Journal of Virology, July 1999, p. 6197-6202, Vol. 73, No. 7
0022-538X/99/$04.00+0
Copyright © 1999, American Society for Microbiology. All rights reserved.
Identification of Biased Amino Acid Substitution
Patterns in Human Immunodeficiency Virus Type 1 Isolates from
Patients Treated with Protease Inhibitors
Robert W.
Shafer,1,*
Phillip
Hsu,2
Amy K.
Patick,3
Charles
Craig,4 and
Volker
Brendel5
Division of Infectious Diseases, Stanford
University Medical Center,1 and Stanford
University,2 Stanford, California 94305;
Agouron Pharmaceuticals, Inc., San Diego, California
921213; Department of Zoology and
Genetics, Iowa State University, Ames, Iowa
50011-32605; and Roche Discovery
Welwyn, Welwyn Garden City, Hertsfordshire, United
Kingdom4
Received 21 July 1998/Accepted 26 March 1999
 |
ABSTRACT |
Human immunodeficiency virus type 1 (HIV-1) amino acid
substitutions observed during antiretroviral drug therapy may be caused by drug selection, non-drug-related evolution, or sampling error introduced by the sequencing process. We analyzed HIV-1 sequences from
371 untreated patients and from 178 patients receiving a single
protease inhibitor. Amino acid substitution patterns during treatment
were compared with inferred substitution patterns arising evolutionarily without treatment. Our results suggest that most treatment-associated amino acid substitutions are caused by selective drug pressure, including substitutions not previously associated with
drug resistance.
| |
TEXT |
To assess the statistical
significance of sequence changes during anti-human immunodeficiency
virus (HIV) therapy, sequences should ideally also be observed in a
control group of patients not receiving drug treatment who have had
samples obtained at baseline and after an equal period of follow-up.
Generally, such data are not available. In this paper, we introduce a
novel method for assessing the statistical significance of HIV-1
protease sequence changes during therapy with a protease inhibitor. We
tested the method by using large sets of published subtype B protease
sequences from untreated HIV-1-infected individuals and from patients
receiving therapy with a single protease inhibitor.
Patients and sequences.
The patient group consisted of
patients from whom published protease sequences were available before
and after treatment. There were a total of 178 patients from eight
different studies. Thirty patients had received indinavir, 44 had
received ritonavir, 53 had received saquinavir (hard-gel formulation
[Invirase]), and 51 had received nelfinavir. The number of patients
per study, the drug treatment regimens, and the sequencing methods used
in each study are shown in Table 1.
Control sequences included the 178 baseline pretherapy sequences and
sequences from 193 other, untreated individuals (371 control sequences
in total [GenBank accession numbers given at the end of the text]).
For patients having more than one posttherapy sequence, only the final
sequence was used. Only sequences obtained by dideoxynucleotide terminator cycle sequencing were included. When multiple clones were
sequenced, the most common residue at each position was used. When a
mixture was present in a sequence obtained by direct PCR sequencing,
the residue associated with the greatest electrophoretic signal was
used. Each control sequence was considered to belong to HIV-1 subtype
B. For 85 isolates (including 48 from patients in South America, the
Middle East, Asia, and Africa), the subtype had been confirmed by
env and/or gag sequencing. In the remaining 286 cases, the isolates were considered to be subtype B based on the
patients' North American or European origin, phylogenetic analyses
demonstrating clustering with known subtype B protease sequences, and
comparison with reference subtypes (10).
To prevent the inadvertent inclusion of more than one sequence per
individual and of laboratory contaminants, the nucleic acid sequences
from the control and treated patients were examined for closely related
pairs of sequences (see also reference 7). Neighbor-joining trees of sequences from the treated and control patients were derived (by using the PHYLIP programs
[4]) and revealed several pairs of identical
sequences. Only one sequence from each of the identical sequence pairs
was included in the study. GenBank accession numbers and isolate names
of the excluded sequences are given below. (The 193 sequence pairs from
treated patients and the 371 control sequences from untreated persons represent the curated data set.)
