Exploring the Mosquito–Arbovirus Network: A Survey of Vector Competence Experiments

ABSTRACT. Arboviruses receive heightened research attention during major outbreaks or when they cause unusual or severe clinical disease, but they are otherwise undercharacterized. Global change is also accelerating the emergence and spread of arboviral diseases, leading to time-sensitive questions about potential interactions between viruses and novel vectors. Vector competence experiments help determine the susceptibility of certain arthropods to a given arbovirus, but these experiments are often conducted in real time during outbreaks, rather than with preparedness in mind. We conducted a systematic review of reported mosquito–arbovirus competence experiments, screening 570 abstracts to arrive at 265 studies testing in vivo arboviral competence. We found that more than 90% of potential mosquito–virus combinations are untested in experimental settings and that entire regions and their corresponding vectors and viruses are undersampled. These knowledge gaps stymie outbreak response and limit attempts to both build and validate predictive models of the vector–virus network.


INTRODUCTION
Arthropod-borne (arbo-) viruses face evolutionary pressures that favor generalism in the range of both vertebrate hosts and arthropod vectors that they can use (reviewed in detail in Ciota and Kramer 1 ). That flexibility can pose a particular problem for public health, as it both enables their spread into new locations and ecosystems and adds a layer of unpredictability to their dynamics upon arrival. Experimental studies simplify real-world complexities of transmission and can be used to test not only the basic compatibility of a given virus and arthropod vector species but also vector competence-the relative ability of arthropod vectors to be infected by a virus and then disseminate and transmit it to a susceptible host. 2 Despite arboviruses' evolutionary tendencies toward broad host and vector range, 1,3,4 there are complex genetic underpinnings that govern vector competence, 5 which can manifest as variation in competence between closely related species of the vector 6 or even among populations of the same species. 7 Vector competence experiments are often conducted in response to the emergence of novel pathogens or the emergence of a known pathogen in a new location with previously untested vectors. The distributions of both mosquito vectors and the viruses they transmit are increasingly in a state of disequilibrium as a result of climate change, global travel and trade, and biotic homogenization. 8,9 Operating in a responsive paradigm, medical entomology is increasingly struggling to keep pace with these shifts, as resources are often abruptly diverted to new study systems to answer questions that support outbreak response and influxes of funding to support such studies are typically reactionary to emergence events. 10 The patchwork of experimental research efforts to date represents the cumulative history of these moments rather than a systematic exploration of the mosquito-arbovirus network, limiting outbreak preparedness and particularly complicating efforts to predict unrealized links in that network using machine learning. 11,12 As there are no standardized repositories that register vector competence experiments or immortalize their findings, it is currently difficult to evaluate the distribution of research efforts so far and identify important gaps that may be relevant to future outbreaks.
Here, we conducted a systematic review of the mosquitoarbovirus literature using keywords associated with vector competence and screened studies in the Web of Science to identify these studies. Our objective was to determine the taxonomic and geographic patterns in these studies and to identify historical trends driving research in this subfield.

MATERIALS AND METHODS
A systematic literature search was conducted on the Web of Science to identify suitable records that described vector competence experiments with mosquito-borne viruses. Our search used the following terms: "("vector competence" OR "extrinsic incubation period" OR "vectorial capacity" OR "dissemination") AND (arbovirus OR virus) AND (experiment* OR trial OR captive* OR laboratory) AND mosquito." Our search in the Web of Science Core Collection via University of Edinburgh institutional access returned 570 records in February 2021.
We performed an initial screen of records based on abstracts, excluding reviews, methodology descriptions, and studies with no experimental infection (N 5 135), studies with non-virus infections, insect-specific infections, infection regimes with confounding treatments (i.e., coinfection with Wolbachia or insect-specific viruses; N 5 71), studies involving infection in non-mosquitoes, experimental vector infection without any reported results describing competence quantitatively or ex vivo data (N 5 55), and studies fitting multiple exclusion criteria (N 5 22). For the remaining 287 records, we undertook a second round of screening to determine suitability (i.e., did the study include some sort of experimental test of mosquito competence for arbovirus transmission, such as infectious bloodmeal, intrathoracic injection, or feeding on viremic animals). After excluding one study for which we could not access the full text, 13 we were left with 265 studies that were within the scope (see Supplemental Table 1). For each of these, we recorded the species pairs of mosquitoes and arthropod-borne viruses that were experimentally tested together, regardless of how competence was measured (including both measurements of virus [infectious virus versus viral RNA] and mosquito tissues analyzed [e.g., body, legs, head]) or whether the vector was found to be competent or not; the country from which wild mosquitoes were originally collected (even if populations were maintained in laboratory settings long-term); and when available, mosquito subspecies and (as applicable) dengue virus serotype. Using the taxize R package, 14,15 we updated mosquito binomial names if a more recent valid name could be found in the National Center for Biotechnology Information taxonomic database; in all other cases, we used taxonomy reported verbatim from source studies, including nonstandard naming conventions (e.g., "Culex sp.," "Culex declarator/mollis").
All analyses were conducted and figures generated in R software version 4.0.3.

