Early attempts of immunotherapy, during the pre-HAART era, included non-specific (e.g. IFN and IL-2) and HIV-specific (e.g. Remune) immune interventions in patients who were either untreated or received mono- or dual drug therapy. In some cases positive effects have been reported, but these trials were small and usually not well controlled. During the last 10–15 years, the concept of immunotherapy as a supplement to full HAART has been further developed, with the ultimate aim to reduce or even replace drug therapy.
Amongst the non-antigen-specific immune approaches, systemic IL-2 has been exhaustively investigated and was ultimately shown not to provide clinical benefit in addition to HAART. It remains to be seen whether systemic use of other common γ chain cytokines, such as IL-7 or IL-21, could be useful for particular indications (e.g. to counteract lymphopenia). The newer “non-specific” approach is to block negative regulatory pathways, such as PD-1 or CTLA-4 signaling, both over-expressed during HIV infection. Experimental treatment in SIV-infected macaques indicated that PD-1 blocking helped to control viremia and to reduce immune activation. In contrast, CTLA-4 blocking resulted in increased viremia, most probably as a result of an unfavorable balance between too much non-specific and insufficient specific immune activation.
The obvious goal of therapeutic vaccination is to stimulate effective HIV-specific T cell responses, primarily cytolytic/virus suppressive CD8 T cells with supporting CD4 T cell help, while avoiding increased HIV-susceptibility of target cells. To that end, protein or particulate antigen is conceptually not the best option, since it primarily results in CD4 T and B cell activation: antibodies are considered ineffective and CD8 T cells seem essential in immune protection post infection.
Repeated im vaccination with plasmid DNA encoding various HIV antigens, along with cytokines, was successful in lowering VL in macaques, but this has not yet been confirmed in humans, despite repeated attempts. However, the immunogenicity of DNA vaccination is being improved by innovative constructs and enhanced delivery systems . Especially for applications in less sophisticated environments, DNA vaccines have the advantage of being stable and to carrying no infectious risk, but because they are considered “genetic therapy”, the development of DNA vaccines is subjected to strict regulations .
Recombinant viral vectors have been applied rather extensively in therapeutic vaccination trials. In most cases poxviruses (e.g. canary pox, fowl pox and MVA) and occasionally Adenoviruses were used. A fair number of encouraging results were observed in infected/treated macaques and patients in terms of T cell responses, and in some cases also VL reduction after ATI was observed. However, the variable results, obtained with the most extensively used canarypox platform, ALVAC, illustrate that apparently similar trials can have quite discrepant outcomes. Again, this may be the result of a delicate balance between the beneficial enhancement of anti-HIV CD8 T cell immunity and the deleterious effect of non-specific immune activation, inducing increased cellular susceptibility to HIV infection.
DC-based therapy has been developed in a therapeutic context, with various antigenic formats and corresponding loading strategies. Remarkably positive results in terms of viral control have been obtained with autologous inactivated virus loaded DC. DC trials with safer and less cumbersome formats (e.g. RNA) are ongoing, but the first results suggest that there is a need for improvement and further simplification of this methodology.
Collectively, it seems that HIV immunotherapy, although still highly experimental, is a viable option to explore. Although some beneficial effects have been described in untreated individuals, most evidence today argues that more effective immune responses can be induced under HAART coverage. In most trials, potentially favorable T cell responses were induced, while effects on VL have at best been transient. This limited success should not come as a surprise, since in most cases antigens from one particular or a limited number of HIV strains were used as immunogen and even in those trials, where the “autologous” virus or sequences thereof were used, epitopes from which the immune system had already escaped may have been presented.
To overcome these restrictions, new options are available. One is to immunize with a set of the most conserved, subdominant epitopes, mainly in the structural core genes (i.e. Gag) in order to focus and “teach” T cells to effectively react against at least some determinants which the virus cannot possibly escape from without completely losing infectivity. Obviously, this strategy might require repeated immunizations with sufficiently large numbers of different stretches of antigens, in view of the requirement to accommodate the HLA restrictions and to allow T cell receptor “maturation” by gradual somatic mutations. One such approach was proposed by Letourneau et al. . With these antigens, strong immune responses were already induced both in mice and macaques, using HIV conserved alphavirus replicon DNA and electroporation . A similar approach, focusing on different epitopes was proposed more recently by C. Brander . A second alternative is the so-called mosaic gag approach, where a large number of natural HIV sequences are included by in silico recombination (ftp://ftp-t10.lanl.gov/pub/btk/mozaïek) for maximum coverage of potential T cell epitopes of HIV-1 group M, including potential escape variations [153, 154]. This approach has successfully been applied in macaques, using DNA + recombinant vaccinia boost [155–157]. Interestingly, this viral construct also induced broad responses in PBMC from HIV-1 B or C infected subjects .
