Evidence of HIV-1 Adaptation to HLA-Restricted Immune Responses at a Population Level
Corey B. Moore,1
Mina John,1, 2
Ian R. James,1
Frank T. Christiansen,2, 3
Campbell S. Witt,1, 2
Simon A. Mallal1, 2*
Antigen-specific T cell immunity is HLA-restricted.
Human immunodeficiency virus-type 1 (HIV-1) mutations that allow
escape from host immune responses may therefore be HLA
allele-specific. We analyzed HIV-1 reverse transcriptase sequences
from a large HLA-diverse population of HIV-1-infected individuals.
Polymorphisms in HIV-1 were most evident at sites of least
functional or structural constraint and frequently were associated with
particular host HLA class I alleles. Absence of polymorphism was also
HLA allele-specific. At a population level, the degree of
HLA-associated selection in viral sequence was predictive of viral
load. These results support a fundamental role for HLA-restricted
immune responses in driving and shaping HIV-1 evolution in vivo.
1 Centre for Clinical Immunology and Biomedical
Statistics, Royal Perth Hospital and Murdoch University, Level 2 North
Block, Royal Perth Hospital, Wellington Street, WA 6000, Australia.
2 Department of Clinical Immunology and Biochemical
Genetics, Royal Perth Hospital.
3 Department of
Pathology, University of Western Australia, WA 6009, Australia.
To whom correspondence should be addressed. E-mail:
Selection of viral mutations
associated with loss of antiviral cytotoxic T lymphocyte (CTL)
responses has been described in humans with acute and chronic HIV-1
infection (1), macaques infected with simian
immunodeficiency virus (SIV) (2, 3), and rhesus
monkeys challenged with simian-human immunodeficiency virus (SHIV)
after vaccination (4). However, the full extent and
importance of CTL escape mutation to HIV-1 evolution remains to be
established. CTL escape mutations occur at critical sites within
HLA-restricted CTL epitopes where an amino acid substitution may abrogate epitope-HLA binding, reduce T cell receptor recognition, or generate antagonistic CTL responses (1). Mutations that
affect proteosome cleavage sites flanking CTL epitopes may also disrupt
cellular processing of the epitope (5). The capacity of the
virus to mutate at any amino acid residue is constrained, however, by
the functional or structural value of the residue to virus survival
We hypothesized that, as CTL epitopes are HLA-restricted, CTL escape
mutations selected within an individual host would be characteristic
for specific HLA class I alleles across an HLA-diverse host population.
We speculated that such polymorphisms would be particularly evident in
viral genes encoding internal proteins such as HIV-1 reverse
transcriptase (RT), which is highly expressed in virions
(7) and immunogenic in the early response to HIV-1
(8, 9). Therefore, we examined the relationship
between HIV-1 RT sequence polymorphisms, known functional constraint,
and HLA genotypes in 473 participants of the Western Australian
(WA) HIV Cohort Study (10-13).
We first determined the population consensus sequence for HIV-1 RT by
assigning the most common amino acid for each position between
positions 20 and 227 of all first pretreatment sequences pooled from
the cohort. The consensus sequence matched the clade B reference
sequence HIV-1 HXB2 at all positions in HIV-1 RT except residues 122 (lysine instead of glutamate) and 214 (phenylalanine instead of
leucine) (14). The percentages of individuals with
an amino acid in their own first pretreatment HIV-1 RT sequence different from that of consensus sequence were calculated for each
amino acid residue. This rate of polymorphism at single residues varied
from 0% to 60% and correlated with the level of known functional constraint at each site (6, 15).
We then analyzed the most recent HIV-1 RT sequences obtained from all
individuals. We considered single amino acid residues one at a time and
the polymorphism of each HIV-1 residue across the host population. We
conducted a multivariate analysis with logistic regression for each
residue to assess the statistical significance of association(s)
between the presence of a polymorphism (defined as any substitution of
the consensus amino acid) and the HLA-A and HLA-B alleles of the
population. A P value and estimated odds ratio (OR) for each
HLA allele-polymorphism association were generated. We incorporated
initial power calculations to limit analyses to only those HLA alleles
and viral polymorphisms that were sufficiently prevalent to have
associations between them detected. We applied a further selection
procedure based on forward selection and backward elimination. We used
randomization tests to determine the "exact" significance levels
for associations and designed a customized software program, Epipop, to
carry out these analyses (15, 16). This
process was repeated for every residue from position 20 to 227 of HIV-1
RT, giving a residue-specific view of the independent selection effects of HLA on HIV-1 RT in vivo and at a population level.
