TIE-DYE: a combinatorial marking system to visualize and genetically manipulate clones during development in Drosophila melanogaster

During the development of an organism, the fate of cells is regulated by both lineage and extracellular signals. Thus, to understand the development of a tissue or an organ there is a need to describe the origin of cells as well as interactions between cells. Additionally, there are modes of regulation affecting organ size determination and tissue homeostasis that can only be appreciated by studying the tissue as a whole rather than by examining the properties of individual cells. Studying these phenomena would be greatly aided by methods that enable tracking the lineage of many cells within an organ and at the same time examining the consequences of local perturbations on the tissue as a whole.

Experiments to track cell lineages during development have been performed in many different ways (reviewed by Kretzschmar and Watt, 2012), including the study of gynandromorphs (Sturtevant, 1929; Garcia-Bellido and Merriam, 1969), by using chimeric embryos (Le Douarin and Barq, 1969), and by injecting dyes or enzymes into individual cells in order to label their progeny (Weisblat et al., 1978). More recently, genetic systems that stochastically label cells and all of their progeny have been used to study cell lineages. In addition, by combining systems that direct site-specific recombination with genes encoding multiple fluorescent proteins, it has become possible to label multiple lineages uniquely in a tissue simultaneously as was achieved with the Brainbow system (Livet et al., 2007) in the mouse nervous system. This sparked the creation of multicolor labeling systems in other model organisms, which have been used to map multiple single cell projections in the nervous system (Livet et al., 2007; Hadjieconomou et al., 2011; Hampel et al., 2011; Boulina et al., 2013) to track developmental lineages in the zebrafish heart (Gupta and Poss, 2012) and to visualize stem cell-based replacement during homeostasis of the mouse gut (Snippert et al., 2010). The use of these systems demonstrates that being able to simultaneously and uniquely label many cells or follow multiple cell lineages at once can lead to novel scientific discoveries.

The study of Drosophila development has led to many important findings about the genetic regulation of pattern formation. A key experimental approach has been to express a variety of genes with the GAL4/UAS binary system (Fischer et al., 1988; Brand and Perrimon, 1993). Both dBrainbow and Flybow (Hadjieconomou et al., 2011; Hampel et al., 2011) use the GAL4/UAS system for multicolor-labeling of cells and cell lineages in Drosophila. This is achieved by stochastic site-specific recombination to place one of several genes encoding fluorescent proteins downstream of a GAL4-responsive promoter in a restricted subset of the marked clones. In addition, these systems only permanently mark cell lineages if the source of GAL4 is ubiquitously expressed (reviewed by del Valle Rodríguez et al., 2012).

Several methods have been developed to manipulate either the genotype or gene expression within clones of cells. In Drosophila, this can be achieved efficiently either by FLP/FRT-mediated mitotic recombination that efficiently generates patches of mutant cells and wild-type sister clones (Golic, 1991; Xu and Rubin, 1993), or by the FLP-out method where FLP/FRT-mediated excision of a STOP cassette upstream of GAL4, results in heritable GAL4 expression (Pignoni and Zipursky, 1997). Systems using mitotic recombination have been further refined by positively marking the GAL4-expressing clones or by marking both sister clones following the recombination (Lee et al., 2000; Griffin et al., 2009; Yu et al., 2009). However, in all of these cases, most of the remaining cells in the tissue are not labeled and hence it is not easy to infer the consequence of generating a patch of mutant cells on the remainder of the organ.

Recently developed heterologous systems, such as the Q-system (Potter et al., 2010) and the lex-GAD system (Szüts and Bienz, 2000; Lai and Lee, 2006; Yagi et al., 2010), could potentially be used together with GAL4/UAS-based methods of multicolor cell labeling, such as dBrainbow and Flybow, so as to combine lineage tracing with gene manipulations in subsets of cells. However, most current collections of transgenic lines that express RNAi constructs or cDNAs are UAS driven and thus their expression cannot be directed by these newer heterologous systems. Here, we describe a way to combine simultaneous multicolor lineage analysis with overexpression or knockdown of UAS-regulated constructs in a subset of the marked clones. We have used this system, first as a method of lineage tracing, to assess the number of founder cells in the wing-imaginal discs under conditions of normal development and compensatory proliferation following X-ray irradiation. We then demonstrate its utility in examining the effects of manipulating genes that regulate cell proliferation or patterning and its applicability to study phenomena that can involve multiple clones of cells, such as the sorting of cells from epithelia.

