|Home | About | Journals | Submit | Contact Us | Français|
Background.University students were studied prospectively to determine the incidence of and risk factors for acquisition of primary Epstein–Barr virus (EBV) infection and the virologic and immune correlates of disease severity.
Methods.EBV antibody–negative freshmen participated in monthly surveillance until graduation. If antibodies developed, proximate samples were assayed for viral load by polymerase chain reaction. Lymphocyte and natural killer (NK) cell numbers and activation were measured by flow cytometry, and plasma cytokine levels were measured by a multiplex assay.
Results.Of 546 students screened, 202 (37%) were antibody negative; 143 antibody-negative students were enrolled. During a median of 3 years of observation, 66 subjects experienced primary infection. Of these, 77% had infectious mononucleosis, 12% had atypical symptoms, and 11% were asymptomatic. Subjects reporting deep kissing with or without coitus had the same higher risk of infection than those reporting no kissing (P < .01). Viremia was transient, but median oral shedding was 175 days. Increases were observed in numbers of NK cells and CD8+ T-cells but not in numbers of CD4+ T-cells during acute infection. Severity of illness correlated positively with both blood EBV load (P = .015) and CD8+ lymphocytosis (P = .0003).
Conclusions.Kissing was a significant risk for primary EBV infection. A total of 89% of infections were symptomatic, and blood viral load and CD8+ lymphocytosis correlated with disease severity.
Epstein-Barr virus (EBV) was discovered in 1964  and conclusively linked to infectious mononucleosis in 1968 . More than 4 decades later, however, the incidence and risk factors for acquisition and correlates of severity of primary EBV infection are incompletely understood. The prevalence of EBV antibodies varies widely by age and geographic location [3–5], and the inference has been that the later in life EBV infection is acquired, the more likely it will be symptomatic . However, the proportion of symptomatic infections in any age group is not well described. Furthermore, infectious mononucleosis has been postulated to be an immunopathologic disease whose symptoms are caused by the CD8+ response to the virus rather than to the virus per se. Indeed, CD8+ lymphocytosis—not viremia—was associated with symptomatic disease in a previous study of primary EBV infection .
An accurate assessment of the burden of disease attributable to primary EBV infection is an important component in the development of a prophylactic vaccine or antiviral therapy. This information can only be obtained by a prospective study, which we were able to perform because we had access to >4000 resident freshmen on the campus of the University of Minnesota, whose rate of natural infection is relatively high . This report describes our findings in students from 2 freshman classes followed prospectively throughout their undergraduate years with frequent reporting of health histories, clinical monitoring, quantitative virologic testing, and immunologic testing. We measured the following factors to define their range, kinetics, and correlation with disease severity during primary infection: blood and oral viral loads; CD8+, CD4+, and natural killer (NK) cell numbers; and T-cell activation markers.
This study, which was approved by the Research Subjects Protection Program of the University of Minnesota, had a screening and a surveillance phase.
Screening was conducted in 3 freshman residence halls during September and October 2006 for the Class of 2010 and during September 2007 for the Class of 2011. Students who gave informed consent completed a history form and donated 10 mL of venous blood. Historical information collected included date and place of birth, birth order in the family, number of siblings, and prior illnesses resembling infectious mononucleosis. Sera or plasma were assayed for EBV antibodies as described below. Results were discussed with the students, and those who were antibody negative were invited to participate in the surveillance phase.
Subjects who consented to participate in surveillance made periodic visits to the research clinic throughout their undergraduate years. They were seen at least every 8 weeks when school was in session and during school breaks if they remained in the vicinity. During the visits a medical history, an oral wash specimen, and 40 mL of venous blood were obtained, and subjects completed a health questionnaire that included information on sexual behavior, exercise, stress, and diet.
Sera collected at surveillance visits were tested immediately for EBV antibodies. If antibodies developed, subjects were seen as soon as possible for a “sick visit.” Subjects also reported signs and symptoms suggestive of an acute EBV infection, either electronically or by phone. Clinical data specifically tracked were sore throat, cervical lymphadenopathy, fatigue, fever, headache, body aches, decreased appetite, abdominal pain, and upper respiratory tract symptoms. Sick visits were scheduled as soon as possible after symptoms were reported. At all sick visits, a medical history, a physical examination, oral wash sampling, and a venous blood draw were performed. These oral and blood samples, together with those from the proximate past 2 visits, were tested for EBV antibodies and viral loads as described below. Primary EBV infection was defined as a positive result of an EBV antibody test and the presence of EBV DNA in the oral and/or blood compartment of a subject who was previously negative for both EBV antibodies and EBV DNA.
