Genetically manipulated mice: a powerful tool with unsuspected caveats
- 1Gene Targeting Laboratory, The John Curtin School of Medical Research, Building 131, Garran Road, The Australian National University, Canberra, ACT 0200, Australia
- Corresponding author K. I. Matthaei: Gene Targeting Laboratory, The John Curtin School of Medical Research, GPO Box 334, Canberra City, ACT 0200, Australia. Email: klaus.matthaei{at}anu.edu.au
Abstract
Although genetic manipulations in mice have provided a powerful tool for investigating gene function in vivo, recent studies have uncovered a number of developmental phenomena that complicate the attribution of phenotype to the specific genetic change. A more realistic approach has been to modulate gene expression and function in a temporal and tissue-specific manner. The most common of these methods, the CreLoxP and tetracycline response systems, are surveyed here and their recently identified shortcomings discussed, along with a less well known system based on the E. coli lac operon and modified for use in mammals. The potential for further complications in interpretation due to hitherto unexpected epigenetic effects involving transfer of RNA or protein in oocytes or sperm is also explored. Given these problems we reiterate the necessity for the use of completely reversible methods that will allow each experimental group of animals to act as their own control. Using these methods with a number of specific modifications to eliminate non-specific effects from random insertion sites and inducer molecules, the full potential of genetic manipulation studies should be realized.
Genetic manipulations in mice have provided a powerful tool for investigating gene function in vivo. However, complications in the interpretation of past and present studies have arisen following the observation that results can often vary substantially depending on the strain of mice in which the manipulation is made (see Matthaei, 2004 for a more extensive discussion). Other unexpected results of genetic deletion studies, such as embryonic lethality (see for example Faast et al. 2001 for H2AZ) and coordinate or compensatory regulation of other related genes (for example Lim et al. 2004; Blackburn et al. 2006 for GST zeta), provide further complications to delineating the specific role of a gene in a particular tissue type. More recent methods, which enable both tissue-specific and temporal regulation of gene expression, now provide a more realistic, if more complicated, approach to understanding gene function. However, unanticipated complications of these methods are now being unearthed. This brief review will critically survey the pitfalls of the currently available methods and point the way forward in this now seemingly unfathomable area.
Problems with transgenic mice due to random integration
The generation of the first transgenic mouse by injection of a DNA construct directly into the pronucleus of a fertilized single cell mouse embryo (Palmiter et al. 1982) has resulted in the generation of thousands of such strains. Importantly, the site at which the DNA is integrated is random, as are the number of copies of the transgene. Although the expression of the construct is faithful for the promoter, on many occasions it may also be significantly influenced by the local environment at the integration site (the ‘position’ effect). This can lead to the promiscuous expression of the transgene (often referred to as ‘leakiness’), due to modification of the specificity of the promoter, or at times to a more severe phenotype, due to disruption of an unknown gene by insertion of the transgene (insertional mutagenesis). A number of different ‘founder’ animals with different copy numbers and different integration sites must therefore be assessed in order to determine the correct/faithful expression of each transgene. Surprisingly, in one example, 24 different founders resulted in 24 different expression patterns (Feng et al. 2000) making it impossible to determine which pattern was correct.
