Human Immunodeficiency Virus Integration Efficiency and Site Selection in Quiescent CD4⁺ T Cells

Abstract

Until very recently, quiescent CD4⁺ T cells were thought to be resistant to human immunodeficiency virus (HIV) infection. Subsequent studies, attempting to fully elucidate the mechanisms of resistance, showed that quiescent cells could become infected by HIV at low efficiency and form a latently infected population. In this study, we set out to identify the sites of viral integration and to assess the efficiency of the overall integration process in quiescent cells. Based on our results, HIV integration in quiescent CD4⁺ T cells occurs in sites similar to those of their prestimulated counterparts. While site selections are similar, the integration process in quiescent cells is plagued by the formation of high levels of incorrectly processed viral ends and abortive two-long-terminal-repeat circles.

Quiescent CD4⁺ T cells have been shown to be resistant to human immunodeficiency virus (HIV) infection, and the resistance is characterized by incomplete reverse transcription (82, 83). However, the permissiveness of other nondividing cell types, such as resting T cells and macrophages, raised further questions regarding the nature of the block (10, 26, 56, 65, 67, 69-71). Later studies further elucidated which subsets of resting cells were refractory to infection. Truly quiescent cells in the G_0/1a phase were resistant to infection, while cells in the G_1b phase, characterized by high levels of RNA synthesis but not DNA synthesis, were susceptible to infection (43). These combined studies suggested that certain nondividing cells could support a productive infection.

Subsequent studies were aimed at further examining the different steps of the viral life cycle in quiescent cells, as well as identifying potential cellular factors or the lack thereof that may block infection. One potential block to infection was hypothesized to be the lack of raw materials such as nucleotides. Treatment of quiescent cells with nucleosides resulted in increased reverse transcription but did not lead to a productive infection following stimulation of the cells (42, 61), suggesting that other factors may contribute to the resistance to productive infection. Recent studies examined the presence or absence of cellular factors responsible for the block in quiescent CD4⁺ T cells. Both Murr1 and APOBEC3G have been shown to influence the viral life cycle in quiescent CD4⁺ T cells (18, 29). However neither factor fully explains the lack of rescue of productive infection if quiescent cells are stimulated subsequent to infection.

A more detailed examination of the viral life cycle in quiescent CD4⁺ T cells can shed more light on the nature of the block presented by quiescent CD4⁺ T cells. A series of recent studies looking at the different stages of the HIV life cycle in quiescent cells have further supported data from our earlier work (82, 83). More specifically, it has been shown that reverse transcription is inefficient in quiescent cells, generating full-length viral transcripts that are very labile (half-life of 1 day) but are integration competent (60, 84). However, whether there is integrated provirus in these cells has not been determined. In another study, provirus can be found in resting cells of HIV-infected individuals, but this was attributed to previously activated cell populations that returned to a resting state after stimulation, such as memory T cells (31, 35). The development of more-sensitive Alu PCR-based assays allowed for the detection of very low copy numbers of integrated HIV (2, 57). Several studies showed, using alternative methods of infection, such as spinoculation, that a latent infection can be established in quiescent CD4⁺ T cells (58, 73, 74). Our group undertook a comprehensive study and examined multiple stages of viral replication (entry, reverse transcription, integration, viral gene transcription, and viral protein synthesis) in quiescent cells and compared them to those of stimulated cells. Quiescent CD4⁺ T cells were infected and then immediately stimulated to determine if this will rescue infection. Replication in these cells stimulated immediately after infection was characterized by a long delay in reverse transcription and integration compared to that in prestimulated T cells (78). Reverse transcription was largely inefficient (30-fold less than stimulated cells), integration efficiency was slightly decreased (twofold) and protein expression was very poor. Thus, immediate stimulation failed to effectively rescue infection of quiescent cells. Interestingly, their kinetics of infection was very similar to that of infected quiescent cells that remained unstimulated. These data suggested that a major block of HIV infection in quiescent T cells may be due to the slow kinetics of the early stages of infection.

The presence of provirus in quiescent CD4⁺ T cells that do not or poorly produce virions can be attributed to several factors. The presence of defective genomes can influence viral gene expression (19, 20). In addition, defects in viral mRNA expression (1, 29, 33, 36-38, 44, 55), lack of NF-κB and NFAT transcriptional factors in the nucleus (7, 76), and poor viral RNA transcript stability (44, 45) can adversely affect viral gene expression.

