Synthetic Virology: Building Viruses to Better Understand Them

lab-led

Frames influenza A virus as a compact genetic circuit whose modules can be disrupted, tagged or replaced, and organizes two decades of virus engineering into a design vocabulary of tracking, override, positioning and silencing modules that turns virus construction into a method of inquiry.

Benjamin R. tenOever

2019 · Cold Spring Harbor Perspectives in Medicine · review

Senior authors

Benjamin R. tenOever

Correspondence

Benjamin R. tenOever

Research areas & themes

Citation

tenOever BR. Synthetic Virology. Building Viruses to Better Understand Them. Cold Spring Harbor Perspectives in Medicine, 2019, volume 10, issue 11, article a038703.

DOI 10.1101/cshperspect.a038703. PMID 31871242. PMCID PMC7605229.

One-sentence contribution

Frames influenza A virus as a compact genetic circuit whose modules can be disrupted, tagged or replaced, and organizes two decades of virus engineering into a design vocabulary of tracking, override, positioning and silencing modules that turns virus construction into a method of inquiry.

Executive summary

Influenza A virus carries eight negative-sense RNA segments encoding roughly ten major products, and the review treats this compactness as an invitation to read the virus as a genetic circuit with discrete functional modules for entry, nuclear import, transcription, the switch to replication, nuclear export and egress. The organizing argument is that understanding a circuit of this kind comes from rebuilding it, an approach the article labels learning by building. The review traces the enabling technology from minireplicon systems that defined the RNA sequences the viral polymerase recognizes, through complete reverse genetics systems that permit rescue of virus entirely from DNA, to the systematic mapping of packaging signals and of mutational tolerance across viral proteins. It then organizes the applications as engineering modules. A tracking module covers fluorescent and luciferase reporter viruses and the persistent problem that most insertions attenuate. A browser history module uses virus-delivered Cre recombinase with LoxP reporter mice to mark cells that survived infection, which identified club cells as a population that clears the virus. A genetic override module covers microRNA target site insertion and drug-controlled degron tags as kill switches and biocontainment devices. A positioning module uses engineered intergenic space in segment 8 to carry RNA barcodes, which established that aerosol transmission passes only two to three virions while contact transmission passes much more. An RNA interference module has the virus deliver artificial microRNAs in an in vivo screen. Throughout, the review distinguishes designs that attenuate from the few that leave fitness unchanged.

Scientific context

Reverse genetics for positive-strand RNA viruses came early, since their genomes launch without accompanying viral proteins. Negative-strand RNA viruses were harder, because rescue requires an RNA with precise termini correctly bound to nucleoprotein in the presence of its cognate polymerase. The review credits the intermediate step to minireplicon systems, DNA-based platforms using small genomic mimics together with nucleoprotein to study polymerase biology, work it attributes to Hsu, Luytjes, Enami and colleagues in the Palese laboratory and to the broader field summarized by Neumann and Kawaoka. Complete rescue of influenza A virus from cloned DNA followed, credited to Fodor and colleagues and to Neumann and colleagues, and the review notes that the same laboratories that built the minireplicon and reverse genetics systems were the first to apply them to influenza RNA biology. What remained unresolved at the start of that period was how much of the viral genome is available for modification, which is the practical question the review treats as prior to every application that follows.

Central question

If influenza A virus is treated as a genetic circuit, which of its modules can be disrupted, tagged, replaced or added to without destroying the circuit, and what can be learned about virus and host biology by building the resulting recombinants and observing them.

Experimental strategy

Not applicable in the experimental sense. This is a single-author review and presents no new data. Its organizing strategy is a design taxonomy. Rather than surveying the literature chronologically or by virus, the article sorts engineered influenza viruses by the function the added element performs within the circuit, borrowing vocabulary from engineering. The taxonomy does real work, because it separates the questions that each class of design can answer. A tracking module reports where the virus is now. A lineage-marking module reports where it has been, including in cells that survived. An override module controls whether the program runs at all, in a host-specific or drug-dependent way. A barcode module reports the identity and number of founding virions without altering their behavior. A silencing module turns the virus into a delivery vehicle for a perturbation of the host. Against each class the review applies a common criterion, whether the design preserves viral fitness, since an attenuated recombinant reports on a different virus than the one under study.

