Expression of endogenous Anopheles gambiae microRNAs using an Anopheles gambiae densovirus (AgDNV) intronic expression system

Abstract

Background

Anopheles gambiae densovirus (AgDNV) is a highly species-specific parvovirus that reaches high titers in adult Anopheles gambiae mosquitoes with few transcriptomic effects and minimal significant fitness effects. Given these characteristics, AgDNV has been proposed as a viral vector for basic research and mosquito control. Previous work created an AgDNV co-expression system with a wild-type AgDNV helper plasmid and a transducing plasmid expressing enhanced green fluorescent protein (EGFP) that can be used to co-transfect cells to generate infectious recombinant transducing AgDNV virions. Generated virions infect the An. gambiae midgut, fat body, and ovaries, yet this viral vector system is limited in the size of transgenes that can be expressed due to capsid packaging limitations.

Methods

Considering these size constraints, we created an artificial intron within the EGFP gene of the transducing construct that can express small pieces of genetic material such as microRNAs (miRNAs), microRNA sponges, or other small sequences. Placement of this intron in EGFP created a fluorescent reporter such that incorrect splicing produces a frameshift mutation in EGFP and an early stop codon, whereas correct splicing results in normal EGFP expression and co-transcription of the intronic genetic cargo. A selection of miRNAs with predicted or demonstrated importance in mosquito immunity and reproduction with expression localized to the fat body or ovaries were chosen as intronic cargo. Construct expression and splicing was evaluated, and the impact of miRNA expression on putative miRNA targets was measured in vitro and in vivo.

Results

The created intron was correctly spliced in cells and mosquitoes; however, miRNA delivery resulted in inconsistent changes to miRNA and predicted target gene transcript levels—possibly due to organ-specific miRNA expression or inaccurate putative target predictions leading to miRNA–target gene sequence mismatch.

Conclusions

Although our results on target gene expression were inconsistent, with optimization this viral vector and developed intron have potential as an expression tool within An. gambiae mosquitoes or cell lines.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s13071-025-06994-7.

Background

Anopheles gambiae is the major vector of Plasmodium falciparum in Sub-Saharan Africa, where most malaria cases occur [1]. While current malaria control efforts rely heavily on bite prevention via bed nets and mosquito reduction through insecticides, insecticide resistance is increasing [2]. Similarly, the efficacy of antimalarials used to treat disease in humans is threatened by drug-resistant parasites. As such, there is an increasing need for novel tools to better investigate An. gambiae biology, as well as new methods for mosquito and pathogen control (such as the development of malaria-resistant mosquitoes capable of replacing susceptible populations) [3].

Although CRISPR-Cas9 editing has great promise, such experiments typically require specialized microinjection equipment, alterations to the mosquito genome, and the time-consuming establishment of mosquito lines. The introduction of genetic material without modifying the genome, such as through genetically modified microbes, offers an alternative approach that is useful for altering the expression of existing genes, introducing new genes, or, in the case of paratransgenesis, using a symbiont to express a transgene within the vector that acts against the pathogen. While the bacterium Wolbachia has been proposed for paratransgenesis, and select strains have been successful at blocking pathogens in Aedes aegypti, Wolbachia has yet to be genetically modified, and generating stable infections in An. gambiae has proven difficult [4–6]. Although there have been reports of Wolbachia infections in wild populations of An. gambiae in Africa, these findings need further verification, and the impact of these infections on P. falciparum in An. gambiae are currently unclear [5, 7–10]. Viruses also have potential for use in paratransgenesis; however, a limited number infect An. gambiae and few are ideal for genetic manipulation or for the introduction of new genes owing to off-target effects, or the danger of transmission to humans and the resultant disease [11, 12]. One of the only nonpathogenic insect-specific viruses known to infect An. gambiae was discovered in 2008 in An. gambiae Sua5B cells: the An. gambiae densovirus (AgDNV) [13]. AgDNV is closely related to other mosquito densoviruses including Culex pipiens pallens densoviruses (CppDNV) and Aedes aegypti densovirus (AaeDNV), and consists of a 4139 nucleotide (nt) ssDNA genome with terminal hairpins at each end that allow for viral genome packaging [13]. AgDNV is highly specific to An. gambiae and has poor infectivity even within closely related mosquitoes [14]. While some densoviruses in other mosquito species do cause mortality, AgDNV is not pathogenic to An. gambiae and increases in titer over the course of the adult lifespan while having little impact on An. gambiae fitness or gene expression [15–17]. These characteristics make AgDNV ideal for use as a late-life acting bioinsecticide or as a viral vector to express genes against the parasites themselves [16].

Previously a co-plasmid expression system was developed consisting of the unaltered AgDNV genome in a pBluescript cloning vector (pWTAgDNV) and a transducing pBluescript construct containing an Actin5C promoter, an EGFP reporter sequence, and a SV40 termination sequence (pAcEGFP) [18]. From here on, “p” designates the plasmid form of the construct, whereas “v” designates the viral form; constructs lacking a label denote the general sequence. As both plasmids in this co-expression system possess the AgDNV terminal hairpins that are crucial for genome packaging, both sequences get packaged into capsids produced by the wild-type construct. Virions produced by this co-expression system localize to mosquito tissues important for pathogen transmission and immunity such as the midgut, ovaries, and fat body when injected into adult mosquitoes [13, 18]. Although AgDNV has potential as a viral vector, the small genome and capsid size limits the length of transgenes that densoviruses can express; for example, in AaeDNV, a size increase of 8% over that of the wild-type genome resulted in a 10% reduction in packaging efficacy [17, 19]. While less deleterious, a shorter sequence can also negatively alter packaging efficacy [18]. These size constraints have led to the development of various expression strategies using AaeDNV, including the use of an artificial intron to express microRNAs or sponges [20]. These constraints and the success of various modification strategies in AaeDNV led us to modify the transducing construct of AgDNV to express microRNAs or miRNA sponges that would increase the size of AcEGFP from 3994 nt to closer to the 4139 nt size of WT AgDNV, but would not exceed the size of the WT genome [18].

