Search

Homo-oligomeric runs were inserted into a spleen necrosis virus-based retrovirus vector to determine the nature and rate of mutations within runs of 10 to 12 identical nucleotides during a single replication cycle. Clones of helper cells containing integrated copies of retroviral vectors were used to produce virus for infection of target (nonhelper) cells. Proviral sequences from target cell clones were compared with proviral sequences from helper cell clones to study mutations that occurred during a single cycle of replication. In addition to the internal region spanning the homo-oligomeric inserts, a naturally occurring run of 10 T's in the long terminal repeat (LTR) also was sequenced. Rates of mutation ranged from < 0.01 to 0.38 frameshift mutations per run per cycle for different nucleotide runs. Frameshift mutations ranged from deletions of 2 bases to additions of 5 bases; the most common mutations were +1 and -1. Frameshift mutation rates did not increase as the run length increased from 10 to 12 bases. Rates of frameshift mutation for runs of T's and A's were significantly higher than rates for runs of C's and G's, and rates for runs of pyrimidines were significantly higher than those for runs of purines. Interestingly, the vast majority of frameshift mutations in the internal region (95%) were positive, suggesting that the primer strand tends to slip backward on the template in this region. LTR runs had a significantly lower number of positive frameshift mutations than the internal runs. By analyzing the types of frameshift mutations within runs and by comparing the patterns of frameshift mutations in the 5' and 3' LTRs of individual proviruses, we conclude that the majority of mutations observed in our system occurred during minus-strand DNA synthesis of reverse transcription.

We tested the effect of copackaging retroviral vectors of different sizes on retroviral replication and recombination. Our results indicate little or no difference in replication or in the rate or pattern of the strand transfers leading to the formation of recombinant proviruses with size. The size difference of the vectors also allowed us to extend our previous analysis of the linkage of markers in the recombinant proviruses. We conclude that the observed linkage is inconsistent with the strand displacement/assimilation model of retroviral recombination.

Retroviruses, as a result of the presence of two identical genomic RNA molecules in their virions, recombine at a high rate. When nonhomologous RNA is present in the dimer RNA molecules, nonhomologous recombination can occur, although the rate is very low, only 0.1% of the rate of essentially homologous recombination (J. Zhang and H. M. Temin, Science 259:234-238, 1993). We found, as is found in naturally occurring highly oncogenic retroviruses (J. Zhang and H. M. Temin, J. Virol. 67:1747-1751, 1993), that the crossovers usually occur at a short region of sequence identity. We modified the previously studied vectors to study the effect of different lengths of short regions of sequence identity in the midst of otherwise nonidentical sequences. We found that the efficiency of recombination depends on the length of this sequence identity. However, the highest rate in such molecules remained lower than for recombination between essentially homologous molecules, even when there was extensive sequence identity. Junction sequences of the recombinants indicated that retrovirus recombination is not an error-prone process as was reported for human immunodeficiency virus reverse transcriptase by using a cell-free system (J. A. Peliska and S. J. Benkovic, Science 258:1112-1118, 1992).

We used previously characterized spleen necrosis virus-based retroviral vectors and helper cells to study the strand transfers that occur during the reverse-transcription phase of a single cycle of retroviral replication. The conditions used selected only for formation of an active provirus rather than for expression of multiple drug resistance markers. In nonrecombinant proviruses the minus- and plus-strand DNA primer transfers were almost completely intramolecular. However, as previously reported, recombinant proviruses contained approximately equal proportions of inter- and intramolecular minus-strand DNA primer transfers. Thus, we conclude that in the absence of recombination, one molecule of retroviral RNA is sufficient for viral DNA synthesis. Large deletions and deletions with insertions were detected primarily at a limited number of positions which appear to be hot spots for such events, the primer binding site and regions containing multiple inverted repeats. At these hot spots, the rate of deletions and deletions with insertions visible with PCR was about 10% per genome per replication cycle. Other deletions and deletions with insertions (detectable with PCR) occurred at a rate of about 0.5%/kb per replication cycle. Crossovers occurred at a rate of about 6%/kb per replication cycle under single-selection conditions. This rate is comparable to the rate that we reported previously under double-selection conditions, indicating that retroviral homologous recombination is not highly error prone. The combined rates of deletions and deletions with insertions at hot spots (10% per genome per replication cycle) and other sites (0.5%/kb per replication cycle) and the rate of crossovers (6%/kb per replication cycle) indicate that on average, full-size (10-kb) type C retroviruses undergo an additional or aberrant strand transfer about once per cycle of infection.

Genetic variation of the more complex retroviruses in the human T-cell leukemia virus/bovine leukemia virus (HTLV/BLV) group is less than in some other retroviral genera. To test whether reverse transcription of HTLV/BLV group members is less error prone than that of members of other groups, we developed an assay for detecting forward mutations in BLV, similar to that developed for the simpler spleen necrosis virus (SNV). We used this system to study the rates and types of mutations that occur during a single replication cycle. We found that BLV reverse transcription is approximately two and one-half times less error prone than SNV reverse transcription (4.8 x 10(-6) versus 1.2 x 10(-5) mutation per bp per cycle, respectively). The relative numbers of all types of observed mutations (that is, base pair substitutions, frameshifts, deletions, and deletions with insertions) were similar for BLV and SNV.

