The effort to make mosaic analysis a household tool

Mosaicism occurs naturally in many species. For example, Morgan reported rare cases of gynandromorphy in Drosophila nearly 100 years ago (Morgan, 1914) and the random inactivation of one of the X chromosomes in mammals was demonstrated 50 years ago (Lyon, 1961). Mosaic animals can be generated by non-genetic means, by transplantation of cells or tissues between animals of differing genotypes. For example, Twitty and Schwind used transplantation between salamanders of different sizes to show that organ size is an intrinsic property (Twitty and Schwind, 1931). However, such techniques are laborious and limited in the range of biological questions that can be addressed.

The first intentional generation of genetic mosaics to study development is attributed to Sturtevant (Sturtevant, 1929), who used an unstable X chromosome in Drosophila to generate individuals comprising X/O and X/X cells. Although Sturtevant wrote that analysis of his data could “give considerable information as to the cell lineage of Drosophila” (Sturtevant, 1929), the data were not fully analyzed until he provided them several decades later to Garcia-Bellido and Merriam (Garcia-Bellido and Merriam, 1969), who then generated a developmental fate map of the embryo. In the following 80 years, mosaic analysis, which provides a way to mark a cell early in development and then trace the fate of that cell and its progeny, has been utilized to address questions ranging from developmental biology to behavior, particularly in Drosophila. Landmark studies carried out by Garcia-Bellido led to the idea of cellular compartments in Drosophila imaginal discs (reviewed by Garcia-Bellido et al., 1979). Hotta and Benzer utilized genetic mosaics to determine the tissue focus of particular behaviors in the fly (Hotta and Benzer, 1972). Perhaps the most frequent application of mosaic analysis has been in determining the cell autonomy of gene action. Morgan and Bridges (Morgan and Bridges, 1919), using gynandromorphs, showed that sex-linked genes are usually ‘autonomous’; that is, each body part develops according to its genetic composition. Sturtevant, through his studies of the vermillion gene, was the first to use genetic mosaics to demonstrate the ‘non-autonomy’ of gene function (Sturtevant, 1920). Mosaic analysis has been particularly important as a method to predict the direction of signal transduction between cells during development (reviewed by Rubin, 1989; Heitzler and Simpson, 1991).

The discovery by Curt Stern (Stern, 1936) of somatic crossing-over between homologous chromosomes provided a reliable method for generating mosaic tissues in Drosophila. A cell heterozygous for a mutation (+/–) normally produces two identical heterozygous daughter cells after mitosis. However, if a crossover occurs between the two homologous chromosomes during mitosis, a heterozygous somatic cell could produce a homozygous mutant cell (–/–) and a twin homozygous wild-type cell (+/+), resulting in a mosaic animal carrying cells with three distinct genotypes (+/–, –/– and +/+). Such mitotic recombination occurs much less frequently than meiotic recombination, but can be induced to a rate of ∼1% with ionizing radiation. Although this technique was successfully used in many important studies, several technical difficulties limited its application. First, the site of the mitotic recombination event is not controlled. Second, the frequency of mitotic recombination induced by radiation is low. Third, the high radiation dose used to induce recombination causes extensive cell and tissue damage. Lastly, to identify the clones of mutant cells, the mutation of interest had to be closely linked with a marker gene that produced a cell-autonomous, easily scored phenotype in the cells of interest. Suitable marker genes, which needed to be located near the mutation of interest on the same chromosome arm, were often limiting.

Two key advances allowed us to overcome the limitations of the ionizing radiation protocol and produce the highly efficient system for mosaic analysis reported in our 1993 paper (Xu and Rubin, 1993): (1) the ability to engineer stable insertions of DNA constructs in the fly genome using P-element transposon (Rubin and Spradling, 1982) and (2) the demonstration that a site-specific recombination system from yeast, comprising FLP recombinase (FLPase) and its targets (FLP recombination targets, or FRTs), can function in the Drosophila genome (Golic and Lindquist, 1989) and catalyze mitotic recombination between FRTs located on homologous chromosomes (Golic, 1991).

In 1990 (when Tian Xu joined Gerry Rubin’s laboratory in Berkeley as a postdoctoral fellow) we embarked on a project with the aim of developing a widely applicable methodology that would allow facile mosaic analysis for every gene in the Drosophila genome (Fig. 1).

Since only the portion of the chromosome arm distal to the mitotic recombination site becomes homozygous, it was important to have an FRT site located close to the centromere on each of the major chromosome arms. To achieve this, we first established a large collection of strains each containing a different randomly inserted FRT-containing P-element. We used a construct carrying multiple tandem copies of the original long FRT sequence as we (correctly) anticipated that this would increase the frequency of FLPase-mediated mitotic recombination. With the help of Todd Laverty and Wan Yu, the sites of insertion were mapped by in situ hybridization to polytene chromosomes to identify those FRT-containing P-elements inserted near centromeres. Inserted elements that caused lethality or other phenotypes were rejected. Finally, proximally located insertions on each chromosome arm were tested for their ability to support mitotic recombination at high frequency. In the end, we were able to identify a suitable FRT line for each of the major chromosome arms. Together, this set of lines enabled the generation of mosaics for more than 95% of the genes in the Drosophila genome.