MicroRNAs (miRNAs) are small, non-coding RNAs that act as post-transcriptional regulators of gene expression through the RNA interference (RNAi) pathway [21]. The RNAi pathway and miRNAs are highly conserved and, in An. gambiae, over 163 miRNAs involved in a variety of processes including mosquito reproduction, immunity, and development have been identified to date [22–28]. Endogenous miRNAs are coded in introns or intergenic sections of DNA and form short, hairpin-shaped, secondary RNA structures following transcription [29]. After processing by Drosha, pre-miRNA hairpins are exported from the nucleus to the cytoplasm where they are cut further by Dicer into ~22 nt duplexes [30]. One strand of each duplex is degraded, and the remaining, more stable strand forms the mature miRNA that is brought to target mRNA sequences by the RNA-induced silencing complex (RISC). The mature miRNA sequence binds to regions of mRNAs where there is sequence complementarity (often the 3′ UTR) and the degree of complementarity controls whether the mRNA transcript is cut and targeted for degradation or whether binding simply blocks translation [21, 31]. Endogenous An. gambiae miRNAs or in silico designed miRNA sponges that are complementary to mature miRNA sequences and “soak” up endogenous miRNAs are ideal for expression via AgDNV owing to their small size [32, 33]. Through the expression and endogenous processing of pre-miRNAs into mature miRNAs, levels of mature miRNAs can be enhanced, whereas expression of miRNA sponges will lead to depletion of endogenous mature miRNAs.

As miRNAs are often encoded in introns or intergenic regions, and we wanted to express small RNA cargo in a way that allowed for tracking of expression, we developed an artificial intron with a reporter phenotype within the EGFP gene of the transducing AgDNV construct (Fig. 1A and B). To test this intron, we identified miRNAs for expression and manipulation that are putatively involved in mosquito functions such as immunity and egg development that are tied to organs that AgDNV is known to infect [34–37]. Selected miRNAs all had predicted or observed functions within An. gambiae, yet many have not been purposefully manipulated in An. gambiae, and functional studies identifying specific mRNA transcript targets and downstream effects are lacking (Table 1) [36, 38, 39].

Fig. 1.

Intron layout and splicing scheme. A Splicing donor and acceptor sites within the EGFP gene flank the pre-miRNA region. MluI and BstBI cut sites allow for swapping of g-block sequences containing EGFP bases removed during digestion, splice sites, and intronic cargo (pre-miRNA sequences, miRNA sponge, or nonsense RNA). Arrows mark cut sites within the splice donor and acceptor sequences. B During transcription, WT AgDNV transcripts are expressed from the WT construct (pWTAgDNV), while EGFP-encoding transcripts containing the intron are expressed by the transducing construct. Intronic splicing of the pre-mRNA (or miRNA sponge) transcript from the transducing construct results in the rejoining of EGFP-encoding transcript halves and the intronic cargo being processed via the RNAi pathway. Translation of WT DNV transcripts results in capsid formation, whereas translation of EGFP mRNA results in EGFP expression if intronic splicing occurred correctly

Table 1.

Selected miRNAs and miRNA sponge along with selected target gene transcripts

Construct	Target gene transcripts and directionality	Target gene functions	Target gene binding	Study species
miR8	↓ Swim	Egg development	Demonstrated	Ae. aegypti An. gambiae
miR8SP	↑ Swim ↓ miR8	Egg development	Demonstrated	Ae. aegypti An. gambiae
miR34	MISO	Egg development	Predicted	Ae. albopictus Ae. aegypti An. gambiae An. stephensi
	Cactus Rel1	Toll immune pathway regulation
	Caspar Rel2	IMD immune pathway regulation
miR305	APL1C	TEP1 regulation and P. falciparum immunity	Predicted	An. gambiae Ae. albopictus
miR375	↑ Cactus ↓ Rel1	Toll immune pathway regulation	Demonstrated	Ae. aegypti
	Caspar Rel2	IMD immune pathway regulation	Predicted

Many miRNA targets are predicted as in vitro and in vivo functional studies are lacking. Directionality, as noted through arrows, indicates an increase or decrease in target gene transcript levels following miRNA or sponge (SP) expression. Predicted target transcripts have no directionality as they have not yet been studied. The binding column notes whether this binding has been predicted or whether it has been observed in functional studies. Study species are noted to indicate which species these miRNAs and targets have been examined or predicted in

The lack of validation of these selected miRNAs within An. gambiae proved challenging when assessing changes in miRNA or mRNA target levels, yet transcript splicing patterns indicate that the developed intron delivery system functioned as expected. This intron represents a new AgDNV viral vector expression strategy and may be useful for the expression of sequences including endogenous miRNAs, miRNA sponges, synthetic short interfering RNAs, or guide RNAs within genetically modified Cas9 mosquito lines.