All Drosophila stocks, unless otherwise mentioned, were obtained from the Bloomington Stock Center. The TIE-DYE marking system was generated by the recombination of the following transgenes: Act <stop <lacZ-NLS (Struhl and Basler, 1993) and Ubi <stop <eGFP-NLS (Evans et al., 2009) onto the same 2nd chromosome; and Act <stop <GAL4 (Pignoni and Zipursky, 1997) and UAS-His2A::mRFP (Emery et al., 2005) onto the same 3rd chromosome. A stock containing these recombinant chromosomes was built, along with a stock with hsFLP to facilitate testing UAS-driven transgenes. Although this stock with hsFLP can be maintained successfully for multiple generations, at room temperature, germ-line FLP-out clones are sometimes observed where all cells in the progeny are labeled with a given marker. The stock is therefore re-assembled using the following stocks: y, w, hsFLP; Sp; Dr / SM5-TM6B and w; Act <stop <lacZ^nls, Ubi <stop <GFP^nls; Act <stop <GAL4, UAS-his2A::RFP / SM5-TM6B.

Experimental animals were generated by the following crosses: (1) y, w, hsFLP × w; ubi <stop <GFP^nls, act5C <stop <lacZ^nls; Act5C <stop <GAL4, UAS-his2A::RFP / SM5-TM6B; or for experiments with an additional UAS-transgene, (2) y, w, hsFLP; ubi <stop <GFP^nls, act5C <stop <lacZ^nls; Act5C <stop <GAL4, UAS-his2A::RFP / SM5-TM6B × UAS-X. The following transgenes were used for overexpression experiments: UAS-ci.3m.HA (Price and Kalderon, 1999), UAS-dpp.S 42B.4 (Staehling-Hampton and Hoffmann, 1994), UAS-Ras85D.V12 (Lee et al., 1996) and UAS-yki.S168A.V5 (Oh and Irvine, 2009). The following transgenes were used for RNAi knockdown experiments and included UAS-DCR2: ci^RNAi=TRiP.JF07715, ex^RNAi=TRiP.JF03120, Myc^RNAi (VDRC), y^RNAi (VDRC) and yki^RNAi=TRIPJF03119.

Larvae were dissected in PBS and fixed in 4% paraformaldehyde for 15 minutes at room temperature. Primary antibodies used were either rabbit anti-β-galactosidase (1:500; MP Biomedicals) or mouse anti-β-galactosidase (1:500; Promega), with ALEXA Fluor-633 (anti-rabbit or anti-mouse; 1:1000; Invitrogen) as secondary antibodies. Tissues were mounted in SlowFade Gold Antifade (Life Technologies). Fluorescent images were taken on a Leica TCS SL confocal microscope with Ar (488 λ), GreNe (543 λ) and HeNe (633 λ) lasers using 16× and 40× objectives. No additional stain was performed to detect the GFP or RFP in the tissues. Images of GAL4(+) and GAL4(-) clones were generated using Adobe Photoshop CS3; the selection tool was used to select areas above a cut-off threshold in the red channel to generate images of both RFP(+)-only clones and non-RFP(+) clones (by subtraction).

Clones were induced by heat shocks in a 37°C circulating water bath, in either vials or on grape-juice plates (placed in a secondary container). For experiments conducted on grape-juice plates, L1 larvae were picked ∼24 hours after egg lay (AEL) and placed in a vial with yeast paste. The durations of the heat-shocks were experimentally determined for different developmental stages. A heat-shock of 30 minutes was used for embryos and 15 minutes was used for early/late-stage larvae to generate a high clone frequency with diverse clone types. Twenty-four-hour egg collections (in vials) were conducted for experiments in Figs 4, 5, 6, clones were induced during 2nd/3rd instar and discs were dissected as late 3rd instar larvae.

