The pipid frog Xenopus laevis has been among the most productive model systems for vertebrate experimental embryology, due to the large size, external development, robustness, and ease of molecular and surgical manipulation of its embryos.  One reason Xenopus embryos are superb for embryological manipulations is that each cell contains its own power source- a yolk supply- enancing its ability to withstand the rigors of transplantation or to continue to differentiate in an explant or as an individual cells in simple salt solutions.
 
    Recent contributions from this system to axis determination, embryonic inductions, morphogenesis, and signal transduction studies have important implications for research on human disease 3, 4, 5,6.  In particular, evolving molecular techniques have helped identify a wide variety of activities through the addition of gene products to X. laevis assays, typically through mRNA injection into embryos or incubation of explants with protein preparations 7, 8, 9, 10, 11. However, in order to determine whether a newly-identified activity is actually required for a given process, it is neccessary to subtract a given gene function.  In complex developmental systems, tools provided by genetics- especially the phenotype of a null mutant- provide the strongest proof of participation.

    The history of one soluble factor illustrates the strengths and weaknesses of the frog system. Smith and co-workers 7,12 were able to characterize the dramatic mesoderm-inducing activity of activin by adding protein fractions of a tissue culture supernatant to a Xenopus ectoderm explant assay.  However, a number of other activities gradually emerged as competing candidate mesoderm inducers13, 14,15. Strategies to identify participants in complex processes by deleting specific activities have been tested in Xenopus, by application of antisense polynucleotides, blocking antibodies, or dominant negative mRNA constructs.  Unfortunately, antisense and antibodies often fail to completely eliminate gene function, and not all gene products may be amenable to construction of dominant negative variants19.  In other cases, dominant negative constructs have lead to misleading results due to interactions between the introduced gene product and parallel or intersecting signaling pathways. To continue with the activin example, injection of mRNA encoding a dominant negative version of the activin type II receptor resulted in a spectacular, but misleading neuralized phenotype20, which probably results from crosstalk with related Bone Morphogenetic Protein Receptor signaling pathways21. Subsequent genetic analysis of homologous recombinant mouse mutants revealed that mesoderm induction was unimpeded by the absence of zygotic activin22, suggesting that the requirement for activin in this process might be very limited. Genetic approaches offer the most direct way of examining a process in the absence of specific gene functions, and also offer the possibility of identifying completely new genes via mutagenesis; availability of mutants permits epistasis experiments which can order the steps of biological pathways and provide important mechanistic insights23.

    Further impetus for developing frogs as a genetic model derives from a recent breakthrough which allows transgenic frogs to be produced cheaply, efficiently, and in large numbers (see Preliminary Results  1).  Transgenesis-related genetic approaches, combined with the frog's embryological advantages and the low cost of husbandry and animal housing, should permit the dissection of many basic developmental processes quickly and inexpensively relative to other vertebrate model systems.  The transgenic protocol underpins many of the experiments described in this proposal. However, Xenopus laevis has two significant drawbacks as a genetic system.  First, its generation time is usually 1-2 years, reducing the practicality of many types of multi-generation experiments (e.g. making a line of frogs bearing reporter or marker constructs for transplantation experiments or breeding loci to homozygosity). Second, its evolutionary history has resulted in allotetraploidy, such that many genes are represented by extra copies which may or may not be functional. This greatly complicates the prospects for creating mutants (since it may be neccessary to inactivate four copies of a gene) and for analysing gene regulation (since it may be difficult to distinguish genomic clones of pseudogenes from those of active loci).
 
    A new model system which is better suited to these studies is Silurana (Xenopus) tropicalis.  X. tropicalis is a close relative of X. laevis24 (viable hybrids between the two species have been reported25), and shares virtually all of X. laevis' advantages as an embryological system. In addition, it features a much shorter generation time (3-4 months), and a smaller diploid genome (twenty chromosomes, with about 1.7 x 109 base pairs, versus thirty-six chromosomes and 3.1 x 109 bp for X. laevis)26.  Adult X. tropicalis, at 4-5 cm, are considerably smaller than the 10cm X. laevis, and consequently can be housed more efficiently; eggs are also somewhat smaller (0.6-0.7 mm vs 1-1.3 for X. laevis), but still amenable to manipulation, and are more abundant (1000-3000 per spawning versus 300-1000)27.  The genome size and number of chromosomes compare favorably to those of mouse (forty chromosomes, 3 x 109 bp)28 and zebrafish (fifty chromosomes, 1.6 x 109 bp29.
 
