ABSTRACT
Conditional mutant mice equipped with heterologous recombination systems (Cre/lox or Flp/frt) are promising for studying tissue-specific gene function and for designing better models of human diseases. The utility of these mice depends on the cell target specificity, on the efficiency and on the control over timing of gene (in)activation. We have explored the utility of adenoviral vectors and transgenic mice expressing Cre under the control of tissue-specific promoters to achieve Cre/lox-mediated somatic recombination of the LacZ reporter gene, using a newly generated flox LacZ mouse strain. When adeno Cre viruses were administered via different routes, recombination and expression of LacZ was detected in a wide range of tissues. Whereas in liver [beta]-galactosidase activity was quickly lost by turnover of expressing cells, even though the recombined allele was retained, [beta]-galactosidase in other tissues persisted for many months. Our data indicate that the flox LacZ transgenic line can be utilized effectively to monitor the level and functionality of Cre protein produced upon infection with adeno Cre virus or upon crossbreeding with different Cre transgenic lines.
Recent advances in mouse embryology and molecular genetics have made it possible to specifically mutate defined genes in the germline of mice. Protocols using transgenesis (1 ) and homologous recombination in embryonic stem (ES) cells (2 ,3 ) permit overexpression, inactivation and modification of genes more or less at will. Mutations, identical to those observed in inherited or somatically acquired diseases such as in cancer, can now be mimicked in a mouse model system. These technologies are invaluable in assessing the role of genes in complex processes such as tumorigenesis, embryonic development and functioning of the immune system. However, there are a number of limitations as well, e.g. nullizygosity appears to be lethal in many instances or causes complex pleiotropic effects and, therefore, does not permit the development of an in vivo model system in which gene inactivation is restricted to a defined subset of cells (4 ).
To overcome these limitations, strategies for conditional, cell type-specific gene targeting (5 ) and inducible gene disruptions (6 ) have recently been developed. These systems take advantage of site-specific recombinases such as the Cre/loxP recombination system of bacteriophage P1, which requires only two well-characterized components: the 38 kDa recombinase protein Cre and the 34 bp long Cre-specific recognition sequence, loxP (7 -9 ). The utility of this system has been shown both by the generation of conditional transgenic mice and the production of conditional gene knockouts. In the latter case, a target construct flanked by two loxP sites (flox) was used to modify the cognate gene by homologous recombination in ES cells. The expression level of the floxed allele is expected to be the same as that of the wild-type and should, therefore, not lead to phenotypic changes. Crossing of the floxed mice with transgenic mice carrying the Cre recombinase gene under the control of a cell type-specific promoter or an inducible promoter then leads to excision of the intervening sequences. This strategy has been shown to work in a number of settings (6 ,10 ).
However, the current systems do not provide optimal flexibility. Three issues are relevant in this respect: (i) recombination in other cells than the desired target cell might occur due to inappropriate or unexpected expression of the Cre transgene; (ii) lack of control over the developmental timing of recombination as a result of the nature of the promoter used or leakiness of the inducible promoter chosen; (iii) complexity of the breeding when (multiple) conditional alleles have to be combined with specific transgenes.
We have started to address these questions by: (i) generating a flox LacZ indicator mouse that permits monitoring of the activity and timing of the Cre recombinase in vivo; (ii) exploring the utility of vectors designed for somatic gene transfer to introduce the Cre recombinase at a desired time into somatic cells.
First, we have explored the utility of a Cre adenovirus (11 ,12 ) to mediate Cre/lox recombination in vivo. Our results confirm and further extend studies using adenoviral vectors to mediate recombination in mice (13 ). Earlier, the use of adenoviral vectors has been explored for their application in gene therapy protocols. From these studies it has become clear that adenoviruses can infect a variety of cell types in vivo irrespective of whether the cells are actively dividing or not (14 -20 ). It has become apparent that in immunocompetent hosts expression of genes transduced by adenoviral vectors lasts, in general, <3 or 4 weeks, because of elimination of the infected cells through a cell-mediated immune response (21 ,22 ).
Here we show that, using the flox LacZ indicator mouse strain, adeno Cre virus can mediate recombination at high frequency in a range of somatic tissues. We also demonstrate that the switch-on of LacZ, rather than expression of Cre or viral structural genes, leads to elimination of cells that have undergone recombination of the chromosomally localized LacZ reporter gene. Finally, the system provides the potential for efficient and simple lineage marking.
