JOURNAL OF VIROLOGY, Feb. 2004, p. 1160–1168
Vol. 78, No. 3
T-Cell Subsets That Harbor Human Immunodeficiency Virus (HIV) In
Vivo: Implications for HIV Pathogenesis
Jason M. Brenchley,1Brenna J. Hill,1David R. Ambrozak,2David A. Price,1Francisco J. Guenaga,1
Joseph P. Casazza,2Janaki Kuruppu,2Javaidia Yazdani,3Stephen A. Migueles,4Mark Connors,4
Mario Roederer,5Daniel C. Douek,1and Richard A. Koup2*
Human Immunology Section,1Immunology Laboratory,2and Immunotechnology Section,5Vaccine Research Center, and
Laboratory of Immunoregulation,4National Institute of Allergy and Infectious Diseases, National Institutes of Health,
Bethesda, Maryland 20892, and Department of Internal Medicine, University of Texas Southwestern
Medical Center, Dallas, Texas 753903
Received 5 August 2003/Accepted 7 October 2003
Identification of T-cell subsets that are infected in vivo is essential to understanding the pathogenesis of
human immunodeficiency virus (HIV) disease; however, this goal has been beset with technical challenges.
Here, we used polychromatic flow cytometry to sort multiple T-cell subsets to 99.8% purity, followed by
quantitative PCR to quantify HIV gag DNA directly ex vivo. We show that resting memory CD4?T cells are the
predominantly infected cells but that terminally differentiated memory CD4?T cells contain 10-fold fewer
copies of HIV DNA. Memory CD8?T cells can also be infected upon upregulation of CD4; however, this is
infrequent and HIV-specific CD8?T cells are not infected preferentially. Naïve CD4?T-cell infection is rare
and principally confined to those peripheral T cells that have proliferated. Furthermore, the virus is essentially
absent from naïve CD8?T cells, suggesting that the thymus is not a major source of HIV-infected T cells in
the periphery. These data illuminate the underlying mechanisms that distort T-cell homeostasis in HIV
Although human immunodeficiency virus type 1 (HIV-1)
infection is characterized by the progressive depletion of
CD4?T cells, the mechanisms underlying this remain contro-
versial (18, 22). Furthermore, the profound qualitative changes
in both CD4?and CD8?T cells that are observed in HIV-
infected individuals are still not well understood (1, 6, 24, 32,
43, 46). It has been estimated that the frequency of CD4?
T-cell infection by HIV in vivo is probably too low to account
alone for the death and dysfunction of T cells throughout the
disease (5, 17). CD4?T-cell loss is likely effected by factors
including increased memory T-cell turnover, damage to the
thymus and other lymphoid tissues, limitations in CD4?T-cell
renewal, and the direct destructive effects of the virus on sus-
ceptible T-cell pools.
Developmental and homeostatic relationships between var-
ious T-cell compartments—thymocytes, naïve T cells, and rest-
ing memory, effector, and terminally differentiated T cells—are
affected by HIV. A fuller appreciation of infection within these
compartments may lead to a better understanding of HIV
disease pathogenesis and events that result in latent infection
(13, 20). For example, it has been shown that memory CD4?T
cells specific for HIV are preferentially infected by the virus,
which may contribute to the loss of HIV-specific CD4?T-cell
responses (14, 17, 24). Furthermore, it has been shown that
developing thymocytes at the CD4?CD8?double-positive
stage can become infected by HIV and thus might contribute
to infection of the peripheral naïve CD4?and CD8?T-cell
pools (7, 9, 25, 28, 33, 34, 38, 44). Such infection of naïve T cells
has important consequences, as they might serve as a quiescent
reservoir for virus (19), supply the larger pool of infected
memory CD4?T cells, contribute to infection of CD8?T-cell
pools (30, 34, 44), and attenuate homeostatic maintenance of
the diminishing memory CD4?T-cell pool (18, 23). Addition-
ally, the evidence that dividing CD4?T cells are infected more
efficiently than resting T cells (55) implies that HIV infection,
through inhibition of proliferation or cytotoxicity, would affect
the in vivo generation of terminally differentiated CD4?T cells
There are limited data on HIV infection of memory and
naïve CD4?T cells in vivo (10, 42) and CD8?T cells (25, 30,
34, 44). While studies have demonstrated the presence of HIV
in these T-cell subsets, the purity of T-cell populations ana-
lyzed ex vivo limits interpretations. However, recent advances
in flow cytometry have allowed us to accurately define T-cell
subsets with many more parameters than previously and to
isolate them at high purity (16).
In this study we used polychromatic flow cytometry to sort
multiple CD4?and CD8?T-cell subsets, defined by 11 phe-
notypic parameters, directly from the peripheral blood mono-
nuclear cells (PBMC) of HIV-infected individuals. We com-
pared the frequency of HIV infection by quantitative PCR
within each of these subsets. Importantly, subjects with viremia
may have increased unintegrated viral DNA compared to in-
dividuals with low viral load. Our quantitative PCR assay can-
not distinguish between provirus and unintegrated viral DNA.
Unintegrated viral DNA would have different physiological
consequences compared to viral DNA; however, our assay
offers a measure of infection history nonetheless. Our data
provide detailed information of the degree to which different
* Corresponding author. Mailing address: Building 40 3504, 40 Con-
vent Dr., National Institutes of Health, Bethesda, MD 20892. Phone:
(301) 594-8585. Fax: (301) 480-2779. E-mail: firstname.lastname@example.org.
T-cell compartments serve as substrates and reservoirs for HIV
in vivo and relationships between these compartments. Such ex
vivo analysis of the differential infection of T-cell subsets pro-
vides a mechanistic framework to comprehend HIV pathogen-
esis in vivo.
MATERIALS AND METHODS
Study subjects. Twenty-two HIV-1-infected subjects were recruited for this
study. These subjects included men and women with CD4 counts varying from
101 to 1,746 and viral load varying from ?50 to 170,004; 11 of the individuals
were being treated with highly active antiretroviral therapy, and 11 of the indi-
viduals were highly active antiretroviral therapy naïve. Clinical details are shown
in Table 1. Viral loads were determined with either the Roche Amplicor Monitor
assay or the Roche Ultradirect assay. The subjects all gave informed consent in
compliance with the appropriate institutional review board.