We plotted all the statistically significant HLA-polymorphism
associations on a map of HIV-1 RT in relation to polymorphism rate,
previously reported functional characteristics of residues (6), and published CTL epitopes
(17) (Fig. 1). There were 64 significant
positive associations between polymorphisms and HLA alleles (OR > 1, P < 0.05 in all cases) (Fig. 1B, fig 51B).
Polymorphisms associated with the same HLA allele appeared to cluster
along the sequence. For example, HLA-B7 was associated with
polymorphism at positions 158, 162, 165, and 169, which are all within
or flanking the known HLA-B7-restricted CTL epitope RT(156-165).
There was also apparent clustering of associations for HLA-B12,
HLA-B35, HLA-B18, and HLA-B15. Fifteen HLA-specific polymorphisms were
at positions within known CTL epitopes, and the HLA allele association
matched the known HLA restriction of the epitope. For example, the
amino acid at position 162 resides within a known HLA-B7-restricted
CTL epitope RT(156-165) and the odds of polymorphism at this site were
significantly increased in those individuals with HLA-B7 (OR = 10, P < 0.001). Four polymorphisms (at positions 101, 135, 165, and 166) were at primary anchor positions within corresponding CTL
epitopes (HLA-A3-, HLA-B51/HLA-B*5101-, HLA-B7-, and
HLA-A11-restricted, respectively) where mutation could abrogate
binding to the HLA molecule. The remaining 11 associations were at nonprimary anchor positions of CTL epitopes. The number of
HLA-specific polymorphisms observed within known CTL epitopes with
corresponding HLA restriction was significantly greater than that
expected if significant positive associations occurred randomly across
residues (15 versus 4.27, P < 0.001). Furthermore, an
excess of associations over that expected was observed for 10 of the 11 HLA specificities with CTL epitopes in this segment of HIV-1 RT
(15). Six HLA allele-specific polymorphisms were not within
but flanked CTL epitopes, including the predicted proteosome cleavage
sites at positions 26 and 28 flanking HLA-A2- and HLA-A3-restricted CTL epitopes (18) [see fig. S1 (15)].
Map of polymorphism rate and HLA
associations at amino acid positions 94 to 215 of HIV-1 RT. Residues
with little power to detect HLA associations are shaded. (A)
Published CTL epitopes are shown as black lines with their known HLA
restriction to the level of broad serological genotype, except the
HLA-B*5101- and HLA-B*3501-restricted epitopes examined in further
detail in the study. HLA restrictions of epitopes are marked in red if
there is a corresponding HLA-specific polymorphism within the epitope.
(B) HLA alleles significantly associated with specific
polymorphisms and ORs for the association. HLA-specific polymorphisms
within published CTL epitopes restricted to the same HLA allele are
red. All other associations are black and may indicate the location of
putative CTL epitopes. Boxed associations are those that remain
statistically significant after correction for total number of residues
examined across the entire gene region. (C) Negative HLA
associations with ORs of residue not varying from consensus.
(D) Percentages of individuals with a viral amino acid
different from that of consensus sequence at each position. Known
functional characteristics of residues (6) are marked as
stability (S), functional (F), catalytic (C), and external
[View Larger Version of this Image (32K GIF file)]
To take account of the multiple comparisons used in the statistical
process over the entire HIV-1 RT protein, we applied Bonferroni-type corrections based on randomization tests. Following this highly stringent correction, 12 associations remained statistically
significant (P < 0.05). Overall, the finding of HLA
associations, taking all positions and multiple comparisons into
account, was statistically significant (P < 0.001). We
propose that those HLA-associated polymorphisms that were not within
published corresponding CTL epitopes may indicate where previously
unknown, untested epitopes are located. This is particularly likely for
those HLA associations that are strong (with high OR), are clustered,
or remain statistically significant after correction for multiple
comparisons. It is important to note that our statistical correction
for the number of residues examined will be overly conservative in some
cases, because the degree of correction depends on the size of the gene
region arbitrarily chosen for study. Such correction results in
decreasing false associations (higher specificity) at the cost of
losing true associations (lower sensitivity). Our gradation of
P values uncorrected for multiple comparisons reflects a
gradation in strength of statistical evidence for associations. We
conclude that independent biological validation rather than statistical
means will best determine what P value cutoffs are optimal
for sensitivity or specificity.
