Abstract
Dengue hemorrhagic fever (DHF), the most severe form of illness following infection with a dengue virus, is characterized by plasma leakage, thrombocytopenia, and hepatic inflammation. The interrelationships among virus burden, immune activation, and development of DHF were examined in 54 children with secondary dengue-3 virus infections participating in a prospective, hospital-based study. DHF was associated with higher mean plasma viremia early in illness and earlier peak plasma interferon-γ levels. Maximum plasma viremia levels correlated with the degree of plasma leakage and thrombocytopenia. Maximum plasma levels of interleukin (IL)-10 and soluble tumor necrosis factor receptor-II correlated with the degree of thrombocytopenia, independently of viremia levels. Hepatic transaminase elevation correlated with plasma soluble IL-2 receptor levels and not with viremia levels. Quantitative differences in virus burden and host immune responses, and the timing of type 1 cytokine responses, have differing influences on the severity of disease manifestations during secondary dengue-3 virus infections.
Dengue is an emerging arboviral disease caused by infection with 1 of the dengue viruses, a group of 4 antigenically related mosquito-borne flaviviruses [1]. It remains a major cause of morbidity throughout tropical and subtropical regions of the world [2, 3]. Classic dengue fever (DF) is characterized by high fevers, retro-orbital headache, severe myalgias, and a rash. The most severe form of illness, dengue hemorrhagic fever (DHF), is characterized by the development of plasma leakage and a hemorrhagic diathesis near the time of defervescence. Although coagulation abnormalities [4–6], thrombocytopenia, and hepatitis [7, 8] are prominent features of the DHF syndrome, plasma leakage is the major pathophysiologic hallmark of DHF. In severe DHF, morbidity and mortality are the result of hypotension and shock, at times accompanied by severe disseminated intravascular coagulation and bleeding [9].
Several studies have observed that sequential heterotypic dengue virus infections (secondary infection) are more likely to produce DHF [10–12]. At least 2 mechanistic concepts have been implicated in this apparent immune enhancement of disease severity. Antibody-dependent enhancement of infection postulates that preexisting cross-reactive antibodies facilitate virus entry on Fc receptor-bearing cells [13]. This may result in an increase in total virus burden leading to DHF. In addition, an exaggerated cellular immune response to dengue virus infection, driven by crossreactive memory T lymphocytes, may lead to increased disease severity and DHF [14]. These 2 mechanisms are not mutually exclusive and may interactwith other viral and host factors [15–19]. Other studies have examined separately the importance of virus burden, immune activation, and cytokine production in the development of DHF [20–28]. These important aspects of DHF pathogenesis have not been previously studied in a comprehensive fashion in a single patient cohort.
The goal of this study was to better define the relative contributions of virus burden and host immune activation to the pathogenesis of DHF. In a prospective, hospital-based study of dengue, we examined the relationships among viremia, circulating cytokine and cytokine receptor levels, and disease severity in children with secondary dengue-3 virus (D3V) infections.
Methods
Study design. Details of the investigational protocol have been published elsewhere [8]. Children were enrolled during 1994–1997 and 1999–2000 at the Queen Sirikit Institute of Child Health in Bangkok or the Kamphaeng Phet Provincial Hospital, Thailand. Enrollment criteria were age 6 months to 14 years, a febrile illness of <72 h duration, no hypotension or shock, and no other obvious source of infection. Children were observed in hospital until at least 1 day after defervescence. Venous blood samples were drawn daily up to the day after defervescence or for a maximum of 5 consecutive days, as well as 8–13 days after enrollment. A complete blood cell count (T540 counter; Coulter) and plasma aspartate aminotransferase (AST) and alanine aminotransferase (ALT) determinations (Clinical System Analyzer, model 700; Beckman Instruments) were obtained daily. Serial hematocrit determinations were done daily and every 6 h for 4 times on defervescence; in addition, a right lateral decubitus chest radiograph was obtained on the day after defervescence. Aliquots of plasma were stored at 270°C.
Those children with a secondary D3V infection and ⩾2 days of fever after enrollment were included in this study. Secondary dengue virus infections were identified by use of previously established serologic criteria for IgMand IgG ELISAs and hemagglutination-inhibition assays to dengue virus serotypes 1–4 and Japanese encephalitis virus in paired plasma specimens [29–31]. D3V infection was identified by virus isolation in Toxorhynchites splendens mosquitoes [32, 33] or by a serotype-specific reverse-transcription (RT) polymerase chain reaction (PCR) assay [34].
