Introduction HIV-1 infection is normally characterized by sustained viral replication and a progressive loss of CD4+ T cells, leading to AIDS. Combined antiretroviral therapy (cART) suppresses viral replication and drastically reduces morbidity and mortality . However, cART does not eradicate infected cells , and plasma viremia generally rebounds quickly after treatment is discontinued . The existence of a few HIV-infected patients who spontaneously controlled HIV replication to undetectable levels for many years (HIV controllers [HICs]) suggests that a functional HIV cure or remission might be possible. However, how or whether other patients can achieve an HIC-like status is unclear. Emerging evidence shows that early treatment has long-term benefits . Treatment initiation during primary HIV-1 infection (PHI) rather than during chronic HIV-1 infection (CHI) may i) further reduce residual viral replication , ii) limit viral diversity  and viral reservoirs , iii) preserve innate immunity and T and B cell functions , , , and iv) accelerate immune restoration . Most relevant studies show that CD4+ T cell counts are higher and that viral rebound occurs later (and at a lower level) after the discontinuation of treatment that began during PHI compared with treatment that began during CHI , . Although in most cases, these advantages wane soon after treatment interruption , the existence of individuals in whom the viral load remains undetectable for several years after the interruption of prolonged therapy that was initiated very early after infection (post-treatment controllers [PTCs]) was reported by our group in 2010 . These individuals hold important clues in the search for a functional HIV cure. Here, we have identified and characterized a group of 14 PTCs. We analyzed whether PTCs shared parameters that have been associated with spontaneous control of viremia in HICs, to explore whether the efficient control of infection in PTCs may indeed be derived from early treatment. In addition, we explored the level and distribution of the PTCs' latent viral reservoir in the blood. Indeed reaching functional cure will likely require reducing not only the size but also the distribution of the HIV reservoirs, particularly among the CD4 T cells with long lifespan or important clonogenic properties, as naïve and central-memory T cells (TCM). Results Study population We studied 14 HIV-1-infected patients with durable viral control following the interruption of effective cART that was initiated during PHI (PTCs). The patients' characteristics are reported in Table 1 and Figure 1. All 14 patients were diagnosed with PHI in the late 1990s or early 2000s. Twelve patients had a symptomatic primary infection. During PHI (1.6 [1.1–2.1] months estimated after initial exposure), the PTCs had higher viral loads (median 5.0 log HIV-1 RNA copies/ml) and lower CD4+ T cell counts (median 502 cells/µl; Table 1) compared with the 8 patients in the ANRS PRIMO cohort who subsequently exhibited spontaneous control of viremia (median 3.0 log HIV-1 RNA copies/ml of plasma and 794 CD4+ T cells/µl at PHI (2.2 [1.7–3.5] months estimated after exposure, p = 0.11 for the delay when compared to PTC)  (Figure 2). In contrast, PTC values during PHI were similar to those of patients in the ANRS PRIMO cohort who did not control their infection afterwards (5.1 log HIV RNA copies/ml and 517 cells/µl; Figure 2). 10.1371/journal.ppat.1003211.g001 Figure 1 Long-term control of viremia and stable CD4+ T cell counts in fourteen patients after interruption of antiretroviral treatment initiated in primary HIV-1 infection. CD4+ T cell counts (in black) and plasma HIV-1 RNA viral loads (in blue) during the follow-up after PHI diagnosis in the 14 PTCs included in the study. The detectable viral loads after treatment interruption are indicated in red. The gray areas represent the periods during which the patients received cART. 10.1371/journal.ppat.1003211.g002 Figure 2 Patients to become post-treatment controllers have higher viral loads and lower CD4+ T cell counts than HIV controllers during primary HIV infection. CD4+ T cell counts (A) and plasma viral load (B) during the primary infection for the patients enrolled in the ANRS PRIMO cohort who later exhibited spontaneous control of infection (preHIC; n = 8) , for the PTCs included in our study (n = 14) and for the patients in the ANRS PRIMO cohort who did not control infection (n = 1,245). The median and the 10th and 90th percentiles are shown for each group. 