JOURNAL OF VIROLOGY, July 2010, p. 6425–6437
Copyright © 2010, American Society for Microbiology. All Rights Reserved.
Vol. 84, No. 13
Establishment of HIV Latency in Primary CD4?Cells Is due to
Epigenetic Transcriptional Silencing and P-TEFb Restriction?
Mudit Tyagi, Richard John Pearson,† and Jonathan Karn*
Department of Molecular Biology and Microbiology, Case Western Reserve University School of Medicine, Cleveland, Ohio 44106
Received 21 July 2009/Accepted 15 April 2010
The development of suitable experimental systems for studying HIV latency in primary cells that permit
detailed biochemical analyses and the screening of drugs is a critical step in the effort to develop viral
eradication strategies. Primary CD4?T cells were isolated from peripheral blood and amplified by antibodies
to the T-cell receptor (TCR). The cells were then infected by lentiviral vectors carrying fluorescent reporters
and either the wild-type Tat gene or the attenuated H13L Tat gene. After sorting for the positive cells and
reamplification, the infected cells were allowed to spontaneously enter latency by long-term cultivation on the
H80 feeder cell line in the absence of TCR stimulation. By 6 weeks almost all of the cells lost fluorescent protein
marker expression; however, more than 95% of these latently infected cells could be reactivated after stimu-
lation of the TCR by ?-CD3/CD28 antibodies. Chromatin immunoprecipitation assays showed that, analo-
gously to Jurkat T cells, latent proviruses in primary CD4?T cells are enriched in heterochromatic markers,
including high levels of CBF-1, histone deacetylases, and methylated histones. Upon TCR activation, there was
recruitment of NF-?B to the promoter and conversion of heterochromatin structures present on the latent
provirus to active euchromatin structures containing acetylated histones. Surprisingly, latently infected pri-
mary cells cannot be induced by tumor necrosis factor alpha because of a restriction in P-TEFb levels, which
can be overcome by activation of the TCR. Thus, a combination of restrictive chromatin structures at the HIV
long terminal repeat and limiting P-TEFb levels contribute to transcriptional silencing leading to latency in
primary CD4?T cells.
The introduction of highly active antiretroviral therapy
(HAART) in the mid 1990s led to a dramatic increase in
patient longevity due to the ability of antiretroviral drugs to
suppress HIV replication to below threshold detection levels
(?50 copies HIV RNA/ml) (23, 52). Unfortunately, despite
the intensive therapy, there is continuing viral replication at
levels below the limits of detection of most clinical assays due
to inefficient antiviral pharmacodynamics that create environ-
ments where drug potency is reduced (12, 13, 41, 43). For
example, there is recent evidence for ongoing HIV replication
in gut-associated lymphoid tissue during long-term antiretro-
viral therapy (7). A second cause of HIV treatment failure is
the creation of a subpopulation of HIV-infected CD4?T lym-
phocytes that harbors latent replication-competent proviruses.
Since no viral proteins are produced, the latently infected cells
cannot be recognized by the antiviral immune response and are
highly resistant to antiretroviral therapy. The development of
these latent and slowly replicating viral reservoirs during HIV
infections has immense practical consequences for treatment
of HIV infections because it provides a mechanism that allows
the virus to evade immune clearance and the effects of antiviral
drugs while still retaining an ability to quickly revert to the
productive state upon interruption of drug therapy or in re-
sponse to cellular activation signals (6, 17).
Multiple complementary mechanisms are required to silence
HIV transcription and permit its entry into latency. Although
HIV silencing can readily occur in transformed cell lines, sev-
eral features of the metabolism of resting CD4 cells ensure that
latent proviruses remain transcriptionally inactive for long pe-
riods. First, a key factor contributing to the restricted tran-
scriptional initiation that is characteristic of HIV transcrip-
tional silencing is the sequestration of the cellular initiation
factors NF-?B and NFAT in the cytoplasm of quiescent T cells
(28, 37). The second major transcriptional block seen in la-
tently infected cells is the incorporation of the P-TEFb elon-
gation factor into an inactive complex containing HEXIM and
7SK RNA (8, 56). This restricts P-TEFb levels in the cell and
creates a block to efficient transcription elongation from the
HIV promoter. In addition, posttranscriptional restrictions
further reduce HIV gene expression. For example, limiting
nuclear levels of the PTB splicing factor in quiescent cells leads
to a block to the export of HIV-specific RNA transcripts (32).
