Aneuploidy-induced delaminating cells drive tumorigenesis in Drosophila epithelia

Abstract

Genomic instability has been observed in essentially all sporadic carcinomas. Here we use Drosophila epithelial cells to address the role of chromosomal instability in cancer development as they have proved useful for elucidating the molecular mechanisms underlying tumorigenic growth. We first show that chromosomal instability leads to an apoptotic response. Interestingly, this response is p53 independent, as opposed to mammalian cells, and depends on the activation of the c-Jun N-terminal kinase (JNK) signaling cascade. When prevented from undergoing programmed cell death (PCD), chromosomal instability induces neoplasic overgrowth. These tumor-like tissues are able to grow extensively and metastasize when transplanted into the abdomen of adult hosts. Detailed analysis of the tumors allows us to identify a delaminating cell population as the critical one in driving tumorigenesis. Cells loose their apical–basal polarity, mislocalize DE-cadherin, and delaminate from the main epithelium. A JNK-dependent transcriptional program is activated specifically in delaminating cells and drives nonautonomous tissue overgrowth, basement membrane degradation, and invasiveness. These findings unravel a general and rapid tumorigenic potential of genomic instability, as opposed to its proposed role as a source of mutability to select specific tumor-prone aneuploid cells, and open unique avenues toward the understanding of the role of genomic instability in human cancer.

Genomic instability (GI) was originally proposed to cause cancer over 100 years ago, because it was discovered in all epithelial cancers investigated (1) and found to cause abnormal and tumor-like phenotypes in developing sea urchin embryos (2). Since then, GI has been observed in essentially all sporadic carcinomas, the most common type of cancer occurring in humans and derived from putative epithelial cells. There are various forms of GI and the most common in cancer is chromosomal instability (CIN), which refers to the high rate by which chromosome structure and number changes over time in cancer cells compared with normal cells (3). Eukaryotic cells have developed quality control mechanisms, collectively called checkpoints, which ensure correct execution of cell cycle events to maintain genome stability. Three checkpoints have been thoroughly documented: the DNA damage checkpoint, which is able to block cells in G1, S, G2, or even mitosis; the DNA replication checkpoint, which monitors progression through S phase, and the spindle-assembly checkpoint (SAC), which monitors attachment of chromosomes to functional spindle microtubules, ensuring equal segregation of genomic material among daughter cells (4–6). Unfortunately, only a few lines of evidence support a potential role of CIN in tumorigenesis. CIN is not sufficient to drive tumorigenesis in Drosophila stem cells (7) and spontaneous tumor development in SAC mutant mice is at relatively low rates (8). Interestingly, mouse strains mutant for SAC genes activate the ataxia telangiectasia mutated (ATM)/p53 pathway and additional depletion of p53 induces tumorigenesis of mouse strains with a dysfunctional SAC (9). So far, the molecular mechanisms and cellular behaviors underlying CIN-induced tumorigenesis remain uncharacterized and perhaps the most accepted hypothesis on the role of CIN in cancer development is the one that proposes that CIN is a source of mutability that helps the tumor cell population to pass through the critical steps of tumorigenesis such as cell delamination, extravasation, and invasiveness (10). To address this further in a genetically tractable system, we have selected a group of genes involved not only in the spindle assembly checkpoint, but also in spindle assembly, chromosome condensation, and cytokinesis, whose depletion leads to CIN (7), and analyzed its impact in highly proliferative epithelial cells of Drosophila.

Results

CIN Leads to Drosophila p53-Independent Programmed Cell Death.