Primary EBV infection was classified clinically as infectious mononucleosis (at least 2 of the following symptoms were present: sore throat, cervical lymphadenopathy, fever, and fatigue), symptomatic (symptoms were present but did not fulfill the definition of infectious mononucleosis), or asymptomatic. The severity of illness was evaluated for subjects with laboratory-documented primary EBV infection, using a categorical scale from 0 (asymptomatic) to 6 (essentially bedridden) as previously published [8, 9]
Subjects were seen once per semester after they had experienced laboratory-documented primary EBV infection. Subjects were followed until they graduated or left the University of Minnesota before graduation and moved out of the vicinity.
The screening tests were enzyme immunoassays (EIAs) for EBV viral capsid antigen (VCA) and EBNA-1 immunoglobulin G (IgG) antibodies . Subjects negative for both antibodies were considered to be EBV naive and eligible to enroll in the surveillance phase.
Subjects were tested at least every 8 weeks by EIA for EBV VCA IgG antibodies. Subjects who developed VCA IgG antibodies or reported symptoms consistent with primary EBV infection were also tested by EIA for EBV VCA immunoglobulin M (IgM) and EBNA-1 IgG antibodies  and for heterophile antibodies against bovine erythrocytes, by means of a commercially available chromatographic immunoassay (Inverness Medical).
Real-time quantitative polymerase chain reaction (PCR) was performed on oral cells, oral supernatant, and whole blood as previously detailed . Briefly, the amplicon was a 71–base pair portion of the EBNA-1 gene. The reliable limit of detection was 4 copies/reaction, which equates to 16 copies/mL for the oral wash cell pellet and 80 copies/mL for the oral wash supernatant and whole blood. Results were expressed as log10 copies of EBV DNA per milliliter of sample.
Peripheral blood mononuclear cells (PBMCs) from each subject were stained with antibodies against CD3, CD4, CD8, CD56, CD38, and HLA-DR (e-Bioscience, Biolegend, or Invitrogen) and with peptide–major histocompatibility class I tetramers (NIH tetramer facility or Beckman Coulter). To detect granzyme B, cells were washed, permeabilized, and stained using the BD Cytofix/Cytoperm Kit per the manufacturer's instructions (BD Biosciences). Samples were analyzed on a LSR II (Becton Dickinson), and all data were processed using the FlowJo Software (Treestar).
Plasma was separated from whole blood cellular components via centrifugation and was stored frozen. Cytokine levels in plasma were determined with a multiplex Luminex kit (Panomics) and read on a BioPlex 200 reader (BioRad). Samples prior to EBV infection were used as subjects' baseline samples.
Statistical analysis was performed using Prism software (Graphpad), SAS, version 9.2 (SAS Institute, Cary, NC), or R (R Foundation for Statistical Computing, Vienna, Austria). Comparisons between groups were performed with 2-sample 2-tailed t tests, for continuous outcomes, and with χ2 analysis, for categorical outcomes. Spearman rank correlation coefficients were calculated to assess associations. Baseline characteristics were assessed as risk factors by a χ2 test and by comparing Kaplan-Meier estimates of time to EBV infection, using a log-rank test. Time-varying characteristics, such as final examination periods, were assessed as risk factors by testing the time-varying predictor within a proportional hazards model. Cumulative annual infection rates were computed by a life table method (SAS Proc Lifetest); cumulative infection rates were compared between groups by the Wilcoxon test. Annual incidence rates were compared by a linear contrast within a Poisson regression model.
Of 546 freshmen from the classes of 2010 and 2011 who were screened for EBV antibodies, 344 (63%) were positive and 202 (37%) were negative. Their median age was 18.6 years (mean, 18.6 years; range, 18.0—22.1 years). The prevalence of EBV antibody was nearly identical for each class: 64% (172 of 267), for the Class of 2010, and 62% (172 of 279), for the Class of 2011.
