Abstract
Hemorrhagic manifestations occur frequently accompanying a wide range of dengue disease syndromes. Much work has focused on the contribution of immune factors to the pathogenesis of hemorrhage, but how dengue virus (DENV) participates in the pathogenic process has never been explored. Although there is no consensus that apoptosis is the basis of vascular permeability in human dengue infections, we showed in dengue hemorrhage mouse model that endothelial cell apoptosis is important to hemorrhage development in mice. To explore the molecular basis of the contribution of DENV to endothelial cell death, we show in this study that DENV protease interacts with cellular IκBα and IκBβ and cleaves them. By inducing IκBα and IκBβ cleavage and IκB kinase activation, DENV protease activates NF-κB, which results in endothelial cell death. Intradermal inoculation of DENV protease packaged in adenovirus-associated virus-9 induces endothelial cell death and dermal hemorrhage in mice. Although the H51 activity site is not involved in the interaction between DENV protease and IκB-α/β, the enzymatic activity is critical to the ability of DENV protease to induce IκBα and IκBβ cleavage and trigger hemorrhage development. Moreover, overexpression of IκBα or IκBβ protects endothelial cells from DENV-induced apoptosis. In this study, we show that DENV protease participates in the pathogenesis of dengue hemorrhage and discover IκBα and IκBβ to be the new cellular targets that are cleaved by DENV protease.
Introduction
Dengue virus (DENV), a mosquito-borne flavivirus, causes illness ranging from flu-like dengue fever to severe dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS) (1). Major features of DHF and DSS include hemorrhage; an increase of vascular permeability with plasma leakage, pleural, or peritoneal effusion; bleeding manifestations; and thrombocytopenia (2, 3). The most important pathophysiological event in DHF and DSS is plasma leakage, which occasionally leads to death (4–6).
We showed previously in an experimental dengue hemorrhage mouse model that DENV targets the endothelium and that both the virus and host factors contribute to endothelial cell death in hemorrhage tissues (7, 8). In addition, administration of pan-caspase inhibitor inhibits hemorrhage development (8). These results showed that endothelial cell apoptosis is key to hemorrhage development in the mouse model. In recent years, much effort in the study of dengue is devoted to the discovery of immune factors, and TNF and vascular endothelial growth factor are the leading candidates that contribute to endothelium damage (4). Less is known about viral component and the molecular mechanism of how it participates in this pathogenic process.
The DENV genomic RNA is translated during viral replication to a single polyprotein that is cleaved into three structural and seven nonstructural (NS) proteins by host cell furin proteases and viral NS2B-NS3 (2B-3) protease complex (9). The NS3 protein has a serine protease domain (activity sites: H51, D75, and S135) that spans 180 residues in the N terminus (10–12). The cleavage sites for DENV protease on the polyprotein have the consensus sequence of dibasic amino acids (R and K), followed by a small amino acid (G, A, or S) (13–15). It has been reported recently that DENV protease cleaves human MITA at LRR/96G through which it modulates human host innate immune response to infection (16, 17). Whether DENV protease is also involved in the pathogenic mechanism of dengue hemorrhage is a question that has never been addressed.
NF-κB is a transcription factor involved in the regulated expression of genes that encode for cytokines, chemokines, anti-apoptotic proteins and proapoptosis proteins (18, 19). The NF-κB complex is a heterodimer (p65/p50 subunits) in association with a family of inhibitory proteins, called IκB proteins (18). Upon degradation of the IκB proteins, the NF-κB complex translocate to the nucleus to activate target genes (18). NF-κB is also shown to be involved in the induction of apoptosis after DENV infection (20, 21). However, the role of NF-κB pathway in DENV pathogenesis is still not clear.
Flaviviral proteases, including that in Langat virus, West Nile virus, and DENV, are functional in inducing host cell apoptosis (22–25), while activity-abolished NS2B-NS3185 (H51 replaced by A, 2B-Pro-H51A) fails to induce apoptosis in Vero cells (24). Human microvascular endothelial cell-1 (HMEC-1) infected by DENV or transfected with DENV NS2B-NS3185 (2B-Pro) undergo apoptosis (8, 25). Caspase-3 activation was observed in transfected cells (25). These reports indicate that DENV protease–induced endothelial apoptosis is activity dependent and involves the apoptotic pathway. In this study, we aimed to investigate the molecular mechanism of how DENV protease induces apoptosis.
In this study, we discovered that DENV protease interacted with both IκBα and IκBβ and cleaved them. The interaction resulted in NF-κB p65 and p50 nuclear translocation and activation of extrinsic apoptotic pathway. While endothelial cells transduced with DENV protease became apoptotic, overexpressing either IκBα or IκBβ protected them from cell death. Interestingly, DENV protease targeted to endothelium by adenovirus-associated virus-9 (AAV9) displaying SLRSPPS (AAV9-SLR) in the capsid (26) induced endothelial cell death and hemorrhage development in mice. The ability of DENV protease to induce IκB-α/β cleavage and the associated events leading to cell death was protease activity dependent. To our knowledge, our results showed for the first time that dengue viral protease is involved in hemorrhage development in mice and that IκBα and IκBβ can be added to the list of cellular proteins that are cleaved by DENV protease.
Materials and Methods
Ethics statement
All experiments involving animals were carried out in strict accordance with the recommendations in the Guidebook for the Care and Use of Laboratory Animals, 3rd Ed., 2007, published by The Chinese Taipei Society of Laboratory Animal Sciences. The experiment protocol was approved by the Ethics of Animal Experiments Committee of the National Taiwan University College of Medicine (permit number 20120041)
Mice
C57BL/6 mice were obtained from National Laboratory Animal Center, National Applied Research Laboratories, Taiwan, and bred at the Laboratory Animal Center of the National Taiwan University. Mice were housed in sterilized cages, bedding, filter cage tops and fed with sterilized food and water. Mice at 6 wk of age were used in all experiments.
