JOURNAL OF VIROLOGY, OCt. 1994, p. 6672-6683
Copyright (C 1994, American Society for Microbiology
Vol. 68, No. 10
Human Immunodeficiency Virus Type 1 Evolution In Vivo
Tracked by DNA Heteroduplex Mobility Assays
ERIC L. DELWART,1t HAYNES W. SHEPPARD,2 BRUCE D. WALKER,3
JAAP GOUDSMIT,4 AND JAMES I. MULLINS"*
Department ofMicrobiology and Immunology, Stanford University School ofMedicine, Stanford,
California 94305-54021; Viral and Rickettsial Disease Laboratory, Califomia Department of
Health Services, Berkeley, Califomia 907042; Infectious Diseases Unit, Massachusetts
General Hospital, Boston, Massachusetts 021143; and Human Retrovirus
Laboratory, Academic Medical Center, Amsterdam 1105AZ, The Netherlands4
Received 1 February 1994/Accepted 18 July 1994
High mutation rates and strong selective pressures imposed on human immunodeficiency viruses in vivo
result in the formation of pools of genetic variants known as quasispecies. DNA heteroduplex mobility and
tracking analyses were used to monitor the generation of HIV sequence diversity, to estimate quasispecies
complexity, and to assess the turnover of genetic variants to approach an understanding of the relationship
between viral quasispecies evolution in vivo and disease progression. Proviral DNA pools were nearly
homogeneous soon after sexual transmission. The emergence and clearance of individual variants then
occurred at different rates in different individuals. High quasispecies complexity was found in long-term-
infected, asymptomatic individuals, while rapid CD4+ cell decline and AIDS were often, but not always,
associated with lower quasispecies complexity. Proviral genetic variation was often low following in vitro
culture, because of the outgrowth of one or a few variants that often became more abundant only later as
proviruses in peripheral blood mononuclear cells. These studies provide insight into the dynamics of human
immunodeficiency virus sequence changes in vivo and illustrate the utility of heteroduplex analysis for the
study of phenomena associated with rapid genetic changes.
The already high (41) and growing (11, 26) level of polymor-
phism between independent human immunodeficiency virus
type 1 (HIV-1) strains represents an important problem in the
design of widely cross-reactive vaccines and antiviral drugs.
The choice of variants to analyze for various characteristics,
such as drug susceptibility, cytopathicity, cellular tropism,
neutralization, or cytotoxic T-lymphocyte recognition proper-
ties, is complicated by the multitude of genetic and phenotypic
variants that can exist sequentially and/or simultaneously
within HIV-1-infected individuals. The high level of viral
genetic variation detected within infected individuals (38, 65,
67, 73) and the frequent emergence of phenotypic variants (1,
3, 16, 18, 27, 36, 60) are likely to be driven by the high mutation
rate of retroviral reverse transcriptase (12, 50, 55) and in part
by the strong selective pressures for structural change imposed
by the immune system (5, 10, 69).
To expand our understanding of the evolution of HIV-1,
simple techniques were developed to allow rapid quantitation
of genetic relationships and sensitive detection of minor vari-
ants in complex quasispecies. These techniques, referred to as
heteroduplex mobility assay (HMA) and heteroduplex tracking
assay (HTA), grew out of the observation that when HIV-1 env
gene sequences are amplified by nested PCR (N-PCR; 40)
from the peripheral blood mononuclear cells (PBMC) of
infected individuals, related DNA products coamplified from
divergent templates can randomly reanneal to form heterodu-
plexes that migrate with reduced mobility in neutral polyacryl-
*Corresponding author. Present address: Departments of Microbi-
ology and Medicine, University of Washington, Seattle, WA 98915.
Phone: (206) 543-5492. Fax: (206) 543-8292. Electronic mail address:
t Present address: Aaron Diamond Center for AIDS Research, New
York University, New York, NY 10016.
amide gels (11). By using these techniques, genetic relation-
ships between multiple viral DNA template molecules can be
rapidly evaluated (11). The heteroduplex assays can also detect
the presence of minor variants in a background of distinct
molecules in a manner more sensitive than anything but
exhaustive, and thereby impractical, DNA-sequencing studies.
Here these assays are shown to afford description of overall
viral quasispecies diversity in vivo, development of viral diver-
sity in newly infected individuals, and turnover of individual
genetic variants over time. A link between increased viral
quasispecies complexity and a prolonged disease course is also
MATERIALS AND METHODS
Study subjects. PBMC from HIV-1-infected homosexual
male individuals examined in this study were drawn from three
cohorts. Subject 145/MA (21, 22) is a homosexual man from
the Massachusetts General Hospital cohort. Samples and
derived clones were taken from the period 4.5 years (month 1,
April 1989; 28) to nearly 7 years (63) postinfection. He was
asymptomatic in early 1994, with a stable CD4+ cell count in
the range of 400/mm3, after more than 9 years of HIV-1
infection. Subjects 537 and 1058 were drawn from a prospec-
tive study cohort from Amsterdam (26). PBMC samples were
obtained from these individuals from a time prior to serocon-
version (SC) through .4.5 years of infection (see Fig. 5). Their
CD4 cell numbers, disease status, and therapy history are
described elsewhere in the text. The remaining 22 samples
were taken as part of a cross-sectional analysis of the San
Francisco Men's Health Study cohort (62). When known, the
time (in months) from SC is indicated (see Fig. 3). For the
individuals determined to be infected at the time of entry into
the study in 1984 and 1985 (see Fig. 3), the time (in months)
HIV-1 EVOLUTION IN VIVO
since SC is preceded by a greater-than (>)sign. The CD4+
count at the time the sample was taken, the average CD4+
loss per year since infection or entry into the study, and the
presence or absence of an AIDS-defining clinical illness are
also shown for each individual (see Fig. 3).
DNA preparation, PCR amplification, and proviral DNA
quantitation. PBMC DNA purification, PCR primers, and
amplification conditions have been described previously (11).
The first-round primers were ED3 and ED12, used with 1.4
mM MgCl2, and the second-round primers were ES7 and ES8,
used with 1.8 mM MgCl2. Primers ES7 and ES8 amplify the V3
to V5 region ofenv, resulting in a product of approximately 702
bp, of which 627 bp are target dependent (11). Semiquantita-
tive estimation of HIV-1 provirus copy number was performed
by duplicate serial dilution of infected PBMC DNA to 0.2,
0.05, 0.02, 0.005, and 0.002 ,ug, followed by N-PCR. The lowest
concentration of infected-cell DNA to yield a positive PCR
signal (using N-PCR with single-proviral-molecule sensitivity
) was used to estimate the proviral DNA load in 1 ,ug of
PBMC DNA. When a low proviral load precluded simulta-
neous amplification ofmore than 20 to 50 proviruses from 1 ,g
of PBMC DNA, multiple PCRs were pooled, denatured, and
Heteroduplex formation andanalysis.Heteroduplex forma-
tion and the gel electrophoresis conditions used have been
described previously (11). Briefly, heteroduplexes are formed
between divergent, PCR-amplified sequences by melting com-
bined DNAs at 94°C and reannealing them by rapid cooling on
wet ice. Heteroduplexes are then displayed by electrophoresis
on a 5% neutral polyacrylamide gel and stained with ethidium
bromide. The mobility of heteroduplexes formed between
sufficiently divergent molecules (usually >1 to 2% mismatches
or single-base gaps due to insertion-deletion)
relative to that of the fully complementary homoduplex mol-
ecules. HMA involves ethidium bromide fluorescence detec-
tion of heteroduplexes formed from a mixture of variants from
a single source or a mixture of variants from different sources.
HTA is a similar process in which one PCR product used as a
labeled probe, here radioactive, is mixed with a 100-fold excess
(driver) of an unlabeled PCR product from a different source
(11). Under these conditions, probe sequences are driven
completely into heteroduplexes with the driver, and thus an
autoradiogram of the resulting polyacrylamide gel reveals only
these heteroduplexes and provides a visual display of the
relationship between the two virus populations under study.
The sequences used (see Fig. 2C) were as follows: PE, 1 to 8;
BU, 1 to 8; MA-1, 5,6, 7,9, 10, 13, 16, and 18; MA-22, 301,303,
305, 306, 307, 311, and 314; subtype B, SF162, SF2, pNL4-3,
PE1, BU1, and MA5; subtype D, MAL and NDK.