Peptides. Fifteen-mer peptides overlapping by 11 amino acids corresponding
to sequences of the chimeric HXBc2/Bal R5 HIV strain were synthesized as free
acids, and lyophilized peptides were resuspended and grouped together in cor-
responding antigen mixtures as previously described (8).
HIV-specific stimulation assay. Stimulation was performed on fresh or frozen
PBMC as described elsewhere (40). Freshly isolated or freshly thawed PBMC
were resuspended at 106/ml in RPMI medium supplemented with 10% heat-
inactivated fetal calf serum and 1 ?g of anti-CD28 and anti-CD49d antibodies
per ml. Overlapping peptides were used to stimulate HIV-specific T cells in the
presence of brefeldin A (1 ?g/ml) (Sigma) for 5 h at 37°C. All cells were surface
stained for phenotypic markers of interest and intracellularly stained for cyto-
kines or surface stained for tetrameric major histocompatibility complexes.
Monoclonal antibodies. Monoclonal antibodies used for phenotypic charac-
terization of T-cell subsets were anti-CD19 conjugated to Cy5-phycoerythrin
(Cy5PE), anti-CD14 conjugated to Cy5PE, anti-CD56 conjugated to Cy5PE,
anti-CD57 conjugated to fluorescein isothiocyanate, anti-CD27 conjugated to
phycoerythrin, anti-gamma interferon (anti-IFN-?) conjugated to allophycocya-
nin (APC; Becton Dickinson Pharmingen, San Diego, Calif.), anti-CD45RO
conjugated to energy-coupled dye (Coulter), CD3 conjugated to Cascade blue,
CD8 conjugated to Cy7PE CD4 conjugated to Alexa 594, and CD11a conjugated
to Cy7APC. Unconjugated antibodies against CD3, CD8, CD4, and CD11a were
obtained from BD Pharmingen and were then conjugated with the appropriate
fluorochrome (Molecular Probes, Eugene, Oreg.; Amersham, Piscataway, N.J.;
Prozyme San Leandro, Calif.) with standard protocols (http://drmr.com/abcon).
In some experiments, anti-CD31 conjugated to phycoerythrin (BD Pharmingen)
and anti-CD27 conjugated to allophycocyanin (BD Pharmingen) were used in
lieu of anti-IFN-? conjugated to allophycocyanin and anti-CD27 conjugated to
Flow cytometric cell sorting. All sorts were performed on stained cells fixed
with 1% paraformaldehyde (Electron Microscopy Sciences, Ft. Washington, Pa.)
with a modified FACS DIVA (BD Pharmingen). Instrument set-up was per-
formed according to the manufacturer’s instructions. All sorts were performed at
25 lb/in2. Instrument compensation was performed with antibody capture beads
(BD Pharmingen) stained singly with individual antibodies used in the test
Viral DNA. HIV DNA was quantified by quantitative PCR with an ABI7700
(Perkin-Elmer, Norwalk, Conn.) as previously described (17). To quantify cell
number in each reaction, quantitative PCR was performed simultaneously for
albumin gene copy number as previously described (17). Standards were con-
structed for absolute quantification of Gag and albumin copy number and were
validated with sequential dilutions of 8E5 and Ach2 cell lysates, which contain
one copy of Gag per cell. Duplicate reactions were run and template copies were
calculated with ABI7700 software. When no viral DNA was amplified from a
given cell population, we report half the lower limit of detection. As the quan-
titative PCR for gag DNA is sensitive to a single copy of gag DNA, half the lower
limit of detection is based on twice the number of cells put into each PCR
(determined by albumin copy number).
Statistical analysis. Correlations and statistical significance were determined
by Spearman rank correlation analysis with Prism 3.0 software (Prism, San
Polychromatic flow cytometry delineates highly purified T-
cell subsets. PBMC from 22 HIV-infected individuals (Table
1) were stained with a combination of 11 monoclonal antibod-
ies, and seven T-cell populations were sorted by flow cytom-
etry. We specifically chose this cohort of individuals because it
represented a cross-section of HIV-infected individuals at dif-
ferent stages of HIV disease. We simultaneously measured
expression of CD3, CD4, CD8, CD11a, CD27, CD45RO,
CD57, and IFN-?. In addition, to minimize background cap-
ture of antibodies, we designated one fluorochrome channel as
the dump (27). This combination of surface and intracellular
molecules was chosen based upon brightness of expression,
stability after the freeze-thaw process, ease of individual anti-
bodies to be conjugated to different fluorochromes, and ability
to distinguish between naïve, memory, and terminally differ-
entiated T-cell subsets. Using polychromatic flow cytometry
technology (Fig. 1), we sorted naïve CD4?T cells, naïve CD8?
T cells, memory CD8?T cells, HIV-specific CD8?T cells, and
terminally differentiated and preterminally differentiated
memory CD4?T cells based on the phenotypic characteristics
shown in Fig. 1 and described in Table 2.
Figure 1 illustrates the necessity of using multiple parame-
ters; for example, naïve CD8?T cells defined only as CD3?
CD8?CD4?CD11adullCD45RO?can contain 30% nonnaïve
CD8?T cells. A representative postsort analysis for naïve
CD4?T cells (Fig. 2) revealed that less than 0.2% contami-
nating cells were present regardless of which phenotypic mark-
ers were examined. Postsort analyses of other cell populations
were consistently as pure (?99.8%, data not shown).
Memory CD4?T cells are infected at highest frequency.
Following flow cytometric sorting, gag DNA frequencies within
T-cell subsets were compared. The number of gag DNA copies
amplified from each sample was normalized to the number of
cells in each PCR and expressed as gag copies per 105sorted T
cells. In all individuals studied, we were able to amplify gag
DNA from CD57?CD4?memory T cells. Between 100 and
TABLE 1. Subject cohort
VOL. 78, 2004T-CELL SUBSETS THAT HARBOR HIV 1161
10,000 copies of gag DNA/105sorted T cells were amplified
from this population (Fig. 3a and b).
