Antiviral efficacy of short-hairpin RNAs and artificial microRNAs targeting foot-and-mouth disease virus

Introduction

Foot-and-mouth disease virus (FMDV) is the causative agent of a highly devastating disease of cloven-hoofed animals leading to enormous economic losses. FMDV belongs to the Aphthovirus genus in the Picornaviridae family. There are seven FMDV serotypes (A, O, C, Asia1, SAT1, SAT2, SAT3) and several subtypes.

The viral particle is ∼25–30 nm in diameter and consists of a single stranded (+)RNA molecule surrounded by an icosahedric capsid composed by the four structural viral proteins VP1 to VP4. Viral infection begins with the interaction of the FMDV capsid with cellular receptors (α_V integrins in animal hosts, additionally heparansulfate and a newly identified receptor in cultured cells) (Dicara et al., 2008; Lawrence et al., 2016). Receptor-mediated endocytosis guides the viral particles to the cytoplasm, where the whole replication cycle takes place. Acidification of endosomes drives disassembly of the viral capsid and release of the FMDV genome into the cytoplasm, where it is immediately translated in a cap-independent manner to produce the viral polyprotein. Co- and post-translational processing of the FMDV polyprotein originates viral protein precursors and ultimately the mature structural and non-structural proteins (Grubman & Baxt, 2004). Finally, capsids are assembled and newly synthesized genomes are encapsidated and released from the cell in a lytic way. The replication cycle is fast, taking approximately 6–8 h in cell culture.

During replication, viral RNA molecules are synthesized by the viral RNA-dependent RNA polymerase (RdRp) 3Dpol. As other viral RdRps, the enzyme lacks proofreading activity, which contributes to the generation of mutated FMDV genomes in every replication cycle. Hence, in an infected host (cell or animal) the virus appears as a swarm of different but closely related viral variants which are called the viral quasispecies (Domingo et al., 2003; Domingo, Sheldon & Perales, 2012; Andino & Domingo, 2015). The quasispecies nature of FMDV facilitates its rapid adaptation to unfavorable environments.

Following current standards of the World Organisation for Animal Health (OIE), countries are divided according to their foot-and-mouth disease (FMD) status in FMD-free countries with or without vaccination (almost all Americas, Europe and Oceania) and FMD-endemic countries (parts of Africa and Asia). Current approved FMD vaccines are inactivated viruses formulated with oil-based adjuvants. Although they are effective, inactivated vaccines show a number of disadvantages including lack of cross-protection between serotypes and even subtypes, inability to induce long-lasting immunity (there is a need of re-vaccination to maintain protection) and risk of incomplete inactivation during manufacturing. Moreover, current vaccines take ∼7 days to induce protection (Golde et al., 2005). Thus, there is a need of developing new tools against FMDV that may complement vaccination and reduce the risk of infection especially during outbreaks.

RNA interference (RNAi) is a well-conserved mechanism in eukaryotic cells that directs post-transcriptional gene silencing (Fischer, 2015). RNAi is mediated through small RNA molecules that guide ribonucleprotein complexes to their cognate mRNAs and lead to their degradation and/or inhibition of translation. There are many types of small RNAs, including microRNAs (miRNAs) and short interfering RNAs (siRNAs). The former are endogenous molecules encoded in the cellular genome that regulate gene expression. In animals, biogenesis of miRNAs involves RNA polymerase II-mediated transcription of a primary miRNA (pri-miRNA), which is further processed by the RNAse III enzyme Drosha to pre-miRNA hairpins. After export from the nucleus, pre-miRNAs are processed by Dicer to produce mature 21–24 nt miRNAs that exert their function in the cytoplasm of the cell. In turn, siRNAs are exogenous molecules that may be introduced into the cell either as mature siRNAs or as DNA precursors by transfection or by virus-mediated transduction (Castanotto & Rossi, 2009). Importantly, the degree of complementarity between the small RNA and the target RNA defines the mechanism of post-transcriptional gene silencing: while perfect base complementarity mainly drives target RNA degradation by endonucleolytic cleavage, partial complementarity can lead to translational repression. In animals, the canonical siRNA pathway involves perfect siRNA:target RNA recognition, whereas binding of most miRNAs demands base pairing of the ‘seed’ region (mostly located at nucleotides 2–8 of the mature miRNA), allowing for imperfect target RNA recognition (reviewed in (Carthew & Sontheimer, 2009)).

