Proc. Natl. Acad. Sci. USA
Vol. 96, pp. 3848–3853, March 1999
HLA alleles determine human T-lymphotropic virus-I (HTLV-I)
proviral load and the risk of HTLV-I-associated myelopathy
KATIE J. M. JEFFERY†‡, KOICHIRO USUKU‡¶, SARAH E. HALL†, WATARU MATSUMOTO§, GRAHAM P. TAYLOR?,
JEANETTE PROCTER††, MIKE BUNCE††, GRAHAM S. OGG‡‡, KENNETH I. WELSH††, JONATHAN N. WEBER?,
ALUN L. LLOYD§§, MARTIN A. NOWAK§§, MASAHIRO NAGAI¶¶, DAISUKE KODAMA§, SHUJI IZUMO¶¶,
MITSUHIRO OSAME§, AND CHARLES R. M. BANGHAM†??
Departments of†Immunology and?Genito-Urinary Medicine and Communicable Diseases, Imperial College School of Medicine, St. Mary’s, Norfolk Place,
London W2 1PG, United Kingdom;¶Department of Medical Informatics,§Third Department of Internal Medicine, and¶¶Division of Molecular Pathology,
Center for Chronic Viral Diseases, Faculty of Medicine, Kagoshima University, 8-35-1 Sakuragaoka, Kagoshima 890-8520, Japan;††Oxford Transplant Center,
Nuffield Department of Surgery, Churchill Hospital, Oxford OX3 7LJ, United Kingdom;‡‡Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3
9DU, United Kingdom; and§§Department of Zoology, University of Oxford, Oxford OX1 3PS, United Kingdom
Communicated by Robert May, Univesity of Oxford, Oxford, United Kingdom, January 15, 1999 (received for review October 30, 1998)
virus infections such as HIV-I, hepatitis B and C, and human
T-lymphotropic virus-I (HTLV-I) is strongly determined by
the virus load. However, it is not known whether a persistent
class I HLA-restricted antiviral cytotoxic T lymphocyte (CTL)
response reduces viral load and is therefore beneficial or
causes tissue damage and contributes to disease pathogenesis.
HTLV-I-associated myelopathy (HAM?TSP) patients have a
high virus load compared with asymptomatic HTLV-I carri-
ers. We hypothesized that HLA alleles control HTLV-I pro-
virus load and thus influence susceptibility to HAM?TSP.
Here we show that, after infection with HTLV-I, the class I
allele HLA-A*02 halves the odds of HAM?TSP (P < 0.0001),
HLA-A*02?healthy HTLV-I carriers have a proviral load
one-third that (P ? 0.014) of HLA-A*02?HTLV-I carriers. An
association of HLA-DRB1*0101 with disease susceptibility
also was identified, which doubled the odds of HAM?TSP in
the absence of the protective effect of HLA-A*02. These data
virus load is associated with prognosis and imply that an
efficient antiviral CTL response can reduce virus load and so
prevent disease in persistent virus infections.
The risk of disease associated with persistent
infectious disease in humans. In 1974, Zinkernagel and
Doherty (1) showed that the cytotoxic T lymphocyte (CTL)
response to virus infections was restricted by class I alleles of
the MHC. Surprisingly, however, it has been difficult to
demonstrate a direct protective effect of class I MHC alleles
against a viral infection in either human or animal populations.
There have been reports of protective class I alleles in HIV-1
infected long-term nonprogressors (2–5), but the results have
not been consistent. HLA class II alleles have been associated
with both susceptibility to and protection from viral diseases,
e.g., hepatitis B and human papilloma virus; the immunoge-
netics of infectious diseases has recently been reviewed by Hill
In chronic virus infections such as HIV-1 and 2, hepatitis B
virus, and hepatitis C virus, virus load is an important deter-
minant of the outcome of infection and disease. Recent
evidence suggests that provirus load is also an important factor
in the outcome of human T-lymphotropic virus-I (HTLV-I)
infection (7–10). HTLV-I is a persistent virus, infecting 10–20
million people worldwide. Most infected people remain
healthy, but 1–2% develop a progressive paralytic myelopathy
(HTLV-I-associated myelopathy; HAM?TSP) and a further
2–3% develop an aggressive T cell leukemia?lymphoma. The
reasons for the different outcomes of infection are unknown.
HAM?TSP is a chronic debilitating inflammatory disease of
the central nervous system, characterized by axonal damage
and demyelination, most pronounced in the midthoracic spinal
cord (11). The HTLV-I proviral load is 10- to 100-fold greater
in HAM?TSP patients than in asymptomatic healthy carriers
(HCs) of the virus (9, 12), although the ranges overlap. The
pathogenesis of this condition is not understood.