One way to overcome this problem is to use a ‘knock-in’ procedure where the transgene is introduced into a specific locus using homologous recombination in embryonic stem cells. In this situation a modified gene can be expressed under its own promoter (see for example Yang et al. 2006), or a transgene can be expressed from a ubiquitous promoter, such as Rosa26 (Soriano, 1999). An alternative approach to improve promoter specificity is to use larger DNA fragments, which can be manipulated in bacterial (BAC) or yeast (YAC) artificial chromosomes, to drive gene expression. These larger fragments can avoid the position effect since they contain more of the local environment of the promoter thereby ensuring its ability to function normally. These fragments can vary in length up to 250 kb as opposed to the 1–2 kb commonly used for transgenic constructs and they result in faithful control of reporter gene expression (see for example Moreira et al. 2006). Indeed a major initiative using this approach estimates that 85% of BAC transgenic constructs express reproducibly in multiple transgenic lines (Gong et al. 2003), thereby reducing the need to generate and assess more than a few lines to identify faithful expression. However, it is not clear whether the presence of other genes, transcription factors, microRNAs or control regions in these fragments may compromise results. In the BAC studies mentioned above, the readout was expression of a reporter gene (EGFP) and not the expression of a functional protein such as a particular cytokine that can produce specific effects. It would be interesting to compare the phenotype of the interleukin(IL)5 transgenic mouse that uses a minimal promoter (Dent et al. 1990) with a mouse expressing IL5 from a BAC which would also contain a number of other cytokines due to their very close proximity, including IL3, IL4, IL13 and granulocyte macrophage colony stimulating factor (Lee & Young, 1989; McKenzie et al. 1999; Avery et al. 2004). This approach, without careful analysis, may therefore not be as widely applicable as first thought and the ‘knock-in’ procedure, although more complex, may still be the method of choice. Again, caution must be taken since the ‘knock-in’ will produce an insertional deletion of the gene controlled by the promoter to be used (see for example Gazit et al. 2006). Moreover, even in the heterozygous state the mice can have a phenotype if the gene exhibits haploinsufficiency (deletion of one functional allele resulting in modified gene expression). However, the insertional deletion can be avoided if the construct uses an internal ribosome entry site (IRES) followed by the transgene and these are placed between the stop codon and the polyadenylation signal of the target gene. In this case a bicistronic transcript is made containing both genes that are then translated into two separate proteins (see for example Rees et al. 1996), resulting in the expression of the transgene without disrupting the endogenous gene.
However, the potential ‘leakiness’ of conventional transgenic mice complicates most current studies involving genetic manipulations, as we will see in the following sections.
Tissue and temporal specific control of gene function
CreLoxP.
The deletion of a gene results in embryonic lethality if there is an absolute requirement for that gene during development. This lethality can be avoided if the gene is deleted later in life or if it is deleted in a tissue-specific manner particularly if the gene has multiple functions. The most common method to obtain tissue-specific control of gene deletion is the CreLoxP system modified from bacteriophage. LoxP sites are short DNA sequences that are recognized by a specific DNA recombinase enzyme called Cre (causes recombination) that deletes any DNA between the two sequences (see Fig. 1). An exon in the gene of interest is flanked by these LoxP sites and the modified gene is introduced into its correct location in embryonic stem cells by homologous recombination. The stem cells are injected into mouse blastocysts and mice generated (for a fuller description of chimæra generation see Matthaei, 2004) in which the gene is ‘floxed’ but fully functional until it is inactivated by the Cre recombinase due to the removal of the exon. This is achieved by breeding the ‘floxed’ mice with a mouse transgenic for the Cre recombinase usually under the control of a tissue-specific promoter (first described in Gu et al. 1994, but see also Rajewsky et al. 1996; Kwan, 2002). The result is deletion of the gene in a specific tissue.
A major problem with this method is the ‘position’ effect as described above, whereby Cre is expressed non-specifically in other tissues, or the promoter driving Cre is active early during development so that the gene is ablated at the wrong time or in the wrong tissue. An improvement is to provide temporal control of Cre expression through the use of an inducible promoter that has been modified to be ‘off’ until the mice are treated with the inducer. This has been achieved using the synthetic steroid RU486 (first described in Kellendonk et al. 1996) in low doses to prevent anti-progesterone effects in mice, such as abortion, although side-effects in other tissues, such as in neurons (Hendry et al. 1987) may continue to be a problem. In some systems these side-effects may be avoided by the topical application rather than ingestion of RU486 (see, for example, the activation of Keratin.Cre, where the effect is limited mostly to the skin (Zhou et al. 2002)). Other Cre inducible systems using Tamoxifen have also been established (Zheng et al. 2002); however, again there are potential side-effects due to the use of this drug. A further difficulty, that has never been addressed, is that the mammalian genome contains many ‘pseudo’ LoxP sites (Thyagarajan et al. 2000) that potentially could be recombined by Cre recombinase with totally unknown effects. Such a consideration requires the inclusion of a Cre-transgenic mouse amongst the control wild-type mice (see also the use of littermates as control below).
Notwithstanding the various drawbacks inherent in the use of conventional transgenic mice, the major failing with the Cre recombinase system is that it is not reversible.
Reversible tissue and temporal specific control of gene over-expression
Tetracycline response system.
