key: cord-0008786-vnynbijl authors: Conzelmann, Karl-Klaus; Meyers, Gregor title: Genetic engineering of animal RNA viruses date: 1999-03-19 journal: Trends Microbiol DOI: 10.1016/0966-842x(96)10062-7 sha: e5949cd130cebafdb37640236ef94aac025be11e doc_id: 8786 cord_uid: vnynbijl The ability to genetically manipulate viruses has led to extraordinary advances in understanding virus biology and to the establishment of useful vector systems. Initially confined to DNA viruses and retroviruses, RNA viruses have more recently become attractive candidates for expression of heterologous genes and offer promising perspectives for biomedical applications. R ecombinant DNA technology makes it possible to specifically modify DNA genomes. Because of the small size of their genomes, viruses are particularly amenable to such manipulations. Since the recovery of the first recombinant virus, simian virus 40 (SV40), 20 years ago I, a variety of DNA viruses and retroviruses has been genetically manipulated to exp~loit their replication machinery' for expression of heterologous genes. When transfected into a cell, the purified DNA of The ability to genetically manipulate viruses has led to extraordinary advances in understanding virus biology and to the establishment of useful vector systems. Initially confined to DNA viruses and retroviruses, RNA viruses have more recently become attractive candidates for expression of heterologous genes and offer promising perspectives for biomedical applications. 4 . In recent years, RNA viruses lacking a DNA phase in their replication cycle have also been manipulated. Since recombinant RNA techniques remain cumbersome, almost all studies with RNA viruses have relied upon cDNA intermediates to produce biologically active RNA molecules. Two major RNA virus groups, the positive strand and the negative strand RNA viruses, are distinguished based on whether their purified RNAs are able to initiate an infectious cycle or not. For the latter group, the minimal infectious unit is not an RNA molecule but a ribonucleoprotein qomplex (RNP). The reconstitution of functional RNPs from proteins and The plus-stranded RNA viruses were the first that were open to direct genetic manipulation s-v because their genomic RNA (vRNA) is able to function as an mRNA, directing the production of all viral proteins necessary for the initiation of virus propagation. To provide a template for the synthesis of additional mRNA molecules, replication starts with the polymerization of a minus strand complementary to the genome (cRNA). Thus, for all positive-strand RNA viruses the components of the replicase complex have to be translated directly from the genomic RNA. For the other viral polypeptides, which mainly constitute structural proteins, two principle expression strategies exist, by which positivestrand RNA viruses can be classified into two groups (Fig. 1) . Viruses in the first group generate only one kind of mRNA, which is of genome length and contains one long open reading frame (ORF). Expression of all viral proteins is achieved by translation of this RNA into a polyprotein that is co-and post-translationally REVIEWS processed by viral and host cellular proteases. The members of the picornavirus and flavivirus families belong to this first group. The genomes of these viruses are 7-16.5 kb, and are organized such that the structural proteins are encoded in the 5' region, whereas the 3' terminal part of the ORF codes for the RNA-dependent RNA polymerase. The second group comprises the families Togaviridae, Caliciviridae, Coronaviridae and Arteriviridae. These viruses are characterized by the subgenomic RNAs used for gene expression (Fig. 1 ). In contrast with the first group, the replicase genes of these viruses are located in the 5" part of the genome upstream of the structural genes. For all of these viruses the subgenomic RNAs are 3' coterminal with the genomic RNA. In the case of alphaviruses and rubivirus (family Togaviridae), one subgenomic RNA of about 4 kb is synthesized from a 12 kb genome cRNA after initiation at a specific promoter region. Translation of the subgenomic RNA results in a polyprotein that is processed into the three structural proteins s. One species of subgenomic RNA is also found in cells infected with caliciviruses. It is colinear with the 3' terminal third of the 7.5 kb genomic RNA and is responsible for expression of the capsid protein of these viruses 9J°. Members of the families Coronaviridae and Arteriviridae have genomes of about 30 and 15 kb, respectively, and express multiple subgenomic mRNAs. The mRNAs form a 3' coterminal nested set, which implies that all but the smallest subgenomic RNAs contain more than one cistron (Fig. 1) . In most cases only