Genetic variation — the existence of at least two forms — is the essential ingredient present in all genetic experiments. Phenotypic variation, in particular, is used as a means for uncovering the normal function of a wild-type allele at many loci. As discussed in the first chapter of this book, it was the availability of many variant phenotypes within the fancy mouse trade that made the house mouse such an ideal organism for studies by early geneticists. In a sense though, the house mouse won by default because in the absence of domestication and artificial selection, variation in traits visible to the eye is extremely rare, and thus, other small mammals were genetically intractable. Although the fancy mouse variants provided material for a host of early genetic studies, the number of different variants was still limited, and the rate at which new ones arose spontaneously in experimental colonies was exceedingly low: it is now known that, on average, only one gamete in 100,000 is likely to carry a detectable mutation at a particular locus.

During the 1920s, several investigators began investigating the effects of X-rays on reproduction and development. In two laboratories, at least, new mutant alleles were recovered in the offspring of irradiated parents, but the investigators failed to make any connection between irradiation and the induction of these mutations (Little and Bagg, 1924; Dobrovolskaia-Zavadskaia, 1927). The connection was finally made by Muller who, in 1927, published his classic paper explaining the induction of heritable mutations by X-rays (Muller, 1927). Since that time, geneticists who study all of the major experimental organisms — from bacteria to mice — have used both ionizing irradiation and various chemicals as agents of mutagenesis to uncover novel alleles as tools for understanding gene function.

Large-scale mouse mutagenesis experiments were first begun at two government-based "atomic energy" laboratories: the Oak Ridge National Laboratory in Oak Ridge, Tennessee, in the U.S. and the MRC Radiobiological Research Unit first at Edinburgh, Scotland, and then at Harwell, England, in the U.K. Both of these experimental programs were begun initially after World War II as a means for quantifying the effects of various forms of radiation on mice and, by extrapolation, humans, to better understand the consequences of detonating nuclear weapons. The U.S. effort was directed by W. L. Russell and the British effort was directed by T. C. Carter (Green and Roderick, 1966). Scientists at both laboratories quickly realized the potential of their newly created resource of mutant animals, and both laboratories have since gone on to generate mutations by chemical agents as well. The very large-scale studies conducted at Oak Ridge and Harwell — where 10,000 to 60,000 first generation animals were typically analyzed in an experimental protocol — have provided most of the empirical data currently available on the mechanisms and rates at which mutations are caused by all well-characterized mutagenic agents in the mouse.

The experiments performed by Russell and Carter, and other colleagues who followed in their footsteps, were designed to obtain discrete values for the mutagenic potential of different radiation protocols. Rather than attempt to examine all animals for all effects of a particular irradiation protocol (as was common in earlier experiments), these mouse geneticists chose instead to look only at the small fraction of animals that were mutated at a small set of well-defined "specific" loci. The rationale for the "specific locus test" was that effects on individual loci could be more easily quantitated and that the limited results obtained could still be extrapolated for an estimate of whole genome effects. Russell decided that mutation rates should be followed simultaneously at a sufficient number of loci to distinguish and avoid problems that might be caused by locus-to-locus variations in sensitivity to particular mutagens. He decided further that the same set of loci should be examined in each experiment performed. The seven loci chosen to be followed in the specific locus test were defined by recessive mutations with visible homozygous phenotypes that were easily distinguished in isolation from each other, and had no effect on viability or fertility. The seven loci are agouti (a is the recessive non-agouti allele), brown (b), albino (c), dilute (d), short-ear (se), pink-eyed dilution (p), and piebald (s). A special "marker strain" was constructed that was homozygous for all seven loci.

In its simplest form, the specific locus test is carried out by mating females from the special marker strain to completely wild-type males that have been previously exposed to a potential mutagen. In the absence of any mutations, offspring from this cross will not express any of the seven phenotypes visible in the marker strain mother. However, if the mutagen has induced a mutation at one of the specific loci, the associated mutant phenotype will be uncovered. This test is very efficient because it only requires a single generation of breeding and visual examination is all that is required to score each animal.

Although recessive mutations at all loci other than the specific seven will go undetected in the first generation offspring from this cross, it is possible to detect a dominant mutation at any locus so long as it is viable and produces a gross alteration in heterozygous phenotype such as a skeletal or coat color change. One should realize that the most common effect of any undirected mutagen will be to "knock-out" a gene and, in the vast majority of cases, the resulting null allele will be recessive to the wild-type. There is, however, a very small class of loci at which null alleles will act in a dominant or semidominant fashion to wild-type. These "haplo-insufficient" phenotypes are presumably caused by a developmental sensitivity to gene product dosage. Among the best characterized of the dominant-null mutations are the numerous ones uncovered at the T locus — which result in a short tail — and the W locus — which result in white spotting on the coat.

Mutations can be induced by both physical and chemical means. The physical means is through the exposure of the whole animal to ionizing radiation of one of three classes — X-rays, gamma rays, or neutrons (Green and Roderick, 1966). The chemical means is to inject a mutagenic reagent into the animal such that it passes directly into the gonads and into differentiating germ cells. The specific locus test has provided an estimate of the relative efficiency with which each reagent induces mutations. Under different protocols of exposure, X-irradiation was found to induce mutations at a rate of 13—50 x 10-5 per locus, which is a 20 to 100 fold increase over the spontaneous frequency, but still not high enough to be used by any but the largest facilities as a routine means for creating mutations (Rinchik, 1991). The mutations created by irradiation are often large deletions or other gross lesions such as translocations or complex rearrangements.