Memory T cells are a heterogeneous population. In HIV
infection this heterogeneity is particularly evident when study-
ing CD57 expression by CD4?T cells (CD57 marks terminally
differentiated T cells) (8, 31, 53). The CD4?CD57?T-cell
subset overlaps the effector memory T-cell subset (8, 45) and
was dramatically expanded in many HIV-infected individuals
(P ? 0.002, data not shown). Hence, we sought to examine the
infection history of memory CD57?and CD57?memory
CD4?T cells. Collectively, memory CD57?CD4?T cells
contained more gag DNA than did CD57?CD4?T cells (Fig.
3a). Assuming virus copy number per cell is distributed simi-
larly between infected cells, memory CD57?CD4?T cells
were always infected more frequently than CD57?CD4?T
cells (Fig. 3b).
The number of infected CD57?CD4?memory T cells cor-
related with plasma viral load (Fig. 3c). In addition, the num-
ber of infected CD57?CD4?T cells correlated with the num-
ber of infected CD57?memory CD4?T cells (Fig. 3d). This
indicates that the plasma virus pool and the pool of infected
memory CD57?and CD57?CD4?T cells are intimately re-
Memory CD8?T cells are rarely infected by HIV. Several
reports show the presence of HIV in CD8?T cells (25, 34, 44).
Therefore, we initially compared the levels of gag DNA in bulk
memory CD8?T cells with that in memory CD57?CD4?T
cells (Fig. 4a and b). We found that although CD8?T cells can
harbor HIV DNA, the frequency was extremely low. Memory
CD8?T cells from many individuals had no detectable viral
DNA (Fig. 4b). However, recently stimulated CD8?T cells
FIG. 1. Flow cytometric sorting strategy for T cells. PBMC from 16 subjects in the cohort were stimulated with overlapping HIV-peptides
stained extracellularly with the antibody combination described in the text and intracellularly for IFN-?. Lymphocytes were defined with forward
and side scatter (I). CD3?T cells were then defined based on expression of CD3 without expression of CD56, CD14, or CD19 (dump) (II). CD4?
T cells were then defined based on expression of CD4 without expression of CD8, CD8?T cells were defined based on expression of CD8 without
expression of CD4 (III). Naïve CD4?T cells were defined based on dull expression of CD11a, no expression of CD45RO or CD57 with expression
of CD27 (IV A). Memory CD4?T cells were defined based on expression of CD45RO with high expression of CD11a. Memory CD4?T cells were
then separated based on expression of CD57. Naïve CD8?T cells were defined under the same constraints as naïve CD4?T cells (IV B). Memory
CD8?T cells were separated into HIV-specific (production of IFN-? or tetramer binding) and other memory CD8?T cells.
TABLE 2. T-cell phenotypes
CD4 CD8 CD11a CD45RO CD27 CD57 IFN-?
Naı ¨ve CD4?
Naı ¨ve CD8?
aAll populations were also defined as lymphocytes (by forward and side
scatter), CD3?, CD14?, CD16?, and CD19?. Blanks indicate that expression of
that marker was not used to define the population.
1162BRENCHLEY ET AL. J. VIROL.
FIG. 2. Postsort analysis of naïve CD4?T cells. In order to ensure the purity of sorted populations, each sorted population (when possible) was
reanalyzed on the same instrument with the same instrument settings. A representative example is shown. Sorted cells must be defined for
lymphocytes because cellular debris results from high-speed sorting. The right four plots are only defined for lymphocytes based on characteristic
forward and side scatter. All sorted populations were routinely ?99.8% pure.
FIG. 3. CD57?CD4?T cells have less viral DNA than memory CD57?CD4?T cells. PBMC from HIV-infected individuals were stained with
the antibody combination detailed in Fig. 2. Memory CD57?and CD57?CD4?T cells were sorted, and quantitative PCR for gag DNA and
albumin was performed. Infection of CD57?CD4?T cells was compared to infection of CD57?memory CD4?T cells in a subject-independent
fashion (A) and a subject-dependent fashion, with white bars representing CD57?memory CD4?T cells and shaded bars representing CD57?
memory CD4?T cells (B). Asterisks mark individual subsets where no gag DNA was amplified, and the values listed are calculated based on half
of the lower limit of detection. Corresponding subjects are listed along the x axis. The plasma viral load was compared to the number of infected
CD57?memory CD4?T cells (C). While there is a correlation between the number of infected CD57?CD4?T cells and the number of infected
CD57?memory CD4?T cells (D), the CD57?population contains significantly less HIV than the CD57?memory CD4?T-cell subset (A and B).
VOL. 78, 2004 T-CELL SUBSETS THAT HARBOR HIV1163
express low levels of CD4 and could be targets for HIV (25,
44). Therefore, we sorted CD8?CD4dullT cells from five
subjects in the cohort. There were more (5- to 100-fold) copies
of viral DNA within the CD8?CD4dullT cells than in memory
CD8?CD4?T cells (data not shown). These data suggest that
memory CD8?T cells are capable of becoming infected after
activation-induced expression of CD4.
HIV-specific CD8?T cells are not preferentially infected by
HIV. We hypothesized that HIV-specific CD8?T cells might
become preferentially infected as they respond to HIV anti-
gens and become activated in vivo. Thus, we sorted HIV-
specific CD8?T cells identified by production of IFN-? fol-
lowing HIV peptide stimulation from seven subjects in the
cohort (17, 40). Only in four of these seven individuals were we
able to amplify any gag DNA. Thus, while HIV-specific CD8?
T cells were capable of becoming HIV-infected, this popula-
tion was clearly not preferentially infected (Fig. 4c and d). We
also used tetramer binding instead of IFN-? production to
define HIV-specific CD8?T cells. This confirmed that their
infection was extremely rare (subjects 17 to 22, data not
shown) and thus unlikely to account for functional defects
within this population (1, 11, 36, 39, 51).
Naïve CD4?T cells are infected at low frequency by HIV.
Naïve T cells may become infected during maturation as thy-
mocytes or as mature naïve T cells in the periphery. To address
each possibility, we studied infection of naïve CD4?and CD8?
