Persistent low viral load on antiretroviral therapy is associated with T cell-mediated control of HIV replication

Background: It is unclear how stable low-level viral replication and CD4 cell numbers can be maintained under highly active antiretroviral therapy (HAART). This study was designed to analyse whether HIV-specific responses in stable partially controlled patients during antiretroviral therapy (ART) differ from those observed in complete HAART failure and whether they contribute to the control of viral load (VL).

Methods: Three groups of patients were selected according to plasma HIV RNA levels during 18 months of ART: persistently low VL (LoVL; HIV RNA <10 000 copies/ml; n = 28), undetectable VL (UnVL; HIV RNA <200 copies/ml; n = 29) and high VL (HiVL; HIV RNA >10 000 copies/ml; n = 14). T-cell responses were studied using lymphoproliferative and interferon (IFN)-γ-ELISpot assays against HIV-p24, -gp160, recall antigens, and 15 pools of HIV-(Gag + RT) peptides.

Results: Frequencies of IFN-γ-producing CD4 T cells against HIV-p24 were higher in LoVL than in UnVL or HiVL groups [median, 131, 47 and 23 spot-forming cells (SFC)/1 × 10⁶ peripheral blood mononuclear cells (PBMC), respectively; P = 0.012 and P = 0.047]. Lymphoproliferative responses to HIV-p24 and recall antigens were similar in LoVL and UnVL groups but lower in HiVL (P = 0.004). Frequencies of HIV-specific CD8 T cells were higher in LoVL than in UnVL (1340 versus 410 SFC/1 × 10⁶ PBMC; P = 0.001). They correlated negatively with VL in the LoVL and HiVL (r, -0.393, P = 0.039 and r, -0.643, P = 0.024, respectively) and positively correlated with anti-HIV CD4 cell frequencies in the LoVL group only (r, 0.420; P = 0.026).

Conclusion: Persistently low viral replication (<10 000 copies/ml) during ART stimulates high frequencies of HIV-specific CD4 and CD8 T cells compared to full virus suppression or complete ART failure. The association of high anti-HIV activity with large numbers of HIV-specific CD8 T cells contribute to the control of viral replication.

Introduction

One of the main treatment aims in HIV infection is to drive plasma viral load (VL) persistently below the limit of detection (usually 200 copies/ml), but this goal cannot be achieved in all patients. Intriguingly, CD4 lymphocyte counts can remain stable ^[1-3], and as previously reported ^[4-6] clinical progression can be markedly slowed in some patients with VL values persistently below the higher threshold of 10 000 copies/ml under highly active antiretroviral therapy (HAART). However, this incomplete viral suppression carries a risk of resistance and treatment failure ^[7,8]. The stability of the CD4 cell counts may reflect the reduced viral fitness of the resistant strains in these patients ^[9], while the CD4 cell count generally falls during treatment interruption in patients infected by wild-type virus ^[10,11]. The clinical prognosis of patients in whom the CD4 T-cell count increases during antiretroviral therapy is similar regardless of whether there is a detectable VL ^[2,12,13]. This suggests that clinical and/or immunological benefits are possible despite persistence of viral replication in plasma.

HIV-specific CD4 and CD8 T lymphocytes play a key role in controlling HIV infection ^[14-16] although the exact function correlating with protection remains unsolved. An early treatment initiation during the acute phase of HIV infection can preserve HIV-specific CD4 T cells capable of proliferation and interferon (IFN)-γ production ^[16-18]. In contrast, when treatment is started during chronic infection, proliferative HIV-specific CD4 T-cell responses are rarely observed ^[19-23] while the IFN-γ production remains detectable though with a weaker intensity ^[24], even though T-cell immunity towards opportunistic pathogens is consistently restored in these patients. In addition, the high HIV-specific CD8 T-cell numbers generally observed in chronic infection exponentially decay with VL after treatment initiation ^[25,26]. These observations led to the theory that the HIV-specific effector immunity wanes during HAART because HIV-specific T cells are no longer optimally stimulated by appropriate levels of HIV antigens. Nevertheless, immune responses to HIV can be restored, at least transiently, when the immune system is re-exposed to HIV antigens during treatment interruption ^[17,27-31]. Antigen dependence of the HIV-specific effector responses has recently been illustrated in some treated patients with persistently low-level virus replication in whom an enhanced HIV-specific immunity could be observed ^[32]. In this study, however, immune responses from successfully or partially controlled patients were compared to those in untreated patients but the question remains as to whether HIV-specific responses in partially controlled patients differ from those observed in complete HAART failure and might participate in controlling the virus when the VL remains stable.

