Microparticles Provide a Novel Biomarker To Predict Severe Clinical Outcomes of Dengue Virus...

Microparticles Provide a Novel Biomarker To Predict Severe ClinicalOutcomes of Dengue Virus Infection

Nuntaya Punyadee,a,b Dumrong Mairiang,a,c Somchai Thiemmeca,a,b Chulaluk Komoltri,d Wirichada Pan-ngum,e Nusara Chomanee,f

Komgrid Charngkaew,f Nattaya Tangthawornchaikul,a,c Wannee Limpitikul,g Sirijitt Vasanawathana,h Prida Malasit,a,c

Panisadee Avirutnana,c

Division of Dengue Hemorrhagic Fever Research, Department of Research and Development,a and Graduate Program in Immunology, Department of Immunology,b

Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok-noi, Bangkok, Thailand; Medical Biotechnology Research Unit, National Center for Genetic Engineeringand Biotechnology (BIOTEC), National Science and Technology Development Agency (NSTDA), Bangkok, Thailandc; Division of Clinical Epidemiology, Department ofResearch and Development, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok-noi, Bangkok, Thailandd; Mahidol-Oxford Tropical Medicine Research Unit,Faculty of Tropical Medicine, Mahidol University, Rajthevee, Bangkok, Thailande; Department of Pathology, Faculty of Medicine Siriraj Hospital, Mahidol University,Bangkok-noi, Bangkok, Thailandf; Pediatric Department, Songkhla Hospital, Ministry of Public Health, Songkhla, Thailandg; Pediatric Department, Khon Kaen Hospital,Ministry of Public Health, Khon Kaen, Thailandh

ABSTRACT

Shedding of microparticles (MPs) is a consequence of apoptotic cell death and cellular activation. Low levels of circulating MPsin blood help maintain homeostasis, whereas increased MP generation is linked to many pathological conditions. Herein, weinvestigated the role of MPs in dengue virus (DENV) infection. Infection of various susceptible cells by DENV led to apoptoticdeath and MP release. These MPs harbored a viral envelope protein and a nonstructural protein 1 (NS1) on their surfaces. Exvivo analysis of clinical specimens from patients with infections of different degrees of severity at multiple time points revealedthat MPs generated from erythrocytes and platelets are two major MP populations in the circulation of DENV-infected patients.Elevated levels of red blood cell-derived MPs (RMPs) directly correlated with DENV disease severity, whereas a significant de-crease in platelet-derived MPs was associated with a bleeding tendency. Removal by mononuclear cells of complement-op-sonized NS1–anti-NS1 immune complexes bound to erythrocytes via complement receptor type 1 triggered MP shedding invitro, a process that could explain the increased levels of RMPs in severe dengue. These findings point to the multiple roles ofMPs in dengue pathogenesis. They offer a potential novel biomarker candidate capable of differentiating dengue fever from themore serious dengue hemorrhagic fever.

IMPORTANCE

Dengue is the most important mosquito-transmitted viral disease in the world. No vaccines or specific treatments are available.Rapid diagnosis and immediate treatment are the keys to achieve a positive outcome. Dengue virus (DENV) infection, like someother medical conditions, changes the level and composition of microparticles (MPs), tiny bag-like structures which are nor-mally present at low levels in the blood of healthy individuals. This study investigated how MPs in culture and patients’ bloodare changed in response to DENV infection. Infection of cells led to programmed cell death and MP release. In patients’ blood,the majority of MPs originated from red blood cells and platelets. Decreased platelet-derived MPs were associated with a bleed-ing tendency, while increased levels of red blood cell-derived MPs (RMPs) correlated with more severe disease. Importantly, thelevel of RMPs during the early acute phase could serve as a biomarker to identify patients with potentially severe disease whorequire immediate care.

Each year, up to 390 million people are infected by dengue virus(DENV), the most significant mosquito-borne virus of the

Flaviviridae family, posing serious socioeconomic and health bur-dens globally (1). Four serotypes of viruses (DENV serotype 1[DENV-1], DENV-2, DENV-3, and DENV-4) are transmitted tohumans by Aedes mosquitoes to establish both endemic and epi-demic transmission cycles primarily in tropical and subtropicalcountries (1). Although the majority of DENV infections in hu-mans are subclinical, a small fraction of infected individuals de-velops clinical symptoms ranging from a self-limiting mild flu-likeillness, namely, dengue fever (DF), which usually resolves withoutany complications, to a life-threatening capillary leakage syn-drome called dengue hemorrhagic fever (DHF) or dengue shocksyndrome (DSS) (2). DHF/DSS is marked by vascular leakage,resulting in hemoconcentration accompanied by thrombocytope-nia and abnormalities in liver function and coagulation, a constel-lation that may result in hemorrhage, shock, organ failure, and,

ultimately, death (3). Currently, there are no vaccines or specifictherapeutics for severe DENV infection. Rapid and reliable diag-nosis along with immediate and appropriate fluid replacement is

Received 29 July 2014 Accepted 10 November 2014

Accepted manuscript posted online 19 November 2014

Citation Punyadee N, Mairiang D, Thiemmeca S, Komoltri C, Pan-ngum W,Chomanee N, Charngkaew K, Tangthawornchaikul N, Limpitikul W,Vasanawathana S, Malasit P, Avirutnan P. 2015. Microparticles provide a novelbiomarker to predict severe clinical outcomes of dengue virus infection. J Virol89:1587–1607. doi:10.1128/JVI.02207-14.

Editor: A. García-Sastre

Address correspondence to Panisadee Avirutnan, [email protected].

Copyright © 2015, American Society for Microbiology. All Rights Reserved.

doi:10.1128/JVI.02207-14

February 2015 Volume 89 Number 3 jvi.asm.org 1587Journal of Virology

on February 3, 2015 by N

icholas White

http://jvi.asm.org/

Dow

nloaded from

http://dx.doi.org/10.1128/JVI.02207-14

http://dx.doi.org/10.1128/JVI.02207-14

http://jvi.asm.org

http://jvi.asm.org/

the key for successful clinical management to achieve a positiveoutcome in patients with DHF/DSS.

Microparticles (MPs) are a heterogeneous population of smallcell-derived vesicles generated by an active energy-dependent cel-lular process called “vesiculation” or “ectocytosis” which can oc-cur either spontaneously or in response to various stimuli, such ascell activation, apoptosis, or stress (4). MPs are characterized bytheir size (diameter, 0.1 to 1 �m), the presence of negativelycharged phospholipids (phosphatidylserine [PS]) on their sur-faces, and an antigenic profile pointing to their cellular origin (4).Under normal physiological conditions, constitutive vesiculationis an ongoing process for the majority of cells, and thus, significantlevels of MPs originating from different cells can always be de-tected in the blood (5). Changes in the level, the cellular origin,and the population composition of MPs in the circulation mayindicate distinct pathological conditions, such as cancer, tissueinjury, autoimmunity, and inflammation, as well as cardiovascu-lar, hematological, and infectious diseases (4). An elevation in thelevel of circulating MPs was found to be associated with systemicinfection, such as sepsis (6), human immunodeficiency virus in-fection (7), malaria (8, 9), and active chronic hepatitis C (10).Growing evidence suggests the potential usage of MPs as diagnos-tic biomarkers (11).

It is unclear whether MPs play a role in dengue pathogenesis.During the acute phase of dengue disease, cellular activation andapoptosis directly induced by DENV as well as inflammatory me-diators and reactive oxygen species generated from the interactionbetween viruses (or infected cells) and the host immune systemmay trigger many types of cells to shed MPs. We therefore per-formed experiments to assess this possibility. The methodologyused to detect and characterize MPs produced by DENV-infectedcells in vitro was set up and extended to the ex vivo analysis of MPsin blood specimens taken from patients at multiple time pointsand with different levels of clinical severity, leading to the novelfindings reported here. Our results point to multiple roles of MPsin dengue pathogenesis and provide a potential novel biomarkerthat can distinguish DHF patients from DF patients.

MATERIALS AND METHODSCells and viruses. All transformed cell lines used in this study were ob-tained from ATCC. HepG2 human hepatocellular carcinoma cells weregrown in Dulbecco’s modified Eagle medium (DMEM; Gibco) supple-mented with 10% fetal bovine serum (FBS; Gibco) with 1% nonessentialamino acids (Gibco) and 1 mM sodium pyruvate (Biochrome). MEG-01human megakaryoblast cells and U937 human myelomonocyte cells werecultured in RPMI 1640 medium (Gibco) supplemented with 10% FBS.EAhy926 human umbilical cord vein endothelial cells were cultured inDulbecco’s modified Eagle medium–nutrient mixture F-12 (DMEM/F-12; Gibco) supplemented with 10% FBS. African green monkey Vero cellswere grown in minimum essential medium (MEM; Gibco) with 10% FBS.C6/36 Aedes albopictus mosquito cells were grown in Leibovitz-15 me-dium (L-15; Gibco) supplemented with 10% FBS and 10% tryptose phos-phate broth (Sigma). Primary human umbilical cord vein endothelial cells(HUVECs) were isolated as previously described (12) and grown in me-dium 199 (M-199; Gibco) containing 10% FBS on 1% gelatin-coatedsurfaces. Primary human peripheral blood mononuclear cells (PBMCs)were isolated from buffy coats obtained from healthy volunteers by Ficoll-Hypaque (Pharmacia) density gradient centrifugation. All cell culture me-dia were supplemented with 2 mM L-glutamine (Sigma), 100 U/ml peni-cillin G, and 100 �g/ml streptomycin sulfate (Sigma).

To generate virus stocks, the Hawaii strain of DENV-1, the 16681strain of DENV-2, the H87 strain of DENV-3, and the H241 strain of

DENV-4 were propagated in C6/36 cells. The viral titer was determined bya focus-forming assay on Vero cells. Briefly, 10-fold serially diluted(range, 10�1 to 10�7) virus stock was added in duplicate to individualwells of tissue culture 96-well microtiter plates containing Vero cells at�90% confluence. Virus was incubated with Vero cells for 2 h at 37°C,after which the wells were overlaid with 1.5% gum tragacanth (Sigma) inMEM with 4% FBS. After culture at 37°C for 72 h, the overlay medium wasremoved and the wells were washed with phosphate-buffered saline(PBS). Cells were fixed with 3.7% formaldehyde in PBS for 10 min at roomtemperature and permeabilized with 2% Triton X-100 (Fluka) in PBS(Triton X-100 buffer) for 10 min at room temperature. Cells were stainedwith the culture supernatant of mouse monoclonal antibody (MAb) spe-cific to flavivirus envelope protein (4G2) at 37°C for 1 h. Following severalwashes, the wells were incubated with horseradish peroxidase (HRP)-conjugated anti-mouse IgG (1:1,000 dilution in Triton X-100 buffer;Dako) for 30 min at 37°C. The wells were washed, and infectious foci werevisualized with 3,3=diaminobenzidine tetrahydrochloride (DAB) sub-strate (Sigma) after a 5- to 10-min incubation at room temperature. Thewells were rinsed with water and dried prior to analysis.

DENV infection and MP isolation. Cells (1 � 106) were infected withDENV-2 at multiplicities of infection (MOIs) of 1 and 5. Supernatantsand cells were harvested at 24, 48, and 72 h after infection. Harvested cellswere washed twice, and the percentage of apoptotic cells was measured byuse of a fluorescein isothiocyanate (FITC)-conjugated annexin V (AnV)detection kit I (BD Biosciences) according to the manufacturer’s instruc-tion and analyzed by use of a FACSCalibur flow cytometer (Becton, Dick-inson). Data were processed with CellQuest software (Becton, Dickin-son). To quantitate the percentage of infected cells, harvested cells werefixed with 2% formaldehyde (BDH, England), permeabilized with 0.1%Triton X-100 (Sigma), stained with 100 �l undiluted hybridoma culturesupernatants of anti-DENV-2 NS1 MAb clone 2G6 (13) followed by 6�g/ml of FITC-labeled rabbit anti-mouse IgG (Dako), and analyzed byflow cytometry. For MP isolation, culture supernatants from mock- andDENV-infected cells were collected at the times indicated below and seri-ally centrifuged at 500 � g and at 13,000 � g (10 min each) to remove celldebris and small aggregates, respectively. A final centrifugation step at20,000 � g for 60 min was performed to pellet the MPs. The MP pelletswere washed thoroughly with PBS and kept at �70°C until MP analysis. Insome experiments, wild-type 293T cells and 293T cells stably expressing Tcell immunoglobulin domain and mucin domain 1 (TIM-1; kindly pro-vided by Ali Amara) were infected with DENV-2 at an MOI of 0.3 for 2days until both cells and supernatants were harvested and processed asdescribed above.

To determine MP generation in cells infected with the four serotypesof DENV, HepG2 cells were infected with each serotype of DENV at anMOI of 5 in duplicate. One set of infected cells was UV irradiated at 3,000J/s for 10 min (Hoefer UVC500 UV cross-linker) to inactivate DENV.Supernatants and cells were harvested at 48 h after infection. Analyses ofinfection efficiency and apoptosis level and MP isolation were performedas described above.

Analysis of MPs generated from DENV-infected cells. MPs from cul-ture supernatants (1 ml) of mock- and DENV-infected HepG2 cells wereisolated as described above. Isolated MPs were stained with 3 �l of FITC-conjugated AnV (BD Biosciences) plus 5 �l of 10-fold-concentrated AnVbuffer for 30 min on ice. AnV buffer (300 �l; 150 mM NaCl, 2 mM CaCl2,10 mM HEPES, pH 7.4) was added prior to flow cytometric analysis. MPswere gated by the use of standard beads of defined size (BD Biosciences).Only events with a size of less than 1 �m were further evaluated for thepercentage of AnV-positive MPs. Events within this size limit werestopped at 10,000 counts.

For analysis of DENV antigens expressed on the surfaces of MPs, 40 �lof 10-fold-concentrated mock- and DENV-infected MPs was incubatedwith 5 �l of Fc receptor blocking solution (BioLegend) for 15 min on ice,followed by a 1-h incubation on ice with 1 �g/ml of FITC-conjugatedanti-NS1 2G6 MAb or 1 �g/ml of allophycocyanin (APC)-conjugated

Punyadee et al.

1588 jvi.asm.org February 2015 Volume 89 Number 3Journal of Virology


icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

anti-envelope protein (anti-E) 4G2 MAb. FITC- and APC-conjugatedmouse MAbs against unrelated antigens were used as negative controls forNS1 and E staining, respectively. After a 30-min incubation with phyco-erythrin (PE)-conjugated AnV (3 �l; BD Biosciences) on ice, AnV buffer(300 �l) was added and the solution was transferred into a tube with aknown density of fluorescent TruCount beads (BD Biosciences) and an-alyzed using a BD FACSCalibur flow cytometer. PE-conjugated mouseIgG specific to unrelated antigens was used as a negative control for AnVstaining. MP acquisition gating was performed as described above. Abso-lute numbers of MPs were calculated according to the following formula:(number of events of MPs)/(number of events of TruCount beads col-lected) � (TruCount Bead count per test/sample volume) � dilutionfactor.