Yeast-two-hybrid assay
Plasmids pGilda, pJG4-5 and pSH18-34 as well as EGY48 yeast were obtained from Origene Technologies in the DupLEX-A Yeast Two Hybrid System kit. DENV 2B-3 fragment was cloned into pGilda (bait) plasmid and mouse cDNA library into pJG4-5 plasmid. Yeast EGY48 were cotransformed with pGilda-2B-3, pJG4-5-library and reporter pSH18-34. Transformants were selected in 5-bromo-4-chloro-3-indolyl β-d-galactoside–containing synthetic defined media without His, Leu, Trp, and Ura in quadruple dropout plates. The potential positive transformants were colonies that turned blue. DNA isolated from blue colonies was used to transform Escherichia coli (KC8), and the library plasmid was recovered. The recovered library plasmids were amplified by PCR and sequenced.
Vector construction and transfection
DENV components 2B-3, 2B-Pro, 2B-Pro-H51A mutant, Pro, and 2B with Flag-tag on the C terminus were cloned into pRV vectors (at restriction sites BglII and XhoI) with GFP under the control of internal ribosomal entry site (IRES). Human IκBα and IκBβ were cloned into pRV vectors with DsRed under the control of IRES. The plasmids were transiently transfected into 293T cells using Maestrofectin transfection reagent (Omics Bio). Human IκBα and IκBβ were cloned into pWPI vectors with DsRed under the control of IRES. DENV 2B-Pro, 2B-Pro-H51A, Pro, and 2B with Flag-tag on the C terminus were cloned into pWPI vectors with GFP under the control of IRES.
Lentivirus production and transduction
To generate lentivirus, pWPI carrying different DENV viral components, IκBα or IκBβ, pSPAX2-1 (CMV-Gag polymerase), and pMD2.G (CMV-envelope plasmid) were mixed in Maestrofectin transfection reagent. The mixture was added to 293T cells, which had been plated the prior day in a 10-cm culture dish. Supernatants were collected at 48 or 72 h after transfection, concentrated with buffer containing PEG-6000 (Sigma-Aldrich) and NaCl and resuspended in 150 μl PBS. To transduce cells, lentivirus preparations (8.83 × 104 transduction unit [TU]) were diluted in M200 medium (Invitrogen) with polybrene at 8 μg/ml. The mixture was added to HMEC-1 cultured in 6-well plate. The plates were centrifuged at 1100 × g in room temperature for 55 min. The supernatant was removed, and the infected cells were washed by PBS and replenished with M200 complete medium (M200 medium plus 2% of low serum growth supplement (Invitrogen)). To determine the titer of purified lentivirus, HMEC-1 cells (2 × 104) cultured in 12-well plates were transduced with 10 μl resuspended virus. The percentage of GFP+ cells was determined by flow cytometry at 48 h postinfection. The TU was calculated by the formula: TU/ml = percentage of GFP+ cells × (2 × 104)/0.01 ml.
Abs
Rabbit anti-IκBα, rabbit anti-IκBβ, rabbit anti-IκB kinase (IKK)α, and mouse anti-IKKβ Abs were obtained from Santa Cruz Biotechnology. Rabbit anti–caspase-3, -8, and -9 Abs were from Cell Signaling Technology. Mouse anti-Flaviviruses NS3 Ab was from Yao-Hong Biotechology. Mouse anti-Flag, mouse HRP-conjugated anti-Flag, and mouse anti-tubulin Abs were purchased from Sigma-Aldrich. Rabbit anti-GAPDH Ab was from GeneTex. Secondary HRP-conjugated anti-mouse Ig and HRP-conjugated anti-rabbit Ig Abs were from Cell Signaling Technology and GeneTex, respectively. HRP-conjugated anti-mouse L chain and HRP-conjugated anti-rabbit L chain Abs were from Millipore.
Reagents and inhibitors
Pan
Immunoprecipitation and Western blotting
Transfected 293T cells or transduced HMEC-1 cells with or without DENV infection were lysed with radioimmunoprecipitation assay buffer containing protease inhibitor mixture (Sigma-Aldrich). Anti-IκBα, anti-IκBβ, or anti-Flag Ab (2 μg) was added to cell lysate. The mixture was left at 4°C for 1–4 h. Protein A/G–conjugated agarose beads were added to the cell lysate, and the mixture was rotated at 4°C overnight. Samples were washed with cold PBS, resuspended in 2× sample buffer (Bio-Rad), and subjected to 12–15% SDS-PAGE, followed by transfer to polyvinylidene difluoride membranes (Millipore). The membranes were blotted with primary Abs and secondary HRP-conjugated Abs, followed by ECL (Millipore).
Measurement of IκBα and IκBβ stability
To assay for the stability of endogenous IκBα and IκBβ, transfected 293T cells were treated with or without cycloheximide (20 μg/ml) at 48 h after transfection. Cell lysates were then subjected to Western blot analysis. The density of the protein bands was analyzed by ImageJ software. The half life was determined from the slopes of ordinary least squares regression where the relative level of IκBα or IκBβ was 0.5.
Detecting the cleavage forms of IκBα and IκBβ
293T cells were cultured in 6-well plate and transfected with increasing amounts of plasmids encoding viral components and fixed amounts of IκBα or IκBβ. Proteasome inhibitor (MG132; 5 μg/ml) (Sigma-Aldrich) and IKK1/2 inhibitor VII (IKK inhibitor VII; 0.5 μM) (Merck Millipore) were added at 36 h after transfection. Cells were cultured in the presence of the inhibitor for an additional 24 h. Cleavage forms of IκBα and IκBβ in cell lysates were detected by anti–C-terminal IκBα (Santa Cruz Biotechnology) and anti–C-terminal IκBβ (Santa Cruz Biotechnology) Abs, respectively.
