PHYSIOLOGICAL REVIEWS   Vol. 78 No. 4 October 1998, pp. 1131-1163
Copyright ©1998 by the American Physiological Society

Using Knockout and Transgenic Mice to Study Neurophysiology and Behavior

MARINA R. PICCIOTTO AND KEVIN WICKMAN

Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut; and Department of Cardiology, Children's Hospital, Harvard Medical School, Boston, Massachusetts

I. INTRODUCTION
II. WHY CREATE A MUTANT MOUSE?
III. GENERATION OF MUTANT MICE
    A. Transgenic Mice
    B. Inducible Expression of a Transgene
    C. Gene Knockout Through Gene Targeting
    D. Inducible and Tissue-Specific Knockouts
IV. MOUSE STRAIN
V. EXISTING PARADIGMS THAT HAVE BEEN USED TO STUDY TRANSGENIC ANIMALS
    A. Physiological Measures
    B. Behavioral Tests
    C. Mouse Models of Human Neurological and Psychiatric Disease
VI. CONCLUDING REMARKS
REFERENCES

	ABSTRACT

Top References

Picciotto, Marina R., and Kevin Wickman. Using Knockout and Transgenic Mice to Study Neurophysiology and Behavior. Physiol. Rev. 78: 1131-1163, 1998.  Reverse genetics, in which detailed knowledge of a gene of interest permits in vivo modification of its expression or function, provides a powerful method for examining the physiological relevance of any protein. Transgenic and knockout mouse models are particularly useful for studies of complex neurobiological problems. The primary aims of this review are to familiarize the nonspecialist with the techniques and limitations of mouse mutagenesis, to describe new technologies that may overcome these limitations, and to illustrate, using representative examples from the literature, some of the ways in which genetically altered mice have been used to analyze central nervous system function. The goal is to provide the information necessary to evaluate critically studies in which mutant mice have been used to study neurobiological problems.

	I. INTRODUCTION

Top Next References

One fundamental goal of biological science is to determine the physiological function of identified proteins. This is a daunting task, particularly since a given protein functions within the context of thousands of other proteins found in every cell. For proteins found in the brain, the center of such complex activity as cognition, emotion, and learning, the task is even more daunting. One approach is to remove the protein from its normal environment and study it in a more simple setting, such as a reconstituted in vitro system. Although inferences with respect to normal physiological role can, and have, been made from such studies, conclusions are necessarily restricted. In vitro and heterologous expression systems are useful precisely because they reduce the complexity of the living cell. Nature has provided some clues, in the form of genetic diseases, to the physiological relevance of certain proteins. Linkage of particular gene defects to genetic diseases has greatly enhanced our understanding of the role of many proteins. Indeed, studies of naturally occurring gene mutations leading to neurophysiological defects in mice and humans have enhanced our understanding of several proteins found in the central nervous system (CNS) (for reviews, see Refs. 11, 57, 203, 237). More recently, technology has evolved that allows the specific modification of the genetic composition of many organisms. This type of reverse genetic approach, in which detailed knowledge of the gene of interest permits in vivo modification of its expression or function, provides a powerful method for examining the physiological relevance of any protein.

When studying proteins involved in biochemical or developmental pathways, simple cellular processes, or behaviors conserved throughout evolution, the classical subjects of genetic research such as yeast, Drosophila melanogaster, and Caenorhabditis elegans are superior to the mouse from the standpoints of ease of genetic manipulation, the number of organisms that can be generated and studied, and the cost and time investment. Unique features of yeast make disruption of specific genes relatively simple, while studies with Drosophila and C. elegans are not restricted by the relatively slow time course of mouse studies. Indeed, the 3-wk gestation period, subsequent 2 mo to reach sexual maturity, and comparably small litter size (usually 2-10 pups) make mouse studies time and space consuming, and therefore costly. Mice possess a distinct advantage over classical genetic subjects, however, in that mice are more closely related to humans from a physiological perspective. Although some genetic studies have been performed using rats (for review, see Ref. 37), the vast majority of transgenic research has used mice, primarily because of the greater technical ease of manipulating the mouse embryo, the lack of classical genetic information (genetic locus markers), and the smaller number of inbred rat strains. In addition, rats lacking a specific gene of interest have not been generated, since it has not been possible to date to create pluripotent rat embryonic stem cell lines that are necessary for this technique (see sect. IIIC for discussion of embryonic stem cells). Thus, when studying the role of a protein in a complex behavior such as learning, or when trying to model a human disease, mice are often the research subjects of choice.

Complex behaviors are determined by the function of many gene products. Thus some of the same features that make mouse models advantageous for studying the molecular basis of complex behaviors can make the interpretation of phenotypes resulting from mouse mutagenesis difficult. For example, several variables, including the expression level and allelic variants of other genes, can influence phenotypes revealed by physiological studies. Thus useful information from studies of genetically altered mice can only be obtained if appropriate controls are employed and data are interpreted within the context of the experimental limitations.

