Disease progression in mice infected with DENV-3 P12/08
IFN-α/β/γRKO mice infected with type 3 dengue virus (DENV-3) P12/08 died at Day 4 post-infection (p.i.). To observe disease progression, mice were dissected at Day 2, 3, and 4 p.i. (Fig. 1a). On Day 3 p.i., both the liver and small intestine started showing mild gross changes, which became more evident on Day 4 p.i. (Fig. 1b). The color of the liver faded to white, and the small intestine was swollen. At Day 4 p.i., mice had diarrhea (data not shown). These observations suggest that pathological changes started to occur between Days 3 and 4, and progressed quickly. Next, we examined vascular leakage by injecting Evans blue. Vascular permeability in both the liver and small intestine increased suddenly at Day 4 p.i. (Fig. 1c, d). Of note, the infection triggered intestinal dilation and shortened the length of the small intestine.
Treatment with an anti-TNF-α antibody
Several studies report protection from lethal infection after treatment with an anti-TNF-α Ab26, 27, 28, 29. The difference between our model and their models was the virus serotype and the genetic background of the mice. In our model, we used a DENV-3 P12/08 clinical isolate, whereas the others used a DENV-2 S221 mouse adapted strain. In addition, our IFN-α/β/γR KO mice have a C57/BL6 genetic background, whereas their AG129 mice have a 129 genetic background. We confirmed that the anti-TNF-α Ab protected all IFN-α/β/γR KO mice from acute lethal infection (Fig. 2b), and reduced body weight loss (Fig. 2c). Anti-TNF-α Ab treatment also prevented vascular leakage both in the liver and small intestine at Day 4 p.i., although the preventive effect in the liver was less marked (Fig. 2d, e).
Pathway analysis
To define the gene expression signatures related to severe dengue, and the effect of anti-TNF-α Ab treatment, we used a mouse microarray to profile global gene expression in the liver and small intestine in five groups of mice (Groups A to E)(DENV-3 P12/08-infected mice with normal IgG treatment collected on Day 3 (Group A) and 4 p.i. (Group B), DENV-3 P12/08-infected mice with anti-TNF-α Ab treatment collected on Day 3 (Group C) and 4 p.i. (Group D), and mock-infected mice (Group E) (Fig. 2f). Differentially expressed genes (DEGs) were defined as those showing a ≥2-fold change in expression in an unpaired Student’s t-test (p ≤0.01 for differential expression analysis). We first performed an exploratory study using principal component analysis (PCA). In the liver, samples were grouped according to the day of infection (Group A and C = Day 3 p.i.; Group B and D = Day 4 p.i.), although they were still segregated according to anti-TNF Ab treatment (Fig. 2g). In the small intestine, infection was a major cause of variable gene expression (PC1 = 28%). Samples from infected mice collected at Day 4. p.i. (Group B) were grouped separately (Fig. 2h), suggesting that induction of significant changes in gene expression at Day 4 p.i. were efficiently repressed by anti-TNF-α Ab treatment. To ascertain the nature of the DEGs, we performed Gene Set Enrichment Analysis (GSEA) based on Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets. In the liver, the number of most differentially expressed genes involved in inflammatory responses, including those encoding cytokines, was highest at Day 3 p.i.; however, anti-TNF-α Ab treatment inhibited these transcriptional changes (Supplementary Fig. 1). In the small intestine, the top ranked pathways were “cytokine-cytokine receptor interaction” and “viral protein interactions with cytokine and cytokine receptors”, indicating major involvement of cytokines in disease progression. Although DEGs in mice treated with the anti-TNF-α Ab were still enriched in these pathways, their numbers were smaller. Interestingly, the IL-17 signaling pathway played a critical role in the small intestine. Similarly, anti-TNF-α Ab treatment reduced the number of DEGs in the IL-17 signaling pathway. Next, we generated a heatmap of the differential expression of genes in the cytokine-cytokine receptor interaction pathway (Supplementary Fig. 2). In the liver of infected mice, expression of most genes was highest on Day 3 p.i. By contrast, expression of these genes in the small intestine of infected mice was highest on Day 4 p.i., and expression fell after exposure to the anti-TNF-α Ab. This suggests that infection induces stronger cytokine production in the small intestine than in the liver, and that anti-TNF-α Ab treatment suppresses gene expression more effectively in the small intestine at Day 4.