Transmission electron microscopy. HepG2 cells infected with DENVat an MOI of 5 were harvested at 48 h postinfection and fixed with 1%glutaraldehyde in 0.1 M sodium cacodylate buffer at 4°C overnight. Fixedcells were then treated with 1% osmium tetroxide in Millonig buffer (120mM Na2HPO4, 100 mM NaOH, 30 mM D-glucose [pH 7.4]) for 1 h at 4°Cand later dehydrated with graded ethanol solutions: 70%, 80%, 90%, and100%. Dehydrated samples were transferred from 100% ethanol to a 50:50mixture of ethanol-propylene oxide and finally to 100% propylene oxidebefore being infiltrated with epoxy resin (Ted Pella Inc., Redding, CA) andpolymerized at 60°C for 48 h. Ultrathin sections were cut on an ultrami-crotome using a Diatome diamond knife (Leica EM UC7) and thenmounted on grids and stained with 2% uranyl acetate and lead citratesolution. Images were collected by a transmission electron microscope(JEM 1230; JEOL, Peabody, MA).

Negative staining. MPs from culture supernatants of mock- andDENV-infected HepG2 cells were isolated as described above. A 10-fold-concentrated MP preparation (3 �l) was adsorbed on a glow-dischargedFormvar-carbon-coated nickel grid (Electron Microscopy Sciences, Hat-field, PA) for 5 min. Excess sample was removed by a filter paper, followedby a wash with distilled water. The grids were then stained with 1.5%phosphotungstic acid (PTA) (pH 6.6) for 2 min, and air-dried grids werevisualized using an FEI Tecnai T20 transmission electron microscope op-erated at 200 kV.

Immunogold labeling. Expression of PS on the surface of DENV MPswas evaluated by immunogold labeling with AnV as previously described,with some modifications (14). Briefly, a 10-fold-concentrated MP prepa-ration was adsorbed on a glow-discharged Formvar-carbon-coated nickelgrid for 5 min. Afterward, the grids were stained with 3 �l FITC-conju-gated AnV in 10-fold-concentrated AnV buffer for 1 h at room tempera-ture. Excess staining was removed by a filter paper, and the grids were thenincubated with blocking buffer (PBS containing 1% bovine serum albu-min [BSA]; Aurion, Wageningen, the Netherlands). After a 10-min incu-bation at room temperature, the grids were washed with PBS containing0.1% BSA and then incubated with 10-nm gold nanoparticle-conjugatedmouse anti-FITC MAb (Aurion) for 1 h at room temperature. Subse-quently, the grids were washed with 0.1% BSA–PBS, stained with 1.5%PTA for 2 min, and then washed with distilled water and air dried. Thespecificity of AnV immunogold labeling on MPs was verified by stainingthe grid that was adsorbed with PBS instead of MPs, followed by theimmunolabeling procedure described above. The samples were observedusing an FEI Tecnai T20 transmission electron microscopy operated at200 kV.

In vitro experiments with immune complexes (ICs). Healthy adultvolunteers provided blood for this study following the provision of writ-ten informed consent approved by the Institutional Review Board of theFaculty of Medicine Siriraj Hospital (protocol number 159/2557). Col-umns, buffers, and the method for construction of a high-gradient mag-netic separator (HGMS) apparatus for isolating cells from whole bloodwere described previously (15). Briefly, the isolation of human erythro-cytes by using HGMS (X-Zell Biotech) was as follows: 2 ml of EDTA-blood from healthy volunteers was centrifuged at 800 � g for 10 min,washed three times with 5 ml of PBS containing 0.1% BSA (Sigma), and

centrifuged at 500 � g for 5 min after each wash. Subsequently, 250 �l ofpacked red blood cells (RBCs) was resuspended in 750 �l of PBS contain-ing 3% BSA, the suspension was mixed with 20 �l CD45 HMxBeads(X-Zell Biotech), and the mixture was incubated for 20 min on a con-stantly rotating shaker at room temperature. After incubation, the sus-pension was centrifuged at 500 � g for 5 min to remove unbound beads,followed by the addition of 1 ml PBS containing 3% BSA and 15 �l ofCD45 HMxBeads for a second round of incubation on a constantly rotat-ing shaker at room temperature for 20 min. During the second round ofincubation, an HMX column was prepared for purification of RBCs.Briefly, the column was filled with PBS containing 1% BSA, and afterremoving the remaining air bubbles by gentle finger tapping, a 26-gauge1-in. needle was connected to the stopcock. The column was placed be-tween the poles of the dipole magnet and equilibrated for 5 to 10 min withPBS containing 1% BSA. After finishing the second round of incubation,the suspension was applied to the top of the column and the flowthroughwas collected in a 15-ml tube. The flowthrough was centrifuged at 500 �g for 5 min, washed once with DGVB�� (2.5 mM barbital sodium, 139mM dextrose, 71 mM NaCl, 0.1% [wt/vol] gelatin, 0.15 mM CaCl2, 1 mMMgCl2), and finally, checked for purity (�99.9%) by flow cytometry(FACSCalibur; BD Biosciences).

In vitro analysis of IC opsonization and vesiculation. Purified DENVNS1 (100 ng/ml) (16) was incubated with 5-�g/ml IgG fractions purifiedfrom pooled dengue patient convalescent-phase serum (PCS) in the pres-ence of 5 � 106 purified RBCs diluted in DGVB�� and 10% autologousserum for 1 h at 37°C. After centrifugation, the supernatants were col-lected for MP analysis and the cells were washed once with 1 ml ofDGVB��. Subsequently, 7.5 � 106 autologous PBMCs diluted in 100 �lof DGVB�� were added and the mixture was incubated for 30 min at37°C. Afterwards, the supernatants were collected for MP analysis and thecells were washed once with 0.1% BSA–PBS, followed by flow cytometricanalysis. Negative controls, which were run in parallel, consisted of puri-fied NS1 diluted with 5 �g/ml of PCS in the presence of 20 mM EDTA (toinhibit complement activation) or with 5 �g/ml of pooled non-denguepatient serum (PND).

To determine the binding of ICs to human erythrocytes before andafter addition of PBMCs, cells were washed with 0.1% BSA–PBS andstained with 20 �g/ml of anti-C3d clone YB2-39HL (17) diluted in 0.1%BSA–PBS for 1 h on ice, followed by 4 �g/ml Alexa Fluor 488 goat anti-ratIgG antibody (Invitrogen) for 30 min on ice. After washing, complementfragment deposition on erythrocytes was analyzed by flow cytometry andthe data were processed with FlowJo software (Tree Star).

MPs were quantified after in vitro IC experiments. Briefly, 34 �l ofsupernatant was incubated with 5 �l of 10-fold-concentrated AnV bufferand 3 �l of FITC-conjugated AnV either with 3 �l of PE-conjugatedanti-CD45 or with 5 �l of PE-Cy5-conjugated anti-CD235a at a dilutionof 1:100. After 30 min, 300 �l of AnV buffer was added and the mixturewas transferred into a tube with a known number of fluorescent TruCountlyophilized pellet beads and analyzed by flow cytometry as describedabove.

Patients. This study was performed at the Khon Kaen Provincial Hos-pital and Songkhla Hospital between the epidemic seasons of 2010 and2013. Written informed consent was sought from the parents or guardiansof all participants in accordance with the guidelines of the ethical com-mittees of the Ministry of Public Health (protocol number 92/2550), Fac-ulty of Medicine Siriraj Hospital (protocol number 349/2550), the KhonKaen Provincial Hospital, and the Songkhla Hospital. Acute dengue virusinfection was diagnosed by using reverse transcription-PCR (RT-PCR)-based DENV gene identification or by using DENV-specific IgG and IgMcapture enzyme-linked immunosorbent assay (ELISA) (18, 19). The in-clusion criteria for the patients enrolled in the study were described else-where (16). Patients who tested positive for DENV were classified into theDF or DHF group according to the World Health Organization (WHO)1997 guideline (2), whereas those with negative RT-PCR and DENV an-tibody results were assigned to the control group, labeled the group with

Roles of Microparticles in Dengue Pathogenesis



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

other febrile illness (OFI). Samples used for this study were derived frompatients that ranged in age from 5 to 15 years (mean age, 10.3 years); 36%of samples were obtained from females, and 64% were obtained frommales. Of a total of 53 samples included in this study, 24 samples werefrom DHF cases, 19 were from DF cases, and 10 were from OFI cases. Of

the 19 DF patients, there were 3 cases with DENV-1 infection, 6 cases withDENV-2 infection, 7 cases with DENV-3 infection, and 1 case withDENV-4 infection. Of the 24 DHF patients, there was 1 case with DENV-1infection, 16 cases with DENV-2 infection, and 3 cases with DENV-3infection. All dengue cases were secondary infections. Data were derived

TABLE 1 Demographic, clinical, and key laboratory characteristics of patients enrolled in the studya

Characteristic

Result for the following patient groups:

P valueOFI patients (n � 10) DF patients (n � 19) DHF patients (n � 24)

No. (%) male patients 5 (50) 10 (53) 19 (79) 0.1012Age (yr) 9.5 (6, 10.8) 10 (8, 11) 11 (8.3, 13.8) 0.1824Body wt (kg) at enrollment 22.5 (18.5, 35.7) 33.0 (23.2, 49.0) 35.5 (25.5, 46.0) 0.0683Platelet nadir (109/liter) 184.5 (161.0, 198.0) 96.0 (52.5, 140.0) 27.5 (21.2, 48.8) 0.0001Day of acute illness of platelet nadir �0.5 (�1.8, 0.0) 0 (0, 1) 0 (0, 1) 0.0454Maximum RBC count (1012/liter) 4.7 (4.5, 4.9) 5.4 (4.9, 5.8) 5.5 (5.2, 6.0) 0.0040Day of acute illness of maximum RBC count 0 (�2, 1) 0.0 (�0.5, 0.5) 0 (�1, 1) 0.5826No. (%) of patients with mucosal bleeding 0 (0) 5 (29) 12 (71) 0.0078Day of acute illness of first episode of mucosal bleeding — �1 (�1.5, �0.5) 0 (�1, 0) 0.2007a Data are presented as the number (percentage) for categorical variables and median (25th, 75th percentiles) for continuous variables. OFI, other febrile illness; DF, dengue fever;DHF, dengue hemorrhagic fever; RBC, red blood cell; —, OFI patients did not have mucosal bleeding.

FIG 1 DENV infection causes apoptotic cell death. (A to D) HepG2 cells (A and B) and HUVECs (C and D) were infected with DENV-2 at MOIs of 1 and 5 andharvested at 1, 2, or 3 days postinfection for intracellular NS1 staining to determine the percentage of infected cells (A and C) and AnV labeling to measure thepercentage of cells that had undergone apoptosis (B and D). (E and F) HepG2 cells were infected with DENV-1 (strain Hawaii), DENV-2 (strain 16681), DENV-3(strain H87), and DENV-4 (strain H241) at an MOI of 5 and harvested at 2 days postinfection to determine the percentage of infected cells (E) and the percentageof cells that had undergone apoptosis (F). In parallel, cells were treated with equal volumes of UV-irradiated DENV (UV-DENV), harvested at 2 days posttreat-ment, and processed in the same way as were cells incubated with live DENV. Data are the mean SD from three to four independent experiments. Asterisksdenote statistically significantly differences between infected or apoptotic cells and mock-infected cells (*, P 0.05; **, P 0.001; ***, P 0.0001).

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

from plasma samples collected daily during the febrile phase and pairedconvalescent-phase samples obtained at 2 weeks and 2 months after de-fervescence. Study day 0 (defervescence) was defined as the calendar dayon which the patient’s temperature fell and stayed below 37.8°C. A sum-mary of the subjects’ characteristics is shown in Table 1.

Blood sample collections. Human peripheral blood was collected inVacutainer K3-EDTA tubes (BD Biosciences) and processed within 1 hafter blood collection. Blood samples were centrifuged at 750 � g for 20min at 4°C to obtain platelet-rich plasma (PRP), followed by a secondcentrifugation step at 1,250 � g for 15 min at 4°C to obtain platelet-poorplasma (PPP). PPP was carefully removed, aliquoted, snap-frozen in liq-uid nitrogen, and stored at �80°C until MP analysis.

Labeling of plasma MPs. The cellular origins of plasma MPs weredetermined using fluorescence-conjugated MAbs specific to cellularmarkers. DENV antigens associated with plasma MPs were identified us-ing fluorescence-conjugated MAbs specific to E and NS1. Briefly, PPP (15�l) was incubated with 3 �l of Fc blocker for 30 min on ice prior to theaddition of 3 �l FITC-conjugated AnV to identify MPs, 1 �l APC-conju-gated anti-DENV E or NS1 MAbs, and 5 �l PE- or peridinin chlorophyllprotein (PerCP) complex-labeled MAbs specific for CD235a (glycophorinA; for red blood cell-derived MPs), CD41a (glycoprotein IIb/IIIa; forplatelet-derived MPs), CD14 (for monocyte-derived MPs), CD66 (for gran-ulocyte-derived MPs), CD3 (for T cell-derived MPs), CD19 (for B cell-de-rived MPs), and CD56 (for NK cell-derived MPs). In some experiments,fluorochrome-conjugated isotype-matched control IgG1 (MOPC-21) andIgG2a (UPC-10) antibodies were used instead of MAbs specific for cellularmarkers. All antibodies except APC-conjugated anti-DENV MAbs were pur-chased from BD Biosciences; APC-conjugated anti-DENV MAbs were gen-erated in-house using an APC conjugation kit (Biolegend) according to themanufacturer’s instructions. After a 30-min incubation in the dark, 300 �l ofAnV buffer was added and the mixture was transferred into a tube with aknown density of fluorescent TruCount beads and analyzed by flow cytom-etry. To avoid any sedimentation of beads, the samples were thoroughlymixed prior to flow cytometric analysis. All of the staining buffers and fluo-rochrome-labeled antibodies were clarified by centrifugation at 13,000 � gfor 10 min to remove small aggregates before staining.

MP procoagulant activity. MP procoagulant activity was assayed witha functional chromogenic method on micro-ELISA plates (ZymuphenMP-activity) according to the manufacturer’s instructions.

Statistical analysis. For the in vitro studies, data were collected from atleast three independent experiments. Data sets were compared using atwo-tailed, unpaired t test. Multiple comparisons were performed usingan analysis of variance (ANOVA) test. Correlation coefficients were esti-mated on the basis of simple linear regressions and Pearson product-moment correlation using GraphPad Prism (version 5) software. Statisti-cal significance was achieved at a P value of 0.05.

For the ex vivo studies, the demographics of three patient groups (theDF, DHF, and OFI groups) were compared using the Fisher exact test andthe Kruskal-Wallis test for categorical and numeric data, respectively. The

levels and dynamics of MPs, RBCs, and platelets among the patient groupsduring the acute phase of infection were analyzed by use of a mixed modelusing a quadratic form. There were some missing data, since the diseaseday at admission of each patient was different. Only patients with data onat least three consecutive days were included in the analysis. Since thelevels of AnV-positive (AnV�) MPs, CD41� MPs, CD235� MPs, RBCs,and platelets were positively skewed, a logarithmic transformation wasapplied. In addition, the ratios in percentage terms between CD41a� MPsand AnV� MPs and between CD235� MPs and AnV� MPs were used asanother means to normalize CD41a� MPs and CD235� MPs, respec-tively.

To determine whether MP, RBC, and platelet levels and dynamicswere significantly different among the patient groups, t tests were per-formed on mixed models and P values were calculated on the basis of theSatterthwaite approximation for denominator degrees of freedom. Todetermine whether MP, RBC, and platelet levels on the same date betweentwo patient groups or between two dates for a single patient group weresignificantly different, general linear hypotheses and multiple compari-sons for parametric models were used on mixed models.

