing levels of antibody. The increased infectivity and replication in ADE-infected THP-1 cells was also confirmed by immunofluorescence in DENV-2 (strain 3295)-infected cells at 48 hpi using the mouse anti-E protein mAb, 4G2. The cells infected with the virus–antibody complex showed more cells with fluorescence (~40%) compared with cells infected with virus alone (3–5%; Figure 1D). Quantitative analysis using ImageJ showed that ADE-infected cells had a relative fluorescence of 0.46 compared with a fluorescence ratio of 0.13 for cells infected without antibody. Not surprisingly, the level of infection of HEK-293 cells was found to be higher compared with ADE infection of THP-1 cells; however, the serotype-specific differences were more prominent in the latter infection (P<0.05).
Induction of UPR by DENV-1–4 serotypes
To investigate the impact of the differential replication on the host cellular pathways, the genes involved in UPR were monitored by real-time RT–PCR at 0, 6, 12, 24 and 48 hpi. The results showed that, although all four serotypes induced the selected five UPR genes, there were differences in the induction, especially in the BiP mRNA level at 48 hpi. HEK-293 cells (Supplementary Figure S1A at http://www.bioscirep.org/bsr/031/bsr0310221add.htm) as well as THP-1 (ADE–infected; Figure 2A) cells infected with DENV-1 and -3 (>1000-fold in THP-1) showed significantly higher levels of BiP mRNA compared with cells infected with DENV-2 and -4 (~40-fold in THP-1; P<0.01). In ADE infection of THP-1 cells, serotype-specific differences were also seen in mRNA levels of GADD34 and CHOP (Figures 2B and 2C). HEK-293 cells showed a similar trend, although the differences were not statistically significant (Supplementary Figures S1B and S1C). Other investigated host genes [XBP1 (X-box-binding protein 1) and eIF2-α kinase] were induced during dengue infection but did not show any serotype-specific differences in either cell type (results not shown).
UPR induction by DENV serotypes in THP-1 cells
To confirm that UPR induction and serotype-specific differences in mRNA expression corresponded to the cellular protein levels, Western blotting was performed on both cell types infected with DENV-1–4 at 48 hpi. The results showed an increase in BiP protein level in infected cells; however, cells infected with DENV-1 and -3 showed a significantly higher level of BiP compared with cells infected with DENV-2 and -4 (Supplementary Figure S1D for HEK-293 cells and Supplementary Figure S2D at http://www.bioscirep.org/bsr/031/bsr0310221add.htm for ADE THP-1). Control cells showed a very low level of basal BiP protein expression. A similar serotype-specific increase was observed for phosphorylated eIF2-α, another marker for UPR induction, when probed with antibody specific to phosphorylated eIF2-α. There was a higher level of phosphorylation of eIF2-α in cells infected with DENV-1 and -3 compared with those infected with DENV-2 and -4 at 48 hpi. However, there was no significant difference in total amount of eIF2-α in cells infected with any serotype, as shown by Western blotting using anti-total eIF2-α antibody. The cells infected with virus alone (without hAb 4G2) showed no significant induction in any of the host mRNAs involved in the UPR pathway, due to low level of infection (results not shown) and were excluded from further studies.
Induction of UPR by DENV clinical isolates
Clinical isolates of strains from a single dengue epidemic were tested to see whether they would show differential UPR induction and if the genes selected in this study can be used to predict disease severity retrospectively. This was done using six low-passage strains of DENV-2 serotypes isolated from Indonesian patients in the 1975–1978 outbreak with known clinical outcome (blinded to us) and that have been fully sequenced . As described above, HEK-293 cells or THP-1 cells were infected directly with virus or the virus–hAb 4G2 complex respectively. Equal NS1 (non-structural 1 glycoprotein) levels at 6 hpi as measured by real-time RT–PCR for the clinical isolates indicated equal infectivity in HEK-293 cells (Supplementary Figure S2A) as well as ADE THP-1 cells (Figure 3A). At 48 hpi the NS1 level of some strains showed slightly higher replication compared with others. This difference was more significant in ADE-infected THP-1 cells (P<0.001) with strains 1172, 1183 and 1127 showing higher replication compared with other strains. In HEK-293 cells, host UPR gene profile showed induction in BiP, GADD34 and CHOP mRNAs after infection with the clinical strains. However, there were no consistent strain-specific variations in the induction of these genes (Supplementary Figures 2B–2D).
