Identification of a Reservoir for HIV-1 in Patients on Highly Active Antiretroviral Therapy

See allHide authors and affiliations

Science  14 Nov 1997:
Vol. 278, Issue 5341, pp. 1295-1300
DOI: 10.1126/science.278.5341.1295


The hypothesis that quiescent CD4+ T lymphocytes carrying proviral DNA provide a reservoir for human immunodeficiency virus–type 1 (HIV-1) in patients on highly active antiretroviral therapy (HAART) was examined. In a study of 22 patients successfully treated with HAART for up to 30 months, replication-competent virus was routinely recovered from resting CD4+ T lymphocytes. The frequency of resting CD4+ T cells harboring latent HIV-1 was low, 0.2 to 16.4 per 106 cells, and, in cross-sectional analysis, did not decrease with increasing time on therapy. The recovered viruses generally did not show mutations associated with resistance to the relevant antiretroviral drugs. This reservoir of nonevolving latent virus in resting CD4+ T cells should be considered in deciding whether to terminate treatment in patients who respond to HAART.

Current understanding of HIV-1 infection has been greatly advanced by measurements of viral burden in infected individuals (1, 2) and of changes in viral burden after treatment with potent new drugs that inhibit the HIV-1 protease or reverse transcriptase (3-8). Wei et al. (3) and Ho et al. (4) have shown that HIV-1 production in infected individuals is largely the result of a dynamic process involving continuous rounds of de novo virus infection of and replication in host cells with rapid turnover of both free virus and virus-producing cells. In patients treated with HAART, plasma virus falls to undetectable levels in 2 to 4 months, and it becomes difficult to culture the virus from the blood (7, 9). Perelson et al. have used mathematical models of decay rates of plasma HIV-1 RNA and peripheral blood mononuclear cell (PBMC) infectivity in treated patients to make the first rational predictions of treatment times required for virus eradication, with the caveat that there may be more stable compartments or viral reservoirs that are not measurable by standard methods (7).

One potentially stable reservoir is composed of latently infected memory CD4+ T cells carrying integrated provirus (5, 10, 11). Postintegration latency appears to result from the reversion of productively infected CD4+ T lymphoblasts to a resting memory state in which there is minimal transcription of viral genes. Recently, inverse polymerase chain reaction (PCR) has been used to demonstrate that latently infected resting memory CD4+ T cells with integrated HIV-1 DNA are present in infected individuals (10, 11). The frequency of these cells is extremely low, but a fraction of them harbor replication-competent provirus (10, 11). Because memory CD4+ T cells can persist for months to years (12), resting memory CD4+ T cells carrying replication-competent viral genomes, although rare, may represent an important long-term viral reservoir in patients on HAART. Our study was undertaken to determine whether replication-competent virus could persist in the resting CD4+ T cells of patients on HAART who had no evidence of active virus replication.

The presence of latent virus in resting CD4+ T cells was examined in 22 patients treated with HAART for up to 30 months (Table1). We selected patients who met each of the following three criteria: (i) patient- and physician-reported strict adherence to aggressive HAART regimens (typically, three or four drugs, including a protease inhibitor), (ii) a rapid decline in plasma HIV-1 RNA to undetectable levels (<200 copies per milliliter) by quantitative-competitive (QC) reverse transcription PCR (RT-PCR) assay after the initiation of HAART, and (iii) continued undetectable levels of plasma HIV-1 RNA on repeated measurements throughout the course of the study. Plasma HIV-1 RNA measurements reached undetectable levels in 2 to 3 months and remained undetectable on repeated sampling, including the day blood was taken for analysis of latently infected cells (Fig.1A). Patients were tested an average of four times each during the months after initiation of HAART. A subset of three more intensively studied patients (patients 16, 19, and 20) had monthly measurements of plasma HIV-1 RNA levels, which were uniformly undetectable. These results suggest that in all of the patients selected for study, long-term suppression of viral replication to undetectable levels was achieved. Consistent with the success of treatment were gradual increases in CD4 counts in most patients (Fig.1B) and the lack of viral evolution (see below).