Genetically altered mice have been used in the fields of immunology and developmental biology with great success for more than a decade. In the last 5 years or so, we have seen increased use of mutant mice to study the complex functions of the CNS. One goal of this review is to illustrate, using representative examples from the literature, some of the ways in which genetically altered mice have been used to analyze CNS function. Because of the logarithmic expansion in mouse genetic studies involving molecules important in the CNS, this review cannot be comprehensive. A second goal of this review is to familiarize the nonspecialist with the techniques and limitations of mouse mutagenesis and to describe new technologies that may overcome these limitations. Interested readers are also directed to recent excellent articles that discuss specific techniques (37, 196, 264, 282) or review in a more comprehensive manner the data that have been obtained from mutant mice (6, 40, 55, 286).

	II. WHY CREATE A MUTANT MOUSE?

Top Previous Next References

Genetically altered mice are used increasingly as tools to define or clarify the in vivo function of molecules that have been studied in vitro. The phenotypes of a mutant mouse can be studied at many levels from biochemistry to cell biology to systems physiology to behavior. One of the advantages of a mutant mouse that survives to adulthood is that the effect of a single gene alteration on a complex behavior can be studied. For example, the role of many signal transduction pathways in learning and memory has been investigated using mutant mice.

In addition to being a tool for understanding the physiological function of particular proteins, mutant mice can also be used to model human diseases. As the genes responsible for human genetic diseases are identified through linkage studies, it becomes feasible to create mouse models of those diseases (reviewed in Refs. 5, 27, 52, 156, 164, 218). Current technology allows modeling of both gain-of-function disorders (in which a mutant protein with altered function is expressed or a wild-type gene is overexpressed) and loss-of-function disorders (in which an endogenous gene is inactivated or a mutant protein is nonfunctional). The hope in these studies is that the altered mice will exhibit symptoms similar to a human patient and that the pathogenesis of the disease is similar between the species. There are several advantages to a mouse model of human disease, including the possibility of studying the gene mutation in multiple individuals with a homogeneous genetic background, the ability to study disease onset and progression in a highly controlled environment, and the ability to study early stages of a disease before the appearance of overt symptoms. Currently, mice are being used to model several human neurological diseases, including Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), and transmissible spongiform encephalopathies (prion diseases). Although no perfect model of any of these disorders exists, invaluable information regarding disease mechanisms has already been gained from these attempts.

Several other benefits can come from mutant mouse studies. Assumptions about the in vivo targets of particular drugs can be tested rigorously using genetically altered mice. By ablating a single protein or subunit of a complex, one can test whether the protein is the target of a commonly used agonist or antagonist. For example, the ₂-subunit of the GABA_A receptor was thought to be a critical site of action for benzodiazepines, widely used anxiolytic agents (217). This hypothesis was confirmed in mice lacking the ₂-subunit, which had a 94% reduction in benzodiazepine binding sites in the brain (98). Similarly, the ₆-subunit of the GABA_A receptor was thought to mediate the behavioral effects of ethanol (163); mice lacking the ₆-subunit, however, exhibited unimpaired ethanol-induced sleep (112), demonstrating that at least one aspect of ethanol function is not mediated through this subunit. When typical pharmacological tools are unavailable, mutant proteins that ablate the function of their wild-type counterparts (dominant negative mutants) can be used as in vivo antagonists. For example, a dominant negative form of the cAMP-responsive transcription factor CREB was expressed in mice by transgenesis to evaluate the role of CREB in the transcription of the tyrosine hydroxylase (TH) gene (154). Overexpression of the dominant negative form of CREB inhibited TH transcription, implying that endogenous CREB, or another CREB family member, normally drives this process.

	III. GENERATION OF MUTANT MICE

Top Previous Next References

For those not intimately involved in the generation and study of mutant mice, it can be difficult to evaluate such studies critically. In this section, we describe the theory, methodology, and limitations involved in the generation of mice expressing (transgenic) or lacking (knockout) a particular gene of interest.

The creation of a transgenic mouse begins with the selection of a transgene, the DNA element to be transcribed, and usually translated, in the mouse. The element is often a cDNA encoding all or part of a particular protein, but it can be a genomic fragment containing all or part of a gene, or an antisense fragment designed to deplete the level of an endogenous mRNA (for examples of transgenic antisense studies, consult Refs. 82, 192, 193). The transgene is subcloned downstream of a suitable promoter element that will drive its expression. To a large extent, the promoter element determines the level, tissue specificity, and temporal pattern of transgene expression. For studies in the nervous system, transgene expression can be controlled by characterized minimal promoter sequences that target the gene to specific cells or brain regions, such as the glial fibrillary acidic protein (GFAP) (25) or myelin basic protein (94) promoters (glia), neuron specific enolase (76) or calmodulin II (177) promoters (neurons), calcium/calmodulin-dependent kinase II (CaMKII)- promoter (forebrain/hippocampus; Refs. 181, 278), or by elements providing a more ubiquitous neuronal expression pattern such as the -actin promoter (134). If a close approximation of the expression pattern of the endogenous gene is desired, as is the case in most disease-modeling experiments, one can use the gene's own promoter. The temporal pattern of expression of the transgene can also be affected by its promoter. Indeed, some studies have utilized promoters that turn on after birth to avoid potential developmental effects of chronic transgene expression during prenatal development (181).