DNA sequencing. In a manner analogous to that of Sim-
monds et al. (66), endpoint dilution of infected-cell DNA was
conducted prior to the first round of N-PCR to generate
products from single HIV proviruses. The N-PCR primer sets
used were ED3-ED14 and ES7-ES8 (11). Dilutions that re-
sulted in amplified products in roughly half of the PCRs were
used to generate templates for sequencing (66). To confirm the
template homogeneity of each reaction, 5 ,lI of the PCR was
examined on a 5% polyacrylamide gel after denaturation and
reannealing alone and after mixing with a PCR fragment
derived from a molecular clone of HIV-1 LAI (pNL4-3). If the
sample product alone showed a single homoduplex band and
only two heteroduplexes were detected when reannealed with
the LAI sequence,
endpoint. For DNA sequencing, amplified fragments were
purified by agarose gel electrophoresis and Spin-X centrifuge
filtration (Costar Inc.). Fluorescent-dye-labeled universal for-
it was considered to be a molecular
ward and reversesequencing primers complementary to the
ends of ES7 and ES8 were then used to sequence the PCR
fragment by cycle sequencingwith Taq polymerase and an
automated DNAsequencerin accordance with the manufac-
turer's(Applied Biosystems,Foster City, Calif.) instructions.
With theexceptionof some of the pre-SC sequences men-
tioned below, thisprocedurewas used to generate all of the
sequencesfrom patients 537 and 1058.
Forpatients PE, BU, and MA, PCR fragments were sub-
cloned into pUC18 following XhoI
fragments generated by usingES7X (5'-ACTGAGTCCTCGA
GCTGTTAAATGGCAGTCTAGC) and ES8S
GAGCT CACT[CTCCAAT[GTCCCTCA) as primers for
the second round ofamplification.These primers were iden-
tical to ES7 and ES8, respectively, except that the ends
complementary to the universal sequencing primers were
replaced byXhoI (ES7X)and SacI (ES8S) restriction endonu-
clease sites(underlined).In all cases, sequencing was per-
formed from both directions with only a short overlap in the
center of thesequenced region. Lastly, four of the five pre-SC
sequencesfrom 1058 and three of the five pre-SC sequences
from 537 were determined from plasmid subclones derived
from the 3' end of the viral genome (3a).
DNAsequence analysis.DNA sequences were aligned by
usingthe GENALIGN program (Intelligenetics, Inc., Moun-
tain View, Calif.)and improved by visual inspection with the
multiple aligned sequenceeditor (14). Simple mismatch fre-
quencies (Hamming distances) were determined with the
programDOTS (64a), not counting sequences within gaps
introduced to maintain alignments.
Nucleotide sequence accession numbers. The nucleotide
sequences reportedhere are available from GenBank under
U00837, U00839, U13240-U13252, and U13373-U13388.
and Sacl cleavage of
Visual display of quasispecies complexity by HMA. To
determine thetypeof distortions of the DNA double helix that
affectelectrophoreticmobility, heteroduplexes were formed by
usingDNAfragmentsderived from previously sequenced,
divergentHIV-1 envgenes.This analysis was confined to an
approximately 620-bp portion of the HIV-1 envelope gene
encodingthe C-terminal half of surface envelope protein SU
encompassingvariable domains V3 to V5 (41).
When differentpairsof amplified products were mixed and
reannealed, singlebands were observed in agarose gels (Fig.
1A). However,inpolyacrylamide gels the same DNA mixtures
resulted in comigrating homoduplex bands plus, in some
instances, two additional heteroduplex bands (Fig. 1B). Mix-
tures of three different sequences yielded six heteroduplex
bands(Fig. 1D). Thus, each possible heteroduplex was de-
tected. The fact that the heteroduplexes migrate with distinct
mobilities indicates that the strand-specific composition of
mismatched andunpairednucleotides affects their mobility.
Heteroduplexesformed from molecules with mismatched
nucleotides but withoutgapswere detected by using nondena-
turingconditions when the degree of divergence exceeded
about 1.4% andgenerallyincreased with the degree of mis-
match(Fig. 1C). These mismatches are thought to result in
"bubbles"(4)in the heteroduplexwhich retard the mobility of
thefragment through the polyacrylamide matrix. Heterodu-
plexes containingasingle 3-bp (Fig. 1B, lanes marked +) or a
single 1-bp (datanotshown)internal gap all displayed mobility
shifts,even in agenetic backgroundof <0.8% mismatches. A
singlethree-base insertion-deletion in a 3.2-kb DNA hetero-
VOL. 68, 1994
DELWART ET AL.
0.16 0.95 1.42 0.79 0.95 1.11
1.31.4 1.83.7 3.9 4.0
FIG. 1. Effects of nucleotide mismatches and gaps on heteroduplex
mobility. Heteroduplexes (-670 bp) with and without a three-base gap
were analyzed on 2.5% agarose (A) and 5% polyacrylamide (B) gels.
The percent mismatch for each heteroduplex and the presence (+) or
absence (-) of a gap are indicated between panels A and B. (C)
Heteroduplexes with mismatches but without gaps analyzed on a 5%
polyacrylamide gel. The percent mismatch for each heteroduplex is
indicated above the panel. (D) Heteroduplexes formed with three
HIV-1 sequences with unique gaps were analyzed on a 5% polyacryl-
amide gel. Lanes M contained molecular size markers of 1.37, 1.08,
and 0.87 kb, respectively.
duplex was also detected by polyacrylamide gel electrophoresis
(data not shown). Mobility shifts due to insertions or deletions
generally increased with the size of the gap (data not shown).
The greater effect of an insertion-deletion, relative to base
substitutions, on electrophoretic mobility is thought to be due
to a "kink" in the double helix required to accommodate the
extra bases in the DNA heteroduplex (4, 70).
PCR amplification products derived from HIV-1 proviruses
in the PBMC of infected individuals were examined next (Fig.
2A). To ensure representative sampling of the viral gene pool,
each quasispecies in Fig. 2, as well as all of the other
quasispecies examined in this study, contained at least 20 to 50
template molecules of HIV-1 DNA, as measured by N-PCR
endpoint dilution titration (see Materials and Methods) (11).
When examined on a 5% acrylamide gel, these samples
revealed complex mobility patterns that were slightly different
between parallel N-PCRs initiated from the same infected
PBMC DNA sample (Fig. 2A, lanes 1 and 7). In contrast, when
a higher input (500 to 1,000 copies) of cellular proviral DNA
was used in the first round of PCR, the patterns obtained were
identical (Fig. 2B). When only second-round PCRs were
repeated with the first-round product as the template, identical
heteroduplex patterns were reproduced, irrespective of the
input copy number (data not shown). This indicates that the
PCR amplification process itself provides faithful and repro-
ducible amplification of a template mixture. The slight differ-
ences noted for N-PCR starting from a small number of
templates from the same PBMC preparations (Fig. 2A, lanes 1
and 7) are therefore likely to be due to amplification of
different sets of template molecules from the quasispecies.
Individual molecules from the mixtures shown in Fig. 2A
were molecularly cloned, their nucleotide sequences were
AABB C C
FIG. 2. HMA of viral quasispecies and component molecules. (A)
Each lane number refers to the month of collection of PBMC from
asymptomatic subject MA beginning approximately 4.5 years after
infection (21, 22, 28). Duplicate N-PCR initiated with different ali-
quots of the same infected PBMC DNA are shown for the 1- and
7-month time points. CC refers to a 4-week coculture of PBMC from
time point 1 with uninfected PBMC (28). BU and PE correspond to
PBMC samples taken from two patients with AIDS. Patient MA lanes
1 to 23 consisted of amplification products initiated with 20 to 50
proviruses as measured by PCR endpoint dilution titration (see
Materials and Methods). The CC and AIDS patient PE and BU PCRs
were initiated with 500 to 1,000 proviruses. (B) The letters above the
lanes refer to PCR products from three individuals in which the input
was PBMC DNA harboring approximately 500 to 1,000 proviruses.
Duplicate reactions are shown. (C) Relative mobilities of intrasubject
heteroduplexes were measured for all possible pairwise combinations
of eight representative sequences (28 pairs) from subjects MA, BU,
and PE and epidemiologically unrelated sequences from U.S. and
African sources (see Materials and Methods). Heteroduplex mobilities
were calculated as the average distance of migration of the two
heteroduplex bands divided by the distance of migration of the
homoduplex bands (O). DNA sequence divergence between pairs (0)
was measured by simple counting of mismatches between aligned
sequences, not counting bases unpaired because of insertions and
deletions. The average value for each group is shown as a horizontal
bar. MA1/22 is a comparison of sequences derived from subject MA at
month 1 with sequences derived at month 22. Subtype B/B is a
comparison of sequences from epidemiologically distinct viruses from
subjects infected within the United States, all corresponding to viruses
belonging to the B envelope sequence subtype (41). Subtype B/D is a
comparison of sequences from six subtype B viruses from the United
States (pNL4-3, SF2, SF162, MA5, BU01, and PE01) and two subtype
D isolates from Zaire (NDK and MAL). Included in the subtype B/D
group is the NDK-MAL (subtype D/D) comparison, which displayed
the fastest mobility and the least sequence divergence in the group.