Methods

Selection of miRNA targets

The first miRNA selected for this work, miR8, was highly upregulated in both the Ae. aegypti and An. gambiae fat body following blood feeding and targets the 3′ UTR region of secreted wingless-interacting molecule (Swim), a molecule involved with the Wnt/Wingless signaling pathway (Table 1) [40–43]. When miR8 was depleted in Ae. aegypti, Swim levels remained high following blood feeding and egg development was inhibited [41]. Another miRNA, miR34, showed differential expression in several different mosquito species during pathogen infection, including in An. gambiae where midgut expression was decreased following an infectious Plasmodium berghei (P. berghei) blood meal [22, 34, 44–46]. Specifically, miR34 was predicted to bind to Relish-like transcription factor 1 (Rel1) and Caspar transcripts, important factors in the Toll and IMD immune pathways, respectively [37, 46]. As Caspar is a negative regulator of the Relish-like transcription factor 2 (Rel2) and Cactus is a negative regulator of Rel1, Rel1 and Rel2 transcript levels were also assessed in target gene quantitative PCR (qPCR) reactions (Table 1). These transcripts, as well as mating induced stimulator of oogenesis (MISO), were predicted target genes of miR34 via the now defunct miRNA–mRNA binding prediction webtool Insectar (http://www.insectar.sanbi.ac.za/) [47]. Previously, knockdown of MISO transcripts using RNAi resulted in reduced egg production, indicating a potential role for miR34 in reproduction [48]. The third selected miRNA, miR305, was elevated in the ovaries and midgut of An. gambiae following blood feeding and was higher in midguts following an infectious Plasmodium-containing blood meal [36, 37]. Inhibition of miR305 decreased the midgut microbiota and increased resistance to P. falciparum, whereas enhancement of miR305 increased P. falciparum infection levels and led to higher levels of midgut microbiota [37]. This miRNA was predicted to target the 3′ UTR of APL1C and as APL1 is part of a complex that stabilizes the immune factor thioester-containing protein 1 (TEP1), which binds to the surface of Plasmodium leading to parasite destruction, miR305 may impact Plasmodium infection [37]. Supporting this, miR305 depletion in An. gambiae led to increased resistance to both P. falciparum and Plasmodium berghei infection and altered the levels of many immunity or anti-Plasmodium genes in mosquito midguts [43]. The final miRNA, miR375, was only detected in blood fed Ae. aegypti mosquitoes and was predicted to bind to the 5′ UTR of Toll pathway immune genes Cactus and Rel1 [49]. Expression of a miR375 mimic in Ae. aegypti mosquitoes or cells led to binding of the 5′ UTRs of Cactus and Rel1 and the upregulation of Cactus and downregulation of Rel1 [49]. Similar changes in target genes and an increase in Dengue virus type 2 titers were observed in Ae. albopictus Aag2 cell lines [49]. Although miR375 has not been studied in An. gambiae, this miRNA has an identical sequence to miR375 in Ae. aegypti and has also been predicted by Insectar to target An. gambiae Cactus and Rel1 along with other gene transcripts including Caspar and Rel2 [47].

Plasmid preparation and production

Sure 2 supercompetent E. coli cells (Agilent Technologies, 200152) were transformed as per kit instructions (SOC media was substituted for NZY + media) with pAcEGFP and pWTAgDNV plasmids [18]. Transformed colonies were plated on Luria broth agar plates with 100 µg/mL ampicillin and incubated at 37 °C overnight. Colony PCR was used to verify transformations and selected colonies were grown in 5 mL of Luria broth in a 37 °C shaker overnight and then preserved as glycerol stocks. Purified plasmids were produced by growing glycerol stocks in liquid culture as before, extracted using an Omega Bio-tek E.Z.N.A. Plasmid DNA Kit (D6942-02), and quantified using a NanoDrop ND-1000 spectrophotometer.

Intron design

A potential splice acceptor site in WT AgDNV was identified at position 463 of the gene encoding the viral protein using the neural-network-based NetGene2 predictive splicing server, which identifies transition sequences between introns and exons [50–52]. This sequence was converted from AG^ACGCAGACAG (with “^” indicating the predicted splicing site) into a splice donor site by replacing the intronic portion with the starting sequence of the second intron of An. gambiae RPS17 such that the new sequence was AG^GTAGGCGCGC. This sequence was further modified by two base pairs to AG^GTAAGTGCGC to match the An. gambiae U1 small nuclear RNA conserved region (Fig. 1A) [53]. This U1 sequence (GTAAGT) represents the binding site for the U1 small nuclear ribonucleoprotein which helps to form the spliceosome [53]. A splice acceptor with the sequence TACTGACATCCACTTTGCCTTTCTCTCCACAG was created to accompany this splice donor at position 464 of the gene encoding the viral protein by adding in the branch point, polypyrimidine tract, and intronic portion from the 3′ end of a chimeric human intron (last 32 nucleotides) preceding an immunoglobulin gene heavy chain variable region that is commonly found in commercial vectors such as in the pRL-CMV plasmid from Promega (Fig. 1A) [54–56]. This splice donor and splice acceptor site were initially chosen within AgDNV to attempt the creation of a nondefective recombinant AgDNV, as previously described for AaeDNV, but we later decided to use a co-transfection system with pWTAgDNV and a transducing plasmid with the artificial intron to create an EGFP reporter phenotype [20]. These developed splice donor and splice acceptor sites were placed within the EGFP gene of pAcEGFP at positions 334 and 337, respectively, to create a reporter phenotype such that improper intronic splicing or a lack of splicing would result in a stop codon within the EGFP gene and correct splicing would result in EGFP expression (Fig. 1B). Predicted splicing was examined at all steps using NetGene2 and the created splice donor and splice acceptor sequences both had a confidence scores of 1.0, indicating a high confidence in splicing [50].

miRNA and sponge selection

For intronic miRNA expression, endogenous pre-miRNA sequences were inserted into the created intron so that upon splicing, the pre-miRNA hairpin would be co-transcriptionally processed alongside EGFP transcripts [22–24]. Selected pre-miRNA sequences for An. gambiae miR8, miR34, miR305, and miR375, as well as a miRNA sponge against miR8 (miR8SP), were added to this developed intron to test the co-expression system and intronic splicing mechanism. A random nonsense RNA sequence (NS) was added to the intron as a control. These miRNAs and the miRNA sponge were chosen, as described above, on the basis of known or predicted effects on genes involved with immunity, pathogen defense, or reproduction in An. gambiae, Ae. aegypti, or relevant mosquito cell lines (Table 1). To test intron functionality and demonstrate that splicing is sequence-dependent, altered splice donor and splice acceptor site sequences were developed using site-directed mutagenesis of the pAcEGFPmiR8 plasmid [50]. When the splice donor site was changed by a single nucleotide (in bold) from AGGTAAGTGCGC to AGATAAGTGCGC, NetGene2 no longer identified this as a splice donor site. Similarly, when the splice acceptor site was changed by one nucleotide (in bold) from TACTGACATCCACTTTGCCTTTCTCTCCACAG to TACTGACATCCACTTTGCCTTTCTCTCCACAT, this site was no longer predicted to be a splice acceptor.