Eggs were collected on grape-juice plates with yeast paste for 2 hours at 25°C, after a 30-minute pre-collection. At 16±1 hours AEL, embryos were heat-shocked for 30 minutes and then exposed to X-rays generated from a Faxitron TRX5200 operating at 125 V and 3.0 mA. The irradiated samples were placed at a distance of 40.3 cm from the X-ray source on a micro-go-round and weight block, producing an exposure rate of ∼3.2 Gy/minute. Third instar larvae were dissected at 96-120 hours AEL. The animals that received the highest dose of X-ray irradiation showed the most delay in development, as previously described (Hussey et al., 1927; Halme et al., 2010).

Wing discs were fragmented and transplanted into an adult female host, as described previously (Bosch et al., 2005). Following the induction of TIE-DYE clones, the mid-3rd instar larvae were sterilized (1 minute in 50% bleach) and dissected. Fragments of the ventral wing disc were removed with tungsten needles, and the resulting three-quarter disc fragments were injected into the abdomen of Oregon-R adult females that were kept at 25°C. Imaginal discs were recovered from hosts after 4 days and stained.

Tracing of the merged images was performed using Adobe Photoshop CS3 Smart Highlighting Tool to follow boundaries of clones. The different channels were visualized independently to determine the number of different markers present in a clone and therefore the appropriate color. The clone area tracings were used to measure the clone area with the Adobe Photoshop CS3 Analysis Record Measurement tool and the particular marker combinations present in the tissue. Statistical analysis of the clone area data was performed using GraphPad Prism (version 4). For comparing GAL4(+) with GAL4(-) clone areas for expanded^RNAi (supplementary material Fig. S4), we used the Mann-Whitney, two-tailed statistical test. For comparing the proportion of GAL4(+) cells between multiple genotypes, we used a one-way ANOVA test with Dunnett’s multiple comparison test between the control and other genotypes.

The probabilities of all possible situations from 1 to 40 founder cells were calculated using a python script and then the probabilities were grouped in one of two ways: (1) the total number of colors represented (one through eight) or (2) specific color outcomes (such as all ways to generate a tissue with only green, yellow and teal marked cells present). From an experimental outcome, we calculated the probability that the disc was generated from 1 to 20 founder cells (i.e. a disc with only red, yellow, purple and white cells was calculated to have the highest probability of being from five founder cells; shown in Fig. 2N).

To visualize the clonal composition of a tissue as well as interactions between mutant cells and wild-type cells throughout the tissue, we developed a system we have named TIE-DYE that generates seven distinct populations of irreversibly marked clones in an unlabeled background. In addition, four out of the seven classes of clones are GAL4 dependent (see below). This allows co-expression of additional transgenes, to manipulate gene function, in a subset of the marked clones, while adjacent clones remain wild type.

The TIE-DYE system takes advantage of the stochastic nature of FLP-induced recombination to excise transcriptional stop-cassettes upstream of lacZ (Struhl and Basler, 1993), GFP (Evans et al., 2009) and/or GAL4 (Pignoni and Zipursky, 1997), such that, following recombination, expression is sustained in that cell and all of its descendants (Fig. 1A). As a result, cells can potentially activate the expression of any one, any combination of two, or all three FLP-out genes, thus generating seven possible combinations, along with an unmarked population (Fig. 1B). Because, unlike the classical dye-injection experiments, the marking of cells is achieved using three independent excisions, we have named the system TIE-DYE. Two of the FLP-out constructs drive marker expression directly and are therefore GAL4 independent. The third FLP-out construct activates the expression of GAL4, which promotes expression of the third marker, UAS-his2A::RFP. Thus, TIE-DYE allows concurrent GAL4-driven overexpression or RNAi knockdown in the RFP-expressing subpopulation. A heat-shock promoter controls the expression of FLP and therefore the duration of a heat-shock affects both the number and type of labeled cells. For simplicity, we will refer to the differently marked clones by color names as displayed in the images. For example, a clone marked only by the expression of β-galactosidase (β-gal) will be called blue; a clone marked by both GFP and RFP will be called yellow. A short heat-shock is likely to induce relatively few recombination events and mostly results in cells marked by a single-marker (green, red, blue). By contrast, a longer heat shock favors many more recombination events, thereby generating doubly (yellow, teal, purple) and triply marked cells (white) (Fig. 1C; supplementary material Fig. S1A-F). Intermediate heat-shock times can produce tissues with all seven classes of cells represented, as well as ‘black’ unmarked cells (Fig. 1D-H).

