All experiments involving animals were performed in animal biological safety level 2 containment laboratories at the National Institute of Infectious Diseases (NIID) in Japan in accordance with the animal experimentation guidelines of the NIID. The protocols were approved by the Institutional Animal Care and Use Committee of the NIID (Nos. 115,064, 118,009, 121,003, and 121,005). Trained laboratory personnel anesthetized the mice via intraperitoneal injection of a mixture of medetomidine, midazolam, and butorphanol prior to viral injection.

The parental virus strains DV1-5 and DV3P12/08, which were derived from patients infected with DENV-1 or DENV-3 in Thailand (Pambudi et al., 2013), were propagated in mosquito C6/36 cells. For adaptation of DENV to mice, DV1-5 or DV3P12/08 was intravenously infected in IFN-α/βR KO mice (Figure 1A). Four to five days post infection (p.i.), the viruses were isolated from the spleen, liver, bone marrow, or serum. The C6/36 cells were inoculated with 10% homogenates of each organ in phosphate-buffered saline (PBS), and the culture supernatants were collected on days 5–7 in order to perform further inoculation to IFN-α/βR KO mice. DV1-5P7Sp treatment was repeated seven times in the spleen. DV1-5P3Serum, DV3P12/08P4Bm, and DV3P12/08P4Li were passaged in the serum, bone marrow, or liver three, four, and four times, respectively. Virus stocks were propagated in C6/36 cells, and culture supernatants were kept at −80°C until use. C6/36 cells were cultured in L-15 medium containing 10% fetal calf serum (FCS) and 0.3% BactoTM Tryptose Phosphate Broth (Becton Dickinson, Sparks Glencoe, MD, United States). Vero cells were cultured in Eagle’s minimum essential medium (Nacalai Tesque, Kyoto, Japan) supplemented with 10% FCS.

IFN-α/βRKO mice, lacking type I IFN receptors, IFN-α/β/γRKO mice, lacking both type I and type II IFN receptors (Phanthanawiboon et al., 2016), and LysM Cre⁺Ifnar^flox/flox mice (Pinto et al., 2015), kindly provided by Dr. Shresta, La Jolla Institute for Immunology, were bred and housed in ventilated cages and kept under specific pathogen-free conditions. Male and female mice aged 8 and 12 weeks were used in this study. Mice were anesthetized by intraperitoneal (i.p.) injection of medetomidine, midazolam, and butorphanol tartrate (final concentrations of 0.3 mg/kg, 4 mg/kg, and 5 mg/kg, respectively) and then challenged intraperitoneally or intravenously with appropriate units (FFU) of DENVs. Following inoculation, the mice were weighed and visually monitored at least once daily to score morbidity. Mice were euthanized with isoflurane at the time of sample collection when they became moribund or died because of difficulty eating or drinking.

Focus forming units (FFUs) were used to estimate infectivity. Supernatants collected from the cultures were serially 10-fold diluted in minimum essential medium (MEM) and transferred to Vero cell monolayers in 96-well microplates. After incubation for 2 h at 37°C, the infected cells were further incubated with 2% carboxyl-methylcellulose in 2% fetal bovine serum-MEM for 72 h. The infected cells were washed five times, fixed with 3.7% formaldehyde solution, and permeabilized by treatment with 0.1% Triton X-100 prior to incubation with an anti-E monoclonal antibody (mAb, D23-1G7C2) (Kurosu et al., 2020). After overnight incubation at 4°C, the infected cells were further incubated with peroxidase-conjugated AffiniPure Rabbit anti-human immunoglobulin G (H + L) (309-035-003, Jackson ImmunoResearch, PA, United States) for 45 min. The culture plates were washed with PBS five times. After washing with PBS, the infected cells were visualized by reacting with H₂O₂-diaminobenzidine (Sigma-Aldrich, St. Louis, MO, United States), and foci were counted using a light microscope.

