Marker rescue of adeno-associated virus (AAV) capsid mutants: a novel approach for chimeric AAV production

Marker rescue, the restoration of gene function by replacement of a defective gene with a normal one by recombination, has been utilized to produce novel adeno-associated virus (AAV) vectors. AAV serotype 2 (AAV2) clones containing wild-type terminal repeats, an intact rep gene, and a mutated cap gene, served as the template for marker rescue. When transfected alone in 293 cells, these AAV2 mutant plasmids produced noninfectious AAV virions that could not bind heparin sulfate after infection with adenovirus dl309 helper virus. However, the mutation in the cap gene was corrected after cotransfection with AAV serotype 3 (AAV3) capsid DNA fragments, resulting in the production of AAV2/AAV3 chimeric viruses. The cap genes from several independent marker rescue experiments were PCR amplified, cloned, and then sequenced. Sequencing results confirmed not only that homologous recombination occurred but, more importantly, that a mixed population of AAV chimeras carrying 16 to 2,200 bp throughout different regions of the type 3 cap gene were generated in a single marker rescue experiment. A 100% correlation was observed between infectivity and the ability of the chimeric virus to bind heparin sulfate. In addition, many of the AAV2/AAV3 chimeras examined exhibited differences at both the nucleotide and amino acid levels, suggesting that these chimeras may also exhibit unique infectious properties. Furthermore, AAV helper plasmids containing these chimeric cap genes were able to function in the triple transfection method to generate recombinant AAV. Together, the results suggest that DNA from other AAV serotypes can rescue AAV capsid mutants and that marker rescue may be a powerful, yet simple, technique to map, as well as develop, chimeric AAV capsids that display different serotype-specific properties.     

Characterization of Wild-Type Adeno-Associated Virus Type 2-Like Particles Generated during Recombinant Viral Vector Production and Strategies for Their Elimination

The pSub201-pAAV/Ad plasmid cotransfection system was developed to eliminate homologous recombination which leads to generation of the wild-type (wt) adeno-associated virus type 2 (AAV) during recombinant vector production. The extent of contamination with wt AAV has been documented to range between 0.01 and 10%. However, the precise mechanism of generation of the contaminating wt AAV remains unclear. To characterize the wt AAV genomes, recombinant viral stocks were used to infect human 293 cells in the presence of adenovirus. Southern blot analyses of viral replicative DNA intermediates revealed that the contaminating AAV genomes were not authentic wt but rather wt AAV-like sequences derived from recombination between (i) AAV inverted terminal repeats (ITRs) in the recombinant plasmid and (ii) AAV sequences in the helper plasmid. Replicative AAV DNA fragments, isolated following amplification through four successive rounds of amplification in adenovirus-infected 293 cells, were molecularly cloned and subjected to nucleotide sequencing to identify the recombinant junctions. Following sequence analyses of 31 different ends of AAV-like genomes derived from two different recombinant vector stocks, we observed that all recombination events involved 10 nucleotides in the AAV D sequence distal to viral hairpin structures. We have recently documented that the first 10 nucleotides in the D sequence proximal to the AAV hairpin structures are essential for successful replication and encapsidation of the viral genome (X.-S. Wang et al., J. Virol. 71:3077–3082, 1997), and it was noteworthy that in each recombinant junction sequenced, the same 10 nucleotides were retained. We also observed that adenovirus ITRs in the helper plasmid were involved in illegitimate recombination with AAV ITRs, deletions of which significantly reduced the extent of wt AAV-like particles. Furthermore, the combined use of recombinant AAV plasmids lacking the distal 10 nucleotides in the D sequence and helper plasmids lacking the adenovirus ITRs led to complete elimination of replication-competent wt AAV-like particles in recombinant vector stocks. These strategies should be useful in producing clinical-grade AAV vectors suitable for human gene therapy.     

Latent infection of KB cells with adeno-associated virus type 2.