Antibody prevalence was higher among women (215 of 326; 66%) than men (129 of 220; 59%), but the difference was not statistically significant. Antibody prevalence was greater among subjects with a history of infectious mononucleosis (12 of 15; 80%) versus those with a negative history (332 of 531; 63%), but the difference was not statistically significant. Birthplace, birth order in the family, household size, and age at screening were not related to antibody prevalence (data not shown).
Of the 202 antibody-negative students, 143 (71%) were enrolled in the surveillance phase. They were similar to the group screened and the eligible pool of EBV-naive subjects in terms of age, sex, race/ethnicity, birthplace, birth order in the family, and number of siblings (data not shown). The surveillance cohort was 93% white (133 subjects), 5% Asian (7 subjects), 1.4% black (2 subjects), and 0.7% native Alaskan (1 subject).
During the surveillance period, the 143 participants made 2549 clinic visits (median, 18.0 visits per subject; mean, 17.8 visits per subject). The overall EBV infection incidence was 14.4 cases per 100 person-years at risk. Sixty-six subjects (45 women and 21 men) experienced a primary EBV infection during 314.3 person-years of observation. The incidence of infection during the freshman year (26 cases per 100 person-years) was more than twice the mean incidence during the following 3 years (10 cases per 100 person-years; P = .002) (Figure 1). The incidence of primary EBV infection was higher in women than in men (23.6 vs 16.1 cases/person-year), but this difference was not statistically significant.
Exercise, stress, and diet were not risk factors for EBV infection, but sexual behavior was a risk factor. As depicted in Figure 2, subjects reporting deep kissing with or without coitus had very similar distributions of time to infection, and both groups were significantly different than the group of subjects who reported no kissing and no coitus and experienced no primary EBV infections (P < .01).
The clinical picture of primary EBV infection was infectious mononucleosis, for 51 subjects (77%); symptomatic but not meeting the definition of infectious mononucleosis, for 8 (12%); and asymptomatic, for 7 (11%). Clinical findings reported by the 59 symptomatic students were sore throat, for 55 subjects (93%); cervical lymphadenopathy, for 45 (76%); fatigue, for 39 (66%); upper respiratory tract symptoms, for 36 (61%); headache, for 28 (47%); decreased appetite, for 26 (44%); fever, for 25 (42%); body aches, for 24 (41%); and abdominal pain, for 5 (8%). The severity of illness scores (Figure 3A) were distributed over the entire range of measurement, from asymptomatic (0) to essentially bedridden (6). The median duration of illness among symptomatic subjects was 10 days (mean, 17.0 days; range, 3–66 days). The duration of illness was significantly longer for subjects whose severity of illness was 2–6, compared with those whose severity of illness was 1 (P < .05; Figure 3B).
Virologic data from the 66 subjects with prospectively identified primary EBV infections are summarized in Table 1. All 66 infected students had EBV DNA in their oral cell pellet. The oral cells were positive for a median of 175 days (mean, 272 days). EBV was present in lower quantities and for a shorter period in the oral supernatant, which essentially represented cell-free virus. EBV DNAemia was documented in 42 subjects (64%), 26 of whom were viremic for <1 week. In 16 subjects, EBV DNA was present in the whole blood for 16–202 days (median, 95.0 days; mean, 96.5 days). Viral shedding was detected in 14 subjects prior to the onset of symptoms. Thirteen subjects had virus in the oral compartment 2–36 days before symptoms developed. Six subjects were viremic 4–18 days before onset of symptoms, and 9 had recurrent DNAemia following the acute illness.
Heterophile antibodies were documented in 50 (77%) of 65 subjects, and VCA IgM antibodies were detected in 54 (83%) of 65 subjects. VCA IgM antibodies were found as early as 8 days before symptoms were reported and persisted for as long as 420 days after onset of symptoms. All 66 subjects with primary EBV infection developed EBV VCA IgG antibody that was first detected as early as the day of onset of symptoms and as late as 91 days after onset of symptoms. EBNA-1 antibodies were documented in 57 (92%) of 62 subjects. They usually were present by 90 days after onset of symptoms, but they appeared as early as 25 days after symptom onset and took >1 year to develop in 2 subjects.
During 125.5 person-years of observation following primary EBV infection, there were 2 reported clinical relapses, only one of which was accompanied by an increase in EBV load.