Retrovirus particles contain a dimer of retroviral genomic RNA. A defined region of the retrovirus genome has previously been shown to be important for both dimerization and encapsidation. To study the importance of the position of this encapsidation and dimerization signal for retroviral replication and homologous recombination, we used a previously described spleen necrosis virus-based helper cell system. This system allows retroviral vectors with multiple genetic markers to be studied after a single cycle of retroviral replication. The sequence responsible for dimerization, the encapsidation/dimer linkage sequence (E/DLS), was moved from its normal location near the 5' end of the retroviral genome to a location near the 3' end of the genome. We characterized four pairs of retroviral vectors: (i) with both E/DLSs at the 5' ends of the genomes, (ii) with both E/DLSs at the 3' ends of the genomes, and (iii) two with one E/DLS at the 5' end of the genome and one at the 3' end of the genome. We found that moving the E/DLS to the 3' end of the genome resulted in at most an approximately factor of 5 reduction in virus titer in a single cycle of retroviral replication. Furthermore, we found no changes that were attributable to the alteration of the position of the E/DLS in the minus-strand DNA primer transfers or the plus-strand DNA primer transfers, the rate of homologous recombination, or the number of internal template switches in recombinant proviruses. These results indicate that any alignment or conformation necessary for retroviral replication or recombination is not the result of the position of the E/DLS.

Recombination occurs at a high rate in retroviral replication, and its observation requires a virion containing two different RNA molecules (heterodimeric particles). Analysis of retroviral recombinants formed after a single round of replication revealed that (i) the nonselected markers changed more frequently than expected from the rate of recombination of selected markers; (ii) the transfer of the initially synthesized minus strand strong stop DNA was either intramolecular or intermolecular; (iii) the transfer of the first synthesized plus strand strong stop DNA was always intramolecular; and (iv) there was a strong correlation between the type of transfer of the minus strand strong stop DNA and the number of template switches observed. These data suggest that retroviral recombination is ordered and occurs during the synthesis of both minus and plus strand DNA.

In the preceding paper we described an experiment that determined the in vivo forward mutation rate in a single replication cycle for spleen necrosis virus. In addition to substitutions, frameshifts, and hypermutations, the mutated proviruses contained two classes of deletions. One class of deletions contained short direct repeats at the deletion junctions. Another class of deletions had short stretches of sequences inserted at the deletion junctions. In this report, we describe the deletion mutations, and we present models for their generation. Detailed analysis of two deletions with insertions indicates that these mutations occurred as a result of template switching during plus-strand DNA synthesis. The analysis also indicates that fragments of viral RNA generated by the viral RNase H endonuclease are used as templates and contribute to the sequences inserted at the deletion junctions.

Somatic gene therapy of human disease with retrovirus vectors is a new technology with potentially important medical benefits. Although it involves recombinant DNA technologies and modified retroviruses, proper design of the vectors and delivery systems removes most potential foreseen risks. Furthermore, even in the very remote possibility that there is a nontherapeutic biological effect of the treatment, it is unlikely to be a harmful one. Thus, once very safe retrovirus vector–helper cell systems are constructed and in use, safety considerations should not hold up further human trials of retrovirus vectors.

We showed previously that retrovirus vector particles can encapsidate RNAs without retroviral cis-acting sequences, that such RNAs are reverse transcribed in infected target cells, and that the cDNA copies are inserted into the host genome resulting in cDNA genes (R. Dornburg and H. M. Temin, Mol. Cell. Biol. 8:2328-2334, 1988). To provide further evidence that this retrovirus-mediated gene transfer occurred through an RNA intermediate, we constructed retroviral vectors containing an intron from a cellular gene. This intron was lost in a cDNA gene formed after infection with retroviral particles, establishing that an RNA intermediate had existed. Retroviral vectors with additional encapsidation sequences were constructed. The presence of a murine leukemia virus encapsidation sequence in an mRNA transcribed from the hygromycin B phosphotransferase gene increased the efficiency of encapsidation into spleen necrosis virus vector particles and the formation of cDNA genes by approximately 2 orders of magnitude.