Next, we used Ingenuity Upstream Regulator Analysis to identify regulators responsible for the inflammatory and immune responses, and found high score of a set of upstream cytokine genes in both organs on Days 3 and 4 (i.e., il1β, tnf, ifn-γ, csf2, il6, and il17a; p-value of overlap = < 10-15) (Figure 3a, b, Supplementary Table 1). Anti-TNF-α Ab treatment reduced the -log (p-value) for most regulators in the liver on Day 3 p.i., but not on Day 4 p.i. However, in the small intestine, anti-TNF-α Ab treatment reduced the -log (p-value) of these cytokines markedly at Day 4 p.i. (Figure 3a, b, Supplementary Table 1). These results indicate that these cytokines play a major role in severe dengue.
We next analyzed peptidases, which are candidate molecules causing vascular leakage. The neutrophil collagenase MMP-8 was ranked in the top 10 genes in the small intestine showing upregulated expression at Day 4 p.i. (Supplementary Table 2). Therefore, we first analyzed peptidases by predicting fold changes in expression using IPA. ClustVis showed the expression of MMP-8 in the liver and small intestine of infected mice at Day 4 p.i. was most highly increased by infection at Day 4 p.i, and that it was suppressed markedly by anti-TNF-α Ab treatment (Fig. 3c, d). In addition, expression of MMP-7 or MMP-3 in the liver or small intestine, respectively, increased upon infection; however, this was suppressed by anti-TNF-α Ab treatment. Thus, these MMPs are candidate effector molecules that trigger vascular leakage.
Time course analysis of host gene expression
To validate the microarray observations, we measured expression of the above-mentioned genes by quantitative RT-PCR. Although both TNF-α and IL-6 mRNA showed maximum increases at Day 4 p.i. (Fig. 4a–d), the fold change in IL-6 expression was much higher, especially in the small intestine. By contrast, increased expression of IL-17A was observed only in the small intestine (Fig. 4e, f). Anti-TNF-α Ab treatment suppressed expression of TNF-α and IL-17A mRNA in both organs on Day 4 p.i., and suppressed expression of IL-6 mRNA in the liver. Strong suppression of IL-6 and IL-17A mRNA in the small intestine by anti-TNF-α Ab treatment (a 91.8% and a 87% reduction, respectively) suggested that they play important roles in severe disease. At the effector level, we observed that fold changes in expression of MMP-8, -3, and -7 mRNA were in accordance with the microarray results (Fig. 4i–l). Anti-TNF-α Ab treatment reduced expression of MMP-8 mRNA in the liver by 79.3% and in the small intestine by 97.2% at Day 4 p.i. Clear induction of MMP-7 or MMP-3 mRNA was observed only in the liver (Fig. 3i) or small intestine (Fig. 4l), respectively, although anti-TNF-α Ab treatment suppressed induction of MMP-3 mRNA only, suggesting strong involvement in severe dengue. We also confirmed that anti-TNF-α Ab treatment did not change expression of viral RNA (vRNA) in the liver and spleen, although vRNA levels fell slightly in the small intestine at Day 4 p.i. (Fig. 4m–o). Thus, protection by the anti-TNF-α Ab is likely due to suppression of immune responses.