We have shown previously that no particular sequence of
HTLV-I is associated with neurological disease (13), and we
therefore concluded that the different outcomes of HTLV-I
infection are caused mainly by differences in the host response
to the virus rather than the virus itself (14, 15). HAM?TSP
patients mount a very vigorous antibody (16) and CTL (17–19)
response to HTLV-I. This has led to the suggestion (17, 20)
that the anti-HTLV-I immune response, in particular the CTL,
contributes to the tissue damage in the spinal cord that causes
the syndrome of HAM?TSP. However, we have found a
chronically activated CTL response almost entirely directed at
the viral transactivator protein Tax (19) in the majority of both
HCs of the virus and HAM?TSP patients. The CTL response
exerts a significant selection pressure on the Tax protein,
(21). However, the variant sequences do not reach fixation in
the viral population, because the putative escape mutations
impair the function of the Tax protein (21). The selection on
We concluded that the CTL response in HTLV-I infection
might be protective rather than pathogenic (23). According to
this hypothesis, HCs are high CTL responders and HAM?TSP
patients low CTL responders to HTLV-I. These conclusions
were supported by mathematical models of the population
made the nonintuitive prediction that the frequency of anti-
HTLV-I CTL could be greater in the HAM?TSP patients than
in the HCs, even though the CTL are responsible for the lower
proviral load in HCs. Thus, polymorphic genes that control the
efficiency of the anti-HTLV-I CTL response might account for
the different outcomes of HTLV-I infection.
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked ‘‘advertisement’’ in
accordance with 18 U.S.C. §1734 solely to indicate this fact.
PNAS is available online at www.pnas.org.
Abbreviations: HTLV-I, human T-lymphotropic virus-I; HAM/TSP,
HTLV-I associated myelopathy; CTL, cytotoxic T lymphocyte; HCs,
healthy asymptomatic HTLV-I carriers; Fp, preventive fraction; OR,
odds ratio; CI, confidence interval; PBMC, peripheral blood mono-
‡K.J.M.J. and K.U. contributed equally to this work.
??To whom reprint requests should be addressed. e-mail: c.bangham@
MHC Class I and Class II as Candidate Genes in HAM?
TSP. Both the class I MHC proteins (HLA-A, -B, and -C),
which present viral peptides for recognition by virus-specific
CTL (25), and class II MHC proteins (HLA-DR and -DQ),
which present peptides to CD4 T cells, are likely to be
important in the immune response to HTLV-I infection. The
results of several previous studies (26–29) have suggested
associations between various HLA class I or class II alleles and
susceptibility to T cell leukemia?lymphoma or HAM?TSP.
However, small sample size, mixed ethnicity, lack of adequate
controls, or lack of staging in these studies precluded a definite
conclusion. No previous study has indicated HLA-associated
protection against HTLV-I-associated disease. We therefore
set out to test two hypotheses, that (i) certain class I MHC
and (ii) certain class I MHC alleles reduce HTLV-I provirus
The most consistent association between HLA and HAM?
TSP has been found with HLA-DRB1*0101 (26, 29–31), an
apparent susceptibility allele for HAM?TSP. This allele is in
linkage disequilibrium in the Japanese population with HLA-
B*07, Cw*07, and DQB1*0501 (32, 33). We therefore set out
to test a third hypothesis, that (iii) HLA-DRB1*0101 is asso-
ciated with susceptibility to HAM?TSP.
We report here the results of a two-stage case-control
association study of HAM?TSP in the population of Ka-
goshima Prefecture (1988 population: 1.7 million), southern
Kyushu, Japan, where the seroprevalence of HTLV-I infection
in adults is ?10% (34, 35). The estimated prevalence of
HAM?TSP in the HTLV-I positive population is ?1% (36).
We show that the HLA-A*02 gene is associated with both
protection from HAM?TSP and a significant reduction in
provirus load in asymptomatic carriers of HTLV-I. The allele
DRB1*0101 is associated with susceptibility to HAM?TSP, but
only in the absence of the protective effect of HLA-A*02. We
conclude that a strong persistent class I-restricted CTL re-
sponse to HTLV-I benefits the host by reducing the viral load.
The results suggest that an effective vaccine against HTLV-I
Study Population. Two hundred and thirty-two cases of
HAM?TSP were compared with 201 randomly selected
HTLV-I seropositive asymptomatic blood donors (HCs) from
the Kagoshima Red Cross Blood Transfusion Service. All
cases and controls were of Japanese ethnic origin and resided
in Kagoshima Prefecture, Japan. The diagnosis of HAM?TSP
was made according to World Health Organization diagnostic
HLA Class I Typing. A two-stage study was performed. In
stage 1, 96 PCR–sequence-specific primer reactions were
performed to detect all known HLA-A, -B, and -C specificities
in an allele- or group-specific manner (92 possible alleles or
groups of alleles) (38). In stage 2, a single pair of sequence-
0225) was used to test an independent sample of cases and
Subsequently, further class I typing was undertaken with a
reduced number of PCR–sequence-specific primer reactions
(42 possible alleles or groups of alleles), designed to detect all
of the HLA-A, -B, and -C specificities occurring at a gene
frequency of ?5% or an odds ratio (OR) of ?0.5 or ?2.0 from
the initial stage 1 study. After typing 100 cases and controls in
total, the alleles HLA-B*0702, Cw*0702, and Cw*0710 were
selected for further analysis in an independent sample. The
PCR primers used were unable to distinguish between
Cw*0702 and Cw*0710, a rare suballele.
HLA Class II Typing. Class II typing was performed in an
unstaged manner by using the methods of Olerup (39, 40) and
HLA-A*02 and HLA-B*07 Subtyping. The design of se-
quence-specific primers for A*0201–A*0225 alleles and
B*0702–0708 was based on published gene sequences (41)
updated from HLA informatics pages available on the internet
methods were as described (38).
