A Brief History of the Discovery of RNA-Mediated Antiviral Immune Defenses in Vector Mosquitos

SUMMARY

Arthropod-borne viruses (arboviruses) persist in a natural cycle that includes infections of humans or other vertebrates and transmission between vertebrates by infected arthropods, most commonly mosquitos. Arboviruses can cause serious, sometimes fatal diseases in humans and other vertebrates but cause little pathology in their mosquito vectors. Knowledge of the interactions between mosquito vectors and the arboviruses that they transmit is an important facet of developing schemes to control transmission. Mosquito innate immune responses to virus infection modulate virus replication in the vector, and understanding the components and mechanisms of the immune response could lead to improved methods for interrupting the transmission cycle. The most important aspect of mosquito antiviral defense is the exogenous small interfering RNA (exo-siRNA) pathway, one arm of the RNA interference (RNAi) silencing response. Our research as well as that of many other groups over the past 25 years to define this pathway are reviewed here. A more recently recognized but less well-understood RNA-mediated mosquito defense against arbovirus infections, the PIWI-interacting RNA (piRNA) pathway, is also described.

INTRODUCTION

This review is a personal overview of gains in understanding by myself and colleagues as we focused on research on the RNA-mediated innate immune responses of mosquitos that are vectors of arthropod-borne viruses (arboviruses). The research that I accomplished during the last 25 years was dependent on working with many talented colleagues; this review also includes the simultaneous explorations of other researchers expanding on our findings. Our work began with attempts to understand how mosquitos defend against arbovirus infection with the goal of devising better schemes to block arbovirus transmission (1). Our research has centered on Aedes species mosquitos, which are vectors of the most globally prevalent and highly pathogenic arboviruses, including dengue, yellow fever, Zika, and chikungunya viruses.

A more comprehensive overview of dipteran insect immunity to virus infections was recently reported by Rosendo Machado et al. (2).

Arboviruses are transmitted between vertebrates, including humans, by arthropod vectors, the most common being mosquitos. Mosquito vectors encounter arboviruses when they are ingested in a blood meal drawn by biting a viremic vertebrate host. Viruses first infect the insects’ midgut epithelial cells, where they are amplified by replication, and then escape to disseminate and infect additional tissues, including salivary glands, via the hemolymph (3). When a female mosquito with a sufficient virus titer in her saliva probes a new vertebrate host for blood, the virus is transmitted, and the transmission cycle (extrinsic incubation period) is completed.

In addition to tissue barriers encountered by an arbovirus as it moves through the mosquito host, innate immune pathways detect and interfere with replication (3). Whereas arboviruses are capable of causing serious diseases and even mortality in vertebrates, sometimes exacerbated by the vertebrate hosts’ immune responses (4), pathology and fitness losses in infected mosquitos are generally negligible (5). This anomaly has been attributed to the differences in the immune responses to arbovirus infections in the two host taxa (6).

The exogenous small interfering RNA (exo-siRNA) pathway is the best-understood and most potent RNA-mediated mosquito antiviral immune response (7) and is my major focus. The PIWI-interacting RNA (piRNA) pathway has been more recently recognized as a potential antiviral response and is incompletely understood in terms of the major protein components involved and the sources and mechanisms of generation of viral genome-derived piRNA (vpiRNA) (2, 8).

BEGINNINGS OF UNDERSTANDING RNA INTERFERENCE

Much of our earliest understanding of mosquito innate immune responses to arbovirus infections was based on research in Drosophila melanogaster. Jules Hoffman was a pioneer of Drosophila innate immunity research and was a corecipient of the Nobel Prize in Physiology or Medicine in 2011 for this work. Hoffmann’s work focused on antibacterial and antifungal innate immune responses and revealed the importance of Toll and immune deficiency (Imd) proteins as “recognition receptors” that set off signaling cascades that led to the activation of transcription factors of the NF-κB family for antimicrobial peptide genes (9). Recognizing the absence in their data of any defense response to virus infection, Hoffmann and colleagues discovered that infection by certain viruses triggered a different signaling pathway involving the Jak kinase Hopscotch and the induction of the transcription factor STAT (10). However, the Jak-STAT response alone was not sufficient to block virus replication, and the Hoffmann group opened another excellent chapter in insect innate immune system research when they began exploring RNA interference (RNAi) in Drosophila (2, 10, 11, 12).