\DI' K lf
HIV-1 EVOLUTION IN VIVO
determined, and they were used as templates for PCR to
generate fragments for HMA (see Materials and Methods). As
shown in Fig. 2C, when individual pairs of sequences were
compared, similar distributions were obtained by plotting
either heteroduplex mobilities or DNA sequence divergence,
the latter determined by simple counting of mismatches be-
tween aligned sequences, discounting bases unpaired because
of insertions and deletions. We found similar concordance of
HMA and DNA sequence divergence when we used molecules
derived from sequential provirus samples from one individual
(Fig. 2C, MA 1/22), as well as when we compared sequences
from different individuals infected with viruses from the same
(Fig. 2C, subtype B/B) or different (Fig. 2C, subtype B/D)
HIV-1 envelope sequence subtypes. Importantly, the patterns
generated by comparison of multiple individual pairs of mol-
ecules were, in aggregate, similar to that observed following
display of the overall quasispecies by HMA (Fig. 2A). HMA
therefore allows rapid visualization of the sequence complexity
ofHIV-1 quasispecies in vivo by direct observations ofmultiple
A quantitative analysis of the relationship between the
mobility of heteroduplexes formed between pairs of molecules
and DNA distance has been presented (11). In that study,
small differences in DNA sequence divergence between closely
related molecules could lead to relatively large differences in
heteroduplex mobility and therefore less reliable inferences of
sequence divergence in this range. The molecules examined in
this study are also from the low end of divergence previously
evaluated. We found that HMA tended to overestimate DNA
sequence divergence between samples from subject MA (an
asymptomatic individual infected for 4.5 to 6.5 years when
these samples were taken) but not from the two subjects with
AIDS (PE and BU). This was likely due to the larger number
of insertions-deletions in MA heteroduplexes relative to PE
and BU heteroduplexes. The average numbers of unpaired
bases (resulting from insertions-deletions) per heteroduplex
were 1.3 bases for PE, 1.8 bases for BU, 12.4 bases for MA-1,
and 10.85 bases for MA-22. Thus, quantitative representation
of individual quasispecies from heteroduplex data is compli-
cated by the particularly large effect of gaps within annealed
sequences on heteroduplex mobility for closely related se-
quences. As a result, the studies presented here focused on
exploration of the ability of heteroduplex assays to provide a
qualitative representation of complex quasispecies.
Complexity of HIV-1 quasispecies in vivo. The general
concordance between heteroduplex mobility and genetic dis-
tance (11) and between the display of multiple coamplified
variants and individual pairwise comparisons of DNA se-
quences (Fig. 2) justified the use of HMA as a rapid means of
assessing viral quasispecies complexity. HMA was therefore
used to examine HIV-1 in uncultured PBMC from 22 individ-
uals whose minimum times since SC were known (from 18 to
more than 93 months; Fig. 3). Samples were arranged accord-
ing to the rate of CD4+ cell decline since the time of infection
or, if the time of time infection was not known, since entryinto
the study. The CD4+ cell count, the occurrence of an AIDS-
defining clinical condition at the time of sampling, and the
minimum number of months since SC are indicated. Four of
the five asymptomatic individuals with the lowest rate of CD4+
cell loss (left side of Fig. 3) showed the highest apparent
degree of quasispecies diversity. Those individuals with a more
rapid CD4+ cell decline, a lower CD4+ cell number, and AIDS
had generally lower quasispecies complexity, although there
was wide variation. Three of the four individuals with the
fastest rate of CD4+ cell decline (right side of Fig. 3) had
among the lowest quasispecies complexities observed. A gen-
*Avg.CD4+ Cell Lass /Year
CD+ Cells mm 3-
s-- s * s s , * S * ,
Yf2 ,29 3 Y93 'I8
)9 3 V
FIG. 3. HIV env DNA sequences present in uncultured PBMC
samplesfrom 22long-term-infected individuals from the San Francisco
Men's Health Study (62). Each quasispecies (composed of -20 to 50
proviruses, as determined by N-PCR endpoint dilution titration [see
Materials and Methods]) is shown along with the rate of CD4+ cell
decline per year that the individual experienced, the CD4+ cell count,
and the presence of AIDS-defining illness at time of sampling. When
known, the time (in months) since SC is indicated. The contents oflane
M are described in the legend to Fig. 1.
eral concordance was therefore noted between high quasispe-
cies complexity and a slow decline in CD4+ cell numbers and,
to a lesser extent, among low quasispecies complexity, a fast
decline in CD4+ cell numbers, and rapid progression to
disease. It should be emphasized that because of their short
survival times, individuals with the fastest progression rates
were sampled only after relatively short periods of infection,
possibly before a high level of genetic diversity could have
Development of quasispecies complexity in HIV-i-infected
individuals. HMA was also used to monitor the generation of
quasispecies complexity over time in newly infected individu-
als. Multiple HIV-1 env gene segments were simultaneously
amplified from PBMC taken from 12 individuals within 1year
of primary infection. As shown inFig. 4,each of foursamples
taken at the time point prior to SC appeared homogeneous
within the resolution limits (1 to 2% mismatches)of thisassay.
Similarly, three of five quasispeciesexamined around the time
of SC and one of three quasispeciesexamined 6 months later
also appeared homogeneous.
Serially acquired PBMC samples from two of the pre-SC
subjects were similarly analyzed (Fig. SA). Quasispeciescom-
plexity increased from the point of initially homogeneous
samples in both individuals. Heteroduplex mobility patterns
suggested that the viruspopulationinsubject1058 diversified
slowly because of the introduction of base substitutions, as
indicated by the gradually increasingwidth of the ethidium
bromide-stained band. Large, discrete mobility shifts were
detected in subject 537, suggestingthe introduction of dele-
tions or insertions into the env gene of amajorfraction of
proviruses present at later time points. To evaluate these
interpretations, DNA molecules from the earliest and latest
VOL. 68, 1994
DELWART ET AL.
| . |
- _ ...
FIG. 4. HIV env DNA sequences in uncultured PBMC samples
from 12 individuals taken early in infection. N-PCR were used to
amplify the -700-bp V3 to V5 fragment from PBMC DNA containing
.20 to 50 proviruses. Viral DNA was amplified at the 6-month interval
before SC (Pre S.C.), at the first seropositive collection (At S.C.), and
at the 6-month collection following SC (Post S.C.). Cohort members
were sampled at 6-month intervals.
time points were PCR amplified and their DNA sequences
over regions V3 to V5 (620 bp of the sequence between
primers) were determined. Five sequences were determined
from each of the pre-SC time points and from the latest
samples analyzed, collected more than 4.5 years later. The
consensus pre-SC sequences were then compared to that of
each molecule derived from the later time point (Fig. 5C).
Endpoint dilution-derived, individual proviral templates from
the later time points were amplified to ensure that the same
proviruses were not sampled multiple times and to prevent
detection of errors introduced by Taq polymerase (66).
As anticipated on the basis of the HMA results, all se-
quences from subject 1058 were of the same length (i.e., no
insertions or deletions were detected) (Fig. SC). The pre-SC
sequences differed by, at most, two point substitutions from the
consensus (0 to 0.3%; data not shown). The average degree of
divergence within the quasispecies collected at 57 months
post-SC was 3.9% (range, 1.8 to 6.5%). Two sequences were
defective because of the presence of multiple stop codons.
All five pre-SC sequences from subject 537 were identical,
except for one sequence with a single substitution (data not
shown). When aligned with the pre-SC consensus sequence, all
five later variants carried the same 3-bp deletion (Fig. SC).
This likely accounts for the heteroduplex first seen 7.5 months
post-SC (Fig. SA and B). Another 3-bp deletion was found in
only two of the five later variants. Nucleotide divergence levels
within the quasispecies in subject 537 collected at 54 months
post-SC displayed an average of 2.6% (range, 1.2 to 4.5%)
substitutions. No inactivating mutations were found. The dis-
tinctions noted within the sequences of later HIV quasispecies
of subjects 1058 and 537, in their patterns of mutations
(insertions-deletions versus substitutions), could therefore be
rapidly approximated by HMA.