the 5" terminal gene of each mRNA is translated ~ 1-13. Usually, the first step towards the generation of a recombinant RNA virus is cDNA cloning of the complete viral genome. This also implies determination of the utmost terminal sequences of the RNA, which contain important signals necessary for replication. These initial steps are followed by attempts to construct fulllength cDNA clones encompassing the complete genetic information of the virus. Although merely a technical problem, this part of the work is challenging since the constructs tend to be unstable when amplified in bacteria and deleterious mutations can be acquired during the cloning procedure. In some cases, problems with insert instability can be circumvented by the use of low copy-number plasmids and bacterial hosts bearing mutations that impair DNA recombination and repair14 16. For the flaviviruses yellow fever virus and japanese encephalitis virus, however, stable full-length clones have not yet been obtained. Nevertheless, recombinant viruses have been generated after in vitro ligation of two fragments that together cover the complete viral genome 17' 18. The picornaviruses poliovirus and coxsackie B3 virus were first recovered after transfection of full-length cDNA constructs into permissive cells 7A9'2°. The efficiency of the respective systems was quite low even when eukaryotic promoters were included. The main difficulties of this approach lie in the generation of the correct RNA termini and, presumably, in problems with nuclear export and unwanted splicing of the in vivo generated RNA. Despite these difficulties, the generation of an alphavirus replicon in cells transfected with a cDNA construct has been published recently zl. Integration of such a construct into the genome of the host cell could lead to inheritable high level expression of a foreign gene. Recombinant nodavirus replicons have been generated after insertion of full-length cDNA constructs into recombinant DNA viruses (vaccinia virus) allowing efficient in vivo generation of infectious RNA within the cytoplasm of the infected cells z2. The strategy of choice for recovery of recombinant viruses is transfection of target cells with genome-like RNA, which is generated in vitro by 'runoff' transcription with an RNA polymerase derived from the bacteriophages T3, T7 or SP6 (Fig. 2) . The simple structure of the promoters for these enzymes, together with an appropriate restriction site used to linearize the template at the 3" end of the viral sequence, allow synthesis of an RNA with precise termini. The RNA yields obtained by this method are high and the specific infectivity of the transcripts in some cases approaches that of virion RNA 23,24. The RNA generated in vitro can be introduced into cells by various methods. Transfection of RNA, bound to either DEAE dextran or cationic lipids, is most commonly used. Recently, electroporation has also been employed for this purpose and results in transfection efficiencies of almost 100% ( Refs 24, 25) . Until now, the large genome size of coronaviruses (28-32kb) has prevented the generation of viruses entirely from cloned sequences. Two different strategies have been pursued to circumvent this problem. One approach is based on the transcription of smaller RNAs that contain deletions and therefore are not able to replicate autonomously. Such RNAs can be a source of defective interfering particles (Dis). Dis are able to replicate with the help of normal viruses, which provide the functions not encoded by the DI itself. Since DI genomes have to contain certain cisacting sequences to be replicated and packaged into infectious particles, construction of synthetic DI genomes can be used for the identification of replication and packaging signals. In addition, recombinant DI-RNAs can serve as replicons designed for the expression of foreign genes 26,27. The second approach is based on the high frequency of recombination observed for coronaviruses. During one round of replication about 25% of the newly generated RNAs undergo recombination 13. This extremely high rate of recombination probably resolves the problem of deleterious mutations resulting from replication of the large genome by an error-prone RNAdependent RNA polymerase. The characteristically efficient homologous recombination can be employed for the genetic manipulation of viruses by transfection of coronavirus-infected cells with, for example, RNA containing a heterologous gene flanked by coronavirus sequences 28,29. Different positive strand RNA viruses have been used as vectors for the expression of heterologous sequences 24. Obvious advantages of these systems include the easy and rapid engineering of the DNA constructs and, in contrast to DNA viruses