Our research group arrived at the study of RNAi by a different route. In searching for mechanisms that could inhibit arbovirus replication in mosquitos, we emulated the studies of plant virologists who had shown that the expression in plant cells of an untranslatable long RNA sequence derived from an RNA virus genome could make the cells resistant to the homologous virus. This phenomenon was called “pathogen-derived resistance” or “posttranscriptional gene silencing” (13, 14). This resistance was highly nucleotide sequence specific.

DISCOVERY AND CHARACTERIZATION OF RNAi IN MOSQUITOS

We found that the insertion of a nontranslatable, complementary (antisense) portion of the dengue virus type 2 (DENV-2) premembrane (prM) protein-coding sequence into the genome of a recombinant double-subgenomic Sindbis virus and then infecting cultured C6/36 (Aedes albopictus) mosquito cells or Aedes aegypti mosquitoes with the recombinant virus rendered the cells and vectors completely resistant to DENV-2 (15, 16). We attributed this resistance to antisense RNA binding to and inhibiting the replication and expression of the DENV genome. However, Fire et al. (17) discovered that the injection of double-stranded RNA (dsRNA) derived from one of the nematode’s genes into Caenorhabditis elegans resulted in the silencing of the cognate gene. This finding was followed promptly by a publication showing that injecting dsRNA into Drosophila melanogaster embryos also resulted in gene silencing and that dsRNA was more effective than single-stranded RNA (ssRNA) with either positive- or negative-sense polarity (18). Since both drosophila and mosquitos are members of the order Diptera, this demonstration had important implications for our work. We reasoned that the recombinant Sindbis virus that we had used to express the DENV-2 genome segment would have generated DENV-specific dsRNA, probably as part of a replicative intermediate, in the infected mosquito cells and that this was the likely source of the silencing of DENV replication.

We tested this hypothesis by transforming C6/36 cells with a plasmid that expressed an inverted-repeat RNA sequence derived from the prM gene of the DENV genome that was capable of forming dsRNA when transcribed (19). Control cells were transformed with either empty plasmids or plasmids expressing ssRNA with either genome-sense or antisense polarity. A large proportion of the cloned cells that expressed dsRNA were highly resistant to DENV infection, whereas the expression of neither genome-sense nor antisense ssRNA effectively blocked DENV replication. Research with plant virus pathogen-derived resistance had found that small, negative-sense RNAs (~25 nucleotides [nt]) derived from the silencing sequence occurred in the plant cell cytoplasm (20). In a stably transformed DENV-resistant C6/36 cell line, we also showed by Northern blotting the existence of small (20- to 30-nt) RNAs with both positive-sense and negative-sense polarity derived from the dsRNA expressed by the transforming plasmid (19). Our subsequent research showed that the exo-siRNA pathway in C6/36 cells was defective (21, 22), so we now assume that the virus-specific small RNAs generated in C6/36 cells were probably piRNAs, and this is discussed in a later section.

GENES AND GENE PRODUCTS INVOLVED IN MOSQUITO ANTIVIRAL RNAi

The discovery of dsRNA-mediated gene silencing in the fly Drosophila melanogaster coupled with the publication of its complete, 120-Mb genome sequence (23) rendered it an ideal model for the study of RNAi in mosquitos. Three proteins central to drosophila RNAi were rapidly identified and characterized: Dicer 2 (Dcr2) (24), an RNase III family nuclease that recognizes and cleaves long dsRNA into 21-nt duplexes; R2D2 (25), which binds cleaved dsRNA and assists in its intracellular movement; and Argonaute 2 (Ago2) (26, 27), an endonuclease that is a key component of the RNA-induced silencing complex (RISC). Ago2 binds the Dicer-produced duplexes, degrades the “passenger strand,” and uses the remaining “guide strand” for the recognition of the target mRNA or its complement (in antiviral RNAi, this is the viral genome, subgenomic messenger, or antigenome template for replication), resulting in its cleavage. However, the much larger and less genetically manipulable genomes of arbovirus vectors, in addition to the lack of genome sequence data, made the identification of orthologs of these important proteins in mosquitos difficult.

