Authors: Galen H. Fisher¹, Sandra Orsulic¹, Eric Holland², Wendy P. Hively¹, Yi Li¹, Brian C. Lewis¹, Bart O. Williams^1,3, Harold E. Varmus¹*

¹Varmus Lab, NIH-NCI-DBS, 49 Convent Drive, Building 49, 4A56, Bethesda MD 20892

² Dept. of Neurosurgery, MD Anderson Cancer Center, 1515 Holcombe Blvd. Houston, TX 77030

³ Current Address: Van Andel Research Inst. Grand Rapids, MI 49503

* correspondence: H. E. Varmus email: varmus@nih.gov

Keywords: avian sarcoma leukosis virus, ASLV; breast cancer; cancer models; gliomagenesis; lung cancer; melanoma; ovarian cancer; pancreatic cancer; retroviral gene vectors

Abstract

To develop models of human cancer we have expressed the avian retroviral receptor, TVA, under a variety of mammalian promoters in transgenic mice, thus rendering mice susceptible to infection with avian leukosis virus-derived gene vectors. TVA-based retroviral gene transfer offers advantages over current murine models of human cancer. A single transgenic mouse line can be used to evaluate multiple genetic lesions, individually and in combination. Furthermore, mutant genes are introduced somatically into animals, as occurs in the majority of naturally occurring tumors. Because the avian viral vectors replicate only in avian cells, the viral receptor in infected transgenic mouse cells remains available for multiple rounds of infection with different ASLV vectors. We discuss the theoretical and practical aspects of using recombinant avian retroviruses with TVA transgenic mice to generate cancer models.
 
 

Introduction

Current research on tumorigenesis seeks to understand the molecular events driving tumor formation. One goal of this research is to identify targets for the development of novel therapeutics. Gene knockout and transgenic mice have yielded significant insights into the contribution of specific mutations to tumorigenesis. However they have several limitations: mutations are germline, with potential to affect tissue development and predispose to neoplasia all cells of a tissue or a mouse. In contrast, in the majority of human cancers, most mutations occur somatically in a single cell . In addition, in most transgenic models, only one or two genetic lesions are controlled by the investigator, but several mutations are likely required to produce most human tumors . Furthermore, studying interactions of germ line mutations requires extensive, expensive, and time-consuming breeding protocols.

We have sought to develop another strategy to circumvent some of the limitations of typical transgenic and gene knockout tumor models. This strategy is based on use of the receptor for subgroup-A avian leukosis virus, TVA, and allows multiple genes to be introduced somatically into a single transgenic mouse strain . TVA-based tumor modeling systems rely on tissue-specific or generalized expression of a transgene encoding TVA; the appearance of TVA on the surface of mammalian cells is sufficient to mediate retroviral infection by subgroup A avian sarcoma leukosis viruses, (ASLVs) . For the purposes of our studies, we have constructed ASLV vectors by adapting the RCAS vectors popularized by Hughes and colleagues to carry dominant negative tumor suppressors (e.g. mutant p53), oncogenes (e.g. c-myc, k-ras), marker genes (e.g. alkaline phosphatase, green fluorescent protein), and recombinases (e.g. Cre). (For a complete listing of our available RCAS vectors, see http://rex.nih.gov/RESEARCH/basic/varmus/tva-web/tva2.html)
 
 

Background and Methods

TVA structure and function: The tv-a gene was cloned from chicken and quail DNA based on its ability to allow infection of mammalian cells by ASLV-A carrying selectable markers . mRNA transcribed from the avian tv-a gene is alternatively spliced to produce at least two proteins-- a transmembrane and a GPI-anchored isoform. These two isoforms have identical 83 amino acid (aa) extracellular domains, and both are sufficient to permit infection of mammalian cells. While there is no recognized mammalian homolog, a 40 aa region in the extracellular domain is related to a cysteine-rich element repeated seven times in the low density lipoprotein receptor, LDLR . TVAís role in avian biology, other than acting as a receptor for subgroup avian leukosis viruses, is unknown.

