DHIM Supporting Immunologic Investigations and the Identification of Immune Correlates of Protection

Abstract

Evidence suggesting that immune responses to dengue virus (DENV) have the potential for both beneficial and detrimental effects on the outcome of infection is a concern for dengue vaccine development. There is thus a great need to define measures of DENV-specific immune responses that reliably indicate when immunity is protective. The existence of 4 main DENV serotypes and the difficulty in defining which individuals have been exposed and to which viruses present challenges to defining immune correlates of protective immunity against DENV in field efficacy studies; experimental infection studies in humans offer a pathway to address these challenges.

Along with the initial isolations of dengue virus (DENV) during World War II, the finding of type-specific protective immunity led to optimism about the development of a vaccine [1]. Less than 2 decades later, the recognition of dengue hemorrhagic fever and its relationship to prior DENV exposure [2] drastically changed expectations. Some 50 years later, the hurdles to DENV vaccine development stubbornly remain. Despite the enormous strides in understanding the different manifestations of dengue disease and its pathogenesis, the scientific community has done poorly at predicting the results of clinical trials of vaccine candidates. Neither cell culture nor animal models have yet shown a reliable path toward selection of the best vaccine candidates for large-scale efficacy studies. Borrowing a tool from an earlier era, the Dengue Human Infection Model (DHIM) has been proposed as a viable strategy to collect data on vaccine efficacy [3–5]. As planning for DHIM studies proceeds, it is worthwhile considering the potential for this model to also address gaps in our knowledge of immunological correlates of resistance and/or susceptibility to DENV infection or to severe illness.

ADAPTIVE IMMUNITY IN DENGUE

Dengue is one of the clearest examples among natural viral infections in humans where the adaptive immune response can have either positive (protective) or negative (pathological) effects [6]. Evidence for protective immunity against DENV observed both in challenge studies performed in the 1940s and 1950s and in epidemiologic studies includes long-lasting resistance to reinfection with the same DENV serotype as well as short-term protection against symptomatic DENV infection with DENV serotypes other than the one that caused the earlier infection [7–10]. On the other hand, numerous epidemiologic studies have confirmed that the risk of more severe dengue illness, principally dengue hemorrhagic fever (DHF), is significantly higher during secondary DENV infections, that is, in individuals who have had prior DENV infections and have DENV-specific immunological memory prior to exposure to the second DENV serotype [11, 12].

Considerable debate is ongoing regarding the principal mechanisms explaining protective immunity and the increased risk for DHF during secondary DENV infections. Primary DENV infection induces both antibodies and T cells to DENV. DENV-specific antibodies, especially high-avidity, serotype-specific antibodies, inhibit DENV infection by blocking viral attachment or entry [13]. That this activity corresponds to protection against infection in vivo makes intuitive sense, although it has been well established that natural infection can occur even in the setting of high neutralizing antibody titers [14]. The role of T cells in protective immunity is less clear, and T cells are unlikely to give sterilizing immunity because some cellular infection would be needed for antigen presentation and recognition by T cells. Once initial infection occurs, production of antiviral cytokines such as interferon γ and lysis of DENV-infected cells could control infection sufficiently to minimize disease. Measurement of T-cell responses in one prospective cohort study [15] and associations of specific HLA alleles with a reduced risk for disease [16] support the in vivo significance of these mechanisms.

Some of the DENV-specific antibodies and T cells induced by a primary infection recognize corresponding epitopes on ≥1 of the other (heterologous) DENV serotypes, and these serotype-cross-reactive immune responses are therefore potentially able to function (either positively or negatively) during a secondary DENV infection. Partial protection afforded by heterologous immunity is presumed to follow similar mechanisms as described above. On the other hand, studies in cell culture and animal models have defined a variety of potential pathological responses that can be induced by serotype-cross-reactive immunity. For antibodies, attention has principally focused on the phenomenon of antibody-dependent enhancement of infection (ADE), which leads to higher levels of viral replication and can also modify cytokine production by DENV-infected cells [17]. Immune complex formation and complement activation, as well as molecular mimicry of endothelial cell antigens and/or components of the coagulation cascade [18], are additional effects of antibodies with potential relevance to disease pathogenesis in vivo. Several of the cytokines elaborated by DENV-specific T cells, such as tumor necrosis factor α, have pro-inflammatory activity and cause an increase in vascular permeability in cell culture and animal models. The phenomenon of heterologous immunity in mice, in which sequential challenge with other viruses results in enhanced tissue pathology [19, 20], provides a possible model for T-cell-mediated immunopathology in secondary DENV infection.