Quantification of D3V viremia. Viremiawas defined as the level of circulating viral RNA copies and was quantified as D3V genome equivalents per milliliter (cDNA copies/mL). After RNA extraction from 100 µL of plasma [35], a fluorogenic RT-PCR assay [36] was used to determine the D3V RNA copy number. A143-nt amplicon of the D3V 3′-noncoding region (the RT-PCR product) [36] was cloned into a pNOT vector. Serial dilutions of this D3V clone were used to construct a standard curve for each 96-well plate. Unknown sample concentrations were expressed as D3V genome equivalent cDNA copies/mL. Interassay precision was monitored by aliquots of healthy control plasma spiked with high and lowconcentrations of a low-passaged D3V isolate and were assayed on each 96-well plate. All unknown samples and controls were assayed in triplicate. TheABI gene detection system 7700 (Perkin-Elmer Applied Biosystems) was used for PCR cycling amplification, real-time data collection, and analysis.
Study protocol definitions. Illness day 1 was the calendar day on which fever began, as ascertained by history. Fever day 0 was the calendar day of defervescence; days before and after this point were numbered consecutively (e.g., −1, −2 and +1, +2, respectively). Pleural effusion index was calculated from the right lateral chest radiograph as follows: 100 × (maximum width of right pleural effusion)/(maximum width of right hemithorax). Maximum percentage of hemoconcentration was calculated as follows: 100×(maximum hematocrit surrounding defervescence)/)minimum hematocrit during acute illness or convalescence)−(minimum hematocrit during acute illness or convalescence). The maximum viremia level was defined as the highest plasma viremia level (D3V genome equivalent cDNA copies/mL) measured during illness. The maximum viremia level was considered to be a peak viremia level only in cases in which viremia rose after study day 1 or decreased by ⩽0.5 log from study day 1 to study day 2.
Clinical categories. DHF (based on World Health Organization criteria [31]) was defined as a platelet count nadir of ⩽100,000/mm3 and hemoconcentration surrounding defervescence with maximum hemoconcentration percentage of>20%, or.1mmof pleural effusion detected on the right lateral chest radiograph. Cases of DHF were further graded as I–IV according to the World Health Organization criteria [31].
DF was defined as maximum hemoconcentration percentage of ⩽15% and no detectable pleural effusion on the right lateral decubitus chest radiograph. There was no restriction on platelet count nadir.
Intermediate DF/DHF (did not meet criteria for DF or DHF) was defined as maximum hemoconcentration percentage of 15%–20% or detectable pleural effusion (but ⩽1 mm) on the right lateral decubitus chest radiograph, or the patient met DHF criteria for plasma leakage but the platelet count nadir was >100,000/mm3.
ELISAs. Plasma levels of the following cytokines or cytokine receptors were determined by ELISA: interferon (IFN)-α, IFN-γ, interleukin (IL)-10, soluble tumor necrosis factor receptor (sTNFR)-II, and soluble IL-2 receptor (sIL-2R). Details of the ELISA protocol for IFN-α have been described else where [37]. The same protocol was used for the IFN-γ ELISA with the following modifications: coating antibody was 0.5 µg/mL mouse anti-human IFN-γ monoclonal antibody 2G1; biotinylated detecting antibody was 0.25 µg/mL mouse anti-human IFN-γ monoclonal antibody B133.5; streptavidin/peroxidase was 1:15,000 dilution of poly-horseradish peroxidase-conjugated streptavidin; and the standard was human recombinant IFN-γ (all from Endogen). Plasma IL-10, sTNFR-II, and sIL-2R levels were measured by use of commercial ELISA kits, according to manufacturers' instructions (IL-10 and sTNFR-II, Quantikine, R&D Systems; sIL-2R, Cellfree, Endogen). All samples and standard dilutions were assayed in duplicate.