As labeling can occur in cells that are not actively undergoing mitosis, recombination can be induced even under periods of cellular quiescence, as occurs during periods of embryogenesis, or following terminal differentiation. In addition to imaginal discs, we have used the TIE-DYE method in a variety of other tissues, including the larval brain and the adult ovary (supplementary material Fig. S1G,H).

Most tissues, including Drosophila imaginal discs, are generated by clonal expansions of relatively small numbers of founder cells. During early embryonic development, multiple cells from two different parasegments are specified to become the wing imaginal disc (reviewed by Serrano and O’Farrell, 1997), and are restricted in fate to either the anterior or posterior compartment (Garcia-Bellido et al., 1973). The wing-imaginal disc primordia invaginate and mostly separate from the rest of the embryonic epithelium between 9 and 11 hours AEL (Bate and Arias, 1991). Soon afterwards, the cells arrest in G₁ for the remainder of embryogenesis (Knoblich et al., 1994). Each compartment of the wing disc is therefore composed of multiple clones of cells and has been described as a ‘polyclone’ (Crick and Lawrence, 1975). Except for a relatively smooth boundary that separates anterior cells from posterior cells, the clonal boundaries within each compartment are irregular and do not obviously correspond to any anatomical landmark. We used the TIE-DYE system to study the wing-imaginal disc under normal and compensatory growth conditions to address two questions: (1) how many founder cells in a late stage embryo normally contribute to the wing-imaginal disc?; and (2) from this invaginated primordium, what is the smallest number of cells that can generate a wing-imaginal disc?

Previous studies estimate that 38-50 cells give rise to the wing-imaginal disc. These estimates were made by counting the cells in the late-stage embryo (Madhavan and Schneiderman, 1977) or by taking the reciprocal of the fraction of the area of the adult wing that is covered by a single large clone (Garcia-Bellido and Merriam, 1971). We wanted to take advantage of having seven uniquely labeled, plus one unlabeled, cell populations to estimate the number of founder cells present in the late stage embryo. Simply counting the number of labeled patches (marked + unmarked) in the wing disc, would underestimate of the number of founder cells, because some of those patches could be composed of more than one clone. If two adjacent clones were to express the same combination of markers, they would form a single contiguous patch. To prevent clones of the same color from touching, one could generate very rare clones by using a short heat-shock; however, this would generate very few labeled clones and leave large portions of the disc unlabeled.

We have developed an alternative method for estimating the number of founder cells that is based on the probability of having each class of clones represented in the disc. As each transgene has a specific probability of being activated by a given heat shock, the probability of all eight classes (each of the seven colors and black) being represented in any single disc increases as the number of founder cells increases. For example, if the wing disc derives from a single founder cell, then the whole disc would always express a single combination of markers. If a disc derives from three founder cells, then the maximum number of different colors that could be present would be three (Fig. 2A). If the number of founder cells were extremely large, then all eight classes would be found in most, if not all, discs (for a heat shock of the appropriate duration).

