Conditional Gene-Targeting in Mice: Problems and Solutions

Conditional mutagenesis using the Cre-loxP system in genetically modified mice is one of the most important tools to help us understand cellular or molecular interactions in in vivo models in biology. In particular, immunologists are fond of this versatile toolbox permitting tissue or cell-specific deletion of genes and studying the impact of this in models of development, homeostasis, and disease. Using the bacterial Cre-recombinase which recognizes the palindromic loxP sequence to induce genetic recombination was initially employed in transgenic mice by Rajewsky and colleagues (Gu et al., 1994) and has become one of the most powerful tools in modern biology. However, in immunobiology research, we frequently come across statements such as “we assessed the targeting efficiency of gene X, by crossing our Cre line with the fluorescent-reporter line..…” or variations thereof. While the use of a reporter system at first glance seems to be the diligent and logical thing to do, to assume that a Cre-line would specifically delete the gene of interest in a particular cell type because another gene is deleted specifically in that cell is wrong. The pitfalls of the Cre/lox system were already discussed 10 years ago (Schmidt-Supprian and Rajewsky, 2007). However, in many publications authors interpret their data assuming that a specific Cre-line permits gene-targeting equally between different loci within the genome. The Cre-lox system has numerous potential pitfalls, and here we wish to point out only a few and provide some simple solutions and guidance to verify the targeting efficiency and recombination frequency of the gene of interest (Table S1).

Limitations and Shortcomings in Different Cre Lines

Problem:

Traditionally, a Cre transgenic mouse line such as Cd4-Cre (Lee et al., 2001) is made by providing the Cre transgene with the specific promoters and enhancers or silencers. Later, bacterial artificial chromosome (BAC)-mediated transgenesis was established to ensure more accurate Cre expression, because BACs likely contain most if not all the required regulatory elements (Gong et al., 2007). Nevertheless, distal cis- or trans-regulatory elements might be present and are not included in the chosen BAC, which could lead to Cre expression in undesired cell types or tissue locations. Finally, Cre can also be directly targeted into the desired endogenous gene locus either to replace the coding sequence, such as Cd19-Cre (Rickert et al., 1997), or at the 3′ untranslated region (UTR) like Foxp3-Cre (Rubtsov et al., 2008) of the target gene. The latter method is often combined with the insertion of an internal ribosomal entry site (IRES) allowing the expression of a bicistronic message encoding both the endogenous gene and Cre. While the “knockin” Cre approach permits Cre expression under the control of the endogenous promoter and gene regulatory elements, it still has its shortcomings. For example, for haplo-insufficient genes, Cre can only be inserted outside the coding region, but as is the majority of post-transcriptional regulatory events occur in the 3′UTR, appending a large gene sequence to the end of the RNA message could potentially impact the endogenous gene expression (Wan and Flavell, 2007).

Solution:

Whenever possible, Cre mouse lines should express Cre fused with a fluorescent reporter gene, allowing for the easy and direct detection of Cre expression. Also, as potential Cre toxicity or altered target gene expression resulting from Cre insertion can both lead to unwanted biological consequences (Wan and Flavell, 2007), to avoid misinterpreting the experimental results, Cre expressing mice without a loxP-flanked target should be used as controls.

Too Much, Too Little, But Almost Never on Target

The Problem:

The assumption that the pattern of Cre-expression equals the pattern of Cre-mediated recombination is erroneous. Reporter mice are useful tools, to determine Cre-activity. Cre-reporter mice carry a loxP flanked stop cassette in front of a reporter beacon such as lacZ or a fluorescent protein inserted into one specific locus (most often the ROSA26 locus). But targeting of the ROSA26 locus occurs with vastly different efficiencies than targeting in distant genetic loci. This is because each locus will have a distinct and unique “sensitivity” to Cre-recombination. While some loci are targeted with an extremely high efficiency, other genetic loci are largely resistant to targeting. In other words, a fluorescent reporter may indicate the desired cell- or tissue-specific targeting using one particular Cre-line. The very same Cre-line might target another genetic locus with only 10% efficiency, whereas yet another gene might be targeted across multiple cell types or even in the entire body. The resulting targeting pattern might thus be the desired 100%, a negligible and biologically irrelevant targeting of a small number of cells, or resemble a full- germ-line knockout mouse. The immunologist might have a perfect Cre-line with nearly perfect targeting in reporters, but the biology of the gene of interest might be invisible if the targeting is poor or vastly exaggerated if the targeting shoots far beyond the desired goal.