The transgene is introduced into single-cell mouse embryos by pronuclear injection (Fig. 1; for an extensive review of practical methods, consult Ref. 111). Once inside the pronucleus, the transgene either integrates into the genome in a random fashion or is degraded by exonuclease activity. Those embryos that integrate the transgene usually acquire several copies in a tandem head-to-tail array. Transgene insertion usually occurs quite early in the developmental process; thus most or all of the cells that comprise the resultant mouse contain the transgene. However, if insertion occurs after several rounds of cell division, the transgene may be present in only a subset of cells. The resulting mouse will be mosaic, with only a subset of cells containing the transgene. If the transgene is present in cells comprising the germ line, it will be transmitted to subsequent generations of mice. Mice that transmit the transgene through the germ line are termed founder mice, and although they can be examined for phenotypic consequences of transgene expression, founders are more often used to generate large numbers of transgenic animals. Occasionally, the transgene integrates into more than one site in the genome, and offspring from such a founder animal may inherit one or the other copy of the transgene, resulting in multiple lines with potentially distinct phenotypes from a single founder (278).

View larger version (15K): [in this window] [in a new window]   FIG. 1.   Classical transgenic construction. A: a segment of DNA to be transcribed (cDNA, genomic fragment, antisense fragment) is subcloned downstream of a neuron-specific promoter (e.g., neuron-specific enolase). B: after separating transgene from plasmid sequence, isolated fragment is injected into pronucleus of a single-cell embryo. Ten to twenty injected embryos are implanted into a foster mother. Some mice (transgenic founders) carry transgene in all cells including germ line (eggs or sperm) and thus can transmit gene to subsequent generations. Gene of interest should be expressed predominantly in brain, although insertion effects may result in ectopic expression.

Not all founder mice are alike, despite the fact that inbred mouse lines, in which all mice are genetically identical, are often the source of the embryos. One key difference between founders is the number of copies of the transgene integrated into the genome. In fact, founder mice vary widely in the number of transgene copies, usually between 2 and 50, that are integrated into the genome. The expression level of the transgene is often positively correlated with its copy number. In some cases, then, mice containing different numbers of transgene copies can exhibit distinct phenotypes. For example, an attempt to model familial ALS using a single transgene yielded several mouse lines that all exhibited progressive muscle wasting leading to death, but the disease progression rate was generally proportional to the transgene copy number, and histopathological features observed in certain lines were not observed in others (54).

Although copy number can affect the level of transcription of the transgene, a more important distinction between founder mice often involves the site of integration of the transgene within the genome. Most transgenic studies show extensive positional effects, including undesirable expression of the transgene in a particular tissue(s) (ectopic expression). The level and spatial distribution of transgene expression are sensitive to its proximity to transcriptional activators or silencers in the genome. If the transgene of interest is expressed under the control of a neuron-specific promoter, but inserts into the genome near a liver-specific enhancer, the transgene could be expressed both in neurons and in the liver. Thus ectopic, inappropriate, or unintended expression of the transgene can lead to phenotypes that nonspecifically affect the nature of the system to be studied. Silencer elements are also distributed throughout the genome, making it possible that some integration sites yield little or no expression of the transgene. To control for positional influence when screening for phenotypes, it is necessary to generate a number of founder mice and compare phenotypes among their offspring.

Just as the genome can influence transgene expression, the transgene can alter the expression of neighboring genes. Although only a small percentage of the genome consists of actual protein coding sequence, genes often span several hundred kilobases of DNA. Therefore, transgene integration can disrupt protein coding sequence, promoter elements, or other less-understood regulatory elements controlling gene expression. In one case, insertion of an interferon transgene disrupted the monoamine oxidase A gene, resulting in hyperaggressive mice with high monoamine levels due to the disruption rather than the overexpression of interferon (34). Random transgene integration has been exploited by other groups to purposely disrupt, and consequently mark, genes that can be cloned later by screening of a library constructed from genomic DNA (61, 78).

Since positional effects related to transgene integration site were first appreciated, attempts to minimize inconsistency in phenotypes between founder mice have been made. One approach to combat ectopic expression and positional effects has been to create mice using transgenes consisting of large (>50 kb) genomic fragments containing intact genes and regulatory elements with desired alterations. Unfortunately, genomic fragments of this size are not easily manipulated or mutated in classical bacterial plasmids or cosmids. In contrast, yeast artificial chromosomes (YAC) are useful cloning vectors that can accommodate very large genomic fragments (up to 2 megabases) that can be modified efficiently in yeast by homologous recombination (213). In brief, a small plasmid construct carrying a portion of the large genomic fragment with the desired alteration, such as a cDNA to be used as a transgene, plus a selectable marker, is introduced into yeast (Fig. 2). The fragment carrying the desired mutation then recombines or exchanges with the homologous portion of the large fragment in the YAC, transferring the alteration to the large fragment (23, 269). Although YAC have been used successfully as transgenes in mouse experiments, several features of this system have limited its utility. First, isolation of intact YAC and introduction into the germline by classical pronuclear injection are difficult because of the unavoidable shearing stress and viscosity of these large fragments. Although more gentle alternatives to the pronuclear injection of YAC into mouse embryos have been developed, YAC also exhibit a high degree of chimerism and clonal instability (189). In other words, a YAC may acquire deleterious or unwanted mutations when passaged in yeast. Bacterial artificial chromosomes (BAC, based on the Escherichia coli fertility factor) and P1-derived artificial chromosomes (PAC, based on bacteriophage P1) are cloning vectors that propagate genomic fragments up to ~300 kb in E. coli and have several advantages over YAC. First, BAC and PAC are stable and exhibit minimal chimerism when passaged in recombination-deficient bacteria (189). Second, purification of PAC and BAC is relatively easy since they exist as supercoiled circular plasmids that are resistant to shearing, making them amenable to pronuclear injection. Third, as was recently demonstrated, BAC can be modified by homologous recombination and utilized as standard transgenes (296). Thus YAC, BAC, and PAC technologies could dramatically improve the quality of mouse transgenesis in the future.