Since the transcriptional activity of identical proviruses at different chromosomal locations can vary (1, 29, 33, 36-38, 44, 55), the sites of HIV integration in quiescent CD4⁺ T cells may play a very important role in viral genome expression. Extensive work on HIV integration has shed light on the distribution and patterns of viral integration (15, 23, 37, 38). A series of elegant studies have shown that HIV integration is not tightly sequence specific but not random (8, 12, 14, 15, 23, 37, 38, 47, 51, 54, 79). HIV integration sites tend to be within active transcription units and gene-rich regions and are less favored into centromeric heterochromatin (8, 11, 12, 23, 37, 38, 51, 53, 54, 66, 79, 80). The mechanisms of viral integration have not yet been fully elucidated. While a host of cellular and viral proteins seems to influence the process, a number of studies have implicated LEDGF/p75, a cotranscriptional activator, to play a major role. LEDGF/p75 has been shown to interact with integrase within the context of a larger complex (3, 17, 21, 22, 27, 49, 50, 62, 68). Elegant knockdown studies have demonstrated the importance of LEDGF/p75 in integration targeting and integrase stability (27, 48, 50, 62, 77). Additional studies have explored the link between chromosomal DNA structure and integration, suggesting that nucleosome-bound DNA may be more susceptible to integration due to conformational changes (16, 25, 63, 64, 79).

Here, we present a study comparing the patterns of integration in quiescent CD4⁺ T cells with those in their stimulated counterparts. The lack of viral protein expression in latently infected quiescent T cells suggests that the integration sites may be distinct in this cell subset. T-cell quiescence is not a default but rather an actively maintained state regulated by lung Krüppel-like factor (LKLF) and Tob (24, 30). Thus, patterns of cellular gene expression in quiescent CD4⁺ T cells could result in different integration patterns, which could impart the ability to rescue virus from quiescent CD4⁺ T cells. To address the above, we isolated quiescent and stimulated CD4⁺ T cells and infected them with the CXCR4-tropic HIV type 1 molecular clone NL4-3 (HIV-1_NL4-3). HIV integration sites were sequenced and mapped. Based on our results, the chromosomal features associated with integration between the two cell types were very similar, but the integration process was less efficient in quiescent cells and characterized by a high frequency of incorrect integration and abortive circular viral forms.

MATERIALS AND METHODS

Cell lines, primary cells, and virus stocks.

CEM is a human T4 lymphoblastoid cell line (ATCC number CCL-119). Human peripheral blood mononuclear cells (PBMC) were obtained from healthy HIV-seronegative donors, separated over a Ficoll-Hypaque gradient, and cultured in RPMI 1640 medium supplemented with 10% fetal calf serum (Gibco-Invitrogen, Carlsbad, CA), 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM l-glutamine. The HIV molecular clone NL4-3 was used in these studies. Virus stocks were generated from 24-h harvests of supernatants from CEM cells, electroporated with full-length infectious cloned viral DNA, and further amplified on CEM cells. Supernatants collected were cleared by centrifugation and treated with DNase I (1 μg/ml) in the presence of 10 mM MgCl₂ for 30 min at room temperature. Subsequently, supernatants were concentrated 10- to 20-fold using Amicon Ultra-15 centrifugal units (minimum molecular-weight retention, 100 kDa) (Millipore, Billerica, MA) according to the manufacturer's instructions. Titers of the viral stocks were determined by limiting dilution assays using CEM cells.

Isolation of quiescent CD4⁺ T cells.

Human PBMC were incubated with a cocktail of mouse monoclonal antibodies against human lymphocyte markers to remove activated CD4⁺ T cells (CD25, CD38, CD69, HLA-DR) and unwanted cell lineages, such as CD8 T cells (CD8), macrophages (CD14, CD16), B cells (CD19), granulocytes (CD123), and NK cells (CD56). All antibodies were purchased from BD Biosciences (BD Biosciences, San Diego, CA). Cells stained with the above antibodies were removed after incubation with magnetic beads coated with goat antibodies against mouse immunoglobulin G (IgG; Miltenyi Biotech, Auburn, CA) and separated using an autoMACS (Miltenyi Biotech) cell sorter. To assess the purity of the quiescent cell population, 5 × 10⁴ cells were stained against the markers listed above, using phycoerythrin- or fluorescein isothiocyanate-conjugated monoclonal antibodies, as previously described. Ten thousand events were acquired on a FACSCalibur flow cytometer (BD Biosciences). Live cells were gated by using forward-versus-side-scatter dot plots. Data were analyzed by using the CellQuest program (BD Biosciences). Quiescent cells contained less than 1% contaminating cell populations. The above-described protocol was used to determine the levels of CD25, CD69, and HLA-DR on stimulated cells.

Infection of quiescent and stimulated CD4⁺ T cells.

The quiescent and stimulated cell populations were infected by incubation at a multiplicity of infection of 1 (unless otherwise specified) for 1 to 2 h with virus in the presence of Polybrene (10 μg/ml). Cells were washed to remove residual free virus and recultured under the appropriate conditions. Cells were stimulated using plate-bound anti-CD3 (OKT3, 1 μg/ml) and soluble anti-CD28 (25 ng/ml). Some cells were cultured in the presence of zidovudine (Sigma-Aldrich) to serve as negative controls. To prevent virus spread and achieve a single-cycle infection, cells were treated with a 100 nM concentration of a protease inhibitor (indinavir) following infection.