We further characterized two strong examples of HLA-specific
polymorphism, I135x (substitution of consensus isoleucine at position
135) and D177x (substitution of aspartate). These were associated with
HLA-B5 and HLA-B35, respectively, in the multivariate models (Fig. 1B).
However, these broad serologically defined HLA alleles both have
subtypes with distinct epitope binding motifs. We performed
high-resolution DNA sequence-based typing on all individuals with
HLA-B5 or HLA-B35 and reexamined the associations between their HLA
subtypes and viral polymorphisms (Tables
1 and 2).
Position 135 is the anchor position of the HLA-B*5101-restricted epitope RT(128-135 IIIB) (17). Six of the other
seven amino acids in the epitope are critical stability residues
(6) and were relatively invariant (Fig. 1D). All but
1 of the 40 (98%) individuals in the cohort with HLA-B*5101 had I135x, compared with 127 of the 431 (29%) without HLA-B*5101
(P < 0.0001; Fisher's exact test) (Table 1).
For the predominant substitution observed, I135T [see fig. S2
(15)], the predicted half-time of dissociation score for
the mutant epitope (TAFTIPST) is 11 compared with 440 for the consensus sequence (TAFTIPSI), which
indicates that binding to HLA-B*5101 in vivo would be abrogated (19). I135T has also been shown to necessitate a 100-fold increase in the peptide concentration required to sensitize target cells for 50% lysis by CTLs in vitro (20). Notably, the one
subject with HLA-B*5101 and consensus sequence at position 135 had
taken highly active antiretroviral therapy only days after exposure to
HIV. He was asymptomatic, highly viremic, and seronegative at the time
of starting treatment, which may have decreased viral replication
before selection of I135x by the acute CTL response. This suggests that
I135x is characteristically selected during the acute
HLA-B*5101-restricted CTL response instead of at transmission or in
chronic infection and that protection from viral escape could
contribute to the effect of therapy in acute HIV infection, leading to
stronger chronic inhibitory CTL responses. Similarly, D177x is within
the epitope RT(175-183) known to bind HLA-B*3501 and contain a binding
motif distinct from that of other HLA-B35 subtypes (21).
After high-resolution HLA typing, D177x was associated with HLA-B*3501
specifically and not with other HLA-B35 subtypes (Table 2).
Associations with other HLA-B35-associated polymorphisms in HIV-1 RT,
I69x, D121x, and D123x were all strengthened by considering molecular
subtypes of HLA-B35.
Distribution of the presence of I135x in HIV-1 RT
within individuals with HLA-B5, HLA-B*5101 subtype, and non-HLA-B5.
High-resolution HLA genotyping strengthens the I135x-HLA allele
association. One HLA-B5 individual did not have sufficient DNA sample
for high-resolution HLA typing.
Distribution of the presence of D177x within
individuals with HLA-B35, HLA-B*3501 subtype, and non-HLA-B35.
High-resolution HLA genotyping strengthens the D177x-HLA allele
association. Seven individuals with HLA-B35 by serology did not have
DNA available for high-resolution HLA typing; in another four, HLA-B35
was not confirmed by sequencing.
In addition to positive HLA associations, we detected 25 negative HLA
associations [Fig. 1C, fig. S1C (15)]. For example,
polymorphism at positions 32, 101, 122, 169, and 210 was negatively
associated with HLA-A2 (OR < 1, P < 0.05 in all
cases). This means that HLA-A2 individuals were significantly less
likely to vary from the consensus at these sites compared with all
non-HLA-A2 individuals. HLA-A2 is the most common HLA-A allele in our
cohort and it had 5 of the 25 negative associations (compared with 3 of
the 64 positive associations). There appeared to be a predominance of
common HLA alleles with negative associations. Unlike
positive selection pressure, which causes demonstrable escape over time
in individuals, negative selection pressure favors preservation of
wild-type virus in vivo and therefore could be made evident only at a
population level. This raises the intriguing possibility that consensus
or wild-type virus is adapted to the CTL responses that it has most
often encountered during primordial evolution (that is, those CTL
responses restricted to the most common or evolutionarily conserved HLA
alleles in the host population). It is possible that the consensus
HIV-1 sequence in our study population has been selected by the host
population's predominant HLA types, as has been shown for weakly
immunogenic variants of Epstein-Barr virus in other host populations
(22, 23). We speculate that this could explain
the apparent lack of HIV-1 escape from HLA-A*0201-restricted responses
in studies that have argued against an important role for CTL escape
(24, 25), and might even explain why surprisingly
few HLA-A2- and HLA-A1-restricted epitopes have been mapped in HIV-1
(26). Primordial viral adaptation to predominant HLA
types may account, at least in part, for HIV-1 clade differences.
