INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Vectors based on retroviruses are currently one of the most widely used systems for gene transfer, both in experimental studies and in clinical trials (13, 27). Their production requires two components: the replication-defective vector and a packaging cell line. Construction of the vector is based on a retroviral genome from which viral genes have been removed, retaining only essential elements required in cis to allow packaging, reverse transcription, and integration into the genome of target cells. The viral functions required for replication are provided in trans by the packaging line. The prototypes of such lines contain the proviral DNA of a retrovirus from which the packaging signal has been excised (10), but this does not prevent encapsidation at low frequencies, resulting in transmission of the viral helper functions to target cells. Moreover, copackaging of the helper genome and vector permit recombination, and only one such event may be required to reacquire the packaging signal and restore replication competence.

The inadvertent production of replication-competent retroviruses constitutes one of the major safety issues concerning the use of retroviral vectors (2, 29). Such viruses may lead to chronic viremia and subsequently to formation of malignant tumors as a result of insertional mutagenesis (4, 19). This hazard has been demonstrated in monkeys which were intentionally given replication-competent murine leukemia viruses and which later developed lymphomas (6). Therefore, there has been much effort to improve the design and construction of packaging lines to minimize such risks. Also of concern is a recent report that amphotropic replication-competent retroviruses are involved in the pathogenesis of a spongiform encephalopathy in mice (16). In many of the packaging lines used currently, the helper genes gag-pol and env are introduced as separate transcriptional units, so that at least three recombination events would be needed to form a replication-competent virus. The widely used amphotropic packaging line GP+envAM12 (11), which is also used routinely in our laboratory, is an example of such a "split-function" packaging line. This line has also been used in clinical trials (22). Since recombination events with the vector are more frequent at regions of homology (9, 30), the risks may be reduced further by using vectors with minimal overlap with the helper sequences, as in the pBabe family of vectors used in our laboratory (15).

Previously, we reported finding, on routine screening by a marker rescue assay, the presence of helper virus in stocks of pBabeNeo vector obtained from a producer cell line derived from GP+envAM12 (3). These vector stocks were used to infect K1735 and B16 murine melanoma cells to obtain G418-resistant clones. Supernatants from G418-resistant K1735 and B16 cells were able to transfer G418 resistance to other cultures of fresh cells, indicating the presence of helper virus in the supernatant from the producer line. G418 resistance could easily be transferred by supernatant to further cultures of fresh cells through several passages, suggesting that the helper virus was replication competent.

This breakout of helper virus was significant since no such viruses have previously or since been detected in our laboratory, where stringent routine screening protocols have always been in operation. Also of significance is the fact that replication-competent retroviruses have never previously been used in our laboratory. In addition, since the discovery of the helper virus, stringent screening of all routinely used cell lines in the laboratory has failed to show any such virus which may have been imported into our stocks without our knowledge. Therefore, the possibility of adventitious cross-contamination by an existing virus was very unlikely. We now describe further characterization of this helper virus, focusing on the recombination events that led to its formation. This information has important implications for the design and use of retroviral vector systems. The findings also emphasise the importance of routine screening of retroviral vector stocks, even when using split-function packaging lines which have been designed to minimize such occurrences.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cell culture. All cell lines were grown in Dulbecco's modified Eagle's minimal essential medium supplemented with 10% (vol/vol) fetal calf serum.

Hybridization of RNA blots (Northern blots). RNA was extracted from cells with an RNAzol B kit (Biogenesis, Dorset, United Kingdom). The RNA was separated in formaldehyde denaturing gels and transferred onto Hybond-N⁺ membranes (Amersham, Little Chalfont, United Kingdom). Double-stranded DNA fragments for probing RNA blots were labelled with mixed hexadeoxyribonucleotide primers of random sequence, using reagents supplied in a kit (Pharmacia Biotech, Milton Keynes, United Kingdom). Denatured radiolabelled probes were incubated with the blots for 16 h at 42°C.