10.1371/journal.ppat.1003211.t001 Table 1 Characteristics of PTC included in the study. Code Sex1 Year of diagnostic PHI2 Fiebig ART initiation ART combination3 Time on cART (months) Time since interruption(months) CD4 T cell counts (cells/µl) Last HIV-1 DNA (copies/106 PBMC) HIV-1 RNA VL4 (copies/ml plasma) HIV-1 RNA VL since treatment interruption First cART discont. Last At PHI (Log) Last during follow up VL 50 400 OR1 M 1996 Sympt V 2 NRTI 81 82 416 1057 959 134 4.3 50 copies/mL prior to treatment interrupted cART while their viral load was 50 copies/mL or 1 viral load >50 copies/ml, followed by cART resumption (Figure 7). The probability of maintaining viral control at 12 months was estimated as 15.3% [4.4–26.3], and it was identical at 24 months post-cART interruption. 10.1371/journal.ppat.1003211.g007 Figure 7 Interruption of long-term treatment initiated at PHI leads to a significant frequency of viremia control. Kaplan-Meier curve of the probability for patients included in the FHDH between 1997 and 2011 to lose control of viremia after interruption of a, at least, one year-long cART initiated within 6 months of HIV infection, and who had at least one viral load determination 12 months after treatment interruption (n = 74). Loss of control was defined by 2 or more viral loads above 50 RNA copies/mL or one viral load above 50 RNA copies/mL followed by resumption of cART. Discussion Numerous efforts have been aimed to achieve a functional cure for HIV infection that would allow treatment to be stopped altogether. We studied 14 patients in whom viral replication was controlled to undetectable levels for several years after the discontinuation of cART. These PTCs with long-term virological remission may hold important clues about a possible functional cure for HIV. The 14 PTCs presented in this study maintained lasting control of viremia after the interruption of prolonged therapy that began early during PHI. We found that most PTCs were readily distinguishable from spontaneous HICs in many respects. In many cases, spontaneous control seems to start very soon after HIV infection , , and most HICs have lower-than-normal viral loads during PHI . In contrast, the PTCs had a more severe primary infection with higher viral loads and were frequently symptomatic, which may have prompted the early treatment in some cases. These observations are consistent with the generally unfavorable HLA genotypes of the PTCs. In particular, the risk alleles HLA B*35 and HLA-B*07, rarely observed in the HICs , were highly prevalent among the PTCs. Furthermore, two PTCs carried the HLA-B*3503 allele, which is associated with accelerated disease progression and impaired HIV-specific T cell function . We cannot rule out the possibility that spontaneous control may have been masked in some cases by early therapy initiation. In particular, it might be possible that some potential HICs who lacked protective HLA alleles were more prone to have higher viral loads in primary infection and, hence, more likely to initiate therapy. However, other differences were observed between the PTCs and the HICs during the chronic phase of infection. In particular, the PTCs had a low frequency and quality of HIV-specific CD8+ T cell responses. Although some HICs do not exhibit strong HIV-specific CD8+ T cell responses , , the overall differences between the HICs and PTCs in our study were striking, even when the HICs carrying the protective HLA B*27 and B*57 alleles were excluded from the analyses. Finally, the PTCs were characterized by a lower CD8+ T cell activation status compared with the HICs. The 5 to 15% of PTCs observed among the patients in the FHDH ANRS CO4 study and in other studies , ,  appears higher than the proportion of HICs with spontaneous viral control in patients followed from primary infection , . Therefore, our results strongly suggest that the infection control in the PTCs was not achieved spontaneously and was favored by the early onset of therapy. Because the interruption of long-term cART initiated early during PHI is not recommended, only a very small proportion (∼2%) of the patients in the FHDH experienced such an interruption, which may explain the rarity of PTCs worldwide. It is also important to consider that the 14 PTCs studied here had exhibited infection control without therapy for a very long period, and they may differ from PTCs with a shorter period of control . The control of viremia following