Furthermore, studies of HIV-1-exposed seronegative individuals suggest
that CTL responses can alter viral infectivity and susceptibility to
established primary HIV-1 infection (27-31). The HLA class
I alleles associated with natural HIV-1 resistance or susceptibility
appear to differ between racially distinct populations (27-29, 32). To some extent, this may reflect
differences in the HLA alleles that are common in different populations
and the degree to which a population-adapted wild-type virus can adapt
to the individual.
We sought to determine whether HLA-specific polymorphisms were
associated with increased plasma HIV RNA levels (viral load) (33). We examined single polymorphisms with sufficient subject numbers for comparison. HLA-A11-associated K166x is at the
anchor position of HLA-A11 epitope RT(158-166 LAI). In HLA-A11 individuals (n = 19), the median pretreatment viral
load was 5.54 ± 0.46 log copies per ml (cps/ml) of plasma
(median ± SD) in those with K166x (n = 4)
compared with 4.31 ± 0.82 log cps/ml in those without K166x
(n = 15; P = 0.045, Wilcoxon test). A second
putative CTL escape mutation within a CTL epitope (but not at a primary anchor position) showed a similar effect. The median pretreatment viral
load in HLA-B7 individuals with S162x (n = 18) was
significantly higher (5.41 ± 1.04 log cps/ml) than in those
without S162x (n = 15, 4.57 ± 0.83 log cps/ml;
P = 0.046, Wilcoxon test ). A global analysis of factors
influencing viral load at a population level showed that the presence
of viral polymorphisms in combination with their positively associated
HLA alleles or consensus amino acids with their negatively associated
HLA alleles was a significantly better predictor of pretreatment
viral load than HLA alleles or viral polymorphisms alone
(P < 0.004) (15). This suggested that the amount
of HLA-associated selection in an individual, as defined by our
analyses, explained the viral load variability in the population better
than HLA alleles.
This study encompasses demographic, clinical, and laboratory data
collected over 2210 person-years of observation. Our findings support a
model of HIV-1 evolution in vivo in which CTL escape mutations are
selected within functional limits within individuals, and this
selection during viral passage through a population determines the
wild-type or consensus viral sequence. Thus, the HLA alleles present in
a population may explain in large part both the polymorphism (viral
adaptation to individuals) and the consensus (primordial adaptation) of HIV-1 sequences in that population. Our data suggest that CTL escape mutation is common and widespread and selected by
responses restricted to a much wider array of HLA alleles than have
been studied to date.
These results are especially notable, considering the factors that
reduce the likelihood of observing significant associations in such
analyses. First, the power to detect associations is not constant for
all combinations of HLA allele and viral residue. Large numbers of
individuals would be needed to observe any polymorphism at residues
under CTL pressure but with strong functional constraint or any
associations with HLA alleles that are rare. The use of formal power
calculations identifies those HLA associations that cannot be excluded
until larger data sets are examined. Second, as suggested by the
enhancement of associations between HLA-B5 and I135x and between
HLA-B35 and D177x by high-resolution HLA typing, the molecular subtype
of an HLA allele better predicts its binding properties in vivo. Other
alleles with multiple splits of similar frequency (HLA-A10 or HLA-A19)
may have had associations that we did not detect because only broad
alleles were considered. Furthermore, molecular splits that have
opposing effects at the same viral residue would negate any association
with the broad allele. Lastly, published epitopes are more likely to be
in conserved regions, because studies tend to use laboratory reference
strains as target antigens and conserved regions are more likely to
generate measurable CTL responses in vivo (34). This
approach, in contrast, preferentially detects putative CTL epitopes in
polymorphic regions and, thus, may be complementary to standard epitope
Application of this method to other HIV-1 genes in larger populations
with more complete high-resolution HLA genotyping is needed. To that
end, an international collaboration to pool data and provide the Epipop
program to participating centers has been initiated. This approach
could be used to screen or prioritize standard testing of candidate
epitopes in all HIV-1 proteins. In the HIV envelope, effects associated
with antibody responses to HIV, CCR5 and CXCR4 genotype, and any other
polymorphisms of genes encoding products that target envelope proteins
could also be considered. Future analyses of HIV-1 RT and other
antiretroviral drug targets should adjust for drug selection effects to
examine the interactions between drug resistance mutations and putative CTL escape mutations. If CTL responses and antiretroviral drugs compete
at sites within viral sequence (35), a greater or lesser
tendency to drug resistance and response may be observed depending on HLA genotype. Individualization of antiretroviral therapy conceivably could be improved if synergistic or antagonistic interactions between immune pressure and drug pressure were better understood. Ultimately, it may be possible to generalize these approaches to examine host immune effects on hepatitis B, hepatitis C,
and other chronic human pathogens.