We generated a simple mathematical model to determine the likely number of founder cells, in a late-stage embryo, that gives rise to the wing imaginal discs, under normal and stressed (irradiated) conditions. An estimated recombination frequency of each of the three constructs (GFP, GAL4, lacZ) was determined by measuring the fraction of area in 3rd instar discs that have flipped a given construct (supplementary material Fig. S2A). The frequencies of clones with multiple FLP-out markers was consistent with each recombination event occurring independently of the others (supplementary material Fig. S2B), suggesting that multiply marked clones are not predisposed to death because of chromosomal rearrangements. For a heat-shock of 30 minutes at 16±1 hours AEL, the average recombination frequencies were calculated to be Act < <GAL4=0.50, Ubi < <GFP=0.49 and Act < <lacZ=0.28, which were used to calculate the probability of each color (see Materials and methods). From these frequencies of each color, the model estimates the likelihood of each possible color combination for 1-40 founder cells. As the founder cell number increases, as expected, the likelihood of generating a disc with seven or eight colors increases, and the likelihood of generating a disc with only three or four colors significantly decreases (Fig. 2A,B). Similar curves are generated even when recombination frequencies are a standard deviation above or below the average (supplementary material Fig. S2C). The background (without heat shock) recombination frequency in the wing imaginal disc was low; we observed single clones in ∼7% of wing discs (n=82). A higher frequency (∼30%) was observed for the myoblasts associated with the wing disc. We would not expect this ∼7% background frequency to have a large effect on our estimates as it would only impact our calculations if this additional recombination event generated a color that was not already present in the disc. However, it could increase the number of different colors of clones present slightly and result in a small overestimation of the number of founders. In addition, the background recombination frequency was not elevated by X-ray irradiation (one disc had a single clone, n=25).

TIE-DYE clones were induced in late-stage embryos at 16±1 hours AEL; for the X-ray irradiation experiments, the embryos were irradiated shortly after clone induction, which is expected to eliminate some of the precursor cells (Fig. 2C). The disc precursor cells that survive the X-ray irradiation would then generate the entire wing imaginal disc. Irradiation would be expected to generate a tissue with fewer and larger clones than normal.

We used our model to estimate the number of precursor cells at a late stage of embryogenesis that contribute to the disc proper (excluding the peripodial epithelium) of the 3rd-instar wing-imaginal disc. A wild-type (non-irradiated) wing imaginal disc is shown in a merged image (Fig. 2D) and a clone area tracing of the differently labeled clones present in the disc proper (Fig. 2E). All possible marked and unmarked clone types (eight in total) are present in this disc, as was seen in 61% of the discs (n=23). In addition, 26% discs contain seven classes and 13% discs contain six classes. Based on the relative probabilities of obtaining specific color combinations from different numbers of founder cells (1-40), we calculated the number of founder cells that would best explain the observed discs (Fig. 2F). The model predicts with >95% confidence interval that the disc is derived from 22-34 precursor cells. This prediction can be refined further with lower levels of confidence to 25-31 founders (∼73% confidence). This is lower than the estimate of 38 founder cells derived by counting the number of cells in the wing imaginal primordium of 1st instar larvae (Madhavan and Schneiderman, 1977). This may be because not all cells in the morphologically recognizable primordium contribute to the disc proper of the wing imaginal disc. Martín et al. (Martín et al., 2009) recently estimated that the disc proper region of the mature larval wing disc is composed of ∼30,000 cells. Based on measured doubling times at different stages of development, they calculated that cells undergo an average of 9.1 doublings, which results in an estimate of 55 founder cells. Our estimate of 22-34 founders would be predicted to generate 30,000 cells if there were, on average, slightly more than 10 cell divisions.