Figure 1

Effectiveness of HAART in suppressing viremia and reversing CD4 depletion. (A) Time course for changes in plasma HIV-1 RNA after initiation of HAART. The zero time point is taken as the day on which the patient began a completely suppressive combination therapy regimen, typically a three- or four-drug regimen with at least one protease inhibitor (Table 1). In patient 19, monotherapy had produced a decline in plasma viremia to undetectable levels even before the initiation of combination therapy. Nevertheless, the zero time point was considered to be the day combination therapy was started. The dashed line indicates the limit of detection of the assay (200 copies per milliliter). Assays in which HIV-1 RNA was undetectable are arbitrarily plotted as 150 copies per milliliter. In patient 17, an acute seroconvertor, plasma HIV-1 RNA was still detectable at 3 months (570 copies per milliliter) and subsequently declined to undetectable levels. In this case, the longer interval to undetectable plasma HIV-1 RNA probably reflects the extremely high levels present before treatment (2.36 × 106 copies per milliliter). For patient 21, at the earliest visit after treatment at 1.5 months, plasma HIV-1 RNA was undetectable by the bDNA assay (limit of detection, 10,000 copies per milliliter). On the basis of the subsequent responses, it is likely that HAART therapy produced more rapid decline than that shown. After viral RNA became undetectable, plasma HIV-1 RNA levels remained undetectable in all patients, with the exception of a single reading of 415 copies per milliliter on one occasion for patient 17. (B) Time course for changes in CD4 count after initiation of HAART.

Table 1

Characteristics of patients studied.

View this table:

We next isolated highly purified populations of resting CD4+ T cells from the peripheral blood of these donors. After depletion of monocytes by adherence, negative selection with monoclonal antibodies and magnetic beads was used to remove CD8+ T cells, B cells, monocytes, and natural killer cells. In addition, activated CD4+ T cells, which represented an average of 10% of the CD4+ T cell populations in these individuals (Table 1), were removed by bead depletion with antibodies to CD25, CD69, and human lymphocyte antigen (HLA)–DR. Additional purification was achieved by flow cytometric sorting of small lymphocytes expressing CD4 but not HLA-DR, which is expressed on all activated T cells. This multistep purification gave populations of resting CD4+ HLA-DR T cells that were, on average, 97% pure (Table 1). Most importantly, the sorted populations were typically contaminated with <1% activated CD4+ T cells.

Specialized PCR approaches have shown that a low frequency of resting CD4+ T cells from untreated seropositive individuals carry HIV-1 DNA in an integrated form (2, 10,11). Culture studies have demonstrated that the fraction of resting CD4+ T cells harboring replication-competent provirus is even lower (11) and quite variable from patient to patient, reflecting the fact that much of the proviral DNA detected by PCR may be defective (11, 13). Therefore, although proviral DNA is readily detectable in most patients on HAART (Table 1) (3, 7, 8, 14,15), the critical issue is whether virus persists in a replication-competent form. After 2 to 12 months on HAART, standard culture assays for replication-competent virus in unfractionated PBMCs are typically negative (7, 14). We reasoned that detection of latent virus in resting CD4+ T cells might require enrichment for the relevant cell population, afforded by purification of CD4+ HLA-DR cells, and a high-efficiency activation step (16) in which all of the resting cells are induced to undergo blast transformation and become permissive for viral replication. Therefore, dilutions of purified resting CD4+ T cells were activated in the presence of phytohemagglutinin (PHA) and irradiated PBMCs from an HIV-1–negative donor. These conditions allow activation of virtually 100% of resting T cells (17). The day after activation and again on day 7, CD8-depleted PHA blasts from an HIV-1–negative donor were added to amplify infectious virus.

In each of 18 cases in which a sufficient number of resting CD4+ T cells were obtained for analysis, replicating virus was isolated by this culture system (Table 1). One of these patients had been on HAART for 30 months with no detectable plasma virus. Two other patients had been on HAART for over 20 months. Data were quantified by limiting-dilution assay. In wells seeded with the highest numbers of purified resting CD4+ T cells (5 × 106 or 1 × 106), large amounts of p24 were present in the supernatants on day 14. Wells containing less than 2 × 105 CD4+HLA-DR cells were usually negative. Control wells lacking patient cells were invariably negative for p24. Virus isolation failed only in the case of four patients from whom an insufficient number of resting CD4+ T cells were isolated to set up replicate wells at 5 × 106 or 1 × 106 cells per well.

In principle, activation of resting CD4+ T cells carrying defective viral genomes could result in the production and release of noninfectious viral particles. However, if all of the viral particles released after activation of resting CD4+ T cells were noninfectious, amounts of p24 in the supernatant should decrease over time as the cultures are fed and split. This was not the case, as levels of p24 in our culture system were generally undetectable on the first day after stimulation and increased markedly between days 7 and 14, indicative of a spreading infection (18). To confirm that resting CD4+ T cells from patients on long-term HAART harbor infectious virus, we further passaged p24 positive supernatants onto CD4+ T lymphoblasts from HIV-1–negative donors, and increasing supernatant p24 values were again observed in these cultures (Fig. 2). In addition, viral isolates from some patients had the ability to induce syncytia in cultures of nontransformed CD4+ T lymphoblasts (19). Taken together, these results demonstrate that purified resting CD4+ T cell populations from patients on HAART harbor replication-competent virus that in some cases is cytopathic.