IKK activity assay
IKKα/β assay/Inhibitor Screening Kit (Abnova) was used to measure IKK activity. Briefly, 293T cell were transfected with pRV vectors carrying different viral components. Cells were harvested at 24, 36, and 48 h after transfection and resuspended in PBS. Cells were lysed through three −20°C/37°C freeze/thaw cycles. After centrifugation, cell lysates were harvested and applied to wells precoated with a substrate corresponding to recombinant IκBα with two serine residues that can be phosphorylated by IKKα and IKKβ. Primary anti-phosphorylated IκBα Ab was added and left at room temperature for 30 min, followed by incubation with secondary HRP-conjugated anti-mouse IgG Ab at room temperature for 30 min. Tetramethylbenzidine was used for chemiluminescence development. Absorbance was measured in a spectrophotometric iMark microplate reader (Bio-Rad) at 450 nm wavelength.
Characterization of NF-κB activation
The cytoplasmic and nuclear protein extracts from transfected 293T cells, transduced HMEC-1 or DENV-infected HMEC-1 were fractionated by reagents provided in NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific Pierce). The protein extracts were subjected to Western blot analysis. Anti-p50 (Santa Cruz Biotechnology), anti-p65 Abs (Santa Cruz Biotechnology), anti-tubulin (Sigma-Aldrich) and anti-Lamin A (Abcam) Abs were used.
Staining for cell death
DENV-2–infected (multiplicity of infection [MOI] = 10) or transduced HMEC-1 was trypsinized at different time points and resuspended in 50 μl allophycocyanin-conjugated Annexin V in Annexin V Binding Buffer (BD Pharmingen) and incubated at room temperature for 30 min in the dark. Cells were resuspended in the same buffer with propidium iodide (PI; Sigma-Aldrich) and acquired by FACSCanto (BD Biosciences) and analyzed by FlowJo software (Tree Star).
Targeted transfer of 2B-Pro and 2B-Pro-H51A to mice by AAV9
2B-Pro-IRES-GFP and 2B-Pro-H51A-IRES-GFP gene fragments were cloned into the pXX-UF1-CB-lacZ (B1268) plasmid by replacing the lacZ fragment. An empty pXX-UF1-CB plasmid was also generated by deleting the lacZ gene from B1268. To generate the endothelial cell-targeting AAV9-SLR vector, 293T cells were transfected with three plasmids: p5E18-VD2/9-SfiI1959-SLRSPPS (capsid plasmid) (26), helper plasmid and the empty B1268 plasmid, or B1268 plasmid carrying 2B-Pro or 2B-Pro-H51A. The purification and quantification of AAV9 were performed as described previously (27). Six-week-old C57BL/6 mice were inoculated intradermally with 2 × 1011 vector genome (vg) of AAV9-SLR carrying empty plasmid, 2B-Pro or 2B-Pro-H51A as described previously (7). Mice were sacrificed on day 2 after inoculation. Skin was exposed to observe hemorrhage development, then excised, snap-frozen, and embedded in optimal cutting temperature medium and kept at 80°C.
Immunofluorescence staining
Frozen skin was cryosectioned (Cryotome SME; Shandon) at a thickness of 5 μm. Sections were stained with TUNEL reagents in the In Situ Cell Death Detection kit (Roche). Alexa Fluor 488–conjugated anti-DYKDDDDK (anti-Flag; Cell Signaling Technology) and allophycocyanin-conjugated anti-CD31 (BD Pharmingen) Abs were added and left at 4°C overnight. Hoechst 33258 (Sigma-Aldrich) was used to stain nuclei. Sections were imaged by confocal microscope (Zeiss) and analyzed using LSM Image Browser software (Zeiss).
Stable expression of IκBα or IκBβ in HMEC-1 cells
HMEC-1 cells were transduced with IκBα or IκBβ and cultured for 48 h. GFP-positive IκBα or IκBβ-overexpressing HMEC-1 cells were sorted by fluorescence-activated cell sorter (FACSAria; BD Biosciences) and expanded in M200 complete medium containing 2% of low serum growth supplement.
DENV infection of HMEC-1 cells
HMEC-1 and mock-transduced and stable IκB-α/β–overexpressing-HMEC-1 cells were infected with DENV-2 16681 at an MOI of 10. The cells were gently shaken every 10 min, harvested after incubation, and subjected to analysis.
Statistical analysis
A two-tailed Student t test was used to compare the difference between groups. Data are presented as mean ± SD.
Results
Cellular IκBα and IκBβ interact with DENV protease
Using yeast-two-hybrid system with DENV 2B-3 as a bait, we screened mouse cDNA library and identified IκBβ as a cellular protein candidate that interacts with 2B-3. To confirm this interaction in mammalian cells and to expand the observation to IκBα, we cloned DENV 2B-3, 2B-Pro, 2B-Pro-H51A mutant, NS3185 (Pro), and NS2B (2B) components as well as cellular IκBβ into retroviral vectors. The vectors carrying DENV components had Flag-tag on the C terminus. Immunoprecipitation of 2B-3, 2B-Pro, and Pro, but not 2B, coprecipitated both IκBα (Fig. 1A, 1B) and IκBβ (Fig. 1C, 1D). Reciprocally, immunoprecipitation of IκBα or IκBβ coprecipitated 2B-3, 2B-Pro, and Pro but not 2B (Supplemental Fig. 1A, 1B), showing that protease-containing DENV components directly interact with IκBα and IκBβ. Interestingly, enzyme activity-dead H51A mutation did not affect the interactions of 2B-Pro with IκBα (Supplemental Fig. 1C, 1D) and IκBβ (Supplemental Fig. 1E, 1F), indicating that protease H51 activity site is not involved in its physical interaction with IκB proteins. We then generated pWPI constructs carrying cellular IκBα and IκBβ to evaluate whether the physical interaction between NS3 and IκB proteins occurs in DENV infection. HMEC-1 cells were transduced with lentiviral vectors expressing either IκBα or IκBβ and infected with DENV. Results showed that immunoprecipitation of either IκBα or IκBβ coprecipitated viral NS3 in endothelial cells infected with DENV (Fig. 1E, 1F). Taken together, these experiments demonstrated that DENV protease physically interacts with IκBα and IκBβ during infection and that this interaction does not involve its H51 activity site.
