The Role of the Mosquito in a Dengue Human Infection Model

Abstract

Recent efforts to combat the growing global threat of dengue disease, including deployment of phase IIb vaccine trials, has continued to be hindered by uncertainty surrounding equitable immune responses of serotypes, relative viral fitness of vaccine vs naturally occurring strains, and the importance of altered immune environments due to natural delivery routes. Human infection models can significantly improve our understanding of the importance of certain phenotypic characteristics of viral strains, and inform strain selection and trial design. With human models, we can further assess the importance of the natural delivery route of DENV and/or the accompanying mosquito salivary milieu. Accordingly, we discuss the use of mosquitoes in such a human infection model with DENV, identify important considerations, and make preliminary recommendations for deployment of such a mosquito improved DENV human infection model (miDHIM).

The global incidence of dengue fever and more severe forms of the disease was recently estimated at 96 million; along with an additional 294 million subclinical infections, this increases the assessment of the worldwide dengue burden by a factor of three [1]. Currently, the only available mitigation strategy is the attempt to interrupt transmission by vector control, though numerous investigations are underway to develop vaccines and other therapies aimed at preventing infection and limiting severe disease [2–4]. The most recent of these developments was an advanced deployment of an experimental dengue vaccine in Thailand [5, 6]. One hypothesis regarding the incomplete success of this effort was the variability of fitness within serotypes, specifically of the vaccine strain of dengue virus (DENV) serotype 2 and the putatively more fit, naturally circulating strain of DENV 2 [6]. Although this effort, along with many others, is in ongoing clinical trials, this result raises the question as to whether the protective effect of these formulations is being adequately measured by current investigational pipelines.

Currently, there is not an opportunity for data collection on therapeutic performance (ie, vaccine efficacy) in humans under near natural conditions before field trial deployment under advanced investigational new drug (IND) phases. In addition, therapy performance measured during early safety phases does not account for preexisting antibody to mosquito salivary antigens, or in vivo protection against related viral strains, which have the potential to confound interpretations [7, 8]. In vitro methods, while critical, cannot be anticipated to directly correlate to the performance of a therapy under natural conditions, such as within the natural host. Further, the results of studies in laboratory animals such as mice and nonhuman primates are also difficult to extrapolate to the natural DENV system, principally due to the lack of a robust animal model that closely approximates human susceptibility, response, and disease [9, 10]. These methods do, however, represent an important collection of results that has informed scientific and pharmaceutical hypotheses.

Historical studies using a human infection model have led to important insight into the mechanisms of transmission and pathogenesis of DENV [11–13]. Consequently, the inclusion of a human infection model would: (1) provide critical information as to the formulation of vaccines and other therapeutic interventions, and (2) when combined with dengue infected mosquito challenge, provide enhanced understanding as to the mechanisms and efficiencies of infection establishment and therapeutic performance under conditions nearer to the natural experience. As item 1 above has been addressed by others [14, 15] and is the specific topic of other elements of this special JID supplement series, we will explore the use of mosquitoes in such a dengue human infection model with DENV, identify important considerations, and make preliminary recommendations for deployment of such a mosquito improved DENV human infection model (miDHIM).

It is critical to recognize that the delivery of a viral challenge by needle to a human might bypass important interactions between the pathogen and the natural transmission environment. During mosquito probing, local tissues in the dermis are being physically damaged while a diverse array of salivary proteins are injected in an effort to locate blood, prevent coagulation, and limit disadvantageous vertebrate immune responses [16–19]. Some of these immune responses are innate, such as shifting the immune response toward a Th2 mediation and down regulation of key antiviral molecules (interferon β [IFN-β], interferon γ [IFN-γ], Toll-like receptor 3 [TLR3], inducible nitric oxide synthase [iNOS]); whereas others are adaptive, including immunoglobulin E (IgE) and immunoglobulin G (IgG) directed against many of these salivary proteins detected during prior mosquito exposures [16, 20]. If the foraging mosquito is infectious, DENV will also be deposited in the dermis during probing and be subjected to this altered immune state, potentially encouraging the infection to establish in cells and tissues other than would be encountered during intradermal (in the absence of mosquito salivary proteins), subcutaneous, or intravenous needle-delivered challenges [21–23]. The development of miDHIM will provide much needed data and insight as to whether mosquito saliva/salivary proteins are essential components in DENV transmission efficiency, infection establishment, and/or disease presentation. This will in turn inform research concerning vaccine and therapeutic development and transmission risk assessment efforts.