Reversible control of gene expression has for some time been possible with the tetracycline response system (Kistner et al. 1996). In this procedure, one mouse strain is made to express a tetracycline-responsive transactivator protein, tTa (tet-off) or rtTa (tet-on), usually under the control of a tissue-specific promoter. A second mouse expresses the gene of interest under the control of a minimal promoter that requires the transactivator protein (tTA or rtTA) and tetracycline. Doubly transgenic mice for both transgenes can then be regulated by the addition (tet-on) or removal (tet-off) of tetracycline in the drinking water allowing control of the gene of interest in a tissue- and temporal-specific manner.
However, this system again is based on conventional transgenic mice and for the reasons given above can also suffer from promiscuous expression. A particular problem occurs due to the lack of complete silence of the minimal promoter when the transactivator protein is absent, hence the system is never fully ‘off’. More recently, a major improvement for the rtTA system was developed involving a second protein (tTS) that is a fusion of the transactivator protein and the KRAB-AB silencing domain of the Kid-1 protein (Zhu et al. 2002). tTS binds to the minimal promoter in the absence of tetracycline and ensures complete silencing of the promoter (Zhu et al. 2002). In the presence of tetracycline, tTS dissociates from the promoter allowing rtTa to bind and drive transcription allowing better temporal control. However, it is important to note that tetracycline is an antibiotic that could affect the bacterial flora of the experimental mice necessitating multiple control groups. Moreover the recent finding that the rtTA protein itself can produce emphysema-like symptoms in the lungs of mice is a further complication (Sisson et al. 2006). Whether the tTS protein also causes similar symptoms in the lung or indeed if there is a general problem of expressing rtTA or tTS protein in different tissues needs clarification.
In spite of these potential problems, the tetracycline system has enabled tissue and temporal control of gene function.
LacO/LacIR system.
Recently, an alternative, reversible and inducible expression system based on the lac operon of E. coli was developed for use in mammals (Cronin et al. 2001). In E. coli, the lac operon functions by a repression mechanism involving the production of an inhibitor protein LacIR, that binds to regulatory sites lacO in the promoter and turns off transcription of the genes required for lactose metabolism. By adding lactose, the LacIR protein changes its binding affinity for the lacO sequences, dissociates and transcription of the lac genes can occur. E. coli uses this system to tightly control the genes required for the utilization of lactose and it is completely reversible. In a proof of principle, Scrable and colleagues were able to modify the tyrosinase promoter with LacO sites and successfully switch coat colour on and off at will when IPTG (an analogue of lactose) was added to the drinking water of mice that also expressed the LacIR protein (Cronin et al. 2001). Tyrosinase activity could also be regulated during discrete periods of embryogenesis (Cronin et al. 2003).
In a subsequent study, similar control of gene expression during embryogenesis was demonstrated in vivo for a luciferase reporter gene under the control of the ubiquitous Huntingtin promoter with introduced LacO sites (HuntingtinLacO-luciferase) (Ryan & Scrable, 2004). In this study it was shown that luciferase activity could be regulated in live offspring in utero by the addition of IPTG to the drinking water of the pregnant dam (Ryan & Scrable, 2004).
In order to study the regulation of this gene in different tissues we obtained mice expressing the HuntingtinLacO-luciferase (HDOSluc) reporter and mice expressing the LacIR repressor protein from Professor Scrable. By crossing these mice we have observed repression of luciferase expression by the LacIR protein to less than 10% in a range of different tissues (Fig. 2A and Matthaei et al. 2006). Importantly, addition of IPTG to the drinking water produced de-repression (re-activation) of luciferase expression within 48 h to 80% of normal activity (Fig. 2B for kidney). This level of repression is similar to that achieved with siRNA (Carmell et al. 2003) but without side-effects in non-target tissues (Jackson et al. 2003). We therefore see tight reversible regulation of a transgene using a common compound (sugar) that should have minimal side-effects. This system warrants intensive investigation since it possibly provides the best method of controlling gene overexpression or deletion (see below).