PCR amplification. Oligonucleotide primers were synthesized on an Applied Biosystems 380B synthesizer. The nucleotide sequences of the primers used in this study are listed in Table 1. Genomic DNA was obtained from cell lines with a Nucleon II DNA extraction kit (Scotlab, Strathclyde, United Kingdom). PCR was performed in all cases with Taq DNA polymerase, except when amplifying sequences of >3 kb, when rTth DNA polymerase from a GeneAmp XL PCR kit (Perkin-Elmer, Norwalk, Conn.) was used. For nucleotide sequencing, PCR-amplified fragments were ligated into the pCRII vector with a TA cloning kit (Invitrogen, NV Leek, The Netherlands).

Reverse transcriptase-PCR (RT-PCR). RNA was extracted from cells as described above. First-strand cDNA was generated with a first-strand cDNA synthesis kit (Pharmacia Biotech). The reaction mixture was then used for PCR analysis, as above.

Nucleotide sequencing of DNA. DNA was prepared with a miniprep or maxiprep kit (Qiagen, Crawley, United Kingdom). Nucleotide sequencing was performed conventionally by a dideoxynucleotide chain termination method with a Sequenase version 2.0 kit (Amersham) or on an Automated Laser Fluorescent ALF DNA sequencer (Pharmacia Biotech) or an ABI 373A DNA sequencer. To facilitate sequencing of long DNA fragments, sets of nested deletions were constructed by controlled digestion of DNA with exonuclease III, using reagents supplied in a nested-deletion kit (Pharmacia Biotech).

Nucleotide sequence accession numbers. The nucleotide sequence data referred to in this report have been submitted to the GenBank nucleotide sequence database and have been assigned accession no. AF034782, AF034783, AF034784, and AF034785.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

The full-length genome of the helper virus is approximately 8 kb long. To examine the genomic structure of the helper virus, RNA was extracted from the helper virus-infected cell line K1735-puro/neo. This line represents a pool of G418-resistant clones resulting from exposure of K1735-puro cells to supernatant from GP+envAM12/pBabeNeo producer cells. K1735-puro cells are K1735 murine melanoma cells which had been transfected previously with the pBabePuro plasmid. RNA was also prepared from pools of G418-resistant clones of NIH 3T3 cells and MeWo human melanoma cells that had been exposed to supernatant obtained from K1735-puro/neo cells. Previously, by using receptor interference studies, we showed that the helper virus displayed an amphotropic host range (3). Therefore, Northern blots of RNA from the infected cell lines were probed with the 4070A env gene, which was excised from plasmid FB4070A SALF (kindly provided by F.-L. Cosset, Villeurbanne, France). Blots of RNA from the infected cell lines showed strongly positive signals at approximately 8 and 3 kb; the infected NIH 3T3 cells also showed a faint signal at approximately 1.5 kb (Fig. 1) (data not shown for MeWo cell lines). Significantly, these signals were not present in the blots of RNA from the corresponding parental lines. The 8-kb RNA corresponds to a size which is compatible with a full-length genome of a retrovirus, while the 3-kb fragment is expected to represent a subgenomic spliced segment which includes the full 4070A env gene (Fig. 2a). The 1.5-kb signal may represent a spliced fragment formed by splicing at cryptic sites.

View larger version (36K): [in this window] [in a new window]   FIG. 1.   Northern blots of RNA extracted from parental and helper virus-infected cells. RNA was obtained from K1735-puro/neo cells, parental K1735 cells, helper virus-infected NIH 3T3 cells, and parental NIH 3T3 cells. Northern blots of the RNA were probed with 4070A env (top panel). The blot was stripped and probed with neo (middle panel) and then, as a control, probed with the gene for murine glyceraldehyde-3-phosphate dehydrogenase (G3PDH) (bottom panel).

