Capricious Cre: The Devil Is in the Details

Over the last few decades, the mouse has supplanted the rat as the primary animal model in biological and biomedical research (1, 2). This shift has occurred for a variety of reasons, including ease and cost of maintenance, short life cycle, and the wealth of classical genetics tools (inbred and congenic strains and spontaneous mutants). But the innovation most central to the ascendance of the mouse is the advent of mouse genetic engineering. Beginning with transgenic mice in the early 1980s (3–7) and the subsequent generation of targeted genetic manipulations (knockin/knockout) in the late 1980s (8–10), the use of genetically engineered mice has grown to become a cornerstone of biomedical research. With the subsequent application of Cre/Lox technology (11), modern day researchers have a mammalian system in which any genetic sequence can be modified in nearly any tissue or cell type. Yet the benefits of any technology also come with limitations, and in this issue of Endocrinology, Padilla and colleagues (12) highlight a critical complication that can arise with Cre/Lox technology, namely the transient expression of Cre in unanticipated sites during development. The result is Cre expression that may not match the adult expression pattern and as such may not meet the investigator's expectations.

The ability to modify genetic sequences proved to be a boon for scientific discovery, and the Cre/Lox system represents an additional refinement of the original knockout approach. Although whole-body deletion is often highly valuable, in many cases, these modifications produce either lethal phenotypes or phenotypes that are so severe that it is difficult to distinguish primary from secondary effects. Cre/Lox technology addresses this limitation by providing a means to produce cell-specific modifications. The Cre/Lox system is derived from the P1 bacteriophage (13), which uses the enzyme Cre recombinase to circularize and replicate its genomic DNA at distinct sequences called LoxP sites. Applying this system to mammalian cells (14), mice expressing Cre recombinase in a cell-specific fashion can be crossed with a separate line in which a targeted sequence is flanked by LoxP sites (termed floxed). The result is a recombination event that is localized to a specific cell type (11, 15, 16). Cre/Lox and similar systems (flp-frt recombination) have had a major effect on biological research, greatly increasing the pace of research and providing a means to test novel hypotheses.

Padilla and colleagues (12) focus on a specific application of this technology, the use of POMC-Cre mice to generate either POMC-specific knockouts or reporter mice. Proopiomelanocortin (POMC) is a precursor protein that is cleaved to several notable peptide products; ACTH, α-MSH, γ-MSH, and β-endorphin. Within the adult brain, POMC is expressed within two locations. The first is the nucleus of the solitary tract in the brainstem, and the second is the arcuate nucleus in the hypothalamus. The expression of POMC has been used as a defining characteristic for these individual neural populations, and a large literature has focused on arcuate nucleus POMC neurons as regulators of food intake and body weight homeostasis (17). Considering the prominence of these neurons in our current understanding of body weight homeostasis, it was virtually a given that Cre/Lox technology would be used in an effort to induce modifications specifically within these neurons. Central to this approach is the use of the Pomc promoter to drive the cell-specific expression of Cre recombinase. Yet it was earlier work by Padilla and colleagues (18) that initially raised concerns about the reliability of Pomc as a Cre driver. These early experiments focused on POMC expression during development, demonstrating that POMC was transiently expressed in a wide range of neurons before eventually becoming localized to a few discrete brain areas in adulthood (18). In other words, a large number of neurons express POMC during development, many of which develop into neurons that are phenotypically distinct from adult POMC neurons.

Although not directly addressing the use of POMC-Cre mice, the implication of this original work was obvious. Even a brief burst of Pomc promoter activity during development would likely be sufficient to drive Cre expression, resulting in recombination in unanticipated locations. In the current article, Padilla and colleagues (12) demonstrate just that result. The key experiment is the generation of a double-transgenic mouse, in which enhanced green fluorescent protein expression is under the control of the Pomc promoter whereas a tomato-red reporter is induced by POMC-Cre dependent recombination. The result is enhanced green fluorescent protein (reflecting current Pomc promoter activity) within a relatively narrow group of cells, but tomato-red (reflecting a previous Cre-dependent recombination event) being expressed much more widely. Many of the neurons labeled by the Cre-dependent reporter were not the classic anorectic POMC neurons, because the Cre-labeled neurons also included orexigenic neurons [expressing neuropeptide Y (NPY) and agouti-related peptide (AgRP)] and other as yet unidentified hypothalamic neurons. These data therefore suggest that any theoretical knockout mouse made using this POMC-Cre line would produce deletions that were not restricted to the adult POMC neuron population.