DENV protease interacts with cellular IκBα and IκBβ. (A) 293T cells were transfected with empty pRV vector (mock) or pRV vector carrying 2B-3, 2B-Pro, Pro, or 2B. The viral components as well as endogenous IκBα were detected by indicated Abs. (B) Cell lysates were immunoprecipitated with anti-Flag Ab and blotted with anti-IκBα and anti-Flag Abs. (C) 293T cells were cotransfected with pRV vectors carrying the same viral components as described in (A) and pRV vector carrying IκBβ. Cell lysates were blotted with indicated Abs. (D) Immunoprecipitations were performed as described in (B), except that anti-IκBβ instead of anti-IκBα Ab was used. (E) HMEC-1 cells were stably transduced with IκBα (left panel) or IκBβ (right panel) by lentivirus and infected with or without DENV (MOI = 10). Cell lysates were collected 36 h later and blotted with indicated Abs. (F) Cell lysates were immunoprecipitated with anti-IκBα (left panel) or anti-IκBβ (right panel) Ab and blotted with indicated Abs. Either rabbit IgG (rIgG) or mouse IgG (mIgG) Ab was used as immunoprecipitation control. Tubulin was used as a loading control. Data shown in (A)–(F) are representative of three independent experiments. Because 2B-3 and 2B-Pro were enzymatically active and contained some autocleavage sites (14, 47, 48), multiple bands were observed on the Western blot. Arrows point to uncleaved 2B-3 and 2B-Pro and white arrowheads to cleaved fragments.
DENV protease reduces the stability of IκBα and IκBβ
We first examined the stability of IκBα and IκBβ in 2B-Pro–transfected cells to determine the effect of DENV protease on IκB proteins. Interestingly, 2B-Pro transfection significantly reduced the stability of endogenous IκBα (t1/2 = 27.7 ± 3.7 min in 2B-Pro–transfected versus t1/2 = 43.4 ± 10.1 min in mock transfected, p < 0.05; Fig. 2A, top and bottom) and that of IκBβ (t1/2 = 26.5 ± 2.5 min in 2B-protransfected versus t1/2 = 51.1 ± 15.1 min in mock transfected, p < 0.05; Fig. 2A, middle and bottom) in cyclohexamide-treated cells. The stability of IκBα and IκBβ in cells transfected with activity-dead H51A mutant was comparable to mock controls (Fig. 2B). These data revealed that the specific interaction between DENV protease and IκB proteins reduces the stability of the latter.
DENV protease reduces the stability of IκBα and IκBβ by cleaving them and activating IKK. (A and B) 293T cells were transfected with empty pRV vector (mock) and pRV vector carrying 2B-Pro or 2B-Pro-H51A. At 48 h after transfection, 293T cells were treated with or without cycloheximide (CHX; 20 μg/ml) for indicated periods of time. Cell lysates were blotted with anti-Flag, anti-IκBα (top panels), or anti-IκBβ (middle panels) and anti-tubulin Abs and quantified by ImageJ software. The relative levels of IκBα and IκBβ were calculated by dividing the density of IκBα or IκBβ by the density of tubulin at the same time point. The relative levels of IκBα and IκBβ at time 0 were set as 1 (bottom panels). Data presented are means ± SD obtained from five independent experiments. *p < 0.05, **p < 0.01 by comparing either 2B-Pro (A) or 2B-Pro-H51A (B) to mock. (C) 293T cells were transfected with IκBα (0.8 μg) or IκBβ (0.5 μg) along with increasing quantities of plasmid encoding Flag-tagged 2B-Pro or 2B-Pro-H51A (1, 3, and 6 μg) (upper panels) and 2B-Pro or 2B-Pro-H51A (1.5, 3.5, and 6.5 μg) (lower panels). At 36 h after transfection, MG132 and IKK inhibitor VII were added. Viral protein was detected with anti-Flag Ab. The cleavage forms of IκBα and IκBβ were detected by anti-C terminus-IκBα and -IκBβ Abs, respectively. (D) 293T cells were transfected with pRV vector (mock) and pRV vectors containing 2B-Pro or 2B3-Pro-H51A and left in culture for 24, 36, or 48 h before harvest. Cell lysates collected at 24 and 48 h were applied to wells coated with IκBα. Phosphorylated IκBα was measured by anti-phosphorylated IκBα Ab (upper panels). Data presented are the mean ± SD of data obtained from three independent experiments with three replicates each. Cell lysates were blotted with anti-IKKα or anti-IKKβ Ab (lower panels). Tubulin was used as loading control. Arrows point to uncleaved 2B-Pro fragment and white arrowheads to cleaved fragments. *p < 0.05, **p < 0.01.
DENV protease cleaves IκBα and IκBβ and activates IKK
To examine whether IκBα and IκBβ are the potential 2B-3 targets, we added MG132 and IKK inhibitor VII to prevent IκB protein degradation by 26S proteasome and phosphorylation by IKK, respectively. We coexpressed 2B-Pro and IκBα or IκBβ in 293T cells and found that 2B-Pro dose-dependently cleaved IκBα and IκBβ (Fig. 2C). The sizes of the cleaved forms of IκBα and IκBβ were ∼30 and 33 kDa, respectively. Coexpressions of 2B-Pro-H51 and IκBα or IκBβ did not result in IκBs cleavage (Fig. 2C). Kinase assay also showed that although transfections of 2B-Pro and H51A mutant did not affect IKK α/β expression, the former but not the latter induced IKK activation (Fig. 2D). These results together indicate that DENV protease interacts with both IκBα and IκBβ and induces the cleavage of these cellular proteins through its enzymatic activity and also activates IKK.