Viral DNA analysis.

Approximately 0.3 × 10⁶ cells from under each condition were used for reverse transcription and integration assays. Quantitative real-time DNA PCR was performed as previously described using a primer/probe pair that amplifies cellular β-globin sequences and one that amplifies full-length HIV reverse transcripts (LTR/gag junction), formed near the completion of the reverse transcription process. All amplifications were performed on the ABI 7700 system (Applied Biosystems, Foster City, CA). All amplifications were performed in parallel with a set of known quantitative standards. The standard curve used to determine HIV DNA levels ranged from 10 to 20,000 copies of cloned HIV DNA. The standard curve used to determine the levels of β-globin gene sequences consisted of DNA derived from 10 to 100,000 normal human peripheral blood lymphocytes. Quantitation of HIV-1 sequences was achieved by extrapolation from these standard curves. Detection and quantitation of integrated viral DNA were assessed by Alu PCR as previously described (57, 78). Briefly, we performed two nested PCR amplification steps. In the first preamplification step, we used primers to Alu and gag. This was followed by real-time PCR using internal primers and probes to the LTR/gag junction. To control for any nonintegrated viral DNA, we performed the preamplification step with only the gag primer (linear preamplification of DNA) or without any primers (no preamplification of DNA). A standard curve representing integrated HIV sequences was generated from cells infected with a nonspreading HIV-based reporter vector. In control experiments, there was no background from nonintegrated viral DNA, and values for integrated DNA copies varied by less than 20% within triplicates. The β-globin standard was used to determine the approximate number of proviruses per cell.

Cloning and sequencing of HIV-1 integration sites and 2-LTR junctions.

Provirus-host junction sequences and two-long-terminal-repeat (2-LTR) junction sequences were cloned and sequenced by using a modified inverse PCR strategy (see Fig. 2A) (19, 41). Briefly, genomic DNAs from quiescent CD4 T cells or stimulated CD4 T cells were digested with BamHI that cuts once at position 8466 of the HIV-1 (NL4-3) genome to fragmentize large genomic DNA. After 94°C denaturation of DNA, a biotinylated (B) primer, BNL9005F (5′ B-GGTACCTTTAAGACCAATGA), was annealed at nucleotide position 9005 to 9024 of virus DNA and extended using Taq DNA polymerase and deoxynucleoside triphosphates. Biotin-labeled DNAs were then enriched by binding to streptavidin-agarose Dynabeads as described in the company manual (Dynal magnetic beads; Invitrogen). The isolated DNA was digested with TaqαI (5′-T⁶CGA), which cuts an average of 250 bp in the human genome and only once within the right LTR (nucleotide [nt] 9411 in U3 LTR of HIV-1), and was then washed from the beads and ligated under conditions favoring intramolecular ligation. Circularized DNA was amplified by a two-step PCR. The first PCR was carried out using primers NL495F (5′ CTGGCTAACTAGGGAACCCACT) and NL9560R (5′ CAGGCTCAGATCTGGTCTAA) in 300 μl PCR mixture with 0.5 μM of each primer, 0.2 mM of deoxynucleoside triphosphates, and 6 units of Taq DNA polymerase under the following conditions: 2 min of preincubation at 94°C, followed by 29 cycles at 94°C for 30 s, 58°C for 30 s, and 72°C for 3 min. Amplified DNA was purified using a PCR purification kit and eluted into 50 μl EB buffer (Qiagen). A total of 0.5% (vol/vol) of eluted DNAs was subjected to the second PCR, in which conditions were identical to those of the first PCR except that it was conducted with only 15 cycles and with two nested primers, NL510F (5′ CCCACTGCTTAAGCCTCAAT) and NL9433R (5′ GAAAGTCCCTTGTAGCAAGC), which anneal downstream of the NL495F and upstream of NL9560R binding sites, respectively. The second PCR product was visualized on 2% agarose gels, and diffuse bands between 100 bp and 2 kbp were extracted for cloning and sequencing.

FIG. 2.

Genomic site preferences and orientation of integration sites in quiescent CD4⁺ T cells. (A) Integration sites were found within active genes (**, P < 0.0001 compared to random integration). In addition, they showed preference to regions within 10 kb of transcriptional start sites (TSS) and CpG islands (*, P < 0.01) over those within Alu and Line-1 (L1) elements. The patterns were similar between quiescent and stimulated CD4⁺ T cells. The percentages are based on the total number of true integration events. (B) Orientation of integration sites in quiescent and stimulated CD4⁺ T cells. The percentages are based on the total number of true integration events. RefSeq, Reference Sequence.

Sequence analysis and mapping integration sites.