To address this question, we examined CD4 and CD8 cell-mediated immune responses to HIV in three groups of patients with stably low, undetectable, or high viral load during long-term HAART.

Methods

Study design

This was a cross-sectional study comparing three groups of HIV-1-infected patients receiving antiretroviral therapy, including at least two reverse transcriptase inhibitors (NRTI), for a minimum of 18 months. The study protocol was approved by the Institutional Review Board of Pitié-Salpêtrière Hospital, and all the patients provided their written informed consent.

Study population

The eligibility criteria were as follows: age >18 years, documented HIV-1 infection, and antiretroviral therapy including at least two NRTI, for at least 18 months. Participants were divided into three groups on the basis of plasma VL values during the 18 months of treatment: persistently low (LoVL; HIV RNA between 200 and 10 000 copies/ml; patients with no more than one blip of VL are eligible to be in this group), undetectable (UnVL; <200 copies/ml), or high (HiVL; >10 000 copies/ml on at least two occasions).

Virological evaluations

Plasma HIV RNA levels were determined with the ultrasensitive Amplicor HIV Monitor test (Amplicor version 1.5; Roche Diagnostics, Meylan, France; detection limit 50 copies/ml). Genotypic resistance testing was performed by population-based sequencing and resistance mutations were identified according to the International AIDS Society Resistance-USA Panel (http://www.iasusa.org) as revised in March 2003.

Immunological evaluations

CD4 and CD8 T cell counts were determined in whole blood by three-colour flow cytometry, using standard methods.

CD4 and CD8 cell-mediated immune responses were assessed in terms of IFN-γ production, measured with an enzyme-linked immunosorbent spot assay (ELISpot) ^[29]. To test CD4 T-cell responses (CD4 IFN-γ ELISpot), fresh peripheral blood mononuclear cells (PBMC) were stimulated with HIV recombinant proteins p24 and gp160 (2 μg/ml; Protein Sciences, Meriden, Connecticut, USA), cytomegalovirus (CMV) crude extract antigens (1:2000; BioWhittaker, Wokingham, UK) and purified protein derived from tuberculin (PPD) (1 μg/ml; Statens Serum Institut, Denmark). Phytohemagglutinin (1 μg/ml; Murex, Paris, France) and medium alone served as positive and negative controls, respectively. Plates were incubated at 37 °C for 40 h [a duration that had been previously optimized to ensure the maximal specific IFN-γ release after such antigen stimulation while maintaining a low background of median 2 spot-forming cells (SFC)/well]. The CD4 T-cell origin of IFN-γ secretion under these conditions had been established prior to the study after CD4 cell depletion that abolished 85% of the IFN-γ production. The T-cell origin of IFN-γ secretion in these conditions was first established (not shown). To adjust the CD4 ELISpot results for CD4 cell count and percentage in peripheral blood, results were also expressed as SFC/ml blood, as follows: SFC × CD4 cell count/1000 × CD4%.

ELISpot assay to detect CD8 T cell IFN-γ production after peptide stimulation (CD8 IFN-γ ELISpot) was performed with frozen PBMC (viability >85%) and 15 pools of 15-mer synthetic HIV peptides overlapping by 11 amino acids; 11 pools covered the three HIV-1 Gag proteins (small proteins, n = 3; p24, n = 5; p17, n = 3) and four pools covered HIV-1 reverse transcriptase (RT; 2 μg/ml; Neosystem, Strasbourg, France). Plates were incubated at 37 °C for 20 h according to previously optimized kinetics for IFN-γ release in these conditions. Antigen-specific SFC frequencies were measured with an automated microscope (Zeiss, Munich, Germany) and expressed per 1 × 10⁶ PBMC. Responses were considered positive when ≥50 SFC/1 × 10⁶ PBMC were detected, after subtracting background. In eight patients for samples with a CD8 ELISpot positive against at least one peptide pool and with cells available the assay was repeated with PBMC depleted of CD8 cells by anti-CD8 monoclonal antibody-coated magnetic beads (Dynabeads; Dynal, Oslo, Norway) in order to confirm the cell origin of IFN-γ production. Over a total of 28 Gag and RT pools thus tested, results showed that 82% of Gag-specific and 100% of RT-specific IFN production originated from CD8 T cells.

