Comparative analysis of DNA vectors at mediating RNAi in Anopheles mosquito cells and larvae

Transmission of human malaria, a disease that causes more than one million deaths a year, is accomplished by a small group of mosquito species belonging to the Anopheles genus. The recent development of an efficient gene transfer technology for Anopheles stephensi and Anopheles gambiae mosquitoes (Catteruccia et al., 2000a; Grossman et al.,2001), combined with the recent completion of the A. gambiae genome sequence project (Holt et al., 2002), has paved the way for studies on the genetics of these important disease vectors. However, progress towards the identification of mosquito molecules involved in the interaction with the Plasmodiummalaria parasite has been hampered by the lack of efficient technologies, such as mutagenesis screens and gene knock-out by homologous recombination,routinely used in other organisms to perform functional genomic studies. Gene silencing by RNA interference (RNAi), an evolutionarily conserved phenomenon triggered by double-stranded RNA molecules (dsRNAs), has recently been envisaged as a powerful tool for studying gene function in different model organisms (Fire et al., 1998). Although the mechanism by which dsRNA mediates gene silencing is not fully understood, it is known to invoke a multistep process that results in the cleavage of long dsRNAs into 21–25 nucleotide fragments that then direct the sequence-specific degradation of homologous endogenous mRNA(Bernstein et al., 2001; Elbashir et al., 2001; Hamilton and Baulcombe, 1999; Hammond et al., 2000; Zamore et al., 2000).

To date, the only reported attempts at using RNAi to study gene function in Anopheles mosquitoes have relied upon the direct delivery of high concentrations of in vitro-synthesised dsRNA molecules in cell lines and adults (Blandin et al.,2002; Levashina et al.,2001). While this approach offers a rapid genetic screen of putative target genes, interference to gene expression is transient,non-inheritable and subject to significant variations between individuals. In some organisms, including Caenorhabditis elegans(Tavernarakis et al., 2000),trypanosomes (Bastin et al.,2000; Shi et al.,2000), Drosophila(Fortier and Belote, 2000; Kennerdell and Carthew, 2000; Piccin et al., 2001) and plants (Smith et al., 2000),stable expression of dsRNA from integrated transgenes exhibiting dyad symmetry has recently been achieved, thus overcoming the problems related to the transient nature of RNAi mediated by the delivery of in vitro-synthesised dsRNA. The stable RNAi approach is anticipated to contribute significantly to the investigation of the function of genes involved in mosquito–parasite interactions in Anophelesmosquitoes.

In this study, we have assessed the molecular requirements for maximising the silencing efficiency of dsRNA-encoding genes in Anopheles. We have developed a series of constructs marked with the red fluorescent protein DsRed (Matz et al., 1999), in which sense and antisense regions of the coding sequence of the green fluorescent protein EGFP (Heim and Tsien, 1996) were connected by different spacers and placed under the control of a constitutive promoter. The silencing activity of these constructs was assessed in A. stephensi larvae and A. gambiae cells transiently expressing the EGFP target gene. Furthermore, an RNAi construct was also analysed for its ability to silence an EGFP gene stably integrated in the genome of A. stephensilarvae.

All EGFP inverted repeats (IR) were produced by duplicating, with opposite direction, the EGFP gene (Clontech, Palo Alto, USA) by directionally inserting a 465 bp (from position 219 to 683) or full-length 720 bp (from 1 to 719) PCR product into the plasmid pCR2.1v (Invitrogen, Paisley, UK). BamHI and EcoRV restriction sites were added to the 5′and 3′ ends of the sense arm product, respectively, and XbaI and XhoI restriction sites to the 5′ and 3′ ends of the antisense arm, respectively. To prepare pIR-EGFP₄₆₅S and pIR-EGFP₇₂₀S, a 67 bp PCR product corresponding to intron 1 from A. gambiae lysozyme gene (Kang et al., 1996; GenBank accession U28809) was amplified from A. gambiae genomic DNA, adding EcoRV and XhoI restriction sites to the 5′ and 3′ ends, respectively, and inserted between the EGFP-IR (Fig. 1B). pIR-EGFP₄₆₅NS was produced by inserting the intron described above in its reverse non-splicing orientation, by inverting the restriction sites(Fig. 1B). pIR-EGFP₄₆₅Linker was produced by inserting a non-palindromic 10 bp oligonucleotide (ATCGTTAACC) between the EGFP-IR(Fig. 1B). In each construct,the IR-EGFP cassette was then inserted between actin5C promoter and Hsp70 terminator elements from Drosophila melanogaster, and cloned as a NotI cassette into pMiRED(Nolan et al., 2002), which contained the DsRed marker gene also cloned under the control of the actin5C promoter (Fig. 1A,C).