The publication of the complete 278-Mb genome sequence of Anopheles gambiae in 2002 (28) presented us with our first opportunity to examine the role and mechanisms of antiarboviral RNAi in mosquitos. Anopheles spp. are primarily known as vectors of malarial parasites, but it had been shown that Anopheles spp. could serve as vectors for the alphavirus O’nyong-nyong virus (ONNV) (29). We therefore inserted a green fluorescent protein marker into the genome of ONNV and used the recombinant virus to infect a laboratory colony of An. gambiae mosquitos by intrathoracic (i.t.) injection (30). We used RNAi (i.t. injection at the time of infection of dsRNA derived from a particular mosquito or virus gene) to knock down gene expression. We found that the injection of dsRNA from a nonstructural gene of ONNV significantly depressed virus titers in mosquitos, demonstrating that anopheles mosquitos have an antiviral RNAi response. We also found that the injection of dsRNA from the An. gambiae Ago2 gene caused a significant increase in the virus titer, demonstrating a role for this enzyme in RNAi. Also, surprisingly, we found that the injection of dsRNA from the An. gambiae Ago3 gene also caused an increase in the virus titer (30).

After the publication of the complete 1,376-Mb draft genome sequence of Aedes aegypti in 2007 (31), we were able to identify and study the components and mechanisms of this major arbovirus vector’s RNAi system. We began by using a dsRNA-specific antibody to show that Ae. aegypti Aag2 cells infected with DENV-2 produced long dsRNA and Northern blotting to show that positive-sense genomic-size (10.7-kb) RNA and both positive- and negative-sense 22- to 24-nt DENV RNA fragments were detectable in the cells, demonstrating that the trigger and effectors of the small interfering RNA (siRNA) pathway were present (32). We then used the siRNA pathway to transiently knock down the expression of Ae. aegypti mosquito genes orthologous to drosophila genes known to be central to the antiviral response by intrathoracically injecting dsRNA cognate to the Dcr2, R2d2, and Ago2 genes 2 days before oral infection with DENV-2. The knockdown of the expression of each of these three genes resulted in increases in DENV-2 RNA synthesis and infectious titers, with Dcr2 knockdown having the most significant effect. However, despite the inhibitory effect of the natural Ae. aegypti RNAi pathway on DENV-2 replication, we noted that virus replication and the release of infectious virus persisted for the female mosquitos’ lifetime, suggesting the ability of arboviruses to partially circumvent the siRNA pathway. This circumvention might be due to several factors. Since the persistence of arboviruses in nature requires their transmission to vertebrate hosts after infection and amplification in the arthropod vector, viruses have likely coevolved with the vector to temper replication dynamics to avoid killing the mosquito host (33). In addition, it has been shown in mosquito cells as well as mammalian cells infected with DENV that replication complexes containing the partially double-stranded virus RNA replicative intermediate as well as newly replicated RNA genomes are enclosed in double-membrane vesicles, thus protecting the RNAi trigger molecule from detection by the recognition receptor Dcr2 (34).

EXAMPLE OF exo-siRNA SUPPRESSION BY THE FLAVIVIRUS GENUS

Pijlman et al. (35) made an important discovery that led to the identification of a more specific mechanism of flavivirus evasion of RNAi by demonstrating that all pathogenic flaviviruses accumulated a small (300- to 500-nt) genome fragment in the cytoplasm during replication in both mammalian and arthropod cells, which they designated subgenomic flavivirus RNA (sfRNA). They determined that the sfRNA was generated from the highly structured 3′ untranslated region (UTR) of the virus genome as a result of the incomplete degradation of the virus RNA by the cellular 5′ exoribonuclease Xrn1. Pijlman and colleagues further showed that sfRNA serves as an RNAi suppressor in both insect and mammalian cells (36). Moon et al. (37) showed that in Culex quinquefasciatus mosquitos infected with the flavivirus West Nile virus, sfRNA was responsible for the suppression of RNAi. They demonstrated that both Dcr2 and Ago2 proteins coimmunoprecipitated with sfRNA, suggesting a model in which the binding of these exo-siRNA proteins to the dsRNA portions of sfRNA provided a mechanism for the suppression of mosquito RNA-mediated immune responses.