DENV protease induces NF-κB activation
NF-κB activation is tightly controlled by a series of inhibitory proteins including IκBα and IκBβ. Hence, we tested the effect of the interaction between DENV protease and IκBα/IκBβ on NF-κB activity. Results showed that expression of DENV protease including that in 2B-3, 2B-Pro, and Pro, but not NS2B, resulted in increased translocation of p50 and p65 to the nucleus and decreased cytosolic p50 and p65 (Fig. 3A, 3B). Expression of H51A mutant failed to induce p50 and p65 translocation (Fig. 3C, 3D). These data demonstrate that DENV protease activates NF-κB by reducing the stability of IκB proteins.
DENV protease induces NF-κB activation. 293T cells were transfected with empty pRV vector (mock), pRV vector carrying 2B-3, 2B-Pro, Pro, or 2B (A and B) or pRV vector carrying 2B-3, 2B-Pro or 2B-Pro-H51A (C and D). (A, left and C, left) At 48 h after transfection, cell lysates were blotted with indicated Abs. Tubulin was used as a loading control. (A, right panel, and C, right panel) Cell lysates were separated into cytoplasmic and nuclear fractions and blotted with anti-p50, anti-p65, anti-lamin A (nuclear marker), and anti-tubulin (cytoplasmic marker) Abs. Arrows point to uncleaved 2B-3 and 2B-Pro and white arrowheads to cleaved fragments. (B and D) The p50 and p65 bands were scanned and quantified by ImageJ software. The relative density was calculated by dividing the nuclear density of p50 and p65 by the total cytosol plus nuclear density of p50 and p65, respectively. Data presented are the means ± SD of data obtained from three independent experiments. *p < 0.05, **p < 0.01 by comparing experimental group to mock control or by comparing the two groups linked by a horizontal line.
DENV protease induces endothelial cell apoptosis through its enzymatic activity
Endothelial cells are one of the targets of DENV and their interaction with DENV plays a critical role in hemorrhage development (8). We studied the effect of DENV protease on endothelial cells in vitro. Lentiviral vectors carrying DENV 2B-Pro, 2B-Pro-H51A and 2B with Flag-tag on the C terminus and GFP under IRES control were constructed. Immunofluorescence staining revealed that DENV 2B-Pro but not 2B induced HMEC-1 apoptosis (Fig. 4A), and H51A mutant failed to induce apoptosis (Fig. 4B). This demonstrates that DENV protease induces endothelial cell apoptosis through its enzymatic activity.
DENV protease induces endothelial cell death and hemorrhage development. (A and B) HMEC-1 cells were transduced with lentivirus carrying empty pWPI (mock) or pWPI vector carrying 2B-Pro or 2B (A) and 2B-Pro or 2B-Pro-H51A (B). Transduced cells were stained with Annexin V/PI at 48 h later and analyzed by flow cytometry. Dot plot and bar graph show percentage of dead cells in total GFP+ cell population. Data shown are the mean ± SD from three independent experiments. **p < 0.01. (C) Gene fragments encoding 2B-Pro or 2B-Pro-H51A with Flag-tag on the C terminus were cloned into the pXX-UF1-CB-lacZ (B1268) plasmid by replacing the lacZ fragment. Expression of the viral protein was detected by anti-Flag Ab in transfected 293T cell lysates. GADPH was used as an internal control. Arrows point to uncleaved 2B-Pro fragment and white arrowheads to cleaved fragments. (D and E) Mice were inoculated intradermally with 2 × 1011 vg AAV9-SLR carrying the empty AAV9-SLR, 2B-Pro, or 2B-Pro-H51A at four diagonal sites on the upper back. The skin was exposed and photographed on day 2 postinfection (D). Skin within the boundary of the lines connecting the four injection sites was excised and snap frozen in optimal cutting temperature medium (E). Cryosections were stained with Alexa Fluor 488–conjugated rabbit anti-Flag Ab (green), allophycocyanin-conjugated rat anti-mouse CD31 Ab (yellow), TUNEL reagents (red), and Hoechst 33258 (blue) stain. White lines outline the endothelium of blood vessel.
DENV protease induces hemorrhage in vivo
We further investigated the effects of DENV protease in vivo. Recent report showed that AAV9-SLR specifically targets human umbilical cord vein endothelial cells in situ (26). Thus, we used the AAV9-SLR transduction system to deliver DENV protease to mice (Fig. 4C). Empty AAV9-SLR or AAV9-SLR carrying 2B-Pro or 2B-Pro-H51A was injected intradermally to mice at four sites on the upper back. Interestingly, mice transduced with AAV9-SLR expressing 2B-Pro developed hemorrhage along the blood vessels on day 2 after inoculation, whereas those transduced with H51A mutant or empty AAV9-SLR had no hemorrhagic manifestation (Fig. 4D). Further analysis of the skin biopsies showed that although 2B-Pro and 2B-Pro-H51A mutant colocalized with CD31+ endothelium, TUNEL+ cells were found only in 2B-Pro–transduced mice where the nuclei of most blood vessel endothelial cells stained positive (Fig. 4E). These results showed that DENV protease induces endothelial cell death and hemorrhage in mice and the effect is dependent on protease activity. It is interesting to note that other cells in the vicinity, possibly infiltrating macrophages, were also apoptotic (Fig. 4E). It is possible that TNF released by apoptotic macrophages also contributes to protease-induced hemorrhage development (7, 8).
DENV protease induces endothelial cell death through extrinsic apoptosis pathway
Viral protein–transduced endothelial cells were analyzed for caspase cleavage. Cleaved caspase-3 and -8 were detected in 2B-Pro- and Pro- but not 2B-Pro-H51A- or 2B–transduced HMEC-1 cells, whereas no caspase-9 cleavage was observed in cells transduced with any of them (Fig. 5A–C). Although treatment with caspase inhibitors did not change the transduction efficiency (Supplemental Fig. 2A), treatment with pan-caspase or caspase-3 or -8 inhibitor significantly reduced 2B-Pro–induced HMEC-1 apoptosis (Fig. 5D), indicating that DENV protease induces HMEC-1 apoptosis by activation of the extrinsic apoptotic pathway.