Proviral Load Measurement. The HTLV-I provirus load in
peripheral blood mononuclear cells (PBMC) was measured in
all patients and HCs as described (9). A quantitative PCR
reaction was performed by using an ABI 7700 sequence
detector (Perkin–Elmer Applied Biosystems). All DNA stan-
dards and samples were amplified in triplicate. A standard
curve was generated by using the ?-actin gene from HTLV-
I-negative PBMC and the Tax gene from TARL-2, a cell line
containing a single copy of HTLV-I proviral DNA. The
amount of HTLV-I proviral DNA was calculated as follows:
copy number of HTLV-I (tax) per 104PBMC ? [copy number
of tax?(copy number of ?-actin?2)] ? 104. The lower limit of
detection was 1 copy per 104PBMC.
Statistical Analysis. The ?2test, the Mann–Whitney U test,
and the odds ratio (GraphPad, San Diego) were used for
statistical analysis. The Bonferroni method (42) was used to
correct for multiple comparisons. The population attributable
risk was calculated according to Schlesselman (43). To identify
the significant independent variables associated with disease
risk, we carried out a standard logistic regression analysis (44).
To calculate the prevented fraction (Fp) of disease, consider
the 2 ? 2 contingency table
where D ? disease, H ? health, G?? positive for protective
genotype, G?? negative for protective genotype. By Bayes’
theorem of conditional probabilities, the fraction (Fp) of
potential cases of disease D in the population that is prevented
by the genotype G?is given by Fp? (1 ? R) ? [1 ? (dr1?br2)],
where R ? prevalence rate of disease D in the population, r1
? a ? b and r2 ? c ? d. In the case of HAM?TSP, R is
estimated as ?1% of the HTLV-I-infected population. Fpis
approximately normally distributed: the standard deviation is
Table 1. HLA-A*02 reduces the odds of HAM?TSP
HAM?TSP, No.HCs, No.
P Odds ratio‡
A*02?and A*02?denote the presence or absence of the A*02 gene in the subjects studied. In total, 232 HAM?TSP patients and 201 HCs were
studied. Stages denote independent, consecutive case-control studies, and do not refer to clinical stage.
†With Yates correction.
‡Using the approximation of Woolf.
Immunology: Jeffery et al.Proc. Natl. Acad. Sci. USA 96 (1999)3849
given by SD (Fp) ? (1 ? R ? Fp) ? ? [(c?dr2) ? (a?br1)]. A
full derivation of these formulae is available on request.
Phenotypic Analysis of Tax11–19-Specific CTL. In the chron-
ically activated CTL response to HTLV-I, several peptides
restricted by HLA-A*02 (45, 46). Tax11–19 is a dominant
A*02-restricted epitope (47). To compare the frequency of
HTLV-I-specific CTL in HAM?TSP patients and HCs, we
used the recently developed technique of soluble peptide-
MHC tetramers (48). Analysis of PBMC for the presence of
Tax11–19-specific CD8?CTL was performed by using fluores-
cent-labeled tetramers of HLA-A*0201 ? ?2microglobulin ?
Tax11–19in 19 HLA-A*02?HAM?TSP patients and 19 HLA-
A*02?HTLV-I HCs from the Kagoshima study population.
The cells were incubated with Tax11–19tetramer at 37°C for 20
minutes and anti-CD8 antibody (Caltag, South San Francisco,
CA) on ice for 30 minutes then washed three times in ice-cold
phosphate-buffered saline before fixation in 1% freshly made
paraformaldehyde for 30 minutes at 4°C. The antigen-specific
T cells were quantified by flow cytometry on a Coulter Epics
XL (Beckman Coulter).
The median age of HAM?TSP patients (59 years) was greater
than that of the HCs (41 years). The sex ratio of males?females
in the HAM?TSP group was 1:2, with a 1:1 ratio in the HCs.
However, there was no correlation between the HTLV-I
proviral load and age at blood sampling (HAM?TSP patients:
r ? ?0.096, P ? 0.22; HCs: r ? 0.081, P ? 0.25, Spearman’s
rank correlation) or duration of disease in the Kagoshima
population (see ref. 9 for details). Because the prevalence of
HAM?TSP in Kagoshima is low (?1%) among HTLV-I
seropositives, very few HCs would be expected to develop
HAM?TSP. ?2and logistic regression analyses confirmed that
the frequency of HLA-A*02, B*0702, and DRB1*0101 and the
proviral load were unaffected by age or sex (44) (data not
In the first 50 cases and 56 controls, the genotype frequency
of HLA-A*02 was significantly lower (24%) among the cases of
HAM?TSP compared with the controls (57%, P ? 0.001,
uncorrected) (Table 1). Because of the large number of alleles
tested for, it was necessary to confirm this association in an
independent sample. We therefore retested the association
with HLA-A*02 in an independent sample by using a single
pair of sequence-specific primers to detect all known subtypes
of A*02. The frequency of HLA-A*02 was again lower in the
patients (33%) than the controls (47%, P ? 0.006) (Table 1),
confirming the association. The two data sets, when combined,
indicate that the possession of HLA-A*02 is associated with a
reduction in the odds of disease by ?2-fold (P ? 0.0001) and
prevents (Fp) ?28% (?5.8% SD) of potential cases of HAM?
TSP in the study population.