APPLICATIONS OF IMMUNE CORRELATES

Immune correlates of protection—in vitro assays that reliably indicate whether an individual has protective immunity—have advanced the development, testing, and utilization of other vaccines [21]. These assays can help to define the most important antigens for inclusion in a vaccine, select the optimal vaccination formulation and regimen, and predict the duration of vaccine efficacy and need for booster immunizations. Similar benefits would be anticipated for development of dengue vaccines from the identification of immune correlates [22]. Several considerations specific to dengue would need to be addressed, however. First, because protection against all 4 of the commonly identified DENV serotypes is the goal of most vaccines, immune correlates will need to be validated for each serotype. Second, given the association of more severe dengue disease with immunity to one or a few DENV serotypes, immune correlates will need to be able to distinguish protective immunity from pathological immunity, that is, an increased risk for severe illness.

CHALLENGES IN DEFINING IMMUNOLOGIC CORRELATES IN THE CONTEXT OF FIELD STUDIES

We and others have undertaken efforts to identify DENV-specific immune responses that protect against infection in cohort studies of natural DENV infection [23]. In this approach, subjects are observed for the occurrence (or not) of DENV infection from natural exposure. Surveillance of the cohort identifies incident symptomatic DENV infections and testing of sequential blood samples can reveal a rise in antibody titer indicating a subclinical infection (Figure 1). Baseline measures of DENV-specific antibodies and DENV-specific T cells can be correlated with the clinical outcome—no DENV infection, subclinical DENV infection, symptomatic mild DENV infection, and severe dengue illness. The same approach can be used in vaccine efficacy trials.

Figure 1.

Defining dengue-related outcomes in cohort studies. Within the study cohort, only a fraction of subjects will be exposed to DENV, some of whom may not be infected. Among cohort subjects infected with DENV, outcomes can be defined based on the occurrence (or not) and severity of symptoms. Without a means of defining exposure, it is not possible to distinguish subjects who are not exposed to DENV from those who are exposed but are not infected. Abbreviation: DENV, dengue virus.

Although potentially of great value, the strategy of defining immune correlates of protection in field studies suffers from several important limitations. Foremost among these is that methods do not currently exist to define exposure to DENV per se (as distinguished from infection). Regardless of the intensity of surveillance, cohort subjects who are classified as “no DENV infection” will therefore include both subjects who have been exposed to DENV and those who have not been exposed during the observation period. Only subjects in the former group can be accurately said to have protective immunity, whereas subjects in the latter group may or may not have protective immunity. If both groups of subjects are included in the analysis, it will profoundly complicate the interpretation of the results (Figure 2). As a consequence of this, most studies have focused on the comparison of subclinical DENV infection to symptomatic DENV infection [10, 15, 24]. These groups are more accurately defined in the context of field studies (Figure 1). However, both groups have, by definition, experienced infection with DENV, meaning that their immunity was not fully protective. Furthermore, in the typical situation where multiple DENV serotypes are circulating within the cohort during the observation period, it is difficult to prospectively ensure all groups being studied have the same risk of exposure and infection over time, and to accurately identify the serotype that caused the subclinical DENV infections, because serologic tests are unreliable [25]. A further limitation of cohort studies of natural DENV infection as compared to vaccine trials is that the preexisting immune responses were induced by ≥1 sequential exposures to individual DENV serotypes rather than simultaneous exposure to multiple serotypes.

Figure 2.

The challenge of identifying immune correlates of protection in the absence of measures of exposure. A, Hypothetical distribution of immune responses (either antibody or T-cell responses) in a study cohort participating in a vaccine efficacy study. B, Same distribution in the subset who are exposed to DENV during the observation period. The black bars represent subjects with protective immunity and the striped bars represent subjects without protective immunity; a perfect immune correlate of protective immunity is assumed in this example. Note the smaller values on the y axis reflecting the minority of subjects exposed (10% of the cohort in this example). C, Resulting distribution of immune responses in subjects defined as having no evidence of DENV infection during the trial; subjects with protective immunity (black bars) contribute a small fraction to the distribution of values. The actual case of 4 DENV serotypes with significant immunologic cross-reactivity further complicates the difficulty. Abbreviation: DENV, dengue virus.

