Ascherio A, Munger KL, Lennette ET, Spiegelman D, Hernán MA, Olek MJ, Hankinson SE, Hunter DJ. Epstein-Barr Virus Antibodies and Risk of Multiple SclerosisA Prospective Study. JAMA. 2001;286(24):3083-3088. doi:10.1001/jama.286.24.3083
Author Affiliations: Departments of Nutrition (Dr Ascherio and Ms Munger), Epidemiology (Drs Ascherio, Spiegelman, Hernán, Hankinson, and Hunter), and Biostatistics (Dr Spiegelman), Harvard School of Public Health, Center for Neurological Diseases–Multiple Sclerosis Unit (Dr Olek), and Channing Laboratory, Department of Medicine (Drs Hankinson and Hunter), Harvard Medical School and Brigham and Women's Hospital, Boston, Mass; and Virolab Inc (Dr Lennette), Berkeley, Calif.
Context Epidemiological studies suggest an association between infection with
Epstein-Barr virus (EBV) and risk of multiple sclerosis (MS).
Objective To determine whether elevation in serum antibody titers to EBV viral
capsid antigen (VCA), nuclear antigens (EBNA, EBNA-1, and EBNA-2), and diffuse
and restricted early antigen (EA-D and EA-R) as well as to cytomegalovirus
(CMV) precede the occurrence of MS.
Design, Setting, and Subjects Prospective, nested case-control study. Of 62 439 women participating
in the Nurses' Health Study (aged 30-55 years in 1976) and Nurses' Health
Study II (aged 25-42 years in 1989) who gave blood samples in 1989-1990 and
1996-1999, respectively, and were followed up through 1999, 144 women with
definite or probable MS and 288 healthy age-matched controls were included
in the analysis.
Main Outcome Measure Serum antibody titers to the specific EBV and CMV antigens, compared
between cases and controls.
Results We documented 18 cases of MS with blood collected before disease onset.
Compared with their matched controls, these women had higher serum geometric
mean titers (GMTs) of antibodies to EBV but not CMV. Elevations were significant
for antibodies to EBNA-1 (GMT, 515 vs 203; P = .03),
EBNA-2 (GMT, 91 vs 40; P = .01), and EA-D (15.9 vs
5.9; P = .04). The strongest association was found
for antibodies to EBNA-2; a 4-fold difference in titers was associated with
a relative risk (RR) of MS of 3.9 (95% confidence interval [CI], 1.1-13.7).
The corresponding RRs were 1.6 (95% CI, 0.7-3.7) for VCA, 2.5 (95% CI, 1.0-6.3)
for EBNA, 1.8 (95% CI, 1.0-3.1) for EA-D, and 1.0 (95% CI, 0.6-1.7) for CMV.
Significant but generally weaker elevations in anti-EBV antibodies were also
found in analyses of 126 cases of MS with blood collected after disease onset
and their matched controls.
Conclusions Our results support a role of EBV in the etiology of MS.
The etiology of multiple sclerosis (MS) is largely unknown, but evidence
supports an autoimmune process triggered by infection or other environmental
factors.1- 3 Epstein-Barr
virus (EBV), a herpesvirus, infects more than 90% of the human population,
establishing a persistent and highly immunogenic infection of B lymphocytes.4 Antigen-specific cytotoxic T cells are massively expanded
in response to primary infection and persist at high levels for several years.5,6 Autoimmunity could result if some of
these cells carry T-cell receptors that recognize self-peptides. Epstein-Barr
virus has been related to nasopharyngeal carcinoma, Burkitt lymphoma, and
Hodgkin disease,7 and a relation to autoimmune
diseases has been proposed8 but remains unproven.