Adeno-associated virus (AAV) is a prevalent human virus whose replication requires factors provided by a coinfecting helper virus. AAV can establish latent infections in vitro by integration of the AAV genome into cellular DNA. To study the process of integration as well as the rescue of AAV replication in latently infected cells after superinfection with a helper virus, we established a panel of independently derived latently infected cell clones. KB cells were infected with a high multiplicity of AAV in the absence of helper virus, cloned, and passaged to dilute out input AAV genomes. AAV DNA replication and protein synthesis were rescued from more than 10% of the KB cell clones after superinfection with adenovirus type 5 (Ad5) or herpes simplex virus types 1 or 2. In the absence of helper virus, there was no detectable expression of AAV-specific RNA or proteins in the latently infected cell clones. Ad5 superinfection also resulted in the production of infectious AAV in most cases. All mutant adenoviruses tested that were able to help AAV DNA replication in a coinfection were also able to rescue AAV from the latently infected cells, although one mutant, Ad5hr6, was less efficient at AAV rescue. Analysis of high-molecular-weight cellular DNA indicated that AAV sequences were integrated into the cell genome. The restriction enzyme digestion patterns of the cellular DNA were consistent with colinear integration of the AAV genome, with the viral termini present at the cell-virus junction. In addition, many of the cell lines appeared to contain head-to-tail concatemers of the AAV genome. The understanding of the integration of AAV DNA is increasingly important since AAV-based vectors have many advantages for gene transduction in vitro and in vivo.     

Second-strand synthesis is a rate-limiting step for efficient transduction by recombinant adeno-associated virus vectors.

The ability of recombinant adeno-associated virus (AAV) to transduce cells with a marker gene in vitro was found to be substantially increased by the presence of adenovirus. Transfection experiments with adenovirus genomic DNA suggest that this increase is not facilitated by adenovirus-mediated viral uptake but is instead dependent on adenovirus gene expression. Using various adenovirus mutants, we were able to map this function to early-region E4 open reading frame 6. Plasmid expression of open reading frame 6 protein in cells infected with recombinant AAV increased transduction between 100- and 1,000-fold. The increase in transduction was not dependent on the recombinant AAV gene cassette but instead appeared to involve an immediate early step of the AAV life cycle. Chemical and physical agents that have been shown to induce helper-free replication of wild-type AAV were also able to stimulate recombinant AAV transduction, suggesting that the phenomenon might affect AAV DNA replication. Further experiments showed that viral uncoating was not affected and that the rate-limiting step involved synthesis of a second strand on the single-stranded genomic AAV DNA. These data suggest that the adenovirus E4 region, as well as chemical and physical agents, can play an essential role in an immediate-early step of the AAV life cycle, specifically in second-strand synthesis, and have important implications for the use of AAV vectors in gene therapy protocols.     

The Mre11/Rad50/Nbs1 complex limits adeno-associated virus transduction and replication

Adeno-associated virus (AAV) is a parvovirus with a small single-stranded DNA genome that relies on cellular replication machinery together with functions supplied by coinfecting helper viruses. The impact of host factors on AAV infection is not well understood. We explored the connection between AAV helper functions supplied by adenovirus and cellular DNA repair proteins. The adenoviral E1b55K/E4orf6 proteins induce degradation of the cellular Mre11 repair complex (MRN) to promote productive adenovirus infection. These viral proteins also augment recombinant AAV transduction and provide crucial helper functions for wild-type AAV replication. Here, we show that MRN poses a barrier to AAV and that the helper function provided by E1b55K/E4orf6 involves MRN degradation. Using a fluorescent method to visualize the viral genome, we show an effect at the viral DNA level. MRN components accumulate at AAV replication centers and recognize the viral inverted terminal repeats. Together, our data suggest that AAV is targeted by MRN and has evolved to exploit adenoviral proteins that degrade these cellular factors.     

Rescue of the adeno-associated virus genome from a plasmid vector: evidence for rescue by replication

In cultured cells, adeno-associated virus (AAV) replication requires coinfection with a helper virus, either adenovirus or herpesvirus. In the absence of helper virus coinfection AAV can integrate its genome site specifically into the AAVS1 region of chromosome 19. Upon subsequent infection with a helper virus, the AAV genome is released from chromosome 19 by a process termed rescue, and productive replication ensues. The AAV genome cloned into a plasmid vector can also serve to initiate productive AAV replication. When such constructs are transfected into cells and those cells are simultaneously or subsequently infected with a helper virus, the AAV genome is released from the plasmid. This process is thought to serve as a model for rescue from the human genomic site. In this report we present a model for rescue of AAV genomes by replication. A hallmark of this model is the production of a partially single-stranded and partially double-stranded molecule. We show that the AAV2 Rep 68 protein, together with the UL30/UL42 herpes simplex virus type 1 DNA polymerase and the UL29 single-strand DNA binding protein ICP8, is sufficient to efficiently and precisely rescue AAV from a plasmid in a way that is dependent on the AAV inverted terminal repeat sequence.     

Adeno-associated virus vector integration junctions.