We next examined the effect of reducing the number of founder cells by X-ray irradiation on the number of different classes of discs (e.g. discs with four colors, five colors, etc.). First, we determined how sensitive the cells of the imaginal disc primordium (at 16±1 hours AEL) were to X-ray irradiation. As the dose of X-ray irradiation increased, the number of viable adults decreased and so did the presence of recognizable imaginal discs (supplementary material Fig. S3A,B). At an X-ray dose of 32 Gy, too few precursors cells survived to generate a wing disc. As embryos were subjected to increasing doses of X-ray irradiation, fewer discs of normal size and shape were observed and in these there was a shift towards classes of discs that had relatively few large clones and few classes of marker expression. Merged images and clone area tracing of the discs for increasing doses of X-ray irradiation: 12.8 Gy (Fig. 2G,H), 19.2 Gy (Fig. 2I,J) and 25.6 Gy (Fig. 2K,L). The number of colors present in the disc proper for each experimental condition is summarized in Fig. 2M. At a low dose of irradiation (12.8 Gy), the majority of larvae still had both wing imaginal discs, but there was limited adult viability (∼17%). At higher X-ray irradiation doses of 19.2 Gy and 25.6 Gy, ∼20-50% of wing discs were absent. Some of the recovered wing discs looked morphologically normal (supplementary material Fig. S3C,D), whereas others were of abnormal shape (supplementary material Fig. S3E,F). At 25.6 Gy of irradiation, the most frequent class of discs was one that only contained three different colors (Fig. 2M). The particular classes of colors observed in the discs recovered from the irradiated embryos were used to calculate the probability of the disc originating from 1-20 founder cells. Example probabilities are shown in Fig. 2N. For a disc that consists of only red, yellow, purple and white regions, the model predicts that there is a ∼25% chance that it originated from five founder cells and approximately equal chances that it originated from four or six founder cells (∼20%). In addition, the model predicts that there is a 95% probability that this disc originated from between four and nine cells. A disc that consists only of blue, purple, white and black has a 95% probability that it originated from four to seven cells. As there is expected to be considerable variation in the number of founder cells following irradiation, we did not attempt to fit all the data to a single number of founder cells for each dose of irradiation. If all the data were fitted to a single number of founder cells, finding a single disc with six colors would result in zero probability of generating a disc from four founder cells. Instead, for each disc we calculated the probability that it could arise from each number of founder cells in the range of 1-20 cells. We then present, the ‘average curve’ for each dose of irradiation (Fig. 2O). Thus, increasing doses of X-irradiation reduce the number of founder cells that can still generate a disc of normal size. Moreover, we have observed multiple instances of discs that are likely to have derived from as few as two to four founder cells, which is ∼1/8th-1/16th the number found in normal embryos. Assuming an exponential increase in cell number, these founder cells are predicted to have undergone three or four rounds of additional cell proliferation to generate a disc of normal size and shape.

Regeneration in Drosophila imaginal discs can be observed either by damage in situ (Smith-Bolton et al., 2009; Bergantiños et al., 2010) or by culturing fragmented discs in the abdomens of adult females (Hadorn et al., 1949; reviewed by Worley et al., 2012). Imaginal discs regenerate by forming a blastema, which is an area of localized proliferation. The Twin-Spot Generator (Griffin et al., 2009) was used in fragmented eye-antenna discs undergoing transdetermination to show that relatively few cells generated the transdetermined tissue (Sustar et al., 2011). We used TIE-DYE to visualize regenerative growth after physical damage to the imaginal disc (Fig. 3A). TIE-DYE clones were induced in 3rd instar larvae. By this stage of development, most cells accumulate in G2 and then undergo two additional rounds of mitosis during the pupal stage of development (Fain and Stevens, 1982; Graves and Schubiger, 1982; Milán et al., 1996; Neufeld et al., 1998). Thus, in undamaged larval imaginal discs recovered 24 hours after mid-3rd instar clone induction, individual cells express different combinations of the three markers and the disc overall has a ‘salt-and-pepper’ pattern of different colors (Fig. 3B,C). To trigger regenerative growth, discs lacking a region from the ventral part of the disc were implanted into the abdomens of adult female flies (see Materials and methods) (Fig. 3A). Discs that were implanted without fragmentation looked similar to discs that had not been transplanted (Fig. 3D,E). However, the damaged discs had several clones, each composed of many cells in the part of the disc where tissue had been removed (Fig. 3F,G). Discs from 3rd instar larvae consistently have a much higher frequency of recombination for the Ubi <stop <GFP construct, possibly owing to a more accessible chromatin state for this transgene. Thus, the recombination frequencies of the different constructs seem to be dependent on developmental stage and the type of tissue.

