The Drosophila Netrin receptor frazzled/DCC functions as an invasive tumor suppressor

Background Loss of heterozygosity at 18q, which includes the Deleted in Colorectal Cancer (DCC) gene, has been linked to many human cancers. However, it is unclear if loss of DCC is the specific underlying cause of these cancers. The Drosophila imaginal discs are excellent systems in which to study DCC function, as it is possible to model human tumors through the generation of somatic clones of cells bearing multiple genetic lesions. Here, these attributes of the fly system were utilized to investigate the potential tumor suppressing functions of the Drosophila DCC homologue frazzled (fra) during eye-antennal disc development. Results Most fra loss of function clones are eliminated during development. However, when mutant clone cells generated in the developing eye were rescued from death, partially differentiated eye cells were found outside of the normal eye field, and in extreme cases distant sites of the body. Characterization of these cells during development indicates that fra mutant cells display characteristics of invasive tumor cells, including increased levels of phospho-ERK, phospho-JNK, and Mmp-1, changes in cadherin expression, remodeling of the actin cytoskeleton, and loss of polarity. Mutation of fra promotes basement membrane degradation and invasion which are repressed by inhibition of Rho1 signaling. Although inhibition of JNK signaling blocks invasive phenotypes in some metastatic cancer models in flies, blocking JNK signaling inhibits fra mutant cell death, thereby enhancing the fra mutant phenotype. Conclusions The results of this investigation provide the first direct link between point mutations in fra/DCC and metastatic phenotypes in an animal model and suggest that Fra functions as an invasive tumor suppressor during Drosophila development.


Background
Loss of heterozygosity (LOH) at chromosome 18q, which includes the DCC gene, was identified in a large percentage of colorectal cancers [1]. LOH at 18q is associated with decreased DCC expression and has been linked to many other types of cancer, including neuroblastomas, hematologic malignancies, and gastric, endometrial, prostate, ovarian, esophageal, testicular, breast, and glial cancers [2,3]. It was therefore hypothesized that DCC functions as a tumor suppressor. However, although elevated levels of the DCC ligand Netrin (Net) have been linked to oncogenic phenotypes [4,5], point mutations in DCC have not been directly associated with tumorigenesis in animal models [2]. For example, loss of DCC is not associated with tumor formation in a murine model [6]. However, although the impact of mutating DCC in every cell of an organism has been investigated [6], somatic DCC mutant clones have not been assessed for cancer phenotypes in animal models. Such an analysis may provide an effective simulation of human tumors and offer insight into the putative tumor suppressing functions of DCC.
Drosophila melanogaster is an excellent system in which to study tumor suppressor gene function [7,8]. Many of the hallmarks of human cancer, including selfsufficiency in growth and proliferative signals, insensitivity to anti-proliferative signals, evasion of apoptosis, and invasion/metastasis can be found in Drosophila [7]. Mutagenesis screens in Drosophila have led to the identification of cancer genes, and studies in flies led to the discovery of important interactions between signaling pathways that function during oncogenesis [7,9,10]. One great advantage of the fly system which is of particular relevance to this investigation is that it is relatively simple to use the FLP/FRT system [11] to create clones of genetically distinct somatic cells that model human tumors. Furthermore, since multiple mutations typically cooperate to generate metastatic tumors, it is important to use animal models such as Drosophila in which multiple genetic manipulations can be performed simultaneously when clones are generated.
It was hypothesized that exploitation of the advantages of the Drosophila system to study DCC/Fra function would provide new insights into the function of this gene in oncogenesis. Our analysis of fra loss of function point mutations during fly development suggest that Fra functions as a tumor suppressor. The results of this investigation directly link loss of function point mutations in fra/ DCC to metastatic phenotypes in an animal model for cancer.

Results And Discussion
Death of fra mutant clone cells fra is expressed in imaginal discs [12], suggesting that Fra might function during imaginal disc development. To test this idea, somatic fra loss of function (LOF) clones were generated during eye development. Two different EMS null alleles, fra 3 and fra 4 , both previously shown to lack Fra protein expression [13] were studied in this manner. When either fra 3 or fra 4 mutant clones are induced during eye disc development, most mutant clones fail to persist through adulthood ( Figure 1A,B). In third instar eye discs, fra 3 and fra 4 mutant clones are typically very small. Staining with anti-cleaved caspase-3, a marker for apoptotic cells, reveals dying fra mutant clone cells in the eye-antennal disc ( Figure 1C). The majority of fra mutant eye clones are eliminated by the end of the third instar.