Finally, miRNAs that inhibit translation of HIV mRNAs may
also play an important role in maintaining HIV latency
Entry into latency is also strongly correlated with the recruit-
ment of histone deacetylases (HDACs) to the HIV long ter-
minal repeat (LTR) (9, 50). For example, we have recently
demonstrated that CBF-1 (for latency C-promoter binding fac-
tor 1), a DNA-binding protein that plays a central role in the
Notch signaling pathway, can direct transcriptional silencing of
the HIV LTR through recruitment of HDAC-1 (49). The
HDACs help to establish restrictive chromatin structures that
limit HIV transcriptional initiation and elongation. Additional
chromatin restrictions due to histone H3 methylation by his-
tone methyltransferase Suv39H1, lead to the accumulation of
* Corresponding author. Mailing address: Department of Molecular
Biology and Microbiology, Case Western Reserve University, 10900
Euclid Ave., Room W200, Cleveland, OH 44106-4960. Phone: (216)
368-3915. Fax: (216) 368-3055. E-mail: email@example.com.
† Present address: Department of Microbiology and Immunology,
Stanford University School of Medicine, Stanford, CA 94305-5107.
?Published ahead of print on 21 April 2010.
HP1 proteins on transcriptionally inactive proviruses (14, 35).
We have also been able to monitor the progressive HIV si-
lencing in isolated Jurkat T-cell clones and showed that latency
is associated with these chromatin modifications (40).
Although all silenced HIV proviruses appear to acquire re-
strictive chromatin structures near the viral promoter, the cel-
lular integration site can have a profound influence on the
extent of proviral silencing. Viral integration into actively tran-
scribed host genes can led to transcriptional interference
caused by the elongating RNA polymerase II (RNAPII) tran-
scribing through the viral promoter (15, 20, 33). Similarly,
integration into heterochromatic regions can accelerate provi-
ral silencing (25, 34).
Whereas there is compelling evidence that silencing through
histone remodeling is a key feature mediating the establish-
ment of HIV latency, the involvement of DNA methylation is
more controversial. Recent studies have shown that proviruses
that are poorly responsive to T-cell activation signals also tend
to be hypermethylated (26). However, in many silenced HIV
clones proviral expression does not correlate with DNA meth-
ylation (42). Thus, it seems likely that although DNA methyl-
ation is a powerful silencing mechanism for retroviruses, in the
natural setting, DNA methylation is not inevitably imposed
and partially methylated promoters can still be reactivated
Molecular studies of HIV latency have been severely ham-
pered by the absence of reliable cellular models. The rarity of
latently infected cells in patients (less than 1 in 106resting
CD4?T cells in the peripheral circulation) makes it almost
impossible to isolate them in sufficient numbers for biochem-
ical studies (41). As a result, virtually all molecular investiga-
tions of HIV latency have involved the use of transformed cell
lines, such as the popular Jurkat T-cell line which carries a
functional T-cell receptor (TCR) signaling apparatus (25, 30,
40). However, because the quiescent phenotype of the latently
infected CD4?T cells found in vivo is drastically different from
the replicating and constitutively activated Jurkat T cells, many
laboratories are working to develop more suitable experimen-
tal models for HIV latency using primary cells. A particularly
informative model system has been developed by Zack and
coworkers, who have effectively used the HIV SCID-hu (Thy/
Liv) mouse model to recapitulate the generation of latently
infected naive T cells during thymopoiesis both in vivo (4) and
in vitro (5). Similarly, Cloyd and coworkers have reported that
latently infected, quiescent CD4?T cells can be obtained by
cultivating HIV-infected, activated normal CD4?T lympho-
cytes on feeder cell layers (45).
In a significant recent set of advances, sophisticated cell
culturing techniques have been used to recapitulate T-cell dif-
ferentiation events in vitro and generate latently infected cells.