The Drosophila primordia of adult wing and eye structures (the wing and eye imaginal discs) are epithelial monolayers that actively proliferate during larval development, give rise to a 1,000-fold increase in number of cells and tissue size, and have proved useful for elucidating the molecular mechanisms underlying tumorigenic growth (11, 12). To induce CIN in these cells, we have selected a group of genes whose depletion leads to CIN and analyzed its impact in highly proliferative epithelial cells of Drosophila. Mutations in genes involved in the spindle assembly checkpoint [bub3 and rough deal (rod)], spindle assembly [abnormal spindle (asp)], chromatin condensation (orc2), and cytokinesis [diaphanous (dia)] have been shown to recapitulate the genomic defects most frequently associated with human cancer, including chromosome rearrangements and aneuploidy (7). Cytokinesis defects lead to polyploidy, which takes place in a substantial fraction of human tumors and has been proposed to constitute an important step in the development of cancer aneuploidy (13). We drove expression of dsRNA forms of these genes in the larval primordia of the adult wing and eye structures. To assess the extent of genetically induced CIN, we first quantified loss of heterozygosity (LOH) of two different genetic markers in adult tissues (see SI Materials and Methods for details). The first assay is based on haploinsufficiency of the Minute genes (14), which encode for ribosomal protein genes. Because the 65 Minute loci are spread throughout the euchromatic genome, reduced copy number of most large genomic regions is likely to result in the Minute phenotype (visualized as thin adult bristles, Fig. 1A). The second assay was generated in a heterozygous background for a recessive allele of the yellow gene, which is located on the X chromosome. LOH can be scored by analyzing yellow bristles in adult tissues (Fig. 1B). In flies expressing dsRNA forms for asp, rod, bub3, and orc2, we observed a significant increase in yellow and/or Minute LOH (Fig. 1 A and B). Because these assays rely on the ability of cells to divide, we were not able to quantify LOH in cells expressing a dsRNA form for dia. LOH was also analyzed in wing primordia and generated in a heterozygous background for two different transgenes located on different chromosomes (UAS-GFP transgene located on the second chromosome and UAS-RFP transgene located on the third chromosome, Fig. S1). Expression of these transgenes was driven under the control of the en-gal4 driver and monitored in normal conditions or upon coexpression of a dsRNA form of rod (Fig. S1). LOH can be scored by the loss of GFP or RFP expression within the en-gal4 domain. In wing pirmordia expressing a dsRNA forms for rod, we observed a significant amount of UAS-GFP and/or UAS-RFP LOH (Fig. S1). Altogether these observations indicate that expression of these dsRNA forms was able to induce CIN in our model system (see also below).

Fig. 1.

Dp53-independent CIN-induced apoptosis in epithelial cells. (A and B) Loss of heterozygocity analyses of the Minute (A) and yellow (B) genes. Adult thorax with Minute thin bristles (A, red arrow) and magnification of an anterior adult wing margin with yellow bristles (B, red arrows) upon expression of asp^RNAi (A) or rod^RNAi (B) under the control of the ci-gal4 driver, and histogram plotting the frequencies of Minute (A) and yellow (B) bristles of flies expressing GFP^RNAi, asp^RNAi, rod^RNAi, bub3^RNAi, or orc2^RNAi under the control of the ci-gal4 driver. ci-gal4 drives expression in anterior (A) cells in the wing and in the whole adult thorax. Genotypes: (A) +/+; ci-gal4 (II)/+; UAS-dsRNA (III) and (B) yellow/+ (I); ci-gal4 (II)/+; UAS-dsRNA (III)/+. Only females were scored. (C–J) Larval wing primordia, adult wings, and eyes from individuals expressing GFP (C and D) and the indicated transgenes (D–J) under the control of the en-gal4 (wings) or ey-gal4 (eyes) drivers, and stained for GFP (green), DAPI (blue), and TUNEL staining (red). The wing primordium shown in I is also mutant for dp53. X-Y sections of the basal side of the wing epithelia are shown. H′ shows a magnification of the squared region in H. H′′ shows a cross-section of the posterior compartment. Note in H′′ that TUNEL positive cells are located on the basal side of the epithelium. Vertical line in E–J depicts the boundary between anterior (A) and posterior (P) cells. (K–O′) Larval wing primordia from individuals expressing the indicated transgenes and GFP under the control of the en-gal4 (K and M–O′) or ci-gal4 (L) drivers, and stained for GFP (green), DAPI (blue), and MMP1 (red in K–M) or TUNEL staining (red in N–O′). X-Y sections of the basal side of the wing epithelia are shown. O shows a magnification of the squared region in N. O′ shows a cross-section of the posterior compartment. Note in O′ that TUNEL negative cells are located in the basal side of the epithelium. en-gal4 drives expression in posterior (P) cells and is depicted in green in the adult wings. ci-gal4 drives expression in anterior (A) cells. ap, apical; bs, basal.

