lab-ledEngineering microRNA response elements for a mammalian-ubiquitous microRNA directly into the influenza A virus nucleoprotein coding sequence attenuates the virus in mice while leaving growth in embryonated chicken eggs intact, yielding live attenuated vaccine candidates for H1N1 and H5N1.
Jasmine T Perez; Alissa M Pham; Maria H Lorini; Mark A Chua; John Steel; Benjamin R tenOever
2009 · Nature Biotechnology · primary research
- Senior authors
- Benjamin R tenOever
- Correspondence
- Benjamin R tenOever
Research areas & themes
Citation
Perez JT, Pham AM, Lorini MH, Chua MA, Steel J, tenOever BR. MicroRNA-mediated species-specific attenuation of influenza A virus. Nature Biotechnology. 2009. Volume 27, issue 6, pages 572-576.
DOI 10.1038/nbt.1542. PMID 19483680.
One-sentence contribution
Engineering microRNA response elements for a mammalian-ubiquitous microRNA directly into the influenza A virus nucleoprotein coding sequence attenuates the virus in mice while leaving growth in embryonated chicken eggs intact, yielding live attenuated vaccine candidates for H1N1 and H5N1.
Executive summary
Live attenuated influenza vaccines in use at the time relied on temperature sensitivity for attenuation, a single mechanism with a defined set of excluded recipient groups. The authors asked whether the host microRNA silencing machinery could provide an orthogonal and tunable attenuation mechanism for influenza A virus, one that restricts replication in the vaccinated mammal without restricting yield in the egg substrate used for manufacture. They first established that influenza A virus infection, and the viral NS1 protein on its own, do not block microRNA processing or microRNA-directed silencing in mammalian cells. Because the influenza mRNA has almost no 3' untranslated region available for insertion, response elements were built into the open reading frame of segment five, which encodes nucleoprotein, at two positions where nucleotide changes preserve the hierarchical class of the encoded amino acids. The targeting microRNA, miR-93, was selected because published small RNA profiles show it in mouse and human but not in chicken. A control virus carrying the same three amino acid substitutions with the microRNA pairing disrupted separated silencing effects from protein effects. Viruses carrying both response elements lost more than two logs of lethality in mice, replicated normally in eggs, and replicated normally in Dicer-deficient fibroblasts. Vaccination protected mice against lethal homologous challenge with a broad immunoglobulin response. The approach offers an attenuation mechanism that can be layered onto existing strategies and tuned by the choice and number of elements.
Scientific context
Live attenuated influenza vaccines such as FluMist are temperature-sensitive reassortants, and the annual reformulation cycle places a premium on attenuation strategies that transfer easily to new strains. MicroRNA-mediated restriction of viral replication had been demonstrated for lentiviruses, picornaviruses and rhabdoviruses by other laboratories, in each case by placing microRNA target sequences into viral RNA. Extending that idea to influenza A virus faced two specific obstacles. Influenza mRNAs terminate shortly after the stop codon, so the untranslated region used in other systems is essentially unavailable and disruption of that region can cause packaging and replication defects. In addition, it was not settled in the literature whether influenza infection, and particularly the RNA-binding protein NS1, leaves the cellular microRNA pathway functional. Most microRNAs are conserved across vertebrates, but a minority are species restricted, which raises the possibility of an attenuation phenotype that depends on the host species.
Central question
Can microRNA response elements placed inside an essential influenza A virus open reading frame attenuate the virus in mammals while preserving replication in the avian system used for vaccine production, and does the resulting virus function as a protective live attenuated vaccine?
Experimental strategy
The design has three linked parts. First, a feasibility test in cell culture asking whether influenza infection or NS1 expression perturbs microRNA biogenesis or silencing, assayed by northern blot for precursor and mature forms of an exogenous tissue-restricted microRNA and an exogenous ubiquitous microRNA, and by a luciferase reporter carrying target sites in its 3' untranslated region. Second, a comparative genomics step to choose a microRNA present in mouse and human lung but absent from chicken, using published deep sequencing profiles and confirming mammalian expression by northern blot, RT-PCR and in situ data for the Mcm7 intron that hosts miR-93. Third, engineering and phenotyping. Response elements were introduced by site-directed mutagenesis into two positions of the nucleoprotein coding sequence chosen for amino acid conservation combined with nucleotide plasticity, viruses were rescued by plasmid-based reverse genetics, and the resulting strains were compared with a parental control carrying the identical amino acid substitutions but disrupted microRNA pairing. Attenuation was then attributed to the silencing pathway by three independent controls, replication in Dicer-deficient fibroblasts, rescue of replication by locked nucleic acid inhibition of miR-93, and growth in eggs where the microRNA is absent. Finally the candidate viruses were tested as vaccines in mice, including an H5N1 six-plus-two reassortant, with weight loss, survival after lethal challenge, and serology as readouts.