The pCAG promoter-loxP-CAT-loxP-LacZ DNA for flox LacZ indicator mice was generously provided by Dr Kimi Araki (23 ). For cell type-specific expression of the Cre recombinase, a 1.4 kb fragment containing the Cre open reading frame with an N-terminal nuclear localization signal (nlsCre) (5 ) and synthetic intron was isolated from pOG231 (a gift of Dr O'Gorman). This fragment was inserted between 706 bp of the rat pro-opiomelanocortin (POMC) promoter 5' flanking sequence (24 ) and SV40 splice and polyadenylation signal or the 1.1 kb segment of 5' flanking sequence from the rat P0 gene (25 ) and rabbit [beta]-globin 3' splicing and polyadenylation signal. nlsCre was inserted between 1.3 kb of 5' flanking sequence of human interphotoreceptor retinoid binding protein (IRBP) (26 ) and rabbit [beta]-globin 3' splicing and polyadenylation signal. The inserts were cut out, freed from vector sequences by agarose gel electrophoresis, purified and injected into fertilized FVB/N mouse (27 ) oocytes essentially as described (28 ). The founder mice were bred to FVB mice to maintain the line. The presence of the transgenes was determined by Southern analysis of tail tip DNA.
DNA was isolated from tissues using standard procedures. Briefly, tissues were minced and digested overnight in SDS, proteinase K buffer. These were followed by two phenol extractions, one phenol/chloroform extraction, chloroform/isoamyl alcohol extraction, ethanol precipitation and resuspension in TE buffer. An aliquot of 10 [mu]g DNA was digested with EcoRI and XbaI and subjected to Southern analysis. The 1.0 kb of SacI-EcoRI LacZ fragment and Cre recombinase coding sequence amplified by PCR were used as probes to estimate recombination of the flox LacZ transgene and the quantity of adeno Cre viral genome. The band intensities were quantified with a Bioimaging analyzer (Fuji Bas 1000). Percentage recombination was calculated by subtraction of the background.
The Cre adenovirus vector was constructed and propagated essentially as described (11 ). Briefly, the HindIII fragment of pBS185 containing the Cre ORF under control of the hCMV immediate early promoter and the metallothionein I polyadenylation signal (29 ) was inserted into the HindIII site of pdE1spIA and co-transfected with pJM17 into 293 cells. The viruses were plaque purified and re-analyzed prior to preparation of large scale stocks. The recombinant adenoviral vector was propagated in 293 cells and purified by cesium chloride density centrifugation.
For PCR analysis, 0.5 [mu]g DNA were submitted with Perfect Match DNA Polymerase Enhancer (Stratagene) to 38 cycles of amplification (each cycle consisting of 1 min at 94oC,1 min at 57oC and 2 min at 72oC). To explore the junction between actin promoter and LacZ, AG2 (5'-CTGCTAACCATGTTCATGCC-3') and Z3 (5'-GGCCTCTTCGCTATTACG-3') primers were used.
All animals were housed in SPF facilities. Mice were moved and maintained in an isolator after administration of adeno Cre viruses. Aliquots of 0.2 ml viral solution (109 p.f.u.) were injected into the tail vein or 107 p.f.u. virus were injected into muscle, lung or under the skin.
Mice were sacrificed and organs were fixed in 0.2% glutaraldehyde solution or mounted in OCT compound. After a rinse in detergent, organs were incubated in 5-bromo-4-chloro-3-indolyl [beta]-D-galactoside (X-gal) staining solution at 37oC essentially as described (28 ). A post-fixation was performed for 24-48 h in 10% formalin. Organs were dehydrated and embedded in paraffin for sectioning.
Frozen sections were fixed in 0.2% glutaraldehyde, 0.1 M phosphate buffer, pH 7.3, 5 mM EGTA, 2 mM MgCl2 on ice for 5 min, rinsed three times in phosphate-buffered saline (PBS) with 2 mM MgCl2 on ice, placed in detergent rinse on ice for 10 min, incubated in staining solution (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl2, 0.01% sodium deoxycholate, 0.02% NP-40 in 0.1 M phosphate buffer, pH 7.3) at 37oC overnight. Then they were washed twice in PBS with 2 mM MgCl2 at room temperature for 5 min each, rinsed in distilled water and counterstained in neutral red for 1 min. After staining, samples were rinsed in distilled water three times, dehydrated with ethanol (5 min each in 70 and 100%), cleared twice in xylene and mounted.