Infection with EBV is usually asymptomatic in childhood but frequently
causes infectious mononucleosis in adolescents and adults. The similarity
in the epidemiology of MS and infectious mononucleosis led to the proposition
that MS could be caused by infection with EBV during or after adolescence
in genetically susceptible individuals.9 This
hypothesis is supported by observations suggesting an increased risk of MS
following infectious mononucleosis,10- 15
the rarity of MS among individuals without serum anti-EBV antibodies,16,17 and the higher serum titers of anti-EBV
antibodies in MS patients than in controls. Elevated titers have been reported
against both the EBV viral capsid antigen (VCA), which is expressed in the
viral replicative cycle, and the Epstein-Barr nuclear antigen (EBNA), which
is expressed in latently infected B lymphocytes.18,19
These elevations in antibody titers are consistent with an association between
infection with EBV and risk of MS16 but could
also simply reflect the immune dysregulation in MS.20
To clarify the temporal relation between antibody titers and MS, antibody
status should be assessed before clinical evidence of disease exists.
To address the possible role of EBV in the etiology of MS, we prospectively
examined the association between serum anti-EBV antibody titers and risk of
developing MS in 2 large cohorts of US women.
The study base for this investigation comprised the subsets of participants
who provided a blood sample in 2 large ongoing US cohorts, the Nurses' Health
Study (NHS) and the Nurses' Health Study II (NHS II). The NHS was established
in 1976, when 121 700 female registered nurses aged 30 to 55 years from
11 states responded to a mailed questionnaire about disease history and lifestyle
items. The NHS II was established in 1989, when 116 671 female registered
nurses aged 25 to 42 years from 14 states responded to a similar questionnaire.
All participants in both cohorts were invited to provide a blood sample to
investigate several biomarkers of chronic disease. Blood was collected from
32 826 participants in the NHS between 1989 and 1990 and from 29 613
women in the NHS II between 1996 and 1999. Follow-up questionnaires are mailed
to the participants of both studies every 2 years to update information on
potential risk factors for chronic diseases and to ascertain whether major
medical events have occurred.
Details on the documentation of MS cases in these cohorts have been
reported previously.21 Briefly, participants
who reported a new diagnosis of MS in 1 of the follow-up questionnaires were
asked permission to contact their neurologists and review their medical records.
After obtaining permission, we sent the neurologists a questionnaire addressing
the certainty of the diagnosis (definite, probable, possible, or not MS),
the date of onset of neurological symptoms related to MS, other aspects of
the clinical history, and laboratory test results. Since 93% of all definite
and probable diagnoses appeared to conform to the Poser criteria22
for diagnosis of MS when applied to the clinical and laboratory data provided
in the questionnaire,23 we classified as cases
women who had a diagnosis of definite or probable MS according to their neurologists.
The date of onset of the disease was determined by asking both women
with MS and their neurologists for the date of first neurological symptoms.
When the 2 dates were discordant, the earliest date was considered valid.
A total of 149 incident cases of definite and probable MS were documented
between baseline and December 1999 among women with available blood samples.
For each case, we randomly selected 2 women without MS, matched by year of
birth and study cohort. One case was excluded because no serological results
were obtained and 4 because of missing dates of onset; thus, 144 women with
MS and 288 controls were included in the analysis.
In 18 of the women with MS, the onset of neurological symptoms occurred
after blood collection (median, 1.9 years; range, 2 months–6.5 years).
The diagnoses of MS in these women were all made by neurologists and supported
by magnetic resonance imaging; the diagnosis was reported as definite MS in
12 women and probable MS in 6. Because of the age distribution of the cohorts
at the time of blood collection, most of these women had a late onset of MS
(median age, 52 years; range, 39-66 years).
Blood was obtained using collection kits that were returned to our laboratory
via overnight courier. Approximately 97% of the samples arrived within 26
hours of being drawn. On arrival in our laboratory, the blood samples were
centrifuged and the blood components aliquotted into cryotubes and stored
in liquid nitrogen freezers until laboratory analysis. Serum samples from
MS cases and controls were sent to the laboratory (Virolab Inc, Berkeley,
Calif) in triplets containing the case and the 2 matched controls in random
order. The laboratory was blind to case-control status and unidentified triplets
were included for quality control.