The level of genetic variation of human immunodeficiency virus type 1 (HIV-1), a member of the lentivirus genus of the Retroviridae family, is high relative to that of retroviruses in some other genera. The high error rates of purified HIV-1 reverse transcriptase in cell-free systems suggest an explanation for this high genetic variation. To test whether the in vivo rate of mutation during reverse transcription of HIV-1 is as high as predicted by cell-free studies, and therefore higher than that rates of mutation of retroviruses in other genera, we developed an in vivo assay for detecting forward mutations in HIV-1, using the lacZ alpha peptide gene as a reporter for mutations. This system allows the rates and types of mutations that occur during a single cycle of replication to be studied. We found that the forward mutation rate for HIV-1 was 3.4 x 10(-5) mutations per bp per cycle. Base substitution mutations predominated; G-to-A transition mutations were the most common base substitution. The in vivo mutation rates for HIV-1 are three and seven times higher than those previously reported for two other retroviruses, spleen necrosis virus and bovine leukemia virus, respectively. In contrast, our calculated in vivo mutation rate for HIV-1 is about 20-fold lower than the error rate of purified HIV-1 reverse transcriptase, with the same target sequence. This finding indicates that HIV-1 reverse transcription in vivo is not as error prone as predicted from the fidelity of purified reverse transcriptase in cell-free studies. Our data suggest that the fidelity of purified HIV-1 reverse transcriptase may not accurately reflect the level of genetic variation in a natural infection.

In order to define bovine leukemia virus (BLV) sequences required for efficient vector replication, a series ofmutationsweremadeinaBLVvector.Testingthereplicationefficiencyofthevectorswithahelpervirusand helperplasmidsallowedforseparationofthemutantvectorsintothreegroups.Thereplicationefficiencyofthe first group was reduced by a factor of 7; these mutants contained deletions in the 5*end of thegaggene. The second group of mutants had replication reduced by a factor of 50 and had deletions including the 5* untranslatedleaderregion.Thethirdgroupofmutantsreplicatedatlevelscomparabletothoseoftheparental vector and contained deletions of the 3* end of the gag gene, the pol gene, and the env gene. Analysis of cytoplasmic and virion RNA levels indicated that vector RNA expression was not affected but that the vector RNA encapsidation was less efficient for group 1 and group 2 mutants. Additional mutations revealed two regions important for RNA encapsidation. The first region is a 132-nucleotide-base sequence within the gag gene(nucleotides1015to1147oftheproviralDNA)andfacilitatesefficientRNAencapsidationinthepresence of the second region. The second region includes a 147-nucleotide-base sequence downstream of the primer binding site (nucleotide 551) and near thegaggene start codon (nucleotide 698;gagbegins at nucleotide 628) and is essential for RNA encapsidation. We conclude that the encapsidation signal is discontinuous; a primary signal, essential for RNA encapsidation, is largely in the untranslated leader region between the primer binding site and near thegagstart codon. A secondary signal, which facilitates efficient RNA encapsidation, is in a 132-nucleotide-base region within the 5*end of thegaggene. Retroviral vectors have been useful for studying various aspects of retroviral replication and for many applications of gene transfer. A retroviral vector contains all the cis-acting sequences necessary for retrovirus replication but is deficient in the production of some or all of the viral proteins necessary for replication and virus production (33). Retrovirus vectors were originally produced as virus particles by using a replication-competent virus as a helper. Retrovirus vectors are now commonly produced as virus particles, with retrovirus helper (or packaging) cells. Helper cells contain the coding sequences

Retrovirus genomes have a conserved modular organization that consists of trans-acting gag, pol, and env genes that function through cis-acting sequences to replicate the RNA genome to the DNA provirus. Genetically more complex retroviruses also encode regulatory genes and cis-acting sequences that are essential for their replication. We sought to convert a more complex retrovirus into a simpler retrovirus derivative that can replicate. We constructed novel, hybrid retrovirus vectors to replicate the gag, pol, and env genes of the more complex bovine leukemia virus (BLV) in the absence of regulatory genes and cis-acting response sequences. Most of the cis-acting sequences involved in the replication and regulation of BLV were replaced by the cis-acting transcriptional control sequences of a simpler retrovirus, spleen necrosis virus. We found that the resulting hybrid BLV derivatives can replicate independently of BLV Tax and Rex and the Tax and Rex cis-acting response sequences, as measured by successive passages of virus on target cells, detection of provirus sequences, and analyses of provirus and encapsidated RNAs.

Human immunodeficiency virus variation is extensive and is based on numerous mistakes in reverse transcription. All retrovirus replication requires two strand transfers (growing point jumps) to synthesize the complete provirus. I propose that the numerous mistakes in reverse transcription are the result of this requirement for the two strand transfers needed to form the provirus.

Human immunodeficiency virus type 1 (HIV-1) is a more complex retrovirus, coding for several accessory proteins in addition to the structural proteins (Gag, Pol, and Env) that are found in all retroviruses. More complex retroviruses have not been isolated from birds, and simpler retroviruses have not been isolated from humans. However, the proviruses of many endogenous simpler retroviruses are present in the human genome. These observations suggest that humans can mount a successful protective response against simpler retroviruses, whereas birds cannot. Thus, humans might be able to mount a successful protective response to infection with a simpler HIV-1. As a model, a simpler bovine leukemia virus which is capable of replicating has been constructed; a simpler HIV-1 could be constructed in a similar fashion. I suggest that such a simpler HIV-1 would be a safe and effective vaccine against HIV-1.

Retroviral vectors are widely used for the study of retroviral replication and to introduce DNA into somatic cells. An exciting new approach in retroviral vector technology is the use of internal ribosome entry sites from picornaviruses to provide stable expression of multiple genes. In addition, strategies are being developed that target the expression of retroviral vectors to specific cell populations.