Vascular leakage was examined by intravascular administration of Evans Blue (Sigma-Aldrich) as previously described (Phanthanawiboon et al., 2016). Briefly, Evans Blue (0.2 mL of a 0.5% solution in PBS) was injected intravenously into moribund mice on day 4 p.i. After 2 h, the mice were anesthetized (by i.p. injection of medetomidine, midazolam, and butorphanol tartrate at final concentrations of 0.3 mg/kg, 4 mg/kg, and 5 mg/kg, respectively), euthanized by exsanguination, and perfused extensively with PBS. Liver and small intestine samples were collected. Evans Blue was extracted from the organs by incubation in 1 mL of formamide (Sigma-Aldrich) for 24 h, followed by centrifugation at 3,000 × g for 10 min, and 150 μL of supernatant was collected. The concentration of Evans Blue in each organ was quantified by measuring the absorbance at 620 nm using a Corona Grating Microplate Reader SH-9000 (Corona Electric Co., Ltd.). The results were expressed as optical density per gram of tissue.

Mice were anesthetized and perfused with 10 mL of 10% phosphate-buffered formalin. The liver and small intestine were harvested and fixed. Fixed tissues were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E). For immunostaining, 4 μm thick tissue sections were deparaffinized and heated at 121°C for 15 min in pH6 citrate buffer as antigen retrieval. After washing with phosphate-buffered saline (PBS), endogenous peroxidase was quenched with 3% hydrogen peroxide in PBS. After blocking with 5% skim milk in PBS for 30 min, the tissue sections were incubated at 4°C overnight with rabbit polyclonal anti-CD11b antibody (EPR1344, Abcam) or rabbit anti-cleaved-caspase3 antibody (5E1E, Cell signaling technology). After washing with PBS, sections were incubated with Histofine Simple Stain mouse MAX PO (R) (Nichirei Bioscience). Positive signals were visualized using peroxidase-diaminobenzidine reaction, and the sections were counterstained with hematoxylin. The sections were then scanned using a virtual slide scanner Olympus VS200 and examined using the virtual slide viewer OlyVIA software. Positive cells in the captured images of the tissue section were counted using QuPath software (RRID: SCR_018257). Three photographs were analyzed for each section.

DV1-5, DV1-5P7Sp, DV3P12/08, and DV3P12/08P4Bm were obtained from the culture medium of C6/36 cells. Viral RNA (vRNA) was extracted using the High Pure Viral RNA Kit (Roche). Viral cDNA was synthesized using SuperScript III (Invitrogen) and specific primers (Table 1). The viral genome was amplified by PCR using PrimeSTAR MAX DNA polymerase (Takara) with a set of primers (Supplementary Table S2). Nucleotide sequencing analysis was carried out using the indicated primer sets (Supplementary Table S2), the BigDye Terminator v3.1 Cycle Sequencing kit, and a 3500xL Genetic Analyzer (ABI). Data were analyzed using the Sequencher 4.9 software (GeneCodes Corp., Ann Arbor, MI, United States). Multiple sequence alignments were performed using the ClustalW software. The GenBank accession numbers for DV1-5, DV1-5P7Sp, DV3P12/08, and DV3P12/08P4Bm are LC793494, LC793495, LC793496, and LC793497, respectively.

The thymus, liver, spleen, kidney, small intestine, large intestine, brain, bone marrow, and peritoneal exudate cells (PEC) were homogenized using a TissueLyser II (Qiagen). For cDNA synthesis, total RNA was isolated from tissue homogenates using the TRIzol reagent (Thermo Fisher Scientific). One-step real-time quantitative RT-PCR amplification with SYBR Green I was performed using a Light Cycler (Roche) and One Step SYBR PrimeScript RT-PCR Kit II (Takara). The final concentration of each PCR primer was 0.08 μM, and the concentration of total RNA was 8 μg/mL, with a reaction volume of 12.5 μL. The conditions for reverse transcription were as follows: 42°C for 5 min, followed by 95°C for 10 s. PCR amplification used 45 cycles of 95°C for 5 s, 55°C for 30 s, and 72°C for 30 s. The conditions for reverse transcription were 42°C for 5 min, followed by 95°C for 10 s. PCR amplification used 45 cycles of 95°C for 5 s, followed by 60°C for 20 s. To quantify RNA derived from organs, the amounts were normalized to that of total RNA from the corresponding organs from mock-infected mice. Data were analyzed using LightCycler 96 Software ver. 1.1.0.1320 (Roche). Primer sets and probes have been described previously (Kurosu et al., 2023b).

