A Genetic Screen in Drosophila for Metastatic Behavior

See allHide authors and affiliations

Science  14 Nov 2003:
Vol. 302, Issue 5648, pp. 1227-1231
DOI: 10.1126/science.1088474


A genetic screen was designed in Drosophila to interrogate its genome for mutations sufficient to cause noninvasive tumors of the eye disc to invade neighboring or distant tissues. We found that cooperation between oncogenic RasV12 expression and inactivation of any one of a number of genes affecting cell polarity leads to metastatic behavior, including basement membrane degradation, loss of E-cadherin expression, migration, invasion, and secondary tumor formation. Inactivation of these cell polarity genes cannot drive metastatic behavior alone or in combination with other tumor-initiating alterations. These findings suggest that the oncogenic background of tissues makes a distinct contribution toward metastatic development.

Most cancer fatalities are due to the ability of later stage tumors to metastasize, or form discontinuous secondary foci (1). Experiments in mammalian systems have indicated the presence of metastasis suppressor genes that can block some aspects of metastasis (2); however, relatively little is known about the genetic alterations sufficient to cause noninvasive tumors to become metastatic. Drosophila melanogaster is an important model system for studying cancer biology. Despite some important differences (i.e., the lack of a vascular circulatory system, and therefore of angiogenesis or vascular tumor cell transport), many biological processes related to tumorigenesis and metastasis are well conserved in flies (3, 4); hence Drosophila could be a model for the genetic basis of metastatic behavior. Using Drosophila, the majority of the genome can be interrogated for metastasis-promoting mutations, something vastly more difficult to accomplish in mammalian systems.

To develop a Drosophila metastasis model, we wished to (i) induce noninvasive tumors in a defined location, through either expression of an oncogene or inactivation of a tumor suppressor gene; (ii) genetically label these tumor cells with a visible marker such as green fluorescent protein (GFP); and (iii) explore whether additional genetic alterations in these cells could elicit metastatic behavior (i.e., the movement of GFP-labeled tumor cells into different tissues) (Fig. 1A). Our genetic scheme used eyeless promoter–driven FLP recombinase expression (eyFLP) (Fig. 1B) (5, 6). This allowed the introduction of multiple genetic alterations (either loss-of-function mutations or gene overexpression) into GFP-labeled cells specifically in the developing larval eye-antennal imaginal discs. Because tumor cells were genetically produced, mechanical disruption of cells and extracellular matrix was avoided. Also, tumor progression and metastatic behavior could be easily monitored and studied in living flies because GFP-labeled tumor cells could be observed in transparent larvae.

Fig. 1.

A Drosophila genetic model for studying tumor progression. (A) Clones of GFP-labeled cells in the eye disc carry a genetic alteration that either activates an oncogene or inactivates a tumor suppressor gene. This results in localized noninvasive tumors that never move from the head region (magnification, ×40). Second-site genetic alterations may cause the development of metastatic behaviors, resulting in the presence of GFP-marked cells in ectopic sites. (B) Expression of the FLP recombinase in the developing eye [eyFLP (6)] mediates mitotic recombination between chromosome arms (34) and produces clones of cells homozygously mutant for a gene that promotes metastatic behavior (i.e., scrib). Only these mutant cells lose the Gal80 repressor (35), which allows Gal4 [under control of the eyFLP-activated Act>y+>Gal4 “flip-out” construct (36)] to direct the constitutive expression of UAS-GFP and UAS-RasV12 [as well as other genes of interest (37)] regardless of their eventual locations and differentiation status. Gal80 expression in nonmutant cells also markedly reduces leaky flip-out construct expression in tissues outside of the eye-antennal region. Expression of GFP, RasV12, and other transgenes is therefore restricted to homozygous mutant cells. Multiple genetic alterations can be combined in the same cell and metastatic behavior can be monitored in vivo by following these GFP-expressing cells.

Clones of GFP-labeled wild-type cells were analyzed in the whole bodies of third-instar larvae, pupae, and adults (Fig. 2A; n > 1000) (5). GFP was observed in the larval eye-antennal imaginal discs as well as in the optic lobes of the brain, but was not detected in other adjacent tissues such as the ventral nerve cord (VNC, arrow in Fig. 2A) (7). GFP was also observed in other tissues and occurred in reproducible locations, depending on the particular eyFLP transgene used [mostly in the gonad (5)].