Recombination between 4070A env from the packaging construct and the 3' LTR of the vector. To identify the site of recombination, a segment linking env and the 3' long terminal repeat (LTR) in the helper virus genome was amplified by PCR from genomic DNA of helper virus-infected NIH 3T3 cells. The choice of oligonucleotide primers for this reaction was based on the predicted structure of the genome. The forward primer, RV152, corresponded to a sequence in the pol/env overlap region (Fig. 2a). The fourth nucleotide in the primer sequence represents the first coding nucleotide of env. At this location, the primer sequence corresponds exactly to that of both 4070A env and Moloney murine leukemia virus (MoMLV) env, but the sequences of these two viruses diverge downstream. The reverse primer RV116 was based on a sequence in the U3 region of the MoMLV LTR (Fig. 2a). A signal of the predicted size (approximately 2 kb) was obtained from infected NIH 3T3 cells but was not present in the reactions with DNA of parental NIH 3T3 cells or GP+envAM12 packaging cells (Fig. 3a). To show that this recombination was also present in viral RNA, RNA was extracted from the helper virus-infected cell lines and used for first-strand cDNA synthesis by reverse transcription and PCR with the same primer pairs (RT-PCR). The reactions with RNA from the infected cells produced signals of the same size, whereas samples from the uninfected parental cells were negative (data not shown).

View larger version (16K): [in this window] [in a new window]   FIG. 2.   Genomic structure of helper virus and the pBabeNeo vector. (a) Schematic diagram of the helper virus genome showing the locations of PCR primers used in this study (not drawn precisely to scale). The subgenomic transcript bearing the env gene is also depicted. s.d., splice donor site; s.a., splice acceptor site. (b) Schematic diagram of the pBabeNeo vector (15). A subgenomic transcript containing neo, driven by the simian virus 40 (SV40) early promoter, would also be obtained.

View larger version (44K): [in this window] [in a new window]   FIG. 3.   (a) Amplification of an env-3' LTR fragment by PCR with DNA from helper virus-infected NIH 3T3 cells. (A) As controls for the presence of genomic DNA, PCR was performed with primers for the promoter of the murine tyrosinase gene, with DNA of GP+envAM12 cells and parental and helper virus-infected NIH 3T3 cells. (B) An env-LTR fragment was amplified with primers RV152 plus RV116. A positive signal was obtained only in the reaction with DNA of infected NIH 3T3 cells. (b) Amplification of a 5' LTR-gag-pol fragment by PCR with DNA from helper virus-infected cells. Primer RV150 corresponds to a sequence in MoMLV LTR, and RV151 is at the pol/env overlap region. PCR with DNA of GP+envAM12 showed two signals, consistent with amplification from the two packaging constructs. The reaction with DNA of infected NIH 3T3 cells showed one signal of approximately 5.7 kb, which was absent in the reaction with parental NIH 3T3 cells. (c) Amplification of a pol-4070A env fragment by PCR with DNA from helper virus-infected cells. For PCR with primers RV169 plus RV170, several signals, the largest and most prominent of which was approximately 2.7 kb, were obtained in the reaction with DNA of infected NIH 3T3 cells. These were absent in the other samples. For PCR with primers RV169 plus RV151, signals were present in the reactions with DNA from infected NIH 3T3 cells and GP+envAM12. A positive signal was also obtained with plasmid pCRII-c11, which bears the cloned PCR-amplified fragment obtained from the reaction shown in panel b.