REFERENCES AND NOTES
9 January 2002; accepted 27 March
A. J. McMichael and
S. L. Rowland-Jones,
D. T. Evans,
T. M. Allen,
D. H. Barouch,
M. Del Val,
H. J. Schlicht,
M. J. Reddehase,
U. H. Koszinowski,
J. A. Wrobel,
Proc. Natl. Acad. Sci. U.S.A.
S. P. Layne,
B. D. Walker,
Proc. Natl. Acad. Sci U.S.A.
S. A. Mallal,
J. AIDS Hum. Retrovirol.
||The Western Australian HIV Cohort Study is a prospective
observational cohort study of HIV-infected patients that was
established in 1983. Comprehensive demographic, clinical, and
laboratory data are collected on all individuals and maintained in an
electronic database. Participants in the cohort typically have had
HIV-1 RT proviral DNA sequencing performed at first presentation and
before antiretroviral therapy and serially posttreatment as required
for clinical care since 1995.
||HIV-1 DNA was extracted from buffy coats (QIAMP DNA blood mini
kit; Qiagen, Hilden, Germany) and codons 20 to 227 of RT were amplified
by polymerase chain reaction (PCR). A nested second round PCR was done,
and the PCR product was purified with Bresatec purification columns and
sequenced in forward and reverse directions with a model 373 ABI DNA
sequencer. Raw sequence was manually edited with software packages
Factura and MT Navigator (PE Biosystems).
||All HLA-A and -B broad alleles were typed by the standard
National Institutes of Health complement-dependent microtoxicity
technique. Fifty-one HLA-B5 individuals and 57 HLA-B35 individuals had
HLA-B sequence amplified with primers to the first intronic dimorphism,
and products were sequenced by automated sequencing.
||For supplementary analyses, detailed description of all
statistical methods, and a full version of Fig. 1, see Science
Online at www.sciencemag.org/cgi/content/full/296/5572/1439/DC1.
||F. L. Ramsey, D. W. Schafer, in The
Statistical Sleuth. A Course in Methods of Data Analysis (Duxbury
Press, Belmont, CA 1997), pp. 10-13, chap. 1.3.
||B. T. M. Korber et al., HIV
Molecular Immunology Database 1999 (Theoretical Biology
and Biophysics, Los Alamos, New Mexico, 1999).
J. Mol. Biol.
A. S. de Groot,
AIDS Res. Hum. Retroviruses
N. V. Sipsas,
J. Clin. Invest.
P. O. Campos-Lima,
P. O. Campos-Lima,
M. P. Imreh,
M. G. Masucci,
J. Exp. Med.
C. M. Hay,
J. Clin. Invest.
J. AIDS Hum. Retrovirol.
K. R. Fowke,
S. L. Rowland-Jones,
J. Clin. Invest.
S. L. Rowland-Jones,
Clin. Exp. Immunol.
||The viral load assay used until November 1999 was the HIV
Amplicor (Roche Diagnostics, Branchburg, NJ) (lower limit of detection,
400 copies per ml). The Roche Amplicor HIV monitor version 1.5, Ultrasensitive (lower limit of detection, 50 copies per ml) was used
P. J. R. Goulder,
||We are indebted to participants in the WA HIV Cohort Study as
well as past and present clinical and laboratory staff of the
Department of Clinical Immunology and Biochemical Genetics, Royal Perth
Hospital, Western Australia. We thank S. Rowland-Jones for helpful
comments on the manuscript. C.B.M. was recipient of an Australian
Postgraduate Award, and M.J. is recipient of a National Health and
Medical Research Council of Australia Postgraduate
Include this information when citing this paper.
Related articles in Science:
HLA Leaves Its Footprints on HIV
Andrew McMichael and Paul Klenerman
Science 2002 296: 1410-1411.
Issue of 24 May 2002,
Copyright © 2002 by The American Association for the Advancement of Science. All rights reserved.