Cloning and intronic cargo

MluI and BstBI sites were introduced into the EGFP-encoding gene of pAcEGFP using site-directed mutagenesis to create synonymous mutations. A MluI site was created by altering position 327 of EGFP from C to G, and position 330 from C to T. A BstBI site was created in EGFP by switching position 348 from G to A. Endogenous An. gambiae pre-miRNA sequences from miRbase (https://www.mirbase.org/) were converted to DNA and used to order g-blocks from Integrated DNA Technologies (IDT) [22]. The mir8SP sequence contained ten repeated blocks of the reverse complement of mature An. gambiae miR8. Each block was separated by four spacer nucleotides and the entire sponge sequence was placed within the intron as with pre-miRNA sequences. A nonsense RNA (NS) was created using a random sequence with no matches to the An. gambiae genome or transcriptome when searched using the Basic Local Alignment Search Tool (BLAST). Each pre-miRNA, miRNA sponge, or nonsense RNA was coded on IDT g-blocks synthesized with flanking MluI and BstBI sites, EGFP segments to replace those removed during digestion, and the splice donor and splice acceptor sites (Table S1; Fig. 1A). G-blocks were subcloned into pJet using a CloneJet PCR Cloning Kit (ThermoFisher Scientific, K1231) and later digested using MluI and BstBI. These inserts were ligated into pAcEGFP that had also been digested with MluI and BstBI, and the resulting plasmid sequences were verified.

Cell culture and transfections

Sua5B and Moss55 An. gambiae cells were grown in 25 cc plug cap flasks at 28 °C and passaged once per week at a 1:5 dilution with Schneider’s Drosophila media with 10% fetal bovine serum (FBS) v/v. For transfections, cells were quantified using a hemocytometer and 6 × 10⁶ cells were added to each well of a 6-well plate along with 3 mL of complete media and incubated overnight. Cells were transfected at ~70–80% confluence with a 1:2 ratio of pWTAgDNV to transducing plasmid with 830 ng pWTAgDNV and 1660 ng transducing plasmid per well using a Lipofectamine LTX with Plus Reagent kit (ThermoFisher Scientific, 15338030). Briefly, plasmids were added to a mix of 500 µL OptiMem media with 3 µL Plus reagent and incubated at room temperature for 10 min. Then, 5 µL Lipofectamine was added and tubes were incubated at room temperature for 25 min before transfecting each well with 500 µL of this mixture. Transducing plasmids were pAcEGFPmiR8, pAcEGFPmiR8SP, pAcEGFPmiR34, pAcEGFPmiR305, pAcEGFPmiR375, pAcEGFPNS, pAcEGFPSA, and pAcEGFPSD. Cells were incubated and imaged at 3 d post-transfection. RNA for splicing validation was also gathered 3 d post-transfection. Preliminary in vitro miRNA and target gene expression experiments harvested RNA at 5 d post transfection. For Sua5B in vitro miRNA and target gene expression, cells transfected with pWTAgDNV and pAcEGFPNS served as controls, whereas in Moss55 in vitro miRNA and target gene expression experiments, cells transfected with pWTAgDNV alone served as a control.

Viral production and quantification

To produce virus particles for mosquito infections, Moss55 cells were transfected with selected transducing and helper viruses (as described) and virions were extracted 3 d post-transfection by removing the media, washing cells with 1× phosphate-buffered saline (PBS), and suspending cells in 1 mL 1× PBS. Cells were lysed using three cycles of freeze-thawing and centrifuged at 5000 rpm for 5 min to pellet debris. The virus-containing supernatant was collected and plasmid DNA and free viral genomes were removed using an Ambion TURBO DNA-free kit (AM1907). DNA was extracted using an Omega Bio-tek E.Z.N.A Tissue DNA kit (D3396-02) and viral genome equivalents were determined using standard curves created using AgDNV-coding plasmids with a single copy of each gene-of-interest. Samples and standards were run using PerfeCTa SYBR Green FastMix (Quantabio, 95,072–012) on a Qiagen Rotor-Gene Q at 95 °C for 2 min followed by 40 cycles of 95 °C for 10 s, 60 °C for 40 s, and 72 °C for 30 s. Runs were finished with a melt step using a ramp of 55–99 °C rising by 1 °C each step. WT AgDNV was quantified using primers against AgDNV nonstructural gene 1 (NS-RT-IIIF: CATTCGATCACGGAGACCAC, NS-RT-IIIR: GCGCTTGTCGCACTAAGAAAC) and a standard curve of pWTAgDNV. Selected transducing viruses (vAcEGFPmiR8, vAcEGFPmiR8SP, vAcEGFPmiR34, vAcEGFPmiR305, vAcEGFPmiR375, vAcEGFPNS, vAcEGFPSA, and vAcEGFPSD) were quantified using primers against EGFP (GFP-RT-II-F497: TCAAGATCCGCCACAACATC, GFP-RT-II-R644: TTCTCGTTGGGGTCTTTGCT) and a standard curve of pAcEGFP. Each production of virus consisted of a mixture of vWTAcEGFP and a transducing virus.

Mosquito injections

Female An. gambiae mosquitoes (Keele strain) that were 3 d post-emergence were injected intrathoracically with 200 nL densovirus mixture containing both wild-type vWTAgDNV and transducing virus (either vAcEGFPmiR8, vAcEGFPmiR8SP, vAcEGFPmiR34, vAcEGFPmiR305, vAcEGFPmiR375, vAcEGFPNS, vAcEGFPSA, or vAcEGFPSD) using a Drummond Scientific Nanoject III (3-000-207) and Drummond Scientific 10 µL microcapillary tubes (3-000-210-G) pulled using a Sutter Instrument Co. Model P-2000 (Heat 400, Fil 4, Vel 40, Del 140 Pul 140). Three biological replicates in mosquitoes were completed. For each replicate, mosquitoes were injected with ~10⁶–10⁷ transducing virus particles and 10⁶–10⁸ WT DNV particles (Table S2). Mosquito treatment groups were kept in separate cardboard cup cages with 10% sugar solution w/v ad libitum until RNA extraction or imaging. RNA was harvested and tested from three biological replicates.

RNA extractions and cDNA production

For both in vitro and in vivo experiments, RNA was extracted using an Omega Bio-tek MicroElute Total RNA Kit (R6831-02). For in vitro experiments, RNA was extracted 3 d post-transfection for intronic splicing assessments or 5 d post-transfection for miRNA and target gene quantification. For in vivo experiments, mosquitoes were individually homogenized 10 d post-injection in lysis buffer using zinc-plated steel BB pellets (Daisy 0.177 cal or 4.5 mm) and a Qiagen TissueLyser II with a lysis program lasting 2 min with a frequency of 30 Hz. Following homogenization, RNA was extracted and DNase treated either on the column using an Omega Bio-tek RNase-free DNase Set I kit (E1091) or following RNA extraction using an Ambion DNA-free DNA Removal Kit (AM1906). For target gene quantification or assessment of intronic splicing, cDNA was synthesized using a Quantabio qScript cDNA synthesis kit (95047–500); whereas for miRNA quantification, samples were converted to cDNA using the HighSpec option in the Qiagen miScript II RT kit (218161) and diluted 1:10.

