Even with the major successes of cART, full life expectancy for patients living with HIV has not been restored. In a prospective study of 3990 HIV-infected individuals and 379,872 HIV-uninfected controls in Denmark, the probability of survival was examined in the period prior to cART (1995-1996), during early cART (1997-1999) and during late cART (2000-2005) 5 . There was a clear and substantial increase in survival following the introduction of cART in the late 1990s. However, even in the late cART period, life expectancy remained significantly less than population controls. In fact, the chance of a person with HIV reaching the age of 70 was 50% that of uninfected population controls. These findings are consistent with observations from other large cohort studies 6 .

The incidence of significant morbidity remains elevated despite successful cART due to complex interactions between drug toxicity 7 , persistent inflammation 8 and risk behaviours 9 . Multiple studies have demonstrated that people living with HIV are at increased risk of cardiovascular disease, metabolic disorders, neurocognitive abnormalities, liver and renal disease, bone disorders, malignancy and frailty (reviewed in 10 ). As a consequence, managing the complex care needs of HIV-infected individuals remains a major challenge.

Finally, despite the clear need for universal access to cART and the ongoing expansion in health systems, there remains a lack of financial resources to support life-long treatment, for everyone in need of treatment. Reaching all those in need of treatment is getting harder as donor contributions stabilize and treatment recommendations shift towards earlier initiation of cART 11 12 , which will increase the population of people judged to be in need of treatment. Furthermore, new HIV infections continue to outpace the number of people starting treatment. Even during the rapid scale up of access to cART in recent years, for every two people starting cART, there were five new infections 13 . This imbalance is unlikely to be reversed in the near future despite evidence that global HIV incidence is now declining 14 and the promise of more effective biomedical interventions, including circumcision and tenofovir-containing microbicides 15 16 .

Recent work, commissioned by the Clinton Foundation as part of the AIDS 2031 Project, has modelled the total projected annual AIDS resource requirements for low-and middle-income countries if cART scale up continues at current rates 17 . If HIV treatment is initiated at a CD4 count of 200 cells/mm³ and 40% cART coverage is achieved, the estimated costs by 2031 are predicted to approach $25 billion per year. If cART coverage instead reaches 80% by 2031, the annual cost of treatment is predicted to reach almost $35 billion 17 . Under this scenario, which is broadly consistent with the international community's commitment to universal access, the predicted cost of HIV treatment alone will account for almost half the US foreign aid budget by 2016 13 .

Following cART, HIV RNA in blood rapidly reduces to undetectable levels (<50 copies/ml). However, regardless of whether the patient has been on treatment for two years or 15 years, whether they have been on three drugs or six drugs, whether they started treatment within one year or 10 years of infection, as soon as treatment is stopped, the virus rapidly rebounds. The question then is: where is the virus sitting while the patient has a viral load of less than 50 copies/ml?

More than 10 years ago, several groups identified the persistence of virus in long-lived latently infected cells, measured as HIV DNA. They demonstrated that upon stimulation, these silent viral genomes can be reactivated and subsequently produce infectious viral particles 18 19 20 . More recently, using a highly sensitive assay that detects HIV RNA in plasma down to 1 copy/ml, several groups have shown persistent low-level viremia of around 3-5 copies/ml in 80% of patients 21 22 . In other words, there is no such thing as an undetectable viral load and the virus clearly persists. Currently, some of the major research questions are: what is contributing to this low-level viremia and persistent DNA, and will it ever be possible to eliminate this residual virus?

There are likely to be at least three major barriers to curing HIV. These include the persistence of long-lived, latently infected cells, residual viral replication and anatomical reservoirs. Latently infected cells are predominantly resting CD4+ T cells 18 19 20 , but also include other long-lived cells, such as monocyte/macrophages 23 and astrocytes 24 25 . Latency represents the biggest challenge to finding a cure.

In vivo, HIV latency occurs in resting CD4+ T cells either as pre-integration or post-integration latency. Pre-integration latency refers to unintegrated HIV DNA that is unstable and will either degrade or will integrate into the host cell genome, usually following cell activation 26 . Post-integration latency refers to the presence of integrated HIV DNA in cells that are not actively producing viral particles. The major reservoir of cells that harbour post-integration latency in vivo are resting memory CD4+ T cells 27 28 . Once integration occurs, the virus can persist in these cells for long periods of time, unaffected by antiretroviral drugs or host immune recognition 19 29 . Post-integration latency is therefore critical for the maintenance of the HIV latent reservoir.

In activated CD4+ T cells, the virus life cycle is efficient, with rapid integration, virion production and subsequent death of the infected cells. In contrast, infection of resting CD4+ T cells is difficult to establish in vitro due to multiple blocks in the viral life cycle 30 31 . However, resting CD4+ T cells are clearly infected in vivo 32 33 , as well as ex vivo, in tissue blocks 34 35 , and contain stable integrated forms of HIV.

In vitro, our group has clearly demonstrated that latent infection can be established in CD4+ resting memory T cells, following incubation with multiple chemokines that bind to the chemokine receptors highly expressed on resting CD4+ T cells 36 37 . Studies such as this support the hypothesis that latency can result from direct infection of resting memory CD4+ T cells, possibly as a result of exposure to soluble factors found in lymphoid tissues. An alternative possibility for infection of resting CD4+ T cells is the reversion of an infected, activated cell to a resting state, which has also been demonstrated in vitro 38 39 40 (Figure 1).