Analysis of fra loss of function clones in adult flies
Although most fra mutant clones do not persist, clones occasionally survive into the adult stages. In~1% of adult flies in which fra 3 mutant clones are generated (n = 500), mutant clone cells form overgrowths (Figure 2A,B). These observations suggested that it would be useful to develop a strategy that would permit more efficient analysis of fra LOF clones. Viability of fra 4 mutant cells was increased through expression of the baculovirus caspase inhibitor P35 in mutant clones. In these experiments, the MARCM system [14,15] was used to drive P35 expression in fra 4 mutant clones in the developing eye. Generation of P35-rescued fra 4 mutant clones typically results in death of the organism (~99% of flies with the appropriate genotype for generating these clones die before eclosion; 2000 F1 progeny were scored from the appropriate P1 cross for generating this genotype). However, interesting phenotypes, which are described in detail below, were observed in 100% of the adult escaper flies that managed to survive to adulthood.
Perhaps the most striking phenotype observed in flies in which P35-rescued fra 4 mutant clones are generated in the eye is the detection of eye cells throughout the bodies of adult flies. One example is shown if Figure 2C, in which GFP-marked fra mutant clone cells generated during eye development with the eyFLP driver are located on the wing of the adult. Mutant eye cells appear to differentiate, at least in part, as they express the neural differentiation marker Elav in third instar discs ( Figure 3B, G) and display red pigmentation in adults ( Figure 2C). This intriguing phenotype is not observed in conjunction with P35 ectopic expression alone (not shown). Thus, experiments with two different LOF fra alleles suggest that Fra may have tumor-suppressing functions in Drosophila, a hypothesis that was explored in more detail during the course of this investigation.

Analysis of fra mutant cells during eye development
The initial results obtained in adult flies suggested that fra mutant eye cells might have the capacity to invade surrounding tissues. To assess this possibility, fra LOF clones were analyzed during larval development. These analyses were performed in the developing eye-antennal disc, where clones were generated with eyFLP. Although the ey driver used in this investigation (see methods) was designed to be very tight, occasional FLP expression has been reported in the brain and gonads [16]. We therefore avoided analyses in these tissues and restricted all analyses to the eye-antennal disc so that the point of mutant clone origin would be certain. When fra 3 ( Figure 3D), fra 4 ( Figure 3F), or P35-rescued fra 4 ( Figure 3B,G) LOF clones were generated in the developing eye, a small number of fra mutant photoreceptor cell bodies, typically restricted to the eye field ( Figure 3A,C), were detected in the optic stalk ( Figure 3D,E,F,G) and other inappropriate regions of the eye-antennal disc ( Figure 3B). In such cases, fra mutant Elav-positive foci were isolated and often located outside of mutant clone boundaries ( Figure  3B,D,E,G). Comparable results were obtained when FLP was expressed under heat shock control ( Figure 3E).
While these results suggested that individual fra mutant cells might become invasive and leave the clone, one could alternatively argue that fra mutant cells outside of the eye field have undergone a transformation toward an eye cell fate. Several pieces of data indicate that this is not the case. First, when P35-rescued fra mutant clones are induced in the antennal disc, there is no evidence of a general conversion of these mutant cells to eye cell fates, as the majority of cells in P35-rescued fra 4 antennal clones do not express the photoreceptor marker Elav ( Figure 3B). Furthermore, cells of fra mutant clones located anterior to the furrow do not appear to undergo precocious differentiation as photoreceptor cells ( Figure 3B). Finally, generation of fra 3 mutant clones in the wing disc with hsFLP does not promote transformation of these cells toward an Elavpositive photoreceptor fate ( Figure 3H); this result indicates that the fra mutant cells located on the adult wing ( Figure 2C) are not simply the result of any rare leaky expression of FLP inducing a fra mutant wing clone that is transformed into eye tissue. Together, these data suggest that eye cells located outside of the eye field do  Loss of fra in the developing eye results in expression of invasive cell markers and altered polarity, adhesion, cytoskeletal organization, and proliferation fra mutant eye-antennal disc clones were next assessed for molecular features typically associated with invasive cell types. Although use of P35-rescue was often utilized to permit analysis of larger clones generated with the eyFLP driver, the use of P35 was experimentally controlled for throughout the investigation. Phospho-JNK ( Figure 3I), Mmp1 ( Figure 3K), and phospho-ERK ( Figure 3M), all of which are upregulated in other Drosophila invasive tumor models [17], are elevated in P35-rescued fra 4 mutant clones. Mutant cells were examined for potential changes in adhesion, cytoskeletal organization, and polarity, additional features of metastatic cell types. Within P35-rescued fra 4 mutant clones, E-cad expression is slightly increased, yet delocalized from the membrane ( Figure 4A,C1). However, little E-cad expression is detected in individual fra mutant cells that are found outside of clone boundaries (not shown). Changes in cytoskeletal organization are observed in fra mutant cells, as illustrated by alterations in the actin cytoskeleton detected in a P35-rescued fra 4 clone ( Figure 4D). None of these phenotypes were found in clones expressing P35 alone ( Figure 3J,L, Figure 4B,C2,E).
Polarity changes were assessed through analysis of Discs-large (Dlg), a basolateral marker and neoplastic tumor suppressor [18]. P35-rescued fra 4 mutant cells display delocalization of Dlg ( Figure 4F,G), suggesting that loss of fra results in disrupted apical-basal cell polarity. Disruption of apical-basal polarity is often accompanied by changes in cell proliferation [18]. Increased expression of the mitotic marker phosphorylated Histone H3 is also detected in P35-rescued fra 4 mutant cells ( Figure 5A; compare to Figure 5B). Expression of this mitotic marker was even observed in~10% of Elav-positive fra mutant cells located in the antennal disc, far outside of the normal photoreceptor field (Figure 5D, n = 20). Again, neither changes in polarity nor proliferation were found in clones expressing P35 alone ( Figure 4H, Figure 5C). In summary, characterization of fra mutant cells during eye-antennal disc development indicates that they display several features that are characteristic of metastatic tumor cells.