Intronic splicing, miRNA expression, and target gene quantification

In vitro and in vivo intronic splicing was assessed using primers spanning the intronic region (GFP-COLPCRF: CTGACCTACGGCGTGCAGTGC, RGFP-COLPCRR: CGGCCATGATATAGACGTTGTGGC). PCR products were run on 2% agarose gels and imaged using a UVP GelDoc-It transilluminator. Spliced transcripts resulted in a product of 274 bp, whereas PCR reactions using DNA plasmid controls or unspliced transcripts produced variably sized amplicons depending on insert size with most being ~480 bp.

Target gene qPCR reactions were run on a Qiagen Rotor-Gene Q using PerfeCTa SYBR Green FastMix (Quantabio, 95072–012) or an Applied Biosystems 7900HTFast Real-Time PCR System with Applied Biosystems PowerUp SYBR Green Master Mix (A25724), with conditions of 95 °C for 2 min, 40 cycles of 95 °C for 10 s, 60 °C for 40 s, and 72 °C for 30 s, and a melt curve with a ramp from 55 °C to 99 °C with 1 °C change per step. Primers for An. gambiae Swim cDNA were developed during this study, whereas others came from published studies (Table S3) [48, 57–60].

Reactions to quantify miRNAs used Qiagen miScript SYBR Green PCR kits (218075) and a Qiagen Rotor-Gene Q with a universal reverse primer and forward primers consisting of the sequences of each mature miRNA (Table S4) [40]. An. gambiae U6 levels served as a reference with which to compare miRNA levels. Conditions for miRNA qPCR reactions were 95 °C for 15 min followed by 40 cycles of 95 °C for 15 s, 60 °C (for all miRNAs during cell culture replicates as well as for in vivo miR34) or 55 °C (all U6 reactions and in vivo miR375) for 60 s, and 72 °C for 20 s. All reactions ended with a melt curve consisting of a ramp from 55 °C to 99 °C that increased 1 °C per step.

Data analysis

All qPCR data was analyzed using the delta-delta Ct method to calculate the fold change in expression relative to reference genes (S7 for mRNA transcripts and U6 for mature miRNA quantification unless otherwise noted). The fold change expression data was log2 transformed and a D’Agostino–Pearson omnibus K2 test was used to assess normality in Graphpad Prism 9. If both the control and experimental groups passed the normality test, a parametric unpaired two-tailed t-test assuming equal standard deviations was used to measure statistical significance. If either or both groups failed the D’Agostino–Pearson normality test, a nonparametric two-tailed Mann–Whitney test was used to compare ranks and to assess significance. Significant P values (< 0.05) were reported on graphs. All graphs report fold change expression using a log2 scale. The mean and standard error of the mean was reported for groups analyzed using an unpaired t-test, whereas median and 95% confidence intervals were shown for groups compared using a nonparametric two-tailed Mann–Whitney test.

Results

In vitro EGFP expression

When An. gambiae Sua5B or Moss55 cells were co-transfected with pWTAgDNV and the original transducing plasmid pAcEGFP that lacked the created intron, strong EGFP expression was observed (AcEGFP; Fig. 2A, B). Sua5B and Moss55 cells that were co-transfected with pWTAgDNV and transducing plasmid pAcEGFPmiR8, pAcEGFPmiR8SP, pAcEGFPmiR34, pAcEGFPmiR305, or pAcEGFPmiR375 had visible EGFP expression indicative of intronic splicing 3 d post-transfection (miR8, miR8SP, miR34, miR305, and miR375; Fig. 2A, B). EGFP expression was also present when cells were transfected with pWTAgDNV and the nonsense-RNA-encoding transducing plasmid pAcEGFPNS (NS RNA; Fig. 2A, B). When splice donor or splice acceptor sites were mutated, EGFP expression was not detectable in Sua5B cells 3 d post-co-transfection with pWTAgDNV and pAcEGFPSA, while greatly reduced splicing was observed in cells co-transfected with pWTAgDNV and pAcEGFPSD, indicating that splicing was largely dependent on splice site sequences (SA mutant and SD mutant; Fig. 2A). In An. gambiae Moss55 cells, similar expression patterns were observed, with less EGFP expression occurring in Moss55 cells co-transfected with pWTAgDNV and pAcEGFPSD than in Sua5B cells co-transfected with the same constructs (Fig. 2B). EGFP signals were generally weaker in Moss55 compared with Sua5B cells; however, results were consistent from both cell lines and indicate that intronic splicing is induced in a sequence specific manner for a wide variety of pre-miRNAs, miRNA sponges, and small RNAs in An. gambiae cell lines of varied lineage.

Fig. 2.

EGFP expression in vitro 3 d after co-transfection with pWTAgDNV and selected transducing plasmids. A EGFP expression in co-transfected Sua5B cells. B EGFP expression in co-transfected Moss55 cells. Panels are labeled with the transducing construct that was co-transfected with pWTAgDNV. Scale = 100 µm