Oncogenes discovered in retroviruses such as Rous sarcoma virus were generated by transduction of cellular proto-oncogenes into the viral genome. Several different kinds of junctions between the viral and proto-oncogene sequences have been found in different viruses. A system of retrovirus vectors and a protocol that mimicked this transduction during a single cycle of retrovirus replication was developed. The transduction involved the formation of a chimeric viral-cellular RNA, strand switching of the reverse transcription growing point from an infectious retrovirus to the chimeric RNA, and often a subsequent deletion during the rest of viral DNA synthesis. A short region of sequence identity was frequently used for the strand switching. The rate of this process was about 0.1 to 1 percent of the rate of homologous retroviral recombination.

To elucidate the mechanism(s) of retroviral recombination, we exposed virions to gamma radiation prior to infecting target cells. By using previously described spleen necrosis virus-based vectors containing multiple markers, recombinant proviruses were studied after a single round of retrovirus replication. The current models of retroviral recombination predict that breaking virion RNA should promote minus-strand recombination (forced copy-choice model), decrease or not affect plus-strand recombination (strand displacement/assimilation model), and shift plus-strand recombination towards the 3' end of the genome. However, we found that while gamma irradiation of virions reduced the amount of recoverable viral RNA, it did not primarily cause breaks. Thus, the frequency of selected recombinants was not significantly altered with greater doses of radiation. In spite of this, the irradiation did decrease the number of recombinants with only one internal template switch. As a result, the average number of additional internal template switches in the recombinant proviruses increased from 0.7 to 1.4 as infectivity decreased to 6%. The unselected internal template switches tended to be 5' of the selected crossover even in the recombinants from irradiated viruses, inconsistent with a plus-strand recombination mechanism.

A broad spectrum of mutations occurs at a high rate during a single round of retrovirus replication (V.K. Pathak and H. M. Temin, Proc. Natl. Acad. Sci. USA 87:6019-6023, 1990). We have now determined that this high rate of spontaneous mutation can be further increased by 5-azacytidine (AZC) treatment or by regions of potential RNA secondary structure. We found a 13-fold increase in the mutation rate after AZC treatment of retrovirus-producing cells and target cells. The AZC-induced substitutions were located at the same target sites as previously identified spontaneous substitutions. The concordance of the AZC-induced and spontaneous substitutions indicates the presence of reverse transcription "pause sites," where the growing point is error prone. An analysis of nucleotides that neighbored substitutions revealed that transversions occur primarily by transient template misalignment, whereas transitions occur primarily by misincorporation. We also introduced a 34-bp potential stem-loop structure as an in-frame insertion within a lacZ alpha gene that was inserted in the long terminal repeat (LTR) U3 region and determined whether this potential secondary structure increased the rate of retrovirus mutations. We found a threefold increase in the retrovirus mutation rate. Fifty-seven of 96 mutations were deletions associated with the potential stem-loop. We also determined that these deletion mutations occurred primarily during minus-strand DNA synthesis by comparing the frequencies of mutations in recovered provirus plasmids containing both LTRs and in provirus plasmids containing only one LTR.

Retroviruses mutate at a high rate during replication. We used a spleen necrosis virus-based vector system and helper cell line to characterize mutations occurring during a single round of retrovirus replication. The vector used, JD216HyNeo, codes for two drug resistance genes, hygromycin resistance (hygro) and neomycin resistance (neo). The downstream neo gene is expressed only when a mutation alleviates a block to splicing which is located in the upstream hygro gene. The mutations allowing splicing were large deletions, ranging in size from about 500 to about 2,000 bp. Most of the mutant proviruses lacked the encapsidation sequence, as shown by our inability to rescue the mutant proviruses with wild-type reticuloendotheliosis virus strain A and confirmed by Southern blotting and direct DNA sequence analysis. We therefore concluded that most of the deletions arose during reverse transcription in the target cell, rather than during transcription in the host cell. The sequence data also indicated that the deletions occurred by at least three different mechanisms: (i) misalignment of the growing point; (ii) incorrect synthesis and termination in the primer-binding sequence during synthesis of the plus-strand strong-stop DNA; and (iii) incorrect synthesis and termination before the primer-binding sequence during synthesis of the plus-strand strong-stop DNA. The second mechanism also led to the incorporation of cellular sequences into the proviral genome, pointing to a potential novel mechanism by which retroviruses can acquire cellular genes.