    Assays which have been described for X. laevis are readily applied to X. tropicalis (see Preliminary Results), but have the prospect of being supported by multigeneration genetic analyses.  X. tropicalis' diploid genome will facilitate uncovering recessive phenotypes, and pseudogenes are far less likely to complicate promoter analysis than in the tetraploid X. laevis.  Screens for new embryological mutants may be possible in X. tropicalis that would be dauntingly challenging and expensive in mice; early events in frog embryogenesis are better understood and more accessible to experimental manipulation than in fish.  Using mutants or transgenic animals in highly-developed tissue transplantation regimes facilitates mosaic analysis, in which individual animals contain tissues of more than one genotype; similar analyses have been strikingly effective in Drosophila.
 
    By way of comparison, two other vertebrate model systems are currently the subjects of extensive genetic analysis: the mouse Mus musculus and the zebrafish Danio rerio.  Zebrafish have been the subject of recent screens for developmental mutations that have at least come within calculating distance of saturation30, 31.  These screens generated a wide variety of useful embryonic phenotypes30, 31; however, molecular characterization of the associated genes is dependent on time-consuming positional cloning.  Mutagenesis with pseudotyped retroviruses greatly simplifies cloning, but suffers from considerably lower efficiency than chemical mutagenesis32.
 
    Fish embryology also presents some implacable problems for analysis of early development. Cell mixing is extensive at early stages, leading to a blurred fate map33, 34, and precluding spatially targeted introduction of gene products as well as analysis of early steps in axis formation.  The remarkable transparency of the zebrafish embryo, while convenient for some types of microscopic observation, actually interferes with high resolution analysis of the cell behaviors driving early morphogenesis; nearly all we know about these processes has been done in Xenopus [Shih, 1992 #786; Shih, 1992 #1770] 35, 36; manipulation of transparent explants and transplants is challenging.
 
    Explant assays are also complicated by the specialized mechanisms of early morphogenesis of teleost fish. The teleost undergoes meroblastic (partial) cleavage to form a blastoderm at the animal end of a very large, delicate, difficult-to-manipulate yolk filled cell37, 38. This large cell provides much of the motive force for early morphogenesis39, 40, 41; such a mechanism is atypical among vertebrates, especially mammals, which show many common features with frog development42, 43, 44.  Zebrafish may also be a poor model for some important developmental processes. For example, neural tube closure occurs in teleosts by a mechanism which may be unique among vertebrates.  Unlike the chick, the mouse, and the frog, which close the neural tube by folding of a neural plate and fusion of neural folds42, fish form the neural tube from a "keel" of tissue, generated as cells from both sides ingress at the midline52. Zebrafish are thus unlikely to provide significant insight into certain genetically-derived and environmentally-sensitive defects of neural tube closure in the human population, such as spina bifida53.
 
    Mice are uniquely valuable for their "reverse" genetics: homologous recombination mutageneses in embryonic stem (ES) cells, not currently available in any other higher eukaryote, have provided spectacular insights into mammalian development54. Mice do suffer from several limitations as a developmental genetic system, however.  First, early mouse embryos are small and inaccessible; embryological assays require considerable expertise and some categories of transplantation and explantation assays remain infeasible28. Second, gene targeting in mice is expensive: current costs are estimated to be approximately $20,000 per knockout in a well-established facilility and considerably higher (approximately $200,000) when setting up for the first time (based on costs calculated by the University of Virginia Transgenic facility). Third, mosaic analysis is technically challenging in mice28; in frog embryos, mosaicism can be readily spatially and temporally controlled by transplantation. Fourth, numbers of embryos per litter, space and budgetary constraints, as well as the inconvenience of internal development, preclude conventional "forward" genetic screens for developmental mutants.
 