First, Bosque and coworkers (26) have taken advantage of
polarizing conditions to force activated T cells to differentiate
and enter quiescence. Similarly, Marini et al. (36) used low
doses of interleukin-7 (IL-7) to generate and maintain latently
infected memory CD4?T cells in vitro. Although promising,
both methods are of limited use for biochemical analyses of
latent proviruses because the yields of viable latently infected
cells are low. Greater quantities of latently infected cells have
recently been obtained by Yang et al. (53), who used Bcl2 to
partially transform primary CD4?T cells isolated from periph-
eral blood mononuclear cells (PBMC). We report here a novel
ex vivo method that permits, for the first time, the generation
of large populations of latently infected primary CD4?T cells.
We used this system to demonstrate that both creation of
heterochromatic structures on the HIV provirus and restric-
tions in P-TEFb levels contribute to the establishment of HIV
latency in primary cells.
MATERIALS AND METHODS
Cell culture and lentiviral vectors. CD4?primary T cells and H80 cells were
maintained in RPMI 1640 medium supplemented with 10% fetal calf serum with
or without IL-2 (20 U of recombinant human IL-2 [R&D Systems, Inc.]/ml). The
lentiviral vectors pHR?P-PNL-mCherry or pHR?P-PNL-d2EGFP vectors were
constructed and pseudotyped HIV particles were packaged as described previ-
Isolation, stimulation, and culture of CD4?T lymphocytes. The CD4?cells
were either isolated from the blood drawn from healthy donors or were isolated
from discarded tonsils. CD4?T cells were purified by negative selection method
using a MACS kit (Miltenyi Biotechnology, Auburn, CA). Purified CD4?T cells
(?98% pure) were stimulated for 4 days in RPMI 1640–10% fetal bovine serum
with 25 ?l of ?-CD3/CD28 antibodies conjugated to magnetic beads (Dynal
Biotech) per million cells, along with 20 U of IL-2/ml. One million cells were
infected with VSV-G-pseudotyped HIV viruses. After 2 days, the fluorescent
cells were purified by fluorescence-activated cell sorting (FACS) and again prop-
agated in the presence of ?-CD3/CD28 antibody-conjugated Dynal beads (25
?l/106cells) and 20 U of IL-2/ml for 2 to 3 weeks. Fresh medium was added every
4 days, and the cultures were maintained at a density of 1.5 ? 106to 2.0 ? 106
cells/ml. Once 0.5 ? 108to 1 ? 108cells were obtained they were placed on 30
to 40% confluent H80 cell layers (45). Every 2 to 3 days, half of the culture
medium was replaced by fresh IL-2-containing medium, and every 2 weeks the T
lymphocytes were transferred to the fresh flasks of H80 feeder cells.
Flow cytometry. Cells were analyzed for fluorescent reporter gene expression
FACSCalibur flow cytometer (Becton Dickinson, Franklin Lakes, NJ). Multi-
color flow analysis of cell surface marker expression was performed by using a
LSRII flow cytometer. Antibodies were conjugated to the fluorophores indicated
in the figure legends. Between 20,000 and 80,000 events were acquired for each
antibody and its appropriate isotype control. The data was analyzed with the
Staining of cells with BrdU and Ki67. Cells were labeled for 18 h with
bromodeoxyuridine (BrdU) using a commercial BrdU incorporation assay pro-
tocol (BD Pharmingen). Approximately 3 ? 105labeled cells were permeabilized
using 500 ?l of Cytoperm (BD Biosciences) for 10 min in the dark at 37°C. After
a washing step, the cells were incubated for 30 to 60 min with antibodies to BrdU
and ?-Ki67 or control antibodies. The unbound antibodies were then removed by
washing, and the cells were fixed and resuspended in 100 ?l of 1% paraformal-
dehyde in phosphate-buffered saline plus 200 ?l of wash buffer prior to FACS
ChIP assays. Chromatin immunoprecipitation (ChIP) was performed as pre-
viously described (49). To activate cells, we used either 10 ng of tumor necrosis
factor alpha (TNF-?)/ml or 25 ?l of ?-CD3/CD28 antibodies bound to Dynal
beads (Dynal Biotech) per 106cells. Most antibodies were purchased from Santa
Cruz, including anti-RNAPII (N-20), CBF-1(H-50), CIR (C-19), mSIN3A (AK-
11), HDAC-1 (H-51), HDAC-2 (H-54), and p65 (C-20). Anti-acetylated his-
tone-3 and histone-4 antibodies were obtained from Upstate.