Confirmation of in vitro intronic splicing

In vitro intronic splicing was further validated using primers spanning the intronic insert. A 274 bp PCR product consistent with splicing was observed in Sua5B cells 3 d post-co-transfection with pWTAgDNV and transducing constructs pAcEGFPmiR8, pAcEGFPmiR8SP, pAcEGFPmiR34, pAcEGFPmiR305, pAcEGFPmiR375, and pAcEGFPNS (cDNA samples 1–6; Fig. 3A). Unspliced plasmid DNA samples had PCR product sizes dependent on intronic length. This was ~480 bp for most constructs, although the miR8SP-expressing plasmid had a larger insert and an amplicon of 572 bp (plasmid samples; Fig. 3A). Sua5B cells co-transfected with pWTAgDNV and constructs containing mutated splice acceptor or splice donor sequences (pAcEGFPSA or pAcEGFPSD) exhibited some level of intron splicing despite absent or greatly reduced visible EGFP expression (cDNA samples 7 and 8, Fig. 3A; SA mutant and SD mutant, Fig. 2A). This indicates that some transcripts are spliced despite the lack of predicted splicing via NetGene2, but that this splicing may be incomplete or in a location that causes a disruption in EGFP expression due to a stop codon. Faint ~480 bp bands, indicating the presence of some unspliced transcript, were also observed in PCR reactions using cDNA from cells co-transfected with pWTAgDNV and pAcEGFPNS (cDNA sample 6; Fig. 3A). This points to some level of splicing disruption in these constructs, yet, given the visible EGFP expression in cells co-transfected with pWTAgDNV and pAcEGFPNS, this may be explained by splicing intermediates (NS RNA; Fig. 2A). Both cDNA and plasmid versions of EGFP lack the intron sequence and have the same PCR product size of 274 bp (cDNA sample 9, plasmid sample 9; Fig. 3A).

Fig. 3.

Intronic splicing in vitro 3 d post co-transfection alongside unspliced DNA plasmid control samples. A Intronic splicing and plasmid controls in co-transfected Sua5B cells. B Intronic splicing and plasmid controls in co-transfected Moss55 cells. In both A and B, matching numbers indicate cDNA and plasmid versions of the same construct. Samples are as follows: (1) miR8, (2) miR8SP, (3) miR34, (4) miR305, (5) miR375, (6) nonsense RNA, (7) SA mutant, (8) SD mutant, (9) EGFP lacking the intron, (10) no template control. Splicing of the intron in cDNA samples resulted in a PCR product of 274 bp, whereas a lack of splicing, as observed in plasmid controls on the right side of the gel, resulted in bands of ~480 bp for most constructs and 572 bp for miR8SP in well 2

Similar splicing patterns were observed in Moss55 cells (Fig. 3B). Cells co-transfected with pWTAgDNV and transducing plasmid pAcEGFPmiR8, pAcEGFPmiR8SP, pAcEGFPmiR34, pAcEGFPmiR305, pAcEGFPmiR375, or pAcEGFPNS all had 274 bp bands, indicative of splicing (cDNA samples 1–6; Fig. 3B). Cells co-transfected with pWTDNV and pAcEGFPmiR8, pAcEGFPmiR34, or pAcEGFPNS also had larger ~480 bp bands, consistent with some level of unspliced transcript or splicing intermediates (cDNA samples 1,3, and 6; Fig. 3B). In cells co-transfected with pWTAgDNV and pAcEGFPNS, another intermediate-sized band was also present (cDNA sample 6; Fig. 3B). When cells were co-transfected with pWTAgDNV and the splice acceptor mutant pAcEGFPSA, a ~480 bp band, representative of a lack of splicing, as well as a 274 bp band, consistent with splicing, was observed despite of a lack of EGFP expression in transfected cells (cDNA sample 7, Fig. 3B; SA mutant, Fig. 2B). Cells co-transfected with pWTAgDNV and the splice donor mutant pAcEGFPSD produced a single strong  ~480 bp band representative of a lack of splicing despite faint EGFP expression observed in transfected cells (cDNA sample 8, Fig. 3B; SD mutant, Fig. 2B). As before, PCRs of plasmid DNA resulted in larger ~480 bp bands for most constructs (plasmid samples 1 and 3–8, Fig. 3B). A band of 572 bp, reflective of a larger intronic segment, was detected for pAcEGFPmiR8SP (plasmid sample 2; Fig. 3B). Plasmid DNA from pAcEGFP, as well as cDNA from cells co-transfected with pWTAgDNV and pAcEGFP, lacked the intron and produced bands of 274 bp (plasmid sample 9, cDNA sample 9; Fig. 3B).

In vivo EGFP expression and intronic splicing

Isolations of vWTAgDNV and each transducing virus purified from Moss55 cells were injected into adult female mosquitoes that were 3 d old, with images of these mosquitoes taken 10 d later. Punctate EGFP expression indicative of splicing was observed in the thorax and abdomen of mosquitoes co-injected with vWTAgDNV and vAcEGFPmiR8, vAcEGFPmiR8SP, vAcEGFPmiR34, vAcEGFPmiR305, vAcEGFPmiR375, or vAcEGFPNS (Fig. 4A). Little to no EGFP expression was observed in mosquitoes co-injected with vWTAgDNV and vAcEGFPSA (SA mutant, Fig. 4A). Mosquitoes injected with vWTAgDNV and vAcEGFPSD exhibited weak EGFP expression that remained localized to the mosquito thorax (SD mutant; Fig. 4A). In vivo intronic splicing was measured as before via PCR of cDNA made from 10 d post-injection mosquitoes. Spliced 274 bp bands were observed in cDNA samples taken from mosquitoes co-injected with vWTAgDNV and vAcEGFPmiR8, vAcEGFPmiR8SP, vAcEGFPmiR34, vAcEGFPmiR305, vAcEGFPmiR375, or vAcEGFPNS (cDNA samples 1–6; Fig. 4B). Faint unspliced bands were also observed in mosquitoes co-injected with vWTAgDNV and vAcEGFPmiR34, vAcEGFPmiR375, or vAcEGFPNS (cDNA samples 3, 5 and 6; Fig. 4B). Plasmid controls resulted in larger bands of ~480 bp for all constructs except for the larger miR8SP insert, which produced a band of 572 bp (plasmid samples 1–6; Fig. 4B).

Fig. 4.