HTA for detection of individual variants in quasispecies. A
modification of the HMA referred to as the HTA can be used
5.5 14.5 33
FIG. 5. (A) Development of HIV-1 quasispecies diversity in sub-
jects 1058 and 537 over time. Time points are shown relative to the
time point (in months) when SC was documented (point 0). (B)
Tracking representation of variants present pre-SC over time in the
evolving quasispecies. Viral DNAs amplified from the pre-SC time
point from subjects 1058 and 537 were radiolabeled and mixed with a
100-fold excess of driver DNA from each time point. Heteroduplexes
formed between probe and driver sequences were then detected after
electrophoresis and autoradiography. The slowed mobility of the
radioactive signal at the later time points for subject 1058 is due to
reduced heteroduplex mobilities and not to a gel smile artifact. Patient
1058 started combination zidovudine and dideoxyinosine chemother-
apy at 54.5 months post-SC. (C) Alignment of viral DNA sequences
from subjects 1058 and 537. The consensus sequence of the pre-SC
time point was used as a reference in each case (pre-SC variants
differed by, at most, two substitutions) for comparison to sequences
from the last time points available 57.5 months post-SC from subject
1058 and 54 months post-SC from subject 537. Identical residues in
other sequences are indicated by dots. Gaps, reflective of deletions,
were introduced to maintain alignment and are indicated by dashes.
Mutations found in all representatives of the later quasispecies are
boxed. The shaded box highlights the deletion found in a subset of
subject 537 sequences.
to determine the representation of related variants within a
quasispecies swarm (11). This is done by mixing, denaturing,
and annealing two sets of amplified gene fragments, one
radiolabeled and used as the probe and the other present in
100-fold excess and used as the driver in the reannealing
reaction. These conditions prevent the formation of homodu-
plexes derived by reannealing of complementary probe se-
quences. Following electrophoresis of the annealed products,
autoradiography of the gel permits detection of only those
heteroduplexes formed between the probe and driver se-
HIV-1 EVOLUTION IN VIVO 6677
.................... T ..= ...................T CCC.....................................
........ ....... ........................................... ........................ ........ ................. ...C.....
........... ........ ..................
.................... .... ...
.................................... a ..................................................TA
TCTAACCTTACTA TACCAACATC CAATAACAC TTTAAACAATACTTACAAAAT
.....A...... . . .
HTA was used to assess the representation of virus found
pre-SC in the evolving viral quasispecies over time (Fig. 5B).
Pre-SC samples were used as probes to determine how long
these early variants were maintained at substantial levels in
vivo. A detection limit of 2 to 5% for minority variants under
these conditions was determined by reconstitution experiments
(data not shown). Sequences closely related to the major
infecting variant were maintained as proviruses in subject
DELWART ET AL.
1058's PBMC for 54 to 57 months before showing a slight but
noticeable sequence divergence, as would be expected from
the level of substitution revealed by sequence analysis (average
divergence from the pre-SC sequence was 3.3%, with a range
of 1.8 to 5.9%).
Consistent with the pattern of deletions detected by DNA
sequence analysis, variants which formed rapidly migrating
heteroduplexes with the pre-SC probe were almost entirely
cleared from the PBMC of subject 537 by 20 months post-SC.
The initial appearance, at 7.5 months post-SC, of the deletion
mutation found in all later variants, as well as that of the
second deletion (within a subset of sequences identified by
HTA and by sequencing) at 47 months post-SC could therefore
be rapidly identified by HTA. Pairwise HMA with the se-
quenced molecular endpoints confirmed the identity of the
intermediate- and slower-mobility bands in Fig. SB (data not
shown). The average divergence of the subject 537 sequences
from the pre-SC sequence was 4.7%, with a narrow range of 4.3
to 5.1%. Subject 537 had a relatively stable CD4+ cell count of
250 to 400/mm3 through 62 months post-SC. The rate of CD4+
cell decline was faster in subject 1058, with a count of 110
CD4+ cells per mm3 by 73 months post-SC.
The representation of individual variants was also monitored
over time in the PBMC of an asymptomatic individual with a
stable CD4 cell count of 260 to 470/mm3 over the period of
analysis (subject MA) (Fig. 6A). Molecularly cloned variants
were obtained from MA PBMC at the initial time point and
after 22 months of follow-up (starting an estimated 4.5 years
after primary infection) which, by sequence analysis, showed a
broad range of representation within their quasispecies (28,
63). Three sequences were chosen for use as heteroduplex
probes for the quasispecies shown in Fig. 2A. Close relatives
(those forming heteroduplexes with mobilities close to that of
the homoduplex) of clone MA20, a rare genetic outlier at the
earliest time point sampled (month 1) (28, 63), were found in
low abundance at the time point at which clone MA20 was
derived and were not detected at later time points (Fig. 6A).
Close relatives of clone MA16, a genetically common variant
from the earliest time point (month 1), were abundant from
the time at which clone MA16 was derived through the
following 22 months and then abruptly diminished the next
month. Close relatives of clone MA305, a common variant
from one of the later time points (month 22), appeared during
the 7 to 22-month interval, were abundant at 22 months, and
were less abundant at later time points. Thus, changes in the
quantitative representation of individual variants can occur
rapidly in complex quasispecies and be easily monitored by
HTA was also used to monitor the evolution of entire
quasispecies over time by using all of the variants amplified
from one time point as probes for other time points. Radio-
actively labeled tracer amounts of amplified products from
subject MA PBMC were generated from months 1 and 27 and
used as probes with quasispecies from months 1, 6, 7, 22, 23,
and 27. In each instance, gradual replacement of the quasi-
species was observed, drifting away from that found at the time
the probes were generated (Fig. 6B). Thus, the turnover
observed through examination of individual variants (Fig. 6A)
was mirrored when comparisons between complex populations
were performed (Fig. 6B).
HIV-1 quasispecies complexity following in vitro coculture
and representation of coculture-replicating variants in vivo.
For a time, it was generally assumed that viruses replicating in
vitro faithfully reflected the in vivo virus populations from
which they were derived. However, a small number of DNA-
sequencing studies demonstrated a marked reduction in the
7 2223 27
7 22 23 27
6 7 2223 27 16 7 222327
FIG. 6. (A) Tracking representation of individual variant HIV
genomes within subject MA PBMC over time. PCR-amplified products
derived from plasmid templates were radiolabeled (11), mixed, heated,
and reannealed with a 100-fold excess of the product from subject MA
PBMC. Individual clone numbers for probes are indicated along with
the month of origin (in parentheses [Ml or M22]). The months of
origin of driver sequences are indicated. Lane C contained probe
sequences alone. Autoradiographs of these products separated on 5%
polyacrylamide gels are shown. The asterisk to the left shows the
position of single-stranded DNA, and the letter H indicates the
position of homoduplexes. (B) Quasispecies evolution in PBMC over
time within an infected individual. PCR products from the subject MA
PBMC samples shown in Fig. 2A for months 1 (Ml) and 27 (M27)
were used as probes along with an excess of the unlabeled product
(driver) from the time points shown.
diversity of HIV-1 env genes, relative to that of the proviral
population in uncultured PBMC, following virus amplification
by in vitro coculture with uninfected PBMC (27, 28, 38, 63).
We therefore used HMA and HTA to examine the generaliz-
ability of diversity reduction and also to relate the identity of
cultured viruses to provirus turnover in vivo.
In a previous study, the cocultured virus population was
found to be highly homogeneous by mismatch divergence
alone and to consist of two main variant populations differing
by a single 3-bp gap (28). Results ofpairwise HMA with cloned
virus templates representing these two subpopulations are
shown in Fig. 1B (gap + lanes). Furthermore, as predicted by
DNA sequence analysis, the same two populations were readily
recognizable in the simple heteroduplex mobility pattern de-
rived from amplification of viral DNA from the bulk coculture
(Fig. 2A, lane CC).
The ability of HMA to visualize the relationship between
large numbers of molecules simultaneously was next employed
HIV-1 EVOLUTION IN VIVO
Probe: MA214 (Ml)
FIG. 7. (A) HIV env DNA sequences present after 3 to 6 weeks of
coculture. Twenty-two independent PBMC cocultures were examined.
More than 200 template molecules of viral DNA were present in each
PCR amplification (see Materials and Methods) and the products
examined following electrophoresis in a 5% polyacrylamide gel. The
contents of the rightmost lane were identical to those of lane M in Fig.
1. (B) Tracking representation of cocultured virus in PBMC over time.