Primary transcripts of retroviruses contain two poly(A) sites, one near the 5' and one near the 3' end of the transcript, but only the 3' poly(A) site is used for 3' end formation of viral RNA. It was hypothesized on the basis of experiments with U3-deleted vectors of spleen necrosis virus that the U3 region contains sequences required for this RNA 3' end formation: the titer of a U3-deleted vector was 150 times lower than that of the parental vector, and the addition of the simian virus 40 poly(A) signal sequence increased the titer of the U3-deleted vector (J. P. Dougherty and H. M. Temin, Proc. Natl. Acad. Sci. USA 84:1197-1201, 1987). However, we now show that the U3 region transcribed from the 3' long terminal repeat is not required for RNA 3' end formation and that the experiments of Dougherty and Temin led to an erroneous conclusion. We show here that the deletion of the U3 region did not decrease the steady-state level of viral RNA or shift the site of poly(A) addition. The added simian virus 40 poly(A) signal sequence was used preferentially over the poly(A) signal of spleen necrosis virus, and it increased the levels of RNA transcribed from vectors with and without deletion of the U3 region. Our results indicate that alteration of regulatory sequences in retroviral vectors can change the steady-state RNA levels and titers of the vectors in an unpredictable manner.

We determined the in vivo forward mutation rate in a single replication cycle for spleen necrosis virus (SNV). A method was developed to clone integrated proviruses of retroviral shuttle vectors by exploiting the tight binding of the lac operator to the lac repressor protein. The vectors contained the lacZ alpha gene as a reporter of mutations. Thirty-seven of the 16,867 proviruses recovered contained five classes of mutations, including substitutions and frameshifts. Runs of 9 and 10 identical base pairs and a direct repeat of 110 base pairs were mutational hotspots. In addition, two copies of a provirus contained 15 G-to-A substitutions. Such proviruses, which we name hypermutants, may arise through the action of an error-prone polymerase and could significantly contribute to the genetic variation in retroviral populations.

Injected DNA proceeds with certain probabilities through the following steps: degradation by serum nucleases, adsorption to cells, uptake into cells, ligation to other DNA, mutation, expression of unintegrated DNA, integration, expression of integrated DNA, and activation of or inactivation of cellular genes. The maximal probability per DNA molecule of each of these steps is estimated based on experimental results in cell culture with transfection of DNA and with infection by retroviruses. A maximum cumulative probability of having a harmful effects is calculated to be less than 10(-16) to 10(-19) per DNA molecule from a cell without activated proto-oncogenes or active viral oncogenes. The most frequent harmful effects considered are inactivation of a tumour suppressor gene and activation of a proto-oncogene. Such inactivation and activation in a cell that could give rise to cancer would increase the age-standardized incidence of cancer by a small amount. The amount of increase would differ among individuals depending upon their genotypes and their environments. Thus, the magnitude of the increase will depend upon the frequency of more sensitive individuals. The probability of an increased incidence of cancer as a possible effect of the vaccination should be compared to the number of DNA molecules to be injected per person and to the protective effects of successful HBV vaccine.

Retroviruses contain two complete viral genomic RNAs in each virion. A system to study in a single round of replication the products of virions with two different genomic RNAs was established. A spleen necrosis virus-based splicing vector containing both the neomycin-resistance gene (neo) and the hygromycin B phosphotransferase gene (hygro) was used. Two frameshift mutants were derived from this vector such that the neo and the hygro genes were inactivated in separate vectors. Thus, each vector confers resistance to only one selection. The vectors with frameshift mutations were separately propagated and were pooled to infect DSDh helper cells. Doubly resistant cell clones were isolated, and viruses produced from these clones were used to infect D17 cells. This protocol allowed virions containing two different genomic RNAs (heterozygotes) to complete one round of retroviral replication. The molecular nature of progeny that conferred resistance to single or double selection and their ratio were determined. Our data demonstrate that each infectious heterozygous virion produces only one provirus. The rate of retroviral recombination is approximately 2% per kilobase per replication cycle. Recombinant proviruses are progeny of heterozygous virions.

Retroviral proteins can encapsidate RNAs without retroviral cis-acting sequences. Such RNAs are reverse transcribed and inserted into the genomes of infected target cells to form cDNA genes. Previous investigations by Southern blot analysis of such cDNA genes suggested that they were truncated at the 3' and the 5' ends (R. Dornburg and H. M. Temin, Mol. Cell. Biol. 8:2328-2334, 1988). To analyze such cDNA genes further, we cloned three cDNA genes (derived from a hygromycin B phosphotransferase gene) in lambda vectors and analyzed them by DNA sequencing. We found that they did not correspond to the full-length mRNA: they were truncated at both the 3' and the 5' ends, did not contain a poly(A) tract, and were not flanked by direct repeats. The 3'-end junctions to chromosomal DNA of five more cDNA genes were amplified by polymerase chain reaction, cloned in pUC vectors, and sequenced. All of these cDNA genes had 3'-end truncations, and no poly(A) tracts were found. Further polymerase chain reaction experiments were performed to detect hygromycin B phosphotransferase cDNA genes with a poly(A) tract in DNA extracted from a pool of about 500 colonies of cells containing cDNA genes. No hygromycin B phosphotransferase cDNA gene with a poly(A) tract was found. Investigation of two preintegration sites by Southern analysis revealed that deletions were present in chromosomal DNA at the site of the integration of the cDNA genes. Naturally occurring processed pseudogenes correspond to the full-length mRNA, contain a poly(A) sequence, and are flanked by direct repeats. Our data indicate that cDNA genes formed by infection with retrovirus particles lack the hallmarks or natural processed pseudogenes. Thus, it appears that natural processed pseudogenes were not generated by retrovirus proteins.