THEORETICAL ADVANTAGES AND LIMITATIONS OF THE DENGUE HUMAN INFECTION MODEL

Given the challenges of field efficacy trials, the DHIM has the potential to fill an important niche in the immunologic evaluation of dengue vaccines (Table 1). The major advantage of the DHIM over field studies is that the exposure to DENV is well characterized; the DENV strain and serotype, inoculum size, and time of exposure are experimentally defined. Assuming that an inoculum is chosen that has a high likelihood of infecting control (DENV-naive) subjects, it should be possible to definitively classify the virologic outcome (occurrence or nonoccurrence of infection) in addition to the clinical outcome (symptomatic or subclinical). Thus, vaccinated subjects who do not experience infection after challenge can be accurately defined as having protective immunity.

Table 1.

Comparison of Natural Infection and the Human Infection Model for Studies of Dengue Viral Immunology

Characteristic	Natural Infection	Experimental Challenge
Study population	Relevant to endemic transmission	Can be controlled
Viruses	Authentic Highly diverse Poorly characterized	Cultured Narrow diversity Well characterized
Prior DENV exposures	Difficult to define	Better defined
Timing of DENV challenge	Uncertain	Known
Outcomes measured:
Uninfected exposed	No	Yes
Infected asymptomatic	Yes	Yes
Symptomatic mild	Yes	Yes
Symptomatic severe	Yes	Unlikely

Abbreviation: DENV, dengue virus.

The DHIM also makes it possible to select DENV serotypes for challenge based on the desired sample size, and sequential challenge could be performed to detect immunological interactions. Since all 4 DENV serotypes circulate in most endemic areas and outbreaks are difficult to predict, these issues would not be easily addressed under field conditions. Study subjects can be selected to test specific scenarios, such as prior exposure to DENV or other flaviviruses. Furthermore, the defined time of exposure will allow more thorough collection of data and blood (and other) specimens that will maximize the detection of viral replication in vivo.

On the other hand, several limitations make it unlikely that the DHIM could completely replace field trials in the immunologic assessment of dengue vaccines. Because of the need to standardize the challenge inoculum, and meet regulatory requirements for its formulation, the DHIM of necessity uses cell culture-derived strains of DENV. Although the challenge strains are being selected based on their ability to induce a dengue-like illness, subtle viral adaptations, particularly in the interaction with innate immunity, are likely and represent sources of error in interpreting the protective and/or pathological potential of vaccine-induced immune responses. A related limitation of the DHIM is that, for ethical reasons, the experimental risk of DHF must be kept to a minimum. As a consequence, defining correlates of pathological immunity will be extremely difficult. To the extent that protective immunity is the dominant effect over pathological immunity, this limitation may be of lesser importance to vaccine development.

REFLECTIONS ON IMMUNOLOGIC CHARACTERIZATION OF THE DHIM

In the context of further DHIM studies focused on the clinical and virologic characterization and standardization of the DHIM to support dengue vaccine testing, it will be invaluable to incorporate characterization of the immune response during the acute phase, including flow cytometry and gene expression analyses. Comparison with similar studies that have been performed in the context of natural DENV infection [26–31] should help to address the limitations noted above and also provide an important bridge to compare vaccine efficacy with results from field studies. In conjunction with data on viremia, changes in gene expression and cellular activation phenotype will reflect the innate immune response to the challenge DENV strains and identify any effects of in vitro passage. It will also be important to further explore the effects of prior DENV exposure(s) in the DHIM. The ability to define the serotypes of sequential DENV infections as well as the interval between infections will greatly advance the interpretation of antibody and T lymphocyte cross-reactivity and will also address potential sources of interference with the typical immune response to candidate vaccines.

CONCLUSIONS

The complexity of natural DENV circulation and the immunologic interactions between sequential DENV infections, along with the difficulty in defining past or current natural exposures to DENV, are likely to continue to complicate the evaluation of candidate dengue vaccines. Because so many potential scenarios exist in nature, and because of the difficulties in defining protective immunity, it is inevitable that there will be questions about the safety and efficacy of candidate dengue vaccines even after field studies to support regulatory approvals have been completed, and these questions could linger for years. Animal models to address these issues do not yet exist. The potential benefits to dengue vaccine development of immunologic studies in the DHIM merit consideration in evaluating the ethical justification of such experimentation.