The class of known chemical agents that can induce mutations (known as mutagens) is very large and expanding all the time. However, two chemicals in particular — ethylnitrosourea (ENU) and chlorambucil (CHL) — have been found to be extremely mutagenic in mouse spermatogenic cells (Russell et al., 1979; Russell et al., 1989). Both of these chemicals produce much higher yields of mutations than any form of radiation treatment tested to date (Russell et al., 1989). Optimal doses of either ENU or CHL can induce mutations at an average per locus frequency which is greater than one in a thousand — 150 x 10-5 with ENU and 127 x 10-5 per locus with CHL (Russell et al., 1982; Russell et al., 1989). Although the rates at which ENU and CHL induce mutations are very similar, the types of mutations that are induced are quite different. In general, ENU causes discrete lesions which are often point mutations (Popp et al., 1983), whereas CHL causes large lesions which are often multi-locus deletions (Rinchik et al., 1990a). Originally, it was thought that the basis for mutational differences of this type was the chemical nature of the mutagen itself (Green and Roderick, 1966), but, this no longer appears to be the case. Rather, it now appears that the germ cell stage in which the mutation arises is the major determinant of the lesion type (Russell, 1990). The correlation observed between chemical and lesion type is a result of the fact that different mutagens are active at different stages of spermatogenesis. Thus, ENU acts upon pre-meiotic spermatogonia where mutations are likely to be of the discrete type, and CHL acts upon post-meiotic round spermatids where mutations are likely to be of the large lesion type (Russell et al., 1990).

ENU was the first chemical to be identified that was sufficiently mutagenic to be used by smaller laboratories in screens for mutations at particular loci or chromosomal regions of interest (Bode, 1984; Shedlovsky et al., 1988). ENU has also been used in screens for non-locus-specific phenotypic variants that could serve as models for various human diseases (McDonald et al., 1990). Several laboratories are beginning to use ENU for saturation mutagenesis of small chromosomal regions defined by deletions as one tool (among several complementary ones) for obtaining a complete physical and genetic description of such a region (Shedlovsky et al., 1988; Rinchik et al., 1990b; Rinchik, 1991). The rationale for these studies is the belief that many (although not all) genes can be uncovered phenotypically by knock-out mutations which are bred to be doubly heterozygous with a deletion. The major limitation to the global use of this approach is the very small number of genomic regions in the mouse at which large deletions have been characterized.

The availability to Drosophila geneticists of deletions (or deficiencies as the fly people call them) that span nearly every segment of the fly genome has played a critical role in the identification and characterization of large numbers of genes and the production of both gross functional maps and fine-structure point mutation maps by the very approach just described above. Clearly, a method to accumulate a similar library of deletions for the mouse would be well-received. The mutations induced by X-rays are often large-scale genomic alterations including translocations, inversions, and deletions. Indeed, most mouse deletion mutations maintained in contemporary stocks were derived in this manner. However, the overall yield of X-ray-induced deletions is quite low, and because of other problems inherent in this approach, it is not ideal for global use.

In 1989, chlorambucil was reported to be an attractive alternative to X-rays as an agent for the high-yield induction of deletion mutations in the mouse (Russell et al., 1989). The per locus mutation rate was found to be on the order of one in 700 in germ cells of the early spermatid class, and of the eight mutations induced at this stage that were analyzed, all were deleted for DNA sequences around the specific locus marker (Rinchik et al., 1990a). This study also showed that CHL-induced mutations were often associated with reciprocal translocations. This last finding is unfortunate because translocations can reduce fertility with consequent negative effects on strain propagation.

There is hope that CHL can be used as a means for generating sets of overlapping deletions that span entire chromosomes (Rinchik and Russell, 1990). Projects of this type will require very large animal facilities and support resources and will consequently be confined to only a handful of labs. However, once mouse strains with deletions have been created and characterized, they can serve as a resource for the entire community.

An advantage to using the mouse as a genetic system is the strong sense of community that envelops most of the workers in the field, and it is in the context of this community that strains carrying many different mutations — both spontaneous and mutagen-induced — have been catalogued and preserved and are made available to all investigators. A catalog containing detailed descriptions of all mouse mutations characterized as of 1989 has been compiled by Margaret Green and is included as the centerpiece chapter in the Genetic Variants and Strains of the Laboratory Mouse edited by Mary Lyon and Tony Searle (Green, 1989). This catalog is now available in an electronic form that is updated regularly (see Appendix B). Of course, many more mutant animals are found and characterized with the passing of each year, and an updated list is published annually in the journal Mouse Genome. This list contains information on the individual investigators that one should contact to actually obtain the mutant mice.

The largest collection of mutant mouse strains is maintained at the Jackson Laboratory under the auspices of the "Mouse Mutant Resource" (MMR) which is currently maintained under the direction of Dr. Muriel Davisson (Davisson, 1990). In 1990, over 250 mutant genes were maintained in this resource, accounting for two-thirds of all known mouse mutants alive at the time (Davisson, 1990). Each year, animal caretakers identify an additional 75 to 80 "deviant" animals among the two million mice that are produced by the Jackson Laboratory’s Animal Resources colonies (Davisson, 1993). Approximately 75% of the deviant phenotypes are found to have a genetic basis and breeding studies are conducted on these to determine whether or not they represent mutations at previously characterized loci. If they do, DNA samples are recovered and the lines are discarded or placed into the Frozen Embryo Repository. If a mutation is novel, its mode of transmission (autosomal/X-linked, dominant/recessive) is determined, the phenotypic effect of the mutation is characterized, and it is mapped to a specific chromosomal location with the use of breeding protocols to be described in section 9.4 (Davisson, 1990; Davisson, 1993). Descriptions of all newly characterized mutations are publicized, and mutant mouse strains are made available for purchase through the standard Jackson Laboratory catalog. In 1992, over 35,000 mice from the MMR were distributed to investigators throughout the world (Davisson, 1993).