To assess the efficiency of Cre-mediated site-specific recombination, we have generated flox LacZ reporter mice that carry the chicken [beta]-actin promoter (30 ) and [beta]-galactosidase ([beta]-gal) transcription unit separated by the CAT gene flanked by loxP sites (Fig. 1 ). The CAT transcription unit prevents [beta]-gal expression. However, when the CAT gene is excised, [beta]-gal expression will ensue (23 ). Fourteen independent transgenic founders were produced and two of them carried one and two copies of the transgene. Expression of the CAT gene was estimated by Northern blot analysis of RNA isolated from the various tissues. Both strains showed high mRNA levels in heart and muscle, moderate levels in kidney, lung, thymus, spleen, stomach, intestine, brain, retina and peripheral nerves and low amounts of mRNA in liver (data not shown). The single copy transgenic line was selected for the experiments described below. No LacZ expression can be detected in this mouse line by X-gal staining without exposure to the Cre recombinase. Therefore, this transgenic mouse line allows evaluation of the recombination frequency not only by LacZ staining in situ, but also by Southern blot analysis on tissue DNA. Upon EcoRI digestion, the single copy transgene gives rise to fragments of 4.4 and 3.1 kb when unrearranged and recombined respectively (Figs 1 and 2 A).
Recombinant adeno Cre virus was constructed as described (11 ). Aliquots of 109 p.f.u. vector or the same volume of physiological saline (220 [mu]l) were infused into the tail vein of 6-week-old flox LacZ reporter mice and non-transgenic control mice; 5, 15, 30 and 60 days later the mice were sacrificed and frozen tissue sections were stained with X-gal and counterstained with neutral red or hematoxylin/eosin. Strong LacZ expression was observed in a wide range of tissues: liver, heart, pancreas, intestine, lung, muscle, kidney and adipose tissue. Furthermore, recombination could be detected by PCR in most tissues except brain (see Fig. 6 ).
The liver appeared the preferred target site after i.v. inoculation of the adeno Cre virus. It has been shown previously that nearly 100% of hepatocytes can be transduced in vivo when a large number of adenoviral particles are infused i.v. (31 ). Inoculation of 109 p.f.u. vector into the tail vein of flox LacZ indicator mice resulted in massive staining of the liver. About 65% of hepatocytes were stained by X-gal 5 days after injection (Fig. 3 b and g), whereas no staining was observed in control livers (flox LacZ mice injected with saline; Fig. 3 a and f). However, the fraction of [beta]-gal-positive hepatocytes was reduced to 40, 4 and finally <1% of hepatocytes on days 15, 30 and 60 post-injection respectively (Fig. 3 c and h, d and i, e and j and Table 1 ), probably as the result of a cellular immune response directed against the transduced cells and/or LacZ-expressing cells (see below). Evidence for a self immune response against hepatocytes was evident from perivascular infiltration of lymphocytes, extensive degeneration and necrosis of the liver and hypertrophy of cells. Whereas only few abnormalities were seen on day 5, extensive degeneration of hepatocytes and large numbers of aberrant giant nuclei were found at later stages. Lymphocyte infiltration increased especially in the portal triad. Liver regeneration started around day 15 and was mostly completed by day 30. Only few lymphocytic infiltrates were observed perivascularly on day 60 (Fig. 3 l-o).
After adeno Cre virus infusion into the tail vein of flox LacZ mice, [beta]-gal-positive cells were observed at all the time points tested (up to 60 days post-injection) in cardiac muscle (Fig. 5 a and b), the muscular layer of the main vessel walls in lung (Fig. 5 e and f), the acinar cells of the pancreas (Fig. 5 g and h), the longitudinal and circular muscular layer of the intestine (Fig. 5 c and d), skeletal muscle and the endothelium of glomerular capillaries and tubular cells of the kidney. Southern blot analysis did not detect recombination in these tissues but PCR analysis confirmed that recombination had occurred (Fig. 6 ). This indicates that the recombination frequency in these tissues was relatively low. Interestingly, in contrast to what we observed in liver, the number of [beta]-gal-positive cells in these tissues did not diminish significantly over time.