View larger version (17K): [in this window] [in a new window]   FIG. 2.   Use of yeast artificial chromosomes (YAC) as transgenes. Very large fragments of genomic DNA can be manipulated using YAC as vectors. In this example, promoter sequences contained within a large fragment of DNA can be used to drive expression of a cDNA of interest. A: first, a YAC is identified containing desired promoter. A plasmid is then constructed containing cDNA of interest flanked by short segments of genomic DNA homologous to site at which transgene will be inserted (in this example, insertion site would be immediately after promoter elements). This transgene plasmid would also contain a selectable marker such as URA3 gene that allows yeast to grow in absence of uracil in medium. B: cDNA-containing plasmid is then introduced into yeast carrying YAC, allowing recombination to occur between cDNA-containing vector and YAC. In this example, cDNA is inserted downstream of large promoter fragment in position of endogenous gene. cDNA-containing plasmid itself does not replicate in yeast; therefore, only those yeast that have undergone recombination with plasmid, and incorporated URA3 marker, will survive when plated on medium lacking uracil. C: recombined YAC can then be isolated and injected into oocytes like a standard transgene. This technique could also be used to express a mutated transgene under its own promoter.

Several inducible systems are available that are based on a common theme (238). These inducible systems are exemplified by the tetracycline (Tet)-regulatory system (Fig. 3; Refs. 79, 92, 137, 253). In this system, the generation and crossing of two different lines of mice are required. The first mouse line expresses the Tet transactivator (tTA) under the control of a promoter that directs its expression in the tissue or tissues of interest. Tetracycline transactivator is a transcription factor that regulates the expression of any gene downstream of its cognate promoter sequence (TetOp). The second mouse line harbors the transgene consisting of the gene of interest downstream of TetOp. Inducibility in the Tet system depends on the presence of Tet or a Tet analog such as doxycycline (Dox). In the "Tet-off system," Tet inhibits the interaction between tTA and TetOp, effectively suppressing transcription of the gene of interest.

View larger version (21K): [in this window] [in a new window]   FIG. 3.   Inducible transgene expression. In tetracycline (Tet)-off system, transcription of gene of interest (X) occurs when Tet transactivator (tTA) binds to TetOp promoter. In presence of Tet, tTA is blocked from activating transcription. System can be introduced into mice by creating 2 independent transgenic mouse lines. First mouse expresses tTA in a specific tissue (brain). Second line harbors gene of interest driven by TetOp promoter (TetOp-X). When lines are crossed, and in absence of Tet, those tissue(s) expressing tTA will also transcribe gene X. When tetracycline is present, however, transcription is blocked. tTA has also been modified so that gene transcription occurs only in presence of Tet (see text).

One disadvantage of the Tet-off system is that induction of the transgene only in the adult animal requires the persistent exposure of mice (beginning at conception) to levels of Tet or Dox sufficient to prevent transgene expression. This can be expensive and may result in unknown behavioral side effects. Indeed, exposure to Dox during development in one study resulted in adult mice with impaired spatial memory and fear conditioning (180). Furthermore, the rate at which induction can occur is limited by the clearance of Tet or Dox from the animal, and this rate varies from tissue to tissue (137). To maximize responsiveness and minimize the side effects of the antibiotics, mice are usually exposed to the minimum amount of Tet or Dox required to repress transgene transcription.

An alternative system that eliminates the necessity for chronic exposure to Tet or Dox utilizes a mutant form of tTA, reverse tTA (rtTA), which is activated rather than repressed by Tet or Dox (93). Thus the expression of the gene of interest occurs in the presence of Tet or Dox ("Tet-on"). Because transgene expression does not depend on the clearance of antibiotic from the tissue of interest, this system might offer more rapid induction. This system was used in transgenic mice, and induction over several orders of magnitude was achieved in 4 h in some tissues and was usually complete after 24 h (137). This study reported strict dependence of transgene expression on the presence of antibiotic, but a high degree of basal expression (leakiness) has been seen in cell culture using this system (Picciotto, unpublished data).

The Tet-off system has been used to direct transgene expression in specific brain regions at particular times in adult mice (180). The CaMKII- promoter was used to restrict tTA expression to the forebrain. These mice were then crossed with another transgenic line harboring a mutated (constitutively active) form of CaMKII- under the control of TetOp, to generate bigenic mice (mice containing both transgenes). In the presence of Dox, which penetrates the blood-brain barrier and regulates tTA activity more effectively than Tet, transcription of the mutant CaMKII transgene in the brain was repressed. When Dox was removed, the mutant kinase was expressed in different subregions of the forebrain, depending on the mouse line used. Interestingly, induction of the mutant kinase in the hippocampus throughout development altered their spatial learning capabilities and the electrophysiological responses of hippocampal neurons to changes in the environment, and this effect was reversed by treatment with Dox in the adult.