like vaccinia virus, the possibility of avoiding any wild-type virus background by de novo generation of viruses entirely from cloned sequences. Among the positive strand RNA viruses that do not possess subgenomic mRNAs, most efforts in vector development have focused on poliovirus. The major aim of these studies was the generation of new vaccines 3°. In this respect, poliovirus is of interest because of the availability of human vaccine strains of defined efficacy and safety, which have the advantages of oral application and the ability to induce mucosal immunity. To this end, poliovirus has been used for the expression of short heterologous peptides integrated into the viral capsid proteins. Detailed knowledge of poliovirus structure and antigenic properties has facilitated the choice of integration sites. The expression of a foreign protein with defined ends using poliovirus is more difficult. Since the ends of poliovirus proteins are generated by processing at specific protease cleavage sites, defined both by amino acid sequence and protein structure, insertion of a foreign sequence into the genomic RNA usually results in expression of a fusion protein. This problem has been overcome by placing foreign sequences, followed by a cleavage site for the poliovirus 3C protease, at the 5' end of the open reading frame. Sequences coding for up to 400 amino acid residues have been inserted this way 31. An elegant alternative method is based on the generation of bicistronic poliovirus RNAs by integration of an internal ribosome entry site (IRES) 32. IRES elements are usually located at the 5' end of picornaviral RNAs, and allow translation in a cap-independent REVIEWS fashion, based on the direct binding of ribosomes to a specific RNA structure~L Integration of a second IRES element at an internal position results in translation of downstream sequences starting at a defined initiation site. By positioning the second IRES together with the foreign gene at the end of the viral ORF, a heterologous protein can be expressed with its ends defined by the translation initiation site and the translational stop codon. For members of the flavivirus family, infectious clones have been described in the past few years14-1L In the near future, genetic manipulation of these viruses will be focused by investigations of virus biology and the need to develop vaccines against these pathogens. Construction of chimeras between pathogenic and attenuated viruses will help to identify mutations responsible for attenuation. In the case of pestiviruses, which cause enormous losses to livestock, an additional challenge will be to introduce a specific marker into the vaccine to permit the differentiation of vaccinated animals from those infected with field viruses. Such 'marker' vaccines are prerequisites for vaccinationbased eradication of these pathogens. Flaviviruses could also be vector candiates. One advantage of these viruses is their slow replication, which results in less severe immediate effects on infected cells. Also, naturally occurring pestivirus recombinants have been isolated that contain more than 4 kb of additional sequences in their genome, suggesting that packaging constraints might be less important for these viruses 34. The most advanced vector systems based on positive strand RNA viruses have been developed for alphaviruses 24,2s. Establishment of these systems has been facilitated by the detailed study of the cis-acting elements in the viral genome, which are responsible for replication, packaging and transcription. Because alphaviruses use a subgenomic RNA for gene expression, a foreign protein with defined ends can easily be generated via translation of an additional mRNA. Three distinct groups of alphavirus expression systems have been described. The first group consists of viruses that have the foreign gene inserted into a nondefective genomic RNA able to replicate and to produce progeny virus. In most cases, the foreign sequence is expressed via a second subgenomic RNA that is transcribed from a duplicated subgenomic promoter. The insert can either be placed upstream or downstream of the structural genes. The system is designed for serial propagation of the recombinant viruses 24. The second group relies on the generation of synthetic Dis that contain the 5' and 3' terminal sequences necessary for replication and packaging. These sequences flank the foreign gene, which again is under the control of the subgenomic promoter. The DI is replicated in the presence of a helper virus that also provides the enzymes for transcription and capping of an mRNA that directs translation of the foreign sequence ~4. For the third type of alphavirus expression system, the heterologous gene is inserted