Classification of amino acid substitution types.
The consensus
of the 371 control amino acid sequences differed from the Los Alamos
HIV Sequence Database subtype B consensus sequence at one position,
residue 63 (L in Los Alamos, P in this data set) (5). 63L is
more commonly considered to represent the wild-type amino acid at this
position and thus was also used in this study. We define five types of
possible amino acid pairs derived from the alignment of two sequences
(Table 2). If the consensus residue at a
given position is designated as C and all others by N (nonconsensus),
then an aligned pair of amino acid residues could be CC (both sequences
have the consensus residue), NN (both sequences contain the same
nonconsensus residue), CN (a substitution from the consensus residue to
a nonconsensus residue), NC (a substitution from a nonconsensus residue
to the consensus residue), or NN' (a substitution from the nonconsensus
residue N to a different nonconsensus residue, N').
View this table:
[in this window]
[in a new window]
|
TABLE 2.
Types of amino acid pairs comprising the alignment of two
HIV-1 protease sequences obtained from an individual before and
after treatment with a protease inhibitor
|
|
Classification of positions according to the distribution of
substitution types.
The distribution of amino acid substitution
types was determined at each of the 99 protease sequence positions for
the group of patients that had received a protease inhibitor (Table
3). Based on this distribution, each
position was classified according to its pattern of substitution types:
(i) conserved/invariant, all residue pairs are CC; (ii)
polymorphic/invariant, all residue pairs are CC or NN; (iii)
conserved/mutated, all residue pairs are CC or CN; (iv)
polymorphic/mutated, all residue pairs are CC, NN, NN', or CN
(substitutions only occur away from the consensus); (v)
polymorphic/revertant, all residue pairs are either CC, NN, or NC
(substitutions only occur toward the consensus); or (vi) polymorphic/bidirectional, CN and NC occur.
View this table:
[in this window]
[in a new window]
|
TABLE 3.
Classification of HIV-1 protease amino acid positions
according to the distribution of substitution types at that
position in patients receiving a protease inhibitor
|
|
In our analysis, a single reversion to consensus (NC) was sufficient
for classifying a position as polymorphic/bidirectional (pattern VI).
An alternative approach would be to establish flexible cutoffs for the
level of NC that distinguish a polymorphic/mutated (pattern IV)
position from a polymorphic/bidirectional (pattern VI) position.
Sequence variability in HIV-1 subtype B protease from untreated
patients.
It is estimated that HIV-1 variants with every possible
amino acid substitution are produced daily in HIV-1-infected
individuals (1). However, only those protease variants with
enzymatic function approximating that of wild-type virus are likely to
ever become the predominant variant in the plasma of an untreated
person. To exclude technical sequencing errors and cases of circulating virus containing unusual variants, we examined polymorphisms that were
present in at least two isolates from untreated individuals and that
were present as the predominant clone whenever multiple clones were sequenced.
Figure 1 shows the polymorphisms present
in the sequences from the 371 untreated individuals. There were 40 polymorphic positions. At 12 positions (residues 12, 13, 15, 35, 36, 37, 41, 62, 63, 64, 77, and 93), nonconsensus residues occurred
with a frequency of >10%. At 15 positions, nonconsensus
residues occurred with a frequency of 2 to 10%. At 13 positions,
nonconsensus residues occurred in two or four sequences (0.7 to
1.4%). This distribution of polymorphisms is similar to that described
in a previous publication (6).
View larger version (10K):
[in this window]
[in a new window]
|
FIG. 1.
Sequence polymorphisms present in published HIV-1
protease sequences obtained from 371 individuals not receiving
treatment with a protease inhibitor. The uppermost sequence contains
the consensus residue at that position among the 371 sequences and is
identical to the subtype B consensus sequence (5). Amino
acid residues beneath the consensus occurred in sequences from at least
two individuals. The number following the residue is the percentage of
sequences with that residue.
|
|
Pairwise differences between sequences of HIV-1 protease from
patients not receiving protease inhibitors.