EGFP expression and intronic splicing in 10 d post-injection mosquitoes. A Punctate expression is visible in mosquitoes co-injected with vWTAgDNV and vAcEGFPmiR8, vAcEGFPmiR8SP, vAcEGFPmiR34, vAcEGFPmiR305, vAcEGFPmiR375, and vAcEGFPNS (panels miR8, miR8SP, miR34, miR305, miR375, and NS miRNA). Little to no EGFP expression was present in mosquitoes co-injected with vWTAgDNV and vAcEGFPSA (SA mutant panel). Weak expression is visible in mosquitoes co-injected with vWTAgDNV and vAcEGFPSD (SD mutant panel). B PCR product gel measuring intronic splicing in female An. gambiae 10 d post-injection with cDNA samples (left side) alongside unspliced DNA plasmid control samples (right side). Matching numbers indicate cDNA and plasmid versions of the same construct. Samples are as follows: (1) miR8, (2) miR8SP, (3) miR34, (4) miR305, (5) miR375, (6) nonsense RNA, and (7) no template control. In vivo splicing resulted in a PCR product of 274 bp for wells 1–6 on the cDNA side. A lack of splicing in plasmid controls for wells 1–6 on the plasmid side resulted in bands of ~480 bp, or 572 bp in the case of pAcEGFPmiR8SP

Preliminary in vitro work to select miRNA targets

Preliminary work in Sua5B cells showed that transfection with pWTAgDNV and pAcEGFPmiR8 or pAcEGFPmiR375 led to higher levels of miR8 and miR375, respectively, 5 d post-transfection (Fig. S1A and C). miR34 levels were reduced rather than elevated when Sua5B cells were transfected with pWTAgDNV and pAcEGFPmiR34, perhaps indicating the processing of this transcript into an anti-miRNA rather than expression of the predicted mature miRNA (Fig. S1B). Transfection with pWTAgDNV and pAcEGFPmiR305 or miR8SP did not result in any significant changes in miRNA levels (data not shown). We also observed significant upregulation of the predicted target gene Cactus in cells transfected with pWTAgDNV and pAcEGFPmiR375 (Fig. S2A). Given the lack of significant change in miRNA levels for miR8SP and miR305, and difficulty in interpreting results from miR8 in light of the miR8SP results, these miRNAs were not evaluated in later experiments. A preliminary in vivo experiment was also carried out where mosquitoes 10 d post-injection with vWTAgDNV and vAcEGFPmiR34 showed significant elevation in miR34 target gene transcripts MISO, Caspar, and Rel2 despite a lack of change in the levels of miR34, Cactus, and Rel1A (Fig. S3A–F).

In vivo miRNA and target gene expression

After down-select, larger-scale in vivo experiments focused on mosquitoes injected with vWTAgDNV and either vAcEGPFmiR34 or vAcEGFPmiR375. Mosquitoes injected with vWTAgDNV and vAcEGPFmiR34 were evaluated for changes in MISO, Caspar, and Rel2 transcript levels, whereas mosquitoes injected with vWTAgDNV and vAcEGPFmiR375 were tested for differences in Cactus and Rel1A transcript levels.

Similar to preliminary in vivo results, levels of miR34 were not significantly different between mosquitoes injected with vWTAgDNV and vAcEGFPmiR34 and control mosquitoes injected with vWTAgDNV and vAcEGFPNS (Fig. 5A). In addition, no differences were seen in Caspar or MISO transcript levels 10 d post-injection (Fig. 5B and C). However, Rel2 transcripts were more abundant 10 d post-infection in mosquitoes injected with vWTAgDNV and vAcEGFPmiR34 compared with those injected with vWTAgDNV and vAcEGFPNS (Fig. 5D). The direction of this change in target gene transcript abundance may indicate miRNA signaling through the lessor known RNA activation pathway rather than the RNAi pathway, or that this miRNA acts on an upstream inhibitor of Rel2 [61–63].

Fig. 5.

miR34 and target gene transcript expression in mosquitoes 10 d post-injection with vWTAgDNV and vAcEGFPmiR34. A Expression of miR34 was unchanged in mosquitoes that were injected with vWTAgDNV and vAcEGFPmiR34. B Expression of Caspar transcripts was not altered following injection. C Levels of MISO were also not changed 10 d post-injection. D Rel2 expression was enhanced in mosquitoes following injection of vWTAgDNV and vAcEGFPmiR34. Dashed lines indicate a fold change of 0. Green dots represent individual mosquitoes co-injected with vAcEGFPmiR34 and pWTAgDNV, whereas blue squares indicate individual mosquitoes co-injected with control vAcEGFPNS and vWTAgDNV. Data in A–D were normal as assessed by a D’Agostino–Pearson normality test and were analyzed using a two-tailed unpaired t-test

When mosquitoes were injected with vWTAgDNV and vAcEGFPmiR375, miR375 levels were reduced 10 d post-infection (Fig. 6A). This was unexpected given the strong increases seen in preliminary in vitro experiments and may indicate that the introduced miRNA is not processed in vivo as expected, and instead produces an anti-miRNA that binds to endogenous miR375. Both Cactus and Rel1A transcript levels were elevated in mosquitoes 10 d post-infection (Fig. 6B and C). This contrasts with the predicted increase in Cactus and decrease in Rel1A from Ae. aegypti (Table 1).

Fig. 6.

miR375 and target gene transcript expression in mosquitoes 10 d post-injection with vWTAgDNV and vAcEGFPmiR375. A Expression of miR375 was slightly decreased in mosquitoes injected with vWTAgDNV and vAcEGFPmiR375. B Expression of Cactus transcripts was elevated in mosquitoes 10 d post-injection. C Levels of Rel1A expression was slightly increased 10 d post-injection with vWTAgDNV and vAcEGFPmiR375. Dashed lines indicate a fold change of 0. Green dots represent individual mosquitoes co-injected with vAcEGFPmiR375 and pWTAgDNV, whereas blue squares indicate individual mosquitoes co-injected with control vAcEGFPNS and vWTAgDNV. Data in A and B was not normally distributed as assessed by a D’Agostino–Pearson normality test and was analyzed using a nonparametric two-tailed Mann–Whitney test. Data in C was normally distributed and analyzed using a two-tailed unpaired t-test