Other inducible systems based on the triggering substances interferon- (144) and glucocorticoids (135) have been reported, but the associated nonspecific effects of these drugs are pronounced in mice and could induce phenotypes unrelated to the transgene. A system based on induction of the progesterone receptor by the synthetic steroid hormone RU-486 has been used in mice to express a reporter gene (59), but once again, RU-486 has powerful effects on behavior and physiology on its own, and this could affect the interpretation of the transgenic phenotype. Another system involves the insect molting hormone ecdysone as an inducer that does not have a target in mammalian cells. Recently, a transgenic mouse line expressing a modified ecdysone receptor dimer was generated and was bred with a second mouse harboring a transgene consisting of the ecdysone responsive promoter and reporter gene (205). Induction reaching four orders of magnitude was achieved with low or undetectable leakage transcription. One disadvantage of this system is that three different transgenes must be present in the cell type of interest. The chances of obtaining mice expressing the receptor subunits in a uniform and overlapping fashion are therefore low, breeding strategies become complex, and controlling for integration site effects is difficult.

The critical variables that differ between existing inducible systems include the amount of transcription that occurs when transgene expression is intended to be off (leakage transcription), the extent to which expression can be increased (dynamic range), the speed with which transgene expression can be up- or downregulated (pharmacokinetics), and side effects related to the exposure of the mouse to the triggering substance. Available systems each have strengths and weaknesses that need to be addressed in the context of the experimental question to be answered.

A useful method to determine whether a gene product of interest is necessary for mediating a particular process is to abolish its expression and evaluate the consequences. Random gene inactivation can be accomplished using ultraviolet irradiation, mutagenic chemicals, viruses, or transgenes, all agents which randomly attack the genome. Coupled with in vivo selection assays based on the function of the gene product of interest, this approach can be used to clone novel genes, particularly in organisms with a short generation time such as C. elegans and D. melanogaster. In contrast to random mutagenesis, targeting via homologous recombination permits directed genomic lesions that either abolish gene expression or alter the gene of interest (for excellent reviews, see Refs. 32, 264, 284). As the phrase suggests, homologous recombination is the in vivo exchange of genomic sequence between homologous fragments of DNA. In gene targeting, two genomic fragments that closely match or are identical to regions in the genomic locus to be targeted flank a selectable marker (making up the targeting vector or targeting construct) (Fig. 4). The targeting construct is introduced into host cells, and when homologous recombination occurs, a part of the genomic DNA of interest will be replaced with DNA found between the homologous flanking domains in the targeting construct. Depending on the design of the targeting construct, several variants of the technique are possible, allowing deletions, point mutations, or replacements (264).

View larger version (23K): [in this window] [in a new window]   FIG. 4.   Classical gene knockout. A: wild-type allele, targeting vector, and targeted allele following homologous recombination. Targeting construct is introduced into embryonic stem (ES) cells by electroporation, and subsequent selection in G418 eliminates all ES cells that do not acquire a copy of this construct. Generally, homologous recombination occurs infrequently compared with random integration of construct. To enrich for homologous recombinants, diphtheria toxin (DTA) or thymidine kinase genes can be included in targeting construct. Homologous recombination can be detected by PCR screening using primers just outside shortest domain of homology used to make targeting construct. Neo-R, neomycin resistant. B: when ES cell clones are identified with a targeted allele, 10-20 cells are injected into blastocysts (3.5-day embryos), and 10-20 blastocysts are reimplanted into each foster mother. In some cases, injected pluripotent ES cells contribute to a large percentage of resultant mouse tissues (chimeric mouse). If ES cells and blastocysts were derived from mouse strains with different coat colors, degree of contribution of ES cells to resultant mouse can be approximated by coat color. Targeted allele can be transferred to subsequent generations if chimeric animal's germ cells were derived from targeted ES cells. Germline transmission is also monitored by following coat color.

From a practical standpoint, gene targeting in mice begins with the isolation and mapping of a mouse genomic fragment encoding all or part of the gene of interest. A combination of Southern blotting, sequencing, and/or PCR is used to determine the gene structure sufficiently to design a targeting strategy, as well as to develop a screening strategy to identify homologous recombination when it does occur. Usually, a strategy is developed so that an exon containing the initiator methionine of the gene of interest or a known functional domain is removed and replaced with a selectable marker such as the neomycin resistance (neo^r) gene driven by a ubiquitous promoter such as the murine PGK-1 promoter (4). Integration of the DNA into the host genome is thus selected for by incubation in neomycin-containing (geneticin, G418) growth medium (positive selection).

In mammalian cells, homologous recombination occurs at low frequency (33, 284). In the majority of transfected cells, then, the targeting construct inserts randomly. To enrich for cells that undergo homologous recombination, many targeting constructs contain a second selectable marker that flanks one of the homology domains (303). Because it is positioned outside of the area of genomic homology, it is spliced out when homologous recombination occurs and remains intact if the construct inserts randomly. In contrast to the positive selection marker, this marker kills the host cell when integrated as part of the targeting vector (negative selection). The most common negative selection markers are the herpes simplex virus thymidine kinase gene (272) and the diphtheria toxin gene (295).