T-cell proliferation assays were performed as described previously ^[29]. Briefly, fresh PBMC were stimulated with the same antigens and controls as in the CD4 IFN-γ ELISpot assays, then labelled with tritium thymidine (CEA, Saclay, France) on day 6. A stimulation index (SI) was calculated as the counts per minute (cpm) ratio of cells + stimuli to cells + medium. Proliferative responses were considered positive when the cpm value exceeded 3000 and the SI exceeded 3.

Statistical analysis

Continuous variables are reported as medians and ranges. Patients with low VL were compared with the other two groups by using the Mann-Whitney test for continuous data and the χ² or Fisher's exact test for qualitative data. Correlations were studied in the LoVL and HiVL groups by using Spearman's rank correlation. Two-sided tests were considered significant when P < 0.05.

Results

Characteristics of the patients

Seventy-one HIV-infected patients receiving antiretroviral therapy were enrolled between April 2000 and October 2001; their baseline characteristics are shown in Table 1. Twenty-eight patients were in the LoVL group, 29 patients were in the UnVL group and 14 patients were in the HiVL group; as expected, VL values differed significantly among the three groups (respectively, 3.2 log₁₀ copies/ml versus 1.7 log₁₀ copies/ml; P < 0.001 and 5.0 log₁₀ copies/ml; P < 0.001). Fig. 1 shows changes in the CD4 cell count and plasma VL during the 18 months prior to the study. CD4 cell counts were stable in LoVL (+11 × 10⁶/l, P = 0.241), UnVL (+47 × 10⁶/l, P = 0.214) and HiVL (-2 × 10⁶/l, P = 0.6) groups. Nadir CD4 cell counts as well as study baseline CD4 counts were similar in the LoVL and UnVL groups (nadir, 149 × 10⁶/l and 142 × 10⁶/l and baseline CD4 counts, 341 × 10⁶/l and 479 × 10⁶/l, respectively) and were significantly higher than in the HiVL group (nadir, 9 × 10⁶/l and baseline CD4 counts, 53 × 10⁶/l; P < 0.001).

Table 1

Fig. 1

More LoVL patients than UnVL patients received dual NRTI therapy alone (seven versus three patients). Likewise, LoVL patients tended to have received a lower total number of antiretroviral drugs than UnVL patients (five versus seven drugs). Neither difference was statistically significant. The median total number of resistance mutations was significantly lower in LoVL patients (seven mutations) than in HiVL patients (13 mutations; P = 0.004). However, the numbers of mutations to NRTI, non-NRTI and protease inhibitors (PI) were similar in the two groups when analyses were restricted to patients exposed to each drug class (NRTI mutations 4 versus 4, P = 0.284; non-NRTI mutations 1 versus 1, P = 0.509; PI mutations 5.5 vs. 7.5, P = 0.077).

IFN-γ production and proliferative CD4 T-cell responses to HIV and recall antigens

Frequencies of CD4 T cells specific for recombinant HIV-p24 and gp160 and recall antigens are shown in Table 2. The median number of cells directed against HIV-p24 was significantly higher in patients with LoVL (131 SFC/1 × 10⁶ PBMC) than in patients with UnVL (47 SFC/1 × 10⁶ PBMC; P = 0.012) and patients with HiVL (23 SFC/1 × 10⁶ PBMC; P = 0.047) (Fig. 2a). As CD4 cell percentages (P = 0.016) and counts (P = 0.095) were lower in patients with LoVL than in patients with UnVL, we also expressed the results as HIV-specific CD4 T cells per ml of blood; similar significant differences were still found (Table 2). The proportion of patients whose CD4 cells responded to HIV-p24 was also significantly higher among patients with LoVL (71%) than patients with UnVL (45%; P = 0.042) or HiVL (31%; P = 0.014) (Table 2). In contrast, the frequencies of cells producing IFN-γ in response to recall antigens (CMV and PPD) did not differ among the three groups, except for PPD-specific responses that were observed only when analysing numbers of cells/ml of blood and did not appear when the proportions (SFC/1 × 10⁶ PBMC) or when frequencies of responders were analysed. These differences between p24 and PPD responses suggested that changes in p24 responses were more consistent and did not simply reflect general losses of immune responses to pathogens.

Table 2

Fig. 2

The proportion of patients whose T cells proliferated in response to HIV-p24 and gp160 did not differ between groups with LoVL and UnVL (Table 3). Similarly, the intensity of CD4 cell proliferation to p24 did not differ, whereas proliferation to gp160 was slightly stronger in patients with LoVL than patients with UnVL (P = 0.047). Proliferative capacity to both p24 and gp160 was significantly lower in patients with HiVL than patients with LoVL (P = 0.004 and P = 0.007, respectively). Proliferative responses to recall antigens were similar in patients with LoVL and UnVL but lower in patients with HiVL.