treatment interruption was associated with very low HIV blood reservoirs in the PTCs. This observation, together with similar observations in the HICs , suggests that limiting the pool of infected cells is crucial for the successful control of viral replication in the absence of therapy. In PTCs, the early cART initiation and the lengthy treatment period likely played an important role in reducing the reservoirs , . Interestingly, five PTCs experienced a progressive decline in their viral reservoir after treatment interruption, which is one of the goals in the search for an HIV cure. However, very small HIV reservoirs do not guarantee infection control off therapy . A key additional element might be a low reservoir distribution in cell subsets with long lifespan as naïve and central-memory T cells. Indeed we found that the cell subsets of all the PTCs analyzed ex vivo carried very low levels of HIV DNA. In particular, long-lived resting CD4+ T cells from the PTCs provided a minor contribution to the total HIV reservoir. Naïve CD4+ T cells were poorly infected, and overall the presence of the virus in these cells could not be detected (via DNA or viral replication) in 40% of the samples. This extremely low reservoir in PTCs' naïve cells contrasts with the massive infection detected at the end of the first month after initial infection with a median of 3 log copies HIV-DNA/million naïve cells (C. Bacchus and A. Cheret, personal communication), as also reported a year after initial infection in the absence of treatment, although the naïve cells contained a log lower level of cell-associated HIV DNA than other memory subsets . These discrepancies suggest that early therapeutic intervention is extremely efficient at decreasing those very long-lived reservoirs. Central memory CD4+ T cells also contributed very weakly to the HIV reservoir because of a skewed resting CD4+ T cell subset distribution with a large proportion of shorter-lived transitional memory cells. The skewed distribution of the resting CD4+ T cells observed in the PTCs is also found in uncontrolled early infection (our own unpublished results), further indicating that early therapeutic intervention strongly contributed to the nature of the viral reservoir in these individuals. The TCM cells have been shown to be heavily infected a year after infection, and the main contributor to the total HIV reservoir in patients treated during chronic infection . Similarly weakly differentiated memory CD4 T cells were shown to contain the majority of the HIV reservoirs in untreated chronically infected patients . In contrast, we recently reported a protection of TCMs that contributed less to the total HIV reservoir in long-term non progressors carrying HLA-B*27 or B*57 alleles , and TCM protection has also been observed in the nonpathogenic SIV infection of sooty mangabeys . Altogether our results suggest that a functional cure would most likely require reducing both the size and the distribution of the HIV reservoirs, particularly among those resting CD4 T cells with a long lifespan or important clonogenic properties, such as naïve and central memory T cells. Early therapy may also limit viral diversity and offer protection of innate and specific immunity from the deleterious effect of chronic immune activation. However, it remains unclear why only a limited fraction of patients is able to control the infection after therapy interruption, and a study of the effectors of control in PTCs is underway. In addition, mechanisms that diminish the susceptibility of host cells to HIV-1 infection  and protect long-lived cell types  have been implicated in the control of HIV/SIV infection and pathogenicity in humans and nonhuman primates and may favor infection control after treatment interruption in some individuals. Finally, it is also possible that properties of the viruses infecting the PTCs studied, along with potential limitation of viral diversity by early institution of cART may play a role in the phenotype reported. We are currently addressing these questions. Arguments against cART initiation during PHI include the potential for long-term toxicity, the development of resistant viruses and the cost. However, new antiretroviral drugs are well tolerated, highly effective and associated with excellent compliance, strongly reducing the risk of resistance . In addition, early treatment initiation improves survival  and reduces the risk of HIV-1 transmission . Here, we show that