Epigenetic complications for genetically manipulated mice
Epigenetics is the transmission of information from a cell or multicellular organism to its descendants without that information being encoded in the nucleotide sequence. Evidence is now emerging that epigenetic changes can occur due to the transmission of RNA or protein in the oocyte or sperm from transgenic parents to non-transgenic offspring. We were first alerted to this problem when we were crossing a ‘β-galactosidase Stop floxed EGFP’ mouse (see Fig. 1) with Cre mice controlled by the Tnap promoter (Lomeli et al. 2000). In this cross we expected to see the Cre delete the β-galStop and allow expression of EGFP (Figs 1 and 3A). Indeed, when the Cre transgenic parent was the male we did see faithful transmission of the Cre transgene to the offspring and deletion of our floxed gene in these mice (Cre/delLoxP, Fig. 3A). However, when the Cre-carrying parent was the female, we saw deletion of our floxed gene in all offspring, even in those that had not inherited Cre (WT/delLoxP, Fig. 3B). This phenomenon has also been observed by others. Tissue-specific deletion occurred when the Cre transgene controlled by the keratin K5 promoter was carried by the male parent, but deletion in all offspring when the Cre transgene was carried by the female parent (Ramirez et al. 2004), including those not inheriting Cre. Moreover, when the Cre parent was female the deletion was no longer tissue specific but occurred in all tissues of the offspring, suggesting deletion very early during embryogenesis (Ramirez et al. 2004). These data suggest that RNA or Cre protein may be present in all oocytes of a Cre transgenic female and this is transmitted to non-transgenic offspring causing generalized deletion. Moreover, since this phenomenon has been seen for two different promoters driving Cre, suggests that the effect may be more general. In this case, non-transgenic littermates would not constitute an appropriate control.
Direct evidence that transgenic RNA may cause phenotypic changes in non-transgenic offspring has also been demonstrated recently. Rassoulzadegan and colleagues observed that male mice carrying the Kittm1Alf transgene had spotted white feet and tail (Fig. 4). When these mice were crossed with wild-type females they produced offspring that all had the spotted white feet and tail phenotype, including those that had not inherited the Kittm1Alf transgene (Rassoulzadegan et al. 2006; see also Fig. 4). Moreover they were able to show that injection of Kittm1Alf RNA into single cell mouse embryos produced offspring that all had the spotted white feet and tail phenotype (Rassoulzadegan et al. 2006). This suggests that Kittm1Alf RNA from transgenic males can be transferred in the sperm to non-transgenic offspring inducing the phenotype in the absence of the transgene itself, a complication not previously recognized.
The two sets of data given above have profound implications since they show that the use of non-transgenic littermates as controls may be inappropriate without careful comparison with wild-type mice. Most importantly the interpretation of data from all transgenic experiments may not be as simple as first thought if transgene products can produce epigenetic effects without evidence of the presence of the transgene in the genome.
The way forward
Future gene-deficient mice.
The ideal system to test the effects of a gene deletion should incorporate complete reversibility and no non-specific side-effects. Currently none of the commonly used methods satisfy these requirements since they rely on the use of transgenic mice with their inherent limitations. Our solution to this dilemma is to insert LacO sites into the promoter of our gene of interest by homologous recombination like those employed to introduce LoxP sites. By crossing these LacO mice with the ubiquitous LacIR-expressing mice, generalized gene expression can be regulated by the addition and removal of IPTG in the drinking water. Moreover, this system will also allow the regulation of mutated forms of genes of interest. In this case the mutation as well as the LacO sites will be introduced into the germline by homologous recombination, then regulated by LacIR and IPTG as above. Tissue-specific regulation of the gene can be obtained by expression of the LacIR protein driven from tissue-specific promoters. However, to avoid the problem of position effect, ‘knock-in’ methods of LacIR into tissue-specific promoters using homologous recombination should be implemented using IRES constructs as indicated above.
Future transgenic mice.
The ideal expression of a transgene should be totally tissue specific, not modify the host genome or be modified by the host genome, and be completely reversible. All current methods apart from site-specific insertions (see above) potentially suffer from modification by the local environment. Moreover, the only commonly used reversible system using tetracycline also suffers from side-effects from the tTA protein and possibly from the use of the antibiotic (see above). To overcome these problems, we are currently trialing a new method in which we are integrating our transgenes of interest into episomes (D. Rangasamy & K. I. Matthaei, unpublished observations). Episomes are circular, self-replicating DNAs often of viral origin that are stably and faithfully inherited during mitosis but are not integrated into the chromosomes of mammalian cells. Moreover, since they are not integrated, episomes carrying a transgene should not only avoid the ‘position’ effect from neighbouring genes but also not cause insertional mutation of an unknown gene thereby making them a possibly ideal transgene carrier.