Space limitations make it impossible for the MMR to maintain breeding stocks of mice that contain every known mutant gene, with the total number expanding each year. Fortunately, mutant stocks that are not currently in demand by investigators can be maintained (at minimal cost) in the form of frozen embryos. The importance of embryo freezing as a storage protocol cannot be over-emphasized. Time and time again, modern-day molecular researchers have reached back to use mutations described long-ago as critical tools in the analysis of newly cloned human and mouse loci.

The basic technology required to obtain preimplantation embryos from the female reproductive tract, to culture them for short periods of time in petri dishes, and then to place them back into foster mothers where they can grow and develop into viable mice has been available since the 1950s (Hogan et al., 1994). Over the ensuing years, this basic technology has been used in a host of different types of experiments aimed at manipulating the process of development or the embryonic genome itself. Embryos can be dissolved into individual cells that can be recombined in new combinations to initiate the development of chimeric mice. Pronuclei and nuclei can be switched from one early embryo to another to examine the relative contributions of the cytoplasm and the genome to particular phenotypes, as well as to investigate aspects of genomic imprinting and parthenogenesis. Foreign DNA can be injected directly into pronuclei for stable integration into chromosomes which can lead to the formation of transgenic animals. Finally, embryonic cells can be converted into tissue culture cells (called embryonic stem [ES] cells) where targeted gene replacement can be accomplished. Selected ES cells can be combined with normal embryos to form chimeric animals that can pass the targeted locus through their germline. These experimental possibilities are discussed more fully later in this chapter. This section is concerned simply with genetic considerations involved in the choice of mice to be used for the generation and gestation of embryos for various experimental purposes.

A number of factors will play a role in the selection of an appropriate strain of females who will contribute the eggs to be used as experimental material. First, in all cases, it is important that the eggs are hardy enough to resist damage from the manipulations that they will undergo. Second, the particular experimental protocol may impose a need for eggs that have special genetically-determined qualities. Third, in those case where very large numbers of eggs are required, it will be important that the strain is one that responds well to superovulation, as discussed in the next section. Finally, there is a question of genetic restrictions on the offspring that will emerge from the manipulation.

As concerns this last criterion, for some experiments it will be important to maintain strict control over the genetic background of embryos to be used for genomic manipulation. In these cases, inbred embryos should be derived from matings between two members of the same inbred strain. If these embryos are used for germline introduction of foreign genetic material, the resulting transgenic animals will be truly coisogenic to the original inbred strain.

For other experiments, strict genetic homogeneity will not be required. In these cases, it is possible to use F2 embryos from superovulated F1 females who have been mated to F1 males of the same autosomal genotype. This breeding protocol is often preferable to the use of either a strictly inbred approach or a random-bred approach. First, in contrast to the random-bred approach, one still maintains a certain degree of control over the genetic input since only alleles derived from one or both of the inbred strains used to generate the F1 parents will be present at any locus in each embryo. Second, in contrast to the inbred approach, the use of both females and embryos with heterozygous genotypes allows the expression of hybrid vigor at all levels of the reproductive process. In particular, heterozygous embryos are less likely to be injured by in vitro manipulations.

One inbred strain that has been developed relatively recently from a non-inbred colony of mice with a long history of laboratory breeding at NIH has special characteristics of particular interest to investigators interested in producing transgenic mice: this strain is called FVB/N. The FVB/N strain is unique in several important ways (Taketo et al., 1991). First, its average litter size of 9.5 (with a range up to 13) is significantly higher than that found with any other well-known inbred strain (see Table 4.1). Second, fertilized eggs derived from FVB/N mothers have very large and visually prominent pronuclei; this characteristic is unique among the known inbred strains and greatly facilitates the injection of DNA. Finally, the fraction of injected embryos that survive into live born animals is also much greater than that observed with all other inbred strains. For these reasons, FVB/N has quickly become the strain of choice for use in the production of transgenic animals.

Although one can recover on the order of 6 to 10 eggs directly from individual naturally-mated inbred or F1 females, it is possible to obtain much larger numbers — up to 60 eggs per animal — by inducing a state of superovulation. For many experiments, it is important to begin with a large number of embryos; with superovulation, one can drastically reduce the number of females required to produce this large number. Superovulation is induced by administering two precisely-timed intraperitoneal injections of commercially-available gonadotropin reagents which mimic natural mouse hormones and initiate the maturation of an aberrantly large number of egg follicles. Superovulation, like normal ovulation, causes both a stimulation of male interest in mating as well as female receptivity to interested males. The protocol is described in detail in the mouse embryology manual by Hogan and colleagues (1994).

Not unexpectedly, the average number of eggs induced by superovulation is highly strain-dependent. Appropriately aged females of the strains B6, BALB/cByJ, 129/SvJ, CBA/CaJ, SJL/J and C58/J can be induced to ovulate 40 to 60 eggs (Hogan et al., 1994). At the other extreme, females of the strains A/J, C3H/HeJ, BALB/cJ, 129/J 129/ReJ, DBA/2J, and C57L/J do not respond well to the superovulation protocol, producing only 15 or fewer eggs per mouse. The response of the FVB/N strain to superovulation is in-between with the production of 25 embryos or fewer per female (Taketo et al., 1991). For generating transgenic mice, however, this single negative feature of FVB/N is outweighed by the other characteristics of this strain discussed above.