Induction of UPR by DENV clinical isolates in THP-1 cells
ADE infection of THP-1 cells with all six clinical isolates showed induction in UPR pathway genes. However, the three clinical isolates, which showed higher NS1 levels at 48 hpi (Figure 3A), induced BiP mRNA more than the other strains at the same time point (>1000-fold compared with 80–300-fold; Figure 3B). Similar results were obtained when real-time RT–PCR was performed for GADD34 and CHOP. Cells infected with clinical isolates 1172, 1183 and 1127 showed higher levels of GADD34 (>1000-fold) and CHOP (600–1200-fold) mRNAs at 48 hpi compared with baseline. However, cells infected with clinical isolates 1023, 1070 and 1017 showed lower level of GADD34 (50–100-fold) increase and CHOP (600–1200-fold) increase over their respective baselines (Figures 3C–3D). These results indicate that there are strain-specific differences in the replication as well as UPR pathway induction, which may account for the variability in dengue pathogenesis during a single epidemic. Interestingly, this difference is only evident during ADE infection of monocytic cells and not in the direct infection model in HEK-293 cells.
Dengue is an acute disease with a wide spectrum of clinical outcomes ranging from DF to DHF and DSS . The reason for this variability in disease presentation is not completely known and is possibly a complex interplay of virus, the host and the environment . In the present study, we have examined the viral factors that may lead to variable host response, using genes of the UPR pathway as markers.
Although peripheral blood leucocytes form the major cell type that is infected in dengue patients , THP-1 monocytic cells were poorly infected by direct primary infection with viruses. HEK-293 cells, on the other hand, could be infected robustly and were used to model the scenario of direct infection . Higher level of infection of THP-1 cells was achieved by using subneutralizing concentrations of humanized mouse mAb 4G2 (hAb 4G2). This infection scenario modelled an in vivo phenomenon called ADE infection [6,40]. The virus binds to the antibody, and is internalized via Fcγ receptor on the cell surface of THP-1 cells using the humanized Fc region of the antibody . Using real-time RT–PCR and immunofluorescence, a significant increase in the level of infection in THP-1 cells in the presence of antibody compared with virus alone was demonstrated; confirming previous reports of ADE infection in these cell types [7,35,36]. Notably, the hAb 4G2 was more robust and reproducible in ADE infection of THP-1 cells than the complex with the parent mouse mAb (results not shown).
Previous studies have shown that viral infection can induce UPR in host cells [21,42]. Umareddy et al.  showed that DENV infection increased the phosphorylation of eIF2-α in A549 cells in a time-dependent manner. The results also showed that there was a difference in the level of ER stress induced by DENV-1 compared with DENV-2. Based on this, we hypothesized that the variability associated with replication rates of different DENV strains in mammalian cells may be correlated with different levels of ER stress. The results of the present study also suggest that the four serotypes of dengue replicated differently in HEK-293 and THP-1 cells and this difference was mirrored in the host UPR gene response. The DENV-1 and -3 serotypes, which replicated faster in mammalian cells, produced a higher level of ER stress as measured by BiP, GADD34 and CHOP mRNA levels compared with DENV-2 and -4. It should be cautioned that it would be incorrect to conclude from our results that only DENV-1 and -3 serotypes lead to severe dengue since it is well known that all four serotypes can cause epidemic outbreaks and severe dengue, and few mutations in the viral genome are needed for significant changes in the disease phenotype . Also, a previous study where DENV-1 and DENV-2 replications were compared by intrathoracic inoculation of mosquitoes suggested that DENV-2 is mostly associated with severe dengue .
Extending the analysis to clinical isolates that were collected during an epidemic in Indonesia showed that different DENV-2 strains also showed different levels of replication in THP-1 cells when infected in the ADE mode, which is consistent with the wide spectrum of disease during dengue outbreak. The faster-replicating virus strains showed higher induction of BiP, GADD34 and CHOP mRNAs in host cells than the slower-replicating strains. These clinical strains were selected since the sequences and the clinical outcomes from these isolates were known (Table 2), although the experiments were performed blind to this information. Two of the Indonesian DENV-2 strains (1172 and 1127) that were fast replicators and produced higher ER stress were isolated from patients with grade IV DHF (Table 2). Strain 1127 was isolated from a fatal case and the clinical records showed that the patient suffered from haemorrhage and convulsions. The patient from whom strain 1172 was isolated produced hepatomegaly along with haemorrhage. Unfortunately, for the strain 1183, which also produced higher ER stress based on host mRNA expression, no clinical data were available. However, based on the results, it can be predicted that strain 1183 produced a severe form of dengue disease. The remaining strains (1070, 1017 and 1023), which induced less ER stress, correlated with the DF, a milder disease.