To assess the effect
of naturally occurring protease sequence variation on the distribution
of genetic distances between HIV-1 isolates from untreated patients, we
calculated all possible pairwise nucleotide and amino acid distances
between the control sequences (Fig. 2).
Nucleotide distances were measured by using the Kimura two-parameter
method (4). Amino acid distances were measured as the number
of amino acid differences between two sequences.
View larger version (24K):
[in this window]
[in a new window]
|
FIG. 2.
Distribution of pairwise genetic distances among the
control patient sequences. (A) Distribution of uncorrected nucleic acid
sequence distances among the 214 control patients from whom nucleic
acid sequences were obtained. (B) Distribution of pairwise amino acid
sequence distances (number of amino acid differences) among all 371 control patients.
|
|
For the 371 control patients in whom amino acid sequences were
available, the mean pairwise sequence distance was 5.4 amino acids
(range, 0 to 16 amino acid substitutions). For the 214 patients in whom
nucleic acid sequences were available, the mean pairwise sequence
distance was 4.4% (range, 0.3 to 11.3%). Uncorrected nucleic acid
sequence distances of <1% were found in 46 (0.2%) sequence pairs,
and sequence distances of 1 to 2% were found in 745 (3.3%) sequence
pairs. Among the 46 sequence pairs with a distance of <1%, seven
pairs included BRU and eight pairs included sequences determined in the
same laboratory. Although the distribution of pairwise sequence
differences is several-fold lower for pol than for
env (7, 12, 16, 17), HIV-1 isolates with pairwise genetic distances of <1 to 2% in their protease genes should be examined for the possibility of laboratory contamination or
epidemiologic linkage.
Estimation of amino acid substitution rates in HIV-1 isolates from
untreated individuals.
Sequence variability at a given position
may indicate the tendency of this position to mutate. Alternatively, it
may represent sequence changes that are ancestral but not necessarily
frequent. To distinguish between these possibilities, we correlated
sequence variability with inferred substitution rates at the 40 variable positions. Neighbor-joining and maximum-parsimony phylogenetic trees of the control sequences were created, and the estimated numbers of inferred amino acid substitutions along the branches of each
tree were calculated (4, 8). Figure
3 shows that percent variability and
inferred amino acid substitution rate were highly correlated
(r2 = 0.91; P < 0.001). This
analysis suggests that the sequence variability at a given
position is a reliable indicator of the tendency of this position to
mutate.
View larger version (11K):
[in this window]
[in a new window]
|
FIG. 3.
Relationship between percentage of variability and
inferred number of amino acid substitutions arising evolutionarily
among subtype B HIV-1 isolates from untreated individuals. The
percentage of isolates with a nonconsensus residue is plotted on the
x axis. The inferred number of amino acid substitutions was
derived from a neighbor-joining tree by using the programs PHYLIP and
MacClade (4, 8). For this analysis, the subtype A isolate
U455 (5) was used as an outgroup. Nearly identical results
were obtained if a maximum-parsimony tree was used instead of the
neighbor-joining tree.
|
|
Comparison of the patterns of amino acid substitutions in HIV-1
isolates from patients receiving protease inhibitors with those
obtained from untreated patients.
Table
4 contains a summary of
those substitutions that occurred at least twice during
protease inhibitor therapy. There were means of 5.8 amino acid sequence
substitutions in the 30 sequence pairs from patients receiving
indinavir, 3.8 substitutions in the 44 sequence pairs from patients
receiving ritonavir, 2.7 substitutions in the 53 sequence pairs from
patients receiving saquinavir, and 2.7 substitutions in the 51 sequence
pairs from patients receiving nelfinavir. The distribution of sequence
patterns included 40 pattern I positions (conserved/invariant), 3 pattern II positions (polymorphic/invariant), 19 positions with pattern
III substitutions (conserved/mutated), 12 positions with pattern IV
substitutions (polymorphic/mutated), 2 positions demonstrating only
reversion to consensus (pattern V), and 23 positions with bidirectional substitutions (pattern VI, polymorphic/bidirectional).