Western blot analysis. Western blotting was performed according to standard
protocols. Anti-NF-?B, anti-CycT1, and anti-CDK-9 antibodies were obtained
from Santa-Cruz. Secondary horseradish peroxidase-conjugated anti-rabbit or
anti-mouse antibodies were from Dako.
Efficient generation of pure populations of latently infected
CD4?memory T cells. We developed a novel ex vivo model
system to study HIV latency in primary CD4?T cells. As
outlined in Fig. 1, pure populations (?97%) of primary CD4?
T cells were isolated from PBMC by using a negative selection
MACS kit from Miltenyi Biotech. The cells were then ex-
panded using ?-CD3/CD28 antibody-coated beads in the pres-
6426TYAGI ET AL.J. VIROL.
ence of IL-2. After 4 days, the cells were infected with HIV-
derived vectors (Fig. 2A) that express fluorescent protein
reporter genes (either the short-lived d2EGFP or mCherry) in
place of the nef gene, as previously described (27, 49). Like
HIV itself, the viruses included the regulatory proteins Tat and
Rev, which provide a positive feedback circuit that enhances
HIV transcriptional elongation and export of mRNA from the
nucleus. In the majority of our experiments, in order to in-
crease the frequency of latently infected cells in the popula-
tion, we utilized Tat carrying the H13L mutation. This partially
attenuated Tat variant was originally identified in the U1 la-
tently infected cell line (16, 44) and was earlier shown by us to
expedite HIV entry into latency (40). The use of an attenuated
Tat mutant is consistent with recent studies which have shown
that latently infected cells accumulate Tat variants with re-
duced transactivation potential (55). However, as described
below, latently infected cells can also be readily generated by
using wild-type Tat.
To obtain a homogeneous population of HIV-infected cells,
cells expressing the fluorescent reporters were purified by cell
sorting. The purified cells were further expanded with mag-
netic beads with ?-CD3/CD28 antibodies. After 4 to 6 weeks,
the cell population reached between 50 ? 106and 100 ? 106
cells; the magnetic beads were removed, and the cells were
placed on H80 feeder cells in the presence of IL-2, as originally
described by Cloyd and coworkers (45).
As shown in Fig. 2B, cultivation of primary T cells infected
with viruses carrying the H13L mutation in Tat on the feeder
layers led to the progressive loss of HIV gene expression and
the entry of the cells into a largely quiescent state character-
ized by a dramatic reduction in cell size. After 6 weeks, 92.13%
of the cells lost virtually all d2EGFP reporter gene expression
(Fig. 2B). As shown in Fig. 2B and 8B, comparison of the
d2EGFP profiles in the silenced cell population to uninfected
control cells shows that the silenced cells display a very low
level of fluorescence, suggesting that there are only minimal
levels of continuing transcription. Very few of the silenced cells
have lost the provirus, since most of them can be efficiently
reactivated by stimulation of the TCR (see Fig. 8B). Impor-
tantly, the quiescent T cells can be maintained on the H80
feeder cells for more than 6 months without any noticeable loss
of viability or reactivation capability.
The progressive silencing of HIV gene expression as cells
enter quiescence can also be observed with viruses carrying
wild-type Tat (Fig. 3A). In this experiment, 86.15% of the cells
carrying proviruses became silenced by day 63. After stimula-
tion of the TCR by antibodies to CD3 and CD28, 90.99% of
the cells were reactivated. As previously observed in our stud-
ies with Jurkat T cells (40), the silencing of proviruses carrying
the wild-type Tat is generally slower and less efficient than the
silencing of proviruses carrying attenuated Tat genes.
The silencing of HIV proviruses was also observed in CD4?
T cells derived from tonsil tissues infected with viruses carrying
either H13L Tat or wild-type Tat and the mCherry fluorescent
reporter (Fig. 3B). After 30 days, 52.35% of cells infected with
viruses carrying the H13L Tat mutation and 46.79% of cells
carrying wild-type Tat were silenced. Because cells derived
from tonsil tissues show variable degrees of activation and are
slower to enter quiescence than CD4?T cells isolated from
PBMC, the remaining experiments in the present study were
performed with PBMC.
