The expression of retroviral vectors in murine stem cells and transgenic mice

The introduction of recombinant DNA into mouse embryos has proved to be a useful technique in addressing a number of important developmental questions, e.g. the identification of DNA sequences controlling tissue-specific gene expression and the role of genes in growth control (for reviews see Palmiter & Brinster, 1985; Wagner & Stewart, 1986). Currently the favoured method for gene transfer into mice is by DNA injection into a pronucleus of the fertilized egg. However, an alternative method is to exploit retroviruses as vectors for introducing genes either directly into embryos or into embryonal carcinoma (EC) or embryonic stem (ES) cells which can then be used to form chimaeric mice (Bradley, Evans, Kaufman & Robertson, 1984). There are a number of advantages to using retroviral vectors, the major one being that they can infect a wide variety of cells at an efficiency approaching 100 %. In several instances they are the only means at present for introducing genes into certain cell types, such as haemopoietic precursor cells (Keller, Paige, Gilboa & Wagner, 1985). In the infected cell the vector usually integrates stably into the chromosomal DNA as a single copy at a random site. This integration as a single copy can be exploited for cell lineage analysis, e.g. in the haemopoietic system (Keller et al. 1985). The expression from the provirus can also be accurately measured, unlike the situation with DNA-mediated gene transfer where the DNA is often integrated in tandemly repeated copies. In addition, retroviruses seem to be potentially useful as insertional mutagens (Jaenisch et al. 1985).

There are, however, disadvantages to using these vectors, some of which are technical and may be overcome in the future. One problem has been the instability of the pro viral structure, in that sometimes rearranged vector sequences are found (Joyner & Bernstein, 1983). The size constraint of approximately 8 kb of DNA, that can be incorporated within a vector is clearly a disadvantage, especially for the expression of genomic DNAs. In addition, the proper functioning of regulatory elements that are positioned in the proximity of the viral long terminal repeats (LTRs), e.g. for tissue-specific or inducible gene expression has yet to be demonstrated.

The infection of mouse embryos and EC cells with retroviruses was initially used as a model system to study the control of gene expression during mouse development (Jaenisch & Jähner, 1984). However, infection of preimplantation embryos or EC cells with the Moloney-murine Leukaemia virus (MLV) did not result in expression of the virus in either instance (Stewart, Stuhlmann, Jâhner & Jaenisch, 1982; Jähner et al. 1982). In infected embryos, this inhibition of expression was usually maintained throughout development except for some Mov strains, where activation of the virus occurred in some unidentified cells (Jaenisch et al. 1981; Jaenisch & Jâhner, 1984).

In this review we will demonstrate that retroviral vectors can be used as a tool for studying gene expression in transgenic mice and embryonic stem cells. We will discuss the unique versatility of these vectors and the various strategies which can be used at present for designing the appropriate expression vector. We will also address the problem of inhibition of expression in early embryonic cells and how this repression can be circumvented in vitro and in vivo. We will conclude by showing the potential of these vectors to express a variety of genes, e.g. an inducible c-fos proto-oncogene, so that the role of these genes in differentiation can be studied both in vitro and in vivo.

The first retroviral vectors used in gene transfer studies were relatively simple constructs. They contained a dominant selectable marker gene (see Fig. 1A), consisting of either the Tn5 neomycin resistance gene (neo^R) or the EcoGpt gene (Gpt) expressed from the 5′ LTR (Mann, Mulligan & Baltimore, 1983; Keller et al. 1985). Infection of fibroblasts usually resulted in a cloning efficiency in the order of 100% (Wagner, Vanek & Vennstrom, 1985a). However, infection of various EC cell lines resulted in a cloning efficiency some three to five orders of magnitude lower, depending on the cell line (Sorge, Cutting, Erdman & Gautsch, 1984; Wagner et al. 19856). The stem cell clones usually contained a single copy of the vector and expressed the selected gene inefficiently compared to fibroblast clones (Stewart, Vanek & Wagner, 1985; Wagner et al. 19856; Fig. 2A).

We have recently shown that infected EC cell clones, selected and characterized in vitro for the expression of a neo^R gene, were able to form viable adult chimaeras when the EC cells were reintroduced into embryos (Stewart et al. 1985). In these animals the expression of the gene under the control of the 5’ LTR was maintained in all the chimaeric tissues (and probably throughout development). This result indicated that the LTR could function as a promoter throughout embryogenesis and probably in all adult tissues. Furthermore, it illustrated the potential use of these cells in allowing them to be characterized with respect to the expression of the introduced DNA in vitro, before their introduction into embryos for subsequent development in vivo. Thus, EC cells selected to express genes in vitro could be used to study the effect that the expression of certain genes (e.g. growth control genes) might have on early embryonic development.