View larger version (22K): [in this window] [in a new window]   FIG. 4.   Nucleotide sequences at recombination sites in helper virus. (a) Recombination site between 4070A env and MoMLV sequences derived from the pBabeNeo vector. The sequences shown here span the 3' end of env, the downstream untranslated segment, and the 5' end of the 3' LTR. The following sequences are depicted: MoMLV (in the pBabeNeo vector, the neo gene is linked to nontranslated MoMLV sequences 3' of env), the final 12 coding nucleotides of 4070A env, Friend MLV (in the packaging construct penvAm, present in GP+envAM12 cells, the 3' end of 4070A env is linked downstream to a sequence derived from Friend MLV), and the sequence of the helper virus. Comparisons of these sequences suggest that a recombination event has occurred between the 3' end of 4070A env and MoMLV sequences in the untranslated region (from pBabeNeo), as indicated by the arrow. The bases in the Friend MLV sequence which are printed in boldface type and underlined indicate differences compared with the helper virus sequence, suggesting that the Friend MLV sequence does not contribute to the genome of the helper virus. (b) Recombination between endogenous polytropic sequences (which contribute to MCF-type viruses) and 4070A amphotropic virus. Nucleotides 5638 to 5696 of the 5' LTR-gag-pol PCR-amplified fragment (from Fig. 3b), corresponding to the 3' portion of pol, are shown and compared with the corresponding sequences from the MCF-type virus pRFM#6 (this represents one of several MCF-type sequences in the database to which the helper virus bears >99% homology at this location) and the 4070A amphotropic retrovirus. The nucleotides which differ between the helper virus and either of these two other viral sequences are underlined and printed in boldface type. The shaded box represents a possible location where recombination between the two viral sequences to form the helper virus sequence may have occurred.

Recombination between gag-pol derived from endogenous retroviral sequences and 4070A env from the packaging construct. To study the recombination events which generated the 5' portion of the helper virus genome, the 5' LTR-gag-pol region was amplified by PCR with DNA from infected NIH 3T3 cells. The forward primer, RV150, was located in the R region of MoMLV LTR, while the reverse primer, RV151, lay in the pol/env overlap region, and was complementary to primer RV152, which had been used previously (Fig. 2a). A single strong band of approximately 5.7 kb was obtained with DNA of infected NIH 3T3 cells (Fig. 3b). The length of this band was consistent with the combined size of the retroviral 5' LTR, gag and pol genes. A number of other smaller weak bands were also noted when the gel was viewed under higher-intensity UV transillumination, as was also seen in the reaction with DNA from parental NIH 3T3 cells. It is most likely that these faint signals represent amplification of endogenous retroviral sequences present in murine cells, resulting from weak hybridization of primers to these sequences. A control reaction performed with DNA from GP+envAM12 cells revealed the expected two bands, which represent amplification of a 0.8-kb product from the penvAm packaging construct and a 5.7- kb fragment, consisting of 5' LTR-gag-pol, from the pgag-polgpt packaging construct.

View larger version (16K): [in this window] [in a new window]   FIG. 5.   Structure of the helper virus genome as suggested by the PCR and RT-PCR studies. The virus arose by recombination events between endogenous retroviral sequences, the transcomplementing packaging construct penvAm, and the pBabeNeo vector, which is derived from MoMLV.

The recombination events which generated the helper virus were recent events. At least two recombination events led to the formation of the helper virus. To investigate whether the exchange between the endogenous retroviral sequence and the 4070A sequence in penvAm had already occurred in the stock of GP+envAM12 cells used to create the producer line, PCR was performed on DNA from these cells with the forward primer HC4, which was located within the endogenous polytropic sequence, and the reverse primer RV170, which was positioned within 4070A env (Fig. 2a). No amplification was achieved with DNA of GP+envAM12 cells, indicating that the link between these sequences was not present in the stock of packaging cells (Fig. 6). This result is also consistent with the lack of amplification when primer pairs RV169 and RV170 were used (Fig. 3c). In contrast, a band of the predicted size was obtained with DNA from infected NIH 3T3 cells and with the primer pair HC4 and RV170 (Fig. 6). It is also significant that no signal was produced in the reaction with DNA from an "early" stock of GP+envAM12/pBabeNeo producer cells. These cells were retrieved from a batch which was frozen at the time when the producer cells were established initially. In contrast, the producer cells which released the helper virus represent a later passage of this line that had been passaged for 3 weeks, during which time the viral supernatant had been harvested. As is routine practice in our laboratory, these cells were discarded after this period, long before it became apparent that they were releasing helper viruses. Therefore, they were not available for inclusion in the above PCR studies. Similar PCR studies were used to examine recombination between 4070A env and the 3' LTR of the vector. These also indicate that this recombination had not occurred in the early stock of GP+envAM12/pBabeNeo cells (data not shown). Therefore, these findings suggest that both exchanges were relatively recent events, which took place during the 3-week period when the producer cells were being passaged. It is likely that the helper virus arose in the producer cells rather than the target cells, since the virus was present in two independent cultures of target cells (3).