Integration site sequences were authenticated when (i) a genomic DNA sequence is flanked by the downstream sequence of NL510F binding site in the U5 LTR and the upstream sequence (TaqαI site) of NL9433R binding site in the U3 LTR and (ii) the genomic DNA sequence matches with 96% or greater identity to the human genomic sequences. The integration site sequences were mapped onto the human genome by using the UCSC Genome Browser (the March 2006 human reference sequence, NCBI build 36.1). All the genomic feature data sets were downloaded from the UCSC genome database (http://www.genome.ucsc.edu). HIV-1 integration sites and 10,000 random integration sites were analyzed as described previously (40). 2-LTR junction sequences (with normal or abnormal formations) were authenticated when at least 15 bp U3 LTR sequences of the region immediately upstream of the TaqαI site in U3 are flanked by the downstream sequence of NL510F binding site in the U5 LTR and the TaqαI site in the U3 LTR. For a statistical analysis, we used chi-square approximation or Fisher's exact test when individual cell counts in the row-by-column (r × c) contingency table were small (<10). Maps comparing the identified integration sites with the distribution of histone methylation patterns, RNA polymerase II, the insulator binding protein CTCF, and DNase hypersensitive sites in human CD4⁺ T cells were generated in the Genome Browser (http://www.ucsc.genome.edu) by obtaining the appropriate tracks (http://dir.nhlbi.nih.gov/papers/lmi/epigenomes/hgtcell.aspx and http://research.nhgri.nih.gov/DNaseHS/May2005/#V) (4, 9) and converting our integration data to the appropriate BED files for Genome Browser viewing.

Bioinformatics analysis.

Quiescent and stimulated T cells from different donors were assayed in 10 μg of leukocyte total RNA, using a high-density oligonucleotide array (U133a; Affymetrix, Santa Clara, CA), at baseline levels and after pooling standard deviation for genome-wide mRNA expression profiles. All assays were performed by the UCLA DNA Microarray Core (Affymetrix) following the manufacturer's standard protocol, with sample quality assured for RNA concentration and purity by using a bioanalyzer (Agilent, Palo Alto, CA) prior to probe synthesis. Following the hybridization of fluorescent cRNA probes, microarrays were imaged using a scanner (Affymetrix), and low-level gene expression values were derived using GeneChip operating software (GCOS) (Affymetrix) and analyzed using dChip software (Harvard School of Public Health). Invariant set normalization was used to normalize arrays at the probe level, and the model-based method was used for calculating expression values. The ontology of genes hosting integration sites was analyzed using GOStat and EASE 2.0 (6, 34).

Nucleotide sequence accession number.

The raw data for transcriptional profiling and integration sites have been deposited in the NCBI Gene Expression Omnibus under accession number GSE14596.

RESULTS

Identification of HIV integration sites in quiescent and stimulated CD4⁺ T cells by inverse PCR.

Quiescent and stimulated CD4⁺ T cells were infected with HIV-1_NL4-3 (multiplicity of infection of 1) and used to examine multiple stages of viral replication (reverse transcription, integration, viral RNA transcription, and protein synthesis). Compared to stimulated cells, viral replication in quiescent cells was characterized by inefficiency and long delays in all stages of the viral life cycle (78). More specifically, we observed a significant relative decrease in the levels of reverse transcription (30-fold). When taking into account differences in reverse transcription, integration efficiency was decreased approximately twofold in quiescent cells. In addition, there is a large decrease in the amount of multiply spliced (tat/rev) viral mRNA in quiescent cells (78). This suggests that while total integration events were decreased, the relative efficiency levels of the integration process itself were comparable between the two cell types. In addition, reverse transcription was delayed by 14 h compared to stimulated cells, integration by 30 h and viral RNA synthesis by 24 h (assessed by the levels of multiply spliced tat/rev viral mRNA) (78). Most intriguing was the lack of viral protein expression in quiescent T cells despite the presence of integrated provirus, suggesting possible defects in either integration site selection or postintegration events.

This led us to examine the patterns of integration in quiescent CD4⁺ T cells and whether they are distinct from stimulated CD4⁺ T cells. To this end, we infected quiescent and stimulated CD4⁺ T cells from six different donors and extracted total DNA to examine the distribution of integrated HIV. Cells were harvested at their peak levels of integration following a single cycle of infection (based on the kinetics described above). Therefore, we harvested quiescent cells at 48 h and stimulated cells at 24 h postinfection. Quiescent CD4⁺ T cells remained quiescent during the incubation period, based on flow cytometry analysis of activation marker levels (CD69, CD25, and HLA-DR) and cell cycle progression (data not shown). Levels of viral DNA synthesis and integration were assessed by real-time PCR and an Alu-based real-time PCR assay, respectively (data not shown). The identification of HIV integration sites was determined by an inverse PCR assay (Fig. 1A) (19, 41). Based on the sequencing analysis, we identified a total of 451 integration site sequences in quiescent cells (G₀) (Table 1). Furthermore, we found 16 sites self-integrated into the provirus genome, 269 2-LTR circles, and 109 sites that were classified as unidentified because they contained sequences corresponding to the junction of the left LTR and primer binding site, which can be derived from 1-LTR, 2-LTR or linear viral DNA (integrated and unintegrated) (Table 1). In the stimulated cell group (Stim), we identified 543 integration sites, 38 autointegrations, and 107 2-LTR circles and isolated 190 unidentified sites (Table 1).