It is likely that HLA-A*02 is associated with protection
against HAM?TSP because HLA-A*02-restricted, anti-
HTLV-I CTL kill HTLV-I-infected cells and so reduce the
provirus load of HTLV-I. We therefore tested the hypothesis
that HTLV-I proviral load is lower in A*02?subjects than
A*02?subjects. In the whole sample, there was a 16-fold
greater median provirus load in the HAM?TSP patients (n ?
222) than in the HCs (n ? 201) (575 copies per 104PBMC,
compared with 35 copies per 104PBMC, P ? 0.0001, Mann–
Whitney U statistic, two-tailed), in agreement with other
studies (7). There was a 3-fold lower median provirus load in
the A*02?HCs compared with the A*02?HCs (Table 2, P ?
An A*02 subtyping method was developed to identify the
A*02 subtypes present in the population and to examine
whether a particular A*02 subtype was associated with disease
protection or a reduction in proviral load. A*02 subtypes
A*0201, 0203, 0206, 0207, and 0210 were detected. The
genotype frequencies (as % of A*02?subjects) were as fol-
lows—A*0201: HAM?TSP 36%, HCs 36%; A*0203: HAM?
TSP 3%, HCs 0%; A*0206: HAM?TSP 43%, HCs 53%;
A*0207: HAM?TSP 20%, HCs 22%; A*0210: HAM?TSP 3%,
HCs 2%. Four HAM?TSP patients and 11 HCs were het-
erozygous for A*02 subtypes. The possession of A*0206 was
significantly associated with disease protection (A*0206?; 30
of 232 HAM?TSP patients, 52 of 201 HCs, P ? 0.001 (two-
tailed), OR ? 0.43, 95% CI (confidence interval) ? 0.26–
0.70). This association remained significant after correction
A*0206 was associated with a 4-fold reduction in median
provirus load in the HCs (P ? 0.004, Table 2), (Fp?15% ?
4.1% SD). The possession of HLA-A*02 subtypes other than
A*0206 (i.e., A*0201, 0203, 0207, and 0210) also was signif-
Table 2.HLA-A*02 and subtype A*0206 proviral loads in HCs
Median proviral load
No. of subjects
PMedian proviral load
No. of subjects
HLA-A*02 and subtype A*0206 were associated with a reduced provirus load in HCs. Other A*02 subtypes were associated with a reduced
provirus load in HAM?TSP patients.
Median proviral load given as proviral copy number per 104PBMC. P reported as two-tailed, uncorrected Mann–Whitney U test.
A*02 positive (A*02?) subjects
HLA-DRB1*0101 increases the odds of HAM?TSP in A*02 negative (A*02?) but not in
?2reported as one-tailed with Yates correction. Odds ratio used the approximation of Woolf.
HAM?TSP, No.HCs, No.
3850 Immunology: Jeffery et al.Proc. Natl. Acad. Sci. USA 96 (1999)
icantly associated with protection against HAM?TSP (A*02?
A*0206?; 39 of 202 HAM?TSP patients, 48 of 149 HCs, P ?
0.008 (two-tailed), OR ? 0.50, 95% CI ? 0.31–0.82). HLA-
A*02 subtypes other than A*0206 also were associated with a
reduction in median provirus load in the HAM?TSP patients
(P ? 0.026, Table 2) but not in HCs (P ? 0.40, Table 2); the
The risk (prevalence) of HAM?TSP at a given proviral load
can be calculated from the data on proviral load in the
HAM?TSP and HC cohorts (Fig. 1). Fig. 1 shows that the risk
of HAM?TSP rises exponentially as the provirus load in-
creases above an apparent threshold of 1% PBMC (log10
proviral copy number per 104PBMC ? 2). However, the risk
of HAM?TSP at any given provirus load was not affected by
the presence or absence of A*02. This result suggests that the
mechanism by which A*02 reduces the risk of HAM?TSP is by
reducing the provirus load of HTLV-I.
These data on the protective effect of A*02 in HAM?TSP
are supported by data from an independent population of
HAM?TSP patients and HTLV-I-infected asymptomatic car-
riers from London. In this population, 4 of 15 HAM?TSP
patients were HLA-A*02?, compared with 10 of 14 HCs [P
subjects were of Caribbean origin and 2 were of Caucasian
microglobulin ? Tax11–19, tetramer-positive CTL were found
in both HAM?TSP patients and HCs: Fig. 2 indicates that
CD8?cells from A*0201?, A*0206?and A*0207?individuals
bound the A*0201 Tax11–19tetramer. The mean frequency of
tetramer-positive cells in the CD8?population in HAM?TSP
patients (1.9% ? 0.5% SE, n ? 19) was not significantly
different from the mean frequency in HC (0.9% ? 0.2% SE,
n ? 19; P ? 0.58, Mann–Whitney). Mean tetramer binding in
HLA-A*02-negative controls of mixed ethnic origin was 0.1%
? 0.1% SE (4 HAM?TSP patients, 1 HC).
HLA-DRB1*0101 was associated with susceptibility to
HAM?TSP [Table 3, P ? 0.049 (one-tailed, because of the
previously observed association)]. Possession of this allele also
was associated with a significantly lower virus load in the
HAM?TSP patients [Table 4; P ? 0.024 (two-tailed)], but not
in HCs. The DRB1*0101-associated susceptibility to disease
and reduced provirus load was not seen in HLA-A*02?