Latently infected cells have resting central memory cell phe-
notype. In HIV patients undergoing HAART, the vast majority
of HIV-infected cells (over 89%) in the peripheral circulation
are resting memory CD4?T cells, although a small but signif-
icant fraction of infected cells (3%) have a naive CD4?T-cell
phenotype (2). To document the changes that occur within the
population during the expansion and subsequent establishment
of the latent virus, the cells were stained with multiple anti-
bodies to cell surface markers and examined using multicolor
FACS. The most striking feature of this analysis is that the
expansion and culturing on H80 cells results in a more homo-
geneous population than the CD4?cells originally isolated
from PBMC (Fig. 4).
As shown in Fig. 4A, freshly isolated T cells contain a mix-
ture of naive T cells (40.43%, CD45RA?CD45RO?), memory
cells (24.61%, CD45RO?CD45RA?), and a small population
(19.01%) of dual-positive cells which represent cells that re-
main in the transitional phase between naive and memory T
cells. In contrast, latently infected cells that have been cultured
on H80 cells for 6 weeks display a uniform CD45RA?
CD45RO?phenotype (92.98%), indicating that they are pri-
marily resting memory cells (Fig. 4A and Fig. 5B).
To further define the phenotypes of the latently infected cells,
we performed multicolor flow cytometric analysis (FACS) using
antibodies for a wide range of different cell surface protein mark-
ers (Fig. 4, 5, and 6). The cells shown in Fig. 4 and 5 were infected
with viruses carrying the H13L Tat gene.
As shown in Fig. 5A, 70.63% of the latently infected cells
used in these experiments did not express the d2EGFP marker
and the remaining 28.77% expressed only low levels of
d2EGFP. However, 82.34% of the cells resumed HIV tran-
scription and expressed high levels of d2EGFP after stimula-
tion of the TCR.
CD38 is only expressed at very low levels on naive T cells and
unactivated memory T cells, but is present at high levels on both
activated T cells and mature thymocytes (11). Consistent with
these observations, we observed that freshly isolated naive and
memory cells express only low levels of CD38 (Fig. 4C). In con-
trast, the latently infected memory (CD45RA?CD45RO?) pos-
itive cells showed uniformly high levels of CD38 (Fig. 4C, and
horizontal axis, Fig. 5A to F).
FIG. 1. Method for obtaining large populations of latently infected
primary CD4?T cells.
VOL. 84, 2010PRIMARY CELL MODEL FOR HIV LATENCY6427
Activation of the cells through the TCR also resulted in the
upregulation of the IL-2 receptor (CD25). In the experimental
results shown in Fig. 4B, 96.32% of the cells upregulated
CD25, whereas 82.91% of the cells were activated in the ex-
perimental results shown in Fig. 5C.
CD27 (Fig. 4E and 5D) is an important T-cell activation and
expansion marker, which is constitutively expressed on central
memory T cells and strongly upregulated after TCR activation
of naive cells (10). Upon acquisition of effector functions, the
levels of CD27 on central memory cells declines as CD27
becomes soluble and detaches from the cell surface (18). CD27
has also been linked with T-cell survival during expansion (21).
In the latently infected population of cells, there is a low level
of CD27 on the cells surface that is modestly reduced after
TCR activation (Fig. 4E and 5D). This is consistent with the
identification of the latently infected cells as resting central
The CCR7 receptors are the hallmark of activated central
FIG. 2. Progressive silencing of HIV expression in infected CD4?T cells. (A) Structure of lentiviral vectors. In some experiments, mCherry
was used in place of the d2EGFP fluorescent reporter depicted in this diagram. (B) Flow cytometric analysis of infected cells placed on H80 feeder
cells. The top panels show a light scatter analysis of the cell size. During proviral silencing, the activated T cells become quiescent and become
reduced in size, as indicated by reductions in both forward and side light scatter. The middle panels display selected histograms showing the fraction
of cells expressing d2EGFP at day 0, day 20, and day 49 after cultivation of sorted cells on the H80 feeder cell line. A gray histogram shows
uninfected control cells used to set the gates indicated by the bars. The percentage of cells in each gate for the various silenced cell populations
is given above the bars. The bottom panels show the time course of proviral silencing during cultivation of CD4?T cells on H80 feeder cells.