Immunoglobulin G antibodies to EBV VCA and anti–early antigen
complex (diffuse [EA-D] and restricted [EA-R]) were determined by indirect
immunofluorescence according to methods of Henle et al.24
Immunoglobulin G antibodies against the EBNA family and 2 of its individual
components, EBNA-1 and EBNA-2, were determined using anticomplement immunofluorescence,
as previously described.25 Antibodies to the
VCA and the EA-D components emerge during the late incubation period or in
the course of the acute phase of infectious mononucleosis, whereas antibodies
to EBNA and EA-R arise only weeks or months after onset of the disease.24 Antibodies to EBNA-2 arise before those to EBNA-1
and usually decline over several months, whereas antibodies to EBNA-1 persist
indefinitely. Persistently high anti–EBNA-2 titers and low anti–EBNA-1
titers have been associated with chronic infectious mononucleosis.24 Antibody titers against cytomegalovirus (CMV) were
also determined to assess the specificity of any association that may be found
between MS and EBV serology. The methods used for CMV serology have been previously
described.26 The intra-assay coefficients of
variation were VCA, 8.6%; EBNA, 13.9%; EBNA-1, 9.8%; EBNA-2, 9.1%; EA-D, 13.0%;
EA-R, 9.2%; and CMV, 7.3%.
We considered in the analyses age, latitude of birthplace (northern,
middle, or southern), ancestry (Scandinavian, Southern European, other white,
or nonwhite), and smoking (never, past, or current), data on which were collected
from all participants as part of the cohort follow-up. The associations of
birthplace latitude, ancestry, and smoking history with risk of MS in these
cohorts have been previously reported.23,27
Antibody geometric mean titers (GMTs) were compared between cases and
controls using generalized linear models that take into account the matched
design of the study and the use of clustered measurements.28
We used conditional logistic regression to estimate the relative risk (RR)
of MS associated with a 4-fold difference in specific antibody titers. In
these regression models, we included the base 2 logarithm of the reciprocal
of the dilution of the antibody titers as a continuous variable. On this scale,
the regression coefficient estimates the logarithm of the RR associated with
a 2-fold difference in titers; we doubled and exponentiated this value to
estimate the RR associated with a 4-fold difference.
The main analyses included only women with MS who provided blood samples
before onset of the disease compared with their matched controls or with all
controls combined. However we also presented results for women with blood
collected after onset of MS to assess whether there is any effect of the disease
on antibody titers. Finally, we classified women according to whether they
had evidence of past infection with EBV or CMV, to estimate the risk of MS
associated with prior infection. For this analysis, women were considered
EBV-positive if the antibody titer to VCA was at least 1:20 or that to EBNA
was at least 1:5; women were considered CMV-positive if the antibody titer
to CMV was at least 1:10. The association between seropositivity and MS was
estimated using the Mantel-Haenszel method for matched data29
or conditional logistic regression for multivariate analyses. All P values are 2-tailed. The statistical software SAS (SAS Institute
Inc, Cary, NC) was used for all analyses.
The GMTs of serum antibodies to EBV were consistently higher among women
with MS compared with their matched controls. In analyses including only the
18 cases with blood collected before disease onset and their matched controls,
these differences were significant for antibodies to EBNA-1 (GMT for cases,
515 vs for controls, 203; P = .03), EBNA-2 (GMT,
91 vs 40; P = .01), and EA-D (GMT, 15.9 vs 5.9; P = .04). The GMTs to VCA and EBNA were also higher among
cases than controls (GMT for VCA, 1613 vs 1036; GMT for EBNA, 667 vs 333),
but these differences only achieved statistical significance in analyses of
the 126 cases with blood collected after onset of MS. There was no difference
between cases and controls in the GMTs of antibodies to EA-R or CMV. The case-control
differences in antibody titers to EBV were of similar magnitude whether the
blood was collected before or after disease onset, although the larger number
of cases with blood collected after disease onset conferred a stronger degree
of statistical significance. A direct comparison of these associations stratified
by time of blood collection is shown in Table 1. In matched analyses, the RRs of MS associated with a 4-fold
difference in antibody titers before onset of MS ranged from 1.6 (95% confidence
interval [CI], 0.7-3.7) for antibodies to VCA to 3.9 (95% CI, 1.1-13.7) for
antibodies to EBNA-2. Significant positive associations with antibodies to
EBNA-1, EBNA-2, and EA-D were also observed in age-adjusted analyses including
all controls (Table 1). A plot
of this association for EBNA-1 is shown in Figure 1; results for EBNA-2 were similar. The corresponding RR
for antibody titers collected after onset of MS were smaller, except for VCA
(1.7 vs 1.6). The RRs associated with antibodies to CMV were close to 1.0
for all comparisons.