HAM?TSP patients, but was significant in HLA-A*02?
patients (Tables 3 and 4). The population attributable risk
conferred by DRB1*0101 (i.e., the excess fraction of cases of
HAM?TSP in the sample that would not have occurred had
DRB1*0101 been absent) was 7%. This increased to 11% in
A*02?–DRB1*0101?subjects, and fell to 0.4% in A*02?–
The association of B*0702 with HAM?TSP was examined in
a staged study. In stage 1, 21 of 100 patients and 8 of 100
controls were found to have B*0702 [P ? 0.016 (two-tailed),
OR ? 3.1, 95% CI ? 1.28–7.28; Pcorrectedwas not significant].
This association was not replicated, however, in an indepen-
dent sample [20 of 130 patients, 15 of 101 controls, P ? 0.46
(one-tailed)], and was only significant in the whole population
by using a 1-tailed test of significance (P ? 0.046, OR ? 1.66,
95% CI ? 0.96–2.88). Among B*0702?subjects, 49 of 57
(86%) also were positive for DRB1*0101, and 49 of 54 (91%)
DRB1*0101?subjects also were positive for B*0702. All
B*0702?subjects possessed Cw*07, and all DRB1*0101?
subjects possessed DQB1*0501. Cw*07 and DQB1*0501 were
not more significantly associated with HAM?TSP than
DRB1*0101 (data not shown).
In the 100 patients and 100 controls examined for all class
I specificities, only HLA-A*02, -B*0702, and -Cw*0702?10
differed significantly in frequency between patients and con-
trols at the P ? 0.05 level (uncorrected). We are therefore
unable to confirm the suggestion made by Nishimura et al. (29)
that HLA-A*31 is associated with HAM?TSP.
This study demonstrates that the risk of HAM?TSP is strongly
associated with the equilibrium provirus load of HTLV-I, and
load. The data reported here and the previous findings in
p(HAM ? L) if A*02?, I, log p(HAM ? L) if A*02?, nd, not detectable
by this assay, —, no A*02?HAM?TSP patient had a logarithmic
proviral load of 0.5–0.9. At a given provirus load, the risk of HAM?
TSP is not affected by the presence or absence of A*02. However, as
the possession of A*02 reduces the proviral load, the effect of A*02 is
to reduce the risk of disease. For the calculation of HAM?TSP risk at
a given proviral load, we used Bayes’ theorem. By using the standard
notation for conditional probability, where p(HAM ? L) denotes the
probability of HAM?TSP in an HTLV-I-infected person with a given
provirus load (L), we write: p(HAM ? L) ? [p(HAM) ? p(L ?
HAM]?[p(HAM) ? p(L ? HAM) ? p(HC) ? p(L ? HC)]. We
estimated p(L ? HAM] and p(L ? HC) from the distribution of proviral
load in the HAM?TSP and HC cohorts in the present study. p(HAM),
the prevalence of HAM?TSP in the HTLV-I-positive population, is
taken as 0.01.
Risk of HAM?TSP according to possession of A*02. Œ, log
patients and HCs. The mean frequency of HTLV-I-specific CTL does
not differ between HAM?TSP patients and HCs (P ? 0.58; two-tailed
Mann–Whitney U-statistic). CD8?cells from both A*0201?, A*0206?,
and A*0207?individuals bound the A*0201-Tax11–19 tetramer. I,
A*0206; Œ, A*0207; E, A*0201; F, A*0210, ? A*0201?A*0206; I,
A*0201?A*0207; ?, mean % tetramer ? SE.
Tetramer-positive CTL were found in both HAM?TSP
Immunology: Jeffery et al.Proc. Natl. Acad. Sci. USA 96 (1999) 3851
healthy HTLV-I carriers of strong anti-HTLV-I CTL re-
sponses, low proviral load, and viral escape mutants, can be
interpreted as follows. HCs mount a strong CTL response to
HTLV-I and so limit the proviral load to a low level. That is,
their virus-specific CTL proliferate rapidly in response to
HTLV-I antigens and?or kill HTLV-I-infected cells rapidly
(24). HAM?TSP patients, on the other hand, make a weak
CTL response to HTLV-I, and the virus is allowed to reach a
high equilibrium provirus load. Thus, in HCs, a high frequency
of CTL is maintained by a relatively low virus load, whereas in
HAM?TSP patients, a high virus load stimulates an inefficient
difference in the mean frequency of HTLV-I-specific CTL
between the HAM?TSP patients and the HCs, as observed in
this study. These results are consistent with our previous
finding (46) that the mean frequency of the Tax peptide-
specific CTL, measured by limiting-dilution assays, did not
differ significantly between HAM?TSP patients and HCs.
This study demonstrates that each of the major HLA-A*02
subtypes present in the Kagoshima population is able to
present an immunodominant peptide from HTLV-I Tax
(Tax11–19) to CD8?T cells, and confer protection from HAM?
TSP. There may be differences between the effects of the
load reduction in the HAM?TSP and HC cohorts (Table 2).
However, HLA-A*02?CTL responders to Tax frequently
recognize more than one A*02-restricted epitope in Tax (49).
Also, A*02 subtypes differ significantly in their peptide-
bind Tax peptides.