Figure 2

Replication competence of viruses derived from latently infected resting CD4+ T cells. Supernatants from positive limiting dilution cultures were used to infect CD4+ T lymphoblasts from an HIV-1–negative donor. Amounts of p24 antigen in a 1/250 dilution of culture supernatant were measured by ELISA. Increasing p24 amounts in the supernatants of these cultures demonstrate the successful passage of infectious virus. Control supernatants were from wells containing mockinfected PHA blasts.

To provide definitive evidence that the infectious virus detected in these cultures was derived from latently infected resting CD4+ T cells rather than from contaminating activated CD4+ T cells, we directly isolated activated CD4+ T cells from the peripheral blood of HIV-1–positive donors on combination antiretroviral therapy and measured the frequency of activated cells that were releasing infectious virus. In a donor whose plasma HIV-1 RNA was undetectable, the frequency of activated CD4+ T cells releasing infectious virus was <0.0002%. In treated patients with small but detectable levels of plasma virus (∼1000 copies per milliliter), the frequency of activated CD4+ T cells releasing infectious virus was <0.002% (18). Because the highly purified resting CD4+ T cell populations used in the present study contain <1% activated CD4+ T cells and because less than 0.002% of activated CD4+ T cells release infectious virus, it is impossible to account for the observed levels of virus production by contaminating activated CD4+ T cells. The only other trace contaminants were CD4-negative cells, including B cells and CD8+ T cells, cells unlikely to be sources of virus. Taken together, these results definitively identify resting CD4+HLA-DR T cells as a reservoir for infectious HIV-1 in infected individuals on long-term HAART.

The replication-competent virus present in these cells is likely to be in an integrated form. When purified resting CD4+ T cells are infected in vitro and then incubated for 6 days, it becomes difficult to recover virus by activation of the cells (11). Thus, neither virions that remain bound to the cell surface nor unintegrated HIV-1 DNA molecules formed in the cells were capable of giving rise to a productive infection after this incubation. When resting cells from patients on HAART were preincubated before activation, the measured frequencies were very similar to those obtained by immediate activation, indicating that the recovered virus was derived from cells carrying stable integrated forms (18).

To determine the frequency of cells carrying replication-competent provirus among resting CD4+ T cells, we analyzed the limiting dilution virus culture data by a maximum likelihood method (Table 1 and Fig.3). Overall, the measured frequencies ranged from 0.2 to 16.2 per 106 resting CD4+ T cells. These frequencies are lower than the frequencies of cells carrying persistent replication-competent forms of HIV-1 DNA previously measured in a different group of patients who were not on HAART [1 to 81 per 106 resting CD4+ T cells, mean 7 per 106 (11)]. In a cross-sectional analysis, frequencies of latently infected CD4+ T cells did not decrease appreciably with increasing time on HAART (Fig. 3). These results suggest a slow decay rate for this compartment, consistent with the long-term survival of memory CD4+ T cells in uninfected individuals (12). These cross-sectional data cannot be used to define the actual decay rate of this compartment because of a potential problem that can occur in cross-sectional analysis under specific conditions (20). Precise determination of the decay rate of latently infected CD4+ T cells will require prospective sequential quantitation of such cells in individual patients over a period of years. In a single patient studied on two occasions (after 5 months and 8.5 months of HAART), no decrease in the frequency of latently infected cells was noted (Fig. 3).

Figure 3

Relation between the frequency of latently infected CD4+ T cells and the time on HAART. Frequencies are expressed in terms of infectious units per 106 resting CD4+ T cells (IUPM). Each square indicates the isolation of virus from a patient on HAART for the indicated number of months. The patient number is indicated inside the square. Black squares indicate long-term infected patients. Shaded squares (patients 17 and 20) are acute seroconvertors treated within 10 weeks of presentation. The dashed line indicates that assays done on patient 1 after 5 and 8.5 months of HAART showed about the same amount of latently infected cells. The dotted line indicates the normal limit of the detection of the assay.

Of the 18 patients from whom virus was isolated, two were newly infected patients treated within 10 weeks of the onset of symptomatic primary infection (patients 17 and 20). The isolation of virus from these individuals suggests that the latently infected CD4+T cell compartment can become established early in infection.