Fig. 2.

Tumor progression phenotypes in Drosophila (magnifications, ×8). (A to G, K to Z) eyFLP-induced mosaic clones expressing GFP in different flies are shown in the top panels; cephalic complexes dissected from third-instar larvae are shown in the bottom panels. The anterior is to the top in all panels; the mouth hook [arrowhead in (A)] is at the top and the ventral nerve cord [VNC, arrow in (A)] is pointing down in all bottom panels. Metastatic tumorcells caused an extended larval period, and flies died as bloated larvae (E) with cells invading the VNC [arrow in (E)]. The relevant genotype of mutant cells in each fly is indicated above the top panel in the following manner: transgenes/loss-of-function mutations. RasV12/scrib–/– (H) and RasV12/lats–/– (I) cephalic complexes dissected at an early time point. (J) RasV12/scrib–/– tumors generated by a weak ey-FLP [eyFLP1.2 (38)].

Alterations in the Ras oncogene or the lats tumor suppressor gene contribute to tumorigenesis in both flies and mammals (811). We generated flies with noninvasive tumors by inducing clones of cells either mutant for lats or expressing activated Ras (UAS-Ras85DG12V or “RasV12”, Figs. 1A and 2B; n > 1000 for each). In either case, the amount of GFP-marked tissue in the eye-antennal discs was noticeably increased, and most mutant flies died before adulthood. However, GFP-labeled mutant cells were always located in the same areas as in the wild type. Mutant cells were not seen outside of the eye-antennal disc/optic lobe region even after eye disc eversion, suggesting that tissue integrity was not compromised. Thus, this system allowed the potential detection of metastatic behaviors caused by additional mutations.

We used the GFP-labeled RasV12 eye expression system to screen for additional preexisting or newly induced mutations that promoted tumor progression (5, 12). Mutations in the scrib gene (13) in conjunction with RasV12 expression (heretofore referred to as RasV12/scrib–/–) caused metastatic behavior. Third-instar RasV12/scrib–/– larvae carried large primary tumors and had small groups of cells floating in the hemolymph and other distant sites, which suggests that tumor cells formed secondary multicellular growths (Fig. 2E; n > 200). Dissection at various time points revealed that the majority of ectopic tumor cells progressively spread from the primary tumors onto the VNC (arrow in Fig. 2E), eventually enveloping it (Fig. 3, A to D), and also spread into other tissues such as the first and second leg discs (Fig. 3E) and tracheal vasculature (Fig. 3, E to G) at a lower frequency. All of the RasV12/scrib–/– flies displayed the VNC invasion phenotype, making it the best measure of metastatic behavior. Confocal sections indicated that cells could invade the inside of the VNC (Fig. 3, H and I), and the leading edge of these cells had a morphology and F-actin–rich periphery common to actively migrating cells (Fig. 3, J and K). In transplantation experiments, the appearance of secondary growths varied considerably, regardless of genotype; however, only RasV12/scrib–/– tumors invaded host tissues (5) (fig. S1, A to L and Q).

Fig. 3.

RasV12/scrib–/– cells are invasive. (A to D) Representative VNCs dissected from RasV12/scrib–/– larvae showing progressive migration of tumor cells (green) at day 6 (A, B), day 10 (C), and day 15 (D) after egg lay. Arrowheads indicate the leading edge of the migrating tumors. Magnification, ×50. (E to G) Serial confocal sections showing RasV12/scrib–/– cells invading the interior of leg discs and a tracheal branch (arrowheads). RasV12/scrib–/– cells are marked by β-galactosidase staining (green pseudocolor), and discs and trachea are outlined using the G454 protein trap (red pseudocolor). Magnification, ×200. (H and I) Serial confocal sections showing RasV12/scrib–/– cells (green) invading into the VNC (visualized with Texas Red–phalloidin staining, red). Magnification, ×60. (J and K) Close-up view of tumorcells in (I) with high levels of peripheral F-actin staining. In (K), projections extending from the leading edge (arrowhead) are highlighted. Magnification, ×300.