T cells. Sufficient naïve CD4?T cells were sorted from 11
HIV-infected individuals. Naïve CD4?T cells were found to
have significantly less viral DNA (on average 10 times less)
than CD57?memory CD4?T cells (P ? 0.005, Fig. 5a and b).
In three of the individuals, no gag DNA was amplified from
sorted naïve CD4?T cells. However, in one subject there was
more viral DNA in the naïve CD4?T cells. We found no
significant relationship between infection of naïve CD4?T
cells and plasma viral load (Fig. 5c, R ? 0.39, P ? not signif-
icant). In addition, the frequency of infection of naïve CD4?T
cells did not correlate with the frequency of infection of
CD57?CD4?memory T cells (Fig. 5d, R ? 0.45, P ? not
significant). Hence, while naïve CD4?T cells were capable of
becoming infected by HIV, infected naïve CD4?T cells did not
significantly contribute to the pool of infected memory CD4?
T cells. Furthermore, the cellular and viral factors that influ-
ence the ability of memory T cells to become infected by HIV
may not similarly influence the ability of naïve CD4?T cells to
become HIV infected (17, 19, 43, 47, 54).
Naïve CD8?T cells are not infected by HIV. Infection of
CD8?CD4?thymocytes could result in the export of infected
mature naïve CD4?and CD8?T cells to the periphery. As
naïve CD4?T cells can contain HIV in vivo (Fig. 5), we wanted
to determine whether infection of naïve CD8?T cells could be
observed. We sorted sufficient numbers of naïve CD8?T cells
from 12 individuals in the cohort, and were able to amplify gag
DNA from only three of the naïve CD8?T-cell subsets (Fig. 6a
and b). Furthermore, more than 107highly purified naïve
CD8?T cells were sorted in total (cumulative for all 12 sub-
jects) but only six copies of gag DNA were detected. Even at
our level of sorting precision (?99.8%) at this extremely low
level of gag DNA we cannot exclude contamination by other
FIG. 4. HIV infection of memory CD8?T cells. The fraction of infected memory CD8?T cells was compared to the number of infected
memory CD57?CD4?T cells for all subjects (A) and on an individual subject basis (B), with white bars representing memory CD8?T cells and
shaded bars representing CD57?CD4?T cells. Asterisks mark individual subsets where no gag DNA was amplified, and the values listed are
calculated based on half of the lower limit of detection. Corresponding subjects are listed along the x axis. Infection of HIV-specific CD8?T cells
(based on production of IFN-? following HIV peptide stimulation) was then compared to infection of other memory CD8?T cells, and no
significant differences were observed (C). The infection frequency of HIV-specific CD8?T cells was then compared to the infection frequency of
other memory CD8?T cells in a subject-dependent fashion (white bars represent HIV-specific CD8?T cells, and shaded bars represent memory
CD8?T cells) (D). Corresponding subjects are listed along the x axis.
1164BRENCHLEY ET AL. J. VIROL.
T-cell populations, and it is likely that these few copies of gag
DNA actually reside in contaminating cells. This suggests that
naïve CD8?T cells carry no HIV and that the thymus exports
no infected naïve T cells.
Naïve CD4?T cells are infected by HIV in the periphery.
Our observation that naïve CD8?T cells rarely, if ever, contain
gag DNA suggested that naïve CD4?T cells are likely to have
become infected in the periphery. Recently it has been shown
that naïve T cells which have proliferated without T-cell re-
ceptor-mediated stimulation lose surface expression of CD31
(29). We used CD31 expression to differentiate between naïve
CD4?T cells that had and had not proliferated. In 7 subjects,
after gating for naïve CD4?T cells as before (Fig. 1), we sorted
naïve CD4?T cells based on CD31 expression (Fig. 6c). We
initially confirmed that the CD31?naïve CD4?T cells had
undergone fewer rounds of proliferation. CD31?naïve CD4?
T cells had, on average fivefold more copies of T-cell receptor
excision circle than CD31?naïve CD4?T cells (data not
shown). In addition, in all 7 subjects the frequency of gag DNA
was higher in the CD31?than the CD31?subset (P ? 0.016,
Fig. 6d). Taken together, these data suggest that infection of
naïve CD4?T cells occurs primarily in the periphery within
naïve CD4?T cells that have or are proliferating and that
infection of double positive thymocytes rarely, if ever, leads to
infection within the naïve T-cell population.
It is generally accepted that activated memory CD4?T cells
are the predominant targets for HIV infection (5, 12). How-
ever, it remains unclear what other sources of infected cells
exist, what factors lead to their infection, and to what extent
these cells contribute to the total pool of infected cells. Un-
derstanding which T-cell subsets contain HIV in vivo could
establish a mechanistic framework to explain the loss of CD4?
T cells and the inability of the HIV-specific immune response
to control HIV replication. Here, we examined in vivo HIV
infection of multiple highly purified and stringently defined
T-cell subsets by quantifying viral DNA without further in vitro
manipulations. The major findings to emerge from these stud-
ies are that central memory CD4?T cells contain the highest
frequency of viral DNA; terminally differentiated effector
memory CD57?CD4?T cells contain, on average, 10 times
fewer copies of viral DNA than central memory CD4?T cells;
memory CD8?T cells rarely contain viral DNA unless acti-
vated to express CD4; HIV-specific CD8?T cells are not
preferentially infected by HIV; naïve CD4?T cells that pro-
liferate, or have proliferated, in the periphery contain more
viral DNA than other naïve T cells; and naïve CD8?T cells are
probably never infected. Importantly, these trends are exactly
the same regardless of disease state or treatment status.