Several factors influence the frequency of homologous recombination, or targeting efficiency, including the extent of identity between the construct and the genomic locus (303). Overall identity is dependent on the length of the genomic fragments (generally a total of at least 6 kb of homologous DNA) and the source of the fragments used to create the targeting construct. A positive correlation between the combined length of the homologous fragments and targeting efficiency has been observed (273). Targeting efficiency is also improved when the DNA fragments in the targeting vector are obtained by screening a genomic library made from the same strain of mouse that was used to derive the cells for the targeting experiment (271, 283). In addition, it seems that targeting efficiency is highly dependent on the targeted locus itself, probably because of factors such as the specific DNA sequence involved (103), extent of DNA methylation (161), or particular chromatin structure (219). Interestingly, the extent of the deletion created by the recombination event does not appear to reduce targeting efficiency (188).

Once made, the targeting construct is linearized and introduced into embryonic stem (ES) cells by electroporation (Fig. 4). The utility of gene targeting in mouse genetic studies was made possible by the isolation and culture of murine ES cells (21, 22), which are nontransformed cells derived from normal mouse embryos that have retained the ability to divide in culture but remain pluripotent. In other words, these cells are capable of differentiating into any tissue of the developing mouse. Embryonic stem cells have been isolated and cultured by several laboratories and are now commercially available. The vast majority of ES cells are derived from the 129 mouse strain, and thus the majority of gene targeting constructs are also prepared using genomic DNA from this mouse strain. However, it is likely that ES cells and genomic libraries from other strains will become increasingly available. The ability to culture ES cells makes them suitable for gene targeting experiments, whereas the pluripotency of ES cells makes it possible to derive mice from these cells, even after they have been cultured. The combination of these features forms the backbone of gene targeting experiments in mice.

After electroporation and antibiotic selection, a suitable number of colonies (generally between 50 and 500) are picked for DNA isolation and are screened by Southern blotting or PCR to detect ES cell colonies in which the DNA is correctly recombined. The method of screening dictates in part the design of the targeting construct (see, for example, Refs. 158, 215) as one arm of genomic DNA must be <2 kb long if PCR screening is to be used. After amplification of the colony, recombined ES cells are then microinjected into 3.5-day-old blastocysts, usually derived from the C57BL/6 mouse strain. After microinjection, blastocysts are reimplanted into the uterus of pseudopregnant females and are carried to term. Usually a fraction of the resulting mice are chimeric, that is, they have developed in part from the targeted ES cells and in part from cells of the donor blastocyst. Chimerism can be assessed visually if the ES cells and donor blastocysts are derived from strains of mice with different coat color. Male chimeras judged to be derived to a large extent (50% or more) from ES cells are then crossed to female mice, and germline transmission is assessed by the coat color of the resultant offspring. Embryonic stem cell gene targeting usually yields cells carrying a single disrupted allele of the gene of interest; thus offspring of chimeras have a 50% probability of harboring a targeted allele. These mice are typically identified by PCR amplification of a portion of the targeted allele from tail biopsy DNA. Mice containing one wild-type and one targeted allele (heterozygotes) can then be intercrossed to generate mice homozygous for the targeted allele (knockouts). Generally, the targeted mutation is inherited in a classical Mendelian fashion.

Depending on the gene-targeting strategy, different phenotypes can be observed. For example, different approaches to inactivating cystic fibrosis transmembrane conductance regulator (CFTR), the gene linked to cystic fibrosis, led to different phenotypes in the resultant mice. One group removed a functional portion of the CFTR coding region and replaced it with neo^r (261). These mice died at a young age and were severely affected by the mutation. A second group disrupted the CFTR coding region by neo^r insertion (65), and the resultant mice exhibited a less severe phenotype. Subsequently, the mice were shown to contain a small amount of functional CFTR protein, presumably because of a low-efficiency splicing event that removed the neo^r block in a portion of RNA transcripts (66).

Gene targeting does not necessarily mean gene ablation. Gene targeting can be used to introduce subtle mutations into a gene of interest (284), or as part of a strategy to generate a tissue-specific or inducible knockout (see sect. IIID). Gene targeting also provides an alternative for combating temporal and spatial ectopic expression that often plague classical transgenic experiments. For example, desired mutations or gene alterations can be introduced, or "knocked in" to the endogenous mouse gene. Because expression of the mutant gene is controlled by the same factors that regulate expression of the wild-type gene, this approach avoids the confounding positional effects associated with random transgene insertion. As such, knock-in mice are potentially powerful models of human genetic diseases, caused by mutations in one (dominant) or both (recessive) alleles (see, for example, Refs. 35, 201, 221). Unfortunately, the investment of time and reagents required to generate a knock-in (or knockout) mouse are considerably greater than those required to create a transgenic mouse.

As with transgenic studies, functional compensation could occur in a knockout mouse to mask or distort the phenotype resulting from the chronic absence of an endogenous gene. This is of particular concern when targeting a member of a large family of related genes and has been observed, for example, in mice lacking the transcription factor CREB. Knockout of CREB and CREB is not lethal to mice, due in part to increased levels of CREB and the CREB-related protein CREM in the knockouts (19, 118). Alterations in development can occur not only at the molecular level through changes in gene expression, but also at the level of neuroanatomy. For example, lack of the dopamine D₁ receptor subtype leads to altered morphology of the striatum, in which the striosomes seen in wild-type mice are absent (293). Chronic disruption of many gene products can also result in premature death (45, 56). All three scenarios make it difficult or impossible to determine the role of a protein of interest in an adult organism.