A permutation method was used to assess the statistical significance of
high levels of CN substitutions (away from the consensus) relative to
NC substitutions (toward the consensus) at a particular position. For
each of the 178 treatment sequence pairs, we determined the
substitution type at each position. We then randomly picked an equal
number of sequence pairs from the control group with the same distances
as those observed in the treated group. Thus the random sequence pairs
differed overall as much as the treatment sequence pairs, but
variability occurred in how the substitutions were distributed among
the substitution types.
The statistical significance of the observed treatment counts was
assessed by repeating the random selection 10,000 times and deriving
the empirical distribution of the aggregate counts of substitution
patterns at each sequence position. The counts observed in the random
pairings of control group sequences were ordered by value. The
uncorrected P value is the tail probability of the count for
the treatment group data relative to this empirical distribution. For
example, the count of 13 CN/2 NC mutations at position 93 ranked tied
29 to 54 in the empirical distribution (28 random pairings had higher
counts, 9,946 had lower counts), giving a conservative tail probability
of 54/10,001 = 0.0054. To assess the significance of CN
substitutions at sites with <1% variability (pattern III positions in
Table 4), these positions were assumed to have a nonconsensus residue
in 1 of 371 (0.0027) untreated individuals (nonconsensus residues at
these positions were actually present in 0 to 1 of 371 sequences).
Of the 37 positions undergoing at least 2 substitutions during protease
inhibitor therapy, 36 had an increased rate of CN substitutions
relative to NC substitutions (residue 64 gives a tie [Table 4]). The
statistical significance of changes at any one position is weakened by
the necessary adjustment for multiple comparisons. However, the large
number of CN substitutions compared with NC substitutions suggests that
most treatment-associated amino acid substitutions are caused by
selective drug pressure.
CN substitutions predominated at polymorphic positions as well as at
highly conserved positions. Many of the CN substitutions occurred at
positions previously shown to confer drug resistance by experimental
methods. However, several positions with high rates of CN mutations,
such as I13V, Q58E, G73S, and T74A/S (and possibly V75I), have not been
previously associated with drug resistance. Given enough sequence data
from patients receiving protease inhibitors, the statistical
association between mutations such as these and protease inhibitor
treatment should be reexamined.
Our analysis included protease sequences obtained in several
different studies, including dose-finding studies in which the drugs were often used at suboptimal dosages. Therefore, we made no attempt to compare the sequence changes observed with
different protease inhibitors. Studies in which sufficient
numbers of sequenced protease isolates are obtained from patients
receiving optimal protease inhibitor therapy are needed to
compare sequence changes associated with different treatments.
Nucleotide sequence accession number.
The accession numbers of
the baseline and follow-up sequences from patients receiving protease
inhibitor therapy are given in Table 1. Accession numbers for the HIV-1
protease sequences from 195 additional untreated patients include the
following: D10112, K02007, K02013, L02317, L08463 to L08464, M17449, M17451, M26727, M38429, M93258, M96155, U12738, U12745 to
U12756, U19411 to U19415, U19417 to U19431, U19436, U19441, U19446 to
U19449, U19457 to U19458, U21122, U21135, U26546, U31385 to U31395, U31399 to U31406, U31408, U31412, U34603, U37270, U43096, U43141,
AF004394, AF005495, AF009369 to AF009375, AF009379 to AF009381,
AF025722, AF025724, AF025726, AF025731, AF025732, AF025734, AF025736, AF025744, AF027708, AF027710, AF027715 to AF027716, AF027718, AF027720,
AF040579, AF040584, AF040591, AF040596, AF040603, AF040608, AF040611,
AF042100 to AF042101, AF042103 to AF042105, AF047306, AF047317, AF013857, AF078556 to AF078606, and AJ006287. Accession numbers of
sequences excluded from analysis include AJ002496, AJ002497, AJ002499,
AJ002500, U19452, U31409, AF025738, AF042100, AF078577,
AF078578, and AF078669. Alignments of the sequences analyzed in
this paper can be accessed at the web address in reference
16.
| |
ACKNOWLEDGMENTS |
Robert Shafer was supported in part by NIH grant AI27666.