Loss of fra promotes basement membrane degradation and invasion
Several features of fra mutant cells, including changes in adhesion, polarity, cytoskeletal organization, and activation of JNK and Mmp-1 (Figures 3 and 4), are consistent with basement membrane invasive tumor models in Drosophila [7,19,20]. We therefore tested the impacts of fra LOF on basement membrane degradation and invasion in the developing eye disc. The basement membrane was degraded adjacent to both fra 3 ( Figure 6A) and P35-rescued fra 4 mutant clones ( Figure 6C). Unlike control cells (Figure 6B), both fra 3 ( Figure 6A) and P35rescued fra 4 mutant cells ( Figure 6C) were found to lose epithelial polarity and invade the basement membrane.
Live imaging experiments captured the invasive properties of fra mutant cells. GFP-positive P35-rescued fra 4 mutant cells detected just outside of the eye-antennal disc were imaged in live tissue preparations. In the resulting videos (Additional File 1 is a still frame reference for the movies in Additional Files 2 and 3), these GFP-positive cells are motile. At higher magnification (Additional File 3), projections that extend and retract as the cell migrates are observed. Such motility was never observed in response to expression of P35 alone. These data indicate that fra mutant cells exhibit invasive properties.

Inhibition of Rho but not JNK signaling suppresses basement membrane degradation and invasion of fra mutant cells
Inhibition of JNK signaling through expression of dominant negative-JNK is known to prevent basement membrane degradation and cell invasion in other fly metastatic tumor models [7,19,20]. However, expression of DN-JNK in P35-rescued fra 4 mutant clones did not prevent the actin cytoskeleton ( Figure 7A) or E-cad expression/localization (not shown) changes, Mmp-1 expression ( Figure 7B), basement membrane degradation, or cell invasion ( Figure 6D) associated with loss of fra. Furthermore, no adult animals in which such clones had been generated were recovered in this investigation. Rho1 activation has also been associated with invasive cell behavior in flies [16,21]. Expression of DN-Rho1 in P35-rescued fra 4 mutant clones repressed basement membrane degradation and partially suppressed fra mutant cell invasion ( Figure 6E). These results suggest that loss of fra promotes Rho1-mediated basement membrane degradation and cell invasion. Although repression of basement membrane degradation and invasion of fra mutant cells in which Rho signaling had been blocked was substantial, it was not complete (Figure 6E), and expression of Mmp-1 was not entirely blocked (not shown). Likewise, co-expression of DN-Rho did not block organismal lethality, as no adult animals in which clones had been generated were recovered.