Three different Cre transgenic mice were produced.P0 Cre transgenic mouse.To selectively direct expression of Cre recombinase to Schwann cells, we utilized the P0 promoter (25 ). P0 is the major structural protein of peripheral myelin and is expressed in myelinating Schwann cells in the later stages of myelination. The region of 1.1 kb of 5' flanking DNA from the rat P0 gene is effective in directing expression of heterologous genes to myelinating Schwann cells in transgenic mice. Five P0 Cre transgenic mice were obtained and crossbred with flox LacZ indicator mice.The sciatic nerve was isolated from double and single transgenic mice and stained for LacZ expression. Extensive staining of the sciatic nerve was observed in double (Fig. 7 b), but not in single, transgenic mice (Fig. 7 a). Histochemical analysis showed that LacZ was expressed in many, but not all, Schwann cells.
Recently, transgenic mouse systems have been described that permit inducible expression of the Cre recombinase which, in turn, can mediate recombination of chromosomally located recombination targets (6 ). While this work marks a milestone in the field of mouse reverse genetics, full exploitation of this system requires further improvements. The control of Cre expression appears critical in this respect: tissue specificity of expression, background activity, level of induction, control over fraction of cells in which expression can be induced and the timing of expression. These features are difficult to incorporate in Cre transgenic lines. Even when Cre transgenic mice meet some of these requirements, time consuming and costly breeding will have to be performed to combine the different alleles (especially when multiple conditional null alleles are being analyzed). Somatic gene transfer to introduce the recombinase into cells in vivo is, therefore, an appealing option. We have generated a flox LacZ reporter mouse that permits evaluation of the performance of Cre-expressing systems based upon transgenic mice or somatic gene transfer vectors.
We have used an adeno Cre virus to monitor recombination in vivo in these LacZ reporter mice. Replication-deficient adenoviral vectors have been developed for gene therapy purposes and were shown to efficiently transduce genes into many different cell types in vivo in animal model systems. In general, gene expression is transient and extinction of gene expression is thought to be the result of the dilution of episomal viral DNA upon cell division as well as the consequence of the elimination of infected cells by the immune system (22 ,32 ,33 ).
Infusion of 109 p.f.u. adeno Cre virus into flox LacZ reporter mice caused extensive recombination of the flox-LacZ gene construct in hepatocytes. Close to 80% of the hepatocytes showed recombination at the DNA level. While it is possible that the percentage of cells infected by the adeno Cre virus was slightly higher, our data clearly show that adenoviral vectors can be used effectively to achieve efficient Cre/lox recombination.
Under conditions in which the vast majority of hepatocytes were infected by the adeno Cre virus, >90% of the viral genomes were cleared from the liver by day 60. Over this time period LacZ expression was reduced from 65 to <1%, although still >50% of the hepatocytes carried a recombined LacZ allele. A possible explanation for these observations is that the floxed transgene is not expressed in all cells. The lower frequency of LacZ staining (65%) as compared with the recombination frequency (78%) observed 5 days after adeno Cre virus inoculation, a time point at which an immune response has not yet been mounted, is in accordance with this. If we assume that [beta]-gal acts as a potent antigen, as has recently been shown (35 ), LacZ-expressing cells might be an efficient target for immune destruction. Consequently, LacZ-expressing cells will be eliminated and hepatocytes not expressing LacZ will divide to substitute for the lost cells. The observation that at day 60 50% of the hepatocytes still carried a recombined allele, while not expressing LacZ, suggests that [beta]-gal, rather than proteins synthesized from the viral template, elicited this immune response. The loss of adenoviral genomes from the regenerated liver would then primarily be the consequence of cell division. This raises an interesting issue with respect to the role of the vector-encoded proteins, including Cre, in the immune response, as 50% of the hepatocytes that survive have a history of expression of vector-encoded proteins. Either under the circumstances of our experiments [beta]-gal is immunodominant and the immune system `ignores' other viral antigens or expression of adenoviral vector-encoded genes and the LacZ transgene is extinguished after a short burst of transcription. Such silencing could be triggered by local cytokine production, which occurs as a consequence of a massive inflammatory response (36 ). Moreover, since liver macrophages (Kupffer cells) are infected by the adeno Cre virus, this may also contribute to a cytokine-mediated silencing of foreign genes in the liver, as has recently been described for inactivation of HBV transgene expression by LCMV-infected Kupffer cells (37 ). However, we would have expected a much further reduction in expression of the unrearranged CAT transgene (we still observed 20% of wild-type levels at 60 days after adeno Cre virus inoculation) if general transgene silencing was the predominant mechanism. Further analyses are in progress to address this question.