Key findings
-
Influenza A virus infection did not disrupt the microRNA pathway. Exogenous miR-93 and miR-124 were processed to precursor and mature forms in infected HEK293 cells, endogenous miR-93 levels were unchanged over a 36 hour infection time course in human and murine fibroblasts, and miR-124 suppressed a target reporter by about 90 percent whether or not cells were infected or expressed NS1 (Figure 1a-d). The authors read this as establishing that microRNA-based attenuation of influenza is feasible in mammalian cells.
-
Two positions in the nucleoprotein coding sequence, at nucleotides 225 and 818 of A/Puerto Rico/8/34, could be converted into near-perfect miR-93 binding sites with predicted free energies of about minus 28 and minus 37.1 kcal per mole, at the cost of three amino acid substitutions, I63L, S262T and I265L, each preserving the side chain class observed in natural variation at those loci (Figure 2a).
-
The three substitutions were largely tolerated by the protein. In a polymerase reconstitution reporter assay the parental control nucleoprotein retained most activity, with an approximately 20 percent reduction in RNA-dependent RNA polymerase output relative to wild type (Figure 2b). The parental virus behaved like wild type in cell culture but showed some attenuation in vivo, which the authors attribute to that reduced polymerase activity being more consequential under in vivo selective pressure. This attribution is an interpretation and is not separately tested.
-
The doubly targeted virus was attenuated in mice by more than two logs. Median lethal doses were 1.7 x 10^3 plaque forming units for the parental control and 2.15 x 10^5 for the response-element virus (Figure 2d).
-
Attenuation was species restricted. Wild type, parental and all response-element viruses, in both H1N1 and H5N1 six-plus-two configurations, grew to roughly 10^8 plaque forming units per milliliter in ten-day embryonated chicken eggs, indicating no loss of yield in the manufacturing substrate (Figure 2e,f).
-
Attenuation required the silencing machinery and the specific microRNA. In wild type murine fibroblasts the singly targeted viruses showed mild reduction in hemagglutinin and the doubly targeted virus produced none, whereas all strains replicated to similar high levels in Dicer-deficient fibroblasts (Figure 3a). Pretreatment with a locked nucleic acid inhibitor of miR-93 more than doubled doubly targeted virus protein relative to scrambled control (Figure 3b).
-
The restriction acted on translation rather than on transcript abundance. In infected cells, nucleoprotein mRNA was elevated while nucleoprotein protein was very low, a pattern also seen in vivo at 48 hours (Figure 3c,d and Supplementary Figure 5). The authors interpret the elevated mRNA as reflecting the role of unbound nucleoprotein in the switch from transcription to replication, which is a proposal consistent with the data rather than a tested mechanism here.
-
No escape mutants were recovered. Sequencing of nucleoprotein after ten serial passages in A549 cells and after in vivo infection, from more than 25 colonies per cohort, did not identify revertants. The authors read this as suggesting that the flexibility of microRNA targeting combined with the conservation constraint on nucleoprotein limits reversion. Absence of recovered escape over this passage regime does not establish that escape cannot arise.
-
The targeted viruses protected mice. Intranasal vaccination with 10^3 plaque forming units of the H1N1 response-element virus caused less weight loss than the parental control and gave 100 percent survival after lethal challenge, with neutralizing serum and IgM, IgG1, IgG2a and IgG2b responses (Figure 4a). For the H5N1 reassortant, the parental control killed half the animals whereas all animals receiving the response-element virus survived vaccination and subsequent lethal H5N1 challenge with no signs of morbidity, while mock-vaccinated animals all died (Figure 4b).
-
Tropism and the innate response were not obviously altered. Additional in vivo characterization reported no change in viral tropism and normal cytokine and transcriptional responses to infection (Supplementary Figure 6).
Mechanistic model
The data support a model in which mature miR-93, present in mouse and human cells and absent in chicken, loads into the silencing complex and engages the two engineered sites within the nucleoprotein open reading frame, suppressing nucleoprotein translation. Because nucleoprotein is required for genome encapsidation and for the transcription to replication switch, loss of the protein caps the amount of infectious progeny and limits pathogenesis in the mammalian host, while the same genome replicates without restriction in the avian egg and in Dicer-deficient cells.
The translational rather than degradative character of the restriction is directly supported by the divergence between elevated nucleoprotein mRNA and depleted nucleoprotein protein. The specificity of the effect for miR-93 and for the silencing pathway is directly supported by the Dicer-deficient and antimiR experiments. What the study does not establish is the downstream consequence of nucleoprotein limitation in molecular terms. The proposal that elevated nucleoprotein mRNA reflects a biased transcription to replication switch is an interpretation drawn from prior work on nucleoprotein function, not a mechanism demonstrated here. The study also does not resolve which cells in the mouse lung are responsible for the in vivo attenuation, nor whether the residual in vivo attenuation of the parental control shares any mechanism with the microRNA-dependent effect.