Definition(s) of Mosquito Improved DENV Human Infection Models (miDHIM)

We broadly define mosquito improved human infection models as any vector-borne pathogen infection model using mosquitoes as vectors to deliver pathogens to (or acquire pathogens from) vertebrate animals, or using the components of mosquito saliva/salivary glands in conjunction with a pathogen in an effort to account for the effect of the vector upon transmission, or vertebrate infection establishment and disease progression. As such, it is necessary to consider the possible methods of miDHIM use, which in its purest form would involve the deposition of DENV into the study subject from the bite of an intact, infectious mosquito. However, as the absolute quantity of the viral inoculum is incalculable by this approach, an alternative method would be to inject DENV intradermally into sites where non-infectious mosquitoes had just probed (and presumably salivated). Further abstracted methods include coinoculation of DENV with ex vivo prepared salivary gland extract, with experimentally collected saliva, or with a priori selected salivary protein(s). Incorporating components of the salivary inoculum into an intradermal needle-delivered challenge could thereby provide some approximation to the natural experience [16]. Table 1 summarizes the general advantages and limitations of these related methodologies.

Table 1.

Comparison of Different Mosquito Improved Dengue Human Infection Model (miDHIM) Options

	Infectious Mosquito	Coinoculation With DENV
		Probing	Saliva	SGE	Proteins
Near natural route of exposure	✓✓✓	✓✓	✓✓	✓	✓
Calculable/known viral titer	✓	✓✓✓	✓✓✓	✓✓✓	✓✓✓
Known concentration of saliva/SGE/proteins	✓	✓	✓✓✓	✓✓	✓✓✓
Specificity of immune response modulation	✓✓	✓✓	✓✓	✓	✓✓✓

Check marks indicate the magnitude of added value or attribution each miDHIM option would bring to the understanding in the far left column.

Abbreviations: DENV, dengue virus; SGE, salivary gland extract.

Additionally, data from other disease systems that make use of these approaches can provide an intuitive framework for their application to DENV experimentation using miDHIM. In the case where miDHIM is defined as the bite from an infectious mosquito, infection studies with malaria found the efficiency of infection of sporozoites improved when inoculated by Anopheles spp. mosquito bite as compared to intravenous infections in mice [24] or intradermal infection in humans [25]. Similarly, mice infected with West Nile Virus (WNV) by the bite of Culex spp. mosquitoes developed more rapid and intense viremias than mice infected by needle [22]. Mechanisms to explain these findings remain elusive; however, in mice infected with chikungunya virus via mosquito bite, significantly down-regulated IFN-γ and interleukin 2 (IL-2) were detected compared with mice infected by needle, again indicating a polarization to the Th2 response and subversion of IFN-γ stimulated antiviral mechanisms [21].

Alternatively, miDHIM approaches using coinoculation of whole mosquito saliva or salivary gland extract (SGE) may be attractive options as they offer increased quantitative control over the viral inoculum, as well as the concentration of saliva/ SGE. These methods would minimize human study subject discomfort and compliance problems that may occur with protocols using live, foraging mosquitoes; however, other potential issues encountered with these ex vivo tissue preparations are the lack of basic formulation and safety data, which could cause significant consideration by institutional review boards (IRB). Several studies have shown that coinoculation of either saliva or SGE can alter the infection dynamics of arboviruses in murine models [7, 23, 26].

Advancing our understanding of the role of specific salivary proteins in DENV infection establishment and perhaps the role of preexisting immunity to some of these proteins, miDHIM could be designed to target putative mechanisms for therapeutic discovery. For example, Aedes aegypti salivary gland protein (SAAG-4) was shown to induce CD4⁺ cells to produce interleukin 4 (IL-4), again pushing the immune response to a Th2 response [27]. Additionally, a synthetic protein derived from a mosquito salivary allergen, shown to have identical qualities when compared to the native protein, (rAed a 2) bound to IgE of individuals with a history of mosquito allergy. Researchers also observed skin reactions in these patients, indicating that this protein induced an allergic reaction (Th2 immune response) [20]. This overwhelming trend toward a Th2 response driven by saliva and its components indicates that DENV infection in the context of these proteins is encountering a different immune-environment than if only the intracellular, antiviral response (Th1) were to be induced, as is likely with a needle inoculation of virus alone. By investigating findings such as these through miDHIM of human DENV, we mirror more closely the natural transmission, making these data more relevant to existing clinical findings related to DENV transmission and pathogenesis. This, in turn, could lead to the more rapid assessment of putative antiviral mechanisms and development of therapies.