CIN induced gross abnormalities in differentiation and tissue growth in adult wings and eyes (Fig. 1 C and D and Fig. S1) and a strong increase in the number of apoptotic cells in the larval primordia (Fig. 1 D–H). Apoptotic cells were located on the basal side of the wing epithelium (Fig. 1H′′). Mammalian cells use the ATM/p53 pathway to remove, through programmed cell death (PCD), aneuploid cells and preserve the structure and function of their genome following CIN (9, 15, 16). Surprisingly, the number of apoptotic cells observed in larval tissues subject to CIN was not significantly reduced in Drosophila p53 (dp53) mutant tissues or upon depletion of Dp53 activity (Fig. 1 I and J and Fig. S1). The c-Jun N-terminal kinase (JNK) pathway is involved in various stress responses in Drosophila tissues including DNA damage (14, 17). Interestingly, ectopic JNK activation, monitored by the expression of matrix metaloproteinase 1 (MMP1), a direct transcriptional target of dFos downstream of JNK signaling (18), was observed upon genetically induced CIN (Fig. 1 K and L and Fig. S1). MMP1 expression was completely blocked upon expression of a dominant negative version of JNK (basket in Drosophila, Fig. 1M) and the number of apoptotic cells was largely, but not completely rescued (Fig. 1 N–O′). Interestingly, we observed a large number of cells located on the basal side of the epithelium, which were nonapoptotic (compare Fig. 1 O and O′ with H and H′). Altogether, these results indicate that genetically induced CIN induces, in contrast to mammalian cells, a p53-independent apoptotic response in Drosophila tissues, most probably to reduce the number of aneuploid cells and maintain genomic stability in the cell population. Although JNK appears to largely contribute to this apoptotic response, other stress response pathways might be involved as well.

Fig. 2.

CIN-induced tumor growth and tissue invasiveness upon additional blockade of programmed cell death. (A–F and I–K) Larval wing primordia from wild-type individuals (A) or from individuals expressing the indicated transgenes and GFP in posterior (P) cells under the control of the en-gal4 (wings) driver and stained for GFP (green or white, A and F), P35 (B–E), DAPI (blue or white, A and I–K), and Ci (red, I and J) protein expression. Ci labels anterior (A) cells in I and J. Wing primordia shown in F and K are also mutant for dronc (F) or heterozygous for Df(H99) (K). (G) Histogram plotting the P/A size ratio of wing primordia expressing p35 and dsRNA for the indicated genes under the control of the en-gal4 driver. Error bars represent SEM. *P < 0.001. Larvae were 4 d old in A–G and 8 d old in I–K. (H) DNA content analysis by fluorescence associated cell sorter (FACS) of wing cells expressing the indicated transgenes under the control of the ap-gal4 driver. Percentage of cells with DNA content higher than 4n is indicated. (L) Adult fly micrographs taken 20 d after implantation (a.i.) of GFP-labeled larval wing tissue expressing the indicated transgenes. Ratios showing the reproducibility of the phenotype are shown. (M) Larval wing tissue 0 and 20 d a.i. (N–P) GFP (green) and P35 (blue) positive cells are observed inside the host ovary stained for DAPI (gray) and actin (red). O is a cross-section of N along the white line, and P is a magnification of N.

CIN-Induced Neoplasic Tissue Growth and Host Invasiveness upon Additional Blockage of PCD.