|Isolate no. .||Location .||Days post onset .||Outcome .||Other symptoms .|
|1172||East Jakarta||3||Grade IV||Hepatomegaly, epistaxis|
|1127||Central Jakarta||3||Grade IV||Death, convulsions|
|Isolate no. .||Location .||Days post onset .||Outcome .||Other symptoms .|
|1172||East Jakarta||3||Grade IV||Hepatomegaly, epistaxis|
|1127||Central Jakarta||3||Grade IV||Death, convulsions|
A recent report showed that silencing BiP, an ER chaperone protein, in host cells decreased the viral titre . The paper showed that BiP acted as a chaperone for viral proteins and co-immunoprecipitated with E protein. It was previously reported that decreasing BiP using SubAB toxin decreased the release of infectious viral particles . This supports our results that an increased replication induced higher BiP levels. Umareddy et al.  showed that decreasing the GADD34 level by Salubrinal also decreased viral titre, indicating that GADD34, which is a cofactor for eIF2-α phosphatase, was required for sustained viral growth. On the other hand, the CHOP activation indicated induction of the pro-apoptotic pathway, which suggested overwhelming viral replication. Accordingly, our results indicated that the strains isolated from mild DF patients showed a lower level of GADD34 expression, while the virus isolated from severe dengue patients showed a higher induction of the CHOP pathway.
To try and pinpoint the residues that may be involved, phylogenetic analysis of the complete genome sequence of the strains (GenBank® Nucleotide Sequence Database accession numbers GQ398261–398263, GQ398257, GQ398259 and GQ398260) was carried out (results not shown). The sequence comparison confirmed that all the strains belong to the DENV-2 serotype and, although several amino acid changes were noted, an unrooted phylogenetic tree analysis did not show clustering of the severe strains along the lines of the clinical records, unlike the clustering observed for the UPR gene biomarkers in THP-1 cells in the ADE mode of infection.
Together, the results suggest that the differential induction in ER stress may be explained by the genetic variability in the sequences of the viral strains from the DENV-2 clinical isolates from Indonesia, although no specific sequences could be identified from the small sample size in the present study. Most importantly, the results obtained from the present study showed that only in the ADE infection of THP-1 cells, and not in the direct infection of HEK-293 cells, could the virus strains from the mild and severe patients be distinguished. Interestingly, Halstead and co-workers have shown previously that the faster growth of DENV-2 isolates in human peripheral blood leucocytes correlated with severe dengue disease  when infection was carried out in the presence of subneutralizing levels of antibodies in serum. It was reported that, during ADE, DENV evades the immune system by activating specific pathways . A similar mechanism may also account for the difference observed in UPR genes during ADE compared with direct infection in HEK-293 cells. Further studies are in progress to determine whether different signalling events are activated during ADE infection compared with direct infection by virus.
To conclude, although the clinical sample size in this study was small, overall the results, for the first time, suggested that genes involved in UPR during ADE infection of cultured monocytes can be used as biomarkers for dengue severity, thus extending the previous observations of Morens et al. . The impact of these data in predicting dengue disease severity in the clinical setting by infecting THP-1 cells with patient serum samples and measuring the mRNA levels of BiP, GADD34 and CHOP has to be examined in prospective dengue studies.
activating transcription factor
immunoglobulin heavy-chain-binding protein
dengue haemorrhagic fever
dengue shock syndrome
eukaryotic initiation factor 2-α
fetal bovine serum
growth arrest DNA damage-inducible protein 34
multiplicity of infection
non-structural 1 glycoprotein
PKR (double-stranded-RNA-dependent protein kinase)-like endoplasmic reticulum kinase
unfolded protein response
Prasad Praradkar designed and performed the experiments, analysed the data and wrote the manuscript; Eng Eong Ooi designed the experiments and wrote the manuscript; Brendon Hanson prepared the humanized mAb against dengue protein E; Duane Gubler provided the clinical isolates and wrote the manuscript; and Subhash Vasudevan designed the experiments, analysed the data and wrote the manuscript.
We thank Dr Soman N. Abraham, Dr Alex Ward and Dr Nicole Moreland for critically reading the manuscript prior to submission.
This work was supported by DUKE-NUS Signature Research Program (funded by the Agency for Science, Technology and Research, Singapore and the Ministry of Health, Singapore) as a start-up grant to S.G.V.