The next important question to address is how to deliver this improved antigen in vivo. Clearly, genetic constructs, either plasmid DNA, mRNA or recombinant viruses are suitable. All three formats have advantages and drawbacks as explained. Our own preference goes to mRNA, as it is flexible and clean, can include a large number of antigens and variants, and does not carry any infection or insertional risk or risk for vector-related adverse immune reactions . This safety advantage of mRNA-based vaccines as compared to plasmid DNA or viral vectors is reflected in their classification by the authorities (FDA in the USA and Paul Ehrlich Institute in Germany) as no gene therapy approaches. This relaxes the requirements concerning preclinical toxicology studies in animals .
Another feature of mRNA is, however, its biological instability (sensitivity to ubiquitous RNAses); and therefore it has mostly been used to load patient-derived DC ex vivo, a strategy which is not applicable at a larger scale. Recent evidence, however, indicates that mRNA can be chemically stabilized to be injected as “naked” RNA in vivo (e.g. intranodally) and/or formulated with either cationic lipids or positively charged proteins. The latter technologies have been refined over the years, and are now ready for clinical application [158–160].
In addition to the antigen and its format or formulation, further signals may be required to optimize the immune responses. Especially if the format is DNA or RNA, a clearly defined genetic adjuvant is most attractive, as it will restrict the costimulatory effect to the site where the immune response is elicited. Th1-skewing (IL-12, IFN-γ), common γ chain or cytokines, TNF ligand members or Toll-like receptor ligands have been used with some success . From our review two new candidates emerge: IL-21  and PD-1 blocking agents .
The desired clinical outcome of a therapeutic vaccine is to keep the VL under control and restore CD4 T cell counts through CD8 T cell-mediated suppression of viral re-emergence. However, there are no easy correlates for this protective effect while the patient is still on HAART. A straightforward ELISPOT using a standard subtype B set of peptides provides a first indication of immunogenicity, but it does not represent a true correlate for protection. Expanding the breadth of the T cell responses by testing many peptide variants, “potential T cell epitopes” (PTE) , increasing avidity by testing low peptide concentrations or improving the “quality” in terms of lytic capacity of CD8 T cells or poly-functionality as well as expansion of HIV-specific effector-memory or central memory CD4 and CD8 T cells have all been proposed as useful [6, 8], but these tests remain somewhat indirect correlates and require rather delicate and cumbersome experimental approaches. As mentioned, several authors, including ourselves, have proposed to measure the vaccine-induced capacity of the patients’ CD8 T cells to suppress various HIV isolates in vitro, as this function has repeatedly been associated with the “elite controller” status [18–20, 111]. Obviously, this type of functional test is also difficult to standardize, but it has the conceptual advantage of being a direct correlate of the desired protective effect.
Another important parameter to consider is whether immunotherapy or any other intervention aiming at a “functional cure” could reduce the size of the latent reservoir. Several assays have been proposed to measure residual replication-competent virus in purified CD4 T cells, based on (enhanced) viral culture as well as on an array of PCR-based methods quantifying intracellular viral mRNA (unspliced or multiple spliced) or proviral DNA. With regard to the latter, it is important to determine integrated versus non-integrated proviral DNA as well as the precise cell subset of the reservoir (activated versus resting T cells, central-memory versus terminally differentiated, macrophages, dendritic cells etc.). Moreover, the measurement should not be limited to blood, but also consider reservoirs in lymphoid tissues (peripheral or gastro-intestinal-associated lymph nodes) as well as “sanctuary sites” (such as the CNS or the genital tract). While the conceptual importance of these measurements is evident, these assays are presently not standardized and it remains to be determined which ones are most predictive of what we really want to know: will our intervention be able to mitigate or abrogate viral rebound after subsequent interruption of HAART [1, 30]?
Since there is no definite ex vivo/in vitro correlate of protection and since the SIV macaque model does not necessarily predict effects in human trials, one needs to proceed to a human trial with analytical treatment interruption to assess the ultimate clinical effect. In view of the observed variability in trial outcome, there is a clear requirement for a carefully matched and sufficiently large placebo group. Treatment interruption, even when guided by CD4 T cell counts and carefully designed to avoid emergence of drug resistance, is being considered as harmful in the long run since the SMART trial . To address this problem, Routy recently reanalyzed the SMART data for the effects of a limited duration of treatment interruption (e.g. 16 weeks) in patients on stable HIV-suppressive HAART with CD4 T counts > 400/μL, a CD4 T cell nadir of > 200/μL and without concomitant diseases. He concluded that ATI under those conditions is safe and thus acceptable in the context of a well monitored therapeutic trial with timely resumption of HAART in case of rebound and/or CD4 T cell drop .