RESULTS
Taxonomic coverage. We found a total of 298 pairs of viruses (N 5 35) and mosquito species (N 5 122) that have been tested experimentally, leaving the majority of all possible pairings of these specific viruses and mosquitoes untested (93%; 3,972 of 4,270 possible pairs; Figure 1). Among these, some viruses were entirely untested in Aedes or Culex vectors (seven and six, respectively), and most were untested in Anopheles (27 viruses) ( Figure 2). Although some of these pairings may not be relevant for experimental vector competence testing (e.g., viruses and vectors that do not overlap geographically), there are many untested pairings of potential public health importance, such as Everglades virus in Aedes aegypti, that warrant further investigation given ongoing and projected range expansions and changing climatic suitability for transmission. 8,9 However, for several pairings, our search parameters yielded no results in the Web of Science, despite the existence of relevant publications, such as Murray Valley encephalitis virus and yellow fever virus in Culex spp. [16][17][18][19][20] or Ross River virus (RRV) and Sindbis virus in Ae. aegypti. 21,22 Even within tested mosquito-virus combinations, effort is distributed unevenly. Geographic coverage. Lineage variation in vector competence can be quite striking, even within the geographic range of a single globalized species such as Ae. aegypti or Ae. albopictus. 34 To explore how arboviral research captures this dimension of natural variation, we recorded where each study's mosquitoes were sourced from. The majority of studies used mosquitoes sourced from the United States (Figure 4), followed distantly by Brazil and Australia. The range of pathogens tested using mosquitoes from a given country is slightly more even, but still the majority of work has focused on the United States, Brazil, Australia, China, India, and western Europe ( Figure 5). By both measures, Africa and eastern Europe have been severely understudied, particularly compared with the Americas, where multiple explosive multinational epidemics have forced researchers to answer questions about broad geographic risk.