    While duplication of ES cell technologies in frogs remains unlikely, making mutant frogs via homologous recombination is not inconceivable.  Nuclear transfer into enucleated eggs has been somewhat permissive with respect to the differentiation state of the donor cell; fully keratinized skin cells have been used to produce viable frogs55.  The remaining technical requirements for making targeted frog mutants would be a euploid, transfectable, selectable cell line from which nuclei could be harvested for transfer; the remarkable ability of ES cells to recapitulate early development on their own does not seem absolutely neccessary.
 
    Recently technical advances ranging from genetic mapping56, 57 to gene trapping58, expression cloning9, and frog transgenesis1 invite novel combinations in an amphibian genetic system.  Some of the projects we discuss, such as transgenic promoter analysis and generating strains of frogs expressing tissue-specific reporter genes, have excellent prospects for success.  Others, such as the proposed gene targeting experiments, highlight the risks involved in promoting a new model.  We recognize that some of the specific challenges may turn out to be insurmountable.  However, the investment of resources required to explore the feasibility of gene targeting, for instance, is relatively small with respect to the enormous potential return from a success.  The ability to analyze specific mutations in an inexpensive, well-characterized, externally-developing embryological system could not only lead to watershed advances in our understanding of vertebrate development, but could greatly reduce the animal care costs hobbling genetic research in mice.
 
    Some potential disadvantages as a genetic system may be inherent in the amphibian embryo.  The same maternal resources that make its embryology robust may shield it from the effects of many early zygotic mutations; important processes, such as the early phases of mesoderm induction, occur before the onset of zygotic transcription and will require analysis of maternal effect mutations.  The frog egg, unlike that of zebrafish, is not optically transparent, though surface gastrulation movements are in some ways more easily monitored in non-transparent embryos and internal development can be visualized by reporter expression in vivo or by in situ hybridization). While X. tropicalis development seems to be highly reproducible (see Preliminary Results), X. laevis embryos are somewhat more variable than those of fish and mice. Finally, while a diverse panel of X. laevis mutations has been described, no X. tropicalis mutants are as yet available.
 
    We have addressed the challenges of a new model by assembling a team of investigators, collaborators, and consultants with broad expertise. Among the principal investigators, Enrique Amaya pioneered the transgenesis technique and can produce large numbers of transgenic embryos; Robert Grainger combines molecular approaches with sophisticated transplantation assays; Ray Keller is a leader in the field of the biomechanics of morphogenesis and a prodigous developer of explant assays; Richard Harland's lab maintains state-of-the-art molecular biology and library-making skills and has elaborated elegant mRNA expression screens; and Douglas DeSimone's group has expertise in creating and characterizing Xenopus cell lines. Members of these laboratories also have experience in homologous recombination and promoter analysis in transgenic mice as well as zebrafish, Drosophila, and C. elegans genetics.  Collaborators and consultants include Charles Kimmel, one of the central figures in the establishment of zebrafish as a genetic model system; Mary Mullins, a principal in the large-scale zebrafish mutagenesis conducted in Christiane Nusslein-Vollhard's laboratory in Tubingen, Germany; Anne Unger, a zebrafish geneticist who will be on-site in the Grainger laboratory; and Bob Tompkins and Dana Reinschmidt, leading Xenopus geneticists, who were responsible for developing genetic management tools, such as production and analysis of  homozygous diploid Xenopus and identification of developmental mutants. Besides supplying broad technical expertise, having such a team permits a rational division of labor and provides a critical mass for attacking this challenge (16 individuals in five frog labs are currently contributing to the project). Although the projects described here are broad in scope, it should be emphasized that a number of the projects described here are extensions of ongoing research programs in the participating labs and not new initiatives solely for this proposal.
 
    As a vertebrate model system, Silurana tropicalis  is likely to provide many of the advantages of mice for genomic manipulation at a fraction of the husbandry expense, and to combine those of zebrafish with the superior manipulability and detailed embryology of the early amphibian embryo.

Frog genetics: Xenopus tropicalis jumps into the future

A draft of the "Trends in Genetics" article by Enrique Amaya, Martin Offield, and Robert Grainger, July 1998, Vol. 14, No. 7