Key findings

This is a review. Entries below identify the laboratory that produced each result, and mark claims that belong to the review's own argument rather than to any cited study.

1. Reported by many laboratories. Packaging signals for each of the eight segments extend beyond the terminal promoter elements and into the open reading frames, with a different length of required material for each segment. The review credits the initial demonstration that foreign RNA could be packaged to Luytjes and colleagues and to Fujii and colleagues, and the systematic mapping to Duhaut and Dimmock, Fujii, Liang, Marsh, Hutchinson, Essere and Gavazzi and colleagues. The idea that packaging is mediated by RNA to RNA interactions is described as a hypothesis, corroborated by more recent work from Dadonaite and colleagues and Majarian and colleagues. Gao and Palese used the same knowledge to swap packaging material between reading frames and build a virus incapable of reassortment with natural influenza.

2. Reported by the Palese laboratory. Genome-wide insertional mutagenesis using bacteriophage Mu transposase, leaving a 15 nucleotide insertion, found most amino acid insertions were not tolerated, with hemagglutinin and NS1 as the exceptions, reported by Heaton and colleagues. The review interprets that exception as reflecting the flexibility those two proteins need in order to adapt to differing host immune defences.

3. Reported by other laboratories. Two complementary mutagenesis approaches refined the picture. Error-prone PCR on segments 4 and 8 mapped residues involved in receptor binding, receptor structure and NS1 antagonism, reported by Wu and colleagues. Codon-based library construction and deep mutational scanning of nucleoprotein and hemagglutinin was reported by Bloom and colleagues. The generalization drawn, that influenza A virus is a highly optimized circuit with little tolerance for random insertion and that the consensus strain generally represents the optimal composition for its host, is the review's synthesis of these datasets.

4. Reported by other laboratories. Affinity tags inserted at the RNA level allow purification of replication intermediates. A 25 nucleotide PP7 hairpin placed in the neuraminidase stalk region in the complementary RNA orientation allowed immunoprecipitation of complementary ribonucleoproteins with the PP7 coat protein, which supported structural and functional characterisation of that intermediate and found it organized as a filamentous double helix, reported by York and colleagues. A parallel protein-level approach generated eight Flag-tagged recombinants, successful for all major products except M1 and nucleoprotein, used to map a virus to host proteome interaction network, reported by Heaton and colleagues.

5. Reported by other laboratories. Early reporter viruses came at a fitness cost. NS1 truncated and fused to green fluorescent protein gives a reporter that is attenuated in any interferon-competent model, from Kittel and colleagues. Replacing the hemagglutinin and neuraminidase entry system with the single rhabdovirus glycoprotein freed a segment for green fluorescent protein and showed the circuit to be partly modular, from Watanabe and colleagues. Use of 2A ribosomal skipping to build polycistronic segments produced replication-competent but attenuated and unstable reporters in segment 8, from Manicassamy and colleagues, and a stable but less replicative neuraminidase-GFP virus from Li and colleagues.

6. Reported by other laboratories. The finding that considerable foreign material can be added to PB2, the largest segment, came from Dos Santos Afonso and colleagues, and led to a suite of split green fluorescent protein and luciferase reporters described as both stable and showing little to no attenuation, from Avilov, Heaton, Tran and Karlsson and colleagues. Serial passage in mice of the NS1 fusion design recovered a variant with improved stability that remained less pathogenic than wild type, from Fukuyama and colleagues. The review reads the recurrence of attenuation across these designs as evidence that the virus occupies an optimal fitness space in which any major change is detrimental.