An interesting aspect of the high versus low response to superovulation is that in two cases, substrains derived from the same original inbred strain (BALB/cByJ versus BALB/cJ and 129/SvJ versus 129/J) express such clearly distinct phenotypes. This finding suggests that subtle changes in genotype can have dramatic consequences on the expression of this particular reproductive trait.

One critical finding of both practical and theoretical importance is that F1 hybrid females do not always express a better response to superovulation then both of their inbred parents. For example, the commonly used F1 hybrid B6D2F1, which is formed by a cross between a high ovulator (B6) and a low ovulator (DBA/2J), expresses the low ovulator phenotype (Hogan et al., 1994). This observation goes against the grain of hybrid vigor and it suggests that the genetic basis for this phenotype may be much more specific and limited than it is for other general viability and fertility phenotypes. In addition, this observation suggests that for the major genes involved, the "high ovulatory" alleles are recessive.

Two F1 hybrids have been determined empirically to express a high level superovulation — [BALB/cByJ x B6] and [B6 x CBA/CaJ] (Hogan et al., 1994). It is also very likely that F1 hybrids derived from matings between any of the high responders listed above will themselves be high responders as well. Many of these F1 hybrids can be purchased directly from animal suppliers; however, in most cases, suppliers cannot provide an exact day of birth which is necessary to determine the optimal time of use.

Females that have undergone ovulation — either naturally or induced — must be mated with a "fertile stud male" to produce zygotes that can be used for nuclear injection or other purposes. As discussed earlier, it is always preferable to use a fertile stud male with the same genotype as the female, whether it is inbred or an F1 hybrid. Obviously, it is important to use visibly healthy animals in the prime of their life, between 2 and 8 months of age. In addition, past experience is often a good indicator of future performance. Males that have mated successfully on demand in the past (as indicated by a vaginal plug) are likely to do the same in the future; for this reason, records should be maintained on the performance of each male used for this purpose. For optimal results, one should place only one male and one female in each cage, and after a successful mating, the male should be given a rest of two to three days.

Once embryos have been manipulated in culture, they must be placed back into the reproductive tract of a foster mother where they can continue their development into fully-formed live-born animals. Since the foster mother contributes only a womb, and not genomic material, to the engineered offspring, her genetic constituency should not be chosen according to the same criteria used for animals in most other experimental protocols. Only two considerations are important in the choice of a foster mother. First, and most important, she should have optimal reproductive fitness and "mothering" characteristics. This can be accomplished with either an F1 hybrid between two standard inbred strains [B6 x CBA is recommended (Hogan et al., 1994), but others will do as well] or with outbred strains available from various commercial breeders. The best foster mothers are those that have already borne and raised at least one litter successfully; thus, it is often useful to try out potential candidates by putting them through one pregnancy/mothering/weaning cycle.

A second consideration is whether the investigator will be able to distinguish natural-born pups from those that have been fostered. This is only a factor when the foster mother has been mated to a sterile male in order to induce the required state of pseudopregnancy, and there is some question as to whether the male has been properly sterilized. The simplest method for distinguishing the two types of potential offspring is by a coat color difference; for example, albino versus pigmented. If the experimental embryos are derived from non-albino-allele-carrying parents, then both the foster mother as well as her sterile stud partner (discussed below) can be chosen from commercially available outbred albino strains such as CD-1 (Charles River Breeding Laboratories) or Swiss Webster mice (from Taconic Farms). When one is certain that the sterile stud male is really sterile, coat color differences are less critical, so long as well-defined DNA differences exist if the unexpected need does arise to distinguish the genotypes of potential natural-born offspring from experimentally transferred offspring.

In human females, the uterine environment becomes receptive to the implantation of fertilized eggs as a direct consequence of the hormonal induction of ovulation. In mice and most other non-primate mammals, the uterine environment becomes receptive to implantation only in response to a sufficient degree of sexual stimulation. In addition, this stimulation also causes hormonal changes which alter the normal estrus cycle under the assumption that a pregnancy will ensue. When a successful stimulatory response has occurred in the absence of fertilization, the female is said to be in a state of "pseudopregnancy." Only pseudopregnant females will allow the successful implantation and development of fostered embryos. Pseudopregnancy can be achieved in one of two ways: (1) by mating to a sterile male or (2) through the use of female masturbation tools such as vibrating rods inserted into the vagina (West et al., 1977). Most investigators have found that natural matings produce a higher percentage of pseudopregnancies than human surrogates.

Sterile males can be derived genetically or surgically. Genetic derivation requires a breeding colony of mice that are doubly heterozygous for pseudo-allelic mutations on chromosome 17 in a region known as the t complex (Silver, 1985). Animals with the genotype T/tw2 (available from the Jackson Laboratory) are intercrossed for the purposes of both maintaining the strain as well as for the production of sterile males as diagrammed in figure 6.1.

For those without the resources or personnel required to breed genetically sterile males, the only other choice is surgical vasectomy, which involves the severing of the vas deferens on both sides of the body (Hogan et al., 1994). The choice of mouse strain to use is based on criteria analogous to those set out for the choice of a foster mother, except that mating ability should be considered in place of mothering ability. The standard F1 hybrids as well as "random-bred" animals can all be used with success. When it comes to choosing between individual males within a particular strain, one should use the same criteria described in section 6.2.2.2 for the choice of fertile stud males. In addition, pre-mating "sterile" males with fertile females serves to confirm the success of the vasectomy.