DISCUSSION
Our results suggest that research efforts in this field have largely been driven by the shifting priorities of arboviral outbreak response. As such, there are numerous gaps in viral taxonomy, vector taxonomy, and vector geography that limit the utility of available data for future outbreak response. Moreover, the vast majority of mosquito-virus pairs (. 90%) are simply untested within our sample of studies, indicating that network-level understanding of arbovirus ecology is incomplete. In particular, existing descriptions of the mosquito-virus network (e.g., Figure 1 in Evans et al. 11 and Figure 2 in Yee et al. 35 ) are challenging to interpret without explicit understanding that the network's architecture is mostly representative of experimental history ( Figure 6). In turn, this severely constrains inference about-and prediction of-the true underlying network. 11,35,36 Some gaps may particularly limit ecological inference; for example, anthropophilic vectors such as Ae. aegypti are far better studied than bridge and sylvatic vectors such as Aedes africanus, Sabethes spp., or Haemagogus spp. (especially because some of these species (e.g., Haemagogus leucocelaenus) have never been colonized, making vector competence experiments especially challenging and reliant on fieldcollected eggs). These species may be less important during epidemics, but they determine the risk of an enzootic virus  BFV   CHIKV  EEEV  EVEV  GETV  KFV  MAYV  RRV  SINV  VEEV  WEEV  CVV  CHITV  INGV  LACV  RVFV  SBV  SHUV  UMBV  BAGV  BYDV  DTV  DENV  JEV  MVEV  RABV  ROCV  SLEV  USUV  WNV  YFV  ZIKV  BEFV  CHPV 35,37 Efforts to fill these gaps could take several approaches. First, researchers can use network science to identify important vector-virus pairs and conduct vector competence experiments that would fill key knowledge gaps (especially those of likely public health importance). Just as machine learning can be used to guide species sampling for viral discovery in nature, 38 predictive models can be used to develop or augment shortlists of the vectors that will be most relevant during emergency scenarios that have previously been flagged (e.g., yellow fever establishment in the Asia-Pacific region 39 ). The benefits of model-experiment feedback are iterative: at present, many models will struggle with the sparsity of the BFV  CHIKV  EEEV  EVEV  GETV  KFV  MAYV  RRV  SINV  VEEV  WEEV  CVV  CHITV  INGV  LACV  RVFV  SBV  SHUV  UMBV  BAGV  BYDV  DTV  DENV  JEV  MVEV  RABV  ROCV  SLEV  USUV  WNV  YFV  ZIKV  BEFV  CHPV   vector-virus matrix, but targeting important sources of uncertainty will lead to better predictions.
Second, researchers can work toward a more cohesive geographic picture of how risk varies, both for cosmopolitan vectors such as Ae. aegypti and Ae. albopictus and for vectors that are locally abundant or are known to feed on amplifying hosts. 40 The geographic biases we describe here reflect where a relatively small number of institutions are able to continuously fund and maintain vector colonies. This process is itself often extractive and inefficient in nature (i.e., mosquito colonies established with species from countries facing public health emergencies are often used by researchers in the United States and Europe to generate high-impact publications). This dynamic can redirect funding from investment in capacity building in the regions where vector-borne disease burdens are highest, perpetuating disparities in both health and scientific research. Deeply linked to capacity and sustainability of institutions housing laboratory colonies, a lack of training in medical entomology has been recently noted as a global issue in the response to outbreaks of vector-borne diseases. [41][42][43] Investment in these kinds of capacity building in undersupported areas such as Africa, the Middle East, and eastern Europe would be a significant contribution to global preparedness for vector-borne disease outbreaks.
Finally, our study highlights that a substantial breadth and depth of vector competence data are published every year  in the peer-reviewed literature, but currently, the results of these experiments have no standardized home. Other recent studies highlight that synthesis of these data is possible, despite the complexity of metadata required to describe variation in experimental protocols 6 ; however, our study highlights the challenges of recovering "findable" data from the vector competence literature. Limitations we encountered included 1) Web of Science is not a comprehensive record of this research area, and older studies in particular appear to be missing; 2) keywords we used may not have captured all of the relevant studies because of variable terminology; 3) a handful of non-English-language publications-in particular, Spanish and French language publications from Latin America and Africa, respectively-were not captured by our search terms; and 4) not all studies reported reusable experimental results and metadata. Developing a standardized database of vector competence experiments with direct user contributions-ideally supported by a universal set of minimum data and metadata standards 44 -would help translate the inconsistent and patchy funding in this field into more immortalized data. In doing so, such a database would help researchers identify geographic and taxonomic knowledge gaps further in advance of public health emergencies and would make more data immediately available to public health agencies in a searchable format once an outbreak begins.
As we show here, data science approaches can be useful to identify trends and gaps in scientific understanding of arboviral ecology and evolution. 12 In some cases, filling these gaps will be more challenging; for example, establishing colonies is harder for some mosquito species than others, and some viruses require higher biosafety levels, limiting the number of researchers with the ability to safely work with them. Future work may also aim to expand our scope to other medically important vectors, including ticks and midges (for which we excluded a handful of studies that were identified by our search terms). Similarly, future work could examine trends in how coinfection dynamics are studied and tested. Both insect-specific viruses and Wolbachia have been considered as potential biological countermeasures to arboviral transmission in mosquitoes [45][46][47][48] ; however, the network of pathogencoinfection-vector combinations has been characterized even less systematically, and these data remain largely unsynthesized. Addressing these types of questions in the future may point to new opportunities for both empirical research and modeling that harnesses these data to predict and prevent arboviral emergence.  Node type Mosquito Virus FIGURE 6. Network visualization of arbovirus (green) and mosquito (yellow) pairs experimentally tested. Notably, any observable network of compatible species would be constrained to "fit" inside this sampling-based network, highlighting how experimental effort determines observable pairs of compatible species more than their biology. Abbreviations follow naming conventions in virology. Nodes represent virus and mosquito species, where node size is proportional to the number of studies involving each (range for mosquitoes: 1-136; range for viruses: 1-109). Edges represent a record of an experimentally tested pair, where edge weight is proportion to the number of studies returned for each pair (range: 1-47). Edges do not record species compatibility.