Statistical analysis. Student's t test was used for comparisons between normally distributed continuous variables; the Mann-Whitney U or Kruskal-Wallis test was used for comparisons between continuous variables not normally distributed. χ2 analysis or Fisher's exact test was used for comparisons among proportional data. Pearson's and Spearman's correlation tests were used to examine associations between continuous variables. Logistic regression was used to calculate the odds ratios (ORs) for maximum viremia levels and DHF or intermediate DF/DHF. Multivariable linear regression was used to examine the associations of >1 variable with continuous measures of disease severity. P ⩽ .05 was considered to be significant; <.05 P ⩽.10 was considered to be a nonsignificant trend. All values are presented as mean ± SE, unless otherwise stated. The statistical software package SPSS (version 10.0; SPSS) was used for all statistical analyses.
Results
Study population characteristics. Over a 6-year period of enrollment, 381 children were identified with dengue virus infections (dengue virus serotypes 1–4). The study population comprised the 54 children who had secondary D3V infections and ⩾2 days of fever after enrollment. The main characteristics of the study population are summarized in table 1. The clinical categories of DHF, intermediate DF/DHF, and DF reflected a spectrum of decreasing disease severity, as illustrated by measures of hemoconcentration, pleural effusion index, degree of thrombocytopenia, and hepatic transaminase levels.
Characteristics of subjects in study of virus burden and host immune responses in secondary dengue-3 virus infection.
Characteristics of the quantitative RT-PCR assay for D3V viremia. A fluorogenic RT-PCR assay was used to quantitate the D3V RNA copy number in plasma samples (expressed as D3V genome equivalent cDNA copy number). The limit of quantitation of the assay was 20 D3V cDNA copies per reaction, and the linear range of the standard curve extended over 5 logs of D3V cDNA concentration. At the upper range of the assay (high positive control), the coefficient of variancewas 1.2% (mean ± SD, 9.35 ± 0.11 log D3V cDNA copies/mL). At the lower range of the assay (low positive control), the coefficient of variance was 7.7% (4.94 ± 0.38 log D3V cDNA copies/mL). Negative controls were always undetectable. Compared with a competitor RT-PCR method [24], D3V genome equivalent cDNA copy number correlated with D3V RNA copy number (Pearson's r = 0.73; P < .001), and the geometric mean ratio of D3V cDNA to RNA was 1.35 (n = 21). Compared with a mosquito inoculation method [22], D3V genome equivalent cDNA copy number correlated with titers of infectious virus (50% mosquito infectious doses/mL) (Pearson's r = 0.89; P < .001), and the geometric mean ratio of D3V cDNA per 50% mosquito infectious dose was 1.54 (n = 18).
D3V viremia levels. Early in clinical illness, mean levels of plasma viremia (log D3V genome equivalent cDNA copies/mL) were higher inDHFand intermediate DF/DHF than in DF(figure 1). Differences in plasma viremia between DHF and DF were 0.86±0:43 log D3V cDNA copies/mL on illness day 2,0.94±0.26 log on illness day 3, and 0.70±0.30 log on illness day 4 (P = .10, P = .003, and P = .03, respectively). The difference in plasma viremia between intermediate DF/DHF and DF was 0.83°0:31 log D3V cDNA copies/mL on illness day 3 (P = .02). There were no significant differences in plasma viremia levels between DHF grade III and DHF grades I/II at any time point (data not shown).
Magnitude of dengue-3 virus (D3V) viremia, by day of illness. Levels of D3V genome equivalent cDNA copies per milliliter were determined in serial plasma samples from 3 groups of patients with secondary D3V infection: dengue hemorrhagic fever (DHF; n = 29, ●), dengue fever (DF; n = 12, □), and intermediate DF/DHF (n = 13, △). Data are mean ± SE. aP = .10, vs. DF; bP < .05, vs. DF.
Although all subjectswere enrolledwithin 72 h of illness onset, a potential confounding variable when comparingmaximum levels of plasma viremia is a difference in the day of illness on study entry. When logistic regression was used to control for illness day on study entry, higher maximum plasma levels of D3V were associated with DHF than with DF (adjusted OR, 2.0/0.5 log D3V cDNA/mL increase; 95% confidence interval [CI], 1.1–3.8; P = .03). Higher maximum plasma levels of D3V also tended to be associated with intermediate DF/DHF than with DF (adjusted OR, 2.1/0.5 log D3V cDNA/mL increase; 95% CI, 0.9–4.8; P = 008).