7. Reported by the tenOever laboratory with a collaborating group, and by another laboratory independently. Delivering bacteriophage P1 Cre recombinase from PB2 was well tolerated, reducing the lethal dose by only one log, whereas expression from NS1 via a 2A site attenuated substantially, reported by Heaton and colleagues with tenOever as a senior author and by Reuther and colleagues in the Schwemmle laboratory. Combined with a LoxP tdTomato reporter mouse, this permits marking of cells that were infected at any point rather than only cells currently infected. Monitoring non-immune tdTomato-positive cells over the infection showed that although alveolar cells succumb, a subset of club cells clears the virus.

8. Reported by other laboratories, following that observation. Club cells that survive infection were characterized as inherently more resistant, with genetic changes altering their response to virus and to interferon for weeks after clearance, from Hamilton and colleagues, and the basis of the resistance was later attributed to unusually strong antiviral gene induction together with evasion of CD8-mediated clearance, from Chambers and colleagues and Fiege and colleagues.

9. Reported by the tenOever laboratory. MicroRNA-mediated targeting was the first described genetic override for influenza A virus, reported by Perez and colleagues. Two target sites for miR-93, which is absent from the chicken egg but ubiquitous in mammalian cells, placed in the nucleoprotein reading frame gave a virus unable to establish productive infection in mammalian cells while growing to wild-type titres in eggs. The same approach was extended to restrict replication to ferrets as a molecular biocontainment device and to restrict it in haematopoietic cells as an immunological tool, reported by Langlois and colleagues. Because no coding material is altered, these designs give stably engineered viruses with unchanged fitness, supported by Benitez and colleagues and Aguado and colleagues.

10. Reported by the tenOever laboratory. As an alternative to host-supplied silencing, influenza A virus can be engineered to express an artificial microRNA processed into a small interfering RNA against a structurally constrained and conserved part of its own genome, producing a self-inactivating virus, reported by Benitez and colleagues.

11. Reported by another laboratory. A drug-dependent override was built by attaching the small molecule-assisted shutoff tag, composed of the hepatitis C virus NS3 protease followed by part of NS4a, to the PA subunit, reported by Fay and colleagues using the tag described by Chung and colleagues. In the presence of asunaprevir, NS3 cleavage is blocked and the retained NS4a peptide directs degradation. The approach reduced titres by one to two logs and prevented disease in animals, and selective mutation of the target site was observed.

12. Reported by the tenOever laboratory. Destroying the natural 3 prime splice acceptor for NEP and duplicating that region downstream of the NS1 stop codon creates an intergenic region in segment 8 while retaining the splicing dynamics of the segment, and does so without loss of fitness, reported by Chua and colleagues. This space accommodates modules that act only at the RNA level.

13. Reported by the tenOever laboratory. Inserting 22 nucleotide barcodes with no function beyond identity produced a library of more than 100 tagged viruses at roughly equal proportion, which allowed population dynamics to be read out by deep sequencing at single-virion resolution, reported by Varble and colleagues. Applied to transmission, contact transmission transferred up to half the viral population whereas aerosol transmission produced founder populations of only two to three virions. The review notes independent support from reassortment work by Fonville and colleagues and Jacobs and colleagues, and related observations from Brooke and colleagues and Russell and colleagues that infections frequently fail to deliver all eight segments, so co-infection is often required.

14. Reported by the tenOever laboratory. The same intergenic space supports microRNA production from a replication-competent virus, reported by Varble and colleagues, and because hairpin processing depends on structure rather than sequence the stem can be rewritten to silence a chosen host factor, reported by Benitez and colleagues. A library in which each virus silenced a different host factor was followed over an in vivo infection, and the greatest enrichment was seen for viruses targeting Ddx58, Tlr7 and Ifih1, Irf1, Irf7 and Stat1, and Adar and Rnasel. The review presents that enrichment as suggestive rather than as a demonstration, and notes the module is constrained by kinetics, since the virus can silence targets only from the point of infection onward and cannot affect protein made earlier.