Conceptual or technical advance
The work makes an attenuation mechanism available that is genetically encoded, orthogonal to temperature sensitivity, and tunable by the number of response elements and the choice of targeting microRNA. Placing the elements inside a conserved open reading frame rather than in an untranslated region solves the specific constraint imposed by influenza mRNA architecture and, by coupling the target site to codons under amino acid conservation, couples escape to a fitness cost. Because the response elements sit on the nucleoprotein segment, the same engineered segment can be moved into reassortants with different surface antigens, which the authors demonstrate by rescuing an H5N1 six-plus-two virus. The result also makes the species distribution of microRNAs a usable engineering parameter, allowing the attenuating signal to be present in the vaccinee and absent in the production substrate.
Relationship to the broader research program
The study sits at the intersection of two recurring interests in this corpus, the interaction between RNA viruses and host small RNA pathways, and the deliberate engineering of viruses as tools and vaccines. The demonstration that influenza infection and NS1 leave microRNA biogenesis and silencing intact in mammalian cells is a foundational observation for later work in the corpus that uses virus-encoded or host microRNAs as instruments. The use of a targeted viral genome as a reagent whose tropism is set by the small RNA content of the cell recurs as a design idea.
Category 3 synthesis is not attempted here beyond that, because this record was built from one paper read in isolation.
- tenOever BR and colleagues, 2007, Science, on multiple functions of IKKepsilon in interferon-mediated antiviral immunity. Methodological foundation. Cited in this paper as the source of the RT-PCR and immunoblot procedures used.
- Makeyev, Zhang, Carrasco and Maniatis, 2007, on miR-124 and neuronal differentiation. Methodological foundation from another laboratory, source of the miR-124 reporter constructs and the miR-124 minigene used in the feasibility experiments.
- Barnes, Kunitomi, Vignuzzi, Saksela and Andino, 2008, and Kelly, Hadac, Greiner and Russell, 2008, and Edge and colleagues, 2008. Predecessor work from other laboratories applying microRNA targeting to picornaviruses, other viruses and vesicular stomatitis virus, which this study extends to a segmented negative-strand virus and to an open reading frame insertion site.
Limitations and boundaries
The attenuation and protection data are in BALB/c mice with intranasal inoculation, and no larger animal, ferret or human data are presented. The H1N1 component is based on A/Puerto Rico/8/34, a laboratory-adapted strain, and the authors state plainly that the H1N1 vaccine described does not protect against strains then in circulation. Challenge experiments are homologous or antigenically matched, so nothing here speaks to heterologous protection or to durability beyond the three-week interval tested. Cohorts are small, with three to four mice per dose in the toxicity studies and four per group in the vaccination studies. The species specificity rests on published microRNA profiles plus the egg and Dicer-deficient controls, and miR-93 expression was not surveyed across all relevant avian or mammalian tissues. Absence of escape mutants is bounded by ten serial passages in A549 cells and by sampling of more than 25 clones per in vivo cohort, which does not exclude rarer escape or escape under different selective regimes. Because MDCK cells express miR-93, plaque assays for the targeted viruses were limited, which constrains direct titre comparison for those strains. The translational repression conclusion rests on mRNA and protein measurements rather than on direct ribosome-level assay.
Audience summaries
25 words
Influenza engineered to carry binding sites for a microRNA found in mammals but not chickens is disabled in mice yet grows normally in eggs, producing protective vaccine candidates.
75 words
Live influenza vaccines are grown in chicken eggs but must be weakened in people. The authors exploited a small regulatory RNA, miR-93, that mammals make and chickens do not. Placing binding sites for it inside the nucleoprotein gene left egg growth untouched while cutting lethality in mice by more than a hundredfold. The effect disappeared when the silencing machinery was removed. Vaccinated mice survived lethal H1N1 and H5N1 challenge with broad antibody responses.
150 words
Influenza mRNAs offer almost no untranslated region for inserting regulatory sequence, so the authors engineered near-perfect miR-93 response elements directly into two sites of the nucleoprotein open reading frame, choosing positions where nucleotide changes preserve the class of the encoded amino acid. miR-93 is expressed in mouse and human but not chicken, giving an attenuation signal present in the vaccinee and absent in the egg substrate. A parental virus with the identical amino acid substitutions and disrupted pairing isolated the silencing effect. The doubly targeted virus showed a greater than two log increase in median lethal dose in mice, grew to normal titres in embryonated eggs, and replicated unimpeded in Dicer-deficient fibroblasts. Nucleoprotein mRNA accumulated while protein did not, indicating translational repression. Vaccination gave complete protection against lethal homologous H1N1 and H5N1 challenge. The strategy is orthogonal to temperature-sensitive attenuation and tunable through element number and microRNA choice.
Discoveries supported by this paper
Discoverylab-led for Perez 2009, Varble 2010, Langlois 2012 in PNAS, Langlois 2012 in Molecular Therapy, Pham 2012, Varble 2013, Schmid 2014, Benitez 2015 on in vivo screening, Møller 2018 and the tenOever 2019 review, and co-led for Langlois 2013, whose ferret transmission work sits with the Perez and García-Sastre groups. Daniloski 2021 in Cell is co-led with the Sanjana laboratory
Documented publication relationships
Pathogens
Technologies