Taken together, our data show that the T-cell subsets most
likely to become infected are those CD4?T cells with a history
of proliferation: CD31?naïve T cells and, to a greater extent,
resting memory T cells. However, our data also reveal that
infection history itself influences proliferative and maturation
capacity in vivo. First, it has been well documented that devel-
oping thymocytes can be infected by HIV (2, 4, 7, 37, 48),
FIG. 5. Naïve CD4?T cells have less viral DNA than memory CD57?CD4?T cells. PBMC from HIV-infected individuals were stained with
the antibody combination detailed in Fig. 2. Memory CD57?and naïve CD4?T cells were sorted, and quantitative PCR for gag DNA and albumin
was performed on sorted T cells. Infection of naïve CD4?T cells was compared to infection of CD57?memory CD4?T cells in a subject-
independent fashion (A) and a subject-dependent fashion (B). White bars represent naïve CD4?T cells, and shaded bars represent CD57?
memory CD4?T cells (B). Corresponding subjects are listed along the x axis. The plasma viral load was compared to the number of infected naïve
CD4?T cells (C). There was no correlation between the number of infected naïve CD4?T cells and the number of infected CD57?memory CD4?
T cells (D).
VOL. 78, 2004 T-CELL SUBSETS THAT HARBOR HIV1165
suggesting that they might give rise to infected naïve CD4?and
CD8?T cells (19, 34). However, our results show that infection
of developing thymocytes is unlikely to lead to infected naïve T
cells in the periphery because we were able to find virtually no
infected naïve CD8?T cells in any HIV-infected individuals.
This is supported by our observation that most infected naïve
CD4?T cells are of the CD31?phenotype, suggesting that
they were probably infected while proliferating in the periph-
ery. Our data do not suggest that developing thymocytes are
not infected by HIV in vivo, rather that such infected thymo-
cytes do not become infected naïve T cells. The importance of
thymic infection would therefore be one of depleting the sup-
ply of new naïve T cells, and not that of supplying HIV-infected
naïve T cells. The ability of HIV to infect naïve CD4?T cells
in the periphery suggests the potential ability of the virus to
maintain long-lived latency due to the long life span of naïve T
cells and that the probability of stimulating an infected naïve
CD4?T-cell by cognate major histocompatibility complex-
peptide is extremely low (35).
Second, the lack of correlation between the infected naïve
CD4?T-cell pool and the infected memory CD4?T-cell pool
implies that infected naïve T cells do not significantly contrib-
ute to the pool of infected memory CD4?T cells, but that they
die following antigenic stimulation. This suggests a mechanism
by which HIV infection can adversely affect maintenance of the
memory CD4?T-cell pool and shows that the predominant
way of producing infected memory CD4?T cells is by their
Finally, although the memory CD4?T-cell pool as a whole
is the most frequently infected, we have shown that CD57?
FIG. 6. Infection of naïve CD8?T cells and peripheral infection of naïve CD4?T cells. Infection of highly purified naïve CD8?T cells was
compared to infection of naïve CD4?T cells, memory CD8?T cells, and CD57?memory CD4?T cells (A). Naïve CD8?T cells are significantly
less likely to carry HIV than any other subset studied. Comparison of infection of naïve CD8?T cells (white bars, B) and naïve CD4?T cells
(shaded bars, B) demonstrates that naïve CD8?T cells rarely contain detectable viral DNA (asterisks mark individual subsets where no gag DNA
was amplified and values are calculated as half the lower limit of detection). Corresponding subjects are listed along the x axis. Naïve CD4?T cells
were stained and defined as before with side scatter, forward scatter, CD3, dump, CD4, CD8, CD45RO, CD11a, CD27, and CD57 (Fig. 2) from
four subjects in the cohort (13 to 16) and were then separated on the basis of surface CD31 expression (C). Sorted T cells were then assayed for
gag DNA by quantitative PCR. The number of infected naïve CD31?CD4?T cells (white bars) was compared to the number of infected naïve
CD31?CD4?T cells (shaded bars) (D).
FIG. 7. T cells that harbor HIV. A pie chart averaged from four
subjects in the cohort demonstrates the individual contributions of all
T-cell subsets studied to the total pool of infected T cells. The mag-
nitude of infection within each subset and the contribution of each
subset to the pool of PBMC were used in the calculation.
1166 BRENCHLEY ET AL.J. VIROL.
memory CD4?T cells, which have undergone the most rounds
of proliferation to achieve terminal differentiation, are in fact
10-fold less likely to have been infected by HIV. These termi-
nally differentiated memory CD4?T cells are expanded in
HIV infection (15, 31), in part, due to polyclonal T-cell acti-
vation (3, 21).
One interpretation of the marked disparity in frequency of
infection is that if T cells become infected at an earlier stage in
their proliferative history (when they are CD57?), they are less
likely to survive and/or divide to become terminally differen-
tiated CD57?T cells. This would provide direct evidence that
infection of memory CD4?T cells in vivo prevents them from
undergoing the normal homeostatic processes that contribute
to the maintenance of the resting memory CD4?T-cell pool. It
is also possible that the CD57?subset contains the same fre-
quency of infected cells as the CD57?subset, but with on
average 10-fold fewer copies of virus per cell. Studies with
single-cell PCR to detect HIV DNA could help to clarify this
possibility but are difficult because the frequency of infected
cells is so low (17, 26). It is unlikely that the reason for the
difference in infection frequency is simply that terminally dif-
ferentiated CD57?T cells are less infectible than other mem-
ory T cells. CD57?T cells express the same levels of CD4 and
CCR5/CXCR4 as CD57?T cells (data not shown). Further-
more, both subsets contain an equally small frequency of T
cells which express activation markers such as CD69 and
CD25, and CD57?T cells die without proliferating after acti-
vation (8); thus, our analysis largely detects infection events
that occurred before T cells became terminally differentiated
CD57?. However, we also found that there are virtually no
CD57?memory CD4?T cells that express Ki67 in the periph-
ery. This finding might also contribute to the greater infection
within the CD57?memory CD4?T-cell subset.