Although each of these options warrants attention, and one may indeed provide a superior approach to the others, the focus of the rest of this article will be on miDHIM where the delivery of virus and the accompanying salivary milieu is via infectious mosquito bite, as this represents the route nearest to the natural experience and requires the most detailed consideration for experimental use.

Lessons From Nonhuman Animal Mosquito Improved DENV Infection Models

Along the limited availability of extant human infection models, research into the transmission and pathogenesis of DENV has been impeded by the lack of robust animal models. However, recent improvements have been made in several mouse models and are anticipated in nonhuman primate models. We briefly review important infection animal models here but recognize there are other important models that investigate alternative transmission routes (eg, intracerebral inoculation) and otherwise do not consider the role of the vector [9, 10, 28]. Immunocompetent mice have been used in DENV studies, but these results are perhaps hard to interpret due to the high dose needed to achieve successful infection and general lack of symptomology [28, 29]. However, these studies have provided valuable insight into the importance of viral dose and the interferon response in DENV infection. An oft-used mouse is the interferon (type I and II) receptor knockout strain, AG129, which has been an important tool for the study of DENV pathogenesis. This model was critical in advancing our understanding of the role of interferons in DENV infection establishment and disease progression [9, 10, 28]. Likewise, humanized mice have also been used to investigate pathogenesis and infection kinetics of DENV. Viremia was enhanced in mice when DENV infection was established via infectious mosquito bite, and the presentation of symptoms, namely, thrombocytopenia and erythema was also associated with mosquito transmission [30]. This study also reported a significant increase in the number of days that DENV RNA was detected in mosquito infected mice vs intradermal injection of virus alone. Interestingly, prolonged RNAemia was also detected in mice injected intradermally with DENV and either coinoculation with mosquito saliva or simultaneous feeding of uninfected mosquitoes near the injection site [30].

Another murine model, this strain lacking type I interferon (IFN-α/β) receptors, was used to generate a much more virulent strain of DENV 2 by serial passage within the mouse [31]. Such adaptation of the virus without the alternating passage in mammal and insect cellular environments has been described in vitro elsewhere and further points to the importance of the mosquito as not only a delivery method but a necessary biological factor in the evaluation of virulence potential of selected therapeutic strains [32–34]. In a similar system, infection of mice deficient in interferon regulatory factors 3 and 7 (which renders a significantly diminished production of type I interferon) with a nonadapted strain of DENV2 by Ae. aegypti achieved a higher average peak of viremia, as well as an extended length of viremia. In addition, there was a down-regulation of the IFN-γ response associated with mosquito feeding vs needle inoculation [35], and presensitization via mosquito bite induced a generalized suppression of the inflammatory response and several antiviral signaling pathways [36]. Importantly, this mouse model was shown to support the recapitulation of the transmission cycle supporting bidirectional transmission [35], an important tool for the further characterization of critical elements in DENV transmission efficiency.

Nonhuman primates (NHP) are an attractive DENV infection model given their relatedness to humans and presumptive role in the tangential sylvatic cycle of DENV [37]. However, pathogenesis and viremia are limited in NHPs subcutaneously infected with DENV, both in magnitude and length, unless immunosuppressive drugs are used [10, 38, 39]. Intravenous injection of higher doses in Rhesus monkeys did increase symptom presentation [40]. Due to the high homology between chimpanzee and human immunoglobulins, chimpanzees have been used to describe robust antibody responses to DENV [9, 41–44], though this model is no longer tenable due to a National Institutes of Health (NIH) ban.

Challenges and Recommendations for miDHIM Development

Development of miDHIM will need to address several challenges. These challenges and accompanying recommendations and proposed preliminary experimentation is summarized in Figure 1. First, miDHIM must be developed with the safety of human participants as a foremost goal. Thus, appropriate IRB approved protocols should be developed by experts in the fields of clinical immunology (including allergy), entomology, and arbovirology. It is important to note that saliva and mosquito bites have not been definitively associated with severe disease and were shown to have no effect on morbidity and mortality in mice infected with WVN or DENV [22, 35]. However, there is evidence of an association between the Th1→Th2 shift and disease progression, clearly calling for more detailed experimentation to clarify this relationship [8].