Vectors derived from adeno-associated virus (AAV) have the potential to stably transduce mammalian cells by integrating into host chromosomes. Despite active research on the use of AAV vectors for gene therapy, the structure of integrated vector proviruses has not previously been analyzed at the DNA sequence level. Studies on the integration of wild-type AAV have identified a common site-specific integration locus on human chromosome 19; however, most AAV vectors do not appear to integrate at this locus. To improve our understanding of AAV vector integration, we analyzed the DNA sequences of several integrated vector proviruses. HeLa cells were transduced with an AAV shuttle vector, and integrated proviruses containing flanking human DNA were recovered as bacterial plasmids for further analysis. We found that AAV vectors integrated as single-copy proviruses at random chromosomal locations and that the flanking HeLa DNA at integration sites was not homologous to AAV or the site-specific integration locus of wild-type AAV. Recombination junctions were scattered throughout the vector terminal repeats with no apparent site specificity. None of the integrated vectors were fully intact. Vector proviruses with nearly intact terminal repeats were excised and amplified after infection with wild-type AAV and adenovirus. Our results suggest that AAV vectors integrate by nonhomologous recombination after partial degradation of entering vector genomes. These findings have important implications for the mechanism of AAV vector integration and the use of these vectors in human gene therapy.     

Adeno-associated virus: a vector system for efficient introduction and integration of DNA into a variety of mammalian cell types.

Adeno-associated virus (AAV) is a single-stranded DNA parvovirus that is dependent on adenovirus or herpesvirus for reproductive functions. We describe the construction of recombinant AAV vectors containing the chloramphenicol acetyltransferase gene or the neomycin phosphotransferase gene. These vectors carried their respective genes into a wide variety of cell types, including primary skin fibroblasts and hematopoietic cells. Infection efficiencies varied with cell type and ranged up to 3.0%. Coinfection of two different recombinant viruses was also used to introduce two different sequences simultaneously into a given cell. Finally, methods for obtaining recombinant AAV vectors with minimal contamination of wild-type virus are described. These various attributes of AAV vectors make them a viable DNA transduction system.     

The origin of adenovirus DNA replication: minimal DNA sequence requirement in vivo.

Adenovirus mini-chromosomes which contain two cloned, inverted adenovirus termini replicate in vivo when supplied with non-defective adenovirus as a helper. This system has been used to define the minimum cis acting DNA sequences required for adenovirus DNA replication in vivo. Deletions into each end of the adenovirus inverted terminal repeat (ITR) were generated with Bal31 exonuclease and the resulting molecules constructed into plasmids which contained two inverted copies of the deleted ITR separated by the bacterial neomycin phosphotransferase gene. To determine the effect of the deletion in vivo plasmids cleaved to expose the adenovirus termini were co-transfected with adenovirus type 2 DNA into tissue culture cells. The replicative ability of the molecules bearing adenovirus termini was assayed by Southern blotting of extracted DNA which had been treated with DpnI, a restriction enzyme which cleaves only methylated and therefore unreplicated, input DNA. Molecules containing the terminal 45 bp of the viral genome were fully active whereas molecules containing only 36 bp were in-active in this assay. Therefore sequences required for DNA replication are contained entirely within the terminal 45 bp of the viral genome. Thus, both the previously described highly conserved region (nucleotides 9-18) and the binding site for the cellular nuclear factor I (nucleotides 19-48) are essential for adenovirus DNA replication in vivo.     

Spatial and temporal organization of adeno-associated virus DNA replication in live cells

Upon cell entry, the genomes of herpes simplex virus type 1 (HSV-1) and adenovirus (Ad) associate with distinct nuclear structures termed ND10 or promyelocytic leukemia (PML) nuclear bodies (NBs). PML NB morphology is altered or disrupted by specific viral proteins as replication proceeds. We examined whether adeno-associated virus (AAV) replication compartments also associate with PML NBs, and whether modification or disruption of these by HSV-1 or Ad, both of which are helper viruses for AAV, is necessary at all. Furthermore, to add a fourth dimension to our present view of AAV replication, we established an assay that allows visualization of AAV replication in live cells. A recombinant AAV containing 40 lac repressor binding sites between the AAV inverted terminal repeats was constructed. AAV Rep protein and helper virus-mediated replication of this recombinant AAV genome was visualized by binding of enhanced yellow fluorescent protein-lac repressor fusion protein to double-stranded AAV replication intermediates. We demonstrate in live cells that AAV DNA replication occurs in compartments which colocalize with AAV Rep. Early after infection, the replication compartments were small and varied in numbers from 2 to more than 40 per cell nucleus. Within 4 to 8 h, individual small replication compartments expanded and fused to larger structures which filled out much of the cell nucleus. We also show that AAV replication compartments can associate with modified PML NBs in Ad-infected cells. In wild-type HSV-1-infected cells, AAV replication compartments and PML NBs did not coexist, presumably because PML was completely disrupted by the HSV-1 ICP0 protein. However, alteration or disruption of PML appears not to be a prerequisite for AAV replication, as the formation of replication compartments was normal when the ICP0 mutants HSV-1 dl1403 and HSV-1 FXE, which do not affect PML NBs, were used as the helper viruses; under these conditions, AAV replication compartments did not associate with PML NBs.     