We thank Dale Kempf and Akhter Molla for providing the complete amino
acid sequences of the isolates published in reference 9. We also thank David Katzenstein for critical
review of the manuscript.
| |
FOOTNOTES |
*
Corresponding author. Mailing address: Division of
Infectious Diseases, S-156, Stanford University Medical Center, Room
S156, 300 Pasteur Dr., Stanford, CA 94305-5107. Phone: (650) 725-2946. Fax: (650) 725-2395. E-mail: rshafer{at}cmgm.stanford.edu.
| |
REFERENCES |
| 1.
|
Coffin, J. M.
1995.
HIV population dynamics in vivo: implications for genetic variation, pathogenesis, and therapy.
Science
267:483-489.
|
| 2.
|
Condra, J. H.,
D. J. Holder,
W. A. Schleif,
O. M. Blahy,
R. M. Danovich,
L. J. Gabryelski,
D. J. Graham,
D. Laird,
J. C. Quintero,
A. Rhodes,
H. L. Robbins,
E. Roth,
M. Shivaprakash,
T. Yang,
J. A. Chodakewitz,
P. J. Deutsch,
R. Y. Leavitt,
F. E. Massari,
J. W. Mellors,
K. E. Squires,
R. T. Steigbigel,
H. Teppler, and E. A. Emini.
1996.
Genetic correlates of in vivo viral resistance to indinavir, a human immunodeficiency virus type 1 protease inhibitor.
J. Virol.
70:8270-8276[Abstract].
|
| 3.
|
Craig, C.,
E. Race,
J. Sheldon,
L. Whittaker,
S. Gilbert,
A. Moffatt,
J. Rose,
S. Dissanayeke,
G. W. Chirn,
I. B. Duncan, and N. Cammack.
1998.
HIV protease genotype and viral sensitivity to HIV protease inhibitors following saquinavir therapy.
AIDS
12:1611-1618[Medline].
|
| 4.
|
Felsenstein, J.
1993.
PHYLIP (Phylogenetic Inference Package). Version 3.5.
University of Washington, Seattle.
|
| 5.
|
Korber, B. T.,
B. Hahn,
B. Foley,
J. W. Mellors,
T. Leitner,
G. Myers,
F. E. McCutchan, and C. L. Kuiken.
1997.
Human retroviruses and AIDS: a compilation and analysis of nucleic and amino acid sequences.
Theoretical Biology and Biophysics, Los Alamos National Laboratory, Los Alamos, N. Mex.
|
| 6.
|
Kozal, M. J.,
N. Shah,
N. P. Shen,
R. Yang,
R. Fucini,
T. C. Merigan,
D. D. Richman,
D. Morris,
E. R. Hubbell,
M. Chee, and T. R. Gingeras.
1996.
Extensive polymorphisms observed in HIV-1 clade-B protease gene using high-density oligonucleotide arrays.
Nat. Med.
2:753-759[Medline].
|
| 7.
|
Learn, G. H., Jr.,
B. T. M. Korber,
B. Foley,
B. H. Hahn,
S. M. Wolinsky, and J. I. Mullins.
1997.
Maintaining the integrity of human immunodeficiency virus sequence databases.
J. Virol.
70:5720-5730[Abstract/Free Full Text].
|
| 8.
|
Maddison, W. P., and D. R. Maddison.
1992.
MacClade, version 3.01.
Sinauer and Associates, Sunderland, Mass.
|
| 9.
|
Molla, A.,
M. Korneyeva,
Q. Gao,
S. Vasavanonda,
P. J. Schipper,
H. M. Mo,
M. Markowitz,
T. Chernyavskiy,
P. Niu,
N. Lyons,
A. Hsu,
G. R. Granneman,
D. D. Ho,
C. A. Boucher,
J. M. Leonard,
D. W. Norbeck, and D. J. Kempf.
1996.