GENETIC MANIPULATION OF MOSQUITOS TO ENHANCE THE RNA-MEDIATED IMMUNE RESPONSE

To take another step toward our long-held goal of devising means to interrupt the DENV transmission cycle, we genetically engineered Ae. aegypti mosquitos by transformation by a nonautonomous Mariner transposon that expressed an inverted-repeat RNA sequence derived from the DENV-2 prM gene (38). The expression of this gene, designed to form an ~600-bp dsRNA, was regulated by the Ae. aegypti carboxypeptidase A promoter, which is induced in the midgut immediately after the mosquito ingests a blood meal. After the mosquitos fed on a DENV-2 infectious blood meal, the titration of infectious virus in transgenic mosquitos showed that an almost complete lack of virus replication had occurred by 7, 10, and 14 days after the blood meal, while >2,500 PFU of DENV-2 were measured per control mosquito. Various control experiments showed that the inhibition of virus replication was due to DENV RNA-derived siRNA expression and was sequence specific. We attributed the complete RNAi-mediated inhibition of DENV-2 replication in the transgenic mosquitos, compared to the control mosquitos, to the early and robust production, via the endo-siRNA pathway, of silencing siRNAs when the virus is in the process of being uncoated and before replication has begun and when only a few target virus genomes are present in the mosquito midgut. A subsequent corroborating publication (39) pointed out that the exo-siRNA pathway, triggered by dsRNA from the viral genome, was inactive in the Ae. aegypti midgut due to the lack of expression of the dsRNA-binding Loqs2 protein in this tissue but that the endo-siRNA pathway could be triggered by dsRNA expressed from a transgene.

WHAT HAPPENS IN MOSQUITO CELLS WITH A DEFECTIVE exo-siRNA RESPONSE?

As newer technologies became available, we used deep sequencing to examine the small RNAs produced by DENV-2-infected Ae. aegypti mosquitos, cultured Ae. aegypti Aag2 cells, and Ae. albopictus C6/36 cells (21). We were especially interested in comparisons with C6/36 cells since they were derived from a cell clone specifically selected for its more robust replication of dengue and chikungunya viruses (40), and they are frequently used for experimental virus production for this reason. We hypothesized that C6/36 cells had a defective RNAi response, which allowed arboviruses to replicate more abundantly. We found that both infected Ae. aegypti mosquitos and cultured Aag2 cells produced small DENV-specific RNAs of a uniform 21-nt length with approximately equal proportions of positive- and negative-sense polarity and with a uniform distribution over the entire virus genome. In contrast, infected Ae. albopictus C6/36 cells produced DENV-specific small RNAs that were about 27 nt long and predominantly positive stranded, with a large proportion coming from only small regions of the virus genome. In vitro assays determined that the C6/36 cells had defective Dcr2 activity (21, 22), and although the Ae. albopictus genome sequence was not available at the time, partial sequencing of the C6/36 Dcr2 cDNA suggested a single nucleotide deletion creating a nonsense mutation. We concluded that the defective Dcr2 activity and the consequent lack of an siRNA response were responsible for the higher arbovirus replication titers in C6/36 cells, and importantly, the characteristics of the aberrant small RNAs produced in these cells suggested that they were products of the PIWI-interacting RNA (piRNA) pathway (21). After further analysis, we concluded that virus genome-derived piRNAs (vpiRNAs) also occurred in DENV-2-infected Ae. aegypti mosquitos and cultured Aag2 cells, but they were present in low proportions relative to vsiRNAs (21). Hess et al. (41) also showed by Northern blotting that virus genome-specific 24- to 30-nt RNAs, in addition to 20-to 24-nt RNAs, were produced in DENV-infected Ae. aegypti mosquitos, and vpiRNAs were also produced in non-germ line cells of Ae. albopictus mosquitos infected with the alphavirus chikungunya virus (42), suggesting that the mosquito piRNA pathway was routinely involved in arbovirus infections in Aedes mosquitos.