Naturally occurring episomes are large since they require the production of an array of viral proteins to allow faithful replication and segregation into daughter cells (herpesvirus for example is 172 kb; Wade-Martins et al. 2000). This makes the genetic manipulation of episomes difficult. However, recently it was found that only one of the multiple proteins is necessary for episomal replication, the scaffold/matrix region of the human β-interferon gene cluster (Piechaczek et al. 1999) resulting in a plasmid of only 6.7 kb. Moreover this episome avoids insertional mutagenesis and methylation (Jenke et al. 2004) suggesting that it is an ideal transgene carrier. It is our intention to use the LacO/LacIR control and episomal expression of the transgene to generate transgenic mice that have fully reversible, tissue-specific and temporal control of transgene expression without modification from the host cell. If successful, this will be a major breakthrough for the generation of transgenic mice.
Conclusion
It is not my intention to suggest that all genetically modified mouse experiments are flawed or that useful information has not been generated from genetically altered mice. I have also limited my discussion to situations where tissue and temporal control of gene expression is required. Simple deletions of gene function will continue to have a place in some situations as will transgenic mice using minimal promoters. However, it is imperative to understand the limitations of these systems so that appropriate controls can be included and correct conclusions can be made. It is for this reason that Hagg (1999) should be congratulated. He was one of the first to point out that there was a problem with mixed genetic backgrounds in gene deletion experiments. Hagg and colleagues had published a specific phenotype of the deletion of the p75 nerve growth factor receptor (Van der Zee et al. 1996) but later discovered that the background strains were responsible for the phenotype and not the gene deletion. The senior author therefore corrected the literature by retracting the original result (Hagg, 1999). Many other published data may be similarly affected by previously unforseen problems. My intention here is to alert the reader to such possibilities in order to make the best use of a powerful tool which can be used to better understand gene expression and function and thereby health and disease. At the same time, discussions like these might stimulate the reader to re-assess their data in the light of some of the complications presented here, particularly when ‘inexplicable’ results are obtained.
In conclusion, the potential to use completely reversible methods like the LacO/LacIR system will allow each experimental group of animals to be their own control and should help detect epigenetic (or other) influences in studies of both gene overexpression and gene deletion thereby providing exciting new developments for the future.
Acknowledgements
I wish to thank Caryl Hill for constructive criticism and many helpful discussions on the content as well as careful editing of the manuscript. I also thank Heidi Scrable for the HuntingtinLacO-luciferase (HDOSluc) reporter and the LacIR- expressing mice as well as candid discussions about the LacO/LacIR system. I finally thank Martin Goulding for use of the N/ZEG reporter mice in Fig. 3.
Footnotes
-
(Received 18 April 2007; accepted after revision 10 May 2007; first published online 10 May 2007)
- 2007 The Author. Journal compilation © 2007 The Physiological Society
References
Figure 1. Deletion of intra-LoxP DNA by Cre recombinase Mice transgenic for the ‘β-galactosidase Floxed stop’ EGFP construct (called FloxStop) express β-galactosidase and not EGFP. When crossed with a Cre mouse the doubly transgenic mice no longer express β-galactosidase due to its deletion (called delLoxP) and the mice are green due to the removal of the stop and de novo EGFP expression.
Figure 2. Reversible regulation of gene activity by the LacO/LacIR system A, luciferase activity in different tissues from HuntingtinLacO-luciferase (HDOSluc) mice and repression of activity in HDOSluc mice doubly transgenic for LacIR. B, de-repression of luciferase activity at 24 and 48 h after addition of 10 mm IPTG in the drinking water.
Figure 3. Epigenetic modification of offspring by transfer of Cre recombinase in oocytes A, male wild-type Cre-positive mice crossed with FloxStop/FloxStop females produce offspring that are all positive for FloxStop and half are also positive for Cre. Only the Cre-positive mice are green due to the deletion of the stop (delLoxP). B, female wild-type Cre-positive mice crossed with FloxStop/FloxStop males produce offspring that are all positive for Floxtop and half are also positive for Cre. However, all the offspring are green due to the deletion of the stop (delLoxP) even in mice not positive for Cre (circled). The phenotype is due to the transfer of Cre protein from the female into the oocyte where it deletes the Floxed DNA.
Figure 4. Epigenetic modification of offspring by transfer of RNA in sperm Male KittmAlf/+ mice with white spotted feet and tail tips crossed with wild-type mice produce offspring that all have the white spotted phenotype, even the wild-type mice not carrying the mutant gene (circled). The phenotype is the result of KittmAlf RNA transfer in the sperm from the KittmAlf male to the wild-type mice.