There are two problems inherent in all methods of classical mutagenesis. The first problem is that the process is entirely random. Thus, one must start out by designing a screening assay to allow the detection of mutations at the locus of interest, and then one must hope for the appearance of mutant animals at a frequency which is at, or above, the usual per-locus rate. If a mutant allele fails to produce a phenotype that can be picked up by the screen, it will go undetected. Finally, even when mutant alleles are detected, the underlying lesion can usually not be ascertained without cloning and further molecular characterization.

The second problem with classical mutagenesis is that induced mutations are not tagged in any way to provide a molecular entry into a locus that has not yet been cloned. Thus, if a novel locus is uncovered by an induced mutation that causes an interesting phenotype, it can only be approached through candidate gene and positional cloning approaches in the same way as any other phenotypically-defined locus. Furthermore, in the case of ENU-induced mutations, the mutant and wild-type alleles are likely to be molecularly indistinguishable with the exception of a single nucleotide that may or may not affect a restriction site.

One can imagine two types of mutagenic approaches that would be most ideal for the two different types of situations in which mutations can provide tools for molecular analysis of development and other aspects of mammalian biology. On the one hand, a random mutagenesis approach is fine for the elucidation of novel loci so long as the mutant allele is tagged to allow direct molecular access. On the other hand, to further analyze a locus which is already cloned and characterized, one would like to generate animals that miss-express the locus in some defined manner. The technologies of transgene insertion and gene targeting have provided geneticists with the tools needed to accomplish both of these goals.

In 1981, five independent laboratories reported the insertion of foreign DNA into the mouse germ line through the microinjection of one-cell eggs (Costantini and Lacy, 1981; Gordon and Ruddle, 1981; Harbers et al., 1981; Wagner et al., 1981a; Wagner et al., 1981b). Although the incorporation of exogenous DNA into the germ line through viral infection of embryos had been reported earlier (Jaenisch, 1976), the 1981 reports implied for the first time that DNA from any source could be used to transform the mouse genome. The complexion of mouse genetics was changed forever with the development of this powerful tool. A strictly observational science was suddenly thrust into the realm of genetic engineering with all of its vast implications. The insertion of genetic material into the mouse germ line has now become sufficiently routine that the methodology is detailed in various "cookbooks" (Wassarman and DePamphilis, 1993; Hogan et al., 1994) and designer animals are even provided as a commercial service by a number of companies.

The term transgenic has been coined to describe animals that have foreign sequences inserted stably into their genome through human intermediaries. Transgenic animals can be created by microinjection or viral infection of embryos, or through the manipulation in culture of embryonic-like "ES cells" that are subsequently incorporated back into the embryo proper for shepherding into the germ line. The latter technology will be discussed in a following section. Here I will focus on transgenic animals created by direct injection of DNA into embryos.

The initial animal that develops from each micro-manipulated egg is called a founder. Even when multiple embryos have all been injected or infected with the same foreign DNA, the integration site — or transgene locus — in each founder will be different. However, all transgenic animals that descend from a single founder will share the same transgene locus. Protocols for the creation of transgenic mice, and extensive reviews of the technology and its uses have been described elsewhere (Palmiter and Brinster, 1986; Wassarman and DePamphilis, 1993; Hogan et al., 1994). Rules for naming transgene loci and transgenic animals are presented in section 3.3.5.

With current protocols for the creation of transgenic mice by embryo microinjection, the site of integration is not pre-determined, and, for all practical purposes, should be considered random. Microinjection allows one to add, but not subtract genetic material in a directed manner; if a particular experiment leads to the insertion of an novel version of a mouse gene into the genome, this novel allele will be present in addition to the normal diploid pair. Consequently, only dominant, or co-dominant, forms of phenotypic expression will be detectable from the transgene.

The embryo microinjection technology can be used to explore many different aspects of mouse biology and gene regulation. One class of experiments encompasses those aimed at determining the effects of expressing a natural gene product in an unnatural manner. By combining the gene of interest with regulatory regions chosen from other genes, one can cause transgenic mice to express the product at a higher than normal level, or in alternative tissues or developmental stages. The mutant phenotypes that result from such aberrant forms of expression can be used to elucidate the normal function of the wild-type gene. Experiments of this type can be used, for example, to demonstrate the capacity of some genes to induce specific developmental changes and the oncogenic nature of others when they’re aberrantly expressed. Many other types of questions can be answered with this approach.

In another class of experiments, one can dissect out the function of a regulatory region by forming constructs between it and a reporter gene whose expression can be easily assayed in the appropriate tissue(s). With a series of transgenic lines that have partially-deleted or mutated forms of a regulatory region, one can pinpoint which DNA sequences are involved in the turning-on and turning-off of genes in different tissues or developmental stages.

A third of class of experiments is aimed at correcting a genetic defect in a mutant mouse through the genomic insertion of a wild-type transgene. This use of the transgenic technology provides the most powerful means available to prove that a cloned candidate gene is indeed identical to the locus responsible for a particular mutant phenotype. Furthermore, the correction of genetic defects in model mammals is a necessary prelude to any attempt to perform similar studies in humans.

An important consideration in all transgenic experiments follows from the observation that the actual chromosomal location at which a transgene inserts can play a determining role in its expression. This will be readily apparent in cases where different founder lines with the same transgene show different patterns of transgene expression. The reason for such strain-specific differences is that some chromosomal regions are normally maintained in chromatin configurations that can act to suppress gene activity. Different transgene constructs will show different levels of sensitivity to suppression of activity when they land in such regions.