Mechanistic model

This article does not establish a mechanism and does not attempt one. Its argument is architectural rather than mechanistic. Influenza A virus is presented as a genetic circuit of interacting modules, with each viral product assigned to a stage of the cycle and NS1 assigned not to the circuit itself but to protecting it from host interference. The proposition that follows is methodological, that a circuit understood well enough can be reprogrammed to run additional modules concurrently with infection, and that the success or failure of a build is itself informative about the circuit.

Two general claims in the article are the author's interpretation of assembled results rather than demonstrated conclusions. The first is that influenza A virus occupies an optimal fitness space, inferred from the recurrence of attenuation across independent reporter designs and from the mutational tolerance datasets. The second is that the consensus sequence of a strain generally represents the optimal composition for the host it came from, inferred from deep mutational scanning rather than measured directly. Both should be read as the framing the review adopts and not as settled results.

Conceptual or technical advance

The article's contribution is organizational and pedagogical rather than empirical. By sorting engineered viruses according to the circuit function of the added element, it makes visible which questions each design class can answer and which it cannot, and it applies a consistent criterion of preserved fitness that separates designs usable for studying wild-type biology from those that necessarily study an attenuated variant. Several specific lessons are made portable in the process, including the observation that insertions into PB2 and into engineered intergenic space in segment 8 are better tolerated than insertions that alter coding material, and that modules acting purely at the RNA level avoid the fitness cost that protein-level modifications incur. The Cre and LoxP application illustrates a broader point, that engineering can convert a virus into a reagent for a host question, in this case the identity of cells that survive infection, which is not addressable by any assay that only detects cells currently infected. The review closes by extending the framing toward therapeutics, naming oncolytic use, biologics production and RNA editing as directions, with work attributed to Schmid, Pizzuto and Hamilton and colleagues.

Relationship to the broader research program

This article is the clearest statement in the corpus of the premise underlying much of the tenOever laboratory's work, that building a virus is a way of understanding it. Most of the designs the review credits to that laboratory were produced in service of a host or population question rather than as tools in themselves, including microRNA targeting to restrict replication to defined species or lineages, barcoding to measure transmission bottlenecks, and virus-delivered artificial microRNAs to screen host factors in vivo.

Category 3 synthesis, visible only across papers. The review's engineering premise is the direct continuation of the argument in the 2013 Nature Reviews Microbiology article, which established why the chordate microRNA machinery is available for this kind of exploitation, and the 2013 review is cited here for exactly that purpose. The in vivo RNAi screen result, where enrichment fell on pattern recognition receptors, interferon-pathway transcription factors and effectors including Adar, connects to the treatment of the interferon-stimulated gene set as a structured output in the 2007 Science report, though neither paper cites the other and the connection is synthesis rather than a claimed lineage.

Related publications

• tenOever 2013, Nature Reviews Microbiology, predecessor and cited here. It supplies the argument that chordate microRNAs are not engaged by RNA viruses and are therefore available for engineering.
• Perez et al. 2009, Nature Biotechnology, application from the tenOever laboratory, microRNA-mediated species-specific attenuation, cited as the first described genetic override for influenza A virus.
• Langlois et al. 2012, PNAS, and Langlois et al. 2013, Nature Biotechnology, applications from the tenOever laboratory on lineage-restricted and host-restricted targeting.
• Chua et al. 2013, Cell Reports, methodological foundation from the tenOever laboratory for the segment 8 intergenic design used by several later modules.
• Varble et al. 2010, PNAS, and Varble et al. 2014, Cell Host Microbe, methodological foundation and application from the tenOever laboratory for virus-encoded microRNAs and for barcoded transmission studies.
• Benitez et al. 2015, Cell Reports, two reports, application from the tenOever laboratory for self-targeting virus and for in vivo RNAi screening.
• Heaton et al. 2014, Journal of Experimental Medicine, application with tenOever as a senior author, Cre-expressing virus and club cell survival.
• Aguado et al. 2018, PNAS, from the tenOever laboratory, cited in support of stable engineered designs with unchanged fitness.
• Neumann et al. 1999, PNAS, and Fodor et al. 1999, Journal of Virology, methodological foundation from other laboratories for influenza reverse genetics.