To maintain aneuploid cells in the tissue subject to CIN, apoptotic cell death was blocked or reduced by three different means. Apoptosis was first, and most generally, blocked by means of expression of the baculovirus protein p35, known to bind and repress the effector caspases DrICE and Dcp-1 (19). The DNA content profile of dissociated cells subject to CIN and expressing p35 revealed a high percentage of cells with DNA content higher than 4n (up to 41%) compared with control p35-expressing cells (Fig. 2H and Fig. S2). In the case of cells subject to CIN but not expressing p35, this percentage was very similar to control-expressing tissues (Fig. S2), reinforcing the role of apoptosis in eliminating aneuploid cells from the tissue. Apoptosis was also reduced in the whole larvae by halving the dose of the proapoptotic genes hid, grim, and reaper (in Df(H99)/+ animals) or blocked by mutation in the initiator caspase Dronc (in dronc^L29 animals) or in clones of cells homozygous for the H99 deficiency. Interestingly, in all cases, a strong overgrowth was observed in the cell population subject to genetically induced CIN (labeled with p35 or GFP), compared with the neighboring wild-type tissue (not labeled) and with age-matched p35-expressing wing discs (Fig. 2 A–G). Similar results were obtained in adults eyes (Fig. S2). Tissues depleted of dia expression and expressing p35 did not grow, most probably due to defects in cytokinesis. Interestingly, larvae containing overgrown tissues kept growing for longer periods of time than GFP-expressing control larvae (Fig. S2), the resulting wing primordia were massively overgrown (Fig. 2 I–K) and the GFP cell population invaded the neighboring wild-type territory (labeled in red, Fig. 2 I and J). To further characterize the growth potential of the tissue subject to CIN and PCD blockade, wing tissue expressing p35, GFP, and the corresponding dsRNA transgenes was transplanted into the abdomen of adult females and maintained for a period of 20 d. Whereas p35-expressing tissue hardly grew after implantation, tissue subject to CIN and expressing p35 could grow many folds larger than the size of the implant (Fig. 2L) and showed disorganized tissue architecture with extensive folding (Fig. 2M and Fig. S2). In some cases, cells of the overgrown tissue were able to invade different organs of the host and micrometastasis of GFP-expressing cells could be found in the ovary of the adult host (Fig. 2 N–P).

CIN-Induced Loss of Apical–Basal Polarity and Junctional DE-Cadherin Levels.

To characterize the cellular mechanisms underlying CIN-induced tumorigenesis, we carefully examined the behavior of aneuploid cells within the larval primordia. The wing primordium is a cellular monolayer that forms a two-sided epithelial sac (Fig. 3A). One side of this sac forms a thin squamous sheet, whereas the apposed epithelial surface (the columnar epithelium) adopts a pseudostratified columnar morphology. The apical side of both epithelia is oriented toward the lumen of the sac, whereas the basal side is the external surface. Cells subject to genetically induced CIN delaminated from the main epithelium, lost their columnar shape, and acquired a rounded and irregular form (Fig. 3 B and C). Epithelial architecture relies on the polarization of the plasma membrane into apical and basolateral domains separated by the adherens junctions. We used antibodies against atypical protein kinase C (aPKC), Disk large (Dlg), and DE-cadherin (DE-cad) to visualize the subapical, basolateral, and junctional domains, respectively. Whereas aPKC and Dlg were still observed at the membrane of delaminated cells, the overall protein levels were severely reduced (Fig. 3 D, E, D′, and E′ and Fig. S3). Interestingly, DE-Cad lost its tight junctional localization and the overall protein levels were clearly reduced (Fig. 3 F and G and Fig. S3). It was no longer detectable in most of the delaminated cells with the exception of some small discrete puncta (Fig. 3 F′ and G′ and Fig. S3). Overexpression of a modified version of DE-cad (DE-CadΔCyt:α-catenin), able to rescue the adhesion properties of E-cad mutant cells and unable to interfere with endogenous β-catenin signaling (20), did not rescue the delamination process and the overexpressed form of DE-cad localized to small discrete puncta in delaminating cells (Fig. 3 H and H′). Thus, dissociation of Drosophila epithelial cells upon CIN is not due to reduced levels of DE-cad and is most probably a consequence of DE-cad mislocalization.