Production of high titer ASLV stock in an avian cell line: To produce high titer recombinant viral stock, we have taken advantage of a Rous Sarcoma Virus (RSV)-derived replication-competent cloning vector, RCAS , and an established chicken fibroblast cell line, DF1 . RCAS was derived from RSV-A by replacement of the src gene with a multicloning site that stably accommodates inserts up to at least 2.5 kb . Transfection of RCAS plasmid DNA into DF1 cells results in production of high titer, replication-competent viral stock (Figure 1A). Successful transfection of only a few DF1 cells is sufficient since the replication-competent virus spreads throughout the DF1 culture .

ASLV-A infection of mammalian cells: Infection of mammalian cells with subgroup-A avian sarcoma leukosis viruses (ASLV-A) occurs at a low efficiency (< 1 x 10⁶) in the absence of TVA. However, mammalian cells engineered to express TVA are highly susceptible to infection by ASLV-A (Figure 1B). After entry into mammalian cells, a newly synthesized DNA copy of the viral genome integrates into the host DNA, and viral LTRs promote high level transcription of integrated provirus. An artificial splice acceptor in RCAS, downstream of env, is used for efficient processing of the mRNA representing the insert at the multicloning site . However, no infectious viral particles are made (S. Hughes, NIH-NCI, personal communication; Figure 1B). This has several advantages for our purposes. First, there is no spread of virus to adjacent or distant cells. Second, the TVA receptor is not blocked by infectious particles containing Env protein, and is thus available for multiple rounds of infection .

A first test: a-actin promoter-tv-a transgenic mice

The first demonstration that expression of a tv-a transgene in specific mouse tissues could be used in combination with recombinant ASLV-A to introduce foreign genes was performed with a tv-a transgene under the control of the chicken a-sk-actin promoter . The mice expressed TVA in skeletal and cardiac muscle. A week following direct injection of RCAS-alkaline phosphatase (RCAS-AP) virus supernatant into the leg muscle of newborn mice, variably intense AP-positive muscle fibers were observed at the site of injection; the variations in staining presumably reflect differences in transcriptional activity at different sites of proviral integration and perhaps different multiplicities of infection. An important observation from these initial experiments was that infection efficiency decreased dramatically after the neonatal period, likely reflecting the known requirement of cell division for efficient infection with retroviruses other than lentiviruses . In newborn mice, myoblasts are numerous, actively dividing, and thus readily infectable with recombinant ASLV-A . This first in vivo experiment showed that the tv-a -based retroviral system was a viable means of introducing foreign genes into newborn mice in a tissue-specific manner.

First use of tv-a transgenic mice to model human cancer: Gliomagenesis

We first used tv-a -based technology to induce cancer in mice in our studies of gliomagenesis. Two lines of tv-a transgenic mice were generated . The first expressed tv-a under the control of the astrocyte-specific glial fibrillary acidic protein (GFAP) promoter. In the second line, tv-a expression was controlled by the nestin promoter, which should direct expression of tv-a in glial precursors earlier in the astrocyte lineage. Both lines permit efficient glia-specific transfer of genes carried by ASLV-A vectors (RCAS), in vitro and in vivo.

As an initial test of the utility of our strategy, we infected newborn GFAP-tv-a mice with a combination of RCAS-basic fibroblast growth factor (RCAS-bFGF) and RCAS-AP. The number and the distribution of AP-positive cells several weeks later revealed that bFGF promotes extensive glial cell proliferation and migration in vivo, but without induction of frank tumors . This implied that bFGF might contribute to the tendency of high grade gliomas to infiltrate into normal brain parenchyma, but that other genetic lesions are required to produce gliomas.

By crossing tv-a transgenic mice with mice carrying targeted mutations in tumor suppressor genes, we studied combinations of gain-of-function with loss-of-function mutations. In particular, we used this to approach to produce glioma-like lesions in mice by recapitulating the combinations of mutations found in human gliomas . Human gliomas have shown mutations in genes encoding proteins in signal transduction pathways mediated by tyrosine kinase receptors and in governing the G1 to S transition in the cell cycle. To recapitulate these alterations, we generated RCAS vectors encoding Cdk4 and a naturally-occurring, constitutively-active mutant form of the epidermal growth factor receptor (EGFR) denoted EGFR*. We also crossed our tv-a transgenics to mouse lines with targeted inactivating mutations in p53 or Ink4a/Arf .