Limitations and boundaries

The article is a review with no new data, so nothing in it is demonstrated here. Its scope is almost entirely influenza A virus, and the circuit framing is presented for a segmented negative-strand virus with roughly ten products, so it does not transfer automatically to other families. Several designs the review describes work only in heavily laboratory-adapted strains, a qualification the article states explicitly for the segment 8 intergenic modification. Many reporter designs carry attenuation, so results obtained with them describe a modified virus, and the review is careful to record which designs preserve fitness and which do not. The transposon mutagenesis approach is noted to select at the level of protein subdomain and to favour surface loops, and the review states that mapping the relative importance of every residue in every gene would require further work. The RNA interference module is limited by kinetics and can only act on transcripts made after infection begins, which restricts it to identifying virus-induced genes that reduce replication. The in vivo RNAi screen result is described as suggesting rather than establishing the role of the enriched targets. The drug-controlled override showed escape by mutation of the target site. Priority and emphasis language in the source, including its praise of certain studies and its description of several as the first of their kind, belongs to the review and is not adopted here. A bibliographic discrepancy should be noted. The inventory records the year as 2019, matching the online publication and PubMed record, while the article itself carries a 2020 copyright line and citation instruction.

Audience summaries

25 words

Treating influenza A virus as a compact genetic circuit, this perspective organizes two decades of virus engineering into functional modules for tracking, override, barcoding and host silencing.

75 words

Influenza A virus carries only about ten gene products, which makes it tractable as an engineered system. This perspective organizes the field by what an added element does within the viral circuit, covering fluorescent reporters, Cre recombinase for marking cells that survived infection, microRNA target sites as host-specific kill switches, RNA barcodes that revealed that aerosol transmission passes only two or three virions, and virus-delivered artificial microRNAs used to screen host factors during infection.

150 words

Reverse genetics made it possible to build influenza A virus from DNA, and this perspective argues that building the virus is itself a way of understanding it. Systematic work on packaging signals and on mutational tolerance, largely from other laboratories, established that the genome has little room for random insertion, that packaging sequences extend into coding regions, and that hemagglutinin and NS1 are unusually permissive. Against that background the article sorts engineered viruses by circuit function. Reporters track location but usually attenuate, except when placed in PB2 or in an engineered intergenic space in segment 8. Cre recombinase combined with LoxP reporter mice marks cells that survived infection and identified club cells as clearing the virus. MicroRNA target sites act as species-restricted or lineage-restricted kill switches without altering coding material or fitness. RNA barcodes resolved transmission bottlenecks, and virus-delivered artificial microRNAs support in vivo host factor screening.

Discoveries supported by this paper

Documented publication relationships

• MicroRNA-mediated species-specific attenuation of influenza A virus — application. Direction: this paper → linked paper. stated in the Related publications section of 2019-tenoever-synthetic-virology-building-viruse
• MicroRNA-based strategy to mitigate the risk of gain-of-function influenza studies — related. Direction: this paper → linked paper. stated in the Related publications section of 2019-tenoever-synthetic-virology-building-viruse
• Influenza A Virus Utilizes Suboptimal Splicing to Coordinate the Timing of Infection — methodological foundation. Direction: this paper → linked paper. stated in the Related publications section of 2019-tenoever-synthetic-virology-building-viruse
• Engineered RNA viral synthesis of microRNAs — methodological foundation. Direction: this paper → linked paper. stated in the Related publications section of 2019-tenoever-synthetic-virology-building-viruse
• Long-term survival of influenza virus infected club cells drives immunopathology — application. Direction: this paper → linked paper. stated in the Related publications section of 2019-tenoever-synthetic-virology-building-viruse

Pathogens

Technologies