To determine whether MP, RBC, and platelet levels among patientgroups were associated with factors other than DENV infection, such asgenetic background, the levels of total AnV� MPs, CD41a� MPs, CD235�

MPs, RBCs, and platelets at 14 days and 60 days after defervescence weretested with Welch’s t test for any significant differences among patientgroups. Finally, receiver operating characteristic (ROC) curves of absoluteCD235� MPs and percentage of CD235� MPs were constructed to assessthe performances of the earliest levels of CD235� MPs before deferves-cence as a biomarker for predicting the severity of dengue virus infection.

All statistical analyses of the ex vivo data were performed with the Rprogramming language (version 3.0.2), with the lme4 package used formixed model construction, the lmerTest package used for P value calcu-lation from the Satterthwaite approximation, the multicomp packageused for general linear hypotheses and multiple comparisons, and theROCR package used for ROC curve construction. Graphs and charts wereplotted with GraphPad Prism 5 software.

RESULTSDENV infection leads to apoptotic death and MP shedding.Analogous to the results of previous studies (20, 21), infections ofa human hepatocyte cell line (HepG2) and primary human um-bilical cord vein endothelial cells (HUVECs) by DENV-2 led toapoptotic death in an MOI-dependent manner (Fig. 1). To assessMP generation from different cell types induced by DENV-2, cul-ture supernatants were harvested at various time points after in-fection. MPs were isolated by sequential centrifugation and la-beled with the PS-binding protein annexin V (AnV) prior to flowcytometric analysis. Standard microsphere beads of 1 and 3 �m indiameter (Fig. 2A) were used as size markers to identify the MP

FIG 2 DENV infection induces MP production from various cell types. (A to D) Representative histogram and density plots showing the gating protocol for MPs.(A) Histogram profiles of FSC of the 1- and 3-�m-diameter calibrator beads which were used as size markers for MP identification. (B) MPs were defined asevents with a size of less than 1 �m, which are gated in the R1 window. (C and D) Size-selected events (gate R1) are plotted as a function of their fluorescence forspecific AnV-FITC binding against FSC. Positively labeled events in the left upper quadrants are considered AnV� MPs. Examples of density plots showing thepercentage of AnV� MPs in culture supernatants of mock-infected (C) and DENV-infected (D) HepG2 cells harvested at 48 h postinfection are depicted. (E toJ) The percentages of AnV� MPs generated by mock- and DENV-infected HepG2 cells (E) and HUVECs (F) at 1, 2, or 3 days, EAhy926 (G) and U937 (I) cellsat 2 days, primary monocytes (H) at 1 day, and MEG-01 cells (J) at 7 days postinfection were measured. (K) The percentage of AnV� MPs generated from HepG2cells treated for 2 days with UV-irradiated or live viruses of all four DENV serotypes was analyzed. All cell types were infected by DENV at an MOI of 5. Data arethe mean SD from three to four independent experiments. Asterisks denote statistically significantly differences between the percentage of AnV� MPsproduced by DENV-infected cells and the percentage produced by mock-infected cells (*, P 0.05; **, P 0.001). (L) Transmission electron micrograph of aDENV-infected HepG2 cell displaying small vesicles of 80 to 200 nm in size (arrowheads) near the cell periphery. (M and N) Budding MPs at higher magnifi-cation. (O to Q) Negative staining on the grids absorbed by buffer (O), isolated MPs (P), and sucrose density-purified virus particles (Q) released fromDENV-infected HepG2 cells are depicted. (R to T) Immunogold labeling of isolated MPs. (S and T) The clusters of 10-nm gold particles (black dots) at theperiphery of bilamellar vesicular structures of MPs indicate the externalization of AnV-bound PS at the outer leaflet of the MP membrane. (R) Grids adsorbedwith buffer instead of MPs that then underwent the same immunogold labeling procedure used for the isolated MPs in the images shown in panels S and T.

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

population consisting of particles smaller than 1 �m, as assessedby the logarithmic amplification of forward scatter (FSC) and sidescatter (SSC) signals (Fig. 2B, region R1) along with AnV binding(Fig. 2C and D, left upper quadrants). Consistent with increasedDENV-induced apoptotic death over time after infection (Fig. 1),the percentages of AnV� MPs produced by DENV-2-infectedHepG2 cells and HUVECs rose and were significantly higher thanthose spontaneously produced by mock-infected cells (Fig. 2E andF). Other potential target cell types for DENV in humans, includ-ing an endothelial cell line (EAhy926), primary monocytes, amonocytic cell line (U937), and a megakaryocyte cell line (MEG-01), also shed MPs in response to DENV infection (Fig. 2G to J).Infection with all four serotypes of DENV of HepG2 cells also ledto apoptotic death (Fig. 1E and F) and MP shedding (Fig. 2K). Theamount of MP generation likely depends on the degree of apop-totic cell death, the process of which requires the active replication

of DENV, as incubation of inactivated UV-irradiated DENV withcells did not induce apoptotic cell death (Fig. 1E and F) or MPshedding (Fig. 2K). In support of this, the cells overexpressing Tcell immunoglobulin domain and mucin domain 1 (TIM-1) pro-tein, which was recently identified to be an entry receptor forDENV (22), enhanced DENV susceptibility, leading to moreapoptotic death and, consequently, levels of MP generation higherthan those in the wild-type control cells (data not shown).

Transmission electron microscopy of DENV-infected HepG2cells at 48 h postinfection demonstrated a heterogeneous popula-tion of MPs of various sizes and shapes around the cell periphery(Fig. 2L). Some MPs appeared to be budding from the cell surface(Fig. 2L, arrowheads). Most MPs had an electron density similarto that of the cytoplasm and clearly defined margins (Fig. 2M andN). To investigate the MPs in greater detail, naturally liberatedMPs were isolated from the supernatants of DENV-infected

FIG 3 MPs generated from DENV-infected cells express E and NS1 antigens on their surfaces. MPs isolated from 10-fold-concentrated culture supernatants of mock-and DENV-infected cells were labeled with PE-conjugated AnV (AnV-PE), APC-conjugated anti-E (E-APC), and FITC-conjugated anti-NS1 (NS1-FITC) MAbs andanalyzed by flow cytometry. The absolute numbers of MPs generated in culture supernatants were measured on the basis of a known number of fluorescent TruCountbeads. (A to C) Representative flow cytometric dot plots of MPs in buffer (A) and culture supernatants of mock-infected (B) and DENV-infected (C) HepG2 cells aredepicted. Region R1 represents the FSC/SSC light scatter gate of MPs (size, 1 �m). Region R2 represents the known density of TruCount beads. Data acquisition wasstopped when the number of TruCount beads reached 2,000 events. (D to G) Examples of density plots of MPs (events in the R1 gate) in mock-infected (D and F) andDENV-infected (E and G) cell culture supernatants stained with an unrelated PE-conjugated IgG (IgG-PE) (D and E) as negative controls or AnV-PE (F and G) aredisplayed. The percentages of total AnV� MPs in the R1 gate (left upper quadrants) detected in the culture supernatants of mock- and DENV-infected cells are depicted.AnV� MPs were further analyzed for E and NS1 expression. (H to J) Representative density plots of AnV� MPs from DENV-infected cells stained with FITC- andAPC-conjugated isotype control Abs (IgG-FITC and IgG-APC, respectively) (H) and AnV� MPs from mock-infected (I) and DENV-infected (J) cells double stainedwith FITC-conjugated anti-NS1 MAb clone 2G6 (anti-NS1-FITC) and APC-conjugated anti-E MAb clone 4G2 (anti-E-APC). The percentages of AnV� MPs positivefor NS1 alone (left upper quadrants), E alone (right lower quadrants), and both E and NS1 (right upper quadrants) generated by mock-infected (I) and DENV-infectedcells (J) are depicted. (K) The absolute numbers of total AnV� MPs and AnV� MPs negative for both E and NS1 (E� NS1�), positive for E alone (E� NS1�), positivefor NS1 alone (E� NS1�), and positive for both E and NS1 (E� NS1�) were determined by using TruCount beads of known density (region R2 in panel A). Data are themean SD from four independent experiments. Asterisks note statistically significantly differences between the percentage of AnV� MPs produced by DENV-infectedcells and the percentage produced by mock-infected cells (*, P 0.05; **, P 0.001).




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

HepG2 cells at 48 h postinfection, processed by negative staining,and observed under a transmission electron microscope. Electronmicrographs of isolated MPs showed intact double-layered mem-branous structures (Fig. 2P) with heterogeneous sizes between 0.1and 1 �m (data not shown). Background negative staining ad-sorbed by buffer on the grids was used as a negative control (Fig.2O). For comparison, virions were purified from the same batchof culture supernatants from which MPs were isolated and werevisualized by negative-staining electron microscopy. Similar towhat had been described by a previous study for the structure ofhuman cell-derived DENV virions generated at 37°C using cryo-electron microscopy (23), negative-staining electron micrographsof sucrose density gradient-purified DENV produced by infectedHepG2 cells revealed bumpy or spiky virions (diameter, �50 nm),which are characteristics of immature or partially mature viruses,as well as some deformed particles (Fig. 2Q). Binding of AnV to PSon the surface of MPs was also confirmed by immunogold labelingof negatively stained MPs (Fig. 2S and T). The specificity of AnVimmunogold labeling on MPs was verified by staining a grid ad-sorbed with buffer instead of MPs, followed by the same immu-nogold labeling procedure described above (Fig. 2R).

MPs released from DENV-infected cells express viral anti-gens on their surfaces. Because DENV-infected cells express viralproteins, i.e., envelope protein (E) (24) and nonstructural proteinNS1 (25), on their surfaces, infected cell-derived MPs were exam-ined for the presence of these viral antigens. To obtain an accuratenumber of MPs, the known density of TruCount beads, whichwere identified by their size on the basis of the logarithmic ampli-fication of the FSC and SSC signals (Fig. 3A to C, region R2), wasused as an internal reference. MPs generated from mock-infected(Fig. 3B) and DENV-infected (Fig. 3C) HepG2 cells were localizedin the R1 region (as described for Fig. 2) and specifically labeledwith phycoerythrin (PE)-conjugated AnV (Fig. 3F and G) but notwith the negative control, an unrelated PE-conjugated IgG (Fig.3D and E). Of note, the buffer used in the experiments showedvery few and negligible background noise signals within the sameMP gate (Fig. 3A). AnV� MPs in DENV-infected culture super-natants were further analyzed for DENV E and NS1 antigen ex-pression by double immunofluorescence staining with an allo-phycocyanin (APC)-conjugated anti-E MAb and a fluoresceinisothiocyanate (FITC)-conjugated anti-NS1 MAb, respectively(Fig. 3J). The number of DENV-infected cell-derived AnV� MPslabeled with IgG isotype controls (Fig. 3H) and the number ofAnV� MPs acquired from mock-infected supernatants labeledwith anti-E or anti-NS1 MAbs were not significant (Fig. 3I). Theabsolute numbers of MPs were also calculated by using the knowndensity of TruCount beads (Fig. 3K). The supernatants fromDENV-infected cells contained a significantly higher total numberof AnV� MPs than the supernatants from mock-infected cells(P � 0.032). Notably, the majority of AnV� MPs derived fromDENV-infected cells were negative for viral E and NS1 (E� andNS1�, respectively) antigens on their surfaces (Fig. 3K). Aboutone-third of the total AnV� MPs produced by DENV-infectedcells were positive for at least one of the viral antigens, E and NS1(Fig. 3K).

The relatively low detectable levels of DENV antigen-contain-ing MPs might reflect the fact that low levels of antigens are ex-pressed on the surface of infected cells. During outward buddingof the plasma membrane to release MPs, the DENV antigen inten-sity on the surface of very small particles like MPs might be too low

to be positively identified by flow cytometry. Alternatively, a frac-tion of DENV antigen-negative MPs might be released from in-fected cell-derived cytokine-induced activation of uninfectedcells, as not all cells (�50 to 60%) were infected by DENV at thetime of the analysis (2 days postinfection at an MOI of 5; Fig. 1A).It is worth noting that the DENV E antigen detectable on thesurface of some MPs might be due to the binding of specific anti-bodies to virions that attach to the MP surface. Our preliminarydata showed that infected cell-derived MPs were positive forDENV RNA, as detected by nested RT-PCR, and contained infec-tious activity which was reduced by half after washing the isolatedMPs with acidic (pH 3.0) glycine buffer (data not shown). It is alsopossible that non-virion-associated E protein might be expressedon the surface of infected cells and thus may be incorporated intoMPs generated from these cells. This explanation is supported bythe study of Ng and Corner (24), as they found by immunoelec-tron microscopy that DENV E was located in clumps on theplasma membrane of infected cells without an association withDENV virions. Likewise, NS1 has been shown to incorporate intothe plasma membrane of infected cells via a glycosylphosphatidy-linositol linkage or by lipid raft association (26, 27). Additionally,secreted soluble NS1 can bind back to the plasma membrane ofuninfected cells through the interaction with specific sulfated gly-cosaminoglycans (28). These phenomena could contribute to thepositivity of NS1 detection on the surface of MPs. Taken together,DENV infection resulted in the shedding of MPs of various sizesthat may or may not harbor viral proteins on their surfaces.

Circulating MPs in dengue patients. We next measured andanalyzed the cellular origins of circulating MPs in the blood ofDENV-infected patients. Flow cytometric analysis of circulatingAnV� MPs was performed in a blinded manner. MPs were iden-tified according to their standard size and AnV labeling as de-scribed above. Demographic information on all subjects enrolledin this study is shown in Table 1, and a summary of the infectingDENV serotype in patients is provided in Materials and Methods.The demographics were not significantly different among the pa-tient groups. To our surprise, the total counts of AnV� MPs weresignificantly reduced in DENV-infected patients compared withthose in the other febrile illness (OFI) group during the febrileillness, except on day �2 (Fig. 4A; Table 2). No significant differ-ences in the results for paired convalescent-phase samples col-lected after defervescence (day 14 and day 60) were observedamong the groups (Fig. 4A; Table 2).

We further went on to identify the cellular phenotypes of thosecirculating MPs using a panel of specific MAbs to CD markersbelonging to platelets (CD41a), erythrocytes (CD235), monocytes(CD14), granulocytes (CD66), T cells (CD3), B cells (CD19), andNK cells (CD56). Unexpectedly, only CD41a� MPs and CD235�

MPs made up the two major populations of AnV� MPs in theblood of acutely ill patients (50 to 75% of total AnV� MPs), whilethe levels of MPs generated from other cell types, especially thosethat have been shown to be targets for DENV replication in vivo,including monocytes/macrophages (29) and lymphocytes (30),were remarkably low (data not shown).

Similar to the total AnV� MP levels, the absolute numbers ofCD41a� (platelet-derived) MPs in the DHF group were signifi-cantly lower than those in the DF and OFI groups on almost everydisease day during the acute phase (Fig. 4B; Table 3). Interestingly,we observed a rise in the number of CD235� (red blood cell-derived) MPs starting as early as day �2 in the DHF group, and

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

the increased numbers reached statistical significance at day �1;the numbers were still significantly higher than those in the othergroups through day 14 (Fig. 4C; Table 4). At day 60 after deferves-cence, the CD235� MP levels were comparable among all groups(Fig. 4C; Table 4). Importantly, during febrile illness the percent-age of CD235� MPs in the DHF group was significantly higherthan those of the DF and OFI groups, whereas the percentage ofCD41� MPs in the DHF was significantly reduced compared withthose of other groups. However, the percentages of both CD235�

and CD41a� MPs were undistinguishable among all groups dur-ing the convalescent phase (Fig. 4D and E).

