The inclusion of a single copy of target sequences for miRNAs expressed in the CNS (mir-9, mir-124a, mir-128a, and mir-218) completely abolished the neurotropism of the TBEV/DEN4 virus in immunocompetent adult mice, but the effects of these miRNA target insertions on the viral neurovirulence in newborn mice and neuroinvasiveness in immunodeficient mice were less evident than those observed for the let-7c target (15
). let-7c is a member of the large let-7 family of miRNAs which are ubiquitously expressed in a multitude of tissues and cells (2
), and the let-7c target inclusion into viral genome was most effective in terms of limiting TBEV/DEN4 neuropathogenesis in both newborn and adult mice. These findings suggest that a variety of factors including immune system maturation and levels and patterns of miRNA expression in the CNS and/or other tissues may play a role in age-dependent susceptibility of mice to miRNA-targeted virus infection. Sequence analysis data of viral isolates derived from the brains of moribund SCID (15
) or suckling mice, which were infected with TBEV/DEN4 viruses carrying a single miRNA target, show that virus escape from the miRNA-mediated suppression can occur through deletions or mutations within the central region of the miRNA target sequence. These findings are in line with data obtained for other viruses (4
) indicating that the perfect complementarity between the target sequence and cellular miRNA is a critical factor for miRNA target recognition and catalytic cleavage of miRNA-targeted RNA by the RISC (6
). As we demonstrated previously (15
), a single-nucleotide mutation within the central region of the let-7c target sequence was sufficient to restore the ability of the let-7cT mutant virus to cause lethal CNS disease in mice and to efficiently replicate in Vero cells or primary neurons expressing the let-7c miRNA.
Virus escape from the miRNA-mediated suppression in the developing mouse CNS was progressively reduced by increasing the number of miRNA target sites in the 3′NCR of the viral genome. The simultaneous tandem targeting of the TBEV/DEN4 genome for two different brain-expressed miRNAs (mir-9 and mir-124a) was more efficient in reducing virus neurovirulence than targeting of two or three tandem targets for mir-124a miRNA alone. In addition, we found that genomic targeting with a tandem repeat of three mir-124a targets (at sites 1, 2, and 3) was less effective in preventing the development of lethal encephalitis than targeting with the same number of mir-124a binding sites from which a third mir-124a target was inserted at a more distant position (at site 4 or 5).
Multiple miRNA target insertions (2, 3, or 4 copies) into the 3′NCR altered virus neuroinvasiveness in SCID mice and significantly attenuated its neurovirulence in immunocompetent mice without a negative impact on the immunogenicity and protective efficacy. Importantly, the miRNA targeting of a large portion of the 3′NCR with three or four miRNA binding sites greatly reduced the probability of virus to escape from the miRNA-mediated suppression even in the developing mouse CNS. However, the suppression of virus replication in the brain was not complete. Analysis of virus escape mutants derived from the brains of moribund mice showed that emerging viruses can overcome miRNA-mediated suppression by acquiring large deletions that included the inserted miRNA target sequences as well as the portion of viral genome located between the two most distant miRNA targets. It should be noted that this mode of virus escape in the CNS was a common trend among all viruses carrying two or more miRNA targets. This observation raises intriguing questions about the potential mechanism(s) involved in the generation of a large deletion in the viral RNA genome. Since each introduced miRNA target was fully complementary to the corresponding miRNA, the viral RNA should be cleaved by Argonaute-2 of the RISC at the miRNA binding sites (23
). These cleavages would release 3′ and 5′ RNA fragments as well as the genomic RNA fragments located between a pair of miRNA targets that could serve as the RNA template segments for the subsequent formation of novel RNA genomes. Most likely, the mechanism involves the covalent joining of two noncontiguous RNA genome portions (the 5′ RNA fragment containing the sequences of the 5′NCR and polyprotein ORF and the 3′ RNA fragment containing the 3′-terminal part of the 3′NCR) by way of a viral polymerase-directed template-switching mechanism similar to that proposed for the generation of viral defective interfering RNAs (19
). The precise mechanism of template switching by the RNA-dependent RNA polymerase (RdRp) is not well-known. The current model suggests that the viral RdRp together with the nascent RNA strand, which had been synthesized on the primary RNA template (the 3′ RNA fragment), jumps to the 5′ RNA fragment where it resumes RNA synthesis. Interestingly, among the nucleotide sequences of the 29 escape mutants analyzed in this study (see Fig. S3 in the supplemental material), the junction sites did not show nucleotide similarity but contained A/U-rich stretches, which are believed to promote RNA recombination by template switching (26
). For each miRNA-targeted virus, the deleted 3′NCR sequences from virus isolates varied in size and localization and were always associated with and included the miRNA target elements (; see also Fig. S3 in the supplemental material). In the eight progenies of TBEV/DEN4 viruses bearing two or three tandem miRNA targets, deletions extended beyond the last miRNA target to the G/C-rich sequences of the 3′NCR (e.g., to the nucleotide 10370, 10374, 10379, 10381, 10382, 10433, 10471, or 10529; ; see also Fig. S3) which can form hairpin structures (31
) and promote viral RNA recombination (26
). The presence of these large deletions in the viral escape mutants suggests that the cellular exo- and endoribonucleases may contribute to the further degradation of the initially RISC-generated RNA fragments prior to recombination.