Cytokine and chemokine levels in blood
Circulating levels of cytokines, chemokines, MMPs, and related factors in sera collected at Day 4 p.i. were measured in a multiplex Luminex Assay. Levels of TNF-α, IL-6, and IL-17A increased significantly at Day 4 p.i. (Supplementary Fig. 3a–c); however, levels were reduced significantly by anti-TNF-α Ab treatment. Similarly, infection with DENV-3 P12/08 increased MMP-8 levels, which were suppressed by anti-TNF-α Ab treatment (Supplementary Fig. 3d). There was no increase in expression of MMP-7 or MMP-9 after infection (data not shown). Furthermore, we examined levels of other cytokines, chemokines, and factors related to stability of the endothelium. Levels of IL-12p70, MCP-1, IL-10, IL-1β, IFN-γ, IL-2, M-CSF, VEGF, ICAM-1, and P-selectin increased after infection; however, expression of all but MCP-1, IFN-γ, and P-selectin was strongly suppressed by anti-TNF-α Ab treatment (Supplementary Fig. 3e–o). Expression of syndecan-1 increased slightly after infection (Supplementary Fig. 3p). This suggests that levels of most of the inflammatory cytokines, anti-inflammatory cytokines, chemokines, and soluble adhesion molecules increased after infection, and that the effect of anti-TNF-α Ab treatment was limited.
Blockade of IL-6 and IL-17A signaling
The above results suggest involvement of IL-6 or IL-17A in severe dengue. Therefore, to block IL-6 signaling, we treated infected mice with an anti-IL-6 receptor (R) Ab, MR16-1; however, this had no protective effect (Supplementary Fig. 4a, b). We wondered whether this was because the antibody did not suppress IL-6 signaling effectively due to the extremely high levels of IL-6 in serum. Levels of serum albumin A (SAA), which reflect the activity of IL-633, were very high, even after anti-IL-6R Ab treatment (Supplementary Fig. 4c).
Next, we examined blockade of IL-17A signaling. Treatment with an anti-IL-17A Ab protected 50% of mice from lethal infection (Figure5a), with no significant difference in body weight loss compared with IgG-treated infected-mice (Fig. 5b). In addition, we found that the anti-IL-17A Ab reduced vascular leakage significantly (Fig. 5c, d). To better understand the protective effects induced by anti-IL-17A Ab treatment, we examined levels of inflammatory cytokines in serum at Day 4 p.i. The α-IL-17A Ab reduced levels of TNF-α and IL-6 to those in mice treated with the anti-TNF-α Ab (Fig. 5e, f), although IL-6 levels were still higher than those in mock-infected mice. Unexpectedly, anti-IL-17A Ab treatment increased IL-17A levels (Fig. 5g). In addition, the α-IL-17A Ab suppressed IL-1β (Fig. 5h) and IL-12p70 (Fig. 5j) completely, but not IL-10 or IFN-γ (Fig. 5i, k). The α-IL-17A Ab also suppressed MMP-8 (Fig. 5l), but not MMP-3 (Fig. 5m), suggesting that MMP-3 plays less of a role in severe dengue. The α-IL-17A Ab did not affect vRNA levels in the liver, small intestine, and spleen (Supplementary Fig. 5a–c).
Next, we measured mRNA encoding major cytokines and effector molecules. Unexpectedly, the increased TNF-α mRNA levels in the liver and small intestine were not suppressed by the anti-IL-17A Ab (Fig. 5n, o), whereas IL-6 mRNA levels were suppressed in both organs (Fig. 5p, q). IL-17A mRNA was suppressed in the small intestine to some degree (Fig. 5r, s), although serum levels of IL-17A increased (Fig. 5g). Expression of MMP-8 mRNA was suppressed in the small intestine but not in the liver (Fig. 5t, u). The inconsistent results between serum levels of TNF-α and expression of mRNA suggest the existence of a major TNF-α producer in other organs, and that TNF-α production by these producers was efficiently suppressed by the α-IL-17A Ab.