An alternative method for expressing genes from retroviral vectors involves the use of internal promoters, positioned between the two LTRs (Fig. 1B). This allows the expression of genes from two different promoters, the LTR and a constitutive promoter. Such a vector is shown in Fig. 1B and is termed MMCV-neo (Wagner et al. 1985a); the v-myc gene is expressed from the 5’ LTR while the neo^R gene is under the control of the herpes simplex thymidine kinase promoter (TK). This vector gave very high cloning efficiencies in the order of 10 % on infection of EC cells. Virtually all infected EC cells were resistant to G418 and expression of the neo^R gene from the TK promoter occurred as efficiently as in fibroblasts (Wagner et al. 1985a). In fibroblasts the v-myc oncogene was active but no transcripts were observed in the selected EC cells showing the 5’ LTR was not functioning. In support of these results it has been shown recently that a vector expressing the neo^R gene from an internal hybrid SV40 TK promoter was also more efficient at expressing the selectable gene in EC cells than the 5’ LTR (Rubenstein, Nicolas & Jacob, 1984).

The MMCV-rceo vector was tested in chimaeras to see if the internal TK promoter could function in all tissues and whether the LTR can be activated in some tissue or in differentiated cells. ES cells were infected with the vector, and cell clones that carried a single copy were then used to form chimaeras (Stewart et al. 1985). In all chimaeras analysed, only the neo^R gene of the vector was expressed. The 5’ LTR positioned in front of the v-myc gene was not transcribed in several tissues analysed. Even when genomic DNA from one of the ES clones was transfected onto fibroblasts, no LTR expression was restored.

The paradox with LTR activity is that LTRs do function efficiently in fibroblasts and haemopoietic cells; yet they do not function in cell types derived from infected embryos or in EC/ES cells, provided they are not selected for activity. The possible reasons will be discussed in greater detail in the next section. However, several attempts have been made at modifying the LTR so that it can function more efficiently in embryonic cells, either by substituting the LTR’s enhancer sequences with an enhancer that can function in the stem cells (Linney et al. 1984) or by isolating spontaneous mutants (Franz et al. 1986). The enhancer modification in the LTR does not appear to be sufficient for efficient expression in embryos although it may do so in some adult tissues (van der Putten et al. 1985). In the case of the mutant LTRs their function in embryos has yet to be shown.

The restriction in retrovirus expression is itself of relevance to understanding general questions concerning the developmental control of gene expression. This inhibition appears to operate at a number of different levels: it has been shown that the LTR can function as an inefficient promoter in the EC cells (Linney et al. 1984; Stewart et al. 1985) and that trans-acting negative regulatory factors may also be operating in these cells (Gorman, Rigby & Lane, 1985). However, it is not clear how the LTR can function, when it is selected for activity. Furthermore, what is the mechanism that inhibits the expression of retroviruses throughout embryogenesis or during differentiation of EC cells? It has been suggested that the LTR can undergo some mutation that allows it to function as a promoter in EC cells (Franz et al. 1986). Alternatively some cis-acting mechanism related to the chromosomal site of integration may also be involved (Sorge et al. 1984). Concerning the maintenance of this inactivation, it has been argued that de novo methylation of the proviral genome is instrumental in suppressing and maintaining the inhibition of expression (Jaenisch & Jâhner, 1984). Thus hypermethylation of the pro viral genome correlated with the absence of expression, whereas hypo-methylation correlated with expression.

Using various EC cell clones that express a single copy of a retroviral vector, we have addressed the question of the role of methylation. Two EC cell lines, F9 and P19, were each infected with a vector either carrying the EcoGpt or the neo^Rgene and cell clones were isolated by selection. In two representative F9 Gpt clones the vector was expressed at about one quarter of the level found in fibroblasts (Fig. 2A, lanes d, e). Interestingly in one of the clones a second mRNA band of almost equal intensity was also seen (Fig. 2A, lane d). Whether this was due to either readthrough from a promoter upstream to the vector or readthrough from the vector to some downstream sequences has yet to be ascertained. Nevertheless, it illustrates the fact that the chromosomal site of integration can influence the pattern of transcription, provided that an intact provirus is integrated. DNA digestion of the F9 clones with a set of methylation-sensitive enzymes showed that the vector was unmethylated for the number of sites tested, since the characteristic genomic length fragment disappeared after digestion (Fig. 2A,B).