This transient, developmental expression is not unique to POMC. Glial-fibrillary acid protein (GFAP) is well known and routinely used as a marker of glia, because in the adult brain, GFAP is strongly expressed by astrocytes but not neurons. GFAP-Cre would therefore seem to be an ideal model with which to generate glial-specific knockouts. Yet GFAP is transiently expressed in nonglial cells during development, including a wide range of neuronal progenitors (19, 20). Similar observations have been made for other promoters: the neuron-specific nestin seems to be expressed in kidney (21); the adipocyte-specific AP2 seems to be transiently expressed in nonadipose cells, including brain (22); the β-cell-specific insulin promoter is expressed in neurons within the central nervous system (23); and the serotonin transporter Sert is expressed in serotonin and nonserotonin neurons during development (24). Lastly, off-target Cre expression is particularly important if it involves germline expression, which in extreme cases can result in whole-body gene deletion instead of the anticipated cell-specific deletion (24, 25).

Another problematic area of mouse genetics is developmental compensation. In many knockout and Cre-inducible strategies, the deletion or modification event is induced early in life, sometimes in the embryonic stage, and it seems that nature has a remarkable ability to compensate for these early alterations. One example is the effect of AgRP neuronal deletion. AgRP neurons are counterparts of POMC neurons within the arcuate nucleus. These neurons (which also contain NPY) are classically associated with the stimulation of food intake, and it was long assumed that these neurons were key drivers of food intake. Yet knockout mice lacking AgRP, NPY, or both neuropeptides, or neonates with deletion of the entire NPY/AgRP neuron, had normal food intake and energy homeostasis (26–28), suggesting that these neurons were not important for normal food intake. Yet in each of these studies, the deletion occurred relatively early in life, and it was later observed that deletion of AgRP neurons in adult animals rapidly reduced food intake and ultimately led to starvation (28). These observations not only indicate that NPY/AgRP neurons are critical regulators of food intake, but they also suggest that the brain can somehow compensate for the loss of these neurons in early development. As such, these data provide another example of how developmental timing can influence the experimental outcome.

Moving forward, there are also a number of strategies that can be employed to further address the issues noted here. For instance, various inducible models (Tet On/Off and Tamoxifen) give the investigator the ability to induce the deletion or modification at any point in the animal's life cycle, avoiding developmental compensation or developmentally transient expression. Combining Cre/Lox with viral approaches provides the ability to acutely manipulate specific cell populations at a specific point in time. The growing use of optogenetic and DREADD (designer receptors exclusively activated by designer drugs) approaches further increases the ability to induce acute, site-specific effects (29, 30). Lastly, there are an increasing number of tools for the independent validation of Cre expression. The Jackson Laboratory (Bar Harbor, ME) currently provides an analysis of both adult and embryonic expression for many of its Cre lines, and databases such as the Allen Brain Atlas provide a public resource for high resolution images of in situ hybridization from adult and embryonic mouse brains. As these resources improve both in quality and coverage, they may prove invaluable to investigators seeking the ideal Cre line for their work.

Mouse genetic engineering was once a cottage industry, with small groups independently making the mice they then used in their own experiments. Since those early days, the development and availability of various mouse models has rapidly increased, and today the procurement of knockout, Cre, and floxed lines is often as simple as ordering from a catalog. Considering the critical role that mouse models have played in a multitude of discoveries, we can expect their increased use to further accelerate the pace of scientific discovery. Yet the work of Padilla and others serves to remind us that genetic engineering is not a panacea. As genetic models become more ubiquitous, both users and reviewers need to be increasingly conscious of potential pitfalls. The scientific community expects both controls and validation, but currently there is little guidance or consensus regarding what constitutes the appropriate validation of a new model. Perhaps we would all benefit by engaging in such a discussion and in so doing ensure that the great power that these models provide is used to its fullest potential.