Histograms are shown on the left. The graph on the right shows the percent d2EGFP?cells (green line; ?2 ? 100) and the percent d2EGFP?
cells (black line; ?2 ? 100) for each time point (C).
6428TYAGI ET AL.J. VIROL.
memory cells since CCR7 enhances the retention of these cells
by lymph nodes (46). In contrast, effector memory cells do not
express CCR7 but do express other receptors that stimulate
migration to inflamed tissues (46). As shown in Fig. 4D and 5E,
CCR7 is upregulated on the latently infected cells, which is
again consistent with their identification as a resting central
memory cell population.
Finally, as d2EGFP levels increase (Fig. 5A), there is a
concomitant decrease in levels of CD4 (Fig. 5F). The down-
regulation of CD4 is probably due to the increased expression
of HIV proteins (Vpu and Env) from the latent proviruses.
Very similar patterns of surface marker expression were
observed in cells obtained from a second donor that were
infected with viruses carrying the wild-type Tat gene (Fig. 6).
In this cell population, 96.38% of the infected cells did not
express the d2EGFP marker, and the remainder only ex-
pressed low levels of d2EGFP. However, 66.45% of the cells
became strongly activated after stimulation of the TCR (Fig.
6A). As in the previous experiment, the latently infected
cells are all CD45RA?CD45RO?(Fig. 6B) and expressed
low levels of CD27 (Fig. 6D) and CCR7 (Fig. 6E), both of
which became somewhat upregulated after TCR activation.
The cells also showed strong upregulation of CD25 (Fig. 6C)
and strong downregulation of CD4 upon stimulation of the
TCR (Fig. 6F).
Latently infected cells show reduced DNA synthesis and cell
proliferation. To analyze the proliferation capability of the
quiescent cells, we compared the incorporation of BrdU into
cellular DNA before and after activating them with anti-CD3/
CD28 antibodies (Fig. 7A) (29). For control cells we utilized
freshly isolated, but not previously activated, CD4?T cells
isolated from PBMC. In parallel, we also performed intracel-
lular staining for the nuclear antigen Ki67 (Fig. 7B) and CCR7
(Fig. 7C). The IL-2 receptor chain, CD25, was included as an
additional dimension for the flow cytometry. Mock-infected
cells were used for these experiments to avoid interference of
the d2EGFP signal expressed by the lentiviral vectors with the
fluorescein isothiocyanate (FITC)-labeled antibodies utilized
in the BrdU incorporation assay.
As shown in Fig. 7, during cultivation on the H80 feeder
cells, the majority of the HIV-infected cells show strongly
reduced CD25 expression. In order to look for cells in the
population that were undergoing low levels of DNA synthesis,
we labeled the cells for 18 h with BrdU, an unusually long
labeling period compared to the more typical 2-h “pulse” that
is used in cell cycle studies. Even under these conditions, in the
latently infected cell population, 32.55% of the cells failed to
incorporate any BrdU (Fig. 7A). The remaining cells showed
measurable levels of BrdU incorporation. As expected, after
TCR activation 96.88% of the latently infected cells incorpo-
FIG. 3. Epigenetic silencing of HIV expression CD4?T cells obtained from PBMC and tonsil tissue. (A) Silencing and reactivation in CD4?
T cells from PBMC. (Left panel) CD4?T cells isolated from PBMC were infected with viruses carrying wild-type Tat and the d2EGFP reporter.
After sorting for d2EGFP?cells the population was cultivated on H80 feeder cells for up to 63 days. (Right panel) At day 63, the latently cells
were reactivated by stimulation of the TCR with ?-CD3/CD28 antibodies. During the next 5 days there was a gradual reactivation of the entire
latently infected cell population. (B) Silencing in CD4?T cells from tonsils. CD4?T cells were isolated from discard tonsils and infected with
viruses carrying either H13L Tat (left) or wild-type Tat (right) and the mCherry reporter. After sorting for mCherry?cells, the populations were
cultivated on H80 feeder cells for the next 30 days. During this period there was progressive silencing of the proviruses.
VOL. 84, 2010 PRIMARY CELL MODEL FOR HIV LATENCY6429