Ethical approval for the study was obtained from the National Ethics Committee of Health Research of Cambodia. Written informed consent was obtained from all participants or the guardians of participants under 16 years of age prior to inclusion in the study.

Blood samples were obtained from hospitalized children (≥ 2 years) who presented with dengue-like symptoms at the Kanta Bopha Hospital in Phnom Penh, Cambodia. The time-point for collection of blood samples was within 96 h of fever onset at hospital admittance. Patients were classified according to the WHO 1997 criteria upon hospital discharge (30) as dengue fever (DF) or dengue haemorrhagic fever/dengue shock syndrome (DHF/DSS). In addition, age and sex-matched healthy donors were recruited from a cluster-based investigation in Kampong Cham province. Demographics and laboratory parameters of the patients included for each type of analysis as described below can be found in Table 1.

Plasma specimens obtained from patients were tested for presence of DENV using a nested RT-qPCR at the Institut Pasteur in Cambodia, the reference laboratory for arboviral diseases in Cambodia (31). NS1 and anti-DENV IgM/IgG positivity was determined using rapid diagnostic test (combo test for NS1 and IgM/IgG detection, SD Bioline Dengue Duo kits from Standard Diagnostics – Abbott, Chicago, IL, USA). Anti-DENV IgM was measured with an in-house IgM-capture ELISA (MAC-ELISA), as previously described (32). Samples from patients positive for DENV were further tested with hemagglutination inhibition assay (HIA) (33) at admittance and discharge to determine primary/secondary DENV infection as per WHO criteria (30).

Foci reduction neutralization test (FRNT) is used as the gold standard to determine the level of neutralizing antibodies against different viruses. Neutralizing antibody titers against the infecting DENV serotype were determined by FRNT assay using reference DENV reference strains: DENV-1 Hawaii (GenBank: AF425619), DENV2 New Guinea C (GenBank: AF038403), and DENV-4 H241 (GenBank: AY947539). The FRNT was performed on samples obtained at discharge, on average 3 days after hospital admittance in order to reach sufficient high titers for detection (Table 1). Briefly, Vero-CCL cells (ATCC CCL-81) cultured in Dulbecco’s modified Eagle medium (DMEM,; Sigma-Aldrich, St. Loius, MO, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Wattham, MA, USA) and seeded in 96-well plates. Heat-treated plasma samples were serially diluted, mixed with equal volume of DENV and incubated for 1 h at at 37°C. Afterwards, plasma-virus mixtures were transferred onto cells in the respective wells. After 1 h of incubation at 37°C, the mixture was replaced with 1.8% carboxymethyl cellulose (Sigma-Aldrich) and incubated at 37°C at 5% CO₂. At 2–3 days post infection, cells were fixed and stained as described (34) using DENV serotype specific polyclonal hyperimmune ascites fluids (Institut Pasteur du Cambodge). The titer of neutralizing antibodies was expressed as FRNT90, i.e. the plasma dilution at which a 90% reduction in the number of virus-induced foci is observed, and was calculated by regression analysis (GraphPad Software, Inc., La Jolla, CA, USA).

PBMCs were isolated from dengue positive patients and age-matched healthy donors using Ficoll-Histopaque (GE Healthcare, Chicago, IL, USA) density gradient centrifugation and stored in liquid nitrogen. The samples were thawed in RPMI supplemented with 10% FBS, washed with sterile PBS and then surface stained using the following antibodies: HLA-DR Alexa Fluor 488 (clone L243), Lineage cocktail APC [clone UCHT1 (CD3), clone HCD14 (CD14), clone HIB19 (CD19), clone 2H7 (CD20), clone HCD56 (CD56)], CD11c PE (clone 3.9), CD123 BV510 (clone 6H6) (all from BioLegend, San Diego, CA, USA). Samples were acquired on BD FACS Canto II (BD Biosciences, Franklin Lakes, NJ, USA) using FACSDiva software (BD Biosciences) and were analyzed by FlowJo v10.0 software (BioLegend, San Diego, CA, USA).