The most probable mechanism for the pathogenesis of
HAM?TSP is bystander damage to uninfected cells caused by
the activated T cells found in HTLV-I infection (14, 15, 23). It
is likely that CD4?cells play an important part in bystander
damage in the central nervous system, because these are the
predominant cells early in the active lesions of HAM?TSP
(52). Moreover, it is now clear that HTLV-I in the inflamma-
tory lesions is present only in the invading CD4?cells (53, 54).
In this case, class II genes, which determine the antigen
specificity of CD4?cells, could be associated with suscepti-
bility to HAM?TSP. It is therefore possible that one of the
class II alleles in the susceptibility haplotype (HLA-B*0702-
Cw*0702-DRB1*0101-DQB1*0501) is responsible for the sus-
ceptibility effect, via an effect on CD4 T cell activation and
increased bystander damage. Possession of this haplotype is
associated with a lower provirus load in HAM?TSP patients.
The explanation for this is that DRB1*0101?individuals only
develop HAM?TSP if (on average) they have a high proviral
load. DRB1*0101?individuals are, however, more susceptible
and can therefore develop HAM?TSP even if they have a low
proviral load. Therefore, the average proviral load of
DRB1*0101?patients will be lower than that of DRB1*0101?
A gene may be associated with a disease because the gene
causes the disease, because it is in linkage disequilibrium with
the causative gene, or because of population admixture (ge-
netic stratification). The protection from HAM?TSP observed
here is likely to be caused by A*02 itself, not a linked gene,
because (i) there is no evidence of linkage disequilibrium
between HLA-A*02 and any other class I or class II alleles in
this population, apart from the known linkage disequilibrium
between A*0207 and B*4601 (B*4601 was not independently
associated with disease protection; data not shown), (ii) there
is a vigorous anti-A*02-restricted CTL response to HTLV-I,
which is a plausible mechanism of protection, and (iii) the
A*02-associated protective effect has been replicated in a
small, HTLV-I-infected population in London (see Results).
There was no evidence of population admixture in this study.
It is not possible to determine which allele on the DRB1*0101-
associated haplotype is responsible for the susceptibility effect
because of the known strong linkage disequilibrium of
DRB1*0101 with other alleles in the MHC region. However, it
may be possible to identify a single susceptibility allele by
carrying out host genetic studies in other populations in which
different alleles are linked on the susceptibility haplotype (55).
The effect of DRB1*0101 was not examined in a staged fashion
because of its previously suggested association with disease
susceptibility. The linked gene B*0702 was significantly asso-
ciated with disease in the first stage (uncorrected), but not in
an independent population examined for B*0702 alone in the
second stage of the study, perhaps because the sample size was
too small (56).
The effects of HLA class I and class II alleles (or genes in
linkage disequilibrium with HLA) found in this study do not
account for all of the observed difference in individual sus-
ceptibility to HAM?TSP. Previous studies have found an
association between HAM?TSP and female sex (57) and early
age at initial sexual activity (58). We believe that the protective
effect of A*02 is evident because A*02 occurs at a high
frequency in the population studied and because there is a
highly dominant CTL target antigen which contains a domi-
nant A*02-restricted epitope. It is likely that class I alleles are
protective in other infectious diseases but do not occur at a
sufficient frequency in the population sizes studied to reach
statistical significance. Other polymorphic genes that may
contribute to the different outcomes of HTLV-I infection
include those that affect the efficiency of the anti-HTLV-I
immune response (e.g., TAP-1 and -2), or the rate of prolif-
eration or migration of leukocytes (e.g., cytokines and their
receptors, or adhesion molecules).
In conclusion, the present results indicate that MHC class
I-restricted CTL reduce the proviral load of HTLV-I and
consequently the risk of HAM?TSP. DRB1*0101 predisposes
to HAM?TSP in the absence of A*02. It is still uncertain, in
both HTLV-I and HIV-1, whether antiviral class I HLA-
restricted CTL benefit the host by reducing virus load or
contribute to disease by damaging host tissues. Our results
Table 4.HTLV-I provirus load associated with HLA-DRB1*0101 in the presence or absence of HLA-A*02
Median proviral load
No. of subjects
P Median proviral load
No. of subjects
A significant reduction in provirus load is observed in the HAM?TSP patients, but not in the A*02?patients or in the HCs.
Proviral copy reported as number per 104PBMC. P level reported using two-tailed Mann–Whitney U test.
3852 Immunology: Jeffery et al.Proc. Natl. Acad. Sci. USA 96 (1999)
strongly favor a protective role and therefore argue that an
effective antiretroviral vaccine should elicit a vigorous CTL
We thank the staff and blood donors of the Kagoshima Red Cross
Blood Center and the staff and patients of the Third Department of
Internal Medicine, Kagoshima University, Kagoshima, Japan, for
valuable samples. This work was supported by the Program for
Promotion of Fundamental Studies in Health Sciences of the Orga-
nization for Pharmaceutical Safety and Research (OPSR) (Japan), the
Wellcome Trust (C.R.M.B., S.E.H., K.J.M.J., and M.A.N.) and the
Medical Research Council (U.K.) (A.L.L. and G.S.O.).
Hill, A. V. S. (1996) Nat. Med. 2, 395–396.
Kaslow, R. A., Carrington, M., Apple, R., Park, L., Munoz, A.,
Saah, A. J., Goedert, J. J., Winkler, C., O’Brien, S. J., Rinaldo,
C., et al. (1996) Nat. Med. 2, 405–411.