Inhibition of Jnk Signaling enhances the fra mutant phenotype
Although blocking JNK signaling is known to prevent basement membrane degradation and cell invasion in fly metastatic tumor models [7,19,20], as discussed above, it did not prevent basement membrane degradation or cell invasion associated with loss of fra ( Figure 6D). In fact, analysis of the impact of inhibiting JNK signaling in fra mutant clones suggested that blocking JNK signaling actually enhances the fra LOF phenotype. This is illustrated by the highly dysmorphic overgrowths that can result when JNK signaling is blocked in fra LOF clones ( Figure 7D).
Bossuyt et al. [22] recently demonstrated that tumorous eyeful cells activate a JNK-mediated stress response pathway that results in cell suicide. It therefore seemed possible that clonal loss of fra, which results in cell death (Figure 1), might occur in response to JNK activation. Indeed, fra 3 mutant clones that have not been rescued from lethality exhibit increased levels of phosphorylated JNK ( Figure 7C). Little cleaved caspase-3 expression is detected when JNK signaling is blocked in P35-rescued fra mutant clones ( Figure 7D). These results suggest that cell death associated with non-rescued fra LOF clones ( Figure 1A-C) is induced by JNK.

Direct connection between loss of function point mutations in fra/DCC and cancer-like phenotypes
Although mutations in human chromosome 18q, which includes DCC and a number of other genes, have been linked to many human cancers [2,3,8], it is unclear if loss of DCC is the specific underlying cause of these cancers, as point mutations in DCC had not been directly associated with cancer phenotypes in animal models [2]. The results of this investigation provide the first direct link between LOF point mutations in fra/ DCC and cancer-like phenotypes in an animal model. It is interesting to consider the results of this investigation in relation to the DCC knockout mouse [6], which does not have a higher incidence of tumors. In addition to DCC, several other proteins function as Net receptors in vertebrates [2,3], suggesting that redundancy could explain the lack of cancer phenotypes in the DCC knockout mouse. However, recent studies have highlighted the importance of studying cancer genes in a clonal context [16]. It is therefore possible that using an experimental design more comparable to the one employed here, one in which clones of mutant DCC cells are generated and rescued from death, could reveal cancer phenotypes in mice.
Although many of the results of this investigation are consistent with other metastasis models in flies [19,20,23], it should be noted that comparable adult phenotypes in which eye cells are detected outside of the eye field of the adult are not typically reported. One exception is the eyeful study described by Ferres-Marco et al. [24]. A potential explanation for the lack of similar phenotypes is that such adult flies, at least in the case reported here, are rarely viable. It should also be noted that the eyeful phenotype, like the P35-rescued fra mutant phenotype, was the result of several simultaneous genetic manipulations. These studies underline the importance of using genetically tractable animal models wherein such complicated genetic manipulations are possible.
It should be noted that although P35 has been associated with conferring an "undead" cell fate in Drosophila, (see [25,26] ), we do not believe that use of P35 was a confounding factor in this investigation. First, no non-cell autonomous impacts were observed. Also, data demonstrating the invasive qualities of non-P35-rescued fra [3] and fra [4] mutant clones were included. Non P35-rescued mutant cells invade the optic stalk ( Figure  3D,F). Basement membrane degradation and invasion is not dependent on P35-rescue of clones ( Figure 6A). Finally, an example in which a non-rescued clone has resulted in a tumor (Figure 2A,B) is included. Rescue with P35 was a useful tool, as it increased the number and size of mutant clones that could be analyzed.

Mutations in fra/DCC promote Rho1-mediated invasion
Developmental characterization of fra mutant cells indicates that they express a number of tumor cell markers ( Figure 3I,K,M) and exhibit changes in expression of E-cad ( Figure 4A,C1), reorganization of the actin cytoskeleton ( Figure 4D), and loss of apical-basal polarity ( Figure 4F,G), characteristics which are typical of invasive tumor cells. fra mutant cells also overproliferate ( Figure 5A,D). Furthermore, loss of fra results in basement membrane degradation and invasion ( Figure 6A,C). A proportion of these invasive cells can divide, even if they retain expression of the neural differentiation marker Elav ( Figure 5D). These data, in conjunction with live imaging assays (Additional Files 1, 2, and 3), support the notion that loss of fra results in a metastatic phenotype.
Although it was hypothesized that JNK, a mediator of metastasis [27] that is upregulated in fra mutant cells (Figures 3I, 7C), might drive their invasion, inhibition of JNK signaling did not suppress basement membrane degradation or invasion of fra mutant cells ( Figure 6D). Instead, inhibition of JNK signaling in fra mutant cells appears to block JNK-mediated cell death ( Figure 7C,D), resulting in enhanced overgrowth of fra mutant cells ( Figure 7D) which retain their invasive qualities ( Figure 6D). These results are somewhat surprising, as JNK-induced Mmp1dependent degradation of the basement membrane was identified as a critical early event during the invasion process of metastasizing cells [27]. However, several groups have recently demonstrated that JNK function is context dependent, and that JNK activity does not always function to promote cell invasion [22,28]. Additionally, JNKmediated apoptosis has been associated with a number of other tumor suppressor mutations in Drosophila (see discussion in [28]), and it is therefore not unexpected that it would play a key role in the removal of fra mutant clone cells. In future experiments, it will be useful to determine if inhibition of JNK signaling in the absence of P35 rescue is sufficient to prevent cell death in fra mutant clone cells.
Rho1 has also been associated with invasive cell behavior in the wing and eye imaginal discs [16,21]. Rho proteins are important regulators of cell shape, motility, and cytoskeletal arrangements that drive epithelial-mesenchymal transitions [29]. Inhibition of Rho1 signaling in fra mutant cells represses basement membrane degradation and invasion ( Figure 6E). These results suggest that Rho1 can induce fra mutant cell invasion independently of JNK, at least in an eye developmental context. Repression of basement membrane degradation and invasion by DN-Rho1 was substantial, though not complete. This could result from incomplete deactivation of Rho1 signaling, but might also signify that basement membrane degradation and invasion could be mediated by additional factors. In support of this idea, Mmp-1 expression is not entirely suppressed when Rho1 signaling is blocked in fra mutant clones. We are currently pursuing global analysis of gene expression in fra mutant cells, which may elucidate additional molecules that are involved. It will be interesting to determine if other signaling pathways, such as the Hippo pathway which has attracted a great deal of attention in recent Drosophila imaginal disc studies [22,28,30], is important in the fra mutant clonal cell context.