PCR-amplified products derived from molecular clones from cocul-
tured virus were radiolabeled and annealed with PBMC amplification
products from each time point. For descriptions of the lane contents,
see the legend to Fig. 6.
to determine whether a reduction in genetic complexity fol-
addition to the comparison shown in Fig. 2A (the cocultured
virus population in lane CC was derived from the PBMC
sample in lanes 1), the diversity of HIV-1 env genes was
examined in 22 cocultures with PBMC from different infected
individuals (Fig. 7A). In most, but not all, of these cocultures,
maintained for 3 to 6 weeks prior to analysis, a low level of
diversity was observed. A single predominant variant was
observed in 10 instances, two predominant variants were
observed in 8, and more predominantvariants were detected in
the remaining 4. In most cases, minor variants were evident as
To characterize the basis for the strong bias that occurs in
most cultures for replication of certain viruses, the represen-
tation of individual coculture-derived variants was tracked in
PBMC provirus populations from which they were derived.
Three clones representing dominant genotypes in a culture
from subject MA (28, 63) were used in heteroduplex tracking
experiments to assess their representation in vivo. As can be
seen in Fig. 7B, close relatives of clones MA214 and MAC06
were not of sufficient abundance in the PBMC to be detected
at the time point from which they were derived but became
abundant at subsequent time points. Close relatives of clone
is indeed a general phenomenon. In
MAC08 were detected in the PBMC from the time point of
coculture and became more abundant at later time points.
Detection and analysis of DNA sequence polymorphisms.
Several methods have been used to evaluate HIV-1 sequence
variability and identify genetic variants without resort to
large-scale DNA sequencing. These include methods based on
restriction fragment length polymorphism of molecularly
cloned genomes (19); DNA fragment length polymorphism on
denaturing sequencing gels of PCR-amplified, single variable
regions ofenv genes (PCR fragment length polymorphism; 65);
RNase A mismatch cleavage analysis of RNA-RNA heterodu-
plexes derived from culture-amplified virus (34); and anchored
primer PCR typing of culture amplified virus (35). Primer
mismatch-sensitive PCR has also been used to identify specific
mutations conferring antiviral drug resistance (30). Single-
strand conformational polymorphism analysis has been used to
localize mutations arising over short regions of the env gene
during adaptation of molecular-clone-derived HIV-1 isolates
to particular cell lines (17) and in simian immunodeficiency
virus-infected primates (23). Denaturing gradient gel electro-
phoresis (42, 43) has been used to estimate the level of
sequence diversity and to isolate sequence variants from
different quasispecies (2). Single-strand conformational poly-
morphism analysis, RNase A mismatch, and denaturing gradi-
ent gel electrophoresis, while extremely sensitive methods for
detecting single nucleotide differences, are considerably more
laborious than HMA. Except for PCR fragment length poly-
morphism and denaturing gradient gel electrophoresis, these
techniques do not easily allow simultaneous analysis of multi-
ple sequence variants and include the laborious and selective
processes of virus cocultivation or molecular cloning prior to
analysis. In contrast to HMA, these methods also do not
provide an easily derived estimate of the extent of genetic
divergence between sequences (11) or, as shown here, repre-
sentation of particular variants in complex gene pools.
Electrophoretic analysis of DNA heteroduplexes has been
used in the field of human genetics for detection of nucleotide
insertion and deletion polymorphisms (7, 13, 51, 56, 71) and
for determination ofHLA allotypes (9). The ability ofHMA to
detect small gaps in both small and large (3-kb) heteroduplexes
may be useful for detection of DNA polymorphisms in genetic
mapping studies. The base-specific effects that the unpaired
nucleotides have on the electrophoretic mobility of gapped
heteroduplexes (70; data not shown) could also be adapted to
detect single-base substitutions in small, pretargeted regions of
interest, such as those activating proto-oncogenes or drug
resistance genes. The ability to detect low levels of sequence
variants by HTA with a labeled DNA probe may be useful in
the detection of rare genetic polymorphisms in clinical sam-
ples, such as frameshift mutations in tumor suppressor genes
(53). The ability of PCR to amplify variant templates in direct
proportion to their fractional representation strengthens the
usefulness of the latter technique (58). The existence of
divergent sequence variants in PCR products can be estab-
lished through the formation of slowly migrating heterodu-
plexes, which can, in principle, be directly isolated by excision
from the gel, reamplified, and cloned. Individual variants can
also be isolated by endpoint dilution of infected DNApriorto
N-PCR amplification (66), and the products can be used to
assess their representation in the sequencemixture examined,
as shown here. Such approaches may be useful intestingfor
multiple lentiviral infections in animal model systems or hu-
mans repeatedly exposed to HIV.
VOL. 68, 1994
DELWART ET AL.
N-PCRamplificationof HIV-1quasispeciesunder routine
conditions(2 x 35amplification cycles)resulted in heterodu-
plexformation. In other studies, we found that when such
heteroduplexeswereligated inplasmidvectors and trans-
formed into Escherichia coli, theysubcloned asefficiently as
homoduplexesbut that 1 of 30 subclonesanalyzedwas a
recombinant, presumablycausedbystrandswitching during
heteroduplexDNArepairin E. coli(lOa). Such E. coli-
mediatedevents, alongwith nucleotidemisincorporationand
recombination by extension ofincompletely copied DNA
strandsduringPCR(39),are therefore an additionalpossible
source oftemplatealteration associated with PCR.
Virus turnover in vivo and selection of viral genotype
variants in culture. The turnover of individual HIVvariants,
thevaryinglevel ofquasispecies diversity,and therelationship
HMA and HTA. Theseassayswereemployedto assess the
apparentselection for virus variants in cell culture. We found
that PBMC coculture mostoften,but notalways,resulted in
theoutgrowthof a limited number ofdistinguishable variants,
whilequasispeciesin vivo weregenerallymorecomplex,con-
firming previousstudies based on DNAsequence analysisof a
limited number of isolates(27, 28, 38, 63).The abundant
replicationof a limited number of variants in vitromayreflect
theirhigherfitness forgrowthunder tissue culture conditions
relative to the other variants found in vivo (10, 20, 46).
However, coculture-derived variantsequenceswere often de-
tected athigherlevels in vivosubsequentto the time ofculture,
indicating that,at least in somecases,coculture revealednewly
emergingvirus variantspresentwithin the blood ofthe infected
individual that areonlylater fixed in abundance asproviruses
insurviving targetcells. Itmay be, therefore, that coculture
isolates are derived from virions from otherbody compart-
ments adhered to the PBMC used to initiate thecocultures,
which arecontemporarily underrepresentedas PBMCprovi-
ruses(67). Alternatively, they mayderive from a fewlytically
infected cells within the PBMC.Replication-competentneu-
tralizationescapevariants would also beexpectedto increase
theirrepresentationin PBMC DNA over time in vivo. The
diversityofproviral envelope genesin PBMCmay therefore
represent,inpart,a record ofpriorinfection insurviving cells,
perhaps due to the retention of a number of replication-
Viruspopulation turnover and apossiblelink to disease
progression. Consistent with recentDNA-sequencing studies
(37, 49, 72, 75, 76),little or noapparent envelope sequence
diversitywas detected withinquasispecies soon after sexual
transmission. The same observation has been made for moth-
er-to-infant transmission and infectionby exposure to HIV-
infected bloodproducts (74, 75).Thismayindicate that a viral
populationbottleneck occurred and asingleinfectious unit was
transferred.Alternatively,thehigherlevels of mutation(most-
ly synonymous changes)inGag p17matrixsequences relative
to those seen in theenvelopeinprimaryinfection(75, 76) may
reflect the transmission ofmultiple sequence variants, followed
byselectiveamplificationof those fewenvelopevariants suited
forreplicationin anewlyinfectedhost, possibly as a result of
theirmacrophage tropism (8, 37, 64, 75, 76). The envelope
sequencevariation that was detected in somesubjects at SC or
soonpost-SC (Fig. 4) mayreflect eitherrapid de novo gener-
ation ofenvelope sequence diversityor infection by multiple
Thehigh percentageofnonsynonymous substitutions per
mutated site seen in vivofollowing primaryinfection (86% for
bothsubjects537 and1058)indicatespossibleselection for the
generationofphenotypic diversity.Ahighratio ofnonsynony-
mous-to-synonymous mutations was similarly observed in se-
rial samples from rhesus monkeys inoculated with molecular-
clone-derived simian immunodeficiency virus (6, 24, 47, 48).