The observation that the sparsely observed LacZ expression in other tissues is not extinguished argues against a cellular immune response towards cells at these other sites. It is possible that the lower local antigen load fails to recruit immune cells to these sites and, therefore, permits survival of the LacZ-expressing cells. Alternatively, these different cell types might be less well equipped to present the antigens to the immune system.
A few points are worth emphasizing. Firstly, the immune response we observe in the liver calls for immune suppressive or tolerizing measures if an appreciable level of a foreign protein has to be produced in liver for a long time. These actions might be needed irrespective of the vector system used to transduce the gene. Secondly, the discrepancy between LacZ expression and the recombination frequency after adeno Cre virus infection points in this case to a mosaic expression pattern of the reporter transgene. Since this behavior of transgenes might be the rule rather than the exception, this could hamper widespread gene recombination in a particular tissue using transgenic Cre mice. Thirdly, somatic gene transfer, while probably not suitable at present to confer massive switching in all tissues, provides excellent control over the timing of recombination, while cell target specificity can be achieved both by the route of virus administration (20 ) and by the use of tissue-specific promoters within the viral vectors. These features are paired to a high transient expression level, no background and simple breeding requirements, making this approach particularly attractive if one wants to study the biological consequences of gene (in)activation in a range of different tissues, e.g. in assessing the oncogenic potential of loss of tumor suppressor gene function.
Studies with immunodeficient mice have demonstrated that prolonged gene expression from adenoviral vectors is possible (22 ,38 ). Therefore, suppression of the immune response permits expression of somatically introduced genes for long periods of time. Several studies have addressed ways to mitigate the immune response in immunocompetent animals. The temperature-sensitive mutant E2A recombinant adenovirus shows improved transgene persistence and a decreased inflammatory response in mouse liver (39 ). The administration of immunosuppressive agents such as cyclosporin A and cyclophosphamide resulted in persistent expression of somatically transferred transgenes (40 ). Moreover, the administration of CTLA4-Ig, which is known to block co-stimulatory signals between T cells and antigen-presenting cells, permitted persistent adenoviral-mediated gene expression in mice without the requirement for long-term immunosuppression (41 ). These strategies might also facilitate the persistence of cells that have undergone somatic recombination through infection by adeno Cre virus.
We have also demonstrated the utility of Cre recombinase transgenic mice to mediate tissue-specific recombination. Our LacZ reporter mouse will be of specific value to estimate the expression characteristics of distinct promoters, as it leaves an imprint that otherwise might escape detection. The utility of Cre transgenic mice could be further increased by the use of regulatable promoters for expression of the recombinases. The interferon-inducible promoter has been successfully used for conditional mutation of endogenous alleles. Other inducible systems have still to prove their suitability. The use of binary drug-inducible transcription (42 -45 ) or direct fusion of the recombinase to hormone binding domains (46 ) will, hopefully, permit further improvement of these conditional systems.
We would like to thank Dr Kimi Araki (Kumamoto University School of Medicine), Dr Jun-ich Miyazaki (Osaka University), Dr Stephen O'Gorman and Dr G.Lemke (Salk Institute) for providing the pCAG-CAT-Z plasmid, pCAGGS plasmid, pOG231 and P0 promoter respectively and Drs Liou and Low for the ITBP and POMC promoter constructs. We also thank Rein Regnerus for helping with genotyping of the mice and the staff of the NKI animal facility for expert assistance.
*To whom correspondence should be addressed. Tel: +31 20 512 1990; Fax: +31 20 512 2011; Email: tberns@nki.nl
Present addresses: +Department of Tumor Genetics and Biology, Kumamoto University, School of Medicine, 2-2-1 Honjo, Kumamoto 860, Japan and [sect]Unité INSERM 434 Tumor Genetics, Institut Curie, 26 rue d'Ulm, 75231 Paris Cedex 05, France
REFERENCES
Return