Alternatively, the differences we observed in infectivity
could arise due to infection of different T-cell subsets by viral
subspecies with distinctive tropism or replicative capacity. Of
particular interest is whether naïve CD4?T cells are infected
with CXCR4-or CCR5-tropic virus. As naïve CD4?T cells do
not express CCR5, we would speculate that naïve CD4?T cells
infected in the periphery would be infected with CXCR4-
We have previously shown that HIV-specific CD4?T cells
are preferentially infected by HIV (17). Since stimulated
CD8?T cells have been shown to express CD4 transiently
following stimulation, leading to marginal infection of CD8?T
cells by HIV (25, 30, 44), we hypothesized that HIV-specific
CD8?T cells might also become preferentially infected by
HIV. However, while we show that memory CD8?T cells are
occasionally infected by HIV, we did not find that the virus
preferentially infected HIV-specific CD8?T cells. In fact, we
found relatively few copies of HIV gag DNA within HIV-
specific CD8?T cells, implying that infection of this subset
neither contributes to the inability to control viral replication
nor accounts for the observed defects within this subset (1, 11,
39, 50). Lack of preferential infection of HIV-specific CD8?T
cells might be explained by a number of possibilities. It is
possible that upregulation of CD4 by stimulated HIV-specific
CD8?T cells is not sufficient to allow HIV infection. Alterna-
tively, HIV-specific CD8?T cells may produce enough ? che-
mokines upon stimulation to prevent HIV infection (1, 41, 52).
In addition, HIV-specific CD8?and CD4?T cells may be
stimulated by different cell types or in different locations in
In summary, our data show which T-cell subsets are infected
in vivo and to what extent each compartment contributes to the
total pool of cellular associated virus (Fig. 7) and suggest what
circumstances can lead to their infection and the consequences
of that infection. Specifically, this approach allowed us to dem-
onstrate the importance of cellular activation and proliferation
in allowing HIV replication in vivo and further to show that
infection of these cells in vivo leads to an altering of their life
span, decreasing their likelihood of reaching terminal differ-
entiation. Collectively, these findings support a mechanism by
which HIV infection exacerbates depletion of CD4?T cells in
the context of homeostatic strain imposed by chronic T-cell
We thank Steven De Rosa for guidance in polychromatic flow cy-
tometry, Joanne Yu for antibody conjugation, and Steven Perfetto for
assistance in instrument operation.
1. Appay, V., D. F. Nixon, S. M. Donahoe, G. M. Gillespie, T. Dong, A. King,
G. S. Ogg, H. M. Spiegel, C. Conlon, C. A. Spina, D. V. Havlir, D. D.
Richman, A. Waters, P. Easterbrook, A. J. McMichael, and S. L. Rowland-
Jones. 2000. HIV-specific CD8(?) T cells produce antiviral cytokines but are
impaired in cytolytic function. J. Exp. Med. 192:63–75.
2. Beltz, L. 1999. Thymic involution and HIV progression. Immunol. Today
3. Bentwich, Z., A. Kalinkovich, Z. Weisman, and Z. Grossman. 1998. Immune
activation in the context of HIV infection. Clin. Exp. Immunol. 111:1–2.
4. Berkowitz, R. D., K. P. Beckerman, T. J. Schall, and J. M. McCune. 1998.
CXCR4 and CCR5 expression delineates targets for HIV-1 disruption of
T-cell differentiation. J. Immunol. 161:3702–3710.
5. Blankson, J. N., D. Persaud, and R. F. Siliciano. 2002. The challenge of viral
reservoirs in HIV-1 infection. Annu. Rev. Med. 53:557–593.
6. Boaz, M. J., A. Waters, S. Murad, P. J. Easterbrook, and A. Vyakarnam.
CD28?IL-2? CD4 T-cell responses is associated with nonprogression in
HIV-1 infection. J. Immunol. 169:6376–6385.
7. Bonyhadi, M., L. Rabin, S. Salimi, D. Brown, J. Kosek, J. McCune, and H.
Kaneshima. 1993. HIV induces thymus depletion in vivo. Nature 363:728–
8. Brenchley, J. M., N. J. Karandikar, M. R. Betts, D. R. Ambrozak, B. J. Hill,
L. E. Crotty, J. P. Casazza, J. Kuruppu, S. A. Migueles, M. Connors, M.
Roederer, D. C. Douek, and R. A. Koup. 2003. Expression of CD57 defines
replicative senescence and antigen-induced apoptotic death of CD8? T
cells. Blood 101:2711–2720.
9. Brooks, D. G., S. G. Kitchen, C. M. Kitchen, D. D. Scripture-Adams, and
J. A. Zack. 2001. Generation of HIV latency during thymopoiesis. Nat. Med.
10. Cayota, A., F. Vuillier, D. Scott-Algara, V. Feuillie, and G. Dighiero. 1993.
Differential requirements for HIV-1 replication in naive and memory CD4 T
cells from asymptomatic HIV-1 seropositive carriers and AIDS patients.
Clin. Exp. Immunol. 91:241–248.
11. Champagne, P., G. S. Ogg, A. S. King, C. Knabenhans, K. Ellefsen, M.
Nobile, V. Appay, G. P. Rizzardi, S. Fleury, M. Lipp, R. Forster, S. Rowland-
Jones, R. P. Sekaly, A. J. McMichael, and G. Pantaleo. 2001. Skewed mat-
uration of memory HIV-specific CD8 T lymphocytes. Nature 410:106–111.
12. Chun, T. W., L. Carruth, D. Finzi, X. Shen, J. A. DiGiuseppe, H. Taylor, M.
Hermankova, K. Chadwick, J. Margolick, T. C. Quinn, Y. H. Kuo, R. Brook-
meyer, M. A. Zeiger, P. Barditch-Crovo, and R. F. Siliciano. 1997. Quanti-
fication of latent tissue reservoirs and total body viral load in HIV-1 infec-
tion. Nature 387:183–188.
13. Chun, T. W., D. Engel, S. B. Mizell, C. W. Hallahan, M. Fischette, S. Park,
R. T. Davey, Jr., M. Dybul, J. A. Kovacs, J. A. Metcalf, J. M. Mican, M. M.
Berrey, L. Corey, H. C. Lane, and A. S. Fauci. 1999. Effect of interleukin-2
on the pool of latently infected, resting CD4? T cells in HIV-1-infected
patients receiving highly active anti-retroviral therapy. Nat. Med. 5:651–655.