Implications for the DCC dependence receptor model
DCC is proposed to function as a dependence receptor that induces cell death in the absence of Net ligand. In the presence of Net, which is expressed in many types of tumors, cells are proposed to escape death through downregulation of DCC [2,5,31]. In support of the DCC dependence receptor model, Mazelin et al. [32] demonstrated that ectopic expression of Net-1 in the mouse gastrointestinal tract results in spontaneous formation of hyperplastic and neoplastic lesions. They also showed that overexpression of Net-1 in an APC mutant background promotes intestinal tumor development by blocking DCC-induced apoptosis [32]. More recently, Fitamant et al. [5] showed that Net-1 promotes cell survival in breast cancer cells. These studies support the theory that DCC functions as a dependence receptor with respect to its ability to induce apoptosis. Based on the dependence receptor model, one might have expected that loss of function mutations in fra would result in increased cell viability. However, in the Drosophila eye/antennal disc, most loss of function fra clones do not persist (Figure 1). One possible interpretation of these data in light of the recent Net literature is that Net ligand, which is expressed throughout the developing eye-antennal disc [12], provides a survival cue for developing cells in this tissue; fra mutant cells would lack the ability to receive this survival cue and ultimately die in response to JNK activation ( Figure 7C).
It is also interesting to consider the dependence receptor model in terms of metastasis. If the dependence receptor model is applicable to metastasis, DCC would be expected to function as a suppressor of metastasis, and as supported by this investigation, mutation of DCC/fra would be expected to promote metastasis. According to the dependence receptor model, binding of Net ligand to DCC would inhibit its ability to suppress metastasis, and an abundance of Net ligand would therefore promote metastasis. Several recent studies have demonstrated that Net-1 does promote metastasis. Rodrigues et al. [33] described metastatic phenotypes associated with Net-1 activation in human colorectal cancer cells. Furthermore, Fitamant et al. [5] found that in comparison to non-metastatic breast tumors, Net-1 levels are elevated in a large proportion of metastatic breast cancers. Their study showed that reduction of Net-1 signaling inhibits metastasis in a mouse model of lung colonization of a mammary cancer cell line, as well as in a model of lung metastasis in xenografted human breast tumors. Net-1 therefore induces metastasis of several types of cancer cells. Taken together, the results presented here in conjunction with these other studies suggest that DCC functions as a suppressor of metastasis, a function that is inhibited by an abundance of Net ligand which can suppress DCC and promote invasion. This model may also apply to regulation of the invasive growth of axons during neural development.

Live imaging
Live imaging was performed with a Zeiss LSM 710 laser scanning microscope. Third instar larvae were dissected and placed in a drop of saline solution on a microscope slide under a number 1 glass cover slip. Motile GFPexpressing clones were checked for auto-fluorescence, which would rule out that they were hemocytes. In Additional File 2, cells were imaged every 5.0 seconds with a frame average of 1 for~2.5 minutes. In Additional File 3, the cell was imaged every 1.0 second with a frame average of 1 for~1 minute. The resulting movies were processed using Zeiss LSM software.