into the genomic RNA and replaces the sequence coding for the structural proteins. The resulting replicon is replication competent but does not generate infectious particles. Efficient expression of foreign genes has been obtained in this way with the polypeptide product representing up to 25% of the total cellular protein 21. The packaging capacity of these alphavirus vectors is thought to be about 5 kb of foreign sequences 24. A similar system based on Semliki forest virus is available commercially (GIBCO BRL Life Technologies Ltd, Gaithersburg, MD, USA). To avoid the need for high efficiency transfection, the replicon can be packaged into infectious virions by providing the structural proteins in trans. The easiest way to do this is by the cotransfection of a DI genome that directs the synthesis of the structural proteins 24,2s. An interesting option is to use a DI RNA without packaging signals that results in a 'dead end' or 'suicide' system; after transfection of the recombinant genome together with the engineered DI genome, high titer stocks of recombinant viruses, but no infectious DI particles, are obtained. The viruses can infect new cells with almost 100% efficiency but do not produce progeny virus since the structural genes are missing. A commercially available Sindbis virus expression system relies on this strategy (Invitrogen Corporation, San Diego, USA). However, the 'dead end' strategy for the alphavirus system is somewhat leaky, as infectious viruses can be reconstituted by recombination between the DI RNA and the recombinant genome. The potential of other positive strand RNA viruses with subgenomic RNAs for expression of foreign sequences has not yet been evaluated. In the near future, genetic manipulation of corona-and caliciviruses will probably be employed mainly for studies on virus biology and vaccine production. There are intrinsic difficulties with the coronavirus system and the first infectious calicivirus clone has been described only recently3L However, the simple construction of the calicivirus capsid, which consists of only one major protein that spontaneously assembles into virus-like particles 36, makes this virus group interesting for biotechnological applications. Because of their simple biology, nodaviruses also represent interesting vector candidates. These insect viruses have a bipartite genome; one segment encodes the coat protein precursor and the other segment directs synthesis of the replicase. Expression via vaccinia virus showed that the latter RNA is self replicative and thus can be used for replication and expression of a foreign gene 22. Negative strand RNA viruses represent a diverse group of enveloped viruses. A major distinction lies between viruses whose genomes consist of a single RNA molecule (Mononegavirales order), such as the Rhabdoviridae, Paramyxoviridae and Filoviridae families, and those possessing multipartite (segmented) genomes comprising the Orthomyxoviridae (six to nine segments), the Bunyaviridae (three segments), and the Arenaviridae (two segments)3L However, both groups share features characteristic for their typical mode of replication and gene expression. All negative strand RNA viruses contain a ribonucleoprotein complex (RNP) in which the RNA is of the growing RNA chain into the RNP is functionally linked. Mechanisms to ensure the synthesis of nonencapsidated mRNAs involve elongation of capped primers originating from cellular mRNAs ('cap-snatching') in segmented viruses 4°, or the production of a short leader RNA carrying the encapsidation signal in viruses with a nonsegmented genome. In both segmented and nonsegmented viruses transcription is stopped preterminally at an internal transcription stop-signal whereas during replication, polymerization proceeds to the 5" terminus of the template. The 3' terminal region of the resulting antigenomic RNP-RNA is highly similar in its nucleotide composition to that of the genomic RNA and functions as a strong promoter for synthesis of genome sense RNPs, but not for transcription of a subgenomic RNAs (Fig. 3 ). Owing to this expression strategy, isolated RNA from negative strand RNA viruses cannot be infectious. Proteins can be expressed neither from the genomic RNA, because of its negative polarity, nor from the complementary positive strand, because of the modular organization of the RNA which does not allow translation of (all) viral proteins by the cellular machinery. However, the most severe obstacle to genetic alteration of negative strand RNA viruses and the most striking difference from positive strand RNA viruses is the fact that the RNA must be encapsidated into nucleoprotein in order to function as a template for the polymerase. tightly encapsidated in a