Solution:

Whereas the fluorescent reporter will give some indication regarding the pattern of Cre expression, deletion of each other gene with that Cre-mouse must be individually verified ideally at the single cell level. If an abundant protein is to be targeted in circulating leukocytes and can be detected easily by flow cytometry, then this would be a simple and easy solution to determine targeting frequencies. If the gene-product cannot be detected through a specific antibody, then RNA expression or the analysis of the genomic structure in purified cells may be a surrogate. We suggest that this should be combined by interrogating genetic recombination across different cell types (in particular those which might contribute to the phenotype observed).

Alternatively, one could use the biological properties of the target gene-product as a surrogate in measuring targeting efficiency. For instance, if the detection of a cell surface receptor is difficult or provides only a weak signal (e.g., Csf2rb), one may use the signal transduction pathway as a detection method (e.g., P-STAT-5) (Croxford et al., 2015). Again, ideally targeting is verified at the single cell level and by comparing other potentially targeted cell populations, but when that is not possible, one could use purified cell populations to determine how well they are targeted compared to other cell types which may be involved in the biological process studied. The best solution for future experiments however is to generate a conditional loxP-flanked allele with the reporter inserted in the target locus itself such that the reporter will only be activated following Cre-mediated recombination (Klein et al., 2006).

Pre-Existing RNA, Cre-Levels, Cellular Longevity, and Repopulation

Problem:

Expression of the Cre-recombinase has to be in a defined population of cells and not in any other cellular popula tion. The Cd19-Cre line, developed in the lab of Klaus Rajewsky (Rickert et al., 1997) is a good example: B cells all express CD19 and practically no other cell type. Also, most B cells are long-lived cells. As a result, conditional gene deletion in B cells with the Cd19-Cre mice is usually both efficient and specific. But this is very different if cells are short lived, like neutrophils, as new cells without deletion arise from the bone marrow constantly. Another problem with conditional gene targeting in general is that sometimes only one allele of the gene is deleted and not both. This is particularly evident for genes, which are vital for cell survival leading to selection pressure for partially targeted alleles. Lastly, when dealing with aberrant gene deletion from a particular Cre line combined with a specific loxP-flanked target, one should take note that Cre-mediated recombination can be variable even across different littermates ranging from cell type or tissue-specific to virtually ubiquitous in non-target tissues (Heffner et al., 2012).

Solution:

When feasible, generate mice with one germ-line deleted allele and one conditional allele to enhance deletion efficiency, with the following caveat: for genes that are haplo-insufficient, this approach cannot be used. Another caveat is that this is not practical for cases where more than one gene is deleted, so we propose this solution only for genes that are haplo-sufficient and only when only one gene is mutated. Regardless what approach is taken, to avoid using mice with apparent off-target deletions, at the very minimum, one should incorporate PCR primer sets that would allow the detection of the deleted allele in addition to both the WT and conditional alleles. To this end, unless a gene is supposed to be deleted in skin epithelial cells, a deleted allele should not be seen in a small piece of skin that was collected for genotyping. This is particularly critical for Cre lines that yield a high level of germline deletion. This simple method can quickly filter out the mice displaying clear nonspecific Cre excision and greatly increase the reproducibility between different experiments. Moreover, the intended target cells along with control cell populations could also be isolated and assessed by this method to confirm the efficiency of Cre/loxP-mediated gene deletion.