IL-17A production from γδT cells in the small intestine and VγTCR usage
We suspected that, due to the speed of cytokine production, γδT, CD8+ T, or ILC3 cells were the likely producers of IL-17A (Fig. 4f). Therefore, to identify IL-17A-producing cells, we isolated cells from the small intestine and analyzed them by examining a plot of side scatter (SSC) versus CD45 (Fig. 6a); this plot showed an increase in the CD45+ cell populations after infection (Fig. 6a). Gated CD45+ leukocytes were further analyzed by measuring CD3 versus γδTCR signals. Three major cell populations were observed in infected mice: CD3+/γδTCR+ cells, CD3+/γδTCR− cells, and CD3−/γδTCR− cells (Fig. 6b). Among these, only CD3+/γδTCR+ cells, regarded as γδT cells, expressed IL-17A (Fig. 6c); the CD3+/γδTCR+ cells (including CD8+ T cells) (Fig. 6d) and CD3−/γδTCR− cells (including ILC3 cells) did not (Fig. 6e). The total number of γδT cells in the small intestine increased after infection (Fig. 6f), as did the number of IL-17A-producing γδT cells (Fig. 6g). The γδT cells were the major IL-17A producers in this model.
Therefore, γδT cells were further divided into subsets based on their expression of certain T cell receptor (TCR) Vγ-chains. The Vγ gene of the mouse γδTCR is encoded by six genes (Vγ1, Vγ2, Vγ4, Vγ5, Vγ6, and Vγ7) 34, and there is a strong correlation between γδ gene usage and tissue localization 35, and the type of cytokine produced36, 37, 38. γδT cells were isolated from the liver, small intestine, thymus, and spleen, and the Vγ genes were amplified by RT-PCR and analyzed by Southern-blotting. Small intestinal γδ T cells from mock-infected mice mainly expressed Vγ1/2, 2 genes, whereas those from infected mice expressed Vγ1/2, 2, 4, and 6 genes (Fig. 6h). Thymic γδ T cells from mock-infected mice expressed all Vγ genes, whereas those from infected mice expressed Vγ1/2, 2, 4, and 6, although expression of Vγ1/2 was weak. The expression pattern of Vγ genes in the thymus of infected mice is quite similar to that of intestinal γδ T cells. Collectively, the data suggest that infection predominantly stimulates polyclonal expansion of Vγ2, 4, and 6 γδT cells in the intestine and thymus.
Doxycycline prolongs survival
Our findings also indicate the importance of MMPs, especially MMP-8, at the effector level. Therefore, we asked whether inhibiting MMPs protects mice from lethal infection. Infected mice were injected intraperitoneally with doxycycline, a broad-spectrum MMP inhibitor, and survival was observed (Fig. 7a). All infected mice treated with PBS died at Day 4 p.i., whereas75% of mice treated with doxycycline survived for 2 more days (Fig. 7b). Although infected mice treated with doxycycline lost body weight (Fig. 7c), they showed no symptoms such as hunched posture with ruffled fur, or diarrhea at Day 4 p.i. (data not shown). Besides, doxycycline treatment suppressed vascular leakage at Day 4 p.i. (Fig. 7d, e). There is a question of whether the protective properties of doxycycline are due to its antibacterial effects; however, administration of a cocktail of antibiotics 2 weeks before challenge did not improve survival (data not shown). Furthermore, doxycycline did not alter expression of MMP-8 mRNA (Fig. 7f, g) or IL-6 (Fig. 7h, i) in either organ, suggesting that the protective effect was due to inhibition to downstream events in these pathways. We also confirmed that doxycycline did not alter viral production in the liver, small intestine, or serum (Supplementary Fig. 6).
MMP-8 is produced primarily by neutrophils39. Therefore, we examined whether neutrophils infiltrated the small intestine. The number of CD11b+Ly6G+ neutrophils increased significantly after infection (Supplementary Fig. 7f). These observations are consistent with the increased levels of MMP-8 after infection (Fig. 3d, 4h, 5l). In addition, the CD11b+Ly6GintLyF4/80+Ly6Chigh cells, which are activated by inflammation40, and CD11b+Ly6GintF4/80+Ly6Clow monocyte/macrophage subsets increased (Supplementary Fig. 7g, h).