Thus, extensive research has defined the exo-siRNA antiviral response in vector mosquitos, and more recent studies have focused on the possible role of the piRNA pathway in regulating arboviral infections.

THE piRNA PATHWAY, ITS PROTEIN COMPONENTS, AND ITS POSSIBLE ROLE IN MOSQUITO ANTIVIRAL DEFENSE

In contrast to the clear role of the drosophila exo-siRNA pathway in antiviral defense, the fly piRNA system appears to have little role in defending against virus infections (43). Drosophila encodes three PIWI family proteins, P-element-induced wimpy testis (Piwi), Aubergine (Aub), and Argonaute 3 (Ago3), which are members of an Argonaute protein subfamily and are expressed only in germ line tissues (44). Their major role is to suppress the mobility of transposable elements by silencing their transcripts. This involves the posttranscriptional production of pre-piRNAs from transcripts of “piRNA clusters” by either a piRNA-bound PIWI protein or the endonuclease Zucchini (Zuc), the loading of the pre-piRNAs onto Piwi and Aub, maturation by 3′-end modification, and the cleavage of transposon transcripts complementary to Aub-bound piRNAs. These cleavage products are loaded into Ago3, matured by Zuc cleavage and 3′ methylation, and used as guides for Ago3 cleavage of Aub-bound pre-piRNAs in what is known as the “ping-pong loop” (44). The primary and secondary, or initiator and responder, piRNAs have identifiable sequence characteristics due to the mechanisms by which they are formed (45).

Campbell et al. (46) compared the genomic structures of the Argonaute family and subfamily genes of Drosophila, An. gambiae, Ae. aegypti, and Culex pipiens genomes and found that in the arbovirus vector mosquitos Ae. aegypti and C. pipiens, Argonaute and PIWI family genes have undergone major expansions to encompass 7 proteins in each genus. Some of these have been found to be expressed in not only germ line but also somatic tissues (47), suggesting additional functions beyond transposon control.

As noted above, vpiRNAs are found in vector mosquitos and cultured mosquito cells after arbovirus infections (21, 42). In Ae. aegypti Aag2 cells, infection with the alphavirus Sindbis virus resulted in the appearance of vpiRNAs with characteristics suggesting production via the ping-pong loop (48). The knockdown of the expression of Piwi5 and Ago3 in infected cells resulted in a significant reduction of vpiRNAs, whereas the knockdown of Piwi4 and Piwi6 expression had little effect on vpiRNAs. Unexpectedly, although Piwi5 and Ago3 have been shown to be necessary for the production of vpiRNAs in acute infection of mosquitos and mosquito cells by several arbovirus families, interfering with the expression of these PIWI family proteins has been shown to have little effect on virus replication (49). Joosten et al. (49) demonstrated that a Tudor family protein, which they designated Veneno, plays a major role in assembling a cytoplasmic protein complex, including Ago3, that is responsible for processing virus RNA via the ping-pong loop to produce vpiRNAs.

Transposon-derived piRNAs are produced from nuclear transcripts of piRNA clusters. These clusters contain repetitive sequences of transposon fragments, and in the case of mosquitos, they also contain fragments of viral genome copy (c)DNA. Nag et al. (50, 51) showed that DNA containing fragments with flaviviral RNA genome sequences was formed following arbovirus infection of mosquito cell cultures and mosquitos. Alphavirus and bunyavirus genome-derived cDNAs have also been found in Aedes mosquitos (52) and Culex species mosquitos (53). It was presumed that this viral cDNA or “vDNA” was formed by endogenous reverse transcriptase expressed from retrotransposons (54), but whether this vDNA was endosomal or integrated into the mosquito genome was not determined, and whether it could serve as a template for transcripts to be processed into vpiRNAs is unknown.