RNA extraction was done from PBMCs of DENV-positive patients and healthy controls by QIAGEN RNeasy Micro Kit (Qiagen, Hilden, Germany) as per manufacturer’s instructions. cDNA was synthetized from extracted RNA using SuperScript II Reverse Transciptase kit (Invitrogen, Carlsbad, CA, USA). Quantitative real-time PCR for IFN-α2, IFN-β, IFN-I receptor subunits (IFNAR1 and IFNAR2), IRF7, and FOXO3 was carried out with the LightCycler 480 instrument (Roche, Basel, Switzerland). Primers and probes for each gene were added to the Probes Master Mix (Roche, Basel, Switzerland) at 500 and 250 nM, respectively, in a final volume of 20 μl. The housekeeping gene β-glucuronidase was used as an internal control. Gene expression values were calculated by the comparative ΔΔCt method. The primers and probe were assayed on demand and were purchased from Integrated DNA Technologies (Coralville, IA, USA). The list of primers and probes is as follows: IFN-α2 (Hs.PT.58.24294810.g), IFN-β (Hs.PT.58.39481063.g), IFNAR1 (Hs.PT.58.20048943), IFNAR2 (Hs.PT.58.1621113), IRF7 (Hs.PT.58.24613215.G), and Foxo3 (Hs.PT.58.5045552.g).

IFNα and IFNβ protein plasma concentrations were quantified by Simoa digital ELISA developed with Quanterix Homebrew kits (Quanterix, Billerica, MA, USA) as previously described (35). For the IFNα assay, the 8H1 antibody clone was used as a capture antibody after coating on paramagnetic beads (0.3 mg/ml), the 12H5 clone was biotinylated (biotin/antibody ratio = 30/1) and used as the detector, and recombinant IFNα17 (Peprotech, Cranbury, NJ, USA) was used as the standard. For the IFNβ assay, the 710322-9 IgG1, kappa, mouse monoclonal antibody (PBL Assay Science, Piscataway, NJ, USA) was used as a capture antibody after coating paramagnetic beads (0.3 mg/ml), the 710323-9 IgG1, kappa, mouse monoclonal antibody (PBL Assay Science) was biotinylated (biotin/antibody ratio = 40/1) and used as the detector antibody, and recombinant protein (PBL Assay Science) was used to quantify IFNβ concentrations. The limit of detection (LOD) of the IFNα and IFNβ assays were 0.005 fg/ml and 0.05 pg/ml, respectively. Plasma IFNα was below the assay detection limit in one patient with DHF/DSS and was excluded from the analysis.

Statistical analyses were done using GraphPad Prism 7.00 software (GraphPad). Since the data included in the study did not pass the criteria for normality as determined using D’Agostino-Pearson normality test, non-parametric Mann-Whitney U test was used to compare data between two groups. Values were expressed as median and interquartile range. Correlations between groups which did not pass the criteria for normality as determined using D’Agostino-Pearson normality test were calculated by Spearman analysis. For all analyses, p values less than 0.05 were considered statistically significant.

In total, we included 115 dengue-positive paediatric patients admitted at Kantha Bopha Children’s hospital, Phnom Penh, during the acute phase of infection with an onset of symptoms less than 96 h before admission for IFN-I analysis and DC phenotyping (Supplementary Figure 1A). These patients were classified for immune history (primary/secondary DENV infection) and severity according to WHO 1997 criteria (30). Of these, 84 were classified as DF, 22 as DHF and 9 as DSS patients (Table 1). Of the DF patients, 15.2% were undergoing a primary DENV infection whereas 70.6% went through were undergoing a secondary infection, and immune status could not be determined in 14.2% of the cases based on the obtained HIA data (30). 77% of DHF and DSS patients encountered a secondary infection while infection status could not be determined for 23% of the patients (Table 1). Of the 107 patients included with measurable viremia, 31% were infected with DENV-1, 63% with DENV-2, and 7% with DENV-4. No DENV-3 infected individuals were included as DENV-3 was not present in circulation at time of patient recruitment. Viral load as determined by qRT-PCR was not significantly different between DF and DHF/DSS patients (Table 1, Supplementary Figure 1B). As viremia is dependent on immune status, infecting serotype, day of fever and possibly comorbidities, we observe a wide variability in viral load in children with DF and DHF/DSS (36, 37). As frequencies of immune cells and the immune response can vary from day to day during acute DENV infection (38), it is important to emphasize that not only were all patients recruited within 96 h of fever onset, there was no difference in day of fever at inclusion between DF and DHF/DSS patients (Table 1). In addition, 43 age and sex-matched healthy donors were recruited from a cluster-based investigation in Kampong Cham province (Table 1).