Another potential problem can result from the insertion of the transgene into a normally-functioning endogenous locus with unanticipated consequences. In approximately 5 to 10% of all cases studied to date, homozygosity for a particular transgene locus has been found to cause lethality or some other phenotypic anomaly (Palmiter and Brinster, 1986). These recessive phenotypes are most likely due to the disruption of some normal vital gene. In less frequent cases, a transgene may land at a site that is flanked by an endogenous enhancer which can stimulate gene activity at inappropriate stages or tissues. This can lead to the expression of dominant phenotypes that are not strictly a result of the transgene itself. For all of these reasons, it is critical to analyze data from three or more founder lines with the same transgene construct before reaching conclusions concerning the effect, or lack thereof, on the mouse phenotype.

In the vast majority of cases analyzed to date, the disruption of endogenous sequences caused by transgene integration has had no apparent effect on phenotype. However, the absence of a detectable phenotype does not necessarily mean that the transgene has integrated into a non-functional region of the genome. As discussed in chapter 5, only a small subset of all mammalian genes are actually vital, and subtle effects on phenotype are likely to go unnoticed if one performs only a cursory examination of transgenic animals. Thus, the actual frequency of insertional mutagenesis resulting from embryo microinjection is likely to be significantly higher than the numbers imply.

Unless a particular transgenic insertion causes an easily detectable, dominant phenotype, the presence of the transgene in an animal is most readily determined through DNA analysis. For testing large numbers of mice, the best source of DNA is from tail clippings or ear punch-outs (Gendron-Maguire and Gridley, 1993). The presence or absence of the transgene can be demonstrated most efficiently with a method of PCR analysis that is based on transgene-specific target sequences.

The founder animal for a transgenic line will be heterozygous for the transgene insertion locus. The second homolog will be associated with a non-disrupted "wild-type" (+) allele at this locus, whereas the disrupted chromosome will carry a transgene (Tg) allele. As long as the transgene is transmitted to offspring from a heterozygous (Tg/+) parent, it will be necessary to test each individual animal of each new generation for the presence of the transgene. For this reason alone, it would be useful to generate animals homozygous for the transgene allele since all offspring from matings between homozygous Tg/Tg animals would also be homozygous and there would be no need for DNA testing.

In rare cases, homozygous Tg/Tg animals will be phenotypically distinct from their Tg/+ cohorts. This observation is usually a good indication that the transgene has disrupted the function of an endogenous locus through the process of integration. If the homozygous recessive phenotype is lethal, it will obviously be impossible to generate a homozygous line of animals. Otherwise, the phenotype may eliminate the need for DNA analysis. In the vast majority of cases, however, homozygous Tg/Tg animals will be indistinguishable in phenotype from heterozygous Tg/+ animals, and without a recessive phenotype, the identification of homozygous animals will not be straightforward.

One approach to confirming the genotype of a presumptive Tg/Tg animal is based on statistical genetics. In this case, confirmation is accomplished by setting up a mating between the presumptive homozygote and a non-transgenic +/+ partner. If the animal in question is only a Tg/+ heterozygote, one would expect equal numbers of Tg/+ and +/+ offspring. Through the method of chi square analysis described in section 9.1.3, one can calculate that if at least 13 offspring are born and all carry the transgene, the probability of a heterozygous genotype is less than one in a thousand. If even a single animal is obtained without the transgene, the parent’s genotype will almost certainly be Tg/+. Statistical testing of this kind must be performed independently for each presumptive Tg/Tg animal. Once homozygous Tg/Tg males and females have been confirmed, they can be re-mated to each other as the founders for a homozygous transgenic strain.

A second approach to demonstrating transgene homozygosity requires the cloning of a endogenous sequence from the mouse genome that flanks the transgene insertion site. This task is often not straightforward because transgenic material can be present in multiple copies that are intermingled with locally rearranged endogenous sequences. Nevertheless, with the cloning of any nearby endogenous sequence, one obtains a mapping tool that, in theory, can be used to distinguish both the disrputed and nondisrupted alleles at the transgene locus through the use of one of the various techniques described in chapter 8 for detecting "codominant" DNA polymorphisms. With such a tool, and an associated assay, homozygosity for the transgene allele would be demonstrated by the absence of the wild-type nondisrupted allele. Unfortunately, this approach would require the generation of a separate endogenous clone for each and every transgenic line to be studied. Protocols for locating the transgene insertion site within the mouse linkage map are discussed in section 7.3.2.

Although the transgene insertion technology described in the previous section provides a powerful tool for the analysis of gene action in the whole organism, it has one serious limitation in that it does not provide a mechanism for the directed generation of recessive alleles. This limitation can be overcome with a technology known variously as gene targeting, targeted mutagenesis, or gene replacement — the subject of this section. This powerful technology allows investigators to generate directed mutations at any cloned locus. These new mutant alleles can be passed through the germ line to produce an unlimited number of mutant offspring, and different mutations can be combined with variants at other loci to study gene interactions.

This ultimate tool of genetic engineering was born through the combination of several technologies that had developed independently over the preceding 10 to 20 years including embryonic stem cell culture and homologous recombination, with mouse embryo manipulation and chimera formation (Sedivy and Joyner, 1992 provide an excellent review of all aspects of this field). Two independent laboratories, headed by Oliver Smithies and Mario Cappechi, finally succeeded in bringing all of these various technologies together during the mid-1980s (Capecchi, 1989; Smithies, 1993).