The number of platelets was also monitored during the acutephase of infection until recovery at day 14. As expected, the plate-let count in the DHF group was distinctly reduced during theacute phase of illness (Fig. 4F; Table 5), consistent with the lowerabsolute number (Fig. 4B; Table 3) and percentage (Fig. 4D) ofCD41a� MPs compared with those in the DF and OFI groups.Platelet numbers returned to within the normal range at 2 weeksafter defervescence (Fig. 4F; Table 5). To test whether the increaseof CD235� MPs exclusively observed in the DHF group wascaused by an increase in the concentration of RBCs, RBC levels inall groups were analyzed over time. Unlike platelets, no statisti-cally significant difference in RBC levels was observed between theDHF and DF groups during the acute phase of illness (day �2 today 1) or during the convalescent phase (day 14) (Fig. 4G; Table6). The elevated RBC levels in DHF patients, especially within theperiod from 24 h before to 24 h after defervescence (days �1, 0,and 1), were likely due to plasma leakage, thereby resulting in arelative rise in the concentration of RBCs due to hemoconcentra-tion. In this study, all DHF patients had clinical evidence of plasmaleakage, including a �20% rise of hematocrit and/or the accumu-lation of fluid in pleural or abdominal cavities, the two majorWHO criteria for DHF (2, 3).

A simple linear regression showed that the log10-transformedplot of platelet counts had a strong positive correlation with the

log10-transformed level of CD41a� MPs (R � 0.702, 95% confi-dence interval [CI] � 0.627 to 0.764, n � 212, P 0.0001; Fig. 4H)These data suggest that the lower platelet numbers (thrombocy-topenia) induced by DENV may partly contribute to the reducedCD41a� MP plasma levels observed in DF cases and especially inDHF cases (Fig. 4B and D). Conversely, there was a weak positivecorrelation between the log10-transformed level of CD235� MPsand the log10-transformed level of RBC counts (R � 0.323, 95%CI � 0.197 to 0.438, n � 212, P 0.0001; Fig. 4I), indicating thatthe increased concentration of RBCs (or hematocrit levels) due toplasma leakage did not primarily account for the highly elevatedCD235� MP levels detectable in DHF patients (Fig. 4C and E).Strikingly, the phenotypic characterization of circulating AnV�

MPs demonstrated that CD235� MPs constituted the major pop-ulation in the blood of the DHF group (�50% of total AnV� MPs)during the acute phase of illness, whereas during the convalescentphase, CD41a� MPs were the most abundant AnV� MPs (�65%)in the blood (Fig. 4J), and this result was similar for all groups.Indeed, CD41a� MPs are the major MP population in the blood-stream of healthy individuals (5).

Elevation of CD235� MPs in DENV infection is associatedwith DHF: a potential novel biomarker for plasma leakage. Theabsolute counts and percentages of total AnV� MPs, CD41a�

MPs, and CD235� MPs were assessed to determine whether theywere affected by disease severity in patients using a mixed model ofmultivariate statistical analysis because there were some missingdata at some time points. Only data for patients with at least 3consecutive measurements were included in the analysis. This re-sulted in data for 31 patients (17 DHF, 10 DF, and 4 OFI patients)and 107 observations for all parameters. The predicted total levelsof AnV� MPs, CD41a� MPs, and CD235� MPs were obtainedfrom the quadratic mixed model against time (Fig. 5B, D, and F).For comparison, the mean and standard deviation (SD) of eachvariable over time, based on all available data, are also displayed(Fig. 5A, C, and E). DENV infection, but not disease severity,

FIG 4 Circulating MP levels in DENV-infected patients. Platelet-poor plasma was obtained from patients with DF, DHF, and OFIs (acute febrile diseases otherthan dengue) on different disease days during the acute phase (days �2, �1, 0, and 1) and the convalescent phase (days 14 and 60). Day 0 (defervescence) wasdefined as the calendar day on which the patient’s temperature fell and stayed below 37.8°C. (A to G) The absolute numbers of AnV� MPs (A) and AnV� MPsexpressing specific platelet marker CD41a (CD41a� MPs) (B) and erythrocyte marker CD235 (CD235� MPs) (C) and the percentages of CD41a� MPs (D) andCD235� MPs (E) from total AnV� MPs in the plasma of DF, DHF, and OFI patients were quantified by immunofluorescent labeling and flow cytometric analysis.(F) Numbers of platelets measured in individual patients from each group on each day over the course of the acute phase and 14 days after defervescence. Eachdot represents the absolute number (A to C) or the percentage (D and E) of total AnV� MPs (A) or AnV� MPs derived from platelets (B and D) or erythrocytes(C and E) and the numbers of platelets (F) and RBCs (G) for an individual patient (n � 3 to 15 patients on each disease day). Solid bars indicate the mean.Asterisks denote the means that are statistically significantly different between the indicated groups of patients (*, P 0.05; **, P 0.001; ***, P 0.0001). (Hand I) The correlation coefficients between the absolute counts of CD41a� MPs and platelet numbers (H) and CD235� MPs and RBC counts (I) were calculated.The linear regression, correlation coefficient, and P value are presented in the graphs. (J) The percentages of AnV� MPs expressing CD41a�, CD235�, and otherCD markers (CD41a� CD235�) in the plasma of OFI, DF, and DHF patients are displayed by disease day.

TABLE 2 Absolute number of AnV� MPs in three groups of patientsa

Day

Total AnV� MPs (mean SD no. of events/�l) P value

OFI patients DF patients DHF patients OFI vs DF patients OFI vs DHF patients DF vs DHF patients

�2 7,027 1,973 5,202 1,068 5,668 774 0.393 0.458 0.744�1 11,060 2,863 4,800 761 4,606 947 0.007 0.011 0.8790 7,429 1,385 4,434 546 4,324 648 0.021 0.031 0.9011 16,370 7,583 5,110 560 3,468 413 0.031 0.004 0.02314 20,410 6,650 14,380 2,629 13,940 2,701 0.317 0.285 0.90660 ND 7,713 1,326 9,604 1,580 ND ND 0.300a MPs, microparticles; OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; ND, not done.

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

resulted in the reduction of the total levels of AnV� MPs (DFversus OFI patients, P � 0.022; DHF versus OFI patients, P �0.002; DF versus DHF patients, P � 0.237; Fig. 5B). However, incontrast to the total levels of AnV� MPs, which were affected onlyby DENV infection, the level of CD41a� MPs was significantlyaffected by both DENV infection and disease severity (DF versusOFI patients, P � 0.003; DHF versus OFI patients, P 0.0001; DFversus DHF patients, P � 0.031; Fig. 5D). Furthermore, the levelof CD41a� MPs was significantly lower in DHF patients than inDF patients on every disease day except day �1 (day �2, P �0.009; day �1, P � 0.062; day 0, P � 0.023; day �1, P 0.001;Fig. 5D).

Interestingly, the level of CD235� MPs was elevated only inDHF patients, while there was no difference in the level ofCD235� MPs between DF and OFI patients (DF versus DHF pa-tients, P 0.001; DF versus OFI patients, P � 0.307; DHF versusOFI patients, P 0.001; Fig. 5F). The elevation of CD235� MPs inDHF patients was found on every disease day during the acutephase of illness (for DF versus DHF patients, day �2, P � 0.005;day �1, P 0.0001; day 0, P 0.001; day �1, P � 0.032; for DHFversus OFI patients, day �2, P � 0.019; day �1, P � 0.010; day 0,P 0.001; day �1, P 0.0001; Fig. 5F). Notably, the levels ofCD235� MPs in DHF patients remained relatively constant dur-ing the course of the acute phase of illness (for DHF patients, day�2 versus day �1, P � 0.118; day �1 versus day 0, P � 0.951; day0 versus day �1, P � 0.146; Fig. 5F).

The percentages of CD41a� and CD235� MPs among allAnV� MPs was also evaluated by mixed model analysis (Fig. 5Hand J). The mean and standard deviation of both parameters overtime are also displayed (Fig. 5G and I). Similar to the absolutecounts, the percentage of CD41a� MPs was affected by both dis-ease severity and infection with DENV (DF versus DHF patients,P � 0.039; DF versus OFI patients, P 0.0001; DHF versus OFIpatients, P 0.0001; Fig. 5H), and differences between the DF andDHF groups were found on every disease day (day �2, P � 0.002;

day �1, P � 0.020; day 0, P � 0.032; day �1, P � 0.009; Fig. 5H).While DHF patients had remarkably low platelet-derived MP lev-els, the percentage of CD235� MPs, on the other hand, was exclu-sively elevated in the DHF group (DF versus DHF patients, P 0.0001; DF versus OFI patients, P � 0.168; DHF versus OFI pa-tients, P 0.0001; Fig. 5J) and significantly different from that inthe other groups on every disease day during the acute phase ofillness (for DF versus DHF patients, day �2, P 0.0001; day �1,P 0.0001; day 0, P 0.0001; day �1, P 0.0001; for DHF versusOFI patients, day �2, P 0.001; day �1, P 0.0001; day 0, P 0.0001; day �1, P 0.0001; Fig. 5J). The percentage of CD235�

MPs in DHF patients was relatively constant during the course ofillness (for DHF patients, day �2 versus day �1, P � 0.238; day�1 versus day 0, P � 0.131; day 0 versus day �1, P � 0.875;Fig. 5J).

The fact that the increase in CD235� MPs was observed inDHF patients as early as day �2 and CD235� MP levels wereconstantly elevated during the febrile phase led us to reason thatthe earliest single measurement of CD235� MPs might be used asa biomarker to predict the severity of illness in dengue patients orthe development of DHF. To test this, the earliest measurements(median disease day of sample collection � day �2; range � day�6 to day �1; Table 7) of the absolute count and percentage ofCD235� MPs prior to defervescence were collected from all pa-tient groups and assessed through receiver operating characteris-tic (ROC) analysis. Interestingly, the areas under the ROC curves(AUCs) indicated a good prediction performance, with AUCs be-ing equal to 0.824 (95% CI � 0.693 to 0.954) and 0.903 (95% CI �0.808 to 0.997) for the absolute count (Fig. 5O) and percentage(Fig. 5P) of CD235� MPs, respectively. If a cutoff absolute countof CD235� MPs was set at 2.982 on a log10 scale, it would predictthe development of DHF with 94.7% sensitivity, 60.0% specificity,a 69.2% positive predictive value, and a 92.3% negative predictivevalue. If a cutoff percentage of CD235� MPs was set at 29.67%, itwould predict severity with 84.2% sensitivity, 80.0% specificity, an

TABLE 3 Absolute number of AnV� MPs expressing the specific platelet marker CD41a in three groups of patientsa

Day

CD41a� MPs (mean SD no. of events/�l) P value


�2 3,503 926 2,911 611 1,485 337 0.589 0.021 0.042�1 7,291 2,414 2,637 695 1,299 329 0.018 0.0002 0.0720 4,784 1,320 2,041 429 1,117 221 0.049 0.0001 0.0491 11,920 6,238 4,182 1,067 811.2 126 0.061 0.002 0.00214 15,060 5,178 9,595 2,476 10,280 2,155 0.288 0.317 0.83460 ND 6,399 1,318 7,696 1,465 ND ND 0.531a MPs, microparticles; OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; ND, not done.

TABLE 4 Absolute number of AnV� MPs expressing the specific erythrocyte marker CD235 in three groups of patientsa

Day

CD235� MPs (mean SD no. of events/�l) P value


�2 876.3 217 1,138 216 2,852 928 0.449 0.199 0.134�1 1,084 218 1,946 208 1,946 208 0.372 0.056 0.00020 534.8 123 1,065 159 2,167 260 0.053 0.001 0.0021 857.3 288 934.7 147 2,092 343 0.795 0.076 0.01214 529.9 104 752.9 104 1,221 185 0.029 0.029 0.04260 ND 731.5 92 926.4 181 ND ND 0.355a MPs, microparticles; OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; ND, not done.




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

84.2% positive predictive value, and an 87.5% negative predictivevalue. Because of the high negative predictive values of the abso-lute count and the percentage of CD235� MPs at the cutoffs de-scribed above, the measurement of CD235� MPs levels at the earlyfebrile phase of dengue virus infection may potentially be used toscreen out a patient with mild disease (DF) who may not requirehospitalization.

The changes of the other parameters (RBC and platelet counts)over time were also analyzed with mixed-effects models (Fig. 5Land N). The mean and standard deviation of each parameter overtime were also displayed (Fig. 5K and M). Platelet counts wereaffected by disease severity and DENV infection (DF versus DHFpatients, P 0.001; DF versus OFI patients, P � 0.009; DHFversus OFI patients, P 0.0001; Fig. 5L). However, on day �2 theplatelet count was not significantly different among the three pa-tient groups (DF versus DHF patients, P � 0.102; DF versus OFIpatients, P � 0.508; DHF versus OFI patients, P � 0.065; Fig. 5L).The platelet counts of DHF patients started to be significantlylower than those of the other groups on day �1 (DF versus DHFpatients, P � 0.002; DHF versus OFI patients, P 0.0001), whilethe platelet counts of DF patients started to differentiate fromthose of OFI patients on day 0 (P � 0.006). There was no differ-ence in the dynamics of RBC counts throughout the acute phase,except between DHF and OFI patients (DF versus DHF patients,P � 0.349; DF versus OFI patients, P � 0.131; DHF versus OFIpatients, P � 0.026; Fig. 5N).

Circulating platelet-derived MPs and red blood cell-derivedMPs carry DENV antigens on their surfaces. In vitro, DENV-infected cells generated MPs that contained E and NS1 antigens ontheir surfaces (Fig. 3). To evaluate the relevance of these findingsin dengue patients, circulating CD41a� MPs and CD235� MPswere analyzed for DENV E and NS1 antigen content by immuno-fluorescence staining (Fig. 6). Interestingly, the absolute numbersof NS1-positive (NS1�) CD235� MPs but not the absolute num-bers of E-positive (E�) CD235� MPs (Fig. 6C) were significantly

higher in the DHF group during the acute phase (for day �1, DHFversus DF patients, P � 0.012; DHF versus OFI patients, P �0.048; for day 0, DHF versus DF patients, P � 0.052; DHF versusOFI patients, P � 0.031; for day 1, DHF versus DF patients, P �0.008; DHF versus OFI patients, P � 0.037; Fig. 6D). However,there was no significant difference in the absolute counts ofCD41a� MPs carrying DENV E (Fig. 6A) or the absolute counts ofCD41a� MPs carrying NS1 (Fig. 6B) in the DHF group from thosein DF and OFI patients.