To determine whether the viruses that persisted in latently infected resting CD4+ T cells had developed resistance to the relevant antiretroviral drugs, we determined nucleotide sequences of the HIV-1 pol gene for 12 viral isolates (Tables2, 3, and4). Protease inhibitor monotherapy results in rapid selection of drug-resistant variants with characteristic mutations at particular codons in the protease gene (21, 22). For patients treated with ritonavir (RTV), multiple mutations arise in a defined sequence beginning at residue V82 (21, 23). Analysis of eight sequences from seven different RTV-treated patients showed little evidence for the evolution of resistance (Table 2). Only one isolate from one patient (9) had a V82 mutation. This isolate also had a characteristic Ile154 → Val (I54V) mutation and may have arisen during the 15 months of RTV monotherapy that the patient received before starting a four-drug regimen of zidovudine (AZT), lamivudine (3TC), RTV, and saquinavir (SQV). Another isolate from the same patient did not show any mutations associated with resistance to RTV. Neither isolate had the L10, G48, I84, or L90 mutations associated with resistance to SQV. The isolate from patient 16 had M36I and A71T mutations that may reflect the RTV monotherapy the patient received before starting combination therapy or may represent preexisting polymorphism at these positions (24). For this patient, comparison with available baseline plasma RNA and proviral DNA sequences revealed no changes in the reverse transcriptase or protease coding regions after 30 months of combination therapy. One additional RTV-treated patient (patient 11) had a single I54L mutation. None of the isolates from RTV-treated patients showed characteristic changes at K20, L33, M46, I84, or L90. Isolates from four other patients treated with RTV combination therapy for between 5 and 22 months showed no resistance-associated mutations. None of the four patients treated with indinavir (IDV) had any of the characteristic resistance mutations seen at L10, K20, L24, M46, I54, A71, V82, I84, or L90 (Table 3). Thus, viruses grown from resting CD4+ T cells of patients on HAART did not show the pattern of multiple mutations characteristic of high-level resistance. Each of the 1.1-kb sequences was distinct from others in the data set, from any published laboratory strain, and from any of the >200 sequences from other clinical isolates analyzed previously in the laboratory. This finding is illustrated by the analysis of positions where polymorphisms not associated with drug resistance occur (Table 4). In the case in which two biological clones were obtained from the same patient (that is, patient 9), the sequences were clearly related to each other. In addition, the appearance of characteristic resistance-associated mutations in isolates from patients who had received prior monotherapy (that is, patient 9) provides strong evidence linking the sequences to the histories of individual patients. Taken together, these results indicate that in the patients studied, combination therapy was effective in suppressing viral replication and the attendant selection of resistant variants.

Table 2

Nucleotide sequences of eight isolates from resting CD4+ T cells of patients on HAART. Sequences at codons associated with resistance to RTV or SQV for patients who received RTV or SQV (or both) are shown. Dashes indicate identity with the consensus sequence shown in the top row. Positive culture supernatants were amplified by RT-PCR and sequenced with the Affymetrix (Santa Clara, California) high-density oligonucleotide array (GeneChip) as described (14, 21) (GenBank accession numbers AF027391 to AF027402).

View this table:
Table 3

Nucleotide sequences of four isolates from resting CD4+ T cells of patients on HAART. Sequences at codons associated with resistance to IDV for patients who received IDV are shown. Dashes indicate identity with the consensus sequence shown in the top row. Procedures used are described in the legend to Table 2.

View this table:
Table 4

Nucleotide sequences of 12 isolates from resting CD4+ T cells of patients on HAART. Sequences at other codons where there are polymorphisms not associated with protease inhibitor resistance are shown. Dashes indicate identity with the consensus sequence shown in the top row. Procedures used are described in the legend to Table 2.

View this table:

The isolation of viruses with drug-sensitive genotypes from the resting CD4+ T cells of patients successfully treated for long periods with HAART suggests that these viruses are derived from long-lived cells that were initially infected before the initiation of therapy. The ability of replication-competent viruses to persist in a latent form in resting CD4+ T cells is also consistent with the finding that frequencies of latently infected CD4+ T cells did not decrease appreciably with increasing time on HAART. The isolation of replication-competent virus from patients who have responded well to long-term HAART has also been achieved by Wonget al. (25) and Chun et al.(26). Whether latently infected CD4+ T cells can be efficiently reactivated to produce virus in vivo is unknown; however, the existence of a small but relatively stable compartment of latently infected cells should be considered in deciding whether treatment should be stopped in patients with no other evidence of residual virus.


Stay Connected to Science

Navigate This Article