Irrespective of clinical diagnosis, the maximum level of viremia correlated with the degree of hemoconcentration (Pearson's r = 0.40; P = .003) and, somewhat less, with the pleural effusion index on fever day +1 (Pearson's r = 0.27; P = .05; figure 2A and 2B). A significant negative correlation was seen with the platelet count nadir (Pearson's r = −0.32; P = .02; figure 2C), but no correlation was seen with maximum AST or ALT levels (Pearson's r = −0.02 and .02; P = .8 and .9, respectively; data not shown).
Relationships between maximum plasma viremia and continuous measures of disease severity in secondary dengue-3 virus (D3V) infections. Maximum plasma levels of D3V genome equivalent cDNA copies per milliliter were compared with maximum percentage of hemoconcentration (A; △, n = 53), pleural effusion index (PEI) on fever day + 1 (B; □, n = 50), and nadir of thrombocytopenia (C; ○, n = 54). Symbols represent individual cases; least-squares regression best-fit line is shown.
Host immune responses. In 22 cases, a peak viremia level was identified. Mean peak viremia levels trended higher in DHF than in DF (DHF, 7.87±0.21 log D3V cDNA copies/mL, n=11; intermediate DF/DHF, 7.83 ± 0.33 log D3V cDNA copies/mL, n=5; DF, 7.27±0.24 log D3V cDNA copies/mL, n = 6; DHF vs. DF, P=.08). Peak plasma levels of IFN-α were generally not seen and occurred prior to peak viremia. Peak plasma levels of IFN-γ and IL-10 occurred a median of 2 days after peak viremia and near the time of defervescence. Plasma sTNFR-II and sIL-2R levels were nearing their peak a median of 3 days after peak viremia and following defervescence (figures 3 and 4A)
Relationships between viremia, immune activation markers, and time of defervescence in secondary dengue-3 virus (D3V) infections. A, Illustrative example of dengue fever. B, Illustrative example of dengue hemorrhagic fever grade II. D3V genome equivalent levels (cDNA copies per milliliter) are presented as means of triplicate determinations; cytokine or cytokine receptor levels are presented as means of duplicate determinations. Fever day 0 is day of defervescence. ◆, D3V; □, interferon (IFN)-α; W, IFN-γ; △, interleukin-10 (IL-10); ▽, soluble interleukin-2 receptor (sIL-2R); ◊, soluble tumor necrosis factor receptor-II (sTNFR-II).
Pattern of host immune response in secondary dengue-3 virus (D3V) infections in which peak viremia level was identified (n = 22). A, All clinical categories combined; solid bars represent median number of days following peak viremia that maximum plasma levels of given cytokine or cytokine receptor were achieved. Box outlines represent 25th–75th percentile range; error bars represent the 5th–95th percentile range. B, Timing of peak plasma interferon (IFN)-γ levels, by clinical category. Bars represent percentage of cases within clinical category. Open bars, dengue fever (DF); solid bars, dengue hemorrhagic fever (DHF); shaded bars, intermediate DF/DHF. IL-10, interleukin-10; sIL-2R, soluble interleukin-2 receptor; sTNFR-II, soluble tumor necrosis factor receptor-II.
Peak plasma IFN-γ levels occurred within 2 days of peak viremia in 100% of the DHF cases (11/11), 80% of the intermediate DF/DHF cases (4/5), and 50% of the DF cases (3/6) (P = .04; figure 4B). Peak plasma IL-10 levels occurred within 2 days of peak viremia in 73% of DHF (7/11), 80% of intermediate DF/ DHF (4/5), and 33% of DF cases (2/6), differences that were not statistically significant.
Disease severity and host immune response. Maximum plasma levels of IL-10 and sTNFR-II were higher in DHF than inDF(P=.03 and P<.001, respectively; table 2) andwere higher in DHF grade III than in DHF grades I and II (P = .02 for both; data not shown).Maximum plasma levels of sIL-2R trended higher in DHF than in DF (P = .08; table 2). There were no significant differences in maximum plasma levels of IFN-α or IFN-γ among any of the clinical groups.
Maximum plasma levels of cytokines or cytokine receptors in secondary dengue-3 illness.