FIG. 1.

Integration site sequence isolation and identification. (A) Schematic diagram of the inverse PCR assay used to isolate integration sites. (B) The different forms and proportions of LTR-host junctions observed in quiescent (G₀) and stimulated CD4⁺ T cells (Stim). Distributions are represented as a percentage of the total integrated sequences (Provirus) or total sequences (integrated and nonintegrated) analyzed (Total).

TABLE 1.

Identification of isolated integration site sequences

Sample	No. of indicated sequencesa
	Integration	Autointegration	2-LTR	Unidentified	Total
G0	451	16	269	109	845
Stim	543	38	107	190	878

^aSequences labeled as “Unidentified” contain DNA sequences corresponding to the junction of the left LTR and primer binding site, which can be derived from 1-LTR, 2-LTR, or linear viral DNA (integrated and unintegrated). Sequences labeled as “2-LTR” are sequences flanked by a U5 LTR end and a TaqαI digestion site in U3 LTR. Sequences in which integration into a provirus genome occurred were classified as “Auto-Integration,” and sequences in which correct integration into the host genome occurred were labeled as “Integration”.

Further analysis of the integration sites (Fig. 1B), revealed that 50 out of the 451 integration site sequences (11.1%) found in G₀ cells and 23 out of the 543 (4.2%) in Stim cells contained abnormal LTR-host junctions. These sequences were removed from our integration site characterization. A closer look at the remaining sites yielded 189 out of the 401 integration sites in G₀ cells, as unique sequences mapped onto the human genome with completely processed LTRs (Table 2). From those, 180 were sequences with a unique location, and they were used for further analysis of the HIV integration sites. In Stim cells, 342/520 were unique sequences, of which 318 were with unique locations.

TABLE 2.

Identification of true integration sites

Sample	No. of duplicated sequences	No. of unique sequences witha:		Total
		Unique locations	Multiple/unclear locations
G0	212	180	9	401
Stim	178	318	24	520

^aUnique sequences from the integration site sequences with completely processed LTRs were mapped onto the genome. Sequences with a unique genomic location (Unique locations) were further analyzed for the characterization of virus integration sites.

HIV integration occurs within transcriptional units.

Following the identification of HIV integration site sequences in G₀ and Stim cells, we went on to characterize the patterns of viral integration in the two cell types. As shown in Fig. 2A, the majority of integration sites were within active transcriptional units as has previously been shown for activated cells and T-cell lines (11, 13, 14, 23, 38, 66, 79). In addition, integration events not found within genes were located in different regions of the human genome, including areas within the first 10 kb of transcriptional start sites and CpG islands and within Alu and Line-1 elements. Interestingly, these patterns of integration were alike and were not statistically significant between the G₀ and Stim cells, but they were statistically significant against our random integration controls. In addition, we examined the orientation of integrants. This can be an important factor in the expression of the latent reservoir (32) and could be hypothesized to explain the lack of protein expression seen in G₀ cells. However, we did not observe any difference in the orientation of integrated provirus between the two cell types (Fig. 2B). Therefore, based on the data seen, HIV integration site selection in quiescent CD4⁺ T cells is similar to that in stimulated cells, suggesting that the nature of the transcriptional units may influence provirus expression in quiescent cells.

Identified transcription units are unaffected by T-cell stimulation.

The similarities found in the integration patterns between stimulated and quiescent CD4⁺ T cells prompted us to examine the expression levels of genes that hosted HIV integration sites in both cell types. Quiescent T cells actively maintain their phenotype by the expression of a set of genes regulated by LKLF and Tob (24, 30). In addition, transcriptional activity in these cells is lower than that in activated cells, and chromatin is tightly packed (5, 28). Therefore, it would be expected that the type of genes in which the provirus is found might be different than those in activated T cells. To this end, we examined the relationship of our integration sites to the transcriptional activity of quiescent and stimulated CD4⁺ T cells. We collected the genes from our integration site analysis, and we extracted their expression values from microarray analysis data collected in a separate experiment.

In G₀ cells, 31 out of 180 (17%) genes hosting provirus were not expressed in either quiescent or stimulated cells. The remainder of the genes were expressed in both cell types. Interestingly, the expression levels of the majority of these genes were not specifically altered with stimulation (Fig. 3A). More specifically, expression of the majority of the genes hosting integration events did not vary by >2-fold between the two cell types. The analysis of genes hosting integration sites in stimulated CD4⁺ T cells showed that 55 out of the 318 (17%) genes were not expressed in either stimulated or quiescent cells. Similarly, the genes hosting integration sites in stimulated cells were not significantly affected by T-cell stimulation (Fig. 3A). Thus, while HIV does integrate in transcriptionally active units in quiescent CD4⁺ T cells, these genes are not restricted to the ones that regulate the cells' activation state. In addition, integration into silent genes does not occur preferentially in quiescent CD4⁺ T cells.