One, although not the only, appeal of the gene targeting technology is the ability to create mouse models for particular human diseases (Smithies, 1993). But, in essence, gene targeting can provide investigators with powerful tools to study any cloned gene. While patterns of RNA and protein expression provide clues to the stages and tissues in which genes are active, it is only with mutations that a true understanding of function can be obtained (Chisaka and Capecchi, 1991).

After heaping such praises on gene targeting, it is important to forewarn potential users of this technology that its application is not problem-free. First, an investigator must achieve a high level of competence and experience with several distinct, technically demanding protocols; this requires a significant investment of time and energy. Second, there is the fickle nature of the technology itself as discussed below. Nevertheless, the handful of laboratories initially able to target genes successfully has expanded quickly with the training of new young investigators, and this expansion is likely to continue much further with the recent publication of several excellent volumes containing detailed chapters on experimental protocols (Joyner, 1993; Wassarman and DePamphilis, 1993; Hogan et al., 1994).

Once a particular gene has been cloned and characterized, the steps involved in obtaining a mouse with a null mutation in the corresponding locus can be outlined briefly as follows. First, one must design and construct an appropriate targeting vector in which the gene of interest has been disrupted with a positive selectable marker; in the most commonly used protocol, a negative selectable marker is also added at a position that flanks the gene sequence. The most commonly used positive selectable marker is the neomycin resistance (neo) gene, and the most commonly used negative selectable marker is the thymidine kinase (tk) gene.

The second step involves the introduction of the targeting vector into a culture of embryonic stem (ES) cells (usually derived from the 129 strain) followed by selection for those cells in which the internal positive selectable marker has become integrated into the genome without the flanking negative selectable marker. The third step involves screening for clones that have integrated the vector by homologous recombination rather than by the more common non-homologous recombination in random genomic sites. Once "targeted clones" have been identified, the fourth step involves the production of chimeric embryos through the injection of the mutated ES cells into the inner cavity of a blastocyst (usually of the B6 strain), and the placement of these chimeric embryos back into foster mothers who bring them to term. A recently developed alternative approach to chimera formation through the aggregation and spontaneous incorpation of ES cells into cleavage stage embryos has the advantage of not requiring sophisticated microinjection equipment (Wood et al., 1993).

The experiment is deemed a success if the ES cells successfully enter the germline of the chimeric animals as demonstrated by breeding. If the disrupted gene is indeed transmitted through the germline, the first generation of offspring from the chimeric founder will include heterozygous animals that can be intercrossed to produce a second generation with individuals homozygous for the mutated gene. The nomenclature rules that are used to name all newly created mutations are described in section 3.4.4.

A second generation of homologous recombination strategies have been developed to allow the placement of specific small mutations into a locus without the concomitant presence of disrupting intragenic selectable markers. The ability to create subtle changes in a gene could provide an investigator with the tools required to dissect apart the function of a gene product one amino acid at a time. A number of different approaches toward this goal have been described. The most promising of these, called "hit and run," is based on the generation of ES cell lines that have undergone homologous recombination with a targeting vector, followed by selection for an intrachromosomal recombination event that eliminates the selectable markers and leaves behind just the mutated form of the gene (Joyner et al., 1989; Hasty et al., 1991; Valancius and Smithies, 1991; Fiering et al., 1993).

Unfortunately, at the time of this writing, the hit and run protocols are still extremely demanding and with each experiment, an investigator will only obtain a single mutant allele at the locus of interest. An alternative strategy is to break the problem into two separate tasks: (1) knocking-out the gene completely in one strain by standard homologous recombination, and (2) the independent production of one or more transgenic lines that contain subtly-altered mutant versions of the gene. By breeding the knock-out line with one of the transgenic lines, it becomes possible to generate a new line of animals in which the original wild-type allele has been replaced (although not at the same site) with a specially-designed transgene allele. There are several advantages to this approach. First, the methodology required for simply knocking-out a gene is more straightforward and better developed at the time of this writing than the hit-and-run methodology. Second, gene targeting in ES cells requires much more time and effort than the production of transgenic mice by nuclear injection. Thus, when an investigator wishes to study a variety of alleles at a particular locus, it will be much easier to create a single line of mice by gene targeting and then breed it to different transgenic lines. The one potential disadvantage to this approach is that the transgene construct may not be regulated properly and accurate patterns of expression may not occur in the animal, even when the transgene is linked to its own promoter/enhancer.

Even when a laboratory has mastered all of the protocols required to perform gene targeting, the difference between success and failure can still be a matter of luck. Some DNA sites appear highly impervious to homologous recombination, whereas sites a few kilobases away may be much more open to integration. But even at the same site, the frequency of homologous versus non-homologous recombination events can vary by a factor of ten from one day to the next (Snouwaert et al., 1992).

At the time of this writing, many of the factors responsible for success remain unknown. However, one critical factor that has recently become evident is the need to use source DNA for the targeting construct that has been cloned from the same strain of mice used for the derivation of the ES cell line into which the construct will be placed (van Deursen and Wieringa, 1992). In other words, the highest levels of gene replacement are obtained when the incoming DNA is isogenic with the target DNA. Apparently, the homologous recombination process is very sensitive to the infrequent nucleotide polymorphisms that are likely to distinguish different inbred strains from each other. In most cases today, ES cell lines have been derived from the 129/SvJ mouse strain (but see the subsection 6.4.5 below), and thus it is usually wise to build DNA constructs with clones obtained from 129 genomic libraries.