Efficient replication of adeno-associated virus type 2 vectors: a cis-acting element outside of the terminal repeats and a minimal size

Recombinant adeno-associated virus type 2 (AAV2) can be produced in adenovirus-infected cells by cotransfecting a plasmid containing the recombinant AAV2 genome, which is generally comprised of the viral terminal repeats flanking a transgene, together with a second plasmid expressing the AAV2 rep and cap genes. However, recombinant viruses generally replicate inefficiently, often producing 100-fold fewer virus particles per cell than can be obtained after transfection with a plasmid containing a wild-type AAV2 genome. We demonstrate that this defect is due, at least in part, to the presence of a positive-acting cis element between nucleotides 194 and 1882 of AAV2. Recombinant AAV2 genomes lacking this region accumulated 14-fold less double-stranded, monomer-length replicative-form DNA than did wild-type AAV2. In addition, we demonstrate that a minimum genome size of 3.5 kb is required for efficient production of single-stranded viral DNA. Relatively small recombinant genomes (2,992 and 3,445 bp) accumulated three- to eightfold less single-stranded DNA per monomer-length replicative-form DNA molecule than wild-type AAV2. In contrast, recombinant AAV2 with larger genomes (3,555 to 4,712 bp) accumulated similar amounts of single-stranded DNA per monomer-length replicative-form DNA compared to wild-type AAV2. Analysis of two recombinant AAV2 genomes less than 3.5 kb in size indicated that they were deficient in the production of the extended form of monomer-length replicative-form DNA, which is thought to be the immediate precursor to single-stranded AAV2 DNA.     

Rescue of adeno-associated virus from recombinant plasmids: gene correction within the terminal repeats of AAV

We have isolated three types of pBR322-AAV recombinant plasmids that contain deletions within the 145 bp AAV terminal repeats. When the plasmids were transfected into human cells, mutants that contained deletions within the left (type I) or right (type II) terminal repeat were viable. Of four mutants examined that contained deletions in both termini (type III), only one was viable. All of the viable mutants produced AAV virions that contained wild-type AAV DNA. Furthermore, the viable type III deletion could be converted to a nonviable mutant by deleting all copies of an 11 bp sequence from its termini. We conclude that there is an efficient mechanism for correcting deletions within the AAV termini. A model that could account for these observations is also discussed.     

A novel terminal resolution-like site in the adeno-associated virus type 2 genome.

The adeno-associated virus 2 (AAV) contains a single-stranded DNA genome of which the terminal 145 nucleotides are palindromic and form T-shaped hairpin structures. These inverted terminal repeats (ITRs) play an important role in AAV DNA replication and resolution, since each of the ITRs contains a terminal resolution site (trs) that is the target site for the AAV rep gene products (Rep). However, the Rep proteins also interact with the AAV DNA sequences that lie outside the ITRs, and the ITRs also play a crucial role in excision of the proviral genome from latently infected cells or from recombinant AAV plasmids. To distinguish between Rep-mediated excision of the viral genome during rescue from recombinant AAV plasmids and the Rep-mediated resolution of the ITRs during AAV DNA replication, we constructed recombinant AAV genomes that lacked either the left or the right ITR sequence and one of the Rep-binding sites (RBSs). No rescue and replication of the AAV genome occurred from these plasmids following transfection into adenovirus type 2-infected human KB cells, as expected. However, excision and abundant replication of the vector sequences was clearly detected from the plasmid that lacked the AAV left ITR, suggesting the existence of an additional putative excision site in the left end of the AAV genome. This site was precisely mapped to one of the AAV promoters at map unit 5 (AAV p5) that also contains an RBS. Furthermore, deletion of this RBS abolished the rescue and replication of the vector sequences. These studies suggest that the Rep-mediated cleavage at the RBS during viral DNA replication may, in part, account for the generation of the AAV defective interfering particles.     