These latently infected CD4+ T cells are thought to be extremely long lived 41 ; however, it is highly likely that this pool of cells is also maintained by homeostatic proliferation 27 . Latently infected cells can intermittently release virus following activation, making them a major barrier to cure. Latency was originally described in one particular subset of CD4+ T cells called central memory T cells. Because these cells can persist for decades, they ensure the maintenance of long-lasting cellular immunity, but also constitute an extremely stable cellular reservoir for the virus 18 19 .

Over the past few years, several groups have identified that latency can exist in a range of CD4+ T cell subsets, including transitional memory T cells, naïve T cells, thymocytes and multipotent progenitor cells or stem cells 27 42 43 44 . Together, these cells constitute the latent reservoir. There is also some indirect evidence that the normal process of homeostatic proliferation maintains the number of latently infected naïve and transitional memory cells. Following mitotic cell division, both daughter cells contain integrated HIV DNA, meaning that this reservoir may be replenished or even increase in size on cART 27 44 .

The alternative explanation for persistent virus is that there is ongoing virus replication in activated T cells. In other words, cART is effective, but not 100% effective. However, it is currently unclear how much residual replication contributes to HIV persistence. There are several pieces of evidence that argue against residual replication, including the very stable sequence of low-level viremia in plasma 45 46 and the absence of drug-resistant virus in either plasma or CD4+ T cells 47 48 .

Finally, we know that HIV can hide in anatomical reservoirs, such as the brain 49 , the gastrointestinal tract 50 and the genital tract 51 . In the gastrointestinal tract of patients receiving cART, persisting infected cells are almost 10 times more frequent than in blood 50 52 . In these anatomical sites, virus can persist in activated, replicating cells, as well as long-lived, latently infected cells, such as dendritic cells, macrophages and astrocytes. These sites may also have unique barriers to entry of cART, which limit the penetration of drugs.

There are two potential strategies for cure. The first is what we might consider an "infectious diseases model" of cure, where the pathogen is treated and it disappears all together. This would require the elimination of all HIV-infected cells and for patients to have an HIV RNA count of less than 1 copy/ml. This is now commonly referred to as a sterilizing cure. The alternative approach would be to aim for remission or what we might consider a "cancer model" of cure, where an individual would have long-term health in the absence of treatment, with perhaps low-level viremia at less than 50 copies/ml. This is commonly referred to as a functional cure (Table 1).

There are examples of both a sterilizing and functional cure that we need to learn from when designing new strategies for curing HIV. The recent case report of a German patient with acute myeloid leukemia, who received a bone marrow transplant from a donor who was resistant to HIV, is the only current example of a sterilizing cure 53 . The bone marrow donor carried a mutation in the CCR5 gene, a 32-base pair deletion, which knocks out expression of CCR5, the major coreceptor for HIV. Following transplantation, the patient stopped cART due to interactions with his chemotherapeutic drugs. Interestingly, virus did not rebound in the blood of this patient, and in more detailed studies, including multiple biopsies of his gastrointestinal tract, analysis of his cerebrospinal fluid (CSF) and lymph nodes, there were no detectable signs of HIV. The patient is now more than three years post transplant and HIV is still not detected. While a strategy of using bone marrow transplantation with a CCR5 mutant donor is not a realistic cure for HIV given the toxicity of the treatment, we need to comprehensively study this patient to fully understand how and why HIV was eliminated.

Elite controllers are another group that will teach us a lot about trying to achieve a functional cure. Elite controllers represent a unique group of patients who are able to achieve a consistent HIV RNA of less than 50 copies/ml in the absence of treatment 54 . There have been multiple studies examining the role of genetics, the virus and the immune response in elite controllers 55 56 57 . A consistent result from this work is the persistence of a robust HIV-specific T cell response in elite controllers, providing supportive evidence that inducing an effective immune response, perhaps via vaccination, may be a strategy to achieving a functional cure.

However, to date, the use of therapeutic vaccination in patients receiving cART has not been successful 58 . It is also important to note that approximately 7% of elite controllers experience a decline in their CD4+ T cells despite maintaining a viral load of less than 50 copies/ml. Ongoing virus replication and evolution, in addition to enhanced immune activation, has also been observed in these patients 55 59 .

First, universal access to cART still remains the major priority for the management of patients with HIV. cART will always be a part of any strategy that may lead to a cure. Second, there is an urgent need for clinical trials. There are several compounds that look promising in the laboratory, including vorinostat and IL-7. It is highly likely that a combination of approaches will be needed together with cART intensification. These studies are likely to have the greatest possibility of success in patients who initiated cART shortly after acute infection.

Importantly, more active community engagement in this work is critical. Basic science issues are often perceived as highly technical and without impact on the daily lives of infected or affected communities. It is, however, crucial for community representatives and basic science researchers to work together to systematically address the barriers and challenges that hold us back from finding a cure.

Clinical trials will be needed to move the field forward and it is essential that affected communities are involved in these efforts as true partners. For example, it is important that community representatives are involved in longer term strategic planning for eradication studies, as well as the planning of individual studies. Community members should be invited to join the steering committees, advisory boards, and data safety and monitoring boards of these studies. Additionally, they should join together with health professionals in raising the awareness and understanding of issues related to HIV persistence and potential eradication. Such an alliance will also be critical for increasing the funding support for basic science research in the field of HIV.

As we move forward into clinical trials, we also need to carefully consider what the most appropriate endpoints should be. Can we use surrogate markers of the reservoir, including HIV DNA and plasma viremia? Are there circumstances in which it will be acceptable to trial treatment interruption with the well-documented risks of viral rebound 84 ?