A. gambiae Sua 4.0 cells were grown in Schneider's Drosophila medium (GIBCO, Paisley, UK) supplemented with 10% foetal calf serum (FCS), 100 units ml^–1 penicillin and 100 μg ml^–1 streptomycin (Invitrogen) at 25°C. A total of 4×10⁵ cells ml^–1 were plated 24 h before co-transfection experiments, and solutions containing 4 μg each of pMinEGFP(Fig. 1A) and either pMiRED or a dsRNA-transcribing plasmid were co-transfected as described previously(Catteruccia et al., 2000b). Cells were examined two days after transfection at a wavelength of 490 nm to detect EGFP expression and 565 nm to detect DsRed expression.

Wild-type and transgenic Anopheles stephensi Liston adult mosquitoes (strain sd 500) were maintained at 28°C, 70% humidity and fed on 5% glucose. To induce egg production, female mosquitoes 3–5 days old were starved for 5 h and allowed to feed on mouse blood. Two days after blood feeding, eggs were laid and transferred into buckets with H₂O containing 5% NaCl. Larvae were grown at 25°C, 70% humidity and fed on fish food. After 10–12 days, pupae were collected and adult mosquitoes were allowed to emerge in cages.

A. stephensi mosquito embryos were injected essentially as described previously (Catteruccia et al.,2000a). For transient RNAi studies, wild-type A. stephensi embryos were injected with a mixture of pMinEGFP (400 μg ml^–1) and either pMiRED or one of the dsRNA-transcribing plasmids (400 μg ml^–1). For RNAi studies against stably expressed EGFP, A. stephensi line V_B(Catteruccia et al., 2000a) was injected with plasmid pIR-EGFP₄₆₅Linker (400 μg ml^–1). The levels of EGFP expression were assessed daily in larvae positive for DsRed at the wavelengths described above.

Cells and larvae were captured on a Nikon inverted microscope with an attached Nikon DXM1200 digital camera. Fluorescent gene expression was quantified using the Lucia G image-processing and analysis software (Version 4.61, Nikon UK, Kingston, UK). The levels of EGFP and DsRed expression were measured using the MeanGreen and MeanRed feature of the software, which calculates the statistical mean of the intensity of the green or red components of pixels, respectively. Using this software, it was possible to calculate the mean EGFP fluorescence in only those cells in which DsRed was co-expressed. The software was validated by measuring the intensity of EGFP expression in homozygous and heterozygous larvae from the A. stephensi line V_B. Briefly, individuals homozygous for the EGFP insertion consistently showed MeanGreen values that were approximately double those from heterozygous larvae of the same age,demonstrating a good correlation between MeanGreen values and transgene copy number. For statistical analyses, unpaired t-tests were performed;the null hypothesis was rejected at P≥0.05.

In order to create a reliable and objective model of RNAi in Anopheles, the green fluorescent protein EGFP from plasmid pMinEGFP (Catteruccia et al.,2000a; Fig. 1A) was selected as the target gene, since it allowed a direct and non-invasive investigation of dsRNA-mediated silencing at different time points without the need to sacrifice individuals. Studies in plants have suggested that the region linking the perfect inverted repeat (IR) sequences of the target gene is critical to silencing efficiency (Smith et al., 2000). In this light, a series of DNA constructs encoding dsRNA molecules were developed, containing sequences of the EGFPtarget gene cloned in sense and antisense orientations and linked in a head-to-head manner by distinct spacer regions. In the tobacco plant Nicotiana tabacum, the process of intron excision from hairpin RNA(hpRNA) has been reported to enhance gene-silencing efficiency(Smith et al., 2000). To assess whether this phenomenon could also be observed in Anopheles, a construct was developed that contained 465 bp sense and antisense EGFP sequences flanking a 67 bp intron from the A. gambiaelysozyme gene (Kang et al.,1996; pIR-EGFP₄₆₅S; Fig. 1B). The same intron was also cloned in its inverted non-splicing orientation in construct pIR-EGFP₄₆₅NS, with the aim of producing a long hairpin loop in the dsRNA molecule (Fig. 1B). Furthermore, to determine whether the length of the hairpin loop affects silencing efficiency, construct pIR-EGFP₄₆₅Linker was developed,which contained a 10 bp-long linker between the sense and antisense EGFP arms (Fig. 1B). An additional construct, pIR-EGFP₇₂₀S, contained 720 bp arms corresponding to the full length of the EGFP gene separated by the A. gambiae intron in its splicing orientation(Fig. 1B), as studies in other organisms have shown that the length of the dsRNA affects RNAi efficiency(Ngo et al., 1998; Parrish et al., 2000; Tuschl et al., 1999). All dsRNA constructs were derived from plasmid pMiRED(Nolan et al., 2002), which contained a red fluorescent protein (DsRed) marker gene cassette inserted within the arms of the minos transposable element(Franz and Savakis, 1991),with the perspective of developing stable transgenic lines expressing dsRNA(Fig. 1A). The EGFPtarget gene from plasmid pMinEGFP (Fig. 1A) and the dsRNA-transcribing genes from each IR construct were all placed under the control of the actin5C promoter from Drosophila melanogaster, to ensure similar transcriptional levels. A schematic representation of all dsRNA constructs is given in Fig. 1C. All IR constructs were in vitro transcribed to prove they could produce dsRNA species (data not shown).