We found that gene transfer of EGFR* alone to either GFAP-tv-a or nestin-tv-a mice did not result in formation of tumors. However, transfer of EGFR* to astrocytes from Ink4a/Arf deficient mice produced glioma-like lesions. The lesions were more frequent in the nestin-tv-a mice, possibly indicating that gliomagenesis is more efficient in less differentiated cells. Furthermore, EGFR* did not induce lesions in p53-deficient mice unless Cdk4 was transferred along with EGFR*. These findings demonstrated a collaboration between disruption of the G1 arrest pathways and expression of a mutant form of EGFR in gliomagenesis.

Infection of cultured primary cells from GFAP-tv-a transgenic animals:

The TVA-based system allows efficient gene transfer to cultured primary cells . This allowed us to study, in primary astrocyte cultures, the G1 cell cycle arrest pathways that are disrupted in most immortalized cell lines and tumors. By comparing the effects of infection with RCAS(Cdk4) with an absence of Ink4a/Arf we found that both an absence ofInk4a/Arf and overexpression of Cdk4 allowed cells to escape senescence and proliferate indefinitely in culture. Cells overexpressing Cdk4 grow more slowly, and importantly shift to a hyperploid status at a high frequency . However, when Cdk4 was overexpressed in Ink4a/Arf deficient astrocytes, the cells did not become hyperploid. This implies that one or both products of Ink4a/Arf are required for Cdk4 to promote hyperploidy in primary cultured astrocytes .

Other tumor models in development

We are developing additional TVA-based models of some common human cancers, including breast, lung, ovarian, pancreatic, and melanocytic cancer. Each model presents unique technical challenges and has spurred the development of several novel approaches that further expand the flexibility of the system.

Breast cancer: To study breast cancer, we have created two lines of TVA transgenic mice, mouse mammary tumor virus LTR-tv-a (MMTV-tv-a) and whey acidic protein promoter-tv-a (WAP-tv-a) (Y. L., unpublished results). The MMTV LTR directs expression in ductal epithelial cells of the mammary and salivary glands, and MMTV-tv-a mice are predicted to express TVA in breast epithelium from early breast development onward. WAP-tv-a mice are expected to express TVA in the breast epithelium only during late pregnancy and lactation . Infection by direct injection of virus-producing DF1 cells into the glands of pregnant females offers two advantages over infection of glands in virgin animals: the glands are larger and thus easier to inject, and many cells are undergoing cell division, a requirement for ASLV infection.

We are currently adapting conditional knockout technology to the TVA-based system in the mammary gland. For example, we have crossed our MMTV-tv-a mice to mice whose BRCA1 gene contains Lox-P sites flanking exon 11 (BRCA1 fl/fl) . MMTV-tv-a /BRCA1 fl/fl mice can be infected with RCAS-Cre, which should mediate excision of sequences between Lox-P recognition sites (exon 11 of BRCA1 in this example). As compared to using adenoviral-Cre vectors, there are two potential advantages of using RCAS-Cre to infect MMTV-tv-a /BRCA1 ^fl/fl mice; Cre will be delivered only to the mammary epithelium by RCAS-Cre, and RCAS does not appear to be a potent immunogen.

Lung cancer: To evaluate the functional contribution of specific mutations to bronchogenic tumorigenesis, we generated transgenic C57BL6/J mice in which TVA is produced under the control of the 3.7 kb human surfactant protein-C (hSPC) promoter . We chose C57BL6/J (as opposed to the FVB strain more commonly used to generate transgenic mice) to minimize the contribution of unidentified loci to bronchogenic tumorigenesis; C57BL6/J mice have a significantly lower predisposition than FVB to develop spontaneous bronchogenic carcinoma . The hSPC promoter directs TVA expression in the type II airway epithelial cells , and AEC-II cells are candidate precursors of bronchogenic adenocarcinoma . Infectability of SPC-tv-a mice was first demonstrated ex-vivo by culturing day 14 embryonic lungs with a mitogen keratinocyte growth factor (hKGF) and a marker virus (RCAS-GFP), which resulted in infection of approximately 30% of the distal lung bud cells (G.H.F., unpublished results). We have begun in vivo infection of neonatal AEC-II by injection of concentrated RCAS viral vectors (titer 1 X 10⁹/ml) into amniotic fluid on gestational day 14. As opposed to infection of adult lung, in utero infection has two main advantages: SPC-expressing cells are actively dividing (Voelker and Mason, 1989), and embryonic exposure to antigen is potentially tolerogenic . Disadvantages include a requirement for timed pregnant females, time-consuming operations, and, most importantly, low pup survival (less than 50%, unless pups are fostered). Because of the cumbersome nature of in utero infection we are currently pursuing efficient ways to infect newborn and adult animals.