Multivariable linear regression analysis was used to determine the relationships among maximum levels of plasma viremia, IFNg, IL-10, sTNFR-II, sIL-2R, and 3 key measures of dengue disease severity: hemoconcentration, thrombocytopenia, and AST elevation (table 3). The potential confounding effects of illness day on presentation and year of study enrollment were also included in the analysis. The maximum level of D3V viremia was independently associated with the degree of hemoconcentration and thrombocytopenia (partial correlation r = 0.45 and r = −0.43, respectively; P = .002, for both). Maximum plasma levels of sTNFR-II and IL-10 demonstrated significant associations with thrombocytopenia, independent of the level of viremia (partial correlation r = −0:31 and r = −0.36, and P=.04 and P=.01, respectively). No other immune activation markers demonstrated significant independent associations with hemoconcentration or thrombocytopenia. By contrast, the only significant independent association with maximum AST levels was plasma levels of sIL-2R (partial correlation r = 0.33; P = .02).
Linear regression analysis of predictor variables associated with 3 key measures of dengue disease severity.
Discussion
Our data show that D3V viremia (measured as D3V genome equivalent levels) and subsequent immune activation were of greater magnitude in more severe clinical disease. The magnitude of plasma leakage was primarily related to the magnitude of viremia. The degree of thrombocytopenia was associated with viremia, IL-10, and sTNFR-II levels, independently. Hepatic transaminase elevation was associated solely with a marker of immune activation and not with levels of viremia.
Previous studies examining dengue-1 virus (D1V) or dengue-2 virus (D2V) infections have reported higher circulating levels of replicating virus [22] or viral RNA copies [23] in DHF than in DF. However, 2 early studies examining D3V infections did not report an association between circulating virus titers and disease severity [20, 21]. These studies did not obtain sequential blood samples from the majority of subjects and, therefore, did not identify maximum viremia levels or control for illness day. Another report, which included some of the same patients in the current study, also did not find a significant association between maximum viremia levels (RNA copy levels) and disease severity in D3V and D1V infections [24]. The latter study used a different method for virus quantification (competitor RT-PCR) and contained fewer patients than did this study.
Our study is unique in several respects. We found an association between higher early and maximum viremia levels and increasing disease severity by use of a sensitive and reproducible quantitative RT-PCR assay, by closely monitoring disease severity, and by examining the pathophysiologic changes in D3Vi nfection without the constraint of typical clinical categorizations. None of the aforementioned studies examined the distinctions among DF, intermediate DF/DHF, and DHF. By following World Health Organization guidelines, dengue disease severity has been classified as DHF or DF. This 2-part classification scheme fails to appreciate the spectrum of plasma leakage and thrombocytopenia that occurs in dengue virus infections [38]. Our prospective study design elucidated the full spectrum of illness, identifying an intermediate DF/DHF group that fell between the 2 polar ends of a spectrum (DF and DHF). Most of the intermediate DF/DHF cases would have been classified as DF in earlier studies, leading to an underestimation of differences in plasma viremia.
This study is also the first to demonstrate a positive correlation between dengue viremia levels and hemoconcentration. A similar correlation between viremia levels (RNA copy levels) and hemoconcentration has been reported in hantavirus pulmonary syndrome [39], another viral illness characterized by plasma leakage and an increase in microvascular permeability. Simultaneous examination of systemic levels of immune activation markers and levels of viremia showed that the virus burden in secondary D3Vinfection is a driving force behind the degree of plasma leakage. A greater virus burden in severe dengue illness reflects a greater burden of virus-infected cells leading to an augmented cascade of antigen-driven innate and adaptive immune responses. The higher magnitude of viremia seen with higher degrees of plasma leakage and development of DHF may be because of antibody-dependent enhancement of infection, but other mechanisms cannot be excluded. As has been reported for D2V [40, 41], genetic variation among D3V strains might contribute to different cellular infection and viral replication capacities and thus affect the virus burden in dengue illness. Cytokine-mediated differentiation of monocytes to macrophages or immature dendritic cells early in secondary infectionmight also enhance cellular infection and dengue virus replication in an antibody-independent manner [42, 43]. Further studies are required to delineate the relative contributions of these mechanisms to determining the magnitude of viremia in D3V infections.
