Selective elimination of cancer cells by the adenovirus E4orf4 protein in a Drosophila cancer model: a new paradigm for cancer therapy

The adenovirus (Ad) E4orf4 protein contributes to efficient progression of virus infection. When expressed alone E4orf4 induces p53- and caspase-independent cell-death, which is more effective in cancer cells than in normal cells in tissue culture. Cancer selectivity of E4orf4-induced cell-death may result from interference with various regulatory pathways that cancer cells are more dependent on, including DNA damage signaling and proliferation control. E4orf4 signaling is conserved in several organisms, including yeast, Drosophila, and mammalian cells, indicating that E4orf4-induced cell-death can be investigated in these model organisms. The Drosophila genetic model system has contributed significantly to the study of cancer and to identification of novel cancer therapeutics. Here, we used the fly model to investigate the ability of E4orf4 to eliminate cancer tissues in a whole organism with minimal damage to normal tissues. We show that E4orf4 dramatically inhibited tumorigenesis and rescued survival of flies carrying a variety of tumors, including highly aggressive and metastatic tumors in the fly brain and eye discs. Moreover, E4orf4 rescued the morphology of adult eyes containing scrib− cancer clones even when expressed at a much later stage than scrib elimination. The E4orf4 partner protein phosphatase 2A (PP2A) was required for inhibition of tumorigenesis by E4orf4 in the system described here, whereas another E4orf4 partner, Src kinase, provided only minimal contribution to this process. Our results suggest that E4orf4 is an effective anticancer agent and reveal a promising potential for E4orf4-based cancer treatments.


Introduction
The adenovirus E4orf4 protein is a multifunctional viral regulator, which contributes to regulation of the progression of viral infection 1,2 and to inhibition of the cellular DNA damage response (DDR), thus increasing the efficiency of adenovirus replication 3,4 . When expressed alone E4orf4 induces a p53-and caspase-independent mode of cell-death, which can lead to classical caspasedependent apoptosis in some cell lines [5][6][7][8] . E4orf4 celldeath signaling is conserved in several organisms, including yeast, Drosophila, and mammalian cells 9-12 . Our studies in normal Drosophila tissues demonstrated that E4orf4 induced both caspase-dependent and -independent cell-death in the fly, but also inhibited classical apoptosis, thereby causing minimal tissue damage and a marginal effect on fly survival 12 . Studies in mammalian cells revealed that E4orf4-induced cell-death was more efficient in oncogene-transformed cells than in normal cells 13 , indicating that investigation of E4orf4 signaling may have practical implications for cancer therapy. The cancer selectivity of E4orf4-induced cell-death may result from a combination of several E4orf4 activities that interfere with various pathways of cell regulation 1,2 .
Several E4orf4 cellular partners that contribute to E4orf4-induced cell-death have been described, including protein phosphatase 2A (PP2A) and Src kinases 2 . E4orf4 binds the heterotrimeric PP2A holoenzyme through direct association with its regulatory B55 subunit 14,15 .
The model organism Drosophila melanogaster has contributed considerably to the study of cancer and to identification of novel cancer therapeutics [16][17][18][19][20][21] . Various mutations are known to cause tumorigenesis in Drosophila including those affecting tumor suppressor genes that are required for normal cell polarity and asymmetric cell divisions, such as scribble (scrib) 22 . Human homologs of these genes are associated with the formation of diverse types of cancers 23,24 . When scrib mutant cells are surrounded by similar cells they develop tumors, but when surrounded by normal cells they do not 25 , unless supplemented by constitutive activation of the Ras pathway, which confers a proliferation advantage and leads to formation of aggressive, metastatic tumors 25,26 . Activated Ras (Ras V12 ) also cooperates with activation of the phosphoinositide 3-kinase (PI3K) pathway to generate more aggressive tumors 27 .
Here, we show that E4orf4 expression dramatically inhibited the development of various aggressive tumors without greatly affecting healthy tissues, and could restore normal tissue morphology even when expressed at a late stage. The role of E4orf4 partners in this process is also described.

Results
E4orf4 counteracts Ras V12 -induced impairment of tissue differentiation and fly survival in a dose-dependent manner The consequences of Ras V12 expression in the presence or absence of E4orf4 were initially examined in eye imaginal discs. Ras V12 was expressed together with either E4orf4 or a control GFP gene using the UAS-GAL4 system driven by the eyeless (ey) promoter, which drives expression in the eye disc, as well as in regions of the brain 28,29 . GAL4-driven expression rises with increasing temperatures, facilitating investigation of dose-dependent effects. As seen in Fig. 1a, the expression of Ras V12 together with GFP at 24 o C resulted in loss of normal eye disc morphology and a significant increase in disc size indicating induction of uncontrolled proliferation and differentiation defects. The differentiation defects were also visualized by the unorganized staining of the neuronal differentiation marker ELAV. The expression of E4orf4 counteracted the Ras V12 -induced over-proliferation of cells and the loss of normal differentiation as evident by the smaller size and normal shape of the E4orf4 Ras V12 expressing discs.