The 3′NCR of the flavivirus genome is required for the formation of subgenomic RNAs (27
) and is comprised of several conserved RNA structures that can modulate virus replication and pathogenicity (13
). Conserved sequence elements such as a tandem of the repeated conserved sequence 2 (RCS2) and conserved sequence 2 (CS2) in the 3′NCR are important for virus replication in mammalian cells. Indeed, we found that a 270-nt deletion between the stop codon and the last 3′-terminal 114 nt of the 3′NCR engineered into the recombinant TBEV/DEN4 Δ270 cDNA, including the RCS2 and CS2 sequences, was lethal, precluding virus recovery in simian Vero cells. Taking this into account, the recovery of brain-derived 3x mir-124aT(1,2,5) and 4x mir-124aT(1,2,3,5) escape mutants with large deletions in the 5′ part of the 3′NCR, including the conserved RCS2 and CS2 structures ( and ), was unexpected. The 3′NCRs of five escape mutant viruses, one for 4x mir-124aT(1,2,3,5) and four for 3x mir-124aT(1,2,5), varied in length (127, 135, 137, 138, or 139 nt; see Fig. S3 in the supplemental material) but were always longer than the 3′NCR engineered in the TBEV/DEN4 Δ270 cDNA (120 nt) and retained the 110 proximal 3′-terminal nucleotides of the parental TBEV/DEN4 genome (nt 10555 to 10664) (). The spacer sequences between the stop codon and these last 3′-terminal 110 nt ranged from 17 to 29 nt in the length and were the remains of the sequence insertions at sites 1 and 5. Thus, the 5′ part of the 3′NCR between the stop codon and nucleotide 10555 could be substituted with short external sequences without loss of virus viability. These findings suggest that the size rather than the primary nucleotide sequence of the spacer region can affect the yield and replication of infectious virus. However, we also identified another unique mutation in the structural envelope E protein (a Gly443
→ Asp substitution) of the Δ367 escape virus carrying a large deletion in the 3′NCR. This mutation occurred in an α-helix H2 element of E protein that is important for the prM-E dimer stability during virion assembly (1
). The exact mechanism by which the selected escape mutant was able to regain the replicative capability and revert to a neurovirulent phenotype remains to be determined. In the absence of a Gly443
→ Asp mutation in the E protein, the parental TBEV/DEN4 virus, carrying the 3′NCR sequence derived from the escape Δ367 virus, replicated very slowly in Vero cells and reached a titer 104
-fold lower than that attained by parental TBEV/DEN4 or Δ367 virus. Thus, these findings suggest that a point mutation (Gly443
→ Asp) in E protein may contribute to efficient replication of the Δ367 virus in Vero cells and possibly, in the mouse brain.
Taken together, our findings with multiple miRNA targeting of neurotropic flavivirus demonstrate that the degree of virus attenuation for the CNS can be modulated by increasing the number of miRNA target sites and targeting a large portion of the viral genome. Insertion of four target copies for the CNS-specific mir-124a in the 3′NCR [4x mir-124aT(1,2,3,5) virus] was the most effective approach for restricting viral replication, reducing inflammatory responses, and preventing neuronal damage even in the highly vulnerable developing CNS. With this approach, virus escape from miRNA-mediated suppression was very infrequent and observed only in the CNS of suckling mice. Sequence analysis of brain-derived progenies of the miRNA-targeted viruses has shown that virus escape occurs exclusively through the deletion of inserted miRNA targets and results in the loss of the viral genome sequence located between the two most distant miRNA targets. Importantly, deletions never extended upstream into the viral polyprotein ORF or downstream into the conserved 3′-terminal 110-nucleotide structure of the 3′NCR signifying the functionally important role of these regions for virus viability. These findings have prompted us to propose a rational vaccine design approach using multiple miRNA targets engineered into many different viral genome regions (5′NCR, ORF, and 3′NCR) that are crucial for virus viability in order to control virus neurotropism and prevent virus reversion to the neurovirulent phenotype. In such a design, the emergence of neurovirulent escape mutants under the miRNA-mediated suppression in the CNS would be unlikely due to the fact that multiple miRNA-induced cleavages would lead to deletions of critically important virus sequences located between inserted miRNA binding sites.