For quantification of viral RNA (vRNA) isolated from the tissue homogenate, total RNA was adjusted to 100 μg/mL for use in real-time PCR. RNA was quantified using a One Step SYBR PrimeScript RT-PCR Kit II (Takara) and the following dengue group-specific primers: DN-F, 5′-CAATATGCTGAAACGCGAGAGAAA-3′, and DN-R, 5′-CCCCATCTATTCAGAATCCCTGCT-3′ (Phanthanawiboon et al., 2016). The reaction conditions were as follows: 50°C for 30 min, 95°C for 15 min, and then 40 cycles of 95°C for 20 s, 55°C for 30 s, and 72°C for 30 s, followed by a melting curve analysis step. PCR was performed using a LightCycler 96 (Roche). The quantity of vRNA in the initial total RNA was determined by interpolation analysis from a standard curve generated from 10-fold serial dilutions of in vitro-transcribed DENV-2 R05-624 RNA prepared using the MEGAscript Kit (Ambion) (Phanthanawiboon et al., 2016).

Mice were injected intraperitoneally with 100 μg of a purified, functional grade anti-mouse TNF-α antibody (Ab) (XT3.11; Bio X Cell) or 100 μg of an isotype control Ab (MOPC-21; Bio X Cell), on the first day p.i.

The small intestine from the duodenum to the small intestine-cecum junction was removed on day 4 p.i. Mesenteric fat was removed, and the intestine was opened longitudinally, washed in PBS to remove fecal matter, and shaken for 20 min at 37°C in HBSS (WAKO) containing 5 mM EDTA. After removing epithelial cells and fat tissue, single cell suspensions were produced by mincing the tissues and incubating them for 1 h at 37°C (with agitation) with 4 mL RPMI1640 (SIGMA) containing 4% bovine serum albumin (BSA) (WAKO), 1 mg/mL collagenase type 2 (Worthington Biochemical Corporation), 1 mg/mL dispase II (WAKO), and 40 μg/mL DNase I (Roche). Digested tissues were filtered through a 70 μm filter and washed with RPMI1640. The resulting cells were pelleted and washed with 20 mL HBSS supplemented with 5 mM EDTA. The cells were resuspended in 5 mL 30% Percoll (GE Healthcare), overlaid onto 4 mL 80% Percoll, and centrifuged at 1,200 × g for 30 min. Isolated lamina propria cells were collected from the Percoll gradient interface and washed with RPMI1640. For FACS analysis, cells were resuspended in FACS buffer (HBSS supplemented with 0.5% BSA), incubated at 4°C for 10 min with TruStain FcX anti-mouse CD16/32 antibody (BioLegend), and then stained at 4°C for 30 min with the following fluorochrome-conjugated antibodies: CD45 (30-F11, BD), CD11b (M1/70, BioLegend), and Ly6G (1A8, BioLegend). Stained cells were analyzed using a FACSLyric cytometer (BD Biosciences), and data were analyzed using FlowJo software (BD Biosciences).

In all the bar graphs and scatter plots, data are expressed as mean ± standard error of the mean (SEM). All data were analyzed using the GraphPad Prism software (GraphPad, San Diego, CA, United States). Statistical analysis was performed using one-way ANOVA, and the significance of differences was assessed using Tukey’s multiple comparison test for multiple comparisons. Comparisons between two groups were performed using two-tailed student’s t-test. For the graphs of IL-6 mRNA, MMP-8 mRNA, MMP-3 mRNA, and vRNA, log-transformed data were used for statistical analysis. Survival data were evaluated using the log-rank (Mantel–Cox) test. A p-value of <0.05 was considered statistically significant; ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, and ^****p < 0.0001. Details of the sample sizes are included in the figure legends.