DENV protease induces endothelial cell death through NF-κB–dependent extrinsic apoptosis pathway. (A–C) HMEC-1 cells were transduced with empty pWPI vector (mock), pWPI vector carrying 2B-Pro, 2B-Pro-H51A, Pro, or 2B. The lysates were blotted with indicated Abs. HMEC-1 treated with lymphotoxin (1:7500) was used as a positive control for caspase activation (Pos Ctrl). (D) HMEC-1 cells were transduced with 2B-Pro-H51A or 2B-Pro and treated with 4 μM control peptide (Z-FA-FMK), pan-caspase (Boc-D-FMK), caspase-3 (Z-DEVD-FMK), -8 (Z-IETD-FMK), or -9 (Z-LEHD-FMK) inhibitor. Transduced cells were stained with Annexin V/PI at 48 h. The percentage of dead cells in GFP+ cell population was shown in bar graph. The mean ± SD are data obtained from three independent experiments. **p < 0.01. (E and F) HMEC-1 cells were transduced with 2B-Pro-H51A or 2B-Pro before treatment without or with different concentrations of Bay 11-7082 (Bay). (E) Tranduced cells were stained with Annexin V/PI at 36 h and analyzed by flow cytometry. Bar graph shows the percentage of dead cells in total GFP+ cell population. Values in bar graphs are the means ± SD obtained from three independent experiments. *p < 0.05. (F) At 48 h after transduction, cell lysates were blotted with indicated Abs (upper panel). Bar graphs show densitometric quantification of cleaved caspase-3 and -8 against GAPDH (lower panel). Data shown are pooled from three separate experiments, one determinant in each experiment. *p < 0.05, **p < 0.01. Arrows point to uncleaved 2B-Pro fragment and white arrowheads to cleaved fragments in (A), (B), (C), and (F). (G and H) HMEC-1 cells were transduced with 2B-Pro-H51A or 2B-Pro and treated without or with different concentrations of Z-DEVD-FMK (G) or Z-IETD-FMK (H). At 48 h after transduction, the lysates were separated into cytoplasmic and nuclear fractions and blotted with anti-p50, anti-p65, anti-lamin A (nuclear marker), and anti-tubulin (cytoplasmic marker) Abs.
DENV protease-induced endothelial cell death is NF-κB-dependent
Bay 11-7082 has been shown to inactivate NF-κB through irreversible inhibition of IκBα phosphorylation (28). 2B-Pro and 2B-Pro-H51A–transduced HMEC-1 cells were treated with Bay 11-7082. Although treatment with Bay11-7082 did not affect transduction efficiency or viral protein expression (Supplemental Fig. 2B, 2C), it dose-dependently inhibited NS2B-protease–induced p50 and p65 nuclear translocation (Supplemental Fig. 3). Importantly, Bay11-7082 treatment blocked 2B-Pro–induced HMEC-1 apoptosis (Fig. 5E), demonstrating that NF-κB activation is critical to DENV protease–induced endothelial cell apoptosis. In addition, Bay11-7082 treatment dose-dependently reduced DENV protease–induced caspase-3 and -8 cleavages in HMEC-1 cells (Fig. 5F). Inversely, although blocking the activity of caspase-3 or -8 did not affect viral protein expression, it also did not inhibit NS2B-protease-induced p50 and p65 nuclear translocation (Fig. 5G, 5H, Supplemental Fig. 2D). These results together indicate that NF-κB activation precedes caspases-8 and -3 cleavages in DENV protease-induced apoptosis.
Overexpressing IκBα or IκBβ protects HMEC-1 from DENV-induced apoptosis
Because DENV protease induces HMEC-1 apoptosis by reducing the stability of IκB proteins, we investigated whether overexpressing IκBα or IκBβ protects HMEC-1 cells from DENV-induced apoptosis. HMEC-1 cells infected by DENV expressed NS3 and the cells underwent apoptosis (Supplemental Fig. 4A–C). Bay11-7082 treatment dose-dependently blocked DENV-induced p50 and p65 nuclear translocation and reduced virus-induced apoptosis, demonstrating that DENV-induced cell death is dependent on NF-κB activation (Fig. 6A, 6B). HMEC-1 cells stably transduced with IκBα and IκBβ were sorted. Although overexpressing IκBα or IκBβ did not affect virus NS3 expression or the percentage of apoptotic cells (Fig. 6C, 6D, upper panel), overexpressing IκBα– or IκBβ–protected cells from DENV-induced apoptosis (Fig. 6D, lower panel). These results showed that IκBα or IκBβ overexpression protects HMEC-1 cells from DENV-induced, NF-κB–dependent apoptosis.
Suppression of NF-κB activity reduces DENV-induced endothelial cell death. (A and B) HMEC-1 cells were treated without or with Bay 11-7082 before infection with DENV (MOI = 10) at 36 h after treatment. (A, left panel) Cell lysates were blotted with anti-NS3 and anti-tubulin Abs. Tubulin was used as a loading control. (A, right panel) Cell lysates were separated into cytoplasmic and nuclear fractions and blotted with anti-p50, anti-p65, anti-lamin A (nuclear marker), and anti-tubulin (cytoplasmic marker) Abs. (B) The cells were stained with Annexin V/PI and analyzed by flow cytometry. The dot plots (left panel) and bar graph (right panel) show the percentage of dead cells. (C and D) HMEC-1 cells were transduced with empty pWPI vector or pWPI vector carrying IκBα or IκBβ. Cells stably expressing GFP, IκBα, and IκBβ were sorted. Sorted cells were infected with or without DENV (MOI = 10). (C) Cell lysates were blotted with anti-NS3, anti-IκBα, anti-IκBβ, and anti-tubulin Abs. Arrows point to uncleaved 2B-3 fragment and white arrowheads to cleaved fragments in (A) and (C). (D) The percentage of dead cells was determined by flow cytometry at 36 h postinfection. Untransduced HMEC-1 cells were used as a baseline control (Ctrl). The bar graphs show the means ± SD percentage of dead cells obtained from three independent experiments. *p < 0.05, **p < 0.01.