The TIE-DYE system potentially allows for a rapid assessment of the effect of gene overexpression or knockdown on tissue growth in imaginal discs. Conventionally, these experiments are carried out by acquiring images from large numbers of discs in which mutant clones have been generated at low density. Clones are typically compared with their wild-type sister clones generated by mitotic recombination using the MARCM method (Lee et al., 2000) or the twin-spot MARCM (Yu et al., 2009) where GAL4 is expressed in one of the two daughter cells following mitotic recombination. Alternatively, FLP-out clones that co-express a gene or RNAi construct together with a marker are compared with discs of the same developmental stage that have FLP-out clones expressing only the marker. However, this type of comparison can be complicated by changes in the rate of developmental progression caused by gene overexpression or knockdown.

To determine whether, using TIE-DYE, the effects of overexpression or knockdown can be discerned even from single discs, we crossed the TIE-DYE tester stock to fly stocks bearing transgenes that are predicted to impact tissue growth. All clones expressing UAS-RFP also express a UAS-RNAi transgene and UAS-dicer2, whereas clones without RFP do not (Fig. 4A-C). Expression of UAS-dicer2 (Fig. 4D) was used as a control to show the relative size of the GAL4/UAS clones (Fig. 4E) and GAL4-independent clones (Fig. 4F). Reducing the expression of the growth-promoting genes yorkie (yki) (Huang et al., 2005) (Fig. 4G-I) and Myc (Johnston et al., 1999) (Fig. 4J-L) resulted in a reduction of clone size and frequency, presumably owing to some clones being eliminated from the tissue. Conversely, knockdown of the negative regulator of growth, expanded (ex) (Boedigheimer and Laughon, 1993), caused a slight increase in clone size (Fig. 4M-O). The data were quantified by averaging the area from multiple GAL4(+) and GAL4(-) clones in the wing pouch from multiple discs. When the areas of labeled clones were measured, the fraction of the area accounted for by GAL4(+) clones is close to 50% in controls and is clearly reduced following the knockdown of yki or Myc (Fig. 4P), and increased slightly with ex knockdown (Fig. 4P; supplementary material Fig. S4). All differences are statistically significant.

The morphogen Dpp plays a crucial role in patterning of the wing disc through induction of different target genes in a concentration-dependent manner (Affolter and Basler, 2007). Dpp is expressed in a strip of cells anterior to the compartment boundary from where it diffuses both anteriorly and posteriorly to regulate growth and patterning. Overexpression of Dpp can lead to overgrowth and changes in wing patterning (Capdevila and Guerrero, 1994; Tabata et al., 1995; Zecca et al., 1995) in a region-specific manner (Rogulja and Irvine, 2005). Hedgehog (Hh) signaling controls the expression pattern of dpp. The posterior compartment produces the ligand Hh whereas only the anterior cells are able to receive the signal. We tested the effect of the knockdown of cubitus interruptus (ci), a transcriptional regulator of the Hh signaling pathway, and observed substantial overgrowth of one side of the wing disc (Fig. 5A-C). Ci can be a transcriptional activator or, after being processed by the proteasome, a transcriptional repressor (reviewed by Jiang and Hui, 2008). The knockdown of ci, which is expressed only in the anterior compartment (Eaton and Kornberg, 1990; Orenic et al., 1990; Schwartz et al., 1995), is predicted to reduce levels of both the activator and repressor forms of Ci. The knockdown of ci leads to overgrowth localized to the anterior of the wing-imaginal disc (Fig. 5B), which is likely due to reduced levels of the repressor form of Ci that normally prevents the expression of Hh-target genes such as dpp in the anterior compartment.