Pancreatic cancer: We have generated two lines of transgenic mice that should express TVA in pancreatic acinar cells, under control of the pancreatic elastase promoter and pancreatic amylase promoter (B.C.L., unpublished results). Mitotic rates of adult pancreatic acinar cells are exceptionally low(Elsasser et al., 1986). Since ASLV infection requires cell division, we are devising an alternate infection strategy that does not require target cell division. Unlike most retroviruses, lentiviruses (such as HIV) do not require mitosis for proviral integration . We are currently adapting a lentiviral HIV vector to our TVA-based pancreatic carcinoma model. To do this, we co-transfect a plasmid encoding the envelope protein for avian leukosis virus subtype-A, Env-A, together with a plasmid encoding HIVgag, pol, and a third plasmid containing only the HIV-LTRs flanking the gene of interest and a packaging signal . The resultant pseudotyped virus vector is able to infect cell lines expressing TVA with a titer of approximately 5 X 10⁴ -- an efficiency approximating that of RCAS vectors generated in the same cell line after transient transfection (B.C.L. and N. Chinnasammy, unpublished results). We are currently testing the ability of pseudotyped lentiviruses to infect non-replicating murine cells.

Ovarian cancer: In our studies of ovarian cancer, we have taken a slightly different approach because there are currently no characterized promoters that specifically direct expression to the ovarian epithelium, the presumptive precursor to ovarian carcinoma. We have used physical, instead of molecular, means to introduce specificity to our TVA-based ovarian cancer studies. To infect the ovarian epithelium, we excise the ovaries from a line of transgenic mice ubiquitously expressing TVA under control of the b-actin promoter , and infect the ovarian epithelial cells in culture (S. O., unpublished results). In this manner, any organ of interest from b-actin-TVA mice (ovary in this case) can be specifically infected in culture. After one or more rounds of infection, the cells can be studied in vitro by monitoring changes in cell growth, morphological alterations, chromosomal abnormalities, and changes in gene expression, or they can be re-implanted into syngeneic or nude (nu/nu) hosts (Figure 2). One caveat to using physical instead of molecular means to introduce specificity is that isolation of the cell type of interest is the limiting factor determining specificity. For example, within an ovary there are multiple cell types besides epithelial cells, including thecal, germ, granulosa, endothelial, and fibroblastic cells.

Melanoma: We generated mice expressing TVA under the control of the tyrosinase-related protein-2 (TRP2) promoter . This promoter directs expression to early melanoblast lineages as well as in early neural crest cells (personal communication, W. Pavan, NHGRI). To date we have been able to demonstrate expression of TVA in neonatal melanocytes and neural crest cells derived from these mice. Furthermore, neural crest cells derived from these mice can be infected with RCAS viruses (B.O.W., unpublished results, in collaboration with the laboratory of W. Pavan, NHGRI). We are currently optimizing methods for in vivo infection of melanoblasts.

TVA-based studies of neural, hemapoietic, and other organ development

TVA-based systems are also proving to be useful in developmental studies. Doetsch and colleagues recently used our GFAP-tv-a transgenic mice to demonstrate that GFAP-positive astrocytes of the subventricular zone (SVZ) act as neural stem cells. They tracked the fate of AP-positive progeny cells following RCAS-AP infection of SVZ astrocytes in vivo. Three and one half days after infection, AP-positive migrating neuroblasts appeared and, by fourteen days after infection, AP-positive mature neurons were found in the olfactory bulb .