Complement activation of NS1–anti-NS1 ICs triggers MPshedding from red blood cells. Next, we explored why and howRBCs of DHF patients shed more MPs, as shown by the elevationof AnV� CD235� MPs (Fig. 4 and 5). To maintain homeostasis,one important function of RBCs is to trap potentially harmfulimmune complexes (ICs) in the circulation. Complement recep-tor type 1 (CR1) on RBCs binds to complement fragment-op-sonized ICs and transfers them to mononuclear phagocytes in theliver and spleen (31, 32). Circulating ICs have been detected inDHF patient serum during the acute phase (33, 34). Additionally,high levels of NS1 have been found in dengue patients, and theywere positively correlated with disease severity (16, 35). We thusreasoned that, in patients with DHF/DSS, which is mostly associ-ated with a secondary infection (3), circulating ICs formed be-tween the virion or NS1 and their specific antibodies should in-crease dramatically due to the presence of an extensive amount ofviral antigens in the bloodstream (viremia and NS1 antigenemia)together with a rapid rise in the levels of antivirion (36) and anti-NS1 (37) antibodies, followed by an anamnestic secondary im-mune response. High levels of circulating ICs in DENV infectionmay overburden the RBCs’ function of IC clearance and, conse-quently, trigger those RBCs to shed MPs. We hypothesized thatRBC vesiculation may occur either during the processes of im-mune adherence (binding of complement-opsonized ICs to CR1on RBCs) or during the transfer of ICs from RBCs to phagocytic

TABLE 5 Number of circulating platelets in three groups of patientsa

Day

Mean SD platelet count (103/�l) P value


�2 169 14 172 20 124 15 0.917 0.146 0.072�1 202 14 133 18 66 9 0.053 0.0001 0.0010 199 10 113 16 44 7 0.003 0.0001 0.00021 233 16 94 13 43 8 0.0001 0.0001 0.00214 406 29 463 30 401 26 0.248 0.912 0.13860 ND ND ND ND ND NDa OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; ND, not done.

TABLE 6 Number of circulating RBCs in three groups of patientsa

Day

Mean SD RBC counts (106/�l) P value


�2 4.8 0.2 5.3 0.3 5.2 0.1 0.185 0.103 0.817�1 4.6 0.1 5.2 0.2 5.3 0.1 0.158 0.004 0.5240 4.6 0.1 5.2 0.2 5.5 0.1 0.059 0.0005 0.2211 4.5 0.1 5.3 0.2 5.5 0.1 0.017 0.002 0.32614 4.7 0.1 5.1 0.2 4.9 0.1 0.152 0.105 0.49260 ND ND ND ND ND NDa RBC, red blood cell; OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; ND, not done.

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

cells in the liver and spleen, thereby resulting in increased RBC-derived MPs in DHF patients (Fig. 4 and 5).

To test this hypothesis, we developed an in vitro model of ICtransfer and investigated whether this process triggers RBCs to

shed MPs (Fig. 7). Interestingly, incubation of RBCs with ICs(consisting of purified NS1 plus IgG from pooled dengue patientimmune serum [PCS]) and complement led to not only the bind-ing of C3d-opsonized ICs to CR1 (Fig. 7B) but also a slight in-

FIG 5 Magnitude and dynamics of platelet, RBC, and MP levels over the course of the acute phase of DENV infection. Patients were grouped according to diseaseseverity (DF, DHF, and OFI). The mean and standard error absolute counts of AnV� MPs from day �1 to day �1 during the acute phase are shown as log10-transformedvalues for AnV� MPs (A), CD41a� MPs (C), CD235� MPs (E), platelets (K), and RBCs (M). The mean and standard error percentages of CD41a� MPs (G) and CD235�

MPs (I) are also shown. Predicted absolute counts of AnV� MPs (B), CD41a� MPs (D), CD235� MPs (F), platelets (L), and RBCs (N) and the percentage of CD41a�

MPs (H) and CD235� MPs (J) were also analyzed by the use of quadratic mixed-effects models. (O and P) The usefulness of CD235� MPs as a biomarker to predict theseverity of disease in dengue patients was evaluated using the area under the ROC curve (AUC). The AUC for the absolute count of CD235� MPs was equal to 0.824 (95%CI � 0.693 to 0.954) (O), and the AUC for the percentage of CD235� MPs was equal to 0.903 (95% CI � 0.808 to 0.997) (P).

TABLE 7 Characteristics of the earliest samples from each patient included in the ROC analysis of CD235� MPs as a biomarkera

Characteristic

Result for the following patients:

OFI patients (n � 6) DF patients (n � 14) DHF patients (n � 19)

Absolute CD235� MP count 743.7 (691.6, 1,099.5) 892.3 (533.0, 1,324.8) 1,708.7 (1,147.2, 2,340.8)Log10 (absolute CD235� MP count) 2.87 (2.84, 3.03) 2.95 (2.73, 3.12) 3.23 (3.06, 3.37)% CD235� MPs 7.98 (7.18, 10.05) 19.97 (11.70, 28.97) 45.45 (35.14, 65.10)Day of disease that samples were collected �2 (�2.75, �1.25) 2 (�2, �1) �2 (�2, �1)a Data are presented as the median (25th, 75th percentile). OFI, other febrile illness; DF, dengue fever; DHF, dengue hemorrhagic fever; MPs, microparticles.




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

crease of CD235� MP generation (Fig. 7E). Inhibition of comple-ment activation by EDTA or pretreatment of RBCs with function-blocking anti-CR1 MAb clone 3D9 (38), but not with an isotypecontrol, prior to the addition of a mixture of NS1 and PCS abro-gated the binding of ICs to CR1, as determined by negative C3dstaining (Fig. 7A and B) and MP shedding from RBCs (P � 0.013;Fig. 7E). Low levels of CD235� MP generation were also observedwhen RBCs were incubated with serum alone or with a mixture ofNS1 and purified IgG from pooled DENV-nonimmune donors(PND) (Fig. 7E).

To mimic the phenomenon of IC transfer, autologous periph-eral blood mononuclear cells (PBMCs) were added to RBCs car-rying complement-opsonized ICs. As expected, C3d-opsonizedNS1–anti-NS1 ICs deposited on CR1 of RBCs were completelyremoved after PBMC exposure (Fig. 7D). More importantly, thisprocess dramatically augmented MP shedding from RBCs (Fig.7F). However, the MP shedding from PBMCs was not changed, asdetermined by the detection of antibodies against common leu-kocyte surface marker CD45 (Fig. 7G). Taken together, these re-sults suggest that removal of complement-opsonized ICs boundto RBCs via CR1 by mononuclear cells can trigger RBCs to shedMPs and might explain the elevated CD235� MP levels in DHFpatients.

A decrease in platelet-derived MPs in dengue patients is as-sociated with a bleeding tendency. Hemorrhagic manifestationsare commonly associated with DENV infection (2). They can po-tentially complicate clinical management and cause significantmorbidity and mortality, especially when occurring massively inthe gastrointestinal tract (3). Despite being more frequently asso-ciated with DHF, a bleeding tendency can independently occur inthe absence of plasma leakage, i.e., in DF patients (3). Many fac-tors have been shown to contribute to bleeding in DENV infec-tion, including thrombocytopenia, reduced thrombin formation,and increased fibrinolysis activity (39, 40), yet the pathogenesis of

bleeding complications is not fully understood. To further inves-tigate whether circulating MPs influence the outcome of bleeding,all dengue virus-infected patients were separated into two groupsaccording to their history of bleeding, i.e., a bleeding group (totalnumber of patients � 18; number of DHF patients � 13, numberof DF patients � 5) and a no-bleeding group (total number ofpatients � 25; number of DHF patients � 11, number of DFpatients � 14), regardless of plasma leakage. The dengue patientswho had evidence of bleeding on any day during febrile illnesswere categorized into the bleeding group. Bleeding episodes in-cluded nose or gum bleeding, ecchymosis, gastrointestinal hem-orrhage (blood in vomitus or in stool), and hypermenorrhea.Platelet counts (Fig. 8A and B), total AnV� MP levels (Fig. 8D andE), CD41a� MP levels (Fig. 8G and H), and CD235� MP levels(Fig. 8J and K) on each disease day were plotted according tobleeding episodes. Declining levels of platelets, total AnV� MPs,and CD41a� MPs were displayed more prominently in the bleed-ing group over the course of the acute phase of illness, reaching thelowest values on day �1, and were statistically significantly differ-ent between the two groups (platelets, P � 0.021; total AnV� MPs,P � 0.013; CD41a� MPs, P � 0.008). On the contrary, in theno-bleeding group, the levels of those parameters started rising onday �1 and returned to normal on day 14, when both groupsexhibited similar levels (P � 0.05). Interestingly, while the levels ofCD235� MPs were clearly different between DF and DHF patients(Fig. 4C and 5E and F), it was not affected by hemorrhagic mani-festations (Fig. 8J and K).

To obtain more accurate results for the analysis of these con-tinuous variables over time, mixed-effects models were examinedto assess whether the overall changes in the levels of these factorswere related to the outcome of bleeding (Fig. 8C, F, I, and L).Using the quadratic models, 11 bleeding and 16 no-bleeding pa-tients whose data were available on at least three consecutive dayswere included in the analysis. In the models, the outcome of bleed-

FIG 6 Circulating MPs derived from platelets and red blood cells harbor DENV antigens on their surfaces. Platelet-derived MPs (AnV� CD41a�) anderythrocyte-derived MPs (AnV� CD235�) were further analyzed for DENV E (A and C) and NS1 (B and D) expression on their surfaces. The absolute numbersof E� CD41a� MPs (A), NS1� CD41a� MPs (B), E� CD235� MPs (C), and NS1� CD235� MPs (D) were quantified by immunofluorescent labeling and flowcytometric analysis. Each dot represents the absolute number of each population of MPs for an individual patient on different disease days. Solid bars indicatethe means. Asterisks denote the means that are statistically significantly different between the indicated groups of patients (*, P 0.05; **, P 0.001).

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

ing and the disease day were treated as fixed effects, while each casewas treated as a random effect. Interestingly, the pattern of thechanges in platelet counts over time was not significantly differentbetween dengue patients with bleeding and those without bleed-ing (P � 0.132; Fig. 8C). Likewise, bleeding status did not affectthe levels of total AnV� MPs during the course of the acute phaseof illness (P � 0.272; Fig. 8F). However, similar to the analysis ofthe mean (Fig. 8A and D), the multivariate analysis showed thatplatelet counts (P � 0.032; Fig. 8C) and total AnV� MP levels (P �0.003; Fig. 8F) were significantly different between the two groupsonly on day �1. Interestingly, the reduction of CD41a� MP levelsover time in the bleeding group was significantly different fromthat in the patients without bleeding (P � 0.043; Fig. 8I). More-over, the levels of CD41a� MPs could be used to differentiatebetween the two groups on almost every disease day except day �1(day �2, P � 0.047; day �1, P � 0.176; day 0, P � 0.035; day �1,P 0.0001; Fig. 8I). Nonetheless, no differences in the levels ofCD235� MPs were observed among the different groups of pa-tients (P � 0.271; Fig. 8L). Overall, these results imply that thelevels of CD41a� MPs but not those of CD235� MPs might beassociated with the outcome of bleeding.

The surface of CD41a� MPs, the most abundant MPs inhealthy human blood (5), has approximately 50- to 100-fold moreprocoagulant activity than the surface of activated platelets (41).They provide a low-grade procoagulant state to maintain homeo-stasis in the blood of healthy individuals (5, 42). We reasoned thatunder circumstances in which the number of platelets signifi-cantly decreases, like in dengue, a small change in the level ofcirculating CD41a� MPs, which could also be caused by factorsother than thrombocytopenia, could largely influence the ten-dency to bleed. To address this, blood samples from each patienttaken on the day of the platelet level nadir were examined to de-termine whether decreased levels of CD41a� MPs directly affectedprocoagulant MP activity in the blood of those patients who hadone or more bleeding episodes during the acute phase of illness. Asexpected, procoagulant MP activity (P � 0.006; Fig. 8Q) alongwith total AnV� MP levels (P � 0.027; Fig. 8N) and CD41a� MPlevels (P � 0.005; Fig. 8O), but not CD235� MP levels (P � 0.807;Fig. 8P), were significantly lower in the bleeding group than inpatients without any bleeding episodes during the acute phase offebrile illness. Of note, the platelet level nadir was also lower in thebleeding group, but the difference did not reach statistical signif-

FIG 7 Complement activation of NS1–anti-NS1 immune complexes triggers MP shedding from red blood cells. (A and B) Complement C3 deposition on RBCsurfaces is dependent on the interaction between C3 cleavage fragments and CR1. RBCs were incubated with isotype control antibody (black histograms) oranti-CR1 MAb clone 3D9 (gray histograms) prior to the addition of a mixture of NS1, purified IgG from PCS, and 40 mM EDTA to inhibit complement activation(A) or a mixture of NS1 and PCS (B). (C to G) Immune complexes bound to RBCs were removed by the addition of autologous PBMCs, thereby resulting in MPshedding from RBCs. Purified RBCs were incubated with 10% fresh human serum alone (None), serum with NS1 plus purified IgG from pooled non-denguepatient serum (NS1 � PND), or serum with NS1 plus purified IgG from PCS to allow complement activation (NS1 � PCS). EDTA was used to block complementactivation (NS1 � PCS � EDTA). (E) Supernatants were collected and analyzed for CD235� MPs. Cells were washed and were subsequently incubated withautologous PBMCs. (F and G) PBMC-induced CD235� MP (F) and CD45� MP (G) shedding was enumerated. (C and D) Before and after PBMC exposure,RBCs were analyzed for C3 deposition by immunofluorescence staining with FITC-conjugated anti-C3d MAb. Data are the mean SD from four to sixindependent experiments. Asterisks denote the means that are statistically significantly different between the indicated groups of treatments (*, P 0.05; **, P 0.001).




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

icance (P � 0.054; Fig. 8M). All parameters measured during theconvalescent phase, including procoagulant MP activity, were in-distinguishable between the two groups (Fig. 8A, D, G, J, and Q).The levels of CD41a� MPs might reflect the procoagulant MPactivity in the blood of dengue patients, especially in the bleedinggroup, as a weak positive correlation was observed (for the bleed-ing group, R � 0.436 and P 0.001 [Fig. 8R]; for the no-bleedinggroup, R � 0.252 and P � 0.025 [Fig. 8S]), but this association wasnot observed with CD235� MPs (for the bleeding group, R ��0.087 and P � 0.492 [Fig. 8T]; for the no-bleeding group, R ��0.067 and P � 0.559 [Fig. 8U]). These data suggest that platelet-derived MPs but not red blood cell-derived MPs might contributeto the bleeding tendency that occurs in dengue patients.

DISCUSSION

MPs have been recognized to play important roles in many bio-logical processes. In response to various stimuli, including virusinfection, cells shed parts of their plasma membrane and cytoplas-mic content into small vesicular structures with different sizes,characteristics, and functions. Herein we show that DENV in-fected various kinds of susceptible cells in vitro, resulting in apop-totic death and MP release. These MPs harbored viral E and NS1on their surfaces. Electron microscopy techniques, including neg-ative staining and immunogold labeling, revealed greater detailabout the MPs generated from hepatocytes (one of the major tar-get cells for DENV replication in vivo [43]), which were morpho-logically distinct from the virions produced in the same culturesystem. Nevertheless, the molecular mechanism underlying MPgeneration after DENV infection is still unknown, that is, whetherit is directly linked to the process of cells undergoing apoptosis orit is a consequence of cell activation in response to viral infectionleading to increased intracellular calcium, which in turn causesmembrane vesiculation.

It is still questionable as to what could be the significance of thisphenomenon in DENV infection. A previous study has also de-scribed membrane vesicular structures within the cytoplasm andin the culture supernatants of DENV-infected cells and has pro-posed their function as being sites of virus replication (44). Alter-natively, DENV may use these secreted MPs as a tool to escapefrom host immune surveillance and disseminate infection. In fact,this immune evasion strategy via secreted MPs has been previ-ously described for other viruses. Microvesicles released from in-fected cells facilitate the spread of human immunodeficiency virusand herpes simplex virus to neighboring uninfected cells by trans-ferring either viral constituents or the host’s membrane proteinsrequired for viral entry to other null cells, thereby increasing thenumber of susceptible cells (45–47).