The use of RNAi as an antiviral tool has been explored against different viruses (reviewed in (Qureshi et al., 2018; Shah & Schaffer, 2011)). Indeed, RNAi mediators such as double-stranded or single-stranded siRNAs and plasmid or virus-encoded short hairpin RNAs (shRNAs) have been evaluated in different experimental settings against animal viruses including FMDV (Chen et al., 2004). Genetically modified animals expressing shRNAs targeting FMDV have also been developed with promising results (Jiao et al., 2013; Hu et al., 2015; Deng et al., 2017). Of note, shRNAs are potent RNAi mediators that usually are expressed at very high levels from RNA polymerase III promoters; however, this can potentially lead to saturation of the cellular RNAi processing machinery. To overcome this drawback, artificial miRNAs (i.e., miRNAs rationally designed to target defined sequences) have been favored as RNAi shuttles since they exhibit increased safety with comparable efficacy (Boudreau, Martins & Davidson, 2009). The efficacy of artificial miRNAs (amiRNAs) against FMDV has been investigated to a lesser extent (Du et al., 2011; Gismondi et al., 2014; Basagoudanavar et al., 2019). In this work, we further assess the efficacy of amiRNAs against FMDV and shRNAs targeting the same viral regions in susceptible cells.

Materials and Methods

Results

Selection of target sequences

Target sequences were predicted within FMDV A01L 3Dpol-coding sequence with RNAxs software (Tafer et al., 2008), which applies several prediction parameters including RNA accessibility. A total number of 178 target sequences were predicted along 3D (Fig. 1, blue areas; Table S2), with top 50 predicted targets concentrated in discrete regions (Fig. 1, red areas).

Next, an additional screening of the target sequences ranking in the top positions of the RNAxs output was performed to take into account the prediction rules proposed by Low et al. (2012). Firstly, 19 nt target sequences were extended to 21 nt in order to adapt RNAxs prediction to amiRNA-compatible length. Then, the resulting target sequences were analyzed in terms of the accessibility window proposed by Low et al., i.e., sequences showing a less structured stretch downstream of the 21 nt target region were favored. Lastly, the total free energy change (ΔG_total) of target:siRNA duplex formation was calculated with OligoWalk program (Reuter & Mathews, 2010) (Table 1). After applying these rules, sequences 444–464, 290–310 and 1055–1075 were selected (termed 444; 290 and 1055; Fig. 1, Table 1 and Fig. S2). The selected target sequences ranked in the first positions of the RNAxs output. Indeed, in some cases the 21 nt target sequences included more than one 19 nt predicted target sequence (Table 1). Of note, sequences 290 and 444 displayed a ΔG_total value higher than the one recommended by Low et al. (−25 kcal/mol). Regarding the accessibility window, no target sequence showed the expected accessible 13 nt window starting at position 14 (Fig. S1A). However, all sequences were located in the region of high linear correlation between RNA structure and shRNA efficacy described by these authors (Fig. S1B).

Finally, conservation of the selected target sequences among prototypic strains of the South American serotype pool was evaluated (Fig. S1C). Sequences 290, 444 and 1055 showed some polymorphisms between serotypes, which were distributed mainly outside the seed region.

Silencing efficacy of shRNAs and amiRNAs targeting selected FMDV sequences

A reporter assay was conducted to test the silencing efficacy of different RNAi mediators targeting the selected sequences. To this end, a plasmid encoding EGFP fused to the 3Dpol-coding sequence of FMDV A/Arg/01 was co-transfected with plasmids encoding either shRNAs or pre-amiRNAs targeting each individual FMDV sequence. Short hairpin RNAs targeting sequences 290, 444 and 1055 induced a significant reduction of EGFP expression in co-transfected cells, as determined by fluorescence microscopy and flow cytometry (Fig. 2A and Fig. S2A). In turn, amiRNAs targeting single FMDV sequences also exhibited significant silencing activity (Fig. 2B and Fig. S2B).