Histopathological changes in infected mice treated with an anti-TNF-α Ab, an α-IL-17A Ab, or doxycycline
Histopathological changes in the liver and small intestine at Day 4 p.i. were observed by hematoxylin eosin (H&E) staining of tissue sections to examine the effect of treatment. Notable hydropic vacuolar degeneration of hepatocytes was observed in infected mice treated with normal IgG (Fig. 8a), as previously reported29. There was no clear restoration of hepatocytes in IFN-α/β/γRKO mice treated with the anti-TNF-α Ab or α-IL-17A Ab (Fig. 8a). Interestingly, almost no damage was observed in the liver of mice treated with doxycycline. By contrast, clear edematous changes were observed in the small intestine of infected mice treated with IgG control (Fig.8b). The anti-TNF-α Ab, the anti-IL-17A Ab, or doxycycline partially improved the intestinal damage, although there were still minor edematous changes. An increased number of infiltrating cells was observed in the lamina propria after infection, which was partially relieved by each of the three treatments. Staining with α-Ly6b Ab showed similar results (Fig. 8c). To identify tight junction (TJ) proteins, we stained samples for the ubiquitously expressed claudin-341. Claudin-3 expression was strong as cell-cell junctions between small intestinal epithelial cells (Fig. 8d); however, the claudin-3 signal was reduced upon infection, suggesting disruption of intestinal epithelial cell to cell interactions. Treatment with the anti-TNF-α Ab or doxycycline effectively prevented the reduction in claudin-3 expression, while anti-IL-17A Ab treatment achieved this only partially. Although the disappearance of claudin-3 may not reflect disruption of TJ between endothelial cells in microvessels, it suggests breakdown of the physiological structure of cell-cell junctions, which may lead to vascular leakage. Next, we investigated infiltrated cells by flow cytometry analysis. Anti-TNF-α Ab treatment suppressed infiltration of γδT cells, neutrophils, and CD11b+Ly6GintF4/80+Ly6Chigh cells, but not that of CD11b+Ly6GintF4/80+Ly6Clow cells (Fig. 8e–h), whereas anti-IL-17A Ab or doxycycline treatment did not clearly reduce infiltration by these immune cells. Anti-TNF-α Ab treatment strongly inhibited infiltration of γδT cells, neutrophils, and Ly6Chi monocytes. However, a fundamental question remained. Inhibition of immune cell infiltration by anti-TNF-α Ab treatment was not strong enough to account for the protective effects (Figure 2b). Thus, anti-TNF-α Ab treatment might inhibit cellular events at the transcriptional level, rather than immune cell recruitment to the small intestine.
NF-κB is a strong transcriptional activator for IL-6, and translocation of NF-κB p65 protein to the nucleus is necessary for its activity42, 43. We observed strong NF-κB p65 signals in stroma-like cells, including endothelial cells and immune cells, and moderate signals in intestinal epithelial cells in the lamina propria of the small intestine, after infection, and many p65 signals were localized to the nucleus (Fig. 9a). The number of p65-positive cells was partly reduced by α-TNF-α Ab or α-IL-17A Ab treatment; surprisingly, almost all p65 signals were localized to the cytoplasm, suggesting that blockade of TNF-α or IL-17A signaling suppresses NF-κB activation. By contrast, doxycycline did not prevent p65 nuclear localization, suggesting that the protective effect of doxycycline was due to blockade of events downstream of NF-κB activation. Furthermore, phospho-STAT-3 (Tyr705) was detected in leukocytes and intestinal epithelial cells in the lamina propria of the small intestine of infected mice (Fig. 9b), indicating strong signaling through the IL-6R. The phospho-STAT-3 signals were reduced partly by α-TNF-α or α-IL-17A Ab treatment. Additionally, we stained for DENV E protein to identify infected cells in the small intestine. The major target of DENV was Iba1-positive macrophages (Fig. 9c).