Fig. 3.

CIN-induced cell delamination and BM degradation. (A) Cartoon depicting the columnar epithelium (ce) of the wing primordium. (B–J) Cross-sections of the posterior compartment of wing primordia expressing the indicated transgenes under the control of the en-gal4 driver and stained for DAPI (blue), fasciclin III (red, B, C, and I), laminin-γ (labels the BM in green, B, C, and I), Dlg (red, D and E), aPKC (green, D and E), E-cad (red, F–H), and viking-GFP (collagen IV, J). ap, apical, and bs, basal. (C) Magnification of the squared regions shown in B. (D′–H′) X-Y sections of the apical and basal sides of the wing epithelia shown in D–H. Yellow arrows indicate delaminating cells. (K and K′) FACS sorted delaminated and nondelaminated cells from wing primordia expressing p35 and a dsRNA form for rod under the control of the en-gal4 driver were subject to DNA content profile analysis (K) and chromosome labeling with chromosome-specific probes (K′). In K, percentage of cells with DNA content higher than 4n is indicated. In K′, yellow arrows indicate aberrant number of chromosomes in delaminated cells. Ratios showing the number of nuclei presenting an aberrant number of chromosomes (more than two chromosomes 2 or 3 and loss of chromosome 4) are shown below each panel.

Strikingly, delaminated cells showed a high degree of aneuploidy compared with nondelaminated cells, as shown by their DNA content profile and aberrant number of chromosomes (Fig. 3 K and K′ and Fig. S2). A high percentage of cells with DNA content higher than 4n was observed in the delaminated cells (up to 35%) compared with the nondelaminated ones (Fig. 3K). These results support the proposal that aneuploidy triggers cell delamination.

CIN-Induced JNK Activation in Delaminating Cells.

Tumorigenesis is usually accompanied by disruption of the basement membrane (BM) and MMPs are well known to contribute to BM degradation both in flies and mammals (18, 21). BM was clearly disrupted in the cell population subject to CIN (labeled with laminin-γ or with a GFP-tagged protein trap in the viking/collagen IV gene, Figs. 3 B, I, and J and 4B) and ectopic expression of MMP1 was observed in the delaminating cell population (Figs. 3J and 4B and Fig. S4). Mmp1 is a direct target of dFos downstream of JNK activation (18), and consistently, expressing a dominant negative version of JNK (basket^DN) blocked the ectopic expression of MMP1 (Fig. 4E). Expression of puckered, a known target of JNK signaling in Drosophila tissues (22), was also observed in delaminating cells (Fig. S4). The fact that MMP1 and puckered expression were restricted to the delaminating cell population opens the possibility of a causal relationship between JNK activation and cell delamination. However, cell delamination and E-cadherin mislocalization was not rescued upon blocking JNK activity (Fig. 4 E–G). These results indicate that genetically induced CIN induces delamination of epithelial cells, JNK-dependent activation of MMP1 expression, and BM disruption.

Fig. 4.

CIN-induced JNK activation in delaminating cells. (A and C) Wing primordia expressing the indicated transgenes in posterior (P) cells under the control of the en-gal4 driver, and stained for MMP1 (red, A), Wingless (red, C), and Ci (green) protein expression. Ci labels anterior (A) cells. (B, D, and E–G) Cross-sections of the posterior compartment of wing primordia expressing the indicated transgenes under the control of the en-gal4 driver and stained in red for MMP1 (B and E), Wingless (D and F), or E-cadherin (G), in green for laminin-γ (B, E, and F), and in blue for DAPI. Yellow arrows indicate delaminating cells. (H) Wing primordium with clones of cells mutant for Df(H99) and expressing a dsRNA form of asp and GFP (green) and stained in red for Wingless and in blue for DAPI. (Right) Cross-section of the clone expressing Wingless. Note in this clone expression of Wg in delaminating cells (yellow arrow). (I) Scheme showing the genomic region of the wingless gene, the location of the BRV118 enhancer, and the deletion in the wingless¹ mutation. (J and K) Wing primordia expressing the indicated transgenes in the posterior (P) compartment under the control of the en-gal4 driver and stained in red for Wg, in green for BRV118-lacZ expression (J) or Ci (K), and in blue for DAPI (K). In K, Ci labels the anterior (A) compartment. Wing primordium shown in K is also mutant for wingless¹. Asterisks represent endogenous Wg expression. ap, apical; bs, basal.