To evaluate the impact of dual amiRNA_FMDV expression on silencing activity in reporter assays, pre-amiRNA₂₉₀, pre-amiRNA₄₄₄ or pre-amiRNA₁₁₆₂ were subcloned in pamiRNA₁₀₅₅ to obtain bicistronic expression constructs and the resulting plasmids were co-transfected with pEGFP.3D in BHK-21 cells. As shown in Fig. 2C, dual expression of amiRNAs did not enhance silencing of the reporter mRNA.

Stable expression of amiRNAs against FMDV

Electroporation of DNA is a well-established method that yields a high frequency of transformants (Potter, 2003); however, transfection efficiency is not 100%. We hypothesized that electroporation of plasmids expressing small RNAs could lead to a population of non-transfected cells that, when infected, would contribute to FMDV replication. To circumvent this possibility, stably-transformed cell lines expressing amiRNA₂₉₀ and  amiRNA₁₀₅₅ were established and characterized in terms of antiviral efficacy. In this case, only amiRNAs were evaluated since they exhibit less toxicity than shRNAs (Boudreau, Monteys & Davidson, 2008).

Pre-amiRNA_FMDV expressing plasmids were stably transfected in BHK-21 cells and transgenic polyclonal cell lines were selected by blasticidin resistance. Mature amiRNA expression was confirmed by RT-stem loop qPCR (Fig. S4A). After several passages, cells were infected with FMDV A/Arg/01 and cultures were monitored to evaluate the development of cytopathic effect. At 48 hpi, control cell monolayers (non-transfected BHK-21 cells and cells stably transfected with a non-specific amiRNA) were completely detached due to FMDV infection. In contrast, infection of cell lines stably expressing amiRNA₂₉₀ and amiRNA₁₀₅₅ caused partial cytopathic effect, with a significant proportion of cells still adherent to the culture surface (Fig. 3A). This effect was also evidenced in lysis plaques assays, where FMDV infection rendered less plaques of smaller dimensions in cell lines expressing amiRNA₂₉₀ and amiRNA₁₀₅₅ than in control cells (Fig. 3B). However, viral titers determined at different times post-infection in supernatants of amiRNA₂₉₀ and amiRNA₁₀₅₅ polyclonal cells did not differ significantly from control cells except for amiRNA₂₉₀ cells at 18 hpi (Fig. 3C). Together, these results suggest a transient amiRNA-mediated constraint on FMDV replication in transgenic cell lines.

Next, polyclonal cell lines expressing amiRNA₂₉₀ and amiRNA₁₀₅₅ were cloned by limiting dilution to consider the potential variability in transgene expression associated with multiple transgene insertion events during the establishment and selection of stable transformants. The resulting cell lines were infected with FMDV A/Arg/01 and their ability to control viral growth was assessed by plaque assay (data not shown). Cloned cell lines showing less or smaller plaques than control cells (BHK-21 and miRneg-expressing cells) were selected for further analysis. Selected cell lines (n = 4 for amiRNA₂₉₀ and n = 3 for amiRNA₁₀₅₅) were infected with FMDV A/Arg/01 and viral growth was determined by viral titration in supernatants at different times post-infection. As shown in Fig. 4, viral growth was delayed in the majority of the cloned cell lines, supporting the potential use of amiRNAs as antivirals. However, at 48 hpi, the viral titers in supernatants of infected transgenic cells did not differ from control cells. Moreover, direct sequencing of a 250–300 nt region encompassing the target site did not show any mutation that could account for virus escape (Fig. S5).

Discussion

To develop an antiviral strategy against FMDV based on RNAi, the 3Dpol-coding sequence of FMDV A/Arg/01 was used to predict target sites. This region of the viral genome is highly conserved within FMDV serotypes. Moreover, the essential role played by the protein product (3Dpol) in the viral replication cycle makes it an attractive target for RNA interference. As shown in the reporter assays, the predicted targets were silenced efficiently independently of the RNAi mediator used, supporting the feasibility of the algorithm applied for target selection. Our results are in accordance with previous data indicating that the 3D region of the FMDV genome is a suitable target for RNAi (Pengyan et al., 2008; Du et al., 2011; Gu et al., 2014; Basagoudanavar et al., 2019) and they represent a step forward in the identification of new target sequences within FMDV, specifically in serotype A, that had been less explored in previous works.