We have previously reported that DV3P12/08 clinical isolate causes lethal infection with severe dengue symptoms in IFN-α/β/γR KO mice (Phanthanawiboon et al., 2016; Kurosu et al., 2023b). Some, but not all, DENVs are lethal in IFN-α/β/γR KO mice, some of which show symptoms of severe dengue, such as vascular leakage (Phanthanawiboon et al., 2016). To search for virulent DENV in mice, we examined lethality of several DENV-1s in IFN-α/β/γR KO mice and found that DV1-5 clinical isolate caused acute death (Supplementary Table S1). Besides, IFN-α/β/γR KO mice infected with DV1-5 showed clear gross changes in the liver and intestine, i.e., color fade of the liver and small intestine swelling (data not shown), similar to DV3P12/08-infected IFN-α/β/γR KO mice (Phanthanawiboon et al., 2016; Kurosu et al., 2023b). We then examined if DV1–5 or DV3P12/08 causes lethal infection in IFN-α/βR KO mice singly deficient in the IFN-α/β receptor. However, neither original DV1–5 nor DV3P12/08 caused lethal infection in IFN-α/βR KO mice (data not shown). Therefore, in anticipation of viral adaptation to IFN-α/βR KO mice, we passaged both viruses in IFN-α/βR KO mice, respectively, by targeting specific organs (Figure 1A). The DV1-5P7Sp, passaged seven times in the spleen of IFN-α/βR KO mice, demonstrated lethality (Table 1). In contrast, DV1-5P3Ser, which was passaged in the serum of IFN-α/βR KO mice, showed no lethality. DV3P12/08P4Bm and DV3P12/08P4Liver, passaged four times in the bone marrow and liver of IFN-α/βR KO mice, respectively, demonstrated lethality. DV1-5P7Sp and DV3P12/08 were selected for further analyses. We then challenged these viruses in LysM Cre⁺Ifnar^flox/flox mice. Intravenous infection with parental DV1–5 or parental DV3P12/08 did not result in mortality in LysM Cre⁺Ifnar^flox/flox mice (Figures 1B,C). However, intravascular infection with even low amounts of DV1-5P7Sp (1.0 × 10⁴ FFU/mouse) demonstrated 100% lethality in LysM Cre⁺Ifnar^flox/flox mice (Figure 1D), and that of DV3P12/08P4Bm caused lethal infection (Figure 1E). Intraperitoneal infection of LysM Cre⁺Ifnar^flox/flox mice with DV1-5P7Sp or DV3P4Bm (3.6 × 10⁷ FFU/mouse) caused lethal infection, while their parental viruses did not (Figures 1D,E).

In a previous model of DV3P12/08-infected IFN-α/β/γR KO mice, we observed clear gross changes in the liver and small intestine, i.e., color fade of the liver and swollen of the small intestine (Kurosu et al., 2023b). LysM Cre⁺Ifnar^flox/flox mice infected with DV1-5P7Sp or DV3P12/08P4Bm showed a faded liver color on day 4 p.i. with faded DV1-5P7Sp or with DV3P12/08P4Bm (Figures 2A–C). The intestines of the mice infected with both viruses individually were swollen (Figures 2B,C) and shortened (Figure 2D). These features suggest possible pathological damage and vascular leakage in these organs. Vascular leakage was evaluated by injecting Evans Blue. Infection with both viruses individually increased vascular leakage in both the liver and small intestine (Figures 2E–I). DV1-5P7Sp appeared to induce more severe vascular leakage in the liver than DV3P12/08P4Bm (Figure 2H).