Discussion

These experiments show that splicing of the developed AgDNV-delivered intron is robust in vitro within two different An. gambiae cell lines of varied lineages, as well as in vivo for a variety of endogenous pre-miRNA sequences, one developed miRNA sponge sequence, and one random RNA sequence. Occasionally, unexpected unspliced transcripts were observed in PCR assays. These likely represent splicing intermediates or pre-splicing transcripts, as the presence of these larger bands did not correlate with altered EGFP expression in vitro or in vivo (Figs. 2A and B, 4A). This demonstrates that splicing is specific to the developed intronic sequence and not altered by the cargo sequence. Further supporting this, intronic splicing can be eliminated or reduced through alteration of splice donor or splice acceptor sequences. Although mutated splice donor and splice acceptor constructs within the two cell lines sometimes produced bands seemingly consistent with some level of splicing, reduced or absent in vitro EGFP expression observed for these constructs indicates that any splicing that occurs is greatly suppressed, modified, or results in a stop codon as predicted (cDNA samples 7 and 8, Fig. 3A; cDNA samples 7 and 8, Fig. 3B; SA and SD mutants, Fig. 2A and B). Although splicing was not assessed by PCR in vivo for splice donor and splice acceptor mutants, mosquitoes injected with vWTAgDNV and vAcEGFPSA had little to no EGFP expression, and mosquitoes injected with vWTAgDNV and vAcEGFPSD exhibited weak EGFP expression (SA and SD mutants, Fig. 4A). Thus, this intron and expression strategy represents a promising new method for introducing small RNAs both in vitro and in vivo.

Despite the success of this expression strategy in vitro, some differences between cell types were observed. Sua5B cells are larger in size and produced noticeably stronger EGFP expression than Moss55 cells (Fig. 2A and B). While not investigated in this study, this difference in observed EGFP expression between cell lines may be due to chronic AgDNV infection already present in Sua5B cells that could enhance the viral replication and packaging efficiency of transducing constructs [13]. Moss55 cells lack natural AgDNV infection, and thus, rely solely on AgDNV constructs expressed from transfecting plasmids. Alternatively, Sua5B cells are considered to have hemocyte-like properties, whereas Moss55 cells have an epithelial origin, and differences in cell lifecycle or rate of transcription may explain the variation in EGFP intensity [64–66]. Despite these differences in AgDNV infection status and EGFP intensity, no differences were observed in the viral titers produced by the different cell lines for mosquito injections. To better control the genetic diversity of purified viruses, Moss55 cells were used to grow viral stocks used for in vivo experiments.

In vivo experiments largely focused on mosquitoes 10 d post-injection with vWTAgDNV and either vAcEGFPmiR34 or vAcEGFPmiR375. While experiments with mosquitoes injected with vWTAgDNV and vAcEGFPmiR34 did not result in any differences in miR34 levels, similar to preliminary results, Rel2 transcript levels were slightly elevated 10 d post-infection (Fig. 5A and D). Although a directionality to Rel2 changes was not predicted prior to experiments, this increase in Rel2 transcript abundance indicates that miR34 may act on Rel2 through the RNA activation pathway or that miR34 acts on another transcript that in turn influences Rel2 abundance, possibly via an upstream inhibitor of Rel2 (Table 1) [61–63]. Similarly, although miR375 was predicted to bind to Cactus and Rel1 and cause an increase in Cactus and a decrease in Rel1, mosquitoes that were injected with vWTAgDNV and vAcEGFPmiR375 had slightly elevated Cactus and Rel1A transcript levels 10 d post-infection (Fig. 6B and C). This was unexpected, as Cactus is a negative regulator of Rel1, and if binding of miR375 results in transcript depletion though RNAi, in theory, binding would decrease the transcript levels of both. As with miR34, it is possible that miR375 is acting through RNA activation or on transcripts upstream of Cactus or Rel1. In addition, as miR375 levels were slightly decreased in mosquitoes infected with vWTAgDNV and vAcEGFPmiR375, it is also possible that this miRNA is being processed into an anti-miRNA and that this binds to mature miR375 or operates in a different way than mature miR375 (Fig. 6A).

There are many possible reasons for differences between what we observed from in vivo expression experiments versus what was expected on the basis of prior predictions (Table 1). Most importantly, although there have been some prior studies for certain miRNAs, the true direct targets of these An. gambiae miRNAs have largely not been identified and most targets have only been bioinformatically predicted (Table 1). Further, muted changes in miRNA or target gene transcripts could be due to organ-specific effects that are diluted when analyzing the whole mosquito body. Inconsistencies could also be due to a lack of effective miRNA expression in vivo or indicate that infection levels of transducing viruses were lower than those needed to induce a significant alteration in endogenous miRNAs. Despite these inconsistencies between predicted and observed alterations, we did observe changes in miR375 levels and in some miR34 and miR375 target transcripts, indicating that the developed artificial intron has potential for use in miRNA expression and may be more successful if specific target genes are better characterized (Figs. 5D, 6A–C).

Conclusions

Future experiments should examine miRNA and target gene transcript levels within organs known to support AgDNV infection, such as the midgut, ovaries, and fat body. By examining specific organs rather than the whole mosquito, noise in the system may be reduced and changes in miRNAs or target genes may be more evident. The developed expression system can conceptually also be useful for expressing other effectors such as small interfering RNAs (siRNAs) developed against specific genes. By expressing siRNAs with specific targets, the true capabilities of this virus and intron expression system could be better tested and assessed for meaningful intervention in biological or field settings. Additional studies of AgDNV viral replication dynamics, dosing requirements for injections, and organ specificity would greatly aid future work and the further development of this symbiont and expression system as a tool for paratransgenesis. Further, it may be that a nondefective recombinant approach, such as that described in AaeDNV by Liu et. al., 2016, may provide an advantage in expression consistency that could be explored in future AgDNV studies [20].

Author contributions

R.M.J., K.J.M., and J.L.R. conceived the study. Y.S. designed plasmids further modified in this study. K.J.M. designed initial intronic sequences that were further modified in this study. R.M.J. performed experiments. J.L.R. provided supervision and experimental oversight. R.M.J. wrote the initial draft of the manuscript. R.M.J., H.C.M., and J.L.R. edited the initial manuscript. All authors reviewed and approved the final manuscript.

Funding

This research was supported by NIH/NIAID grant R01AI128201, NSF/BIO grant 1645331, USDA Hatch Project 4769, a grant with the Pennsylvania Department of Health using Tobacco Settlement Funds, and funds from the Dorothy Foehr Huck and J. Lloyd Huck endowment to J.L.R.

Data availability

Data supporting the main conclusions of this study are included in the manuscript.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.