The mitogenic signaling molecule Wingless (Wg) is induced in a JNK-dependent manner by a number of insults in Drosophila epithelia (23–25), opening the possibility that wg might also be a target of JNK activity in CIN-induced delaminating cells. Indeed, ectopic expression of Wg was observed in these cells (Fig. 4 C, D, and H and Fig. S4) and, interestingly, expression of MMP1 and Wg were also observed in transplanted tissues subject to CIN (Fig. S2). CIN-induced Wg expression was abolished upon expression of a dominant negative version of JNK/Basket (Fig. 4F). Remarkably, a wg enhancer previously shown to drive Wg expression upon tissue injury (Fig. 4I) (26) was activated in cells subject to genetically induced CIN (Fig. 4J) and in cells expressing an activated form of JNKK (hemipterous in Drosophila, Fig. S4). We next tested whether this enhancer was absolutely required to induce JNK-dependent expression of Wg upon CIN. Luckily, the first ever identified mutation in the wg gene (wg¹) carries a deletion including this enhancer (Fig. 4I) (26, 27), and wg¹ mutant tissue did not show ectopic expression of Wg upon CIN, and the tissue was less overgrown (Fig. 4K). We noticed that the endogenous expression of Wg was unaffected in the wg¹ mutant condition. These results support the notion that JNK activates wg expression in delaminating cells and that the JNK responsive enhancer is required to induce wg expression upon CIN. JNK activation in CIN-induced delaminating cells is most probably a consequence of the loss in the apical–basal polarity. Delaminating cells showed reduced aPKC protein levels, and genetic depletion of components of the Cdc42/Par6/aPKC polarity complex has been previously reported to induce JNK-dependent apoptosis and hyperplastic growth upon additional blockage of PCD (28). Similarly, CIN-induced delaminating cells showed reduced Dlg and junctional DE-cad protein levels, and genetic depletion of individual components of the Scribbled/Disk Large/Lgl polarity complex has been previously shown to induce a reduction in junctional DE-cad levels and activate the JNK pathway (18, 29, 30). JNK activation relies, in these experimental conditions, on increased endocytosis rates and on translocation of Eiger/TNF to endocytic vesicles thus leading to apoptotic JNK activation (31). Molecular mechanisms linking CIN and loss of apical–basal polarity remain to be elucidated.

Central Role of the Delaminating Cell Population in Inducing Tumorigenesis.