Next, we performed histological analysis of H&E-stained sections from the organs of LysM Cre⁺Ifnar^flox/flox mice. The livers of mice infected with DV1-5P7Sp and DV3P12/08P4Bm showed edematous changes and swollen hepatocytes (Figure 3A). In the small intestines of mice infected with both viruses, edematous changes and infiltration of inflammatory cells were observed. The number of infiltrating CD11b-positive cells was increased by infection (Figure 3B). The spleens of mice infected with both viruses showed apoptotic lymphocytes with a marked decrease in lymphocytes in the white pulp. The increased number of apoptotic cells was examined by caspase-3 staining (Figure 3C). The above observations were quite similar to those in the previous DV3P12/08-infected IFN-α/β/γR KO mouse model (Phanthanawiboon et al., 2016). In the thymus, there were many apoptotic cells, tingible body macrophages, and a massive loss of cortical lymphocytes, suggesting damage to the thymus. The large intestines of mice infected with both viruses showed apoptotic epithelial cells and infiltrated immune cells. The bone marrow showed a drastic reduction in hematopoietic cells, suggesting bone marrow suppression, which has been observed in dengue patients (La Russa and Innis, 1995).

The next question was whether the adapted viruses could improve viral production in mice. Virus titers in each organ of LysM Cre⁺Ifnar^flox/flox mice infected with adapted viruses were compared with those infected with their parental viruses. DV1-5P7p increased virus production in the thymus (2.9 × 10⁵ vs. 1.2 × 10³ FFU/mL), small intestine (2.2 × 10⁴ vs. 3.2 × 10³ FFU/mL), large intestine (5.5 × 10⁴ vs. 5.5 × 10³ FFU/mL), bone marrow (5.2 × 10⁴ vs. 3.9 × 10² FFU/mL) and PEC (3.5 × 10³ vs. 1.0 × 10¹ FFU/mL) (Figure 4A). DV1-5P7Sp also significantly increased the serum virus titer (4.3 × 10⁴ vs. 3.1 × 10⁰ FFU/mL). Contrarily, DV3P12/08P4Bm showed higher virus production only in the thymus (6.7 × 10⁴ vs. 1.5 × 10⁴ FFU/mL) and bone marrow (3.8 × 10⁴ vs. 3.0 × 10³ FFU/mL) (Figure 4B). Both viruses grew well in the thymus and bone marrow, and DV1-5P7Sp appeared to increase viral production more efficiently than DV3P12/08P4Bm.

The replication of DV1-5P7Sp and DV3P12/08P4Bm was compared to that of their parental viruses in Vero (African green monkey kidney) cells. DV1-5P7Sp showed higher virus production (2.1 × 10⁵ FFU/mL) at day 1 p.i. and reached maximum virus production (8.3 × 10⁷ FFU/mL) at day 3 p.i. (Figure 5A), whereas DV1–5 showed no detectable virus production at day 1 p.i. and lower production at day 3 p.i. (7.3 × 10⁵ FFU/mL). At 5 and 7 days p.i., the virus titers of both viruses were similar (1.4 × 10⁷ and 4.3 × 10⁷ FFU/mL at day 5 p.i., 4.0 × 10⁷ and 4.5 × 10⁷ FFU/mL at day 7 p.i.). On the other hand, DV3P12/08P4Bm and DV3P12/08 started increasing from day 1 p.i. (5.3 × 10² and 9.2 × 10¹ FFU/mL, respectively) and reached peak titers at day 5 p.i. (1.9 × 10⁷ and 3.0 × 10⁶ FFU/mL, respectively) (Figure 5B). DV3P12/08P4Bm increased in a manner similar to that of DV3P12/08, although the DV3P12/08P4Bm titer was slightly higher than that of DV3P12/08 at day 5 p.i. (2.1 × 10⁶ vs. 2.6 × 10⁵ FFU/mL, respectively). Collectively, DV1-5P7Sp showed faster replication ability, even in Vero cells.

Sequence analysis was performed to determine substitutions in the adapted viruses. DV1-5P7Sp had 14 nucleotide substitutions in the 5′NTR, envelope, NS1, NS3, NS4A, NS4B, NS5, and 3′NTR genes, and 5 amino acid substitutions in E, NS1, NS4A, and NS5 (Figure 5C). Some substitutions included two types of nucleotides, indicating quasispecies. In contrast, DV3P12/08P4Bm had eight nucleotide substitutions in the E, NS1, NS2A, NS3, and NS5 genes and 7 amino acid substitutions in E, NS2A, NS3 and NS5 (Figure 5C).