Because of the noted ability of HIV to adapt rapidly to
selective pressures (1, 3, 31, 44, 54), different patterns of
quasispecies evolution in vivo may reflect differences in im-
mune response efficiency between individuals (15, 25, 33, 57,
61, 62), which may, in turn, affect the rate of CD4+ cell
depletion and disease progression (29, 52, 59).Thepersistence
of highly related viral genotypes exhibiting little sequence
change could reflect a poor ordecliningimmuneresponsewith
little selective force being applied to the residentquasispecies
(73). Likewise, rapidly changing quasispecies may reflect an
efficient immune response. An alternative hypothesis, that
surpassing a high level of antigenic diversity is the cause of
CD4+ cell decline, has been put forward (45). However, the
high level of genetic variation observed in long-term-asymp-
tomatic individuals and the low level of genetic variation in
those who progress very rapidly are not necessarily consistent
with this model. Furthermore, the low levels of viralsequence
diversity observed in the PBMC of somelong-term-infected
AIDS patients (Fig. 3) may be due to the emergence ofrapidly
replicating syncytium-inducing variants, reported following in
vitro culture of PBMC from approximately half of the individ-
uals progressing to AIDS (18, 60, 68). Longitudinal analysis of
the entire course of infection of individuals with different rates
of disease progression is required to test these hypotheses
The extent and pattern of HIV sequence diversification
following primary infection varied in two individuals studiedby
both DNA sequence analysis and HMA. The loweraverage
rate of accumulated nucleotide substitutions in HIV env in
subject 1058 versus that of subject 537 (0.7 versus 1%/year),
the appearance of syncytium-inducing virus and the faster rate
ofCD4+ cell decline in subject 1058 suggest an environment in
which natural selection pressures on the virus were weaker
than in subject 537. A more effective immune response, with a
consequently stronger selective pressure for escape mutants,
may be responsible for the larger number of fixed mutational
changes found in subject 537. When aligned with theirprogen-
itor pre-SC sequences, the same 15 substitutions and a codon
deletion were seen in every genome sampled from the later
time point in subject 537. In contrast, only four substitutions
were found fixed in all later genomes from subject 1058 (boxed
in Fig. SC). While only five envelope genes of the later subject
537 quasispecies were sequenced, the absence of radioactive
heteroduplexes with homoduplex-like mobility seen upon
probing of the latest quasispecies with the pre-SC sequence
also indicates substantial clearance of the pre-SC variant from
Individual sequence variants within asymptomatic subject
MA also appeared to turnover rapidly 5 to 7 years after
infection (Fig. 6A). This turnover could result from repopula-
tion of the quasispecies by descendants of a few genetic
variants with selective advantages (e.g., a neutralization escape
variant) such that all members of the later quasispecies carry
both advantageous and genetically linked mutations. The
higher number of fixed mutations in the later subject 537 viral
pool may indicate the more frequent emergence and fixation of
selectively advantageous mutations than in subject 1058 as a
result of stronger selective pressures.
In summary, our data are consistent with the hypothesis that
individuals capable of mounting a strong immune response
against the virus drive the successive evolution ofmultiple virus
variants, while those individuals showing a rapid CD4+ cell
decline with progression to AIDS mount or maintain a weaker
HIV-1 EVOLUTION IN VIVO
immune response against contemporary HIV variants or are
confronted with a more virulent strain of the virus, with a
resulting lower level of selection-driven changes in the virus
pool. These initial observations require follow-up with longi-
tudinal analyses of more individuals over the entire course of
HIV infection to determine whether the links among increased
virus diversification, quasispecies turnover, and slow progres-
sion are consistent and to evaluate the possibility that a drop in
quasispecies complexity occurs in some subjects during the late
stages of HIV disease. If these hypothesized links can be
substantiated, then HMA may be useful for monitoring the
effects of therapeutic intervention strategies and for providing
a prognostic indication of the length of survival.
We thank Eugene G. Shpaer for assistance with sequence analysis
and Michael V. Gallo and Wilco Keulen for DNA sequencing, Martin
S. Hirsch for ongoing collaboration in the analysis of subject MA, and
Francine McCutchan aiid Mark Holodniy for some of the virus
samples used in this study. We also thank subject MA, volunteers in
the San Francisco Men's Health Study, and the Amsterdam cohort for
consistent participation in these studies and J. Arthos, D. Sodora, C.
de Noronha, and R. McClelland for helpful discussions.
This work was supported by Public Health Service award R01
A132885 and grants from the Pediatric AIDS Foundation (500170-11-
PG) and the Stanford Program in Molecular and Genetic Medicine.
E.L.D. was supported during the early stages of this study by NIH
NRSA AI07328 and by a grant from the Center for AIDS Research at
Stanford (NO1 A182515).
1. Albert, J., B. Abrahamsson, K. Nagy, E. Aurelius, H. Gaines, G.
Nystrom, and E. M. Fenyo. 1990. Rapid development of isolate-
specific neutralizing antibodies after primary HIV-1 infection and
consequent emergence of virus variants which resist neutralization
by autologous sera. AIDS 4:107-112.
2. Andersson, B., J.-H. Ying, D. E. Lewis, and R. A. Gibbs. 1993.
Rapid characterization of HIV-1 sequence diversity using dena-
turing gradient gel electrophoresis and direct automated DNA
sequencing of PCR products. PCR Methods Appl. 22:293-300.
3. Arendrup, M., C. Nielsen, J.-E. S. Hansen, C. Pedersen, L.
Mathiesen, and J. 0. Nielsen. 1992. Autologous HIV-1 neutraliz-
ing antibodies: emergence of neutralization-resistant escape virus
and subsequent development of escape virus neutralizing antibod-
ies. J. Acquired Immune Defic. Syndr. 5:303-307.
3a.Bex, F., W. Keulen, J. Goudsmit, and J. I. Mullins. Unpublished
4. Bhattacharyya, A., and D. M. J. Lilley. 1989. The contrasting
structures of mismatched DNA sequences containing looped-out
bases (bulges) and multiple mismatches (bubbles). Nucleic Acids
5. Brown, L. A., and P. Monaghan. 1988. Evolution of the structural
proteins of human immunodeficiency virus: selective constraints
on nucleotide substitution. AIDS Res. Hum. Retroviruses 4:399-
6. Burns, D. P. W., and R. C. Desrosiers. 1991. Selection of genetic
variants of simian immunodeficiency virus in persistently infected
rhesus monkeys. J. Virol. 65:1843-1854.
7. Cai, S.-P., B. Eng, Y. W. Kan, and D. H. K. Chui. 1991. A rapid and
simple electrophoretic method for the detection of mutations
involving small insertion and deletion: application to P-thalasse-
mia. Hum. Genet. 87:728-730.
8. Chesebro, B., K. Wehrly, J. Nishio, and S. Perryman. 1992.
Macrophage-tropic human immunodeficiency virus isolates from
different patients exhibit unusual V3 envelope sequence homoge-
neity in comparison with T-cell-tropic isolates: definition of critical
amino acids involved in cell tropism. J. Virol. 66:6547-6554.
9. Clay, T. M., J. L. Bidwell, M. R. Howard, and B. A. Bradley. 1991.
PCR-fingerprinting for selection of HLA matched unrelated mar-
row donors. Lancet 337:1049-1052.
10. Coffin, J. M. 1992. Genetic diversity and evolution of retroviruses.
Curr. Top. Microbiol. Immunol. 176:143-164.
10a.Delwart, E. L., R. Oh, and J. L. Mullins. Unpublished data.
11. Delwart, E. L., E. G. Shpaer, F. E. McCutchan, J. Louwagie, M.
Grez, H. Rubsamen-Waigmann, and J.I. Mullins. 1993. Genetic
relationships determined by a heteroduplex mobility assay: analy-
sis of HIV env genes. Science 262:1257-1261.
12. Dougherty, J. P., and H. M. Temin. 1991. Determination of
retroviral mutation rates using spleen necrosis virus-based vectors
and helper cells. Bio/Technology 16:21-33.
13. Farrar, G. J., P. Kenna, S. A. Jordan, R. Kumar-Singh, M. M.
Humphries, E. M. Sharp, D. M. Sheils, and P. Humphries. 1991.
A three-base-pair deletion in the peripherin-RDS gene in one
form of retinitis pigmentosa. Nature (London) 354:478-480.
14. Faulkner, D. V., and J. Jurka. 1988. Multiple aligned sequence
editor (MASE). Trends Biochem. Sci. 13:321-322.
15. Fenouillet, E., N. Blanes, A. Coutellier, J. Demarquest, W. Rozen-
baum, and J.-C. Gluckman.
against human immunodeficiency virus type 1 p25 core protein as
prognostic marker. J. Infect. Dis. 166:611-616.
16. Fouchier, R. A. M., M. Groenink, N. A. Kootstra, M. Tersmette,
H. G. Huisman, F. Miedema, and H. Schuitemaker. 1992. Pheno-
type-associated sequence variation in the third variable domain
(V3) of the human immunodeficiency virus type 1 gpl20 molecule.