McNeil, A. J., Yap, P. L., Gore, S. M., Brettle, R. P., McColl, M.,
Wyld, R., Davidson, S., Weightman, R., Richardson, A. M. &
Robertson, J. R. (1996) Q. J. Med. 89, 177–185.
Klein, M. R., van Baalen, C. A., Holwerda, A. M., Kerkhof
Garde, S. R., Bende, R. J., Keet, I. P., Eeftinck-Schattenkerk,
J. K., Osterhaus, A. D., Schuitemaker, H. & Miedema, F. (1995)
J. Exp. Med. 181, 1365–1372.
Hill, A. V. S. (1998) Annu. Rev. Immunol. 16, 593–617.
Kira, J., Koyanagi, Y., Yamada, T., Itoyama, Y., Goto, I.,
Yamamoto, N., Sasaki, H. & Sakaki, Y. (1991) Ann. Neurol. 29,
Gessain, A., Saal, F., Gout, O., Daniel, M. T., Flandrin, G., de
The, G., Peries, J. & Sigaux, F. (1990) Blood 75, 428–433.
Nagai, M., Usuku, K., Matsumoto, W., Kodama, D., Takenouchi,
N., Moritoyo, T., Hashiguchi, S., Ichinose, M., Bangham,
C. R. M., Izumo, S., et al. (1998) J. Neurovirol., in press.
Yoshida, M., Osame, M., Kawai, H., Toita, M., Kuwasaki, N.,
Nishida, Y., Hiraki, Y., Takahashi, K., Nomura, K., Sonoda, S.,
et al. (1989) Ann. Neurol. 26, 331–335.
Iwasaki, Y. (1993) Brain Pathology 3, 1–10.
Bangham, C. R. M. (1993) Curr. Opin. Neurobiol. 1995, 50–61.
Daenke, S., Nightingale, S., Cruickshank, J. K. & Bangham, C. R.
(1990) J. Virol. 64, 1278–1282.
Daenke, S. & Bangham, C. (1994) Clin. Exp. Immunology 96,
Ijichi, S., Izumo, S., Eiraku, K., Machigashira, R., Kubota, M.,
Nagai, M., Ikegami, N., Kashio, N., Umehara, I., Maruyama, I. &
et al. (1993) Med. Hypotheses 41, 542–547.
Gessain, A. & Gout, O. (1992) Ann. Intern. Med. 117, 933–946.
Jacobson, S., Shida, H., McFarlin, D. E., Fauci, A. S. & Koenig,
S. (1990) Nature (London) 348, 245–248.
Kannagi, M., Harada, S., Maruyama, I., Inoko, H., Igarashi, H.,
Kuwashima, G., Sato, S., Morita, M., Kidokoro, M., Sugimoto,
M., et al. (1991) Int. Immunol. 3, 761–767.
Parker, C., Daenke, S., Nightingale, S. & Bangham, C. (1992)
Virology 188, 628–636.
Kubota, R., Kawanishi, T., Matsubara, H., Manns, A. & Jacob-
son, S. (1998) J. Immunol. 161, 482–488.
Niewiesk, S., Daenke, S., Parker, C. E., Taylor, G., Weber, J.,
Nightingale, S. & Bangham, C. R. (1995) J. Virol. 69, 2649–2653.
Niewiesk, S., Daenke, S., Parker, C. E., Taylor, G., Weber, J.,
Nightingale, S. & Bangham, C. R. (1994) J. Virol. 68, 6778–6781.
Bangham, C., Kermode, A., Hall, S. & Daenke, S. (1996) Semin.
Virol. 7, 41–48.
Nowak, M. A. & Bangham, C. R. (1996) Science 272, 74–79.
Townsend, A. & Bodmer, H. (1989) Annu. Rev. Immunol. 7,
Sonoda, S., Fujiyoshi, T. & Yashiki, S. (1996) J. Acquir. Immune
Defic. Syndr. Hum. Retrovirol. 13, S119–S23.
Manns, A., Hanchard, B., Morgan, O. S., Wilks, R., Cranston, B.,
Nam, J. M., Blank, M., Kuwayama, M., Yashiki, S., Fujiyoshi, T.,
et al. (1998) J. Natl. Cancer Inst. 90, 617–622.
Usuku, K., Sonoda, S., Osame, M., Yashiki, S., Takahashi, K.,
Ann. Neurol. 23 Suppl., 143–150.
Nishimura, Y., Okubo, R., Minato, S., Itoyama, Y., Goto, I.,
Mori, M., Hirayama, K. & Sasazuki, T. (1991) Tissue Antigens 37,
30.Usuku, K., Nishizawa, M., Matsuki, K., Tokunaga, K., Takahashi,
K., Eiraku, N., Suehara, M., Juji, T., Osame, M. & Tabira, T.
(1990) Eur. J. Immunol. 20, 1603–1606.
Kitze, B., Usuku, K., Yamano, Y., Yashiki, S., Nakamura, M.,
Fujiyoshi, T., Izumo, S., Osame, M. & Sonoda, S. (1998) Clin.
Exp. Immunol. 111, 278–285.
Ando, H., Mizuki, N., Ando, R., Miyata, Y., Miyata, S.,
Wakisaka, K. & Inoko, H. (1996) Tissue Antigens 48, 55–58.