To address the possibility that the increased anti-EBV antibody titers
among the cases with blood collected before onset were due to preclinical
disease, we calculated the mean antibody titers among 12 women with MS who
provided a blood sample at least 1 year before the onset of MS and plotted
the antibody titers of women with MS according to the temporal relation between
blood collection and disease onset (Figure
2). Serum antibody titers to EBV antigens were already elevated
among women who gave blood at least 1 year before onset of MS and were unrelated
to the time of blood collection, except for a decline in anti–EA-D after
MS onset (Figure 2).
The 18 women with MS who gave blood before onset of disease were all
positive for both VCA and EBNA, whereas 2 of their matched controls were negative
for VCA and 1 was negative for EBNA (P = .30 and P = .50, respectively). Using the combined sample of MS
subjects, only 1 of 144 women with MS had negative anti-VCA titers vs 18 of
287 controls (OR, 9.0; 95% CI, 1.8-45.2; P = .008)
(Table 2). Results for anti-EBNA
titers were similar. In contrast, the proportions of women with serum anti-CMV
antibodies were similar in the case and control groups (Table 2). These associations were not materially changed in multivariate
In this prospective study, we found that significant elevations in serum
anti-EBV antibody titers were present before onset of MS. Elevations were
observed for antibodies against antigens expressed during replication of the
virus (EA-D) as well as in latency (EBNAs). The strongest association was
found with antibodies to EBNA-2; a 4-fold difference in titers of antibodies
to this antigen was associated with a 4-fold increase in risk of MS.
The nested case-control design of our investigation makes a biased selection
of controls unlikely. Some degradation of the IgG antibodies or desiccation
of blood samples may have occurred during shipping and storage, but these
are probably modest because most samples reached our serum bank within 26
hours and were kept in closely monitored liquid nitrogen freezers. Most importantly,
blood samples from cases and controls were handled in the same manner throughout
the study, and triplets composed of a case and 2 matched control samples were
assayed in the same run and in random order by technicians who were blind
to disease status. Under these conditions, any laboratory error should be
unrelated to disease status and would attenuate the difference in antibody
titers between cases and controls.
A potential limitation of our study is the relatively short period between
blood collection and onset of MS (median, 1.9 years). The onset of the autoimmune
process that leads to demyelination may precede the recognition of neurological
symptoms by several months and so may the immune dysregulation that accompanies
the disease. Thus, the elevation in anti-EBV titers could be an early manifestation
of the preclinical phase of the disease. The fact that antibody titers in
serum samples collected 1 or more years before onset of symptoms were similar
to those in serum samples collected in women with fully clinical disease provides
some evidence against this explanation, but confirmation of these results
in a larger number of serum samples collected years before MS onset will be
important. The late age of onset of MS among women in our study is atypical,
but it is consistent with the age distribution of our cohorts at the time
of blood collection.