It is well known that target accessibility may limit RNAi efficacy (Westerhout, 2005; Tafer et al., 2008; Low et al., 2012; Gismondi et al., 2014). In a paper on HIV inhibition by RNAi, Low et al. (2012) analyzed the impact of target RNA structure and target:siRNA duplex thermodynamics on shRNA efficacy. They found a maximal negative correlation for a 13-nt window that starts at position 14 of the target RNA and extends over 7 nucleotides downstream of the ‘seed region’ binding site. In the context of RNAi, a negative correlation implies that maximal shRNA inhibition is achieved for less structured target RNAs. Moreover, ΔG_total of target:siRNA duplex formation strongly correlated with shRNA inhibition. Thus, in this work RNA accessibility was used as a central criterion for target selection.

The available programs for prediction of RNAi targets can be classified as first- and second-generation algorithms. The latter are based not only on a number of selection criteria but also they were optimized to include results from real RNAi experiments to validate their predictions (Griger & Tisminetzky, 1984; Devaney et al., 1988; Filhol et al., 2012). In the present work we used a first-generation RNAi target selection program but we also included the experimental data derived from the work by Low et al. (2012) to predict RNAi targets. Interestingly, none of the selected regions had been targeted previously in other similar studies with other FMDV serotypes (Chen et al., 2004; Kahana et al., 2004; Mohapatra et al., 2005; Du et al., 2011). Furthermore, our analysis with RNAxs did not retrieve the targets selected by other authors within the 15 best ranking sequences or it did even not predict them at all. This may indicate that, despite sequence conservation between FMDV serotypes in the 3D region, accessibility of specific target sequences may not be equivalent between serotypes due to structural constraints in the context of the whole genome, as we previously reported.

Indeed, conservation of the target sequence among different viral strains should be considered during the development of RNAi-based antiviral tools (Mcintyre et al., 2009). In this work, selected sequences 290, 444 and 1055 showed partial conservation among South American serotypes; however, the sequences used for small RNA design were the most frequent at polymorphic sites (Fig. S1C). Interestingly, other stretches within 3D showing 100% conservation were not predicted in the top-ranking target sites by RNAxs software, suggesting these sequences may not be accessible. Initial reports on determinants of miRNA:mRNA recognition highlighted the existence of two categories of target sites. 5′  dominant sites, which are strongly complementary to the miRNA 5′  end, need little or no pairing of the miRNA 3′  end to achieve silencing. On the other hand, miRNAs with partial complementarity at the 5′  end require strong 3′  compensatory pairing for function (Brennecke et al., 2005; Bartel, 2009). Thus, the existence of polymorphisms in the target sequence should not preclude miRNA efficacy against different FMDV serotypes. Moreover, potentially reduced RNAi due to weak amiRNA:target RNA pairing at polymorphic sites could be circumvented by co-expression of multiple amiRNAs targeting different target sequences of the viral genome.

Interestingly, the shRNA directed against sequence 1162 was effective in the reporter assay and the amiRNA trended to perform in a similar way, although with no statistical significance. Contrarily, shRNA₁₁₆₂ did not affect FMDV replication significantly, whereas amiRNA₁₁₆₂ showed moderate antiviral activity. This target region was selected as a putative positive control, since the same sequence of FMDV serotype O/CHA/99 (Fig. S3) had been efficiently silenced by an shRNA in a previous paper (Gu et al., 2014). Noteworthy, as mentioned before, this region was not predicted as a suitable RNAi target by RNAxs software. It may be argued that shRNA₁₁₆₂ did not work due to structural constraints of the target region in the context of the whole viral genome, which were not present in the reporter mRNA. However, the results obtained with amiRNA₁₁₆₂ indicate that target site accessibility should not be limiting shRNA₁₁₆₂ efficacy. Alternatively, it is possible that this target region is not completely accessible to the mature small RNAs, leaving only some nucleotides available for small RNA:target recognition. In this case, pairing of the seed region of amiR₁₁₆₂ would be sufficient for activation of the RNAi machinery, whereas incomplete pairing between shRNA₁₁₆₂ and its target RNA would not lead to RNA silencing. In addition, since it has been shown that siRNA derived from shRNA precursors are more heterogeneous in cleavage sites and length than small RNAs originating from miRNA precursors (Maczuga et al., 2013), differences in shRNA and pre-amiRNA structures and processing that may lead to slightly different mature small RNAs with singular silencing efficacy cannot be ruled out.