The design of the system provided many technical advantages over radiation-induced mitotic recombination. First, mitotic recombination only occurs at the FRT site, thus excluding the possibility of segregation of the mutation and the marker used to identify the cell clone (even when the two were not closely linked). Second, the markers used to identify the cell clones could be introduced as transgenic constructs. By placing the yellow⁺ and white⁺ transgenes onto each of the FRT chromosome arms, mosaic clones of any mutation in the genome could be marked with the visible yellow or white marker. A mini-white⁺ transgene was also placed onto these arms so that the mutant (–/–) and wild-type (+/+) twin-spot clones could each be identified in the heterozygous background (+/–); a clone of mutant cells in the eye would appear unpigmented, whereas the wild-type twin-spot clone would be a darker shade of red than the surrounding heterozygous tissue. The ability to identify wild-type clones provides an internal control for studying mutations that either result in growth advantage or cause cell death. Most traditional cell markers could, however, only be scored in terminally differentiated cells. Introducing epitope-tagged markers in transgenic constructs allowed non-terminally differentiated cells to be identified in mosaic clones, a capability crucial to the study of genes involved in developmental decisions. In addition, the drug-resistant gene neo^r was engineered into the P-element constructs to genetically label the FRT sites and hence facilitate strain construction. Third, the expression of FLPase has little or no damaging effects on cells and tissues carrying these FRT chromosomes. Moreover, FLPase expression can be controlled so that clones of mutant cells can be generated at specific developmental stages (via heat shock) or in a given tissue or cell type. Finally, the frequency of mitotic recombination of these FRT chromosomes is so high that almost every animal contains mosaic clones. This not only facilitates mosaic analysis of external tissues, but also made mosaic analysis of internal tissues or developing tissues possible, as identification of animals containing internal mutant clones cannot be accomplished by simple visual inspection without dissection. More importantly, the exceptionally high frequency of generating mosaic animals led to new genetic applications, in particular allowing genetic screens to be conducted in mosaic animals in which essential genes can be identified that are required for the development and functions of adult tissues. The desired phenotypes in this type of mosaic screen can be ascertained in the F1 individuals, which is much more efficient than traditional genetic screens involving three generations of breeding.

There have been continuous modifications and improvements of the system for mosaic analysis since our 1993 publication that provide additional sophistication in the generation and analysis of mutant clones. One limitation with our original strains was that markers for internal tissues could only be visualized after antibody staining. The introduction of GFP-based markers in transgenic constructs allowed live cells to be identified in mosaic clones of developing tissue (e.g. Martin et al., 2003), a crucial capability in the study of genes involved in developmental decisions. Combination of the FLP/FRT technology with the UAS/GAL4 system allows for the control and visualization of gene expression in mosaic clones or for killing off wild-type cells such that only mutant cells contribute to the final organ or tissue of interest (Brand and Perrimon, 1993; Chou and Perrimon, 1996; Stowers and Schwarz, 1999; Pagliarini and Xu, 2003; Yu et al., 2009). Methods for positively marking clones allow detailed analysis of single-cell clones, greatly facilitating mosaic studies in complex neural tissues (Lee and Luo, 1999).

The application of the FRT/FLP system in genome-wide mosaic analyses and screens has had a significant impact in a variety of fields in Drosophila research resulting in milestone discoveries. Such studies have helped elucidate many signaling pathways and their functions; for example, the regulation of growth and tissue size by the Hippo/Lats and PTEN/TSC signaling pathways (Xu et al., 1995; Huang et al., 1999; Potter et al., 2001; Potter et al., 2002; Harvey et al., 2003; Huang et al., 2005). The system has also been instrumental in the study of cancer biology in flies, and other biological processes such as cell competition, including interactions between tissues of differing genotypes (Pagliarini and Xu, 2003; Cova et al., 2004; Moreno and Basler, 2004; Li and Baker, 2007; Wu et al., 2010). It has led to important discoveries in defining stem cell niches (e.g. Xie and Spradling, 1998; Xie and Spradling, 2000). The pervasiveness of these Drosophila mosaic studies has also influenced the use of genetic mosaics in other model organisms. Indeed, similar strategies for inducible mitotic recombination in mice have been developed and successfully utilized (Zong et al., 2005; Muzumdar et al., 2007; Wang et al., 2007; Sun et al., 2008). We are pleased that the mosaic analysis system that we developed two decades ago and its evolving descendants continue to enable researchers to address many new questions regarding the genetic regulation of cellular behavior and function.

We thank Ulrike Heberlein, Tzumin Lee, Kevin Moses and Duc Nguyen for comments on the manuscript.