FIG. 3.

Genes hosting HIV integration do not change with stimulation. (A) The genes hosting HIV integration events in both quiescent and stimulated cells were extracted from microarrays to examine their expression levels. Scatter-plot presentation of the expression values for all probe sets derived from genome-wide microarray expression data of indicated cell types. (B) Same sets of genes were extracted from microarrays of HEK cells and compared against quiescent and stimulated CD4 T cells. Over a third of the genes were upregulated in T cells. Green triangles represent genes with a threefold-increased gene expression or higher, red squares those with a two- to threefold change, and blue diamonds those with a one- to twofold change. The black spots represent the total gene population.

Since the expression of the genes hosting provirus is not linked to cell activation state, we examined the gene ontology of these genes. Analysis revealed that most of the genes belonged to groups related to housekeeping cell functions, such as ubiquitin activity, nuclear transport, and nucleotide metabolism (see Table S1 in the supplemental material). Also, we examined whether the genes hosting integration sites were lymphoid specific or had a lymphoid trend (Fig. 3B) by looking at the expression patterns of our gene lists in HEK cells, a human kidney cell line, relative to those of our stimulated and quiescent T-cell data sets. Approximately 30% of the identified genes in both T-cell types were lymphoid specific. Therefore, integration in quiescent CD4⁺ T cells occurs in actively transcribing genes, which are not distinct from stimulated cells; however, many appear to be specific to hematopoietic cells.

Chromosomal mapping of genes.

In addition to gene expression analysis, we examined the chromosomal distribution of the integration sites. As shown in Fig. 4A, overall integration occurs in gene-dense regions in both subsets of cells. The majority of integration events were observed in chromosomes 1, 3, 6, 11, 16, 17, and 19. In addition, in quiescent CD4⁺ T cells, we saw a preference for chromosomes 16 and 17, while, in stimulated cells, we observed a preference for chromosomes 11 and 19. DNase-hypersensitive sites have been mapped on resting and activated CD4⁺ T cells (9) as well as markers for transcriptionally active chromatin. We used the UCSC Genome Browser and constructed a series of tracks that included our integration sites and a series of chromatin modifications. In both cell types, we observed a strong association that was statistically significant with markers of transcriptionally active chromatin such as methylation of H3K4 (P < 0.0001) and DNase-hypersensitive regions (P < 0.01) and a negative association with H3K27 trimethylation (P < 0.0001), a marker associated with heterochromatin (4, 9). In addition, we observed a strong association with regions of steady-state RNA expression as demonstrated by bound RNA Pol II (P < 0.0001) (4). In Fig. 4B, we show a representative example of chromosome 11 that illustrates the above-described trends. All chromosomes annotated with our integration sites and chromatin modifications are shown in Fig. S2 to S25 in the supplemental material. There were a number of chromosomes that did have some gene-rich regions but did not host any integration events. This may reflect the somewhat limited number of sites examined; however, as a trend, it seems that chromosomes that had decreased amounts of DNase I-hypersensitive sites were the ones that did not have any or very few integration events. Therefore, HIV integration events in quiescent cells are strongly associated with regions of high transcriptional activity.

FIG. 4.

Chromosomal distribution of integration sites. (A) The integration sites identified in quiescent (G₀) and stimulated (Stim) cells were mapped onto the human genome, along with the gene density and cytoband patterns. PBMC stands for all HIV integration events identified in PBMC to date. (B) Chromosome 11 as a representative sample, indicating the preferences of HIV integration in transcriptionally active regions. The top row indicates the corresponding chromosome analyzed, while the remaining rows indicate the integration sites from each group (Quiescent and Stimulated), gene density (RefSeq genes), and the markers of chromatin transcriptional activity (CTCF, Pol II/RNA polymerase II, H2A.Z, H3K4me3, H3K4me2, H3K4me1, H3K27me3, CGIs, CpG Islands, Duke DNase Sig, and Duke DNase Sites).

Quiescent CD4⁺ T cells display a higher frequency of 2-LTR circles with abnormal junctions.

The integration site assay strategy that we undertook enabled us to perform a semiquantitative comparison of integrated and unintegrated viral DNAs as well as a comparison of different types of 2-LTR forms, normal and abnormal. During our integration analysis, we came across a significant number of sequences that were 2-LTR circles. The presence of these abortive integrated species provided us with valuable information regarding the efficiency of integration in HIV-infected quiescent and stimulated CD4⁺ T cells. As shown in Fig. 5A and B, while the number of 2-LTR circles formed in quiescent CD4⁺ T cells was significantly higher than that in stimulated cells (31.8% of the total sequences for the quiescent cells and 12.2% of the total sequences for the stimulated cells [P < 0.0001]), the proportion of 2-LTR circles with normal junction in quiescent cells (0.7%) was significantly lower than that in the stimulated cells (20.6% [P < 0.0001]), and the proportion of 2-LTR circles with abnormal junctions was significantly higher in quiescent cells (99.3%) versus in stimulated cells (79.4% [P < 0.0001]).