In the experiments described in Fig. 1a, we noticed a reduced proportion of adult eclosion upon Ras V12 expression at high temperatures, possibly resulting from ey-driven Ras V12 effects in the brain. We therefore examined the ability of E4orf4 to rescue the Ras V12induced lethality at various temperatures (Fig. 1b). Two different UAS-E4orf4 fly strains were utilized, one expressing higher E4orf4 levels than the other when driven by ey-GAL4 (Fig. 1c). In a WT background (ey > GFP), even high levels of E4orf4 ("strong") had an insignificant effect on adult fly viability at 18 o C and 24 o C (p > 0.1) and reduced viability by 13% at most at 29 o C (p = 0.013). These results are consistent with our previous findings showing that expression of E4orf4 in normal fly tissues caused only mild effects 12 . In contrast, strong E4orf4 expression rescued the 100% lethality induced by ey > Ras V12 at 24 o C and 29 o C to 89% and 24% viability, respectively (p < 0.012). Weaker E4orf4 expression led to rescue of 50% viability at 24 o C (p < 0.0001), but could not rescue the Ras V12 -induced lethality at 29 o C. These findings confirm that the dose-dependence of E4orf4 activity, which was previously reported in WT flies 12 is also sustained in a Ras V12 background, and reveal that E4orf4 counteracts Ras V12 -induced lethality.
E4orf4 inhibits the formation of highly aggressive tumors in the fly The preliminary experiments described in Fig. 1 investigated Ras V12 and E4orf4 effects occurring when these genes were expressed broadly in the developing eye. However, since tumor growth is significantly influenced by the microenvironment, additional investigation of E4orf4 effects on tumorigenesis was conducted in clonally induced tumors. For this purpose, the mosaic analysis with a repressible cell marker (MARCM) method was utilized as described previously 26 to induce mitotic GFPlabeled clones in the eye discs and brain. The overexpression of Ras V12 induces non-invasive tumors in flies 26,30 . Since we wished to examine the ability of E4orf4 to eliminate various types of tumors, including highly aggressive, metastatic tumors, we analyzed mitotic clones carrying genetic alterations including Ras V12 activation and/or inactivation of the tumor suppressor scribbled (scrib), a regulator of cell polarity 23 . The clones were produced by ey-driven FLP recombinase, which mediated recombination between chromosome arms, leading to homozygous loss of scrib and to loss of the UAS-GAL4 system inhibitor Gal80, facilitating GAL4-driven expression of UAS-regulated genes within the clones 26 . Flies generated using this system carried one of four types of mitotic clones, including WT clones, scrib − clones, clones expressing Ras V12 , and scrib − clones that expressed Ras V12 . Each of these clones also expressed either the strong E4orf4 transgene or a control RFP gene. As shown in Fig. 2a, eye discs and brains of 3rd instar larvae containing WT or scrib − clones were similar in size and morphology, and E4orf4 expression did not greatly affect organ size, although it reduced clone sizes in the scrib − background. Brains and eye discs of some larvae with Ras V12 -overexpressing clones were slightly enlarged, but the alterations were not significant in comparison with larvae carrying Ras V12 , E4orf4-coexpressing clones. There were also no dramatic changes in the overall organ morphology. In contrast, Ras V12 -overexpression in combination with scrib deletion resulted in massive growth of brain and eye discs and loss of recognizable morphology, creating a huge tumor mass. Furthermore, whereas the GFP-marked mitotic clones with other genotypes were restricted to parts of the eye discs and to limited areas in the brain hemispheres, the Ras V12 , scrib − mitotic clones colonized the whole eye disc and brain area and invaded the ventral nerve cord (VNC), as described previously 26,30 . E4orf4 expression had a remarkable effect on these highly aggressive tumors, resulting in complete restoration of normal size and morphology of both eye discs and brains, as well as nearly complete elimination of GFP-marked clones, reducing their size to that of WT clones.
Next, we determined how generation of the mitotic clones, with or without E4orf4, affected fly viability at 24 o C. As seen in Fig. 2b, 100% of flies with WT mitotic clones were viable, and E4orf4 expression reduced viability in the WT background by only 3%. The generation of scrib − homozygous clones reduced fly viability by 12% in the absence of E4orf4, but the flies were 96.5% viable when E4orf4 was expressed in these clones. Clonal Ras V12 overexpression led to a complete loss of fly viability in both WT and scrib − backgrounds. However, E4orf4 expression could rescue 97% of flies with Ras V12 -expressing clones and 80% of flies with Ras V12 , scrib − mitotic clones.
The eye phenotypes of adult flies that emerged from this experiment demonstrated an improvement in eye morphology in flies with scrib − mitotic clones expressing E4orf4, and similar eye morphologies were evident in flies expressing E4orf4 in Ras V12 and Ras V12 , scrib − mitotic clones (Fig. 2c).
Altogether, the results shown in Fig. 2 indicate that E4orf4 is highly potent in preventing the development of even very aggressive tumors, and consequently reduces the ensuing lethality and eye deformities caused by the tumors.