J. Virol. 66:3183-3187.
17. Fujita, K., J. Silver, and K. Peden. 1992. Changes in both gpl20
can account for increased growth potential and ex-
panded host range of human immunodeficiency virus type
18. Groenink, M., R. Fouchier, S. Broersen, C. Baker, M. Koot, A.
Van't Wout, H. Huisman, F. Miedema, M. Tersmette, and H.
Schuitemaker. 1993. Relation of phenotype evolution of HIV-1 to
envelope V2 configuration. Science 260:1513-1516.
19. Hahn, B. H., G. M. Shaw, M. E. Taylor, R. R. Redfield, P. D.
Markham, S. Z. Salahuddin, F. Wong-Staal, R. C. Gallo, E. S.
Parks, and W. P. Parks. 1986. Genetic variation in HTLV-III/
LAV over time in patients with AIDS or at risk for AIDS. Science
20. Hirsch, V. M., P. Edmonson, M. Murphey-Corb, B. Arbielle, P. R.
Johnson, and J. I. Mullins. 1989. SIV adaptation to human cells.
Nature (London) 341:572-574.
21. Ho, D. D., T. R. Rota, R. T. Schooley, J. C. Kaplan, J. D. Allan,
J. E. Groopman, L. Resnick, D. Felsenstein, C. A. Andrews, and
M. S. Hirsch. 1985. Isolation of HTLV-III from cerebrospinal
fluid and neural tissues of patients with neurologic syndromes
related to the acquired immunodeficiency syndrome. N. Engl. J.
22. Ho, D. D., M. G. Sarngadharan, L. Resnick, F. Dimarzoveronese,
T. R. Rota, and M. S. Hirsch. 1985. Primary human T-lympho-
tropic virus type III infection. Anal. Intern. Med. 103:880-883.
23. Hynes, N. A., D. Adger-Johnson, G. Dapolito, and V. M. Hirsch.
1993. Rapid screening for simian immunodeficiency virus variants
using single-strand conformational polymorphism of PCR-ampli-
fied DNA fragments. AIDS Res. Hum. Retroviruses 9:803-806.
24. Johnson, P. R, T. E. Hamm, S. Goldstein, S. Kitov, and V. M.
Hirsch. 1991. The genetic fate of molecularly cloned simian
immunodeficiency virus in experimentally infected macaques. Vi-
25. Keet, L. P. M., P. Krinen, M. Koot, J. M. A. Lange, F. Miedema,
J. Goudsmit, and R A. Coutinho.
progression to AIDS in HIV-1 seroconverters. AIDS 7:51-57.
26. Kuiken, C., G. Zwart, E. Baan,RCoutinho, J. van den Hoek, and
J. Goudsmit. 1993. Increasing antigenic and genetic diversity of
the HIV-1 V3 domain in the course of the AIDS epidemic. Proc.
Natl. Acad. Sci. USA 90:9061-9065.
27. Kuiken, C. L., J.-J. Dejong, E. Baan, W. Keulen, M. Tersmette,
and J. Goudsmit. 1992. Evolution of the V3 envelope domain in
proviral sequences and isolates of human immunodeficiency virus
type 1 during transition of the viral biological phenotype. J. Virol.
28. Kusumi, K., B. Conway, S. Cunningham, A. Berson, C. Evans,
A. K. N. Iversen, D. Colvin, M. V. Gallo, S. Coutre, E. G. Shpaer,
D. V. Faulkner, A. DeRonde, S. Volkman, C. Williams, M. S.
Hirsch, and J. I. Mullins. 1992. Human immunodeficiency virus
1992. Monitoring of antibodies
1992. Predictors of rapid
VOL. 68, 1994
DELWART ET AL.
type1envelope gene structure and diversityin vivo and after
cocultivation in vitro. J. Virol. 66:875-885.
29.Lang, W.,H.Perkins,R. E.Anderson,N.Jewell,and W. Winkel-
stein. 1989. Patterns ofT-lymphocyte changeswith human immu-
nodeficiencyvirus infection: from seroconversion to thedevelop-
ment of AIDS. J.AcquiredImmune Defic.Syndr.2:63-69.
30. Larder, B. A., P. Kellam, and S. D. Kemp. 1991. Zidovudine
resistancepredicted bydirect detection of mutations in DNA from
31. Larder,B.A.,and S. D.Kemp.1989.Multiplemutations in HIV-1
reverse transcriptase confer high-level resistance to zidovudine
32. Li, Y., J. C.Kappes, J. A.Conway,R. W.Price,G. M.Shaw,and
B. H. Hahn. 1991. Molecular characterization of human immuno-
deficiencyvirus type 1 cloned directly from uncultured human
brain tissue: identification ofreplication-competentand defective
viralgenomes.J. Virol. 65:3973-3985.
33. Lifson,A.R.,S. P.Buchbinder,H. W.Sheppard,A. C.Mawle, J. C.
Wilber,M.Stanley,C. E.Hart,N. A.Hessol,and S. D.Holmberg.
1991. Long term human immunodeficiency virus infection in
asymptomatichomosexual and bisexual men with normallympho-
cytecounts:immunologicandvirologiccharacteristics. J. Infect.
34.Lopez-Galindez, C., J. M.Rojas,R.Najera,D. D.Richman, and
M. Perucho. 1991. Characterization of genetic variation and
3'-azido-3'-deoxythymidine-resistancemutations of human immu-
nodeficiencyvirusbythe RNase A mismatchcleavage method.
Proc. Natl. Acad. Sci. USA 88:4280-4284.
35. McCutchan,F.E.,E.Sanders-Buell,C. W.Oster,R. R.Redfield,
S. K.Hira, P. L. Perine, B. L.Ungar,and D. S. Burke. 1991.
isolatesby polymerasechain reaction. J.AcquiredImmune Defic.
36.McKeating,J. A., J. Gow, J. Goudsmit,L. H.Pearl,C.Mulder,and
R. A. Weiss. 1989. Characterization of HIV-1 neutralization
escapemutants. AIDS 3:777-784.
37. McNearney, T., Z. Hornickova, R. Markham, A. Birdwell, M.
Arens, A. Saah, and L. Ratner. 1992. Relationship of human
disease. Proc. Natl. Acad. Sci. USA 89:10247-10251.
38. Meyerhans, A., R. Cheynier, J. Albert, M. Seth, S. Kwok, J.
Sninsky,L.Morfeldt-Manson,B.Asjo,and S. Wain-Hobson. 1989.
Temporalfluctuations in HIVquasispecies in vivo are not re-
flectedby sequentialHIV isolations. Cell 58:901-910.
39.Meyerhans, A., J.-P. Vartanian,and S. Wain-Hobson. 1990. DNA
recombinationduringPCR. Nucleic Acids Res. 18:1687-1691.
40. Mullis,K.B.,and F. A. Faloona. 1987.Specific synthesisof DNA
in vitro via apolymerase-catalyzedchain reaction. Methods En-
41.Myers, G., B.Korber, S.Wain-Hobson, R. F. Smith, and G. N.
Paviakis. 1993. Human retroviruses and AIDS 1993. Acompila-
tion andanalysisof nucleic acid and amino acidsequences. Los
Alamos NationalLaboratory,LosAlamos, N. Mex.
42.Myers,R.M.,S. G.Fischer,L. S.Lerman,and T. Maniatis. 1985.
Nearlyallsinglebase substitutions in DNAfragments joined to a
GC-clampcan be detectedby denaturing gradient gel electro-
phoresis.Nucleic Acids Res. 13:3131-3145.
43.Myers,R.M.,S. G.Fischer,T.Maniatis,and L. S. Lerman. 1985.
Modification of themelting properties ofduplex DNAby attach-
ment of a GC-rich DNAsequence as determined by denaturing
gradient gel electrophoresis. Nucleic Acids Res. 13:3111-3128.
44. Nara,P.L.,L.Smit,N.Dunlop,W.Hatch, M. Merges, D. Waters,
J. Kelliher,R. C.Gallo,P. J. Fischinger, and J. Goudsmit. 1990.
Emergence of viruses resistant to neutralization by V3-specific
antibodies inexperimentalhumanimmunodeficiency virus type 1
IIIB infection ofchimpanzees. J. Virol. 64:3779-3791.
45. Nowak, M.A., R. M. Anderson, A. R. McLean, T. F. Wolfs, J.
Goudsmit, and R. M.May. 1991. Antigenic diversity thresholds
and thedevelopmentof AIDS. Science 254:963-969.