ENDOGENOUS VIRAL ELEMENTS IN MOSQUITO GENOMES

In Drosophila, the majority of piRNAs are derived from transposons, and, of course, their major function is to silence transposon expression (44). Arensburger et al. (55) found that although Ae. aegypti has a much larger genome (1,380 Mb, compared to 144 Mb for Drosophila) and a much larger transposon load (47% of the genome, compared to 15.8% in Drosophila), only 19% of mosquito endogenous piRNA sequences mapped to transposons. The remaining Ae. aegypti piRNAs mapped to mosquito gene sequences or, surprisingly at the time, various virus gene sequences.

Several other investigators (56–58) have explored the integrated arbovirus or insect-specific virus (ISV) genome sequence fragments in Ae. aegypti and Ae. albopictus, known as endogenous viral elements (EVEs). The proportion of these Aedes species genomes occupied by EVEs is far higher than those found in other vector mosquitos such as Culex quinquefasciatus and Anopheles gambiae (56), leading to speculation that the integrated virus sequences are involved in the biogenesis of vpiRNAs following arbovirus infection. However, a closer analysis of the actual sequences shows that most of the EVEs are homologous to ISVs and only distantly related to circulating arboviruses (59).

Recent studies have pursued the idea that vpiRNAs derived from EVEs are components of an adaptive immunity system for mosquitos. Crava et al. (59) compared 80 genome sequences from wild-caught Ae. aegypti mosquitos from five widespread geographic regions to the current reference sequence, AaegL5, using population genomics and evolutionary approaches. In the wild-caught mosquitos, they found five EVEs derived from two currently circulating ISVs, cell-fusing agent virus (CFAV) and Aedes anphevirus (AeAV), most of which were newly integrated into piRNA clusters near retrotransposons. They concluded that they had observed a dynamic landscape of new virus integrations into vector mosquito genomes, resulting in EVEs that could generate vpiRNAs to mediate adaptive antiviral immunity.

Suzuki et al. (60) discovered a newly integrated Ae. aegypti EVE derived from a currently circulating CFAV in Thailand. They first showed that the CFAV EVE generated vpiRNAs that targeted infecting CFAV, producing ping-pong-loop vpiRNAs with secondary piRNAs derived from the virus genome. Using CRISPR-Cas9 genome editing, they removed the CFAV EVE and then infected the genetically altered mosquitos with CFAV. They noted increased replication of CFAV in the mosquito ovaries, suggesting that immune memory in the form of EVE-generated vpiRNAs protected mosquitos from homologous virus infections.

Research to date suggests that the role of the mosquito piRNA response is the protection of the germ line tissues from arboviral and, probably more importantly, ISV infections. Although the mosquito vpiRNA pathway is clearly active in arbovirus acutely infected somatic tissues, whether or not it plays an important role in RNA-mediated defenses is not clear and is the subject of active ongoing research.

CONCLUDING REMARKS

The vector mosquito exo-siRNA pathway, its components and mechanisms, and its role in modulating acute virus infections are well defined. It is frequently used as a tool for studying the molecular biology of mosquitos and, along with the endo-siRNA pathway, has promise for developing strategies to interrupt arbovirus transmission cycles.

The vector mosquito piRNA pathway and its potential role in RNA-mediated antiviral defense are less well understood. The vpiRNA pathway appears to have a diminished role in antiviral defense in comparison to exo-siRNA. Research to date suggests that the role of the vpiRNA pathway in innate immunity to acute infections is minor but that it may constitute an adaptive immune response, particularly if EVEs are integrated into the genomes of germ line cells. Current and future research on RNA-mediated antiviral immune responses in vector mosquitoes will aim to further define the roles of each member of the PIWI family and other proteins potentially involved in vpiRNA generation by examination of their structure, intracellular localization, and expression during virus infection (61–63); the sources/templates for vpiRNAs; and the precise mechanisms for vpiRNA production.