In previous mouse models for severe dengue, blockade of TNF-α protected mice from lethal infection (Atrasheuskaya et al., 2003; Shresta et al., 2006; Phanthanawiboon et al., 2016; Kurosu et al., 2023b), suggesting a central role in lethality. When LysM Cre⁺Ifnar^flox/flox mice were intraperitoneally infected, anti-TNF-α Ab treatment protected all mice from lethal infection with DV1-5P7Sp (Figure 6A). On the other hand, anti-TNF-α Ab treatment completely protected LysM Cre⁺Ifnar^flox/flox mice from intraperitoneal lethal infection with DV3P12/08P4Bm (Figure 6B). However, anti-TNF-α Ab only partially protected LysM Cre⁺Ifnar^flox/flox mice from DV1–5 upon its intravenous injection (Figure 6C). Contrarily, anti-TNF-α Ab completely protected LysM Cre⁺Ifnar^flox/flox mice even when DV3P12/08P4Bm intravenously injected (Figure 6D). Although the routes of infection affected protective efficacy of anti-TNF-α Ab treatment, these results suggest that the TNF-α signaling pathway is strongly implicated in the pathogenesis of these models.

Vascular leakage, driven by a cytokine storm, occurs in two steps: a cytokine event and an effector event. TNF-α is an upstream regulator of many cytokines and is a key regulator in this model (Figures 6A–D). We have previously reported very high levels of IL-6 induction, which is considered a key cytokine linking both cytokine and effector events, leading to vascular leakage (Kurosu et al., 2023b). Therefore, we measured IL-6 mRNA expression in the liver and small intestine using qRT-PCR. Both DV1-5P7Sp and DV3P12/08P4Bm infections increased IL-6 mRNA in the liver (8.9 × 10 or 1.4-fold) and small intestine (2.8 × 10² or 4.5× 10-fold) (Figures 7A,B). The increase of IL-6 mRNA in the small intestine of LysM Cre⁺Ifnar^flox/flox mice infected with DV1-5P7Sp or DV3P12/08P4Bm was much higher than that in the liver. Compared to the increase in IL-6 mRNA by DV3P12/08P4Bm, that by DV1-5P7Sp was higher in both the liver and small intestine. We also observed higher induction of matrix metalloprotease-8 and -3, which are thought to be important effectors responsible for vascular leakage (Kurosu et al., 2023b). Infection with DV1-5P7Sp or DV3P4Bm increased MMP-8 mRNA (7.2 × 10³ or 1.3 × 10⁴-fold) in the liver (6.5 × 10² or 3.0 × 10²-fold) and small intestine (Figures 7C,D). Furthermore, these infections increased MMP-3 mRNA levels in the liver (6.0 or 7.1-fold) and small intestine (3.1 × 10² or 9.1× 10-fold), although only subtle increases were observed in the liver (Figures 7E,F). These results suggest that the current model is similar to the previous model in terms of downstream events following a cytokine storm.

MMP-8 was suspected to be the most important effector that leads to vascular leakage because it was found to be increased commonly in the liver and small intestine of DV3P12/08-infected IFN-α/β/γR KO mice (Kurosu et al., 2023a). Neutrophils are considered a major source of MMP-8. We also previously reported the high level of neutrophil infiltration in the small intestine of DV3P12/08-infected IFN-α/β/γR KO mice and the corresponding value for pathogenesis (Kurosu et al., 2023b). When we examined the infiltrating cells in the small intestine of LysM Cre⁺Ifnar^flox/flox mice infected with DV1-5P7Sp or DV3P12/08, we observed significantly increased neutrophil numbers in the small intestine (Figure 8). We confirmed the involvement of neutrophils in the pathogenesis of this disease.