Summary Retroviral packaging cell lines were constructed by using the gag-pol gene of spleen necrosis virus, the gag-pol gene of Moloney murine leukaemia virus and the env gene of bovine leukaemia virus. The plasmids containing the gag-pol genes and the plasmid containing the env gene were cotransfected into NIH/3T3 and D17 cells. The cells containing the helper virus constructs were tested for their ability to package replication-defective murine leukaemia and avian reticuloendotheliosis retrovirus vectors. The titre of vector virus produced by each of the retroviral packaging cell lines was about 102 colony-forming units per ml of medium. Tests for events that might result in intact replication-competent retroviruses showed no evidence for the generation of such viruses. The vector viruses were able to infect dog and rat cells. Bovine cells were infected only after their cocultivation with the retroviral packaging cell lines producing murine leukaemia virus vectors, perhaps as a result of a low concentration of receptors.

etiology of human cancer. In fact, pathologists called the chicken leukemias "leukoses" to emphasize their feeling that a disease caused by a virus was not a true cancer. However, as a result of studies in the last 20 years, the hypothesis that viruses directly cause much human cancer has received a great deal of attention (Gross 1970). But, in spite of this attention, we have learned at most that a virus may be directly involved in Burkitt's lymphoma and in nasopharyngeal cancer and that it is possible viruses may be involved in some human leukemias. However, for the great majority of human cancers there is at present no evidence for direct involvement of infectious viruses. Was it then correct to neglect for so many years the viruses that cause cancer in animals? In this article I shall attempt to answer that question by looking at our present understanding of the role of viruses in causing some animal cancers, the role of chemicals in causing cancer, and a possible analogy between cancer viruses and other carcinogens. VIRUSES AND CANCER IN ANIMALS

Stationary chicken embryo fibroblasts exposed to Rous sarcoma virus (RSV) did not synthesize virus RNA until after division of the infected cells. Initiation of RNA transcription after cell division resulted in transcription of the entire viral genome and was followed shortly thereafter by appearance of progeny virus. Addition of colchicine to RSV-infected stationary chicken embryo fibroblasts resulted in inhibition of cell division and a 24-h delay in the appearance of virus RNA. RSV-infected chicken embryo fibroblasts made stationary 10 to 15 cell generations after infection continued to produce an amount of progeny virus similar to that produced in dividing RSV-infected cells. Teh virus produced from cells made stationary after virus production had begun contained RNA synthesized in stationary cells. Therefore, division of the RSV-infected chicken embryo fibroblast is required for the initiation of transcription of virus RNA, but it is not required for the maintenance of transcription of virus RNA.

The minimal cis-acting sequences outside the long terminal repeat (LTR) required for formation of an infectious retrovirus cloning vector were determined with recombinants of spleen necrosis virus (SNV) DNA and herpes simplex virus type 1 thymidine kinase gene. The 3' end of SNV DNA was removed to within 40 base pairs (bp) from the 3' LTR with only a 2-fold effect on the recovery of infectious recombinant virus. However, when the 5' end of SNV DNA was removed to within 100 bp from the 5' LTR, infectious recombinant virus was not recovered. Deletion mutants constructed around this latter region showed that nucleotides between 100 and 285 bp from the 5' LTR are necessary for encapsidation of genomic viral RNA. We call this region required for encapsidation E.

Recombination between viral and cellular genes can give rise to new strains of retroviruses. For example, Rous-associated virus 61 (RAV-61) is a recombinant between the Bryan high-titer strain of Rous sarcoma virus (RSV) and normal pheasant DNA. Nucleic acid hybridization techniques were used to study the genome of RAV-61 and another RAV with subgroup F specificity (RAV-F) obtained by passage of RSV-RAV-0 in cells from a ring-necked pheasant embryo. The nucleotide sequences acquired by these two independent isolates of RAV-F that were not shared with the parental virus comprised 20 to 25% of the RAV-F genomes and were indistinguishable by nucleic acid hybridization. (In addition, RAV-F genomes had another set of nucleotide sequences that were homologous to some pheasant nucleotide sequences and also were present in the parental viruses.) A specific complementary DNA, containing only nucleotide sequences complementary to those acquired by RAV-61 through recombination, was prepared. These nucleotide sequences were pheasant derived and were not present in the genomes of reticuloendotheliosis viruses, pheasant viruses, and avian leukosis-sarcoma viruses of subgroups A, B, C, D, and E. They were partially endogenous, however, to avian DNA other than pheasant. The fraction of these nucleotide sequences present in other avian DNAs generally paralleled the genetic relatedness of these avian species to pheasants. However, there was a high degree of homology between these pheasant nucleotide sequences and related nucleotide sequences in the DNA of normal chickens as indicated by the identical melting profiles of the respective hybrids.