in some HIV-infected individuals with symptomatic primary infection and no favorable genetic background, off-therapy viral control for several years may be associated with a very early and prolonged antiretroviral treatment. These findings argue in favor of early cART initiation and open up new therapeutic perspectives for HIV-1-infected patients. Methods Ethics statement All of the subjects provided their informed written consent to participate in the study. The CO6 PRIMO, CO15 ALT and CO18 HIV controller cohorts are funded and sponsored by ANRS and were approved by the ethics review committees of Ile de France III, VI and VII, respectively. The institutional review board of Institut Pasteur and Pitié-Salpêtrière Hospital (Paris, France) also approved the study protocol. The VISCONTI study was funded by ANRS (EP47), sponsored by Orléans Regional Hospital and approved by the Tours ethics review committee. Subjects The post-treatment controllers (PTCs) were defined as patients who initiated cART within 10 weeks of PHI and whose plasma HIV RNA levels remained less than 400 copies/mL for at least 24 months after cART interruption. Primary infection was defined as symptoms associated with an incomplete HIV-1 Western blot and a positive p24 antigen test or detectable plasma HIV RNA, and/or seroconversion documented by a positive HIV antibody test that was preceded by a negative test less than 3 months before. Fourteen PTCs were included in this study. Four had been identified in a previous study , six were recruited from the ANRS CO6 PRIMO cohort of patients diagnosed during PHI , and four were recruited from patient follow-up at Hôpital de la Croix Rousse in Lyon, CHRU Gui de Chauliac in Montpellier, and CHU de Saint Louis in Paris, France. The HIV controllers (HICs) were patients from the ANRS CO15 and CO18 cohorts who had been infected for more than 5 years, were naïve of antiretroviral treatment and whose last 4 consecutive plasma HIV RNA values were less than 400 copies/ml. Viremic (VIR) patients were defined as patients who were HIV-1-infected for more than 6 months, were not receiving antiretroviral therapy and had HIV-1 plasma viral loads greater than 7500 RNA copies/ml. cART-treated individuals (ARTs) were HIV-1-infected patients whose viral load had been less than 50 RNA copies/ml of plasma for at least 6 months on cART initiated either on PHI or CHI. HLA typing The subjects were serologically HLA-typed using complement-mediated lymphocytotoxicity testing (InGen One Lambda, Chilly Mazarin). High-definition genotyping of the HLA-B*35 alleles was conducted by direct exon sequencing. HIV-specific CD8+ T cell response Interferon (IFN)-γ secretion by HIV-specific CD8+ T cells was quantified ex vivo with an ELISPOT assay . For each subject, the optimal peptides (2 µg/mL) corresponding to known optimal CTL epitopes derived from the HIV-1 Env, Gag, Pol and Nef proteins were tested, depending on the results of the HLA typing. The method used to assess the CD8+ T cells' capacity to suppress an ex vivo HIV-1 infection of autologous CD4+ T cells has been thoroughly previously described . Activation phenotyping The following antibodies were used: CD8-APC-H7 or -PerCPCy5.5 (SK1), CD3-APC or -APC-H7 (SK7), HLA-DR-PE-Cy7 (L243) and CD38-PerCPCy5.5 (HIT2) (BD Biosciences). The cells were fixed and analyzed with a FACSCanto I flow cytometer (BD Bioscience). Sorting of the PBMC subpopulations PBMCs that were cryopreserved and stored in liquid nitrogen and had more than 80% viability after thawing were sorted as live monocytes (CD3−CD4+) and activated and resting CD3+CD4+ T cells on a 5-laser FACS ARIA II cell sorter (Becton Dickinson) on the CyPS platform (UPMC), after staining with the following combination: Live-Dead Fixable Aqua (Life Technologies), CD3-Pacific Blue, CD4-AlexaFluor700, CCR7-PE Cyanine7 (3D12), CD27-APC, CD69-FITC and HLA DR-FITC (BD Pharmingen), CD45RA-ECD and CD25-FITC (Beckman Coulter). The resting CD4 T cells (CD25−CD69−HLADR−) were further sorted into the following categories: naïve (TN, CD45RA+CCR7+CD27+), central memory (TCM, CD45RA−CCR7+CD27+), transitional memory (TM, CD45RA−CCR7−CD27+), and effector memory (TEM, CD45RA−CCR7−CD27−) cells (Supplementary Figure S2). The collected cell numbers varied from 0.01 to 2 million cells among subsets and patients, and the purity of the sorted subsets was greater than 90%. The data were analyzed using Flowjo software (Treestar). HIV DNA quantification The total