As E4orf4 was reported to induce cell-death in flies and mammalian cells 12,13 , we examined whether cancer clone elimination could potentially result from E4orf4-induced cell-death manifested by activation of the Dcp1 caspase. Dispersed caspase activation, which was not concentrated near the strongest GFP-marked clones, was evident in control WT clones, as well as in Ras V12 , scrib − clones (Fig.  3a, c). Dcp1 activation was observed around strong E4orf4expressing clones in the WT background (Fig. 3b), but the staining was somewhat dispersed in comparison to a robust accumulation of Dcp1 staining around Ras V12 , scrib − clones with strong E4orf4 expression (Fig. 3d). It should be noted that most of the E4orf4-expressing Ras V12 , scrib − clones have already been eliminated by this stage (Fig. 2a) and therefore it is likely that only remnants of the E4orf4induced cell-death are still evident. Nonetheless, the results suggest that cell-death could explain, at least in part, the elimination of the large Ras V12 , scrib − tumor clones.
In addition to the highly tumorigenic collaboration between activated Ras V12 and deletion of cell polarityregulating genes, a collaboration between Ras V12 and activation of the insulin-PI3K pathway induces significantly more tumorigenesis than Ras V12 alone, possibly due to increased metabolism that supports cancer cell proliferation 27,31 . Activation of the insulin-PI3K pathway in Drosophila can be achieved by overexpression of Chico, the fly insulin receptor substrate, or by reduced expression of PTEN, the PI3K pathway inhibitor. To determine how E4orf4 affects PI3K-regulated tumorigenesis, E4orf4 or a control RFP transgene were expressed under ey-GAL4 regulation together with various combinations of (see figure on previous page) Fig. 1 E4orf4 counteracts Ras V12 -induced impairment of tissue differentiation and fly survival in a dose-dependent manner. a Eye-antennal imaginal discs expressing Ras V12 under the regulation of ey-GAL4 together with GFP (ey-GAL4 > UAS-Ras V12 ; FRT82B, UAS-GFP) or E4orf4 (ey-GAL4 > UAS-Ras V12 ; FRT82B, UAS-E4orf4-17.22) are shown. Discs were dissected from age-matched 3rd-instar larvae grown at 18 o C or 24 o C. The discs were stained with antibodies to E4orf4 (blue) and ELAV (red) and were analyzed by confocal microscopy. All images were taken at the same magnification (x40) and represent projections of multiple sections. A representative eye disc is shown in each picture. The 50 μm scale bar applies to all the pictures. It should be noted that age-matched larvae were utilized, but E4orf4-expressing larvae were consistently slower to differentiate. b Percentage of adult eclosion ("Live") and of non-eclosed pupae ("Dead") was determined in flies expressing GFP (labeled as WT) or Ras V12 under the regulation of ey-GAL4 together with another copy of GFP (ey-GAL4 > UAS-GFP;FRT82B, UAS-GFP, and ey-GAL4 > UAS-Ras V12 ;FRT82B, UAS-GFP) or E4orf4. Two different fly strains harboring an E4orf4 transgene were tested. These strains express the viral protein to different levels: "strong": UAS-E4orf4-17.22 (ey-GAL4 > UAS-GFP; FRT82B,UAS-E4orf4-17.22 and ey-GAL4 > UAS-Ras V12 ;FRT82B,UAS-E4orf4-17.22) and "weak": UAS-E4orf4-attP2, also called "weak1" in c below (ey-GAL4 > UAS-GFP;UAS-E4orf4-attP2 and ey-GAL4 > UAS-Ras V12 ;UAS-E4orf4-attP2). The flies were grown at the indicated temperatures. N > 300, collected from three independent experiments. c Protein extracts were prepared from ten 3rd instar larvae expressing "strong" E4orf4 and two types of "weak" E4orf4 under the regulation of ey-GAL4 (ey-GAL4 > UAS-E4orf4-17.22, ey-GAL4 > UAS-E4orf4-attP2 (weak#1), and ey-GAL4 > UAS-E4orf4-attP40 (weak #2)). A representative western blot is shown, stained with antibodies to E4orf4 and to Tubulin, which served as a loading control Ras V12 , chico, and PTEN interfering RNA (IR), and fly viability was assayed. To detect differences between the effects of Ras V12 alone and Ras V12 supplemented by activation of PI3K, we counted separately empty pupal cases from which adult flies have eclosed, pupae containing pharate adults that failed to eclose, and pupae in which no metamorphosis took place. Figure 4a demonstrates that chico overexpression with or without concomitant expression of E4orf4 did not reduce significantly fly viability. Co-expression of Ras V12 with two copies of the RFP transgene caused 100% lethality at 24 o C, whereas Ras V12 , E4orf4 co-expression facilitated adult eclosion from 80% of the pupae. Co-expression of chico with Ras V12 aggravated the effect of Ras V12 , increasing the percent of non-metamorphosing pupae from 83% to 94% (p = 0.005). However, even in this background, E4orf4 expression dramatically increased adult eclosion up to 74%.
Similar results were obtained when activation of the PI3K pathway was induced by knockdown of PTEN (PTEN-IR, Fig. 4b). Whereas PTEN-IR alone reduced fly viability to 76%, co-expression of E4orf4 rescued