nucleoprotein (N or NP) and associated with the viral RNA-dependent RNA polymerase. In the case of nonsegmented viruses, the polymerase consists of a catalytic subunit (L) and a noncatalytic cofactor, a phosphoprotein (P). In the virions, RNPs that contain the anti-messenger (negative) sense RNA are enclosed into simple envelopes containing an internal matrix protein (M) and one or two transmembrane spike proteins. The initiation of an infectious cycle requires the presence of a complete RNP. The RNP serves as a template for two distinct RNA synthesis functions, namely the replication of full length RNAs and the transcription of subgenomic mRNAs from internally located cistrons (Fig. 3) 38,39. Since only one polymerase entry site (promoter) at the 3' end of the RNAs exists, the polymerase has to act in a processive mode for the synthesis of full length RNAs and in a nonprocessive mode for transcription (Fig. 3) . Most simply, this can be explained by the presence of a 'replicase' and 'transcriptase' form of the polymerase complex, with only the latter being able to recognize the internal cis-acting stop and restart signals that define the cistron borders. In contrast to transcription, the product of replication is not a free RNA, but an RNP. Since constant protein synthesis is a prerequisite for replication of all negative strand RNA viruses, it is assumed that RNA polymerization and encapsidation Successful encapsidation of a preformed RNA into nucleoprotein ('illegitimate' encapsidation) and formation of a biologically active RNP was first achieved by Peter Palese and colleagues 4~. Transcripts that contained the terminal promoter sequences from an influenza virus genome segment and an internal chloramphenicol acetyl transferase (CAT) reporter gene were encapsidated in vitro by purified influenza virus nucleoprotein (NP) and the viral polymerase proteins PA, PB1 and PB2. After transfection of the reconstituted RNPs into influenza (helper-) virus-infected cells, the RNPS were replicated and transcribed. By reassortment of genome segments, the synthetic segment was incorporated into progeny virus (Fig. 4) Probably because of a tighter RNP structure, attempts to encapsidate preformed RNAs of nonsegmented viruses in vitro have failed. RNPs of these viruses are only formed inside a cell that provides the required viral proteins, as first demonstrated for a monocistronic Sendai virus minigenome 4s (Fig. 3) . By using REVIEWS similar model genomes, intracellular encapsidation systems were subsequently established for several rhabdo-and paramyxoviruses (reviewed in Ref. 46 ). The most well-developed approach makes use of the recombinant vaccinia virus-T7 RNA polymerase expression system 47 and involves simultaneous intracellular expression of both viral proteins and RNA from transfected plasmids (Fig. 4) . A major breakthrough in optimizing the system was exploitation of the autolytic activity of a hepatitis delta virus ribozyme, which is indiscriminate with regard to sequences 5' of its cleavage site, to develop plasmid vectors that allowed intracellular generation of RNAs with discrete 3' termini 22,48. RNAs ending with the correct 3' nucleotide can be generated by autolytic cleavage from primary transcripts containing the ribozyme sequence immediately downstream of the virus sequences. The approach appears to be applicable to most of the negative strand RNA viruses, including the segmented Bunya- Our lab was the first to show that for the rabies rhabdovirus, the RNA of a negative strand RNA virus can be made infectious by coexpression with the N, P and L proteins s5. An important feature of the approach was initiation of the infectious cycle by expressing the antigenomic RNA rather than the genomic RNA. This strategy avoided a potentially deleterious antisense problem, in which RNA transcripts encoding the N, P and L proteins would hybridize to the complementary viral genome sequences and interfere with correct encapsidation. As anticipated, rescue of a 12-kb fulllength RNA into a functional RNP is inefficient and therefore virus production is observed in only one in 106-107 transfected cells. The same strategy was used subsequently for successful recovery of the prototype rhabdovirus vesicular stomatitis virus (VSV) s<57, and also of members from all three genera of the Paramyxoviridae family: recombinant measles (morbilli-) virus 5*, Sendai (paramyxo-) virus sg, and human respiratory syncytial (pneumo-) virus 6° were recovered from full-length antigenome RNAs. Most of the approaches to the recovery of recombinant rhabdo-and paramyxoviruses have been based on vaccinia (vTF7-3 or MVA-T747#1) driven expression of full-length RNA and viral N, P and L proteins. One of the features of