After D3V infection, proinflammatory (type 1) and anti-inflammatory (type 2) immune responses were induced. Increased levels of IL-10 and other type 2-effectors in DHF likely play a role in down-regulating the prior, augmented, release of IFN-γ and type 1-effectors. Greater type 1- and type 2-cytokine responses with increasing dengue disease severity have also been reported in other studies [25, 26, 28, 44–48]. Our observation that peak plasma IFN-γ levels occurred earlier with increasing disease severity is novel and suggests that the timing of a type 1-cytokine response to dengue viremia also plays an important role in disease pathogenesis. Greater CD8+ T cell and NK cell activation have been found in children with DHF than in children with DF [49], and both cell types are potential sources of early IFN-γ production. Through effects on dendritic cells or other antigen-presenting cells, an early burst of IFN-γ can lead to greater T cell stimulation, IL-12 release, increases in tumor necrosis factor (TNF)-α and IFN-γ secretion, and potential dysregulation of the type 1 cytokine response [37]. The cascade of events initiated by an early, poorly controlled type 1 cytokine response likely contributes to multiple aspects of DHF pathogenesis.
The immune response cascade to secondary dengue virus infection is driven by virus burden (antigen) and preexisting host factors (genetic polymorphisms and cross-reactive memory T and B cell responses). As previously discussed, the degree of plasma leakage in secondary D3V infection was primarily related to the virus burden. The degree of thrombocytopenia was related to both virus burden and host factors affecting IL-10 or TNF-α (and sTNFR-II) production. Administration of recombinant IL-10 or TNF-α to human subjects has produced thrombocytopenia because of decreased platelet production (IL-10) [50] or increased platelet consumption (TNF-α) [51, 52].
The aspect of dengue illness related primarily to host factors, independent of virus burden, was hepatic inflammation. The degree of hepatic inflammation in acute D3V infection correlated with plasma levels of sIL-2R, a marker of T cell activation and proliferation [53–55], and not with levels of viremia. A central role for CD4+ T lymphocytes in hepatocyte injury during dengue virus infection has been previously suggested [56], and the degree of hepatic inflammation seen in hepatitis C has also been characterized as primarily immune-mediated with poor correlations with virus levels [57, 58].
Our results emphasize that clinical illness is the end result of a multifaceted interaction between dengue virus and the host. Key manifestations of disease, such as plasma leakage, thrombocytopenia, and hepatic inflammation, occur along a spectrumof severity. Quantitative differences in virus burden and host immune responses are at the heart of varying disease severity in secondary D3V infections. Other factors, such as the timing of a type 1-cytokine response to acute viremia, also play a role in determining disease severity. The contribution of host immune mediators to pathogenesis, independent of virus burden, varies depending on the aspect of dengue illness examined. Enhanced knowledge of the steps involved in DHF pathogenesis will assist in the formulation of new vaccination and therapeutic strategies to combat this emerging, global infectious disease.
Acknowledgements
We thank the doctors and pediatric nurses of Queen Sirikit National Institute of Child Health and the Kamphaeng Phet Provincial Hospital, for providing exceptional patient care; the Armed Forces Research Institute of Medical Sciences virology research nurses, Pranom Vangnai, Nathada Plavooth, Somnuk Lumjiak, Sumetha Hengprasert, Sumolvadee Saravasee, Wipa Chawachalasai, and Suttiman Watanasrirote, for specimen and data collection; Suchitra Nimmannitya, for assistance with the investigational protocol; Sumitda Narupiti, Vipa Thirawuth, and Chonticha Klunthong, for the reverse transcriptase polymerase chain reaction assays; Chuanpis Ajariyakhajorn, for ELISAs and specimen processing; Songdej Saengsi, for specimen processing and storage; Naowayubol Nutkumhaeng and Somsak Imlarp, for virus identification and providing virus stocks; Nonglak Ongsakorn, for virus isolation by mosquito inoculation; Panor Srisongkram, Somkiat Changnak, and Wichien Sanguan-suk, for serologic assays; and Tipawan Kung vanrattana, Warinda Srikham, and Chitchai Hemachudha, for data entry and database management.
References
The investigational protocol was approved by the Institutional Review Boards of the Thai Ministry of Public Health, the Office of the US Army Surgeon General, and the University of Massachusetts Medical School. Parents or guardians of all study subjects gave written informed consent.
Financial support: National Institutes of Health (AI-34533); US Army Medical Research and Materiel Command.
The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or the Department of Defense.
Author notes
Present affiliation: Division of Virology, US Army Medical Research Institute of Infectious Diseases, Fort Detrick, Maryland.