Discussion
Mild hemorrhagic manifestations occur quite frequently accompanying a wide range of dengue disease syndromes. The etiology of these hemorrhagic phenomena is not well understood. We demonstrated in mouse hemorrhage model that DENV infecting endothelial cells and infiltration of TNF-producing macrophages are both important to hemorrhage development. How DENV participates in the pathogenic mechanism has never been studied, and neither has it been reported how a viral component can be involved in this pathogenic process. We found that DENV protease induces endothelial cell apoptosis through interaction with IκBα and IκBβ and causes hemorrhage development when delivered to mice. DENV protease cleaves IκBα and IκBβ after binding to them and activates IKK, thus triggering NF-κB activation, which leads to caspase-mediated endothelial cell apoptosis. Furthermore, overexpression of either IκBα or IκBβ protects endothelial cells from DENV-induced apoptosis. To our knowledge, our results demonstrate for the first time that DENV protease induces hemorrhage in mice, likely through regulating molecules in the NF-κB pathway that controls cell death.
In the mouse model, we demonstrated that DENV targets endothelium and endothelial cell death is critical to hemorrhage development (8). In this study, we used the modified AAV9-SLR system to deliver DENV protease to mouse endothelium in vivo (26). This strategy successfully targets DENV protease to mouse CD31+ vascular endothelial cells and causes hemorrhage to develop along the blood vessel. Interestingly, although activity-dead DENV protease mutant was also successfully targeted to endothelial cells, it does not induce cell death and no hemorrhage was observed. Our results revealed that SLRSPPS-displaying AAV9 is efficient in delivering genes to endothelium in mice, and importantly, that DENV protease, which causes endothelial cell death, is the key viral component that is involved in dengue hemorrhage. On the basis of our findings, we provide a new molecular pathogenic mechanism of viral hemorrhage and we demonstrate how viral protease contributes to the pathogenic process in mouse model. However, whether the phenomena observed in our mouse model can relate to human dengue remains to be demonstrated.
Early in our study employing yeast-two-hybrid system, we identified IκBβ as one of the cellular proteins that interacts with DENV 2B-3. Coimmunoprecipitation/Western blot assay further revealed that DENV protease not only physically interacts with IκBβ but also IκBα. In the same way, the N-terminal protease (Npro) of classical swine fever virus, another member of the flaviviridae that causes hemorrhage in pigs, was shown to interact with IκBα (29). In the study of Flavivirus NS3 and NS5 protein interaction networks, IκBα is found to be one of the NS3 interacting proteins (30). These studies independently support our finding that DENV protease interacts with cellular IκB-α/β. IκB proteins are characterized by the presence of many ankyrin (ANK) repeats that mediate protein–protein interaction (31–34). Through the ANK repeats, IκBβ interacts with the Rel homology region of NF-κB (34). DENV NS3 is recently shown to bind to the 19 ANK repeat-containing ANK repeat domain protein 50 (35). Thus, we postulate that DENV protease binds to IκB-α/β through the ANK repeats. Work is under way in our laboratory to test this hypothesis. The question of whether interaction with cellular IκB-α/β is a shared attribute among the NS3 proteases of other flaviviruses remains to be addressed.
DENV protease is known to cleave viral polyprotein at specific junctions that contain two basic amino acids (KR, RR, RK, and QR), followed by a small amino acid (G, A, or S) (13–15). Interestingly, both IκBα and IκBβ contain the putative cleavage sequence (PR62GS in IκBα; ERR127G in IκBβ). Our results showed that expression of DENV protease reduces the stability of endogenous IκBα and IκBβ in the host cells. We further discovered that DENV protease cleaves both IκBα (37 kDa) and IκBβ (43 kDa) into respective 30- and 33-kDa fragments, thus matching the size prediction based on the putative cleavage sites. Thus, we speculate that PR62GS and ERR127G are the DENV protease cleavage sites on IκBα and IκBβ, respectively. However, we could not rule out the possibility that IκBα and IκBβ are cleaved by a cellular protease, which may potentially be regulated by the activity of DENV protease.
During Coxsackie B3 virus (CVB3) infection, IκBα is translocated to the nucleus and recombinant protease (3Cpro) cleaves cellular IκBα (36). In cells transfected with the cleaved IκBα fragment, the TNF-induced NF-κB p65 fragment enters the nucleus. The cleaved IκBα fragment functions to retain p65 in the nucleus, inhibits NF-κB transactivation, and increases host cell apoptosis (36). Thus, through interaction with IκBα, CVB3 3Cpro increases host cell death (36). We found that DENV protease, like CVB3 3Cpro, interacts with IκB proteins but works through a different mechanism to induce cell death. DENV protease cleaves IκBα and IκBβ after binding to them and also activates IKK. Through both of these modes of action, DENV protease activates NF-κB and induces host cell death.
NF-κB activation normally leads to the production of antiapoptotic proteins, and cells are protected from apoptosis (18, 19). However, in DENV infection, that NF-κB plays a proapoptotic role has been reported in neuroblastoma (20) and hepatoma cell lines (21). In this study, we showed that DENV protease overexpression induces NF-κB activation and that NF-κB activation precedes caspase-8 and -3 activation but not vice versa. These data suggest that product(s) of the NF-κB signaling pathway triggers the extrinsic apoptotic pathway and cell death. Host factors, including TNF, reactive nitrogen, and oxygen species, are important to DENV-induced endothelial cell death and hemorrhage development in DENV infection (4, 7, 8). Our unpublished observations showed that intradermal delivery of 2B-Pro but not 2B-Pro-H51A to mice induced infiltration of TNF-producing macrophages (J.-C. Lin and B.A. Wu-Hsieh, data not shown) which may contribute to protease-induced hemorrhage development. It is possible that NF-κB-mediated endothelial cell apoptosis trigged by DENV protease is through enhancing macrophage recruitment and also possibly through enhancing the sensitivity of endothelial cells to TNF by expressing TNF receptors or by inducing reactive nitrogen or oxygen species or both. These possibilities are currently being tested.