Table 3

HIV-specific CD8 T-cell responses

Frequencies of CD8 T cells specific for HIV-gag proteins and the immunodominant regions of RT protein were measured using an IFNγ ELISpot assay (Table 4). As shown in Fig. 2b, the frequency of HIV-gag specific CD8 T cells was significantly higher in the patients with LoVL (1120 SFC/1 × 10⁶ PBMC) than in patients with UnVL (319 SFC/1 × 10⁶ PBMC; P = 0.003); the frequency was slightly but not significantly higher in patients with LoVL than in patients with HiVL. HIV-gag-specific responses were equally directed against each protein (p17, p24, p1, p2, p6 and p7) in the three groups (Table 4).

Table 4

Median numbers of CD8 cells responding to RT protein were also significantly higher in patients with LoVL (189 SFC/1 × 10⁶ PBMC) than in patients with UnVL (38 SFC/1 × 10⁶ PBMC; P = 0.037) and non-significantly higher than in patients with HiVL (53 SFC/1 × 10⁶ PBMC; P = 0.260) (Table 3). Accordingly, the sum of all HIV-gag- and RT-specific T cells detected in the IFN-γ ELISpot assay again showed stronger HIV-specific CD8 T-cell responses in patients with LoVL than in the other two groups.

Correlations between HIV-specific responses and biological characteristics of HIV infection

Higher frequencies of CD8 T cells directed against pools of HIV-gag and RT peptides were associated with lower plasma viral load (r, -0.393, P = 0.039 and r, -643, P = 0.024, respectively) (Fig. 3a and c) in both the LoVL and HiVL groups; by definition, this correlation could not be studied in patients with UnVL. Higher frequencies of HIV-specific CD8 T cells were associated with larger numbers of HIV-p24 specific CD4 T cells producing IFN-γ (r, 0.420, P = 0.026) in patients with LoVL ((Fig. 3b) but not in patients with HiVL (r, 0.000, P = 1) ((Fig. 3d) or UnVL (r, -0.226, P = 0.247) (data not shown). Neither CD4 nor CD8 HIV-specific responses correlated with CD4 cell counts in patients with LoVL, while the frequency of HIV-specific CD4 T cells correlated positively with the CD4 cell count in patients with UnVL and HiVL (r, 0.474, P = 0.009 and r, 0.795, P = 0.001, respectively) (data not shown).

Fig. 3

Discussion

During antiretroviral therapy, we found that patients with stable low VL had significantly stronger HIV-specific CD4 and CD8 T-cell responses than both patients with full viral suppression and patients with uncontrolled viral replication under HAART.

While persistent viral replication during antiretroviral therapy favours the emergence of drug resistance mutations ^[7,8], most such mutations reduce viral fitness. This could conceivably account for the lower level of replication in patients with persistently low VL, but the numbers of RT gene mutations were similar in our patients with LoVL to those with HiVL. Our findings are in keeping with previous reports ^[32,33] suggesting that a low replication level of resistant viruses can induce significant HIV-specific CD4 and CD8 T-cell responses as compared with the low CD4 and CD8 T-cell responses to HIV that we detected in patients with full viral suppression during HAART. This latter finding is in accordance with the partial restoration previously reported in such controlled patients ^[19-24]. Our patients with stably low viral load on antiretroviral therapy clearly differed from long-term non-progressors (LTNP) who also have strong CD4 and CD8 cell-mediated immunity to HIV ^[34]. Indeed the LoVL group had progressed prior to treatment initiation to the same median CD4 nadir <150 × 10⁶/l than the UnVL group. In addition, our LoVL patients differ from LTNP since they had no gene polymorphisms known to slow virus replication, had normal frequencies of the delta 32-CCR5 deletion (15%), and did not have particular HLA alleles such as HLA-B57, known to protect against HIV disease (data not shown) ^[35,36]. In contrast, we show that patients with complete treatment failure, high viral loads and very low CD4 T-cell counts displayed almost no CD4 T-cell responses to HIV but intermediate levels of HIV-specific CD8 responses that did not differ from those of either fully suppressed or partially suppressed patients. These lower HIV-specific CD4 T-cell numbers in patients with HiVL, may also reflect the much lower concurrent CD4 counts or nadir CD4 counts.