To overcome developmental phenotypes, spatial or temporal control of gene ablation can be achieved using the E. coli bacteriophage P1 enzyme Cre recombinase. Cre recombinase recognizes and binds to a 34-bp DNA sequence called a loxP site, which contains two 13-bp repeated sequences in opposite orientation flanking an 8-bp spacer (1, 108-110). If two loxP sites are present in a DNA fragment, that fragment is said to be "floxed" (flanked by loxP sites). Cre recombinase-mediated recombination results in the removal of the entire floxed segment, leaving one loxP site behind (Fig. 5). Once demonstrated to function in mammalian cells (242-244), the Cre-loxP system was applied to mouse genetic studies (Fig. 5; Refs. 18, 97, 209, 304). For a more detailed overview of this approach, please consult Reference 282. 

View larger version (25K): [in this window] [in a new window]   FIG. 5.   Conditional/tissue-specific gene knockout. To generate a tissue-specific knockout, slight modifications in targeting construct are required. In one of several possible approaches, 2 loxP sites (surrounding selectable marker cassette and gene fragment of interest) are introduced. Correct recombination can be determined by PCR. ES cells can then be injected into blastocysts to generate mice with a floxed allele. When mice homozygous for floxed gene are crossed with a transgenic mouse expressing Cre recombinase in a specific tissue (brain), resultant mice will delete floxed gene only in that tissue. These ES cells can also be used to generate a "classical" knockout through transient transfection with Cre recombinase. NEO, neomycin; CRE, Cre recombinase; DTA, diphtheria toxin.

Recently, this method was adapted to confine gene disruption to a particular brain region and to further restrict gene disruption to postnatal ages (278, 279). Transgenic mouse lines were generated using a portion of a forebrain-specific promoter (the promoter for CaMKII-) to drive expression of Cre recombinase (278). Mice derived from 1 of 14 founder mice expressed Cre recombinase selectively in the CA1 subfield of the hippocampus at sufficient levels to mediate recombination. Because this promoter was active only after birth, Cre recombinase expression was regulated both temporally and spatially. These mice were then crossed with mice containing a floxed N-methyl-D-aspartate (NMDA) glutamate receptor gene, NMDAR1, whose product comprises part of the multimeric NMDA glutamate receptor (279). The presence of the loxP sites, introduced by homologous recombination, did not disrupt NMDAR1 expression; however, the bigenic mice expressed Cre recombinase at high levels in the CA1 subfield of the hippocampus and therefore exhibited postnatal disruption of the NMDAR1 gene selectively in this brain region. This approach allowed the researchers to overcome the early lethality seen in classical knockouts of the NMDAR1 gene (75, 160) and to conclude that NMDA receptor function in this subregion of the hippocampus is critical to the formation of spatial memory (183, 279).

Creating targeting constructs that function in a Cre-loxP system is generally no more difficult than designing a classical targeting construct. One of the distinct advantages of the Cre-loxP system over the classical approach, however, is the flexibility it provides the researcher. Constructs designed with the Cre-loxP system in mind can be used to generate mice lacking a protein in a particular tissue to avoid early lethality or severe developmental consequences, or alternatively, in a slightly modified version of the classical knockout approach (i.e., floxed alleles can be ablated in ES cells after introduction of Cre recombinase). As the technique becomes more popular, the increasing availability of transgenic mouse lines expressing Cre recombinase in various tissues and time points will allow the researcher to control gene ablation precisely. The Cre-loxP system has already been used to make inducible knockouts by linking Cre recombinase expression to the interferon-regulated system, further enhancing the potential for temporal regulation of gene ablation (144). In addition, Cre recombinase has been fused to a modified estrogen receptor that renders it inactive in the absence of tamoxifen, an estrogen receptor ligand (72, 73, 247). This allows temporal control over gene deletion. Although not yet reported in a neuronal system, inducible knockouts specific to brain will likely be common in the near future.

	IV. MOUSE STRAIN

Top Previous Next References

Many strains of inbred mice have been developed and maintained. An inbred strain is, by definition, one in which brother-sister matings have been performed for at least 20 generations, and the line can be traced to a single ancestral breeding pair (166) (Table 1). This inbreeding results in a population of genetically homogeneous animals that are homozygous at every gene allele. The effect of mouse strain on behavior has recently received a lot of attention (86). Knockout of a particular gene can result in very different phenotypes depending on the strain used, reflecting the complex, multigenic nature of most phenotypes. In other words, if a phenotype is influenced by many genes, knockout of one relevant gene could have different effects depending on which alleles of the other genes are present. The effects can be particularly profound in behavioral paradigms (49). For example, the C57BL/6 strain is one of few inbred mouse strains that performs well in the Morris water maze, a standard test of spatial learning (286). Mice of other common strains, such as 129/Sv, perform poorly in such tests (Table 1). If a mutation in a gene of interest is studied in the context of the 129/Sv strain, and compared with wild-type mice of the C57BL/6 strain, it will seem as though the mutation attenuates learning, when in fact it is the overall genetic background rather than the specific mutation that impairs the behavior. It is, therefore, critical to use siblings of the same sex and same litter when making conclusions about the behavioral effects of a particular mutation. Even so, if genes linked (close by in the genome) to the ablated gene influence the observed phenotype, sibling pairs could show artifactual differences in behavior. In addition, many inbred mouse strains have congenital sight or hearing deficits that should be kept in mind when choosing the background strains for mutant mice (Table 1). For example, many albino strains have visual problems, and DBA/2 mice have developmental hearing impairments (49).