A high-throughput hybridization method for titer determination of viruses and gene therapy vectors.

One of the challenges facing researchers working with viruses and gene therapy vectors is the need to rapidly assay for infectious virus. Current methods used to titer many viruses are cumbersome and are not amenable to handling large numbers of samples. Here we describe the development of an assay that can rapidly quantify infectious viruses and gene therapy vectors. The assay relies on biological amplification of viral sequences and hybridization of labeled probes to immobilized nucleic acid from infected cells. The amplification of the viral genome makes this a highly sensitive method. The assay is configured in a high-throughput format that has been used to detect recombinant adeno-associated virus (AAV), wild-type AAV and infectious adenovirus. The assay is quantitative, and can be used to titer virus preparations with or without a known standard.     

Transposase-mediated construction of an integrated adeno-associated virus type 5 helper plasmid

Adeno-associated viruses (AAVs) are replication-defective parvoviruses that require helper virusfunctionsfor efficient productive replication. The AAVs are currently premier candidates as vectors for human gene therapy applications. In particular; much recent interest has been expressed concerning recombinant AAV serotype 5 (rAAV-5) vectors, as they appear to utilize cellular receptors distinctfrom those of the prototypical AAV serotype (AAV-2) and have been reported to have transduction properties in vivo that differ significantly from those of the prototype. One of the most popular current methodsfor the production of rAAVs involves co-transfection of human 293 cells with three plasmids: (i) an adenovirus (Ad)-derived helper plasmid containing Ad genes required for AAV replication, (ii) an AAV-derived plasmid encoding complementing AAV genes (ie., the viral rep and cap genes), and (iii) a target plasmid containing a transgene of interestflanked by AAV inverted terminal repeats (ITRs) that confer packaging and replication capabilities upon the ITR-flanked heterologous DNA. Here we describe novel plasmid reagents designed for convenient and efficient production of rAAV-S. An integrated helper plasmid containing all Ad genes requiredfor the efficient production of recombinant AAV as well as the complementing AAV genes on the same plasmid backbone, was constructed via transposase-mediated insertion into an Ad helper plasmid of a transposable element containing the AAV-5 rep and cap genes linked to a selectable marker This simple strategy can be used in the rapid and efficient construction of integrated helper plasmids derived from any reported AAV serotype for which a molecular clone exists.     

Efficient expression of small RNA polymerase III genes from a novel simian virus 40 vector and their effect on viral gene expression.

In the past, simian virus 40 (SV40) has been used as a cloning vehicle to clone foreign genes by substituting portions of the viral genome vital for viral replication. Propagation of these defective viruses required a helper virus and the recombinant viruses obtained could be grown only as a mixture. In this study, we describe a novel nondefective SV40 vector to clone small RNA polymerase III genes. Two small RNA polymerase III genes, an amber suppressor human serine tRNA gene and the adenovirus (Ad) VAI RNA gene, were cloned in the intron region of the large-T antigen gene of SV40 after deleting DNA sequences coding for the small-t polypeptide. The recombinant viruses grew to wild type levels and showed no growth defects. When CV-1p cells were infected with these viruses, the cloned RNA polymerase III genes were expressed at high levels at late times. Interestingly, large amounts VAI RNA in CV-1p cells infected with SV40-VA recombinant virus, did not enhance translation of viral mRNAs significantly but did lead to a 3 to 4 fold increase in the steady state levels of large-T mRNA suggesting a novel function for VAI RNA in SV40 infected monkey cells. Furthermore, VAI mutants which fail to function in Ad infected human cells also failed to enhance the levels of large-T mRNAs in monkey cells infected with SV40. The simple SV40 vector described here may be useful to study the structure and function of small RNA polymerase III genes in the context of a eucaryotic chromosome. In addition, the nondefective recombinant SV40 which expresses the suppressor tRNA gene at high levels may provide a useful helper system to propagate animal viruses with amber mutations in essential genes.     