Changes in the levels and origin profiles of circulating MPshave been described in many pathological conditions. Our resultsfrom the ex vivo analysis of MPs in the circulation of DENV-infected patients were initially surprising. The total numbers ofMPs identified by AnV labeling of surface-exposed PS and plate-let-derived (AnV� CD41a�) MPs were decreased during the acutephase. Thrombocytopenia (in which the platelet number in bloodis decreased), a key clinical manifestation of DENV infection (2),may contribute to these surprising results. In support of this, DHFpatients whose platelet counts were significantly lower than thoseof DF patients, especially at the critical period (days 0 and �1; Fig.4F), also had significantly lower platelet-derived MP levels (Fig.4B). In fact, platelets are the major source of MPs in healthy indi-

viduals (5). In our study, a strong positive correlation between thenumbers of platelets and CD41a� MPs was observed (Fig. 4H).Notably, various factors could enhance MP shedding from plate-lets during the acute phase of DENV infection, such as plateletactivation (48), platelet apoptosis (49), and direct infection ofplatelets by DENV (50). Indeed, a small number of circulatingAnV� CD41a� MPs carried DENV E and NS1 antigens, similar tothe findings for infected cell-derived MPs, as demonstrated in ourin vitro model of DENV infection. However, the existence ofDENV antigen-carrying CD41a� MPs in the blood during theacute phase of illness (Fig. 6A and B) was not direct evidence forDENV infection of platelets or megakaryocytes (which also ex-press CD41a). Alternatively, Fc�II receptors on the surface ofplatelets might bind to DENV-antibody complexes, as previouslyshown by Wang et al. (51), or NS1-antibody complexes, resultingin shedding of E� or NS1� CD41a� MPs, respectively, withoutactual DENV infection.

In contrast to our findings, a previous study by Hottz et al.reported a higher number of platelet-derived MPs in dengue pa-tients than in healthy controls (52). This discrepancy may be dueto distinct characteristics of dengue patients between the twostudies, including age (children versus adults), the degree ofthrombocytopenia, and disease severity, as well as the method forMP enumeration. The protocol used by Hottz et al. (52) to obtainthe number of circulating CD41a� MPs per 100 platelets did notrepresent the absolute number of MPs in the patients’ circulation.Using this method of flow cytometric analysis, the plasma plateletconcentration could influence the number of MP counts; i.e., forsamples with higher platelet concentrations, like those fromhealthy controls, the amount of time required to acquire plateletcounts of 100 is less, and thus, MP counts are lower than those indengue patients, whose circulating platelet numbers are usuallylow during the acute phase of illness even when they have milddiseases (2). The present study provided more accurate informa-tion on the total circulating MP concentration by using TruCountbeads to obtain the absolute number of MPs per �l plasma (53).Moreover, comparative analysis among groups of patients (DFpatients versus DHF patients versus nondengue OFI patients) atmultiple time points has revealed similar dynamic changes be-tween platelet-derived MP levels (Fig. 4B) and platelet counts (Fig.4F), which are decreased over the course of the acute phase ofillness (for day �2 to day �1, DHF patients DF patients OFIpatients) and rise back to within normal limits during the recoveryperiod and are equal in DHF, DF, and OFI patients on day 14 andday 60.

The functional significance of decreased AnV� CD41� MPlevels in DENV infection may relate to the procoagulant propertyof the negatively charged PS exposed on MPs, which provides thebinding site for coagulation factors and serves as a platform forassembly of the prothombinase complex, resulting in the amplifi-cation of prothrombin cleavage and thrombin production (54).Circulating CD41a� MPs possess up to 100-fold more procoagu-lant activity than activated platelets (41) and help to maintainhomeostasis in healthy individuals (5, 42). Increased amounts ofcirculating platelet-derived MPs are linked to hypercoagulablestates in several pathological conditions (4, 55–58). Conversely,patients with an inherited bleeding disorder, Scott syndrome, de-velop severe hemorrhagic manifestations due to a defect in scram-blase activity, leading to the lack of surface exposure of PS (thusresulting in an inadequate ability to promote a procoagulable

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

state) and to the inability of platelets and other blood cells togenerate MPs (59). Another bleeding disorder that results in adecreased capacity of platelets to shed MPs has also been described(60). In the same line of thought, our results showed that the levelsof CD41a� MPs reflected the procoagulant activity of circulatingMPs, especially in the bleeding group. Therefore, together withthrombocytopenia, diminished total circulating AnV� MP levelsand especially AnV� CD41a� MP levels might also contribute toabnormal hemostasis in some cases of DENV infection.

Surprisingly, our ex vivo analysis of circulating MPs showed asignificant increase in the levels of AnV� CD235� MPs in DHFpatients (but not DF and OFI patients) during the acute phase(Fig. 4 and 5). However, the role of RBCs in DENV infection hasnot been previously described, and importantly, RBCs could notsupport DENV replication. Of note, the increase in AnV�

CD235� MP levels was not due to a difference in RBC levels,which indicates the degree of hemoconcentration. One of the keyfunctions of RBCs is to maintain homeostasis by trapping poten-tially harmful ICs which could deposit in tissues and induce in-jury. This is accomplished by complement opsonization, followedby binding to complement receptor type 1 (CR1), expressed onthe surface of RBCs, the so-called immune adherence process (31,61, 62). ICs bound to RBCs are then transferred to phagocytic cellsin the liver and spleen (31, 62). The finding that the level of NS1containing AnV� CD235� MPs was significantly increased inDHF patients (Fig. 6D) prompted us to hypothesize that ICs thatform between NS1 and its specific antibody are involved in the MPshedding process of RBCs.

We have shown that removal of complement-opsonized NS1–anti-NS1 ICs bound to erythrocytes by mononuclear cells trig-gered in vitro RBC-derived MP shedding. These results suggestthat the IC transfer mechanism might cause RBC vesiculation indengue patients. However, the precise cellular mechanisms in-volved in the transfer of RBC-bound ICs from RBCs to macro-phages or phagocytic cells and the generation of CD235� MPsduring this process remain to be elucidated. The transfer mecha-nism could be driven by a higher number of CR1 copies on themonocyte than on the RBCs (63) or by the close juxtaposition ofthe macrophage to the RBC-bound ICs that leads to macrophage-associated protease cleavage of CR1 and a release of the IC fromthe RBC (64). It is possible that instability of RBC membranes mayoccur during transfer of CR1-bound ICs from erythrocytes to ac-ceptor phagocytic cells via Fc receptors (64, 65) and thus triggerRBCs to shed MPs.

Our data on MP shedding from RBCs in response to the re-moval of surface-bound ICs by mononuclear cells point to thepossibility of increased circulating red blood cell-derived MPs(RMPs) in other IC diseases, such as systemic lupus erythemato-

sus (SLE). Increased levels of MPs from various cellular origins,including platelets, leukocytes, and endothelial cells, but not fromRBCs have previously been reported (66–69). The reason whyDHF patients but not SLE patients showed a significant increase inAnV� CD235� MP levels remains to be answered. The level ofcirculating MPs likely depends on the balance between their ratesof generation and clearance. It is possible that the transport of ICsfor clearance in the liver and spleen in dengue patients might bedefective due to the pathology of those organs directly caused byDENV infection. Indeed, a recent pathology study at autopsy oftissues obtained from patients who have died from DHF/DSS, ledby our group, has demonstrated that the liver and spleen are thetwo major organs found to be heavily infected (43). Alternatively,the amount and nature of ICs, including their size and whetherthey are soluble or cell associated, determine the magnitude andefficiency of complement fixation (70) and may also influence thedegree of MP shedding from RBCs. The marked inefficiency ofcomplement C5 cleavage relative to the efficiency of C3 cleavageand the inefficiency of C5b-9 generation during complement ac-tivation by soluble ICs compared with the efficiency of C5b-9generation by ICs on particles or cells likely result in the detectionof lower levels of C5a and soluble C5b-9 (SC5b-9) complexes inthe plasma of patients with ongoing, complement-consuming ICdiseases like SLE (33). In support of this idea, in contrast to SLEpatients, high levels of circulating SC5b-9 were observed inDENV-infected patients, especially in those with DHF/DSS (16),suggesting more efficient complement activation and possibly alsodifferences in the nature/characteristics of complement activatorsgenerated during the acute phase of DENV infection. Consistentwith these findings, it has been demonstrated that the large anti-gen-antibody complexes formed on the surface of infected cellsefficiently fix complement reaching the terminal pathway, as de-termined by the generation of fluid-phase SC5b-9 as well as thedeposition of C5b-9 membrane attack complexes on cell surfaces(16, 20). Taken together, these factors may also contribute to thediscrepancy between the RMP levels observed in dengue patientsand those observed in SLE patients.

The physiological roles of RMPs have not been well elucidated.Many physiological and pathological conditions have been shownto induce RMP generation. During their life span, RBCs lose ap-proximately 20% of their hemoglobin content through vesicle for-mation (71). Stored RBCs also lose their membrane integrity,thereby resulting in hemolysis and RMP generation that may con-tribute to complications associated with transfusion (66, 72, 73).Additionally, RMP formation could be triggered with differenttypes of stimuli, such as shear stress, complement attack via theinsertion of the terminal complement C5b-9 in the membrane ofRBCs, oxidative stress, and proapoptotic stimulations (74–76).

FIG 8 Blood levels of platelet-derived MPs but not red blood cell-derived MPs correspond to procoagulant MP activity and bleeding predisposition. DENV-infected patients with at least one episode of bleeding and patients with no mucosal bleeding during the admission were classified as the bleed and no-bleedgroups, respectively. Bleeding episodes included epistasis, nose bleeding, gastrointestinal hemorrhage, and hypermenorrhea. (A to Q) The numbers of platelets(A), total AnV� MPs (D), CD41a� MPs (G), and CD235� MPs (J) of DENV-infected patients with and without bleeding episodes during the acute phase andat day 14 after defervescence (convalescent phase) are depicted. The means and standard errors of log10-transformed values were analyzed for platelets (B), totalAnV� MPs (E), CD41a� MPs (H), and CD235� MPs (K). Predicted platelets (C), total AnV� MPs (F), CD41a� MPs (I), and CD235� MPs (L) in each groupwere also analyzed by the use of quadratic mixed-effects models. The levels of platelets (M), total AnV� MPs (N), CD41a� MPs (O), CD235� MPs (P), andprocoagulant MPs (Q) in DENV-infected patients with and without bleeding episodes on the day of the platelet count nadir were plotted. Asterisks denote themeans that are statistically significantly different between the indicated groups of patients (*, P 0.05; **, P 0.001; N.S., not significant). (R to U) Thecorrelation coefficients between the absolute counts of procoagulant MPs and the numbers of CD41� MPs (R and S) and CD235� MPs (T and U) in patients with(R and T) or without (S and U) bleeding episodes were calculated. The linear regression lines, correlation coefficients (R), and P values are presented in the graphs.

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://jvi.asm.org

http://jvi.asm.org/

Increased RMP levels have been documented in different patho-logical conditions, including sickle cell disease (77), hereditaryhemolytic anemia (78), paroxysmal nocturnal hemoglobinuria(73), atherosclerosis (79), and malaria (9, 80). RMPs might also beinvolved in immunomodulation by inhibiting macrophage re-lease of proinflammatory cytokines (81) or, conversely, acting aspotent stimulators of cells of the innate immune system (82). Nev-ertheless, the functional roles of RMPs in dengue require furtherinvestigation.

Although the biological process behind RMP elevation in DHFpatients is not yet fully understood, RMP elevation itself has astrong potential for use as a biomarker for differentiating poten-tially severe dengue disease (DHF) from milder dengue fever.During the early febrile phase, clinical signs and symptoms as wellas currently available laboratory tests cannot differentiate the twoclinical syndromes. We have shown that the elevation of RMPlevels occurred at a very early time point of the disease and theelevated RMP level remained constant throughout the acutephase. Consequently, an early measurement of RMP levels is po-tentially useful to identify patients who will subsequently developplasma leakage (DHF) as well as prevent patients with mild diseasefrom unnecessary hospitalization. This is important, as during adengue epidemic where limited resources are available, overhos-pitalization has been shown to contribute to a significant increasein morbidity and mortality, especially in countries where dengueis endemic (83, 84).

In summary, the present study has provided some evidence ofthe roles of MPs in dengue pathogenesis. MPs generated fromerythrocytes and platelets comprised two major populations inthe circulation of dengue patients. Elevated levels of RMPs directlycorrelated with DENV disease severity, whereas a significant de-crease in platelet-derived MP levels was associated with a bleedingtendency. Removal of complement-opsonized NS1–anti-NS1 ICsbound to erythrocytes by mononuclear cells triggered erythro-cytes to shed MPs in vitro, a process that might explain the in-creased RMP levels in severe dengue. RMPs are a potential bio-marker to predict the development of DHF.

ACKNOWLEDGMENTS

This work was supported by a Mahidol University research grant (to P.A.)The clinical cohorts were supported by the Office of the Higher EducationCommission and Mahidol University under the National Research Uni-versities Initiative (to P.M.). P.A. has been supported by a Siriraj Chal-ermprakiat grant and a research lecturer grant, Faculty of Medicine SirirajHospital, Mahidol University. N.P. and S.T. are Ph.D. scholars in theRoyal Golden Jubilee Ph.D. Program.

We thank Ali Amara for kindly supplying wild-type 293T cells and293T cells stably expressing TIM-1, Chunya Puttikhunt and WatcharaKasinrerk for providing anti-DENV NS1 MAb, Bunpote Siridechadilokfor advise and help in electron microscopy experiments, Prapat Suriya-phol for constructive comments on data analysis, and Richard Hauhartand John Atkinson for helpful discussion and critical reading of the man-uscript.

REFERENCES1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL,

Drake JM, Brownstein JS, Hoen AG, Sankoh O, Myers MF, George DB,Jaenisch T, Wint GR, Simmons CP, Scott TW, Farrar JJ, Hay SI. 2013.The global distribution and burden of dengue. Nature 496:504 –507. http://dx.doi.org/10.1038/nature12060.

2. World Health Organization. 1997. Dengue haemorrhagic fever: diagno-sis, treatment, prevention and control. World Health Organization, Ge-neva, Switzerland.

3. Nimmannitya S. 1987. Clinical spectrum and management of denguehaemorrhagic fever. Southeast Asian J Trop Med Public Health 18:392–397.

4. Piccin A, Murphy WG, Smith OP. 2007. Circulating microparticles:pathophysiology and clinical implications. Blood Rev 21:157–171. http://dx.doi.org/10.1016/j.blre.2006.09.001.

5. Berckmans RJ, Nieuwland R, Boing AN, Romijn FP, Hack CE, Sturk A.2001. Cell-derived microparticles circulate in healthy humans and sup-port low grade thrombin generation. Thromb Haemost 85:639 – 646.

6. Nieuwland R, Berckmans RJ, McGregor S, Boing AN, Romijn FP,Westendorp RG, Hack CE, Sturk A. 2000. Cellular origin and procoagu-lant properties of microparticles in meningococcal sepsis. Blood 95:930 –935.

7. Aupeix K, Hugel B, Martin T, Bischoff P, Lill H, Pasquali JL, FreyssinetJM. 1997. The significance of shed membrane particles during pro-grammed cell death in vitro, and in vivo, in HIV-1 infection. J Clin Invest99:1546 –1554. http://dx.doi.org/10.1172/JCI119317.