46.Overbaugh, J.,P. R.Donahue, S. L. Quackenbush, E. A. Hoover,
and J. I. Mullins. 1988. Molecular cloning of a feline leukemia
virus that induces fatalimmunodeficiency disease in cats. Science
47. Overbaugh, J., and L. M. Rudensey. 1992. Alterations in potential
sites for glycosylation predominate during evolution of the simian
immunodeficiency virus envelope gene in macaques. J. Virol. 66:
48. Overbaugh, J., L. M. Rudensey, M. D. Papenhausen, R. E.
Benveniste, and W. R. Morton. 1991. Variation in simian immu-
nodeficiency virus env is confined to Vl and V4 during progression
to simian AIDS. J. Virol. 65:7025-7031.
49. Pang, S., Y. Shiesinger, E. S. Daar, T. Moudgil, D. D. Ho, and
I. S. Y. Chen. 1992. Rapid generation of sequence variationduring
primary HIV-1 infection. AIDS 6:453-460.
50. Pathak, V. K., and H. M. Temin. 1990. Broad spectrum of in vivo
forward mutations, hypermutations, and mutational hot spots in a
retroviral shuttle vector after a single replication cycle: substitu-
tions, frameshifts, and hypermutations. Proc. Natl. Acad. Sci. USA
51. Paw, B. H., P. T. Tieu, M. M. Kaback, J. Lim, and E. Neufeld.
1990. Frequency of three Hex A mutant alleles among Jewish and
non-Jewish carriers identified in a Tay-Sachs screening program.
Am. J. Hum. Genet. 47:698-705.
52. Phillips, A. N., C. A. Lee, J. Elford, G. Janossy, A. Timms, M.
Bofill, and P. B. A. Kernoff. 1991. Serial CD4 lymphocyte counts
and the development of AIDS. Lancet 337:389-392.
53. Powell, S. M., N. Zilz, Y. Beazer-Barclay, T. M. Bryan, S. R.
Hamilton, S. N. Thibodeau, B. Vogelstein, and K. W. Kinzler.
1992. APC mutations occur early during colorectal tumorigenesis.
Nature (London) 359:235-237.
54. Reitz, M. S., C. Wilson, C. Naugle, R C. Gallo, and M. Robert-
Guroff. 1988. Generation of a neutralization resistant variant of
HIV-1 is due to selection for a point mutation in the envelope
gene. Cell 54:57-63.
55. Roberts, J. D., K. Bebenek, and T. A. Kunkel. 1988. The accuracy
of reverse transcriptase from HIV-1. Science 242:1171-1173.
56. Rommens, J., B.-S. Kerem, W. Greer, P. Chang, L.-C. Tsui, and P.
Ray. 1990. Rapid nonradioactive detection of the major cystic
fibrosis mutation. Am. J. Hum. Genet. 46:395-396.
57. Roos, M., J. Lange, R. de Goede, R. Coutinho, P. Schellekens, F.
Miedema, and M. Tersmette. 1992. Viral phenotype and immune
response in primary human immunodeficiency virus type 1 infec-
tion. J. Infect. Dis. 165:427-432.
58. Ruano, G., and K. K. Kidd. 1992. Modeling of heteroduplex
formation during PCR from mixtures of DNA templates. PCR
Methods Appl. 2:112-116.
59. Schellekens, P. T. A., M. Tersmette, M. T. L. Roos, R. P. Keet, F.
deWolf, R. A. Coutinho, and F. Miedema. 1992. Biphasic rate of
CD4+ cell count decline during progression to AIDS correlates
with HIV-1 phenotype. AIDS 6:665-669.
60. Schuitemaker, H., M. Koot, N. A. Kootstra, M. W. Dercksen,
R. E. Y. de Goede, R. P. van Steenwik, J. M. A. Lange, J. K. M. E.
Schattenkerk, F. Miedema, and M. Tersmette. 1992. Biological
phenotype of human immunodeficiency virus type 1 clones at
different stages of infection: progression of disease is associated
with a shift from monocytotropic to T-cell-tropic virus popula-
tions. J. Virol. 66:1354-1360.
61. Sheppard, H. W., M. S. Ascher, B. McRae, R. E. Anderson, W.
Lang, and J. Allain. 1991. The initial immune response to HIV
and immune system activation determine the outcome of HIV
disease. J. Acquired Immune Defic. Syndr. 4:704-712.
62. Sheppard, H. W., W. Lang, M. S. Ascher, E. Vittinghof, and W.
Winkelstein. 1993. The characterization of non-progressors: long
term HIV-1 infection with stable CD4+ T-cell levels. AIDS 7:
63. Shpaer, E. G., E. D. Delwart, A. K. N. Iversen, M. V. Gallo, R.
Oh, A. Berson, J. Brojatsch, M. S. Hirsch, B. D. Walker, and
J. I. Mullins. Selection, gene conversion and successive evolu-
tion of the HIV-1 envelope gene in vivo. Submitted for publica-
64. Shpaer, E. G., C. L. Kuiken, J. Goudsmit, M. Bachmann, E. L.
Delwart, and J. I. Mullins. Conserved V3 loop sequences and the
transmission of human immunodeficiency virus. Submitted for
64a.Shpaer, E. G., and J. I. Mullins. Unpublished data.
65. Simmonds, P., P. Balfe, C. A. Ludlam, J. 0. Bishop, and A. J. L.
HIV-1 EVOLUTION IN VIVO
Brown. 1990. Analysis of sequence diversity in hypervariable
regions of the external glycoprotein of human immunodeficiency
virus type 1. J. Virol. 64:5840-5850.
66. Simmonds, P., P. Balfe, J. F. Peutherer, C. A. Ludlam, J. 0.
Bishop, and A. J. Leigh-Brown. 1990. Human immunodeficiency
virus-infected individuals contain provirus in small numbers of
peripheral mononuclear cells and at low copy numbers. J. Virol.
67. Simmonds, P., L. Q. Zhang, F. McOmish, P. Balfe, C. A. Ludlam,
and A. J. L. Brown. 1991. Discontinuous sequence change of
human immunodeficiency virus (HIV) type 1 env sequences in
plasma viral and lymphocyte-associated proviral populations in
vivo: implications for models of HIV pathogenesis. J. Virol. 65:
67a.Tersmette, M. Personal communication.
68. Tersmette, M., R. A. Gruters, F. de Wolf, R. E. Y. de Goede,
J. M. A. Lange, P. T. A. Schellekens, J. Goudsmit, H. G. Huisman,
and F. Miedema. 1989. Evidence for a role of virulent human
immunodeficiency virus (HIV) variants in the pathogenesis of
acquired immunodeficiency syndrome: studies on sequential HIV
isolates. J. Virol. 63:2118-2125.
69. Wain-Hobson, S. 1993. Viral burden and AIDS. Nature (London)
70. Wang, Y.-H., and J. Griffith. 1991. Effects of bulge composition
and flanking sequence on the kinking of DNA by bulged bases.
71. White, M. B., M. Carvalho, D. Derse, S. J. O'Brien, and M. Dean.
1992. Detecting single base substitutions as heteroduplex polymor-
phisms. Genomics 12:301-306.
72. Wolfs, T. F., G. Zwart, M. Bakker, and J. Goudsmit. 1992. HIV-1
genomic RNA diversification following sexual and parenteral virus
transmission. Virology 189:103-110.
73. Wolfs, T. F. W., J. J. de Jong, H. van den Berg, J. M. G. H.
TQnagel,W. J. A. Krone, and J. Goudsmit. 1990. Evolution of
sequences encoding the principal neutralization epitope of HIV-1
is host-dependent, rapid and continuous. Proc. Natl. Acad. Sci.
74. Wolinsky, S. M., C. M. Wike, B. T. M. Korber, C. Hutto, W. P.
Parks, L L Rosenblum, K. J. Kunstmann, M. R. Furtado, and J. L
Munoz. 1992. Selective transmission of human immunodeficiency
virus type-1 variants from mother to infants. Science 255:1134-1137.
75. Zhang, L. Q., P. MacKenzie, A. Cleland, E. C. Holmes, A. J. Leigh
Brown, and P. Simmonds. 1993. Selection for specific sequences in
the external envelope protein of human immunodeficiency virus
type 1 upon primary infection. J. Virol. 67:3345-3356.
76. Zhu, T., H. Mo, N. Wang, D. S. Nam, Y. Cao, R. A. Koup, and D. D.
Ho. 1993. Genotypic and phenotypic characterization of HIV-1 in
patients with primary infection. Science 261:1179-1181.
VOL. 68, 1994