It remains controversial as to whether IFN-I responses are protective or contribute to immunopathogenesis during DENV infection (39). Detection of IFN-I responses are challenging since in healthy donors and during viral infections only trace amounts of IFN-I mRNA are present in PBMC and determination of protein concentrations in plasma by conventional methods has proven to be unreliable (40–44). We performed gene expression analysis of a set of IFN-I related genes [IFNα, IFNβ, IFN-I receptor subunits (IFNAR1 and IFNAR2), IRF7 and FOXO3] in PBMCs isolated from DENV-infected patients with variable disease severity and healthy donors. Increased transcription of IFNAR2 and IRF7 mRNA were observed in DENV positive patients (IFNAR2: p=0.0006; IRF7: p <0.0001, Figures 1A, B) compared to age and gender matched healthy donors. In parallel, we analyzed protein concentrations of IFNα and IFNβ in plasma from by ultra-sensitive single-molecule array (Simoa) digital ELISA technology in the same cohort (35). Here, we observed enhanced protein concentrations of IFNα and IFNβ in acute DENV-infected patients versus healthy donors (IFNα: p=<0.0001 and IFNβ: p=0.0138) (Figure 1C). Whereas IFNβ protein concentrations were correlated with IFNβ mRNA (r=0.35; p<0.01), no such correlation was observed for IFNα (r=0.10; p=0.42) during acute DENV infection (Figure 1D). DENV viral load positively correlated with levels of plasma IFNα/β (r=0.28; p<0.01 and r=0.50; p<0.0001) (Figure 2A), IFNAR1-mRNA (r=0.30; p<0.01), and IFNAR2-mRNA (r=0.31; p<0.01) (Figure 2B). Furthermore, when patients were stratified based on disease severity, DENV viral load correlated positively with plasma IFNα/β in patients with DF (r=0.28; p=0.02 and r=0.52; p<0.0001) and DHF/DSS (r=0.47; p=0.04 and r=0.57; p<0.01) (Figures 2C, D). Evaluating the different infecting serotypes, we observed that the IFNα and IFNβ plasma concentrations are similar across all DENV serotypes included in the cohort (DENV-1, DENV-2, DENV-4) (Supplementary Figure 2). Taken together, these data show that an elevated viral load is correlated to an increase in IFNα/β protein synthesis and downstream induction of expression of IFN-I related genes (IFNAR2, IRF7) during acute DENV infection.

pDC and virus-infected cells are the major sources of IFN-I. DENV has tropism for dendritic cells of the myeloid lineage in blood and therefore will contribute to the IFN-I response (25, 45–47). Hence, we aimed to evaluate if changes in circulating frequencies of mDCs and pDCs could account for the changes in IFN-I observed during acute DENV-infection. The gating strategy used for the identification of dendritic cell subsets is outlined in Supplementary Figure 3. A representative dot plot showing the gating strategy of CD11c⁺ mDCs and CD123⁺ pDCs in healthy donors and DENV patients is shown in Figure 3A. The percentages of CD11c⁺ mDCs were found to be significantly decreased in dengue patients compared to healthy controls (median: 5.4% versus 44.1%; p<0.0001) (Figure 3B) whereas no difference was observed in the relative proportions of pDCs between dengue patients and healthy donors (Figure 3C). There was no correlation between the percentages of CD11c⁺ mDCs or CD123⁺ pDCs respectively with viral load in all patients or patients stratified for disease severity (Supplementary Figure 4). However, of major interest, increased percentages of circulating CD123⁺ pDC positively correlated to increased IFNα plasma protein concentrations in DENV-infected patients (r=0.43; p<0.001) (Figure 3D) whereas CD11c⁺ mDC frequencies were not correlated with IFNα plasma concentrations (Figure 3E). No correlation was observed between CD11c⁺ mDC or CD123⁺ pDC frequencies and IFNβ (Figures 3F, G).