When clones isolated from the EC cell line P19, that either contained the EcoGpt or neo^R vector were analysed, expression was similar to that found in the F9 clones (Fig. 2A, lanes b, c). However, DNA analysis of the two EcoGpt clones revealed that the entire vector (including both LTRs) was methylated at the sites analysed as shown by their resistance to digestion with three methylation-sensitive enzymes (Fig. 2A, lane 1). A similar observation was made with three of the neo^R clones (Fig. 2B, lanes 4, 5, 6). In both sets of experiments the vector had been simultaneously introduced with wild-type helper viruses. All clones contained at least one copy of the helper virus which was not expressed and was also methylated as determined by the resistance to digestion with the methylation-sensitive enzymes (data not shown). Thus expression of the vectors in these cells was probably not due to a cellular mutation resulting in the expression of all integrated retroviruses or vectors.

A second series of experiments was directed to analyse the methylation and expression of the neo^R gene in infected but unselected F9 and 3T3 cells at various times after infection. In 3T3 fibroblasts the vector containing the neo^R gene under control of the LTR was efficiently expressed 5 days after infection whereas at the same time point and also after 10 days postinfection no neo^R expression, as measured by the neomycin phosphotransferase (NPTII) activity, was found in the F9 cells (Fig. 3A). DNA analysis showed that even at 20 days postinfection the vector in both cell types remained unmethylated for the internal Hpall sites (Fig. 3A).

In a comparable set of experiments, using a vector where the neo^R gene was under the control of an internal promoter, expression was again maintained in the infected 3T3 fibroblasts. In addition, expression in the infected F9 cells at 20 days (postinfection) was still detectable (Fig, 3B). DNA analysis from both populations of infected cells showed again that the vectors remained unmethylated for the internal sites.

These results show that the level of methylation of two different vectors in EC cells, either under selectable or nonselectable conditions and as well for two different types of selection, cannot be easily correlated with the expression of the vectors. It appears that the role of methylation in regulating retroviral gene expression remains unclear, despite the possible importance of methylation regulating gene expression in other systems (for a review see Dorfler, 1985). It can still be argued that there is a crucial site within the LTR that, if methylated, would result in no expression of the virus or vector. It may be that this site is not detectable by the panel of enzymes we have used here, and thus the results presented here might only be clarified by genomic sequencing of the LTRs which should detect the methylation of every cytosine within these regulatory sequences.

The experiments described in section 11(B) showed that the internal TK promoter expressing the neo^R gene was functional in EC cells as well as in chimaeras obtained from selected ES cell clones. We therefore decided to test its ability to function in vivo by directly infecting embryos with the same vector, MMCV-neo (Fig. 1B). In initial experiments, the infected embryos were allowed to develop to the 14th day of gestation. Fifty foetuses were recovered and were explanted into culture under selective conditions. Eleven produced cell lines which grew under these conditions in G418-containing medium. Subsequent analysis showed that these foetuses contained between one to five copies of the vector and, as seen in the chimaeras, only the neo^R gene of the vector was expressed (Wagner et al. 1985b). These experiments were repeated and the infected embryos were allowed to develop to term. Out of 17 mice born, 6 were positive for the vector sequences and, of these, 3 transmitted a single copy of the vector to their offspring. The 3 transgenic mouse lines were called M-TKneol-3 (C. Stewart et al. in preparation).

It was next of interest to determine if expression of the vector can be found in the adult tissues of these mice that had been infected at the preimplantation stage when usually retroviral gene expression is permanently inhibited. RNA analysis from F₁ offspring showed that the neo^R gene expressed from the TK promoter was active in all adult tissues in two mouse lines, M-TKneo2 and 3 (Fig. 4). In the third transgenic line (M-TKneol), the vector had integrated on the X chromosome as shown by breeding analysis and expression of the neo^R gene was only found in some organs, notably the muscle, kidney and brain (C. Stewart et al. in preparation). No evidence for transcripts from the LTR containing the v-myc oncogene was found, which is in agreement with the former discussion of LTR suppression in early embryos.