8. Combes V, Taylor TE, Juhan-Vague I, Mege JL, Mwenechanya J,Tembo M, Grau GE, Molyneux ME. 2004. Circulating endothelial mi-croparticles in Malawian children with severe falciparum malaria compli-cated with coma. JAMA 291:2542–2544. http://dx.doi.org/10.1001/jama.291.21.2542-b.

9. Nantakomol D, Dondorp AM, Krudsood S, Udomsangpetch R, Pat-tanapanyasat K, Combes V, Grau GE, White NJ, Viriyavejakul P, DayNP, Chotivanich K. 2011. Circulating red cell-derived microparticles inhuman malaria. J Infect Dis 203:700 –706. http://dx.doi.org/10.1093/infdis/jiq104.

10. Kornek M, Popov Y, Libermann TA, Afdhal NH, Schuppan D. 2011.Human T cell microparticles circulate in blood of hepatitis patients andinduce fibrolytic activation of hepatic stellate cells. Hepatology 53:230 –242. http://dx.doi.org/10.1002/hep.23999.

11. Burger D, Schock S, Thompson CS, Montezano AC, Hakim AM, TouyzRM. 2013. Microparticles: biomarkers and beyond. Clin Sci (Lond) 124:423– 441. http://dx.doi.org/10.1042/CS20120309.

12. Marin V, Kaplanski G, Gres S, Farnarier C, Bongrand P. 2001. Endo-thelial cell culture: protocol to obtain and cultivate human umbilical en-dothelial cells. J Immunol Methods 254:183–190. http://dx.doi.org/10.1016/S0022-1759(01)00408-2.

13. Puttikhunt C, Kasinrerk W, Srisa-ad, Duangchinda ST, Silakate W,Moonsom S, Sittisombut N, Malasit P. 2003. Production of anti-dengueNS1 monoclonal antibodies by DNA immunization. J Virol Methods 109:55– 61. http://dx.doi.org/10.1016/S0166-0934(03)00045-4.

14. Heijnen HF, Schiel AE, Fijnheer R, Geuze HJ, Sixma JJ. 1999. Activatedplatelets release two types of membrane vesicles: microvesicles by surfaceshedding and exosomes derived from exocytosis of multivesicular bodiesand alpha-granules. Blood 94:3791–3799.

15. Bhakdi SC, Ottinger A, Somsri S, Sratongno P, Pannadaporn P,Chimma P, Malasit P, Pattanapanyasat K, Neumann HP. 2010. Opti-mized high gradient magnetic separation for isolation of Plasmodium-infected red blood cells. Malar J 9:38. http://dx.doi.org/10.1186/1475-2875-9-38.

16. Avirutnan P, Punyadee N, Noisakran S, Komoltri C, Thiemmeca S,Auethavornanan K, Jairungsri A, Kanlaya R, Tangthawornchaikul N,Puttikhunt C, Pattanakitsakul SN, Yenchitsomanus PT, Mongkolsa-paya J, Kasinrerk W, Sittisombut N, Husmann M, Blettner M, Va-sanawathana S, Bhakdi S, Malasit P. 2006. Vascular leakage in severedengue virus infections: a potential role for the nonstructural viral proteinNS1 and complement. J Infect Dis 193:1078 –1088. http://dx.doi.org/10.1086/500949.

17. Lachmann PJ, Oldroyd RG, Milstein C, Wright BW. 1980. Three ratmonoclonal antibodies to human C3. Immunology 41:503–515.

18. Innis BL, Nisalak A, Nimmannitya S, Kusalerdchariya S, ChongswasdiV, Suntayakorn S, Puttisri P, Hoke CH. 1989. An enzyme-linked im-munosorbent assay to characterize dengue infections where dengue andJapanese encephalitis co-circulate. Am J Trop Med Hyg 40:418 – 427.

19. Yenchitsomanus PT, Sricharoen P, Jaruthasana I, Pattanakitsakul SN,Nitayaphan S, Mongkolsapaya J, Malasit P. 1996. Rapid detection andidentification of dengue viruses by polymerase chain reaction (PCR).Southeast Asian J Trop Med Public Health 27:228 –236.

20. Avirutnan P, Malasit P, Seliger B, Bhakdi S, Husmann M. 1998. Denguevirus infection of human endothelial cells leads to chemokine production,complement activation, and apoptosis. J Immunol 161:6338 – 6346.

21. Thongtan T, Panyim S, Smith DR. 2004. Apoptosis in dengue virus




icholas White

http://jvi.asm.org/

Dow

nloaded from

http://dx.doi.org/10.1038/nature12060

http://dx.doi.org/10.1038/nature12060

http://dx.doi.org/10.1016/j.blre.2006.09.001

http://dx.doi.org/10.1016/j.blre.2006.09.001

http://dx.doi.org/10.1172/JCI119317

http://dx.doi.org/10.1001/jama.291.21.2542-b

http://dx.doi.org/10.1001/jama.291.21.2542-b

http://dx.doi.org/10.1093/infdis/jiq104

http://dx.doi.org/10.1093/infdis/jiq104

http://dx.doi.org/10.1002/hep.23999

http://dx.doi.org/10.1042/CS20120309

http://dx.doi.org/10.1016/S0022-1759(01)00408-2

http://dx.doi.org/10.1016/S0022-1759(01)00408-2

http://dx.doi.org/10.1016/S0166-0934(03)00045-4

http://dx.doi.org/10.1186/1475-2875-9-38

http://dx.doi.org/10.1186/1475-2875-9-38

http://dx.doi.org/10.1086/500949

http://dx.doi.org/10.1086/500949

http://jvi.asm.org

http://jvi.asm.org/

infected liver cell lines HepG2 and Hep3B. J Med Virol 72:436 – 444. http://dx.doi.org/10.1002/jmv.20004.

22. Meertens L, Carnec X, Lecoin MP, Ramdasi R, Guivel-Benhassine F,Lew E, Lemke G, Schwartz O, Amara A. 2012. The TIM and TAMfamilies of phosphatidylserine receptors mediate dengue virus entry.Cell Host Microbe 12:544 –557. http://dx.doi.org/10.1016/j.chom.2012.08.009.

23. Zhang X, Sheng J, Plevka P, Kuhn RJ, Diamond MS, Rossmann MG.2013. Dengue structure differs at the temperatures of its human and mos-quito hosts. Proc Natl Acad Sci U S A 110:6795– 6799. http://dx.doi.org/10.1073/pnas.1304300110.

24. Ng ML, Corner LC. 1989. Detection of some dengue-2 virus antigens ininfected cells using immuno-microscopy. Arch Virol 104:197–208. http://dx.doi.org/10.1007/BF01315543.

25. Winkler G, Maxwell SE, Ruemmler C, Stollar V. 1989. Newly synthesizeddengue-2 virus nonstructural protein NS1 is a soluble protein but becomespartially hydrophobic and membrane-associated after dimerization. Virology171:302–305. http://dx.doi.org/10.1016/0042-6822(89)90544-8.

26. Jacobs MG, Robinson PJ, Bletchly C, Mackenzie JM, Young PR. 2000.Dengue virus nonstructural protein 1 is expressed in a glycosyl-phosphatidylinositol-linked form that is capable of signal transduction.FASEB J 14:1603–1610. http://dx.doi.org/10.1096/fj.14.11.1603.

27. Noisakran S, Dechtawewat T, Avirutnan P, Kinoshita T, Siripany-aphinyo U, Puttikhunt C, Kasinrerk W, Malasit P, Sittisombut N. 2008.Association of dengue virus NS1 protein with lipid rafts. J Gen Virol 89:2492–2500. http://dx.doi.org/10.1099/vir.0.83620-0.

28. Avirutnan P, Zhang L, Punyadee N, Manuyakorn A, Puttikhunt C,Kasinrerk W, Malasit P, Atkinson JP, Diamond MS. 2007. Secreted NS1of dengue virus attaches to the surface of cells via interactions with hepa-ran sulfate and chondroitin sulfate E. PLoS Pathog 3:e183. http://dx.doi.org/10.1371/journal.ppat.0030183.

29. Halstead SB, O’Rourke EJ. 1977. Dengue viruses and mononuclearphagocytes. I. Infection enhancement by non-neutralizing antibody. J ExpMed 146:201–217.

30. King AD, Nisalak A, Kalayanrooj S, Myint KS, Pattanapanyasat K,Nimmannitya S, Innis BL. 1999. B cells are the principal circulatingmononuclear cells infected by dengue virus. Southeast Asian J Trop MedPublic Health 30:718 –728.

31. Davies KA, Hird V, Stewart S, Sivolapenko GB, Jose P, Epenetos AA,Walport MJ. 1990. A study of in vivo immune complex formation andclearance in man. J Immunol 144:4613– 4620.

32. Schifferli JA, Ng YC, Peters DK. 1986. The role of complement and itsreceptor in the elimination of immune complexes. N Engl J Med 315:488 –495. http://dx.doi.org/10.1056/NEJM198608213150805.

33. Malasit P. 1987. Complement and dengue haemorrhagic fever/shock syn-drome. Southeast Asian J Trop Med Public Health 18:316 –320.

34. Theofilopoulos AN, Wilson CB, Dixon FJ. 1976. The Raji cell radioim-mune assay for detecting immune complexes in human sera. J Clin Invest57:169 –182. http://dx.doi.org/10.1172/JCI108257.

35. Libraty DH, Young PR, Pickering D, Endy TP, Kalayanarooj S, GreenS, Vaughn DW, Nisalak A, Ennis FA, Rothman AL. 2002. High circu-lating levels of the dengue virus nonstructural protein NS1 early in dengueillness correlate with the development of dengue hemorrhagic fever. JInfect Dis 186:1165–1168. http://dx.doi.org/10.1086/343813.

36. Koraka P, Suharti C, Setiati TE, Mairuhu AT, Van Gorp E, Hack CE,Juffrie M, Sutaryo J, Van Der Meer GM, Groen J, Osterhaus AD. 2001.Kinetics of dengue virus-specific serum immunoglobulin classes and sub-classes correlate with clinical outcome of infection. J Clin Microbiol 39:4332–4338. http://dx.doi.org/10.1128/JCM.39.12.4332-4338.2001.

37. Shu PY, Chen LK, Chang SF, Yueh YY, Chow L, Chien LJ, Chin C, LinTH, Huang JH. 2000. Dengue NS1-specific antibody responses: isotypedistribution and serotyping in patients with dengue fever and denguehemorrhagic fever. J Med Virol 62:224 –232. http://dx.doi.org/10.1002/1096-9071(200010)62:2224::AID-JMV14�3.0.CO;2-C.

38. Nickells M, Hauhart R, Krych M, Subramanian VB, Geoghegan-BarekK, Marsh HC, Jr, Atkinson JP. 1998. Mapping epitopes for 20 monoclo-nal antibodies to CR1. Clin Exp Immunol 112:27–33. http://dx.doi.org/10.1046/j.1365-2249.1998.00549.x.

39. Chuansumrit A, Chaiyaratana W. 2014. Hemostatic derangement indengue hemorrhagic fever. Thromb Res 133:10 –16. http://dx.doi.org/10.1016/j.thromres.2013.09.028.

40. Orsi FA, Angerami RN, Mazetto BM, Quaino SK, Santiago-BassoraF, Castro V, de Paula EV, Annichino-Bizzacchi JM. 2013. Reduced

thrombin formation and excessive fibrinolysis are associated withbleeding complications in patients with dengue fever: a case-controlstudy comparing dengue fever patients with and without bleedingmanifestations. BMC Infect Dis 13:350. http://dx.doi.org/10.1186/1471-2334-13-350.

41. Sinauridze EI, Kireev DA, Popenko NY, Pichugin AV, Panteleev MA,Krymskaya OV, Ataullakhanov FI. 2007. Platelet microparticle mem-branes have 50- to 100-fold higher specific procoagulant activity thanactivated platelets. Thromb Haemost 97:425– 434. http://dx.doi.org/10.1160/TH06-06-0313.

42. Key NS, Kwaan HC. 2010. Microparticles in thrombosis and hemostasis.Semin Thromb Hemost 36:805– 806. http://dx.doi.org/10.1055/s-0030-1267033.

43. Aye KS, Charngkaew K, Win N, Wai KZ, Moe K, Punyadee N, Thiem-meca S, Suttitheptumrong A, Sukpanichnant S, Prida M, Halstead SB.2014. Pathologic highlights of dengue hemorrhagic fever in 13 autopsycases from Myanmar. Hum Pathol 45:1221–1233. http://dx.doi.org/10.1016/j.humpath.2014.01.022.

44. Bargeron Clark K, Hsiao HM, Noisakran S, Tsai JJ, Perng GC. 2012.Role of microparticles in dengue virus infection and its impact on medicalintervention strategies. Yale J Biol Med 85:3–18.

45. Kadiu I, Narayanasamy P, Dash PK, Zhang W, Gendelman HE. 2012.Biochemical and biologic characterization of exosomes and microvesiclesas facilitators of HIV-1 infection in macrophages. J Immunol 189:744 –754. http://dx.doi.org/10.4049/jimmunol.1102244.

46. Dargan DJ, Subak-Sharpe JH. 1997. The effect of herpes simplex virustype 1 L-particles on virus entry, replication, and the infectivity of nakedherpesvirus DNA. Virology 239:378 –388. http://dx.doi.org/10.1006/viro.1997.8893.

47. Mack M, Kleinschmidt A, Bruhl H, Klier C, Nelson PJ, Cihak J, PlachyJ, Stangassinger M, Erfle V, Schlondorff D. 2000. Transfer of the chemo-kine receptor CCR5 between cells by membrane-derived microparticles: amechanism for cellular human immunodeficiency virus 1 infection. NatMed 6:769 –775. http://dx.doi.org/10.1038/77498.

48. Hottz ED, Oliveira MF, Nunes PC, Nogueira RM, Valls-de-Souza R, DaPoian AT, Weyrich AS, Zimmerman GA, Bozza PT, Bozza FA. 2013.Dengue induces platelet activation, mitochondrial dysfunction and celldeath through mechanisms that involve DC-SIGN and caspases. JThromb Haemost 11:951–962. http://dx.doi.org/10.1111/jth.12178.

49. Alonzo MT, Lacuesta TL, Dimaano EM, Kurosu T, Suarez LA, MapuaCA, Akeda Y, Matias RR, Kuter DJ, Nagata S, Natividad FF, Oishi K.2012. Platelet apoptosis and apoptotic platelet clearance by macrophagesin secondary dengue virus infections. J Infect Dis 205:1321–1329. http://dx.doi.org/10.1093/infdis/jis180.

50. Noisakran S, Onlamoon N, Pattanapanyasat K, Hsiao HM, Songprak-hon P, Angkasekwinai N, Chokephaibulkit K, Villinger F, Ansari AA,Perng GC. 2012. Role of CD61� cells in thrombocytopenia of denguepatients. Int J Hematol 96:600 – 610. http://dx.doi.org/10.1007/s12185-012-1175-x.

51. Wang S, He R, Patarapotikul J, Innis BL, Anderson R. 1995. Antibody-enhanced binding of dengue-2 virus to human platelets. Virology 213:254 –257. http://dx.doi.org/10.1006/viro.1995.1567.

52. Hottz ED, Lopes JF, Freitas C, Valls-de-Souza R, Oliveira MF, BozzaMT, Da Poian AT, Weyrich AS, Zimmerman GA, Bozza FA, Bozza PT.2013. Platelets mediate increased endothelium permeability in denguethrough NLRP3-inflammasome activation. Blood 122:3405–3414. http://dx.doi.org/10.1182/blood-2013-05-504449.