RasV12/scrib–/– metastatic behavior was blocked with a UAS-scrib transgene, confirming the causal role of scrib inactivation (Fig. 2F; n = 25). scrib inactivation causes the presence of overgrowths in homozygous flies (5, 14). However, cells mutant for scrib alone grew poorly in the eye discs of either mosaic or zygotically mutant flies, and did not invade other tissues (Fig. 2, C and D; n = 100 and 10, respectively), which suggests that scrib inactivation and RasV12 expression are both needed to cause metastatic behavior in eye disc cells.

We considered the possibility that excessive proliferation may cause a breakage of tissue boundaries, resulting in the appearance of ectopic GFP cells. However, other mutations identified in our screen caused excessive tumor growth without the appearance of ectopic cells. For example, clones of cells expressing RasV12 and mutant for lats (RasV12/lats–/–) resulted in immensely overgrown tumors in the eye-antennal discs (Fig. 2G; n = 50). As shown by dissection at an early time point, RasV12/lats–/– tumors grew faster than RasV12/scrib–/– tumors (Fig. 2, H and I), but ectopic cells were never seen on the VNC or other locations. Also, a weakly expressed eyFLP (eyFLP1.2) produced much smaller primary RasV12/scrib–/– overgrowths than did eyFLP1, but these cells still exhibited metastatic behavior (Fig. 2J; n = 10), indicating that excessive tumor growth alone does not cause metastatic behavior in RasV12/scrib–/– larvae.

Basement membrane degradation contributes to tumor metastasis (15). To determine whether the basement membrane was compromised in RasV12/scrib–/– tumors, we generated antibodies to laminin-A (5) and used a GFP-tagged protein trap in the viking/collagen IV gene (16, 17). The basement membrane was smooth and continuous on the outer surface of eye discs with wild-type, RasV12-expressing, or scrib–/– cells (Fig. 4, A to C). In discs containing RasV12/scrib–/– cells, however, there were many points of discontinuity in the basement membrane, and mutant cells spread from these areas (Fig. 4, D to F). Thus, like human malignant tumors, Drosophila metastatic tumors can acquire the ability to degrade basement membranes.

Fig. 4.

Degradation of basement membrane. Confocal sections through eye discs of mutant clones of scrib cells (A), RasV12 cells (B and C), or RasV12/scrib–/– cells (D to F); Magnification, ×250. Mutant cells are labeled with GFP [green, (A) to (D)] or β-galactosidase staining [green pseudocolor, (E) and (F)]. Wild-type cells are unlabeled. The basement membrane is visualized by staining with antibody to laminin [red, (A) to (D)] ora collagen IV protein trap [red pseudocolor, (E) and (F)].

Mammalian invasive tumors commonly down-regulate E-cadherin through a variety of mechanisms, and this may be a causal factor in driving tumor progression (18). E-cadherin expression was lowered in RasV12/scrib–/– cells, a phenotype likely due to inactivation of scrib (fig. S2). Expression of full-length E-cadherin (UAS-DECH) in RasV12/scrib–/– cells suppressed their metastatic behavior (Fig. 2K; n = 58). Expression of a truncated E-cadherin lacking its extracellular domain (UAS-CADHintra5) no longer suppressed VNC invasion (Fig. 2L; n = 15), indicating the importance of the cell adhesion function of E-cadherin in preventing metastatic behavior. We next examined whether down-regulation of shotgun (shg), the Drosophila E-cadherin homolog (19, 20), was sufficient to induce metastatic behavior in RasV12-expressing cells (RasV12/shg–/–). Although RasV12/shg–/– cells caused disc eversion defects during metamorphosis, mutant cells were not observed invading the VNC (Fig. 2M; n = 21). This was confirmed with transplantation experiments (fig. S1, M to P), together suggesting that loss of E-cadherin is necessary but not sufficient for metastatic behavior in RasV12-expressing tumors.

scrib interacts with lethal giant larvae (lgl) and discs large (dlg) (14), and mutations in lgl or dlg when combined with RasV12 expression caused metastatic behavior similar to that seen in RasV12/scrib–/– flies (Fig. 2, N and O; n = 25 and 28, respectively). Aside from overgrowth phenotypes (5), mutations in scrib, lgl, and dlg also cause defects in cell polarity and epithelial monolayer formation (16). Other genes such as bazooka (baz), stardust (sdt), and cdc42 are also necessary for maintaining cell polarity, cell shape, and/or epithelial morphology, but their mutation does not result in overgrowth (21, 22). Inactivation of any one of these genes caused metastatic behavior when combined with RasV12 expression (Fig. 2, P to R; n = 14, 10, and 10, respectively). Thus, alterations disrupting cell polarity and epithelial morphology play a key role in the development of metastatic behavior in noninvasive RasV12-expressing cells, perhaps through the abrogation of intercellular junctions or the mislocalization of plasma membrane–targeted signaling molecules. Interestingly, the identified genes function in an interdependent genetic hierarchy (23, 24), which suggests a concerted signaling pathway capable of suppressing metastatic behavior in tumor cells expressing RasV12.