Therefore these distinct levels of HIV-specific T-cell numbers demonstrated that a threshold VL value of between 200 and 10 000 copies/ml might be necessary to maintain strong CD4 and CD8 T-cell responses during HAART, while not depleting absolute numbers of CD4 T cells or their capacity to respond to recall antigens, as previously suggested by mathematical models ^[37]. This stimulatory effect appears to be specifically mediated by HIV antigens, as similar differences are not observed for other pathogens such as CMV and mycobacteria. The potency of anti-HIV responses obtained with this continuous low level HIV replication is comparable or higher than those observed during treatment interruptions ^[17,26-31]. They are also higher than those obtained during treatment when immunity to HIV is restimulated by HIV with non-replicating HIV candidate vaccines such as DNA ^[38] or recombinant canarypox ^[37,39]. Whether these immune responses to HIV play a role in the stability of both HIV RNA levels and CD4 cell numbers/ratios is a key issue.

When comparing groups of patients, both the CD4 responses to the recombinant HIV-p24 protein and the CD8 responses to the Gag and RT peptides were significantly higher in the LoVL group than in the UnVL group, thus apparently positively correlated to the viral replication. When comparing the LoVL group with patients in failure, however, both types of responses were also higher in patients with LoVL, although significant so only in the case of CD4 responses. This first set of data suggested that the relationship between immune responses and viral replication differed in the low and in the high VL zones, as shown previously Deeks et al. ^[11,32]. In addition, when studying the correlations at the single-patient level, it is clear that only the CD8 responses correlate negatively with VL, both in patients with LoVL and in those with HiVL. Altogether our findings show that both CD4 and CD8 responses to HIV are immune correlates of protection and provide clear evidence that one cannot provide simplistic explanations. The relationship between viral and immune parameters depends on the intensity of virus replication: a minimal HIV replication is required to stimulate efficiently HIV-specific CD8 T cells which in turn contribute to establish a partial immune control.

Other studies have shown a negative relationship between VL and HIV-specific T-cell responses during antiretroviral treatment ^[32,33,40], also suggesting that these immune responses control the virus or, alternatively, that high levels of viral replication deplete HIV-specific T cells ^[41]. Here, we found that CD8 T-cell responses to HIV correlated negatively with plasma VL in patients with both LoVL and HiVL, but were only efficient in patients with LoVL, and were very strongly dependent on the presence of HIV-specific CD4 T-helper cells. Therefore, within certain limits, low-level ongoing virus production appropriately stimulates both CD4 and CD8 T-cell responses to HIV and contributes to viral control. In addition, as we have previously observed that some HAART-induced virus mutations induce immunogenicity for CD8 T cells ^[42], the stronger immune responses to RT in patients with persistently low VL might contribute to better control of mutated resistant viruses. Anti-HIV CD8 T-cell responses did not significantly differ between patients with LoVL and those with HiVL, whereas the two groups differed markedly in terms of CD4 cell responses, suggesting that CD4 help is crucial for CD8 response in this setting. We also found that HIV-specific CD4 and CD8 T-cell responses correlated positively with each other in patients with stably low virus replication, further supporting the key role of CD4 help ^[34]. The determining factor appears to be the threshold level of viral load, above which appropriate CD4 T help is depleted and below which it is maintained. We found that the proliferative capacity of CD4 cells was more strongly impaired by high-level virus replication than was their capacity to produce IFN-γ. Whether this difference is due to technical artefacts (e.g., destruction of proliferative CD4 T cells by HIV in vitro) remains to be determined. However, our data are in keeping with recent reports that interleukin-2 production disappears more rapidly than IFN-γ production in a context of continuous virus production ^[43,44]. We show here that IFN-γ-producing CD4 T-helper cells are nevertheless associated with prolonged protection, as viral load remained low and stable for at least 18 months.

Finally, we found that immune defences against other pathogens were also preserved despite persistent low-level HIV replication. This is in keeping with the very low incidence of opportunistic infections in patients with CD4 cell repletion despite persistent virus replication on HAART ^[45].

In conclusion, our results suggest that continuous moderate immunologic stimulation by low-level viral replication helps to maintain potent HIV-specific responses that may play a key role in virus control. These findings may have implications for defining immune correlates of protection. A better understanding of this phenomenon may also help to adequately manage patients with this low level of viral replication in a context of antiretroviral therapy.

Acknowledgements

Sponsorship: This study was promoted by the ANRS (French National Aids Agency). M. Astriti is the recipients of Scholarship with CDC of Greece (KEEL).