Hashimoto, M., Kinoshita, T., Yamasaki, M., Tanaka, H., Iman-
ishi, T., Ihara, H., Ichikawa, Y. & Fukunishi, T. (1994) Tissue
Antigens 44, 166–173.
Osame, M., Usuku, K., Izumo, S., Ijichi, N., Amitani, H. & Igata,
A. (1986) Lancet 1, 1031–1032.
Nakagawa, M., Izumo, S., Ijichi, S., Kubota, H., Arimura, K.,
Kawabata, M. & Osame, M. (1995) J. Neurovirol. 1, 50–61.
Kaplan, J. E., Osame, M., Kubota, H., Igata, A., Nishitani, H.,
Maeda, Y., Khabbaz, R. F. & Janssen, R. S. (1990) J. Acquir.
Immune Defic. Syndr. 3, 1096–1101.
Osame, M. (1990) in Human Retrovirology: HTLV, ed. Blattner,
W. A. (Raven, New York), pp. 191–197.
Bunce, M., O’Neill, C. M., Barnardo, M. C. N. M., Krausa, P.,
Browning, M. J., Morris, P. J. & Welsh, K. I. (1995) Tissue
Antigens 46, 355–367.
Olerup, O. & Zetterquist, H. (1992) Tissue Antigens 39, 225–235.
Olerup, O., Aldener, A. & Fogdell, A. (1993) Tissue Antigens 41,
Bodmer, J. G., Marsh, S. G., Albert, E. D., Bodmer, W. F.,
Bontrop, R. E., Charron, D., Dupont, B., Erlich, H. A., Fauchet,
R., Mach, B., et al. (1997) Tissue Antigens 49, 297–321.
Bland, J. M. & Altman, D. G. (1995) Br. Med. J. 310, 170.
Schlesselman, J. J. (1982) Case-Control Studies (Oxford Univ.
Hosmer, D. W. & Lemeshow, S. (1989) Applied Logistic Regres-
sion (Wiley, New York).
Koenig, S., Woods, R. M., Brewah, Y. A., Newell, A. J., Jones,
G. M., Boone, E., Adelsberger, J. W., Baseler, M. W., Robinson,
S. M. & Jacobson, S. (1993) J. Immunol. 151, 3874–3883.
Daenke, S., Kermode, A. G., Hall, S. E., Taylor, G., Weber, J.,
Nightingale, S. & Bangham, C. R. (1996) Virology 217, 139–146.
Kannagi, M., Shida, H., Igarashi, H., Kuruma, K., Murai, H.,
Aono, Y., Maruyama, I., Osame, M., Hattori, T., Inoko, H., et al.
(1992) J. Virol. 66, 2928–2933.
Altman, J. D., Moss, P. A. H., Goulder, P. J. R., Barouch, D. H.,
McHeyzer-Williams, M. G., Bell, J. I., McMichael, A. J. & Davis,
M. M. (1996) Science 274, 94–96.
Parker, C. E., Nightingale, S., Taylor, G. P., Weber, J. &
Bangham, C. R. (1994) J. Virol. 68, 2860–2868.
Sudo, T., Kamikawaji, N., Kimura, A., Date, Y., Savoie, C. J.,
Nakashima, H., Furuichi, E., Kuhara, S. & Sasazuki, T. (1995)
J. Immunol. 155, 4749–4756.
Barouch, D., Friede, T., Stevanovic, S., Tussey, L., Smith, K.,
Rowland-Jones, S., Braud, V., McMichael, A. & Rammensee,
H. G. (1995) J. Exp. Med. 182, 1847–1856.
Iwasaki, Y., Ohara, Y., Kobayashi, I. & Akizuki, S. (1992) Am. J.
Pathol. 140, 1003–1008.
Moritoyo, T., Reinhart, T. A., Moritoyo, H., Sato, E., Izumo, S.,
Osame, M. & Haase, A. T. (1996) Ann. Neurol. 40, 84–90.
Matsuoka, E., Takenouchi, N., Hashimoto, K., Kashio, N., Mori-
toyo, T., Higuchi, I., Isashiki, Y., Sato, E., Osame, M. & Izumo,
S. (1998) Acta Neuropathol. 96, 340–346.
Todd, J. A. (1999) in The Metabolic and Molecular Bases of
Inherited Disease, eds. Scriver, C. R., Beaudet, A. L., Sly, W. S.,
Valle, D., Vogelstein, B. & Childs, B. (McGraw–Hill, New York),
Suarez, B. K., Hampe, C. L. & van Eerdewegh, P. (1994) in
Genetic Approaches to Mental Disorders, eds. Gershon, E. S. &
Coninger, C. R. (Am. Psychopathol. Assoc., Washington, DC),
Mueller, N. E., and Blattner, W. A. (1996) in Viral Infections of
Humans: Epidemiology and Control, eds. Evans, A. S. & Kaslow,
R. (Plenum, New York).
Kramer, A., Maloney, E. M., Morgan, O. S., Rodgers-Johnson,
P., Manns, A., Murphy, E. L., Larsen, S., Cranston, B., Murphy,
J., Benichou, J., et al. (1995) Am. J. Epidemiol. 142, 1212–1220.
Immunology: Jeffery et al.Proc. Natl. Acad. Sci. USA 96 (1999) 3853