The JNK pathway has a critical role in the tumorigenic behavior of Drosophila epithelial cells (18, 29–31). So far we have provided evidence that cell delamination and JNK activation correlate with high levels of aneuploidy. These results open the possibility that aneuploidy-induced JNK activation in the delaminating cell population has a critical role in organizing the hyperplastic overgrowth of the tissue. Remarkably, blocking JNK activity rescued the tissue overgrowth caused by CIN (Fig. 5A) as well as the capacity to grow extensively in allograft transplantations (Fig. 5B). Similarly and consistent with a role of Wg in promoting proliferative growth (32), a dsRNA form of wg largely rescued the tissue overgrowth (Fig. 5C) and the capacity of the tissue to grow extensively in allograft transplantations (Fig. 5D). These results support the notion that the highly aneuploid delaminating cell population has a critical role in the tumorigenic behavior of the tissue upon CIN induction. Whereas tissue hyperplasia relies on the restricted expression of Wg in this cell population, expression of MMP1 in delaminating cells degrades the underlying basement membrane thus facilitating tissue invasiveness (Fig. 5E). Cells subject to a high dose of ionizing radiation, known to induce aneuploidies in Drosophila tissues (14), or cells subject to segmental aneuploidies and maintained in the tissue by additional blockade of the apoptotic pathway, also delaminated from the main epithelium and showed MMP1 and Wingless expression (Fig. S5). These data reinforce the proposed role of aneuploidy-induced delaminated cells in driving tumorigenesis. The molecular and cellular mechanisms underlying the tumor-like behavior induced by aneuploidy upon additional blockage of PCD resemble, in a striking fashion, those caused by the cooperative action of Ras oncogene activation and mutations in the scribbled or Discs Large tumor suppressor genes (11, 18, 29, 30). Whereas JNK is being primarily used to remove, through PCD, cells whose genomic stability or apical–basal polarity is compromised, an additional blockade of the apoptotic pathway unravels a subversive role of JNK in driving tumorigenesis. Remarkably, JNK activation is not sufficient to drive tumorigenesis in epithelial cells even upon additional blockage of PCD, and apical–basal polarity disruption appears to play a fundamental role in this process (30). The identification of the delaminating cell population as the one with a critical role in promoting JNK-dependent deregulated growth and host invasiveness upon aneuploidy induction is consistent with these observations. In cancer development, an epithelial-to-mesenchymal transition (EMT) converts epithelial cells into migratory and invasive cells and is established as an important step in the metastatic cascade of epithelial tumors. Whether aneuploidy-induced cell delamination shares the properties of a real EMT in molecular and cell behavioral terms remains to be elucidated. However, it is interesting to note that aneuploidy-induced delaminating cells are motile (Fig. 2C, note isolated GFP delaminated cells far from the GFP-expressing domain) and are most probably the ones capable of invading the host.

Fig. 5.

A central role of the delaminating cell population in inducing tumorigenesis. (A and C) Histograms plotting the P/A size ratio of wing primordia expressing, under the control of the en-gal4 driver, p35 alone (red bars), or p35 and dsRNA forms for asp, orc2, bub3, or rod (green and blue bars). Blue bars express also basket-DN (A) or a dsRNA form for wingless (C). Error bars represent SEM. *P < 0.001. (B and D) Adult fly micrographs taken 20 d after implantation (a.i.) of GFP-labeled larval wing tissue expressing the indicated transgenes. Ratios showing the reproducibility of the phenotype are shown. (E) Cartoon depicting CIN-induced tumorigenesis. CIN induces cell delamination of epithelial cells, and delaminating cells activate a JNK-dependent transcriptional response (dark blue nuclei) that leads to the expression of MMP1 (green) and Wingless (red) protein expression. MMP1 has an active role in degrading the basement membrane (purple), a critical step in tissue invasiveness, and Wingless contributes to the hyperplastic growth of the monolayered epithelium.

Discussion

Altogether, our results indicate that genetically induced CIN by different means leads to Drosophila p53 (dp53) independent apoptosis of proliferating epithelial cells, as opposed to mammalian cells (9, 15). Activation of the stress responsive JNK signaling cascade contributes to this apoptotic response and maintenance of the aneuploid cells in the tissue through additional blockage of PCD leads to tissue overgrowth and host invasiveness. This paper conducts a thorough analysis of the relationship between CIN and tumor-like growth, it provides an ideal genetic experimental setup to identify new molecular elements involved in CIN-induced tumorigenesis, and it reinforces the proposed role of programmed cell death in suppressing CIN-induced tumorigenesis.