FIG. 5.

Increased levels of abnormal 2-LTR circles in quiescent CD4 T cells. (A) Semiquantitative assessment of levels of 2-LTR circle formation in G₀ and Stim cells and distribution of the different 2-LTR species identified. The distribution of 2-LTR circles is represented as a percentage of the total 2-LTR (2-LTR) and as a percentage of the total sequences analyzed, integrated and nonintegrated (Total). A small percentage of LTR junctions had insertions, some with truncated ends that were from neither the primer binding site nor the polypurine tract (Others). (B) Normal 2-LTR circles were more prevalent in Stim cells, while the abnormal 2-LTR circle species were more prevalent in G₀ cells (*, 0.01< P < 0.001; **, P < 0.0001). The distribution of 2-LTR circles is represented as a percentage of the total 2-LTRs (% of total 2-LTRs) and as a percentage of the total sequences analyzed, integrated and nonintegrated (% of total sequences). The percentages for LTR truncations were derived from all truncated ends. (C) A proposed model of the HIV reverse transcription and integration events based on our integration site and 2-LTR circle analysis.

The results indicate that the cellular environment of the quiescent cells promotes the generation of viral DNAs with abnormal LTR ends through insertions or deletions of nucleotides that would preferably serve as substrates for the 2-LTR circle formation rather than those for the integration reaction with cellular chromosomes. Previous studies also demonstrated that when RNase H or integrase activity was moderated by mutations or inhibitory drugs, the number of 2-LTR circles can increase with different frequencies for different types of 2-LTR circles (39, 52, 59, 72).

The major types of abortive insertions were derived from a part of a sequence or a full sequence of viral polypurine tract or the primer binding site, as well as sequences adjacent to it, which likely result from aberrant cleavages by RNase H during reverse transcription. The portion of 2-LTR circles with insertions at the junction was significantly reduced from 23.0% for quiescent cells to 8.4% for stimulated cells (P = 0.0001) (Fig. 5B). In addition, the portion of 2-LTR circles with truncated LTRs (truncation ranging from 1 to 321 bases from the end) was also significantly higher in quiescent cells (92.9%) than in stimulated cells (77.6% [P < 0.0001]) (Fig. 5B). A high tendency of LTR truncation in quiescent cells may be explained by the elevated host nuclease activity or an extended exposure to the host nucleases resulting from inefficient nuclear localization or cellular conditions unfavorable for viral integration into the host chromosome.

Based on the above-described data, we can propose a model of HIV integration in quiescent and stimulated CD4⁺ T cells (Fig. 5C). While site selections are quite similar between the two cell types, the process in quiescent cells is riddled with processing errors, which do occur in stimulated cells but are exaggerated in quiescent T cells. These errors result in the formation of LTR circles, many of them abnormal. This may partially explain the lack of protein expression in quiescent T cells and the poor rescue observed when they are immediately stimulated after infection (78).

DISCUSSION

The study of HIV infection in quiescent CD4⁺ T cells has made significant advances in recent years. Studies have examined closely and in more detail, thanks to the development of more-sensitive techniques, the different stages of the viral life cycle (2, 45, 57, 60, 61, 73-75, 78, 84). Recent advances in the field suggest that HIV infection in quiescent cells is very inefficient (2, 45, 60, 61, 73-75, 78, 84). While viral entry is not significantly affected, reverse transcription is severely inhibited, contributing to the lack of a productive infection (2, 73, 74, 78). Interestingly, the full-length viral products do integrate in quiescent cells, making these cells another potential viral reservoir (2, 73, 74, 78). Yet, viral production is very poor (73-75, 78). Based on these findings from our group and others, we decided to examine the patterns of viral integration in quiescent CD4⁺ T cells and how they compare to those in activated T cells. Furthermore, we examined the state of all viral DNA species involved in integration, including LTR circles. Based on our results, the patterns of integration in quiescent T cells seem to follow the trends seen in other studies. However, the process of integration in these cells is quite inefficient and is characterized by abnormal processing of the LTR ends as well as increased formation of LTR circles. This could contribute to the twofold decrease that we observed in the levels of integration (2, 73, 74, 78).

In our study, quiescent and stimulated CD4⁺ T cells were infected and used to examine the patterns of viral integration. This is the first study in which HIV integration sites were examined in CD4⁺ T cells while in their quiescent state. Most studies have focused on viral integration sites in patient samples receiving highly active antiretroviral therapy or those that have been infected over a long period of time (31, 32, 81). Furthermore, while these cells were at a resting state when isolated from the infected subjects, it is plausible that they were infected while activated and reverted to quiescence, as they became memory cells. However, our studies suggest that these cells may have been infected while in a resting state. It is plausible that quiescent cells in blood and lymph nodes encounter free or cell-bound virions, leading to infection. We undertook our in vitro analyses because it is of interest to examine where the virus integrates in the absence of selective pressures appearing in vivo or during extended culturing, as this may provide us with insight to the generation of a latently infected cell type.