View larger version (34K): [in this window] [in a new window]   FIG. 6.   Amplification of a fragment linking endogenous polytropic retroviral sequences and 4070A env. Primer HC4 identifies with the endogenous polytropic retroviral sequence, while RV170 lies in 4070A env. A signal was obtained with DNA from infected NIH 3T3 cells. No amplification was achieved in reactions with DNA from parental NIH 3T3 cells, GP+envAM12 cells, and an early stock of GP+envAM12/pBabeNeo producer cells. As a positive control, plasmid pCRII-c3 was used, which contains the cloned PCR-amplified fragment linking the endogenous ecotropic sequence to 4070A env, as obtained from the reaction in Fig. 3c.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Previously, we described the presence of helper virus in vector stocks from a producer line derived from a split-function packaging line (3). The viral packaging functions were spread easily to fresh cultures of cells, suggesting that the virus was replication competent. Here, we show that the virus was generated by recombination events which involve the vector, a transcomplementing packaging construct, and endogenous retroviral sequences (Fig. 5). It is very unlikely that this virus represents cross-contamination by a preexisting virus, since replication-competent retroviruses have never been used in our laboratory. Routine screening protocols of all vector stocks have always been performed in our laboratory, and helper viruses have never before or since been identified. Screening of all cell lines which are cultured regularly in our laboratory did not indicate the presence of such viruses, making it unlikely that the helper virus we identified preexisted in one of the cell line stocks. Moreover, all the experiments described in this study included control reactions with DNA or RNA derived from the corresponding parental cell lines, and the recombinant structures were not identified in any of these.

A number of improved packaging cell lines which yield higher viral titers or pseudotyped particles for a variety of specific purposes have been constructed (5, 14, 20, 25). All of these are based on the safety concept of split function to minimize the chances of producing replication-competent viruses. Our findings do not devalue this safety device. In contrast, the extreme rarity of replication-competent viruses arising from split-function lines emphasises the success of this approach. A spleen necrosis virus-based split-function packaging line has been reported to release a helper virus which spread inefficiently (12). However, a 1.3-kb overlap region existed between the two packaging constructs in this line, and these had been introduced into the cells simultaneously, increasing the chances of recombination. Indeed, recombination between gag-pol and env had already occurred before transfection with the vector construct. Others have found that an avain leukosis virus-based split-function packaging line released replication-defective particles bearing portions of the packaging constructs (7, 8). It may be expected that transfer of such helper sequences will increase the chance of further recombination events which could lead to replication competence.

Since even short stretches with partial homology, of 8 to 10 nucleotides, favor recombination (18), homologous regions between vector and transcomplementing sequences have been virtually eliminated in improved retroviral vector systems (5, 20). Although this measure would be expected to reduce markedly the chances of recombination between these elements, such events may still occur in nonhomologous regions (8). Also, it is difficult to avoid a small region of homology between the two separate packaging constructs, since the 3' end of pol overlaps with the 5' end of env. It is not possible to obviate completely the infrequent encapsidation of helper sequences (7) or endogenous retroviral sequences (21, 23). The genomes of murine cells contain endogenous retroviral sequences (1), some of which are expressed, and since these bear some homology to vector or packaging sequences, they provide opportunities for recombination (26). Recently, packaging cell lines have been constructed from nonmurine cells (5, 20), and since these carry fewer endogenous retroviral sequences, such occurrences may be minimized. However, it is unlikely that these cell lines would be completely devoid of retroviral or retrovirus-like elements, although the elements would probably have less homology to MLV-based vectors. Therefore, even as the design of retroviral systems continues to improve and the chance of replication-competent virus breakout becomes ever smaller, it does not seem possible to prevent such incidents with absolute certainty. Our findings underscore the importance of screening routinely for helper viruses, even when using a split-function packaging line which is widely regarded as safe.