We quantified CD8+ T-cell numbers and activation over time in study subjects (Figure 4). CD8+ T-cell numbers increased most dramatically during the first 2 weeks following symptom onset (Figure 4A). The emergence of EBV-specific CD8+ T cells during this time frame was confirmed using HLA-restricted class I tetramers . CD4+ T-cell numbers did not increase (Figure 4B). We performed phenotypic analysis of both total and EBV-specific CD8+ and CD4+ T-cell activation, using HLA-DR and CD38 as markers of activation, in addition to a marker of cytolytic activity, granzyme B. These markers have been used to characterize the T-cell activation in YF-Vax and Dryvax vaccine studies  and in other viral infections [12, 13]. We observed a significant upregulation of CD38, HLA-DR, and granzyme B on total CD8+ T cells and, to a lesser extent, on CD4+ T cells (Supplementary Figure 1E–H). This occurred primarily within the first 2 weeks of symptom onset. Additionally, downregulation of granzyme B on CD8+ T cells occurred at a slower rate relative to that of HLA-DR and CD38 (Supplementary Figure 1).
Virus was usually undetectable in the blood or oral fluids prior to the onset of symptoms (Figure 4D and Supplementary Figure 1C and 1D). For most individuals, EBV in the blood was no longer detectable after 200 days; however, virus in the oral cavity was detectable up to 3 years later (Supplementary Figure 1D).
We also determined the expansion and activation of NK cells during acute EBV infection (Figure 4C). While the activation of NK cells, as determined by the granzyme B level, was not significantly different during EBV infection (data not shown), we observed an expansion of NK cell numbers in the blood during acute disease (Figure 4C and Supplementary Figure 1B). The window of NK cell expansion was similar to that observed for CD8+ lymphocytosis (Figure 4).
We observed a positive association between severity of illness and the number of CD8+ but not CD4+ T cells in the blood (Figure 5A and and55B). Severity of illness was likewise positively correlated with CD8+ but not CD4+ T-cell granzyme B expression (Supplementary Figure 2A and 2B). Therefore, consistent with previous data, we found a strong positive correlation between CD8+ lymphocytosis and disease severity. However, contrary to a smaller clinical study , we observed a positive correlation between disease severity and the level of EBV in whole blood (Figure 5C), but the correlation was less strong in the oral cavity (Figure 5D). If most or all of the CD8+ T-cell expansion and activation is EBV specific, as has been suggested by others [14, 15], it is possible that a high viral load drives CD8+ lymphocytosis. In agreement with this, we detected a strong positive correlation between whole blood (but not oral) EBV load and CD8+ lymphocytosis (Supplementary Figure 2C and 2D).
Finally, NK cell numbers correlated positively with CD8+ T-cell numbers (Supplementary Figure 2E and 2F) and with blood but not with oral viral loads (data not shown). For each individual, the maximum NK cell expansion occurred in those with the maximum CD8+ T-cell expansion (Supplementary Figure 2F), suggesting that the same mechanism was responsible for driving expansion of both populations to create the immunopathological response characteristic of infectious mononucleosis.
In addition to having cytotoxic potential, EBV-specific T cells make interferon γ (IFN-γ) during acute EBV infection . Indeed, IFN-γ can cause symptoms similar to those observed during infectious mononucleosis . Thus, we hypothesized that IFN-γ levels might correlate with disease severity. Since levels of several cytokines are elevated during infectious mononucleosis , we identified and measured 8 cytokines that might contribute to disease pathogenesis. The cytokines studied consisted of antiviral cytokines (type I and II IFNs), innate immune/inflammatory cytokines (interleukin 6 [IL-6], interleukin 12, and tumor necrosis factor α), and suppressive cytokines (interleukin 10 and transforming growth factor β).
While levels of several cytokines were elevated during acute infection, only the IL-6 level was correlated significantly with severity of illness (Figure 6 and Supplementary Figure 3). Also, in subjects whose maximum viremia was >3.5 log10 copies/mL, IL-6 levels were significantly higher, compared with those in subjects with low or absent viremia (63% increase; P = .008). Interestingly, IFN-γ levels did not correlate significantly with severity of illness or viremia (Figure 6 and data not shown).