Leavitt and colleagues generated mice expressing TVA under control of the megakaryocyte-lineage restricted promoter, GP-Iba . Cells of the megakaryocyte lineage from these mice were readily infectable both in vitro and in vivo. Simple intra-peritoneal injection of viral supernatant was sufficient to infect a large proportion of megakaryocyte precursors. This TVA-based model was used to demonstrate that IL-3 stimulates cell division, while inhibiting maturation of early megakaryocyte-lineage cells.

Another developmental application of TVA-based gene delivery is in the study of lung development (personal communication, J. Yingling, Vanderbilt University). In this model, lung buds from SPC-tv-a transgenic mice are cultured on embryonic day 11 and infected with different RCAS viruses to assess effects on branching morphology in vitro culture systems.

Practical considerations

Construction of TVA-Transgenic mice:

Choice of Promoter: Generalized, tissue-specific, and lineage-specific promoters have all been successfully used to construct TVA-transgenic mice . Each approach offers distinct advantages as outlined previously.

TVA isoform: As mentioned earlier, at least two TVA isoforms have been identified; a GPI-linked isoform encoded by an 800bp cDNA and a trans-membrane isoform encoded by a 950bp cDNA. The GPI-linked isoform has been used in the majority of tv-a-transgenic mice reported to date . The transmembrane isoform was recently used by Leavitt and colleagues in their studies of the megakaryocyte lineage .Definitive comparisons of these two isoforms have not been performed under any experimental conditions, but it is apparent that both have conferred infectivity on mammalian cells in vitro and in vivo. In the most extensive side-by-side comparison, using both transiently and stably transfected human T-cells (Jurkat cells), the transmembrane form appeared to promote infection slightly more efficiently (1.2 fold) than the GPI-linked form (unpublished results G.F.)
 
 

Delivery of virus vectors: As indicated in the preceding sections, there are now several means to deliver RCAS vectors to TVA transgenic mice, and the strategy chosen will be influenced by several factors, including accessibility and replication of target cells, the pattern of expression of TVA, the goals of the experiment, and the need for multiple infection of individual cells. Here we briefly consider some of the factors and how they might guide the strategy for delivery of vectors.

1. Mitotic activity of target cells: Target cells must be actively replicating for successful infection with RCAS viruses. In some organs, such as the skin, differentiated cells are regenerated throughout adult life from stem cells, and higher proliferation can be induced by skin abrasion. In other organs, such as the breast and the ovary, epithelial cells proliferate in response to hormonal stimulation. To optimize infection, virus delivery should be coordinated with maximum cell proliferation or enhanced by treatment with hormones. In most organs, such as the muscle and the lung, mitotic activity during embryonic development subsides dramatically soon after birth. In these cases, the optimal time for virus delivery is in the neonatal period.
2. Injecting viruses or virus-producing cells: The simplest way to deliver RCAS virus is to inject DF-1 producer cell supernatant into a mouse. It may help to concentrate virus by centrifugation or other means to infect small targets (e.g. developing organs in mouse embryos or newborn mice). Viruses can be also delivered by injecting virus-producing DF-1 chicken cells, which will continue to produce RCAS vectors for several days, thus prolonging the time of infection and increasing the exposure of target cells to the vectors. A disadvantage to this method is the possibility of an inflammatory response against the chicken producer cells.

4. Delivering virus in vitro or in vivo: In most cases, we have infected TVA transgenic mice by injecting RCAS viruses systemically or by injecting viruses or virus-producing cells directly into an organ of interest. Alternatively, the organ of interest can be isolated and its cells grown in vitro where they can be subjected to repeated infections with the same or different viruses (Figure 2). Injection in the animal more accurately resembles tumor formation in its natural microenvironment, since the organ structure is preserved. But the efficiency of infection (and multiple infection) is dependent on the accessibility of the organ and the proliferation rate of target cells. More efficient viral delivery can be achieved in vitro due to induced cell proliferation and the possibility of repetitive exposure of all cells to high titer stocks. This is the preferred method for infections with multiple viruses, which can be introduced simultaneously or sequentially. In addition, changes in cell growth, apoptosis, adhesion, protein expression and chromosome number and structure can be easily monitored at multiple time points, even without returning the cells to the animal. Cells infected in vitro with various combination of viruses can also be re-introduced into genetically identical hosts ? and, in some cases (e.g. ovary), even into the host from which the cells were obtained ? to ask whether tumors occur and how they behave.