Ectopic expression of dpp leads to overgrowth on both the anterior and posterior sides, especially in the lateral parts of the wing-imaginal disc, when compared with the center of the wing pouch (Fig. 5C-F). The medial region of the disc is composed of many small clones, whereas the lateral regions of the disc contain much larger clones (Fig. 5G,H). Importantly, the effect on clone size is observed equally in clones that express GAL4 and those that do not, thus indicating that the effect is not cell-autonomous (Fig. 5I). Thus, at a single glance, it is possible to observe both the region-specific effects as well as the non-cell-autonomous nature of the phenotype.

One clear advantage of the TIE-DYE system is the ability to visualize both homotypic and heterotypic clone boundaries. Round clones with smooth boundaries are seen in some cases of clonal overgrowth, possibly because of different adhesive properties of the mutant cells and surrounding wild-type cells (Lawrence and Struhl, 1996; Lawrence et al., 1999). In the wing imaginal disc, clones are elongated along the proximodistal axis because of oriented cell divisions. Mutations in genes that regulate planar cell polarity, such as fat and dachsous, disrupt oriented cell division and also generate rounder clones (Baena-López et al., 2005). Use of the TIE-DYE system allowed us to determine whether the rounded clones observed by expression of activated forms of Yki, Ci or Ras are due to alterations in cell affinity or perturbations in oriented cell division. Unlike conventional FLP-out clones, the TIE-DYE system allows us to visualize boundaries between individual mutant clones. If the mutant cells have a distinct cell affinity that is different from that of wild-type cells, then we would expect that the boundaries that separate mutant clones to be irregular while those that separate them from wild-type clones would be smooth. By contrast, if these clones were round simply because of defects in oriented cell division, we would expect individual mutant clones to be round. Expression of an activated form of the transcription factor Yki^S168A (Oh and Irvine, 2009) (Fig. 6A-E), a non-cleavable form of the transcription factor Ci (Price and Kalderon, 1999) (Fig. 6F-J) or a constitutively active form of Ras (Ras85D^V12) (Lee et al., 1996) (Fig. 6K-O) causes clonal outgrowths that appear to sort out from the epithelium. For all three genes, we observed that mutant cells from multiple clones aggregate together in a way in which the boundaries between mutant clones are irregular whereas the boundaries between these cells and wild-type cells are smooth. Thus, application of the TIE-DYE system shows clearly that the shape of the GAL4(+) clones is likely due to perturbation in cell affinity or in cell-cell adhesion.

We have developed a method, TIE-DYE, which allows the simultaneous visualization of multiple marked clones in Drosophila tissues with the ability to express GAL4-driven transgenes in a subset of clones. However, in its current form, TIE-DYE labeling is ubiquitous. The addition of tissue-specific or temperature-sensitive versions of GAL80 could be used to prevent GAL4-driven expression in specific tissues or at specific developmental stages where such expression might be deleterious to the organism. In addition, fluorescent proteins that localize to specific cellular locations (e.g. membranes) could be used for particular applications (e.g. visualizing neuronal processes). Although we have used TIE-DYE mostly in the wing disc, it can be used to study interactions between populations of cells in any tissue, even those that are composed of post-mitotic cells such as neurons. Finally, although the TIE-DYE system was put together for use in Drosophila, analogous systems could be used in most multicellular eukaryotes.

We thank F. Serras and his lab for teaching L.S. disc transplantation. We thank U. Banerjee, J. Knoblich and G. Struhl for Drosophila fly stocks. We thank A. Gerhold for advice and assistance with the X-ray irradiation experiments, N. Ellis for assistance with experiments, and J. Sadler and C. Worley for help with modeling the number of founder cells. We thank L. Pachter and T. Nystul for advice and discussions, along with Hariharan and Bilder lab members. We also thank N. Patel and the students of the Embryology Course at the Marine Biological Laboratory at Woods Hole (2011) for testing out this system. We thank the TRiP at Harvard Medical School for providing transgenic RNAi fly stocks used in this study and the Bloomington Stock Center.