14. Demoustier, A., B. Gubler, O. Lambotte, M. G. De Goer, C. Wallon, C.
Goujard, J. F. Delfraissy, and Y. Taoufik. 2002. In patients on prolonged
highly active antiretroviral therapy, a significant pool of HIV infected CD4
T cells are HIV-specific. AIDS 16:1749–1754.
VOL. 78, 2004T-CELL SUBSETS THAT HARBOR HIV 1167
15. De Paoli, P., A. Carbone, S. Battistin, M. Crovatto, N. Arreghini, and G. Download full-text
Santini. 1987. Selective depletion of the OKT 4? 4B4? subset in lymph
nodes from HIV? patients. Immunol. Lett. 16:71–73.
16. De Rosa, S. C., L. A. Herzenberg, and M. Roederer. 2001. 11-color, 13-
parameter flow cytometry: identification of human naive T cells by pheno-
type, function, and T-cell receptor diversity. Nat. Med. 7:245–248.
17. Douek, D., J. Brenchley, M. Betts, D. Ambrozak, B. Hill, Y. Okamoto, J.
Casazza, J. Kuruppu, K. Kunstman, S. Wolinsky, Z. Grossman, M. Dybul,
A. Oxenius, D. Price, M. Connors, and R. A. Koup. 2002. HIV preferentially
infects HIV-specific CD4? T cells. Nature 417:95–98.
18. Douek, D. C., L. J. Picker, and R. A. Koup. 2003. T cell dynamics in HIV-1
infection. Annu. Rev. Immunol. 21:265–304.
19. Eckstein, D. A., M. L. Penn, Y. D. Korin, D. D. Scripture-Adams, J. A. Zack,
J. F. Kreisberg, M. Roederer, M. P. Sherman, P. S. Chin, and M. A. Gold-
smith. 2001. HIV-1 actively replicates in naive CD4(?) T cells residing
within human lymphoid tissues. Immunity 15:671–682.
20. Finzi, D., J. Blankson, J. D. Siliciano, J. B. Margolick, K. Chadwick, T.
Pierson, K. Smith, J. Lisziewicz, F. Lori, C. Flexner, T. C. Quinn, R. E.
Chaisson, E. Rosenberg, B. Walker, S. Gange, J. Gallant, and R. F. Siliciano.
1999. Latent infection of CD4? T cells provides a mechanism for lifelong
persistence of HIV-1, even in patients on effective combination therapy. Nat.
21. Grossman, Z., M. B. Feinberg, and W. E. Paul. 1998. Multiple modes of
cellular activation and virus transmission in HIV infection: a role for chron-
ically and latently infected cells in sustaining viral replication. Proc. Natl.
Acad. Sci. USA 95:6314–6319.
22. Grossman, Z., M. Meier-Schellersheim, A. E. Sousa, R. M. Victorino, and
W. E. Paul. 2002. CD4? T-cell depletion in HIV infection: are we closer to
understanding the cause? Nat. Med. 8:319–323.
23. Grossman, Z., and W. E. Paul. 2000. The impact of HIV on naive T-cell
homeostasis. Nat. Med. 6:976–977.
24. Harari, A., G. P. Rizzardi, K. Ellefsen, D. Ciuffreda, P. Champagne, P. A.
Bart, D. Kaufmann, A. Telenti, R. Sahli, G. Tambussi, L. Kaiser, A. Lazza-
rin, L. Perrin, and G. Pantaleo. 2002. Analysis of HIV-1- and CMV-specific
memory CD4 T-cell responses during primary and chronic infection. Blood
25. Imlach, S., S. McBreen, T. Shirafuji, C. Leen, J. E. Bell, and P. Simmonds.
2001. Activated peripheral CD8 lymphocytes express CD4 in vivo and are
targets for infection by human immunodeficiency virus type 1. J. Virol.
26. Jung, A., R. Maier, J. P. Vartanian, G. Bocharov, V. Jung, U. Fischer, E.
Meese, S. Wain-Hobson, and A. Meyerhans. 2002. Multiply infected spleen
cells in HIV patients. Nature 418:144.
27. Kantor, A., and M. Roederer. 1997. FACS analysis of lymphocytes, p. 49.1–
49.13. In L. A. Herzenberg, D. M. Weir, L. A. Herzenberg, and C. Blackwell
(ed.), Handbook of experimental immunology, 5th ed., vol. 2. Blackwell
Science, Cambridge, England.
28. Keir, M. E., M. G. Rosenberg, J. K. Sandberg, K. A. Jordan, A. Wiznia, D. F.
Nixon, C. A. Stoddart, and J. M. McCune. 2002. Generation of
CD3?CD8low thymocytes in the HIV type 1-infected thymus. J. Immunol.
29. Kimmig, S., G. K. Przybylski, C. A. Schmidt, K. Laurisch, B. Mowes, A.
Radbruch, and A. Thiel. 2002. Two subsets of naive T helper cells with
distinct T-cell receptor excision circle content in human adult peripheral
blood. J. Exp. Med. 195:789–794.
30. Kitchen, S. G., Y. D. Korin, M. D. Roth, A. Landay, and J. A. Zack. 1998.
Costimulation of naive CD8?lymphocytes induces CD4 expression and
allows human immunodeficiency virus type 1 infection. J. Virol. 72:9054–
31. Legac, E., B. Autran, H. Merle-Beral, C. Katlama, and P. Debre. 1992.
CD4?CD7-CD57? T cells: a new T-lymphocyte subset expanded during
human immunodeficiency virus infection. Blood 79:1746–1753.
32. Lieberman, J., P. Shankar, N. Manjunath, and J. Andersson. 2001. Dressed
to kill? A review of why antiviral CD8 T lymphocytes fail to prevent pro-
gressive immunodeficiency in HIV-1 infection. Blood 98:1667–1677.
33. Livingstone, W. J., M. Moore, D. Innes, J. E. Bell, and P. Simmonds. 1996.
Frequent infection of peripheral blood CD8-positive T-lymphocytes with
HIV-1. Edinburgh Heterosexual Transmission Study Group. Lancet 348:
34. McBreen, S., S. Imlach, T. Shirafuji, G. R. Scott, C. Leen, J. E. Bell, and P.
Simmonds. 2001. Infection of the CD45RA?(naive) subset of peripheral
CD8? lymphocytes by human immunodeficiency virus type 1 in vivo. J. Vi-
35. Michie, C. A., A. McLean, C. Alcock, and P. C. Beverley. 1992. Lifespan of
human lymphocyte subsets defined by CD45 isoforms. Nature 360:264–265.