Figure 1.

Summary of the challenges, recommendations, and preliminary experimentation needed to further the development of miDHIM. MS = mosquito saliva to be either introduced by probing of infectious mosquitoes during DENV delivery, probing of noninfectious mosquitoes shortly before/after inoculation with DENV via needle, or coinoculation of whole saliva and DENV via needle; MSC = mosquito salivary components, which could mean coinoculation of salivary gland extract which contains saliva and other structural/ tissue proteins, or coinoculation of specific isolated proteins derived from mosquito saliva. Abbreviations: DENV, dengue virus; IRB, institutional review boards; miDHIM, mosquito enhanced infection model.

Second, we must recognize that in endemic areas and regions where DENV is likely to emerge, there will be a significant portion of the population that have preexisting antibody to DENV as well as mosquito salivary components, and thus a rather mature immune response to one or both. Certainly, the level of response to both DENV and mosquito salivary components will be highly variable and dependent on many things including socioeconomic status and related human behaviors [45–48]. Baseline levels need to be determined and relative differences assessed in these individuals, or else trials should be conducted in areas where the primary vector (Ae. aegypti) and its close relatives (Aedes (Stegomyia) spp.) are absent.

Third, there is the question of what mosquitoes and strains to use. Studies have shown that colonies of Ae. aegypti established from populations separated in space and time can have significant differences in their vector competencies [49, 50]. Thus, we recommend the use of reagent grade mosquitoes from a long-standing, established colony, such as the Rockefeller strain [51]. In addition, there is the question of secondary vectors. Ae. albopictus has been implicated in autochthonous DENV transmission in such areas as France and Croatia, and human DENV cases were positively associated with the presence of Ae. albopictus in households in Bangladesh [52–54]. Preliminary experiments are needed, perhaps in the alternative miDHIM discussed above, to determine whether Ae. albopictus saliva is similar enough to rationalize its dismissal.

Fourth, the method whereby mosquitoes become and are determined infectious should be a standardized procedure: oral challenge of virus or intrathoracic inoculation. Intrathoracic inoculation of mosquitoes with arboviruses usually leads to more rapid, and higher proportions of, disseminated infections compared to oral challenge [55–57]. However, this is accomplished by bypassing the midgut and the potential there for an important infection barrier [58]. If true strain fitness is being investigated, then likely the bypassing of such a critical barrier is not ideal. However, if only the delivery of DENV via mosquito (or with salivary proteins) is desired, then this method may be sufficient. Conversely, there is no doubt that the more natural route of exposure— oral challenge via infectious bloodmeal—is best when characterizing strains, but, as stated above, the uncertainty of the infectiousness of an individual mosquito used for such exposures could result in variable exposure doses. In addition, dissemination of DENV through the mosquito after an oral challenge takes many days longer than intrathoracic inoculation [59]. Notably, the peak concentration of virus detected in the salivary glands is often similar regardless of which method is initially used [60]. There are pros and cons to each method, and again, preliminary experimentation could assist in determining whether one method should become standard or how the 2 methods could be compared and how to interpret such comparisons.

A central challenge for miDHIM is to standardize methods for determining transmission capability of DENV exposed mosquitoes that are being prepared for use in human challenge studies. Researchers test for DENV (or viral RNA) in several different body parts to determine whether a mosquito is likely infectious: head squashes, legs, or salivary glands; however, as the others are terminal, only leg-based testing can potentially be used in advance of the experimental exposure. Other methods include capillary feeding where mosquito saliva is harvested and can be directly tested for the presence of virus via molecular methods (quantitative reverse-transcription polymerase chain reaction) or infection assays (cell culture infection, naive mosquito intrathoracic inoculation of saliva) [61–67], although these methods also invariably cause serious damage to the mosquito during collection, greatly reducing the probability of successful feeding in subsequent human challenge studies. The use of sugar feeding to detect mosquito-borne pathogens has been used successfully and is a nonlethal method but again provides only an estimate of virus in the sample population of mosquitoes and not the absolute number [68]. To better support these studies with miDHIM, research is needed to identify and further define the correlative relationship between transmission success and these measures, so that some quantification of uncertainty around transmitting mosquitoes can be made.