The profound effects of strain in a transgenic study aimed at modeling AD were recently reported. The transgene consisted of the hamster prion promoter driving expression of the amyloid precursor protein (APP) containing mutations linked to the human disorder. When C57BL/6 mice were used to generate the transgenic line, mice resulted that recapitulated many of the age-dependent hallmarks of human AD, including the appearance of amyloid-containing senile plaques with corresponding behavioral consequences (115, 190). C57BL/6 mice have been used widely for aging studies and are one of the few mouse strains that exhibit plaquelike formation (127); however, these do not contain amyloid deposits and do not correlate with neurodegeneration or deficits in learning. When the FVB strain was used to express the identical transgene, no senile plaques were observed, but the mice did exhibit a fear of novel environments (114). The differences between strains probably reflects the participation of additional genes other than APP in the formation of plaques.

Some recommendations regarding possible strategies that can be used to overcome the variability induced by strain differences have recently been proposed that may be useful in developing a breeding strategy (256). Briefly, this involves crossing mice carrying a particular mutation or transgene for several generations to mice from a defined inbred strain that shows the phenotypic characteristics relevant for the planned experiments, and using the animals only once the mutation is on a more homogeneous genetic background. If the mutation is crossed into two separate inbred backgrounds for several generations, and individuals carrying the mutation from each background are then crossed to each other, the advantages of genetic homogeneity can be achieved as well as the advantage of strong performance on behavioral tests that is usually seen in hybrids between two inbred strains. In this case, the resulting animals would be heterozygous at most genetic loci but would still be identical to each other.

	V. EXISTING PARADIGMS THAT HAVE BEEN USED TO STUDY TRANSGENIC ANIMALS

Top Previous Next References

Once a mutant mouse has been generated, there are any number of methods that can be used to test the neurobiological phenotype of the animals. It is often clear from previous biochemical or pharmacological work that complex functions might be altered in a particular mouse. In some cases however, no endogenous function has been identified for a gene product, or initial experiments have not yielded an obvious phenotype. In that case, it is of interest to consider using a battery of different physiological or behavioral tests to study the animals. An unexpected phenotype can lead to a new proposed function for a protein. In addition, many mouse models that are initially described as having no phenotype, based on results from the most obvious tests, are impaired in less expected functions. In this section we discuss some of the ways in which mouse models have been examined. This section is not designed to be comprehensive, but rather to give some examples of what has been done to analyze mutant mice, with the understanding that new paradigms are being developed constantly.

View larger version (41K): [in this window] [in a new window]   FIG. 6.   Role of synapsin I and synaptotagmin I in nerve terminal. A prototypical nerve terminal is shown. Releasable and nonreleasable pools of synaptic vesicles are segregated by association of nonreleasable pool with actin cytoskeleton through synapsin I. A vesicle in process of releasing its contents into synaptic cleft in response to a Ca2+ signal is also shown.

One of the great strengths of genetically altered mouse models is that one can determine how a defined protein contributes to behavior in a living animal (Table 3). Changes in behavior are rarely all or none, however, making it necessary to test large numbers of animals to obtain statistically significant results. It is also impossible to make firm conclusions from these experiments without being careful to control for differences in behavior due to background strain, developmental compensation, and age- or sex-related effects (see sect. IV). In general, behavioral tests should be carried out on same-sex pairs of wild-type and mutant littermates born to parents heterozygous for the mutation of interest. Ideally, mutant mice will be back-crossed for five or more generations to a mouse strain that has already been characterized in the behavioral test of interest, but this is very time consuming and the mutation could be lethal on the inbred background. Because of these difficulties, initial characterization is usually carried out on littermates from a mixed genetic background. Because most inbred strains show deficits in performance of some complex behavioral tasks, strain selection for back-crossing can be critical.

A recent review has proposed a battery of tests that can be used to determine the neurological health of a mouse and is a useful resource for those planning to test the behavior of a new mutant mouse (50). Before behavioral testing, it is important to determine whether the animals have any sensory or motor deficits that might affect performance in complex tasks. Sensory deficits in hearing or sight can be analyzed in either very simple or more technological ways. For example, a simple test to examine hearing is to determine whether an animal startles to a loud noise like a clap. It is also possible to examine startle quantitatively using varying frequencies of sound (see, for example, Ref. 139). A simple test of vision is to observe whether an animal stops at the edge of a table or whether it simply tumbles off the edge. Performance on the visible platform test of the Morris water maze can also indicate visual acuity (see sect. VB2). Even if a deficit is found, it is possible to design tasks that deemphasize the deficiency. For example, one could test blind mutant mice in learning tasks that require pairing of a tone with a stimulus, rather than in a task that required visual cues.