36. Mueller, Y. M., S. C. De Rosa, J. A. Hutton, J. Witek, M. Roederer, J. D.
Altman, and P. D. Katsikis. 2001. Increased CD95/Fas-induced apoptosis of
HIV-’pecific CD8(?) T cells. Immunity 15:871–882.
37. Namikawa, R., H. Kaneshima, M. Lieberman, I. L. Weissman, and J. M.
McCune. 1988. Infection of the SCID-hu mouse by HIV-1. Science 242:
38. Napolitano, L. A., J. C. Lo, M. B. Gotway, K. Mulligan, J. D. Barbour, D.
Schmidt, R. M. Grant, R. A. Halvorsen, M. Schambelan, and J. M. McCune.
2002. Increased thymic mass and circulating naive CD4 T cells in HIV-1-
infected adults treated with growth hormone. AIDS 16:1103–1111.
39. Oxenius, A., A. K. Sewell, S. J. Dawson, H. F. Gunthard, M. Fischer, G. M.
Gillespie, S. L. Rowland-Jones, C. Fagard, B. Hirschel, R. E. Phillips, and
D. A. Price. 2002. Functional discrepancies in HIV-specific CD8? T-lym-
phocyte populations are related to plasma virus load. J. Clin. Immunol.
40. Pitcher, C. J., C. Quittner, D. M. Peterson, M. Connors, R. A. Koup, V. C.
Maino, and L. J. Picker. 1999. HIV-1-specific CD4? T cells are detectable
in most individuals with active HIV-1 infection, but decline with prolonged
viral suppression. Nat. Med. 5:518–525.
41. Price, D. A., A. K. Sewell, T. Dong, R. Tan, P. J. Goulder, S. L. Rowland-
Jones, and R. E. Phillips. 1998. Antigen-specific release of beta-chemokines
by anti-HIV-1 cytotoxic T lymphocytes. Curr. Biol. 8:355–358.
42. Riley, J. L., B. L. Levine, N. Craighead, T. Francomano, D. Kim, R. G.
Carroll, and C. H. June. 1998. Naive and memory CD4 T cells differ in their
susceptibilities to human immunodeficiency virus type 1 infection following
CD28 costimulation: implications for transmission and pathogenesis. J. Vi-
43. Roederer, M., P. A. Raju, D. K. Mitra, and L. A. Herzenberg. 1997. HIV does
not replicate in naive CD4 T cells stimulated with CD3/CD28. J. Clin.
44. Saha, K., J. Zhang, A. Gupta, R. Dave, M. Yimen, and B. Zerhouni. 2001.
Isolation of primary HIV-1 that target CD8? T lymphocytes with CD8 as a
receptor. Nat. Med. 7:65–72.
45. Sallusto, F., D. Lenig, R. Forster, M. Lipp, and A. Lanzavecchia. 1999. Two
subsets of memory T lymphocytes with distinct homing potentials and effec-
tor functions. Nature 401:708–712.
46. Shankar, P., M. Russo, B. Harnisch, M. Patterson, P. Skolnik, and J.
Lieberman. 2000. Impaired function of circulating HIV-specific CD8(?) T
cells in chronic human immunodeficiency virus infection. Blood 96:3094–
47. Spina, C. A., H. E. Prince, and D. D. Richman. 1997. Preferential replication
of HIV-1 in the CD45RO memory cell subset of primary CD4 lymphocytes
in vitro. J. Clin. Investig. 99:1774–1785.
48. Stoddart, C. A., T. J. Liegler, F. Mammano, V. D. Linquist-Stepps, M. S.
Hayden, S. G. Deeks, R. M. Grant, F. Clavel, and J. M. McCune. 2001.
Impaired replication of protease inhibitor-resistant HIV-1 in human thymus.
Nat. Med. 7:712–718.
49. Sullivan, Y. B., A. L. Landay, J. A. Zack, S. G. Kitchen, and L. Al-Harthi.
2001. Upregulation of CD4 on CD8? T cells: CD4dimCD8bright T cells
constitute an activated phenotype of CD8? T cells. Immunology 103:270–
50. Trimble, L. A., L. W. Kam, R. S. Friedman, Z. Xu, and J. Lieberman. 2000.
CD3zeta and CD28 down-modulation on CD8 T cells during viral infection.
51. Trimble, L. A., P. Shankar, M. Patterson, J. P. Daily, and J. Lieberman.
2000. Human immunodeficiency virus-specific circulating CD8 T lympho-
cytes have down-modulated CD3zeta and CD28, key signaling molecules for
T-cell activation. J. Virol. 74:7320–7330.
52. Wagner, L., O. O. Yang, E. A. Garcia-Zepeda, Y. Ge, S. A. Kalams, B. D.
Walker, M. S. Pasternack, and A. D. Luster. 1998. Beta-chemokines are
released from HIV-1-specific cytolytic T-cell granules complexed to proteo-
glycans. Nature 391:908–911.
53. Weyand, C. M., J. C. Brandes, D. Schmidt, J. W. Fulbright, and J. J.
Goronzy. 1998. Functional properties of CD4? CD28- T cells in the aging
immune system. Mech. Ageing Dev. 102:131–147.
54. Woods, T. C., B. D. Roberts, S. T. Butera, and T. M. Folks. 1997. Loss of
inducible virus in CD45RA naive cells after human immunodeficiency vi-
rus-1 entry accounts for preferential viral replication in CD45RO memory
cells. Blood 89:1635–1641.
55. Zack, J. A., S. J. Arrigo, S. R. Weitsman, A. S. Go, A. Haislip, and I. S. Chen.
1990. HIV-1 entry into quiescent primary lymphocytes: molecular analysis
reveals a labile, latent viral structure. Cell 61:213–222.
1168BRENCHLEY ET AL. J. VIROL.