The number of mosquitoes to be used must also be standardized, as it is comparable to dose-response of DENV inoculation. In a malaria challenge model, it was determined that 5 mosquitoes achieved the most reliable infection results [69]. DENV studies have reported conflicting results of whether a single mosquito can be used to infect mice, but are in agreement that the number of mosquitoes sufficient may be similar to that used in the malaria challenge [30, 35].

Lastly, the use of miDHIM comes with the caveat of a lack of control of DENV delivery dose when whole mosquitoes are used. Studies have suggested that Culex spp. mosquitoes inoculate 1 × 10³–1 × 10⁵ infectious particles of WNV per bite [57]. Indeed, the same variability no doubt exists in the titer of DENV transmitted from Ae. aegypti mosquitoes. In some cases, this may be an argument for the use of miDHIM as the coinoculation of a known DENV titer by needle along with saliva or salivary proteins. However, there is some benefit in allowing for the natural variability of dose delivery, depending on the question. Certainly in the interrogation of transmission dynamics, the inherent heterogeneity of dose from mosquito bite delivery is not only sufficient but also ideal. If investigating infection dynamics and immune response to therapeutics, however, control of dose is important. Again, this challenge folds into the previous issue of the number of foraging mosquitoes as it relates to DENV infection establishments and kinetics.

DISCUSSION

The Added Value of Mosquito Improvement to the Dengue Human Infection Model

Scientific curiosity compels us to understand the natural mechanisms of pathogen transmission and the resulting infection dynamics; in this case, the role of the mosquito as the natural infection mechanism of DENV. Beyond these basic scientific questions though, there are many potential uses for miDHIM to improve therapeutic development, provide avenues into personalized medicine approaches, and increase the safety of dengue vaccine trials.

In addition to the accumulation and subsequent waning of human immunity to DENV in a population, viral fitness in the mosquito is a critical determinant of strain and serotype specific transmission potential [61, 63, 64, 70–73]. For example, a miDHIM could be vital when assigning appropriate attribution to critical sources of variability within the transmission system. Recently there has been a push to account for heterogeneity of infectiousness in vector-borne disease and a study in Vietnam demonstrated the variability of DENV acquisition by mosquitoes during dynamic viremias [74, 75]. It is important to note that the study of DENV pathogenicity in humans should not be the primary focus of any experimental design but with careful planning could be a valuable addition to studies aimed at assessing transmission interruption mechanisms or therapeutic performances.

Although appreciation for the milieu of salivary proteins naturally inoculated simultaneously with DENV is necessary when considering the innate immune response to infection, it is important to recognize that in some cases, miDHIM may not be more or less robust than viral inoculation alone. For example, if the scientific question involves therapeutic performance during the viremic phase, then a traditional human infection model of needle-inoculation may be sufficient. Similarly, if the antibody response to DENV alone is of interest, then the delivery method may not be as important. These scenarios assume that the pathological activity of mosquito saliva is involved not in disease progression (downstream viremia intensity, disease presentation), but more in the success of infection establishment, though as mentioned above, there is still some question as to how mosquito saliva may affect DENV infection progression. However, there is arguably much to be gained from the judicious use of mosquito improvements in DENV challenge studies.

Although existing experimental data offer many valuable insights into DENV infection and transmission, some further preliminary experiments are needed to reduce the uncertainty in this system, both with regards to the implementation of miDHIM strategies and the human infection model generally. Specifically, the following 3 areas of development appear critical: (1) to determine the role of inoculation dose in the DENV human infection model and the magnitude of precision needed to achieve consistent results, (2) to achieve standardization or cross-interpretability of the methods used to determine DENV dissemination within the mosquito that predict transmission capability, and (3) to provide a rigorous decision process to account for miDHIM approach-specific performance under different investigational scenarios for each of the subtypes of miDHIM (eg, live mosquito exposure, coinoculation with saliva/SGE, coinoculation with salivary proteins).

Conclusion

In summary, development and use of miDHIMs has the potential to lead to fundamental discoveries regarding DENV transmission and infection dynamics, and has the potential to accelerate the pace of therapeutic innovations. Specifically, as it more closely capitulates the natural transmission experience, miDHIM may allow for more thorough scientific investigations leading to opportunity to evaluate therapeutics early in the development and formulation pipeline, and allow for more thorough scientific investigations of DENV fitness, the immune context of infection establishment and persistence, and the potential for mosquito-based pharmaceutical advances.