The rabbit C family of short, interspersed repeats. Nucleotide sequence determination and transcriptional analysis

When the entire adeno-associated virus (AAV) genome is inserted into a bacterial plasmid, infectious AAV genomes can be rescued and replicated when the recombinant AAV-plasmid DNA is transfected into human 293 cells together with helper adenovirus particles. We have taken advantage of this experimental system to analyze the effects of several classes of mutations on replication of AAV DNA. We obtained AAV mutants by molecular cloning in bacterial plasmids of naturally occurring AAV variant or defective-interfering genomes. Each of these mutants contains a single internal deletion of AAV coding sequences. Also, some of these mutant-AAV plasmids have additional deletions of one or both AAV terminal palindromes introduced during constructions in vitro. We show here that AAV mutants containing internal deletions were defective for replicative form DNA replication (rep-) but could be complemented by intact wild-type AAV. This indicates that an AAV replication function, Rep, is required for normal AAV replication. Mutants in which both terminal palindromes were deleted (ori-) were also replication defective but were not complementable by wild-type AAV. The cis-dominance of the ori- mutation shows that the replication origin is comprised in part of the terminal palindrome. Deletion of only one terminal palindrome was phenotypically wild-type and allowed rescue and replication of AAV genomes in which the deleted region was regenerated apparently by an intramolecular correction mechanism. One model for this correction mechanism is proposed. An AAV ori- mutant also complemented replication of AAV rep- mutants as efficiently as did wild-type AAV. These studies also revealed an unexpected additional property of the deletion mutants in that monomeric single-stranded single-stranded DNA accumulated very inefficiently even though monomeric single-stranded DNA from the complementing wild-type AAV did accumulate.     

Effect of deletions in adenovirus early region 1 genes upon replication of adeno-associated virus.

The growth of adeno-associated virus (AAV) is dependent upon helper functions provided by adenovirus. We investigated the role of adenovirus early gene region 1 in the AAV helper function by using six adenovirus type 5 (Ad5) host range mutants having deletions in early region 1. These mutants do not grow in human KB cells but are complemented by and grow in a line of adenovirus-transformed human embryonic kidney cells (293 cells); 293 cells contain and express the Ad5 early region 1 genes. Mutants having extensive deletions of adenovirus early region 1a (dl312) or regions 1a and 1b (dl313) helped AAV as efficiently as wild-type adenovirus in 293 cells, but neither mutant helped in KB cells. No AAV DNA, RNA, or protein synthesis was detected in KB cells in the presence of the mutant adenoviruses. Quantitative blotting experiments showed that at 20 h after infection with AAV and either dl312 or dl313 there was less than one AAV genome per cell. In KB cells infected with AAV alone, the unreplicated AAV genomes were detected readily. Apparently, infection with adenovirus mutant dl312 or dl313 results in degradation of most of the infecting AAV genomes. We suggest that at least an adenovirus region 1b product (and perhaps a region 1a product also) is required for AAV DNA replication. This putative region 1b function appears to protect AAV DNA from degradation by an adenovirus-induced DNase. We also tested additional Ad5 mutants (dl311, dl314, sub315, and sub316). All of these mutants were inefficient helpers, and they showed varying degrees of multiplicity leakiness. dl312 and dl313 complemented each other for the AAV helper function, and each was complemented by Ad5ts125 at the nonpermissive temperature. The defect in region 1 mutants for AAV helper function acts at a different stage of the AAV growth cycle than the defect in the region 2 mutant ts125.     

Isolation of cis-acting vaccinia virus DNA fragments promoting the expression of herpes simplex virus thymidine kinase by recombinant viruses.

Recombinant TK- vaccinia viruses containing the pBR322 sequence inserted in either orientation within the coding sequence of the viral thymidine kinase gene were constructed. They were characterized by genomic analysis, hybridization studies, reversion to wild-type virus by in vivo recombination, and rescue from their genomes of plasmids which contained all or parts of the pBR322 sequence. TK- cells were infected with one of these recombinant viruses and then transfected with pools of chimeric plasmids composed of a cloned herpes simplex virus thymidine kinase gene which contained upstream inserts of different vaccinia DNA fragments prepared by restriction or sonication. Recombination between homologous pBR322 sequences within infected cells generated selectable recombinant viruses in which expression of the herpes simplex virus thymidine kinase gene was promoted by the upstream vaccinia insert. These viruses were characterized by genomic analysis, hybridization, and in vivo or in vitro phosphorylation of (5-[125I]deoxycytidine as a specific assay for the expressed herpes simplex virus thymidine kinase. Vaccinia DNA inserts were isolated conveniently for transfer to bacteria by rescuing appropriate plasmids from the genome of recombinant viruses. The sequence of 100 nucleotides adjacent to the upstream region of the herpes simplex virus gene was determined in nine different inserts measuring 0.17 to 1.07 kilobase pairs.