Multiple and sequential in vivo infections: While it is theoretically possible to perform multiple sequential infections in vivo, we have found that we usually infect only a subset of cells within the target tissue (e.g. a subset of the cells that surround the needle tract in the CNS). If we assume a 15% random infection rate in a target tissue and infect sequentially with three viruses (a,b,c), eight cell types would be generated-- non-RCAS infected, a, b, c, ab, ac, bc, abc. However, only 0.34% of the cells would contain all three viruses. Sequential in vivo infections (as opposed to facile sequential infections performed in vitro) are extremely unlikely to produce pure populations of multiply infected cells. However, a heterogeneous pool of multiply infected cells within one target tissue is advantageous for studies of tumorigenesis because any tumors that arise can be excised and genotyped. Thus, one mouse can yield multiple tumors with multiple genotypes and phenotypes.

Limited insert size and RCAS stability: One potential limitation of this system is the approximately 2.5 kb insert size limit of RCAS. This has not been a significant problem for our current projects because cDNAs derived from most oncogenes we study are less than 2.5 kb, or well characterized, naturally-occurring oncogenic variants of less than 2.5 kb have been available. This may not be the case for all genes that an investigator wishes to study. One solution to this limitation is to pseudotype (with ASLV-A Env) viral vectors derived from replication-defective Moloney murine leukemia virus, MLV. Pseudotyped MLV vectors have carrying capacity up to 6-7 kb can be efficiently pseudotyped with ASLV envelope protein . Although pseudotyping allows generation of virus with large inserts viral titers may be lower than those achieved in DF1 cells with replication competent RCAS vectors.

Summary

We have made significant advances in the development of a novel retrovirus-based system to investigate the roles of specific mutations in tumorigenesis. The TVA-based gene delivery system offers several advantages over transgenic and gene knockout tumor models. First, the vector is replication-competent in avian cells, allowing for the generation of high titer viral stock. (Titers of 1 X 10⁹/ml are possible after concentrating the DF1 cell supernatant). Second, infected mammalian cells do not produce infectious virus, allowing multiple infections. Third, a single transgenic line of mice can be used for the study of multiple mutations; this simplifies breeding, saving both time and money. Fourth, TVA transgenic mice can be bred with tumor-suppressor-deficient mice to study potentially cooperative mutations. Fifth, specificity of infection is determined by the pattern of expression of the TVA transgene, allowing for developmental and tissue-specific infection. Furthermore, cells from a TVA transgenic mice can be easily infected in vitro, allowing for facile introduction of genes to study their effects on primary cells. The future of TVA-based gene delivery, in combination with the current transgenic technologies, may lead us closer to a molecular understanding of tumorigenesis.

Acknowledgments: B.C.L. is a Helen Hay Whitney Foundation Post Doctoral Fellow, and Y.L. is supported by the Army Breast Cancer Program. B.O.W. was supported by a fellowship from the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation, and E.H. was supported by the Howard Hughes Medical Institutes Fellow Training Program.

REFERENCES

Adams JM and Cory S. (1991). Science, 254, 1161-1167.

Bates P, Young JA and Varmus HE. (1993). Cell, 74, 1043-1051.

Belinsky SA, Devereux TR, Foley JF, Maronpot RR and Anderson MW. (1992). Cancer Res., 52, 3164-3173.

Boerkoel CF, Federspiel MJ, Salter DW, Payne W, Crittenden LB, Kung HJ and Hughes SH. (1993). Virology, 195, 669-679.

Doetsch F, Caille I, Lim DA, Garcia-Verdugo JM and Alvarez-Buylla A. (1999). Cell, 97, 703-716.

Elsasser HP, Lutcke H and Kern HF. (1986) The Exocrine Pankreas: Biology, Pathobiology and Diseases. New York, Raven, 1986, pp. 45-53.

Federspiel MJ, Bates P, Young JA, Varmus HE and Hughes SH. (1994). Proc. Natl. Acad. Sci. U S A, 91, 11241-11245.