cell-associated HIV DNA was quantified using ultrasensitive real-time PCR (Biocentric, Bantol, France) in the PBMC, monocyte and CD4 T cell subsets, as previously described (ANRS assay ). The entire HIV DNA extract was tested in two to four PCRs. The results are reported as either the actual HIV DNA copy numbers/million cells or as an estimated value calculated as 50% of the detection threshold value when the cell HIV DNA was lower than the threshold. The thresholds varied according to the available cell numbers and were calculated for each assay . Detection and amplification of HIV-1 from peripheral blood CD4+ T cell subsets A first fraction of sorted resting CD4+ TN, TCM, TTM and TEM subsets from 7 PTCs was tested for the total cell-associated HIV DNA level (see above). A second fraction of the same samples was cultured in variable numbers in 10% FCS supplemented RPMI 1640 medium for 13 days after stimulation at Day 0 with anti-CD3/anti-CD28 plus IL-2 (Roche, 5 µg/ml) and human recombinant IL-7 (Cytheris, 1 ng/ml) or with human recombinant IL-7 alone. At Days 3, 6, 8, and 10, half of the supernatants were removed to quantify the HIV RNA, and IL-2 and IL-7 were added. The viral production kinetic in the supernatants was measured using real-time PCR HIV RNA quantification (Biocentric, Bandol, France). The viral production capacity of each subset was expressed as the ratio between the HIV RNA copies in the supernatants at a given day of culture and the level of cell-associated HIV DNA of each subset measured at Day 0 of culture. Statistical analyses The Kruskal-Wallis nonparametric test was used to compare continuous variables between groups. A Wilcoxon matched-pairs signed rank sum test was used to compare variations in values (CD4+ T cell counts, HIV DNA levels) over time or to compare cell subsets in the sorting experiments. The allele frequencies in the different groups of patients were compared using Fisher's exact test. A Kaplan-Meier estimate was used to assess the probability of post-treatment control in patients who discontinued early cART. All values given in the text are medians and (range) or [IQR]. The SigmaStat 3.5 software (Systat Software Inc.-SSI, CA) or SAS software package, Version 9.2 (SAS Institute, Cary, NC, USA) was used for all analyses. Supporting Information Figure S1 Similar CD4+ T cell activation levels in post-treatment controllers and HIV controllers. A. Percentage of CD4+ T cells from ART (n = 5), HIC (n = 58) and PTC (n = 8) that expressed CD38, HLA-DR or both CD38 and HLA-DR ex vivo. Means and standard deviations are shown. B. Surface CD38 and HLA-DR expression on CD4+ and CD8+ T cells from one representative ART, HIC and PTC. (PDF) Click here for additional data file. Figure S2 Cell sorting scheme. Resting CD4 T cell subsets were selected out of a singlet lymphocyte population as assessed by cell size and structure. Live (Aqua-) resting CD3+CD4+ T cells (CD25−CD69−HLADR−) were further sorted as Naïve (TN, CD45RA+CCR7+CD27+), Central-Memory (TCM, CD45RA−CCR7+CD27+), Transitional-Memory (TM, CD45RA−CCR7−CD27+), and Effector-Memory (TEM, CD45RA−CCR7−CD27−) cells. (PDF) Click here for additional data file. Figure S3 Different frequency of resting CD4+ T cell subsets in post-treatment controllers and HIV controllers. The frequency of TN, TCM, TTM and TEM cells among circulating resting CD4+ T cells in the PTCs and HICs. (PDF) Click here for additional data file. Figure S4 Weak contribution of long-lived resting CD4+ T cells to the HIV reservoir in the post-treatment controllers with declining levels of cell associated HIV-DNA. CD4+ T cell subset contribution to the resting HIV reservoir for 4 PTC for whom we observed a diminution overtime on their HIV blood reservoir levels and HIC, taking into consideration both the cell infection levels and their frequency. Results are expressed as the median percentage of the resting CD4 HIV reservoir with interquartile range [25%–75%] and minimum and maximum values. (PDF) Click here for additional data file. Table S1 HLA-class I alleles and characteristics of the CD8+ T cell response in the post-treatment controllers. (PDF) Click here for additional data file. Table S2 Comparisons of relevant HLA allele frequencies in post-treatment controllers, HIV controllers and the reference French population. (PDF) Click here for additional data file. Text S1 List of scientists and clinicians who are associated to the VISCONTI study. (PDF) Click here for additional data file.