Heterologous secondary infection increases the risk of a more severe outcome after DENV infection, which may result in DHF or DSS (5, 6, 48). The exact mechanism remains unknown, but an unbalanced and excessive immune response to the infecting serotype seems to contribute to disease severity (49). Hence, we compared levels of IFN-I related genes, IFNα/β protein concentrations and DC subsets between primary infected DF patients (n=9) and secondary infected DF patients (n=49). Here, a decrease in FOXO3-mRNA levels in patients with secondary infection compared to primary infection was observed (p=0.015) whereas levels of other IFN-I related genes were similar in primary infected DF patients compared to secondary infected DF patients (Figure 4A and data for IFN-α2, IFN-β, IFNAR1, and IFNAR2 not shown). Individuals undergoing a secondary DF infection showed decreased IFNα protein concentrations compared to primary infected patients (p=0.016) (Figure 4B). As shown in Figure 4C, the percentages of circulating CD11c⁺ mDCs and CD123⁺ pDC were not different between primary and secondary DF. Taken together, these data show that the type I IFN response is modulated by the infection history of the patients.

Few and conflicting data are available on the IFN-I response in severe dengue cases (15–19). Thus, we compared mRNA levels of IFN-I genes in DENV infected children who developed classical DF (n=50) with patients classified as DHF or DSS (n=26). As we observed minor differences in IFN-I responses in primary versus secondary infected DF patients, we only included secondary infected patients for this analysis. Reduced expression of IFNα (p<0.001), IFNβ (p<0.0001), IRF-7 (p<0.0001), and FOXO3 (p<0.001) transcripts were observed in patients with severe dengue (DHF/DSS patients) than patients with non-severe manifestations (DF patients) (Figures 5A–C). In accordance to the gene expression data, plasma protein concentrations of IFNα (p=0.09) and IFNβ (p=0.013) were lower in DSS/DHF patients compared to DF patients (Figure 5D). A decrease in platelet counts is one of the hallmarks of severe DENV infection and is used in the WHO classification criteria (30). In accordance with the observation of a decreased IFN-I response in severe disease, we observed that lower platelet counts are correlated to lower IFNα protein plasma concentrations in DENV-infected patients (p<0.05, r = 0.27, Figure 5E). As frequencies of circulating pDCs are associated to IFNα protein concentration during DENV infection (Figure 3D), we examined the influence of dengue disease severity in patients with secondary infection on the distribution of dendritic cell subsets. A significant decrease in the relative proportions of pDCs (median: 17.2% vs. 9.1%; p<0.05), but not mDCs (Figure 5F) was observed in secondary DHF/DSS compared to secondary DF cases. Hence, decreased pDC frequencies and an associated decreased IFN-I response are observed in patients developing severe disease during secondary DENV infection, indicating that an early IFN-I response mediated by pDCs is beneficial for clinical outcome after DENV infection.

As IFN-I play an important role in the induction of the humoral immune response during infection or vaccination (50, 51), we sought to determine the relationship between type I IFN and development of anti-DENV antibodies in infected patients. As previous infection history has a major impact on the developing antibody titers during the acute phase of DENV infection (52), we included only secondary infected patients for this analysis. Functional neutralizing antibodies were measured by the focis reduction neutralization test (FRNT90) (34). Even though their presence does not always correlate with protection from severe disease (53), the assay remains the gold standard for measuring humoral protection during DENV infection. Antibody titers were measured at discharge in order to reach sufficient titers for detection. Of interest, lower concentrations of IFNα and IFNβ proteins correlated with higher total anti-dengue antibodies measured at hospital discharge by hemagglutinin inhibition (HI) assay (p<0.0001, r=-0.57 and p<0.0001, r =-0.54) (Figure 6A). In parallel, lower IFNα (r=-0.37; p<0.01), IFNβ (r =-0.34; p <0.01), IFNAR1 (r=-0.27; p<0.05), and IRF7 (r=-0.28; p<0.05) transcripts measured at hospital admittance correlated with higher FRNT90 titers at hospital discharge (Figure 6B, Supplementary Figures 5A, B). Of note, these correlations were independent of the amount of days between hospital admittance and discharge (Supplementary Figure 6). As type I IFN correlates with viral load (Figure 2) and is dependent on disease severity (Figure 5), we also analyzed the correlation of neutralizing antibodies to viral load. Here, neutralizing antibody titers did not correlate to viral load (Supplementary Figure 7) in DF or DHF/DSS patients. Taken together, we observed strong negative correlations between IFN-I response measured at hospital admittance and the development of anti-DENV antibody titers in our cohort of acute DENV-infected Cambodian children.