Ordered accumulation of mutations in HIV protease confers resistance to ritonavir.
Nat. Med.
2:760-766[Medline].
|
| 10.
| National Center for Biotechnology Information.
1998. HIV-1 Subtyping Tool [Online.]
www.ncbi.nlm.nih.gov/retroviruses/HIV1. [6 May 1999, last date
accessed.]
|
| 11.
|
Patick, A. K.,
M. Duran,
Y. Cao,
D. Shugarts,
M. R. Keller,
E. Mazabel,
M. Knowles,
S. Chapman,
D. R. Kuritzkes, and M. Markowitz.
1998.
Genotypic and phenotypic characterization of human immunodeficiency virus type 1 variants isolated from patients treated with the protease inhibitor nelfinavir.
Antimicrob. Agents Chemother.
42:2637-2644[Abstract/Free Full Text].
|
| 12.
|
Quinones-Mateu, M. E.,
A. Holguin,
J. Dopazo,
I. Najera, and E. Domingo.
1996.
Point mutation frequencies in the pol gene of human immunodeficiency virus type 1 are two- to threefold lower than those of env.
AIDS Res. Hum. Retroviruses
12:1117-1128[Medline].
|
| 13.
|
Ruiz, L.,
M. Nijhuis,
C. Boucher,
T. Puig,
A. Bonjoch,
J. Martinez-Picado,
S. Marfil,
D. de Jong,
D. Burger,
A. Arno,
M. Balague, and B. Clotet.
1998.
Efficacy of adding indinavir to previous reverse transcriptase nucleoside analogues in relation to genotypic and phenotypic resistance development in advanced HIV-1-infected patients.
J. Acquired Immune Defic. Syndr. Hum. Retrovirol.
19:19-28[Medline].
|
| 14.
|
Schapiro, J. M.,
M. A. Winters,
F. Stewart,
B. Efron,
J. Norris,
M. J. Kozal, and T. C. Merigan.
1996.
The effect of high-dose saquinavir on viral load and CD4+ T-cell counts in HIV-infected patients.
Ann. Intern. Med.
124:1039-1050[Abstract/Free Full Text].
|
| 15.
|
Schmit, J.-C.,
L. Ruiz,
B. Clotet,
A. Raventos,
J. Tor,
J. Leonard,
J. Desmyter,
E. De Clercq, and A. M. Vandamme.
1996.
Resistance related mutations in the HIV-1 protease gene of patients treated for 1 year with the protease inhibitor ritonavir (ABT-538).
AIDS
10:995-999[Medline].
|
| 16.
|
Shafer, R. W.,
D. Stevenson, and B. Chan.
1999.
Human immunodeficiency virus reverse transcriptase and protease sequence database.
Nucleic Acids Res.
27:348-352[Abstract/Free Full Text].
|
| 17.
|
Shafer, R. W.,
T. K. Chuang,
P. Hsu,
C. Bodley White, and D. A. Katzenstein.
1999.
Sequence and drug susceptibility of subtype C protease from human immunodeficiency virus type 1 seroconverters in Zimbabwe.
AIDS Res. Hum. Retroviruses
15:65-69[Medline].
|
| 18.
|
Zhang, Y.-M.,
H. Imamichi,
T. Imamichi,
H. C. Lane,
J. Falloon,
M. B. Vasudevachari, and N. P. Salzman.
1997.
Drug resistance during indinavir therapy is caused by mutations in the protease gene and in its Gag substrate cleavage sites.
J. Virol.
71:6662-6670[Abstract].
|
Journal of Virology, July 1999, p. 6197-6202, Vol. 73, No. 7
0022-538X/99/$04.00+0
Copyright © 1999, American Society for Microbiology. All rights reserved.
![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Top
Abstract
Text