Perhaps, the most accepted hypothesis on the role of CIN in tumorigenesis is the one that proposes that CIN is a source of mutability (e.g., loss or gain of certain chromosomes or chromosome regions carrying tumor suppressor genes or oncogenes) that helps the tumor cell population to pass through the critical steps such as cell delamination, extravasation, and invasiveness (10). Although the Drosophila tumor suppressor gene l(2)gl has been well documented to be prone to spontaneous deletions (33), overexpression of l(2)gl did not rescue the CIN-induced tumorigenic behavior of the tissue (Fig. S3). Amplification or deletion of other genomic regions might contribute to cell delamination and to the tumorigenic behavior of the tissue. In this context, we would like to speculate that resistance to apoptosis can also be a direct consequence of CIN (e.g., LOH of proapoptotic genes), thus contributing to the survival of aneuploid cells and to cancer development.

The identification of the delaminating cell population as the critical one in driving tumorigenesis adds another layer of complexity to our current understanding of CIN in human cancer. Whereas aneuploidy-induced delaminating cells become motile and degrade the basement membrane, supporting a cell-autonomous role of CIN in driving tissue invasiveness (10), delaminating cells play a nonautonomous role in promoting tumor growth. Wingless expression is restricted to the delaminating cell population and promotes hyperplastic growth of the nondelaminating cells. Interestingly, the nondelaminating cell population subject to CIN grows to a bigger extent than the neighboring wild-type territories (Fig. 2 I and J). Given the absolute requirement of Wingless in tumor growth (Fig. 5 C and D), these results support the proposal that low levels of aneuploidy in the nondelaminating cells might make these cells more responsive to the mitogenic action of Wingless. Whether this behavior is a consequence of LOH of genes involved in Wingless signaling requires further investigation. In addition, these observations uncover a self-reinforcement mechanism that might have implications in tumor initiation and progression in human cancer. Wingless induces hyperproliferation of signal-receiving (nondelaminating) cells, which is expected to increase aneuploidy levels and delamination rates and should lead, consequently, to a net increase in the number of signal-sending (delaminating) cells.

Besides the proposed role of CIN as a source of mutability, our results indicate that CIN also induces a general and rapid response of the tissue in terms of compromised apical–basal polarity, cell delamination, JNK activation, and basement membrane degradation, a prerequisite for tissue invasiveness. This quick response is very unlikely due to the sole accumulation of CIN-dependent oncogenic mutations and unravels, again, a subversive role of the stress-associated JNK pathway in driving tumorigenesis in Drosophila epithelia (11, 18, 29, 30). We support the proposal, based on work in yeast and human cells, that stress, most probably as a result of aneuploidy-induced protein stoichiometry imbalances (34, 35), contributes to this rapid response. Further investigation is required to address whether this is the case.

Stimulation of cell death signaling by different insults and additional depletion of the activity of the apoptotic effector caspases has been previously shown to result in the production of the so-called undead cells, which express mitogenic signals that promote cell proliferation in the surrounding tissue (23, 25). Activation of the apical cell death caspase Dronc has been reported to be necessary and sufficient to drive this proliferation through a feedback mechanism that involves the activity of dp53 and the proapoptotic genes hid and reaper (Fig. S6). Interestingly, the tumorigenic response of the tissue upon CIN and additional blockage of PCD is independent of the activity of Dronc (Fig. 2F and Fig. S6), Dp53 (Fig. S6), and the proapoptotic genes hid and reaper (Fig. 2K and Fig. S6). Moreover, cells depleted for the proapoptotic genes hid, grim, and reaper delaminated from the main epithelium and showed Wingless expression (Fig. 4H), indicating that CIN-induced cell delamination and JNK activation is not a simple consequence of apoptosis induction. Although we cannot rule out a contribution of the dp53- and Dronc-dependent feedback loop in enhancing the tumorigenic response of the tissue upon CIN, our results support a rather specific oncogenic role of genomic instability in Drosophila epithelial cells that cannot be explained by the sole production of stress-induced undead cells.

Materials and Methods

Drosophila strains were obtained from Vienna Drosophila RNAi or the Bloomington Stock Center and are described in FlyBase. Antibodies were obtained from the Developmental Studies Hybridoma Bank or other private sources. These and other experimental details are described in SI Materials and Methods.