We have constructed retrovirus vectors containing two genes, each with its own promoter, whose products can be selected in vivo. In cell clones harboring a single copy of this bicistronic provirus, we find that the 3' gene is usually suppressed when there is selection for expression of the 5' gene, and the 5' gene is usually suppressed when there is selection for expression of the 3' gene. We find that the suppression is epigenetic, reversible, and cis-acting, and is influenced by the strength or position of the different promoters. This type of gene suppression is novel for its action on a gene that can be either upstream or downstream to an adjacent active gene.

Different experiments with viruses and transfection now support the classical view that cancer is the result of a multistep process. This analysis further indicates that some of these steps involve mutations affecting the qualitative and quantitative expression of dominant transforming genes or oncogenes. These mutations are spontaneous or induced and of various kinds, including base pair changes, deletions, translocations, and amplifications. The actions of the active transforming genes or oncogenes lead to the properties of the tumor cell. However, these activities are effective only in the appropriate cell with targets for the products of the oncogenes and without inhibitors. Because there will be multiple genetic changes in tumor cells, it is difficult to determine which changes are significant for the oncogenesis. Retrovirus vectors may be useful in this determination. In addition, our present methods of analysis may be missing certain of the multiple steps in oncogenesis, in particular, those involved with tissue-, organ-, and organism-specific controls.

We prepared retrovirus packaging cell lines containing gag-pol genes from spleen necrosis virus (expressed from a cytomegalovirus promoter and the simian virus 40 (SV40) polyadenylation sequences) and, on a separate vector, either the env gene from spleen necrosis virus (expressed from the Rous sarcoma virus promoter and the SV40 polyadenylation sequences) or the env gene from amphotropic murine leukemia virus (expressed from a cytomegalovirus promoter and the SV40 polyadenylation sequences). The nucleotide sequences in these packaging cell lines have almost no homology to the retrovirus vectors we used. Retrovirus vectors were produced from these new helper cell lines without any genetic interactions between the vectors and sequences in the helper cells and without transfer of the packaging sequences.

The infectious DNAs of a number of avian leukosis-sarcoma and reticuloendotheliosis viruses were digested with six nucleotide-specific restriction endonucleases, and the digests were tested for infectivity. All of the enzymes inactivated the viral infectivities except for EcoRI, which did not inactivate the infectivity of the DNA of two of the reticuloendotheliosis viruses, spleen necrosis and chick syncytial viruses. The infectious DNA of spleen necrosis virus after digestion with EcoRI had a buoyant density in CsCl solution greater than the density of the high-molecular-weight infectious viral DNA. The infectious EcoRI-digested spleen necrosis virus DNA from chronically infected chicken cells was uniform in size, 10 megadaltons, which indicated a single site of integration. The infectious EcoRI-digested spleen necrosis virus DNA from acutely infected cells was heterogeneous in size, ranging from 8-14 megadaltons, which indicated multiple sites of integration. These results are consistent with the hypothesis that cells that integrate infectious spleen necrosis virus DNA at a single site survive and multiply, whereas cells that integrate infectious viral DNA at additional sites either die or selectively lose or inactivate the DNA in the additional sites.

I have discussed the observations and experiments that led to the formulation and establishment of the provirus hypothesis and the DNA provirus hypothesis, which includes RNA-directed DNA synthesis for the formation of the provirus. I have also discussed some aspects of the present status of our knowledge of the mechanism of formation of the DNA provirus both to point out the work remaining to be done and to illustrate hypotheses for the origins of ribodeoxyviruses and the origins of other animal enveloped RNA viruses and of animal small DNA viruses. Finally, I have indicated that I do not believe that infectious viruses cause most human cancers, but I do believe that viruses provide models of the processes involved in the etiology of human cancer.

A retroviral vector system was developed to study the retrotransposition of RNAs lacking all cis-acting sequences required for normal retroviral replication. Our experiments indicate that such RNAs can be encapsidated in retroviral proteins, reverse transcribed, and integrated to form functional cDNA genes in infected cells. The frequency of this process, however, was approximately 8 orders of magnitude less than that of normal retroviral replication. The efficiency was limited at each step in this process. Investigation of seven cDNA genes by Southern blot analysis revealed that all of them were truncated at either the 3' or the 5' end or both. These truncations are not seen with natural cDNA genes and raise the question of retroviral involvement in their formation.

The genome of the highly oncogenic avian retrovirus reticuloendotheliosis virus strain T (REV-T) differs from that of the helper virus reticuloendotheliosis virus strain A by a substitution ( rel ) and a large deletion. Further deletions, constructed in vitro, within the helper-virus-related sequences of REV-T have little effect on the ability of the virus to transform chicken spleen cells in vitro. However, deletions that extend into rel abolish transformation. Substitution of helper-virus-related sequences for the deleted region in the non- rel portion of REV-T also abolishes transformation. Viruses with revertant phenotype were obtained both spontaneously and by construction in vitro from these substituted recombinants. The revertant viruses have various mutations, including deletions and insertions, in the helper-virus-related sequences. Thus the additional helper-virus-related sequences suppress expression of transformation in cis, and the deletion in REV-T seems necessary for expression of the transforming properties of the virus.