The identification of integration sites was made possible by the use of an inverse PCR method. Based on this method, we were able to identify 180 unique locations at which HIV was integrated in quiescent cells and 318 in stimulated cells. In both cell types, integration sites were found mostly within transcriptional units, had no preference in terms of orientation, and exhibited similar patterns in sites outside transcriptional units. Further analysis revealed that the transcriptional units were specific for neither stimulated nor quiescent CD4⁺ T cells. Rather, these are genes that are not specifically affected by T-cell costimulation. This was a bit unexpected since quiescent cells actively maintain their phenotype by the expression of LKLF-regulated genes (24, 30). None of these genes were targets for integration in this system. Therefore, HIV integration in quiescent cells occurs in genes that are not specifically upregulated in quiescent cells or upregulated after costimulation.

In addition, we looked at the chromosomal distribution of these integration sites. HIV, as expected, was integrated in gene-dense regions in both quiescent and stimulated cells. While there were some minor preferences by each group, the distribution was overall balanced. A preliminary look into the gene composition of some of the regions revealed that chromosomal locations hosting integrated virus were made up of genes involved in immune functions. However, more importantly, the distribution of HIV integration is strongly associated with chromatin structure as related to transcriptional activity (11, 46, 66, 79). Based on studies done on resting and stimulated CD4 T cells mapping DNase-hypersensitive sites as well as markers of transcriptionally active chromatin (4, 9), we assembled chromosomal maps by using custom tracks for the above-described markers and our integration sites. Based on the analysis, there was a strong association with markers of transcriptionally active chromatin and DNase-hypersensitive sites. Furthermore, there was a strong correlation with steady-state RNA levels, as indicated by Pol II binding. While there was a higher number of integration sites in certain chromosomes, the low number of events we had for analysis preclude us from making any conclusions regarding these trends.

The integration profile of quiescent cells still does not explain the lack of protein expression in these cells, as integration events do occur in active transcriptional units. Several reasons may account for this. First, the orientation of integrants is not a factor since the trends are similar in both quiescent and stimulated CD4⁺ T cells. Orientation of HIV integration has been suggested to influence viral gene expression (31, 32). Another possibility may be problems in the expression of viral RNA (1, 29, 33, 36-38, 44, 55). Serious defects in viral mRNA transcription may arise in quiescent cells, resulting in a lack of virus expression and poor rescue. In addition, posttranscriptional events, such as stability of viral RNA, may affect productive infection. Recent work by Lassen et al. (45) has suggested that the lack of PTB1 protein in resting cells latently infected is responsible for the lack of protein expression. However, while defects in the process of transcription seem to be an amenable explanation, they do not fully account for the lack of viral expression. While integrated virus is closely linked to viral gene expression, our results need to be viewed from the perspective of global RNA expression in quiescent versus stimulated cells. Quiescent T cells have less total RNA expression than stimulated T cells. When we look at the levels of total RNA, stimulated cells express at least 5- to 10-fold higher levels of total RNA per cell compared to quiescent cells (data not shown). Therefore, the vast majority of genes that HIV would integrate in quiescent cells would express at much lower levels than those seen in activated cells. Thus, this level may not be sufficient to support productive virus replication. While the exact mechanism for this is unknown, certainly lower basal levels of expression may result in lower expression of tat as well as the previously mentioned tat cofactors (45).

Furthermore, a careful examination of extra-chromosomal viral DNA showed that quiescent cells formed more 2-LTR circles than did activated cells. This may be due to the delayed viral life cycle seen in quiescent cells. As shown by our group, HIV integration peaks 24 h after infection in stimulated cells, while in quiescent cells, it begins 36 h after infection. In addition, the integrated provirus and the 2-LTR circles in quiescent cells were plagued by abnormal junction formation, suggesting that the ends of the full-length viral DNA were either degraded or processed by enzymes in the cytoplasm. Thus, viral reverse transcription and cDNA processing are very inefficient. Aberrant RNase H processing may also occur in quiescent cells, further deteriorating the quality of viral cDNA. Therefore, it is quite possible that the viral genome, even though it is integrated, is mutated in regions that were not investigated in this study.

Our studies shed light on several questions important to the field and have relevance to the study of HIV latency. Our studies clearly establish that integration site selection is not affected by the activation state of the target cells. These studies also suggest that the ability of reverse-transcribed viral genomes to appropriately target to cellular DNA is limited in quiescent CD4⁺ T cells. Thus, these cells could serve as potential latent viral reservoirs, albeit at a reduced state.