Because oncogenic Ras promotes proliferation, survival, and cell growth in Drosophila (9, 25, 26), one or more of these processes may be sufficient to make scrib–/– cells become metastatic. The p21 cyclin kinase inhibitor blocks RasV12-dependent proliferation in the Drosophila eye (11). Its expression in RasV12/scrib–/– cells decreased tumor size but did not inhibit the ability of these cells to spread into the VNC (Fig. 2S; n = 20). Expression of the oncogenic proliferation-stimulating proteins E2F and Dp (27) could not substitute for RasV12 in scrib–/– mutant cells (Fig. 2T; n = 8). Expression of the p35 cell death inhibitor (28) also could not permit the metastatic progression of scrib–/– mutant cells (Fig. 2U; n = 10). Coexpression of E2F, Dp, and p35 visibly increased the amount of scrib–/– mutant tissue, but again could not cause an induction of metastatic behavior (Fig. 2V; n = 14). Expression of the growth-promoting oncogene homologs of c-Myc (29) or Akt (30) in scrib–/– cells visibly increased the amount of mutant tissue, but neither of these could promote metastatic behavior (Fig. 2, W and X; n = 12 and 9, respectively). Inactivation of lats and scrib together (scrib–/–, lats–/–) resulted in tumors that did not exhibit metastatic behavior (Fig. 2Y; n = 16). Expression of dAkt in scrib–/–, lats–/– cells further increased tumor size, but again did not induce metastatic behavior (Fig. 2Z; n = 16). This indicates that increased proliferation, growth, and survival by means other than RasV12 are not sufficient to cause the metastatic progression of scrib–/– cells. Thus, the metastasis-promoting effect of scrib inactivation is highly dependent on its specific cooperation with the RasV12 allele. Moreover, aside from its known effects on proliferation, growth, and survival, RasV12 may function through an as yet undefined cellular mechanism to elicit metastatic progression in scrib–/– cells.

It has proven difficult to systematically study the genetic basis of metastasis with the currently available techniques. The Drosophila system described here circumvents the complication of acquired background mutations, which can occur through repeated passaging of cell lines or during the typically long latent period of mammalian tumor progression. In our initial screen, we found that mutations in different genes affecting the same physiological process—epithelial cell polarity maintenance—are sufficient in combination with RasV12 to promote metastatic behavior in vivo. Interestingly, later stage human cancers typically lose cell polarity markers and epithelial structure during epithelial to mesenchymal transition (31). Also, E-cadherin loss, basement membrane degradation, and induction of cell migration and invasion relate well to observations made in human metastasis (15, 31, 32), which suggests that the ongoing screen will uncover genes and general mechanisms relevant to malignancy in humans.

It has been proposed that oncogenes such as Ras may play a dual role in tumorigenesis and metastasis (33); however, this has not yet been rigorously proven in mammalian systems, as the effects of Ras in cell culture depend greatly on the particular cell line used. We provide experimental evidence that genetic alterations promoting noninvasive tumor growth can indeed make additional contributions to the development of metastatic behavior, as RasV12 expression is a crucial factor in making cell polarity–deficient cells metastasize. Furthermore, we show that oncogenic Ras specifically cooperates with inactivation of cell polarity genes to promote metastatic behavior. This may provide an explanation for the different metastatic potential observed in tumors of distinctive origins. The Drosophila genetics techniques described here should make it easier to analyze the specific targets of RasV12 in metastatic cells, to identify other genes that cooperate with RasV12 or other oncogenic alterations in promoting metastasis, and to elucidate the cellular processes that go awry during metastatic progression.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 and S2

References and Notes

References and Notes

View Abstract

Navigate This Article