To assess the ability of a DNA-based system to mediate RNAi in Anopheles cells, dsRNA-transcribing constructs, marked with the DsRed marker gene, were co-transfected with target plasmid pMinEGFP into the A. gambiae Sua 4.0 cell line in a series of consecutive experiments. In this experimental system, the occurrence of RNAi against the target gene will be indicated by a decrease in the intensity of EGFP fluorescence in those cells that express the DsRed marker. Control experiments were performed by co-transfecting the target construct pMinEGFP with plasmid pMiRED, containing the DsRed marker cassette but not the EGFP-IR(Fig. 1A). In these cells,perfect co-localization of the green-fluorescent target and red-fluorescent control plasmids was observed (Fig. 2A). The intensity of fluorescence of the EGFP and DsRed proteins was not affected by the expression of the other marker (data not shown). In all samples, quantitative analysis was performed using an image-processing software that allowed the quantification of the mean intensity of green and red fluorescence of each cell. In cells co-transfected with pMinEGFP and pIR-EGFP₄₆₅S, in which the A. gambiae intron was placed in its splicing orientation, a 93.6% reduction of EGFP expression was observed as compared with control experiments (Fig. 3A). Cells transfected with pIR-EGFP₄₆₅NS, containing the intron in its inverted, non-splicing orientation, exhibited varying degrees of EGFP silencing in different cell subsets, ranging from almost complete to very limited silencing and averaging 70.7% with respect to controls (Fig. 3A). In cells transfected with pIR-EGFP₄₆₅Linker, in which the large hairpin loop was replaced with a short 10 bp spacer region, silencing of the target gene expression increased to 98.2% in all transfected cells (Figs 2B, 3A). High-level silencing(96.4% inhibition) was also observed when the dsRNA corresponding to the full-length EGFP sequence was delivered (pIR-EGFP₇₂₀S; Fig. 3A). In all samples analysed, dsRNA transcription did not inhibit DsRed expression, and the levels of the DsRed protein were consistently comparable to controls(Fig. 3A).

To examine whether the RNAi-mediated gene silencing observed in A. gambiae cells could also be achieved in vivo, the activity of the dsRNA constructs was analysed in mosquitoes. A. stephensi embryos were microinjected with a mixture of the target vector pMinEGFP and one of the dsRNA-transcribing plasmids. Larvae surviving the injection procedure were analysed for DsRed fluorescence as an indicator of successful delivery of the DNA species. To ascertain the perfect co-localization of target and targeting constructs, control experiments were performed by injecting a mixture of the pMinEGFP and pMiRED plasmids. In all control larvae analysed, the expression of the green and red fluorescent protein largely coincided, as indicated by their overlapping profiles (Fig. 2C). However, when pIR-EGFP₄₆₅S was co-injected with the target plasmid, a 76% decrease in the level of EGFP expression was observed with respect to control larvae(Fig. 3B). A similar effect was observed after the injection of construct pIR-EGFP₇₂₀S, which caused a 79% reduction in EGFP expression(Fig. 3B). When construct pIR-EGFP₄₆₅Linker was injected, reduction of EGFP expression was almost complete, averaging 93% as compared with control larvae (Figs 2D, 3B). Although some EGFP expression was observed in larvae injected with all three constructs, it appeared to be confined to those few cells that did not exhibit expression of the DsRed gene. In agreement with the cell data, plasmid pIR-EGFP₄₆₅NS had a less significant silencing effect, and EGFP inhibition averaged 58% with respect to controls(Fig. 3B).