Some animal viruses that contain RNA replicate through a DNA intermediate. The molecular details of the replication of these viruses, which are called ribodenoxyviruses, are starting to be known. The ribodenoxyviruses belonging to a single species may either cause sarcomas, leukemia or no disease. The viruses belonging to a single species differ only in whether or not they contain genes for disease formation. In the case of Rous sarcoma virus, the virus causes sarcomas by adding a gene for sarcoma formation to the genome of infected cells. Ribodeoxyviruses appear to undergo different kinds of genetic changes at extraordinarily high rates. In addition, nucleotide sequences related to ribodeoxyvirus RNA are present in the DNA of many uninfected cells. These nucleotide sequences may represent a virus precursor, and ribodeoxyviruses are hypothesized to have evolved from these nucleotide sequences in uninfected cells. These data have led us to hypothesis that non-viral carcinogens act to mutate a cellular gene(s) that is involved in the same types of information transfer and genetic variation as ribodeoxyviruses and thus give rise to the formation of cancer gene(s).

Loss of intervening sequences in genomic mouse α-globin DNA inserted in an infectious retrovirus vector

We mutagenized cloned spleen necrosis virus DNA to identify a region of the retrovirus genome encoding a polypeptide required for integration of viral DNA. Five plasmids bearing different lesions in the 3' end of the pol gene were examined for the ability to integrate or replicate following transfection of chicken embryo fibroblasts. Transfection with one of these DNAs resulted in the generation of mutant virus incapable of integrating but able to replicate at low levels; this phenotype is identical to that of mutants bearing alterations in the cis-acting region, att. To determine whether the 3' end of the pol gene encodes a protein that interacts with att, we did a complementation experiment. Cells were first infected with an att- virus and then superinfected with the integration-deficient virus containing a lesion in the pol gene and a wild-type att site. The results showed that the att- virus provided a transacting function allowing integration of viral DNA derived from the mutant bearing a wild-type att site. Thus, the 3' end of the pol gene serves as an "int" locus and encodes a protein mediating integration of retrovirus DNA through interaction with att.

The pattern of integration of spleen necrosis virus (SNV) DNA in DNA from a large population of SNV-infected chicken cells was studied by nucleic acid hybridization with iodinated viral RNA by the blotting technique of Southern. SNV DNA was found to be integrated at multiple sites in acutely infected chicken cells. Concomitant with the transition from acute to chronic infection, a shift in the pattern of integration was observed. The majority of integrated SNV DNA found in acutely infected cells was absent from chronically infected cells. This result is consistent with the hypothesis that the cell death that occurs after infection of avian cells with reticuloendotheliosis viruses is a consequence of the multiple integrations of the provirus. Viral DNA was also integrated at multiple sites in chronically infected cells. However, infectious viral DNA molecules in chronically infected cells migrated in a uniform manner in agarose gel electrophoresis after EcoRI digestion (which does not cut viral DNA), indicating that not all integrated SNV copies are equally infectious.

Previously, we described a retrovirus vector that contained two genes: a 5' gene under transcriptional control of the viral long terminal repeat and a 3' gene under transcriptional control of the herpes simplex virus thymidine kinase promoter. By using a biological assay, we found that expression of the 5' gene is suppressed when there was selection for the 3' gene and expression of the 3' gene is suppressed when there is selection for the 5' gene (M. Emerman and H. M. Temin, Cell 39:459-467, 1984). In the present study, we replaced the thymidine kinase promoter with stronger promoters, and we measured expression of the genes in the retrovirus vectors by enzyme activity and RNA analysis. We found that all of the vectors displayed gene suppression when analyzed biochemically, although not when analyzed biologically. The suppressed genes produced about 10 to 50% as much product as when they were selected. The amount of suppression depended on whether the suppressed gene was 5' or 3' to the selected gene and from which promoter the suppressed gene was transcribed. The amount of suppression correlated with a decrease in the amount of steady-state RNA transcribed from the suppressed gene's promoter.

Specific serological relationships were found among the partially purified DNA polymerases of the two groups of avian viruses whose virions contain RNA and a DNA polymerase—the avian leukosis-sarcoma viruses and the reticuloendotheliosis viruses—and three avian species which are natural hosts for these viruses: chickens, turkeys, and Pekin ducks. No relationships were found to DNA polymerases of HeLa cells or Escherichia coli. These results are consistent with the hypothesis that RNA viruses with a DNA polymerase originated from normal cellular components.

Summary After an initial acute infection with cell killing, chicken or duck embryo fibroblasts infected in culture with reticuloendotheliosis viruses set up a chronic infection with no cell killing or morphological transformation. Essentially all of the chronically infected cells produced virus. The virus production was not sensitive to cytosine arabinoside or mitomycin C as was virus production in an acute infection. The chronically infected cells had a strong group-specific resistance to the c.p.e. of superinfecting reticuloendotheliosis viruses. However, they were sensitive to vesicular stomatitis virus and avian leukosis-sarcoma viruses. After double infection, single cells produced reticuloendotheliosis virus and avian leukosis-sarcoma virus.