Federspiel MJ, Swing DA, Eagleson B, Reid SW and Hughes SH. (1996). Proc. Natl. Acad. Sci. U S A, 93, 4931-4936.

Glasser SW, Korfhagen TR, Wert SE and Whitsett JA. (1994). Am. J. Physiol., 267, L489-497.

Greenhouse JJ, Petropoulos CJ, Crittenden LB and Hughes SH. (1988). J. Virol., 62, 4809-4812.

Hanahan D. (1988). Annu. Rev. Genet., 22, 479-519.

Holland EC, Hively WP, DePinho RA and Varmus HE. (1998a). Genes Dev., 12, 3675-3685.

Holland EC, Hively WP, Gallo V and Varmus HE. (1998b). Genes Dev., 12, 3644-3649.

Holland EC and Varmus HE. (1998). Proc. Natl. Acad. Sci. U S A, 95, 1218-1223.

Hughes SH, Greenhouse JJ, Petropoulos CJ and Sutrave P. (1987). J. Virol., 61, 3004-3012.

Ichimura K, Schmidt EE, Goike HM and Collins VP. (1996). Oncogene, 13, 1065-1072.

Kinzler KW and Vogelstein B. (1996). Cell, 87, 159-170.

Korfhagen TR, Glasser SW, Wert SE, Bruno MD, Daugherty CC, McNeish JD, Stock JL, Potter SS and Whitsett JA. (1990). Proc. Natl. Acad. Sci. U S A, 87, 6122-6126.

Macleod KF and Jacks T. (1999). J. Pathol., 187, 43-60.

Murphy GJ and Leavitt AD. (1999). Proc. Natl. Acad. Sci. U S A, 96, 3065-3070.

Nagasawa H, Mizuno M, Hasegawa M and Harigaya T. (1996). Lab Anim., 30, 127-131.

Naldini L, Blomer U, Gallay P, Ory D, Mulligan R, Gage FH, Verma IM and Trono D. (1996). Science, 272, 263-267.

Oomen LC, van der Valk MA and Demant P. (1991). Exp. Lung Res., 17, 283-304.

Petropoulos CJ and Hughes SH. (1991). J. Virol., 65, 3728-3737.

Re FC, Manenti G, Borrello MG, Colombo MP, Fisher JH, Pierotti MA, Della Porta G and Dragani TA. (1992). Mol. Carcinog., 5, 155-160.

Rehm S, Ward JM, ten Have-Opbroek AA, Anderson LM, Singh G, Katyal SL and Rice JM. (1988). Cancer Res., 48, 148-160.

Roe T, Reynolds TC, Yu G and Brown PO. (1993). EMBO J., 12, 2099-2108.

Sarzotti M. (1997). Curr. Opin. Hematol., 4, 48-52.

Schaefer-Klein J, Givol I, Barsov EV, Whitcomb JM, VanBrocklin M, Foster DN, Federspiel MJ and Hughes SH. (1998). Virology, 248, 305-311.

Schmidt EE, Ichimura K, Reifenberger G and Collins VP. (1994). Cancer Res., 54, 6321-6324.

Soneoka Y, Cannon PM, Ramsdale EE, Griffiths JC, Romano G, Kingsman SM and Kingsman AJ. (1995). Nucleic Acids Res., 23, 628-633.

Steel KP, Davidson DR and Jackson IJ. (1992). Development, 115, 1111-1119.

Voelker DR and Mason RJ. (1989). Lung Cell Biology. Massaro D. (ed.) Marcel Dekker, New York, pp. 487-538.

Ward JM, Singh G, Katyal SL, Anderson LM and Kovatch RM. (1985). Am. J. Pathol., 118, 493-499.

Wert SE, Glasser SW, Korfhagen TR and Whitsett JA. (1993). Dev. Biol., 156, 426-443.

Xu X, Wagner KU, Larson D, Weaver Z, Li C, Ried T, Hennighausen L, Wynshaw-Boris A and Deng CX. (1999). Nat. Genet., 22, 37-43.

Young JA, Bates P and Varmus HE. (1993). J. Virol., 67, 1811-1816.