A causal association between EBV and MS was originally suggested 20
years ago,9 primarily because of the similarities
in the epidemiology of infectious mononucleosis and MS and of the increased
risk of MS among individuals with a history of infectious mononucleosis.10- 15
If EBV had a critical role in the etiology of MS, then it would be expected
that few or no MS cases would occur among individuals who are not infected
with EBV. This expectation is indirectly supported by the results of several
case-control studies. In a meta-analysis of published investigations, we estimated
that the odds of disease are more than 10 times higher among EBV-positive
than EBV-negative persons.16 Since primary
EBV infection is rare among patients with MS,17,30
this finding suggest that either EBV itself increases the risk of MS or that
they both are related to some common factor. A microorganism transmitted in
a manner similar to EBV could be involved, but the EBV-MS association in so
strong that it is unlikely to be fully explained by confounding. A common
genetic predisposition may be more plausible, because of the known association
between HLA class II polymorphism and risk of MS and the recent observation
that the HLA class II protein acts as a cofactor in EBV infection of B lymphocytes.31 A study of long-term EBV-negative adults, however,
revealed a distribution of the DR2 alleles commonly associated with MS similar
to that of EBV-seropositive adults,32 and higher
titers of anti-EBV antibodies have been reported in individuals with MS than
in HLA-DR2–matched controls.33 These
observations suggest that genes are unlikely to explain the strong associations
reported herein. Further evidence that EBV has an active role in MS recently
has been provided by the observation that active viral replication occurs
more commonly in MS patients with exacerbations than in patients with stable
The elevation of anti-EBV antibody titers before onset of MS that we
found in this study provides further support for an important role of EBV
in the etiopathogenesis of MS. The observation of elevated anti-EBV serum
antibody titers before diagnosis has contributed to establishment of the causal
link between EBV and Burkitt lymphoma34 and
has been reported in other EBV-related diseases, including nasopharyngeal
carcinoma35 and Hodgkin disease.36
The pattern of antibody response that we observed among women with MS is different
from that observed in immunocompromised hosts or in chronic infectious mononucleosis,
which is characterized by elevated anti-EBNA-2 and reduced anti-EBNA-1 titers.25 The simultaneous elevation of titers to VCA, EBNA-1,
EBNA-2, and EA-D rather suggests a more severe or more recent primary infection
among women who developed MS than in women who remained healthy.
Evidence on potential mechanisms by which EBV could be causally related
to MS is limited. The failure to demonstrate EBV in MS plaques by in situ
hybridization37 or polymerase chain reaction38 suggests that direct central nervous system infection
is not involved. Rather, the T-cell response to EBV infection could include
clones that are potentially cross-reactive with self-antigens. In acute infectious
mononucleosis, there is a massive expansion of activated circulating T cells
that are specific for EBV latent and lytic antigens.6
A high proportion of these cells is committed to a single viral epitope,5 and populations of cells that are specific for both
latent and lytic antigens are still present at frequencies of more than 1%
up to 3 years after recovery.6 Other potential
mechanisms proposed to explain a potential role of EBV in MS include cross-reactions
of anti-EBNA antibodies with epitopes of a neuroglial antigen39
and the EBV-induced expression in B lymphocytes of α-β-crystallin,
a small stress protein that has been reported to be an immunodominant myelin
antigen in MS patients.40 It has recently been
reported that variant A of human herpesvirus 6 is more commonly reported in
MS patients than in controls.41- 43
Unlike the more common human herpesvirus 6 variant B, variant A may infect
EBV-positive B-cell lines and activate the latent EBV genome.44,45
These observations suggest that interactions between herpesviruses may have
a role in the pathogenesis of MS.
In conclusion, our results, in conjunction with those of case-control
studies, offer evidence that EBV infection may increase the risk of MS. Because
few individuals infected with EBV develop MS, other cofactors are required.
These may include genetic predisposition and, perhaps, age at primary infection
or infection with other microbes. Final proof of causality could come from
the demonstration that a suitable vaccine prevents MS. Available antiviral
drugs have little effect on the number of infected B lymphocytes and may thus
be ineffective in MS treatment.7 However, a
better understanding of the mechanism that relate EBV to MS may also lead
to novel therapeutic approaches.