this system has become evident, namely its striking potential for homologous DNA recombination. Replacement of a genetic marker in the full-length Sendal virus cDNA with wild-type sequence was observed at high frequency and was triggered by vaccinia-induced recombination of the fulllength cDNA with a protein-encoding plasmid. This feature was exploited to create a novel Sendai virus whose genome was derived from the sequences of two independent plasmids sg. For recovery of measles virus, cell lines constitutively expressing T7 RNA polymerase and the N and P proteins were used, and viral antigenome and L protein were provided transiently from transfected T7-polymerase-driven plasmidsSL Infectious RSV could be recovered after expression of the elongation factor encoded by the viral M2 gene, in addition to the N, P and L proteins 6°. The tools to manipulate virus genomes allow investigation of the cis-acting sequences that control gene REVIEWS expression, as well as the function of each viral protein in virus replication, assembly and interaction with the host, and will rapidly advance knowledge in this field of virology. In addition to the feasibility of designing attenuated and effective virus vaccines, the exciting potential of negative strand RNA viruses as vectors to express foreign genes can be exploited. Influenza virus and, more recently, rhabdoviruses have been shown to have the capacity to express additional protein sequences. Since nonessential genome segments of influenza virus are rapidly lost by reassortment, selective pressure or linkage of foreign sequences to one of the essential segments is required. Because of the monocistronic nature of the influenza virus genome segments, strategies for the construction of vectors are similar to those for picorna-flavi-and pesti-viruses. These include the insertion of foreign epitopes into suitable sites in the viral glycoproteins NA and HA, and the expression of polyproteins that are autolytically cleaved by, for example, at picornaviral 2A protease target sequences 62. The function of a mammalian IRES element has been exploited to create a bicistronic NA segment expressing two distinct open reading frames 63. Influenza A and B virus vectors have been obtained that express both B and T cell epitopes from different pathogens% Since influenza virus is one of the few RNA viruses replicating in the nucleus of infected cells, attention has to be paid to potential splicing of mRNAs. Engineering of additional genes into nonsegmented viruses is considerably facilitated by the modular organization of their genomes (Fig. 3) . By insertion of an additional gene end and start signal into the rabies virus genome, a virus was generated containing an extra transcription unit and transcribing an extra mRNA% This approach has allowed successful expression of reporter genes such as CAT from recombinant rabies virus 65 and also from VSV 66. Most remarkably, the foreign, and nonessential genes are maintained and expressed through, so far, more than 50 cell culture passages. This finding has confirmed the previous observation that genetic material with no apparent function, particularly large noncoding regions in the glycoprotein genes, are maintained stably in rabies and measles virus genomes 5s,sS. Continuous streamlining of positive strand RNA virus genomes and elimination of sequences not strictly required is probably achieved through recombination by a copy-choice mechanism, which involves detachment of the polymerase from the template, attachment to another template molecule and resumption of polymerization 67,6~. The tight encapsidation into nucleoprotein, which renders the genomes of rhabdoviruses and paramyxoviruses inaccessible for RNase degradation, will impede this mechanism. The tight RNP structure of nonsegmented negative strand RNA viruses had represented the major obstacle to their genetic manipulation, but now appears to be a most welcome feature of recombinant vectors. By selecting the site for introduction, as well as manipulation of genome and antigenome promoters 59,69, appropriate adjustment of the transcription level of foreign genes appears possible. Moreover, the helical nature of RNPs suggests that minimal constraints on the amount of additional RNA might exist in this virus group. Perhaps the most exciting potential of rhabdovirus vectors stems from the simple structure of their envelopes and assembly mechanism. The generation of viruses carrying novel proteins in their envelopes and delivering genes only to the target cells of interest now appears realistic. It has already been shown for VSV and rabies virus that the glycoproteins of other strains or serotypes as well as chimeric proteins can replace homologous proteins 5