The original 129/SvJ mouse, and the one still available from the Jackson Laboratory, has an off-white coat color caused by homozygosity for the pink-eye dilution (p) mutation, and forced heterozygosity for the chinchila (cch) and albino (c) alleles at the linked albino locus (cch p/c p). In contrast, the "129 mouse" that serves as the source of most ES cell lines used for homologous recombination has a wild-type agouti coat color. What is the basis for this difference?

The answer is a historical one that centers on the work of Leroy Stevens, a cancer geneticist, now retired, who worked at the Jackson Laboratory. Stevens had observed that the 129/SvJ strain was unique in the occurrence of spontaneous testicular teratomas at an unusually high rate of 3 to 5% in male animals. As means to better understand the genetic parameters responsible for tumor incidence, Stevens set out to determine whether any of a variety of well-characterized single locus mutations that affect either tumor incidence or germ cell differentiation would interact with the 129 genome in a manner to increase or decrease the natural frequency of tumor formation in this strain. One of the mutations that he tested was Steel (Sl), which plays an important role in the differentiation of germ cells as well as melanocytes and hematopoetic cells. The Sl mutation expresses a dominant visible phenotype — the lightening of the normal wild-type black agouti coat color so that it has a "steely" appearance, and a reduction of pigment in the distal half of the tail. Unfortunately, it is impossible to see this phenotypic alteration on the cch p/c p coat of 129/SvJ mice which already have a nearly complete loss of pigment production. Thus, to follow the backcrossing of Sl onto 129/SvJ, it was necessary to replace the mutant alleles at the c and p loci with wild-type alleles. It is the triple congenic 129/Sv-Sl/+ , +c +p line produced by Stevens that acted as the founder for all "129 mice" that have been used in ES cell work.

As indicated earlier in this chapter, one side product of many transgenic experiments is the generation of mice in which a transgene insertion has disrupted an endogenous gene with a consequent effect on phenotype. Unlike spontaneous or mutagen-induced mutations, "insertional mutations" of this type are directly amenable to molecular analysis because the disrupted locus is tagged with the transgene construct. Unexpected insertional mutations have provided instant molecular handles not only for interesting new loci but for classical loci, as well, that had not been cloned previously (Meisler, 1992).

When insertional mutagenesis, rather than the analysis of a particular transgene construct, is the goal of an experiment, one can use alternative experimental protocols that are geared directly toward gene disruption. The main strategies currently in use are based on the introduction into ES cells of beta-galactosidase reporter constructs that either lack a promoter (Gossler et al., 1989; Friedrich and Soriano, 1991) or are disrupted by an intron (Kramer and Erickson, 1981). The constructs can be introduced by DNA transfection or within the context of a retrovirus (Robertson, 1991). It is only when a construct integrates into a gene undergoing transcriptional activity that functional beta-galactosidase is produced, and producing cells can be easily recognized by a color assay. Of course, the production of "beta-gal" will usually mean that the normal product of the disrupted gene can not be made and thus, this protocol provides a means for the direct isolation of ES cells with tagged mutations in genes that function in embryonic cells. Mutant cells can be incorporated into chimeric embryos for the ultimate production of homozygous mutant animals that will display the phenotype caused by the absence of the disrupted locus. This entire technology, referred to as "gene trapping" (Joyner et al., 1992), is clearly superior to traditional methods for the production of mutations at novel loci that use chemical mutagens or irradiation.

A computerized database (called TBASE) has been developed to help investigators catalog the strains that they produce and find potentially useful strains produced by others (Woychik et al., 1993). The database is available over the Internet through the Johns Hopkins Computational Biology Gopher Server and is linked to the on-line mouse databases maintained by the Jackson Laboratory Informatics Group.

Although the gene replacement technology has been employed with success by more and more laboratories, it is still the case, and likely to remain so, that an enormous amount of time and effort goes into the production of each newly engineered mouse strain. Clearly, it does not make sense to derive strains with the same gene replacement more than once. However, with the high costs of animal care and maintenance, it is often difficult for researchers to maintain strains that they are no longer actively using. Furthermore, many individual research colonies are contaminated by various viruses, and as such, virus-free facilities are reluctant to import mice from anywhere other than reputable dealers. The Jackson Laboratory has recently come to the rescue by setting up a clearing house to preserve what are likely to be the most useful of these strains for other members of the worldwide research community. For the first time, JAX will be importing mice from large numbers of individual researchers. Each strain will be re-derived by cesarean section into a germ-free barrier facility, and will be made available for a nominal cost, without experimental restriction, to all members of the research community.

With the various technologies that have now been developed to manipulate the genomes of embryonic cells combined with ever-more sophisticated molecular tools, it can be stated without exaggeration that the sky is the limit for what can be accomplished with the mouse as a model genetic system. It is always impossible to predict what the future holds, but one can imagine the use of both gene addition and gene replacement technologies as routine tools for assessing the functions of sequential segments of DNA obtained by walking along each mouse and human chromosome. With recent reports of success in the insertion of intact YAC-sized DNA molecules of 250 kb or more in length into the germline of transgenic animals, it becomes feasible to analyze even larger chunks of DNA for the presence of interesting genetic elements (Jakobovits et al., 1993; Schedl et al., 1993). In fact, it is only with experiments of this type that it will be possible to completely uncover all of the pathways through which a gene is regulated, and all of the pathways through which a gene product may function. Just as the study of neurons in isolation can not possibly provide a clue to human consciousness, the study of individual genes outside of the whole animal can not possibly provide a clue to the network of interactions required for the growth and development of a whole mouse or person.