Undergraduate students acquired primary EBV infections at the rate of 14.4 cases per 100 person-years, and 89% of these infections (59 of 66) were symptomatic. Freshmen had a significantly higher incidence of infection as compared with sophomores, juniors, and seniors. Deep kissing with or without coitus was the only significant risk factor we identified for acquisition of infection. Crawford et al  suggested that spread of EBV is enhanced by penetrative sexual intercourse, but we found that subjects who reported deep kissing with or without penetrative sexual intercourse had the same higher risk of primary infection as those who reported no kissing. The persistence of large quantities of virus in the oral cells and supernatant after primary EBV infection supports our view that oral secretions are the major source of transmission. EBV has been recovered from the male and female genital tracts [19, 20], but neither the quantity nor the duration of genital viral shedding has been described.
We observed that the quantity of EBV in whole blood, CD8+ lymphocytosis, and increased NK cells all correlated positively with disease severity. Since CD8+ T cells produce IFN-γ, which has an elevated level in the serum during infectious mononucleosis, and because IFN-γ treatment can result in mononucleosis-like symptoms, it could be argued that CD8+ T-cell production of high levels of IFN-γ might cause some of the symptoms classically associated with infectious mononucleosis. However, we did not observe a positive correlation between IFN-γ levels and disease severity. Furthermore, unlike a previous study by Silins et al , we found that blood viral load also correlated positively with severity of illness, as it did with CD8+ and NK cell numbers. This is consistent with our earlier study, which reported that the rate of viral elimination from whole blood was similar to the rate of improvement in severity of illness . Thus, it is possible that high viral load may induce multiple immune, inflammatory, and cytopathic effects, making it difficult to conclude that infectious mononucleosis is simply a CD8+ T cell–mediated immunopathologic disease.
A study of NK cell numbers and severity of pharyngitis during primary EBV infection correlated NK cell expansion with less severe symptoms . There was also an inverse correlation between NK cells and blood viral load, suggesting that NK cell elimination of EBV-infected B cells contributed to milder illness. In contrast, we describe a positive correlation between NK cell expansion, blood viral load, and disease severity. It should be noted that interpretation of the data presented by Williams et al  is confounded by the fact that NK cell numbers were examined in association with disease severity, but NK cell frequency (ie, percentages) was examined in association with viral load in the blood. Because of the substantial increase in CD8+ T cells during acute infectious mononucleosis, use of NK cell percentages may underestimate the changes in NK cell expansion. Second, Williams et al  examined viral load per 106 PBMCs, whereas we represented viral load as the log-transformed value of the EBV load per milliliter of whole blood. Importantly, our conclusions are based on a larger set of subjects with internal controls for NK cell numbers and frequency before, during, and after EBV infection. Nonetheless, NK cells are a heterogeneous population, and it is possible that expansion of a particular subset of NK cells would be associated with better viral control and reduced symptoms. Further work on this important question is required.
A positive correlation between the blood viral load and host response factors such as increased CD8+ and NK cell numbers may not seem surprising. However, our finding is inconsistent with a paradigm often discussed in the context of EBV infection, which asserts that preexisting aspects of the CD8+ T-cell repertoire, such as memory T cell reactivity to other viruses, would result in a more severe primary EBV infection, independent of the viral load . Rather, it seems that host response factors that allow higher blood viral loads in the 4–6 weeks following initial infection result in more severe symptoms. Unfortunately, very little is understood about the virus-host dynamics during the long incubation period of this virus, which has compelled computer-based modeling of infection dynamics [23, 24]. A prospective study with more frequent observations during the incubation period may shed light on this important period of host-virus interaction.
In summary, acquisition of primary EBV infection in undergraduate students was associated with kissing, the vast majority (89%) of infections were symptomatic, and disease severity was related to high numbers of NK and CD8+ T cells and elevated blood viral loads. Our data suggest that preventing or reducing viremia by immunization or antiviral therapy could lessen the morbidity of primary EBV infection.
Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Acknowledgments.We thank Tony Thomas and Richard C. Brundage, for ideas regarding risk factors, and Juan Reyes, for technical support.
Financial support.This work was supported by the University of Minnesota International Center for Antiviral Research and Epidemiology, the Minnesota Medical Foundation, and the National Institutes of Health (F31AI084524 and T32 CA009138 to O.A.O, and PO1AI35296 to K. A. H.).
Potential conflicts of interest.All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.