These results show for the first time that it is possible to produce transgenic mice by embryo infection using retroviral vectors and that the introduced genes are expressed. A number of other groups have reported that embryos can be infected with retroviral vectors and that the vectors were also transmitted to the offspring (Jâhner et al. 1985; Huszar et al. 1985). However, expression of the vectors has not been reported either when the gene in question was under the control of the 5’ LTR (with the exception of a virus containing the polyoma enhancer; Van der Putten et al. 1985) or when the SV40 promoter itself was used as an internal promoter (Huszar et al. 1985).

One obstacle in using retroviral vectors containing intact LTRs for studying gene regulation is their presence in the proviral DNA. These elements encoding the promoter and enhancer functions can interfere in an unpredictable manner with other regulatory signals, e.g. inducible or tissue-specific promoters. One possible solution is the use of self-inactivating (SIN) retroviral vectors (formerly called suicide vectors; Wagner et al. 1985b). These vectors contain a deletion in the 3’ LTR, which abolishes the enhancer and promoter functions (Yu et al. 1986; Figs 1C, 5). The rationale for such a vector with a deleted 3′ LTR is based on two observations: (i) upon DNA transfection into packaging cell lines, the intact 5’ LTR is used to produce genomic RNA, which can be packaged into an infectious virus particle, (ii) During viral replication and integration in the target cell, the 3′ deletion is duplicated and transferred to the 5′LTR leading to an inactive provirus. Expression of a selectable or regulatable gene is therefore dependent on the internal signals and is independent of the viral regulatory regions.

Three prototype SIN vectors were constructed that carry two genes: the selectable neo^R gene expressible from the TK promoter and either the c-fos protooncogene, the Adl2ElA gene or the CAT gene, all under the control of the internal human metallothionein (hMT) promoter (Yu et al. 1986; Fig. 5A). Both neo^R resistant F9 and 3T3 cell clones were obtained, which were then analysed with respect to their proviral DNA structure and for inducibility of expression using CdCl₂ or dexamethasone. As shown by DNA analysis in Fig. 5B, correct transfer of the 3′ deletion to the 5′ LTR was seen in bulk infected and selected cells, and with some clones inducible c-fos expression was also observed (Fig. 5C).

A similar finding was obtained with the viruses carrying the El A gene. Using the Sp6-RNase protection assay, inducible and correct RNA initiation from the hMT promoter was detected (Fig. 6) as was also observed with the SIN virus containing the CAT gene. However, in this latter construct we were unable to detect an enzymatically active protein, although no obvious DNA rearrangements have been detected (data not shown). One possible explanation for this observation was that the insertion of a gene in the opposite transcriptional orientation to the LTR might produce splice sites that can form small internal deletions in the gene during the replication process leading to the synthesis of an inactive protein.

It seems that the different inserts have to be tested thoroughly for each individual construct and that these novel viruses have to await testing in the in vivo environment in mice.

The results reviewed in this article demonstrate the versatility of retroviral vectors in transferring and expressing genes efficiently in EC/ES cells and transgenic mice. These mice can be obtained by embryo infection or via chimaera formation from in vitro selected and characterized EC/ES cell clones. The vectors also provide us with the potential to introduce genes into a variety of different stem cells either to study the consequences of gene expression on these cells or to use the viral integration site as a molecular marker to analyse cell lineages. We have also outlined the different strategies for the derivation of efficient expression vectors. The genes of interest can be expressed from the viral LTR and from internal promoters or through the use of SIN vectors, that allow expression without interference from the enhancer/promoter elements in the LTRs. In embryonic stem cells and in mice derived from infected embryos the LTR suppression can be circumvented by using an internal promoter. The molecular basis for this suppression in EC cells is still unknown and in our experiments we were unable to relate the inhibition of expression causally to the methylation status of the integrated proviruses.

We have described the derivation of transgenic mice by infection of preimplantation embryos that express the neo^R gene from the internal TK promoter in all tissues and which also stably transmit this phenotype over many generations. Such neo^R mice expressing a dominant selectable marker in any cell may be of benefit to a wide range of research fields, since they can provide the experimenter with a predetermined marked cell type. The demonstration that an internal promoter can function efficiently in mice has been a stimulus to investigate whether inducible or tissue-specific regulatory elements within a retroviral vector can be used. We hope that these elements, in conjunction with the efficiency and versatility offered by these vectors, will allow us to study the function that (proto-)oncogenes have in cell proliferation and differentiation both in an in vivo and in vitro environment.

The authors wish to thank their colleagues within the department for fruitful discussions. They also thank Ines Benner for help in the preparation of the manuscript.