Short hairpin RNAs are highly transcribed from RNA polymerase III promoters such as the U6 promoter used in this study. In addition, it has been demonstrated that shRNAs are more potent than amiRNAs (Boudreau, Monteys & Davidson, 2008) when they are expressed from the same promoter. However, shRNAs exhibit higher cellular toxicity than amiRNAs since the high levels of shRNA precursors produced in a cell may saturate the endogenous miRNA processing machinery (Boudreau, Martins & Davidson, 2009). In our reporter assays, shRNAs and amiRNAs were transcribed from different promoters (U6 and CMV, for RNA polymerases III and II, respectively) and we did not detect any difference in cell viability between cells transfected with pshRNA_FMDV and cells transfected with pamiRNA_FMDV as determined by flow cytometry.

Our results show that amiRNAs induce a significant inhibition of viral replication in cultured cells; however, silencing of FMDV RNA is not complete leading to sustained viral replication. Notably, amiRNAs were poorly expressed in the majority of cloned cell lines as compared to polyclonal cells (Fig. S4B). Thus, the lack of sustained viral inhibition may be associated with insufficient amiRNA expression. However, a ∼24 h delay in viral replication has been also reported by De Los Santos et al. (2005), Chen et al. (2004) and Lv et al. (2009) using shRNAs or siRNAs. Conversely, Kahana et al. (2004) showed a sustained inhibition of FMDV infection in BHK-21 cells transfected with siRNAs against FMDV; however, the authors only analyzed viral replication for 24 h. In turn, by using an siRNA transfected in BHK-21 cells, Mohapatra et al. (2005) reported a reduction of viral titer of >99% at 24 hpi and >87% at 48 hpi. The transitory character of viral inhibition may be explained by several factors. First of all, it should be kept in mind that in the infected cell the FMDV RNA is associated with multiple cellular and viral proteins during the processes of genome translation and replication. Thus, target sequences should not be expected to be continuously accessible for the silencing complexes during the whole FMDV replication cycle. In addition, it is well known that FMDV proteases induce a rapid shut-off of cellular transcription and cap-dependent translation (Kim et al., 2008; Qureshi, Thakur & Kumar, 2013). Given that amiRNAs are expressed from the strong CMV promoter, it is expected that pre-amiRNA transcription is also inhibited in infected cells in the context of viral replication. Thus, the incomplete inhibition of FMDV replication by amiRNAs may be explained by an FMDV-mediated reduction in transgene expression.

Other authors have demonstrated that simultaneous expression of multiple small RNAs directed against different regions of the same target RNA can increase silencing efficacy (Kahana et al., 2004; Brake et al., 2008; Liu et al., 2008; Kim et al., 2010; Saha et al., 2016). This approach becomes particularly relevant when the target RNA is a viral genome, since multiple small RNAs acting on the same genome reduce the chance of emergence of escape mutants thus increasing safety. However, in this work dual amiRNA expression did not enhance silencing of the reporter mRNA or the complete FMDV genome as compared to individual amiRNA expression. Similar results have been reported by other authors using amiRNAs against FMDV (Basagoudanavar et al., 2019). Noteworthy, amiRNA expression did not differ significantly in cells stably transfected with monocistronic or bicistronic vectors (Fig. S4A). Further research is needed to determine whether it is due to structural constraints of the target RNA or to the intrinsic regulation of the RNAi machinery.

Conclusions

In light of the results obtained, we conclude that the selection algorithm proved to be highly accurate in the identification of target sequences. FMDV replication may be transiently controlled by amiRNAs targeting 3D. Ongoing work in our laboratory is aimed at finding new targets in other genomic regions of FMDV to develop additional amiRNAs that could potentiate the antiviral effects observed in the present work. Our results highlight the potential application of amiRNAs as antiviral tools which may complement vaccination, specifically to prevent the spread of the disease during outbreaks in endemic or non-endemic countries.

Supplemental Information