Using activity-dead DENV 2B-Pro-H51A, we found that protease activity site is not physically involved in the interaction between DENV protease and IκB proteins, but protease activity is required for the cleavage of them, activation of IKK, and the events that follows. Furthermore, DENV protease activity is critical to its ability to induce hemorrhage development. It is recently shown that DENV protease cleaves human MITA to inhibit type I IFN response through its protease activity (16, 17). To our knowledge, these reports are the first ones to show that DENV protease interacts with and degrades cellular protein. Our report identifies IκB-α/β as the new cellular targets that are cleaved by DENV protease and demonstrates that DENV protease and IκB-α/β interaction has pathogenic consequences.
Vascular leakage and shock are the major causes of death in patients with DHF/DSS. The endothelium forms the primary fluid barrier of the circulation system. Therefore, it is important to understand the relationship between DENV-induced endothelial cell dysfunction and vascular leakage. It is reported in Ebola virus and Lassa virus infections that dysfunction of endothelial cells can affect vascular permeability and leakage (37–39). In vitro studies showed that DENV productively infects primary HUVEC and microvascular endothelial cell line HEMC-1 and causes cell death, which is enhanced by TNF (8). Several reports claimed that DENV Ags are found in endothelial cells in autopsied tissues of patients who died of DHF/DSS (40–44). However, in these studies, neither the viral negative strand RNA nor NS3 Ags was demonstrated. Neither was the identity of the endothelial cells established by their specific markers. Therefore, it remains unclear whether DENV productively infects endothelium in human cases.
In our dengue hemorrhage mouse model, we showed that introdermal inoculation of DENV results in hemorrhage development (8). Both DENV nucleic acid and Ag are detected in the endothelium within one day after inoculation. Temporal study showed that there is macrophage infiltration into the vicinity of endothelium, increased TNF production, and endothelial cell apoptosis in the hemorrhage tissues by day 2. Administration of caspase inhibitor inhibits hemorrhage development. On the basis of these published results and our current findings, we propose a working model that depicts the molecular pathogenic process of dengue hemorrhage involving both viral and immune factors (Fig. 7): 1) DENV infects the endothelium after inoculation through mosquito bite or by intradermal injection (Fig. 7A, 7B) (7, 8). 2) DENV protease is expressed in the endothelial cells, and it interacts with cellular IκBα and IκBβ (Fig. 7C, 7D). 3) DENV protease reduces the stability of IκBα and IκBβ by enzymatic activity–dependent cleavage (Fig. 7E) as well as by inducing IKK activation (Fig. 7F). 4) P50 and p65 are translocated to the nucleus as a result of IκBα and IκBβ cleavage and NF-κB is activated (Fig. 7G). 5) NF-κB activation triggers the extrinsic apoptotic pathway directly or by its product(s) and induces endothelial cell death (Fig. 7H–M). 6) Hemorrhage develops as a result of endothelial cell death. In the presence of TNF, endothelial cell death is enhanced and more severe hemorrhage develops (Fig. 7N, 7O) (7).
A proposed model for the mechanism of DENV protease-induced endothelial cell apoptosis and hemorrhage development. (a) DENV is introduced to the host intradermally by mosquito bite (in nature) or through injection (experimentally). (b) DENV infects endothelial cells after entering the host. (c) DENV protease is expressed in the infected endothelial cells. (d) DENV protease interacts with cellular IκBα and IκBβ in the cell cytoplasm. (e) IκBα and IκBβ are cleaved into smaller fragments by DENV protease. (f) DENV protease also activates IKK. IKK phosphorylates IκBα and IκBβ by which facilitates their degradation. (g) IκBα and IκBβ cleavage/degradation enables p50 and p65 translocation into the nucleus, thereby activates NF-κB. (h–l) NF-κB activation triggers extrinsic apoptotic pathway involving caspase-8 and -3 cleavages directly or through production of molecule(s) that renders endothelial cells susceptible to host factors. (m–o) Endothelial cells become apoptotic resulting in hemorrhage development. In the presence of TNF, endothelial cell death is enhanced, and more severe hemorrhage develops.
The vascular endothelium is considered to be the ultimate battlefield where damage will lead to severe consequences (4, 45, 46). On the basis of the study in our mouse model, in this study, we propose that DENV protease is the key viral component that participates in the complex pathogenic process of hemorrhage development. By interacting with IκBα and IκBβ and activating NF-κB pathway, DENV protease directly triggers the cellular machinery that can cause endothelial cell death or indirectly by making them more susceptible to the detrimental effect of host factors.
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We thank Dr. O.J. Muller (Department of Internal Medicine III, University Hospital Heidelberg, Heidelberg, Germany) for the gift of the p-5E18-VD-2/9-SfiI1759 vector. We thank the Cell Sorting Core Facility at the National Taiwan University College of Medicine for cell sorting service.
Footnotes
This work was supported by National Research Program for Genomic Medicine Grants 98, 99, and 100-3112-B-002 and National Science Council (Republic of China) Grants NSC-98-2745-B-002-003 and NSC-101-2321-B-002-062.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- AAV9
- adenovirus-associated virus-9
- ANK
- ankyrin
- 2B-3
- NS2B-NS3
- 2B-Pro
- NS2B-NS3185
- 2B-Pro-H51A
- activity-abolished NS2B-NS3185 (H51 replaced by A)
- CVB3
- Coxsackie B3 virus
- DENV
- dengue virus
- DHF
- dengue hemorrhagic fever
- DSS
- dengue shock syndrome
- HMEC-1
- human microvascular endothelial cell-1
- IKK
- IκB kinase
- IRES
- internal ribosome entry site
- MOI
- multiplicity of infection
- NS
- nonstructural
- PI
- propidium iodide
- TU
- transduction unit.
- Received October 9, 2013.
- Accepted May 21, 2014.
- Copyright © 2014 by The American Association of Immunologists, Inc.