We then assessed the effects of the dsRNA-transcribing construct pIR-EGFP₄₆₅Linker on a stably expressed EGFP transgene. This construct was chosen as it had consistently mediated the strongest inhibition of target gene expression in both cell lines and larvae. A. stephensi embryos from transgenic line V_B, stably expressing the EGFP gene under the control of the actin5C promoter,were injected with pIR-EGFP₄₆₅Linker, and EGFP expression was monitored throughout larval development. As a control, homozygous embryos from line V_B were also injected with plasmid pMiRED(Fig. 1A). Quantitative analysis of fluorescence was performed between three days and five days post-hatching, over which period control larvae showed a linear increase in intensity of EGFP (Fig. 3C). Injection of pIR-EGFP₄₆₅Linker significantly decreased the overall intensity of green fluorescent protein expression (Figs 2E, 3C), with the most noticeable effects observed five days post-hatching when 80% silencing was achieved(Fig. 3C). Late larval stages showed more varied silencing effects. In some cases, EGFP expression was slowly restored to normal levels, while in approximately 50% of larvae gene silencing continued to be observed (data not shown). Injections were performed into the posterior end of the embryos, thus the DsRed marker was mainly localised in the last few segments of the larval abdomen(Fig. 2E).

In this study, we performed a systematic analysis of the ability of different DNA constructs to induce RNAi in Anopheles cells and larvae. The advantage of using DNA constructs to generate dsRNAs is based upon their inherent stability over in vitro-synthesised dsRNA molecules. Furthermore, the development of an efficient heritable RNAi technique to study the function of genes expressed at late developmental stages in Anopheles mosquitoes will eventually rely on the use of dsRNA-transcribing constructs capable of producing high-degree silencing combined with good stability of the inverted repeats. A careful analysis of the most suitable construct is therefore particularly important at this preliminary stage to assess the molecular requirements determining the degree of gene silencing.

The results reported here show that inhibition of target gene expression depended considerably upon the spacer region separating the inverted repeats,while the length of the IR itself did not seem to significantly affect silencing efficacy. In particular, the presence of a 67 bp-long spacer(corresponding to the non-splicing orientation of the intron) dramatically reduced inhibition of EGFP expression, while the same spacer in its splicing orientation mediated high-level silencing of the target gene. While the basis for the high efficiency of gene silencing by intron-spliced dsRNA in vivo is not known, it is believed that the process of intron excision may transiently increase the amount of dsRNA either by promoting its formation or by creating a smaller, less nuclease-sensitive loop(Smith et al., 2000). Furthermore, the presence of an intron might increase the genomic stability of RNAi transgenes, as genomic IR have been shown to be unstable(Bi and Liu, 1996; Leach, 1994) and increasing the distance between the IR has been postulated to reduce their recombinogenic potential (Lobachev et al.,2000). On the other hand, a long spacer may obstruct the annealing of the IR, causing a modest number of dsRNA molecules to be formed, or make the dsRNA molecules more nuclease sensitive. Construct pIR-EGFP₄₆₅Linker, containing the smaller spacer, consistently mediated the highest silencing in both cells and larvae. This finding could be explained on the basis of faster kinetics of dsRNA formation, while the intron-containing constructs required splicing to occur before export to the cytoplasm where they could mediate silencing. The two intron-splicing constructs containing different IR lengths performed equally well, with no significant differences between them. This finding seems to indicate that the length of the dsRNA construct does not play a major role in silencing efficiency in Anopheles.

The EGFP target gene transiently introduced into Anopheles cells and embryos provided a simple and impartial model for testing the gene-silencing efficacy of the targeting constructs. This fluorescent marker allowed analysis at the protein level both visually and quantitatively using image-processing software. RNAi was also observed when a stably integrated EGFP gene was targeted. The analysis of RNAi against an endogenous gene could have been compromised by the occurrence of selection against constructs inducing a high degree of inhibition, as this could have had a deleterious effect on fitness and viability of the organism. The levels of EGFP inhibition depended on the amount of pIR-EGFP₄₆₅Linker delivered. In individuals injected with a 10-fold lower concentration of pIR-EGFP₄₆₅Linker, no significant silencing of EGFP expression was observed (not shown). Plasmid DNA persisted for many days after injection, and in most cases DsRed was still visible in fourth instar larvae. This could represent an advantage over using in vitro-synthesised dsRNA molecules, whose half-life has been postulated to be too limited to study late developmental genes(Kennerdell and Carthew, 1998; Misquitta and Paterson, 1999; Montgomery and Fire, 1998; Wianny and Zernicka-Goetz,2000).

The results described here provide the first comparison of DNA constructs in mediating gene silencing in Anopheles and represent an important step towards the development of stable and heritable RNAi in these important malaria vectors.

We would like to thank T. Nolan for stimulating discussions and advice throughout these studies. A.E.B. was supported by the BBSRC, F.C. was supported by the Wellcome Trust.