53. Nielsen MH, Beck-Nielsen H, Andersen MN, Handberg A. 2014. A flowcytometric method for characterization of circulating cell-derived micro-particles in plasma. J Extracell Vesicles 3:20795. http://dx.doi.org/10.3402/jev.v3.20795.

54. Lacroix R, Dignat-George F. 2012. Microparticles as a circulatingsource of procoagulant and fibrinolytic activities in the circulation.Thromb Res 129(Suppl 2):S27–S29. http://dx.doi.org/10.1016/j.thromres.2012.02.025.

55. Owens AP, III, Mackman N. 2011. Microparticles in hemostasisand thrombosis. Circ Res 108:1284 –1297. http://dx.doi.org/10.1161/CIRCRESAHA.110.233056.

56. Siljander PR. 2011. Platelet-derived microparticles—an updated perspec-tive. Thromb Res 127(Suppl 2):S30 –S33. http://dx.doi.org/10.1016/S0049-3848(10)70152-3.

57. Ay C, Freyssinet JM, Sailer T, Vormittag R, Pabinger I. 2009. Circulat-ing procoagulant microparticles in patients with venous thromboembo-

Punyadee et al.



icholas White

http://jvi.asm.org/

Dow

nloaded from

http://dx.doi.org/10.1002/jmv.20004

http://dx.doi.org/10.1002/jmv.20004

http://dx.doi.org/10.1016/j.chom.2012.08.009


http://dx.doi.org/10.1073/pnas.1304300110

http://dx.doi.org/10.1073/pnas.1304300110

http://dx.doi.org/10.1007/BF01315543

http://dx.doi.org/10.1007/BF01315543

http://dx.doi.org/10.1016/0042-6822(89)90544-8

http://dx.doi.org/10.1096/fj.14.11.1603

http://dx.doi.org/10.1099/vir.0.83620-0

http://dx.doi.org/10.1371/journal.ppat.0030183

http://dx.doi.org/10.1371/journal.ppat.0030183

http://dx.doi.org/10.1056/NEJM198608213150805

http://dx.doi.org/10.1172/JCI108257

http://dx.doi.org/10.1086/343813

http://dx.doi.org/10.1128/JCM.39.12.4332-4338.2001

http://dx.doi.org/10.1002/1096-9071(200010)62:2%3C224::AID-JMV14%3E3.0.CO;2-C

http://dx.doi.org/10.1002/1096-9071(200010)62:2%3C224::AID-JMV14%3E3.0.CO;2-C

http://dx.doi.org/10.1046/j.1365-2249.1998.00549.x

http://dx.doi.org/10.1046/j.1365-2249.1998.00549.x

http://dx.doi.org/10.1016/j.thromres.2013.09.028


http://dx.doi.org/10.1186/1471-2334-13-350

http://dx.doi.org/10.1186/1471-2334-13-350

http://dx.doi.org/10.1160/TH06-06-0313

http://dx.doi.org/10.1160/TH06-06-0313

http://dx.doi.org/10.1055/s-0030-1267033

http://dx.doi.org/10.1055/s-0030-1267033

http://dx.doi.org/10.1016/j.humpath.2014.01.022

http://dx.doi.org/10.1016/j.humpath.2014.01.022

http://dx.doi.org/10.4049/jimmunol.1102244

http://dx.doi.org/10.1006/viro.1997.8893


http://dx.doi.org/10.1038/77498

http://dx.doi.org/10.1111/jth.12178

http://dx.doi.org/10.1093/infdis/jis180

http://dx.doi.org/10.1093/infdis/jis180

http://dx.doi.org/10.1007/s12185-012-1175-x

http://dx.doi.org/10.1007/s12185-012-1175-x


http://dx.doi.org/10.1182/blood-2013-05-504449


http://dx.doi.org/10.3402/jev.v3.20795

http://dx.doi.org/10.3402/jev.v3.20795



http://dx.doi.org/10.1161/CIRCRESAHA.110.233056

http://dx.doi.org/10.1161/CIRCRESAHA.110.233056

http://dx.doi.org/10.1016/S0049-3848(10)70152-3

http://dx.doi.org/10.1016/S0049-3848(10)70152-3

http://jvi.asm.org

http://jvi.asm.org/

lism. Thromb Res 123:724 –726. http://dx.doi.org/10.1016/j.thromres.2008.09.005.

58. Shantsila E, Kamphuisen PW, Lip GY. 2010. Circulating microparticlesin cardiovascular disease: implications for atherogenesis and athero-thrombosis. J Thromb Haemost 8:2358 –2368. http://dx.doi.org/10.1111/j.1538-7836.2010.04007.x.

59. Toti F, Satta N, Fressinaud E, Meyer D, Freyssinet JM. 1996. Scottsyndrome, characterized by impaired transmembrane migration of pro-coagulant phosphatidylserine and hemorrhagic complications, is an in-herited disorder. Blood 87:1409 –1415.

60. Castaman G, Yu-Feng L, Battistin E, Rodeghiero F. 1997. Characterizationof a novel bleeding disorder with isolated prolonged bleeding time and defi-ciency of platelet microvesicle generation. Br J Haematol 96:458–463. http://dx.doi.org/10.1046/j.1365-2141.1997.d01-2072.x.

61. Taylor RP, Ferguson PJ, Martin EN, Cooke J, Greene KL, Grinspun K,Guttman M, Kuhn S. 1997. Immune complexes bound to the primateerythrocyte complement receptor (CR1) via anti-CR1 mAbs are clearedsimultaneously with loss of CR1 in a concerted reaction in a rhesus mon-key model. Clin Immunol Immunopathol 82:49 –59. http://dx.doi.org/10.1006/clin.1996.4286.

62. Davies KA, Peters AM, Beynon HL, Walport MJ. 1992. Immune com-plex processing in patients with systemic lupus erythematosus. In vivoimaging and clearance studies. J Clin Invest 90:2075–2083.

63. Emlen W, Carl V, Burdick G. 1992. Mechanism of transfer of immunecomplexes from red blood cell CR1 to monocytes. Clin Exp Immunol89:8 –17.

64. Hepburn AL, Mason JC, Wang S, Shepherd CJ, Florey O, Haskard DO,Davies KA. 2006. Both Fcgamma and complement receptors mediatetransfer of immune complexes from erythrocytes to human macrophagesunder physiological flow conditions in vitro. Clin Exp Immunol 146:133–145. http://dx.doi.org/10.1111/j.1365-2249.2006.03174.x.

65. Craig ML, Bankovich AJ, McElhenny JL, Taylor RP. 2000. Clearance ofanti-double-stranded DNA antibodies: the natural immune complexclearance mechanism. Arthritis Rheum 43:2265–2275. http://dx.doi.org/10.1002/1529-0131(200010)43:102265::AID-ANR14�3.0.CO;2-J.

66. Duval A, Helley D, Capron L, Youinou P, Renaudineau Y, DubucquoiS, Fischer AM, Hachulla E. 2010. Endothelial dysfunction in systemiclupus patients with low disease activity: evaluation by quantification andcharacterization of circulating endothelial microparticles, role of anti-endothelial cell antibodies. Rheumatology (Oxford) 49:1049 –1055. http://dx.doi.org/10.1093/rheumatology/keq041.

67. Antwi-Baffour S, Kholia S, Aryee YK, Ansa-Addo EA, Stratton D,Lange S, Inal JM. 2010. Human plasma membrane-derived vesiclesinhibit the phagocytosis of apoptotic cells—possible role in SLE.Biochem Biophys Res Commun 398:278 –283. http://dx.doi.org/10.1016/j.bbrc.2010.06.079.

68. Sellam J, Proulle V, Jungel A, Ittah M, Miceli Richard C, Gottenberg JE,Toti F, Benessiano J, Gay S, Freyssinet JM, Mariette X. 2009. Increasedlevels of circulating microparticles in primary Sjogren’s syndrome, systemiclupus erythematosus and rheumatoid arthritis and relation with disease activ-ity. Arthritis Res Ther 11:R156. http://dx.doi.org/10.1186/ar2833.

69. Pereira J, Alfaro G, Goycoolea M, Quiroga T, Ocqueteau M, Massardo L,Perez C, Saez C, Panes O, Matus V, Mezzano D. 2006. Circulatingplatelet-derived microparticles in systemic lupus erythematosus. Associa-tion with increased thrombin generation and procoagulant state. ThrombHaemost 95:94 –99. http://dx.doi.org/10.1160/TH05-05-0310.

70. Bhakdi S, Fassbender W, Hugo F, Carreno MP, Berstecher C, MalasitP, Kazatchkine MD. 1988. Relative inefficiency of terminal complementactivation. J Immunol 141:3117–3122.

71. Willekens FL, Roerdinkholder-Stoelwinder B, Groenen-Dopp YA, BosHJ, Bosman GJ, van den Bos AG, Verkleij AJ, Werre JM. 2003. Hemo-globin loss from erythrocytes in vivo results from spleen-facilitatedvesiculation. Blood 101:747–751. http://dx.doi.org/10.1182/blood-2002-02-0500.

72. Almizraq R, Tchir JD, Holovati JL, Acker JP. 2013. Storage of red bloodcells affects membrane composition, microvesiculation, and in vitro qual-ity. Transfusion 53:2258 –2267. http://dx.doi.org/10.1111/trf.12080.

73. Kozuma Y, Sawahata Y, Takei Y, Chiba S, Ninomiya H. 2011. Proco-agulant properties of microparticles released from red blood cells in par-oxysmal nocturnal haemoglobinuria. Br J Haematol 152:631– 639. http://dx.doi.org/10.1111/j.1365-2141.2010.08505.x.

74. Willekens FL, Werre JM, Groenen-Dopp YA, Roerdinkholder-StoelwinderB, de Pauw B, Bosman GJ. 2008. Erythrocyte vesiculation: a self-protective mechanism? Br J Haematol 141:549 –556. http://dx.doi.org/10.1111/j.1365-2141.2008.07055.x.

75. Moskovich O, Fishelson Z. 2007. Live cell imaging of outward and inwardvesiculation induced by the complement C5b-9 complex. J Biol Chem282:29977–29986. http://dx.doi.org/10.1074/jbc.M703742200.

76. Simak J, Gelderman MP. 2006. Cell membrane microparticles in bloodand blood products: potentially pathogenic agents and diagnostic mark-ers. Transfus Med Rev 20:1–26. http://dx.doi.org/10.1016/j.tmrv.2005.08.001.

77. Shet AS, Aras O, Gupta K, Hass MJ, Rausch DJ, Saba N, KoopmeinersL, Key NS, Hebbel RP. 2003. Sickle blood contains tissue factor-positivemicroparticles derived from endothelial cells and monocytes. Blood 102:2678 –2683. http://dx.doi.org/10.1182/blood-2003-03-0693.

78. Alaarg A, Schiffelers RM, van Solinge WW, van Wijk R. 2013. Red bloodcell vesiculation in hereditary hemolytic anemia. Front Physiol 4:365. http://dx.doi.org/10.3389/fphys.2013.00365.

79. Blum A. 2009. The possible role of red blood cell microvesicles in athero-sclerosis. Eur J Intern Med 20:101–105. http://dx.doi.org/10.1016/j.ejim.2008.06.001.

80. Pankoui Mfonkeu JB, Gouado I, Fotso Kuate H, Zambou O, Amvam ZolloPH, Grau GE, Combes V. 2010. Elevated cell-specific microparticles are abiological marker for cerebral dysfunctions in human severe malaria. PLoSOne 5:e13415. http://dx.doi.org/10.1371/journal.pone.0013415.

81. Sadallah S, Eken C, Schifferli JA. 2008. Erythrocyte-derived ectosomeshave immunosuppressive properties. J Leukoc Biol 84:1316 –1325. http://dx.doi.org/10.1189/jlb.0108013.

82. Mantel PY, Hoang AN, Goldowitz I, Potashnikova D, Hamza B, Vorob-jev I, Ghiran I, Toner M, Irimia D, Ivanov AR, Barteneva N, Marti M.2013. Malaria-infected erythrocyte-derived microvesicles mediate cellularcommunication within the parasite population and with the host immunesystem. Cell Host Microbe 13:521–534. http://dx.doi.org/10.1016/j.chom.2013.04.009.

83. Srikiatkhachorn A, Rothman AL, Gibbons RV, Sittisombut N, Malasit P,Ennis FA, Nimmannitya S, Kalayanarooj S. 2011. Dengue—how best toclassify it. Clin Infect Dis 53:563–567. http://dx.doi.org/10.1093/cid/cir451.

84. Gan VC, Lye DC, Thein TL, Dimatatac F, Tan AS, Leo YS. 2013.Implications of discordance in World Health Organization 1997 and 2009dengue classifications in adult dengue. PLoS One 8:e60946. http://dx.doi.org/10.1371/journal.pone.0060946.




icholas White

http://jvi.asm.org/

Dow

nloaded from



http://dx.doi.org/10.1111/j.1538-7836.2010.04007.x

http://dx.doi.org/10.1111/j.1538-7836.2010.04007.x

http://dx.doi.org/10.1046/j.1365-2141.1997.d01-2072.x

http://dx.doi.org/10.1046/j.1365-2141.1997.d01-2072.x

http://dx.doi.org/10.1006/clin.1996.4286

http://dx.doi.org/10.1006/clin.1996.4286

http://dx.doi.org/10.1111/j.1365-2249.2006.03174.x

http://dx.doi.org/10.1002/1529-0131(200010)43:10%3C2265::AID-ANR14%3E3.0.CO;2-J

http://dx.doi.org/10.1002/1529-0131(200010)43:10%3C2265::AID-ANR14%3E3.0.CO;2-J

http://dx.doi.org/10.1093/rheumatology/keq041

http://dx.doi.org/10.1093/rheumatology/keq041

http://dx.doi.org/10.1016/j.bbrc.2010.06.079

http://dx.doi.org/10.1016/j.bbrc.2010.06.079

http://dx.doi.org/10.1186/ar2833

http://dx.doi.org/10.1160/TH05-05-0310



http://dx.doi.org/10.1111/trf.12080

http://dx.doi.org/10.1111/j.1365-2141.2010.08505.x

http://dx.doi.org/10.1111/j.1365-2141.2010.08505.x

http://dx.doi.org/10.1111/j.1365-2141.2008.07055.x

http://dx.doi.org/10.1111/j.1365-2141.2008.07055.x

http://dx.doi.org/10.1074/jbc.M703742200

http://dx.doi.org/10.1016/j.tmrv.2005.08.001

http://dx.doi.org/10.1016/j.tmrv.2005.08.001


http://dx.doi.org/10.3389/fphys.2013.00365

http://dx.doi.org/10.3389/fphys.2013.00365

http://dx.doi.org/10.1016/j.ejim.2008.06.001

http://dx.doi.org/10.1016/j.ejim.2008.06.001

http://dx.doi.org/10.1371/journal.pone.0013415

http://dx.doi.org/10.1189/jlb.0108013

http://dx.doi.org/10.1189/jlb.0108013



http://dx.doi.org/10.1093/cid/cir451



http://jvi.asm.org

http://jvi.asm.org/

Microparticles Provide a Novel Biomarker To Predict Severe Clinical Outcomes of Dengue Virus...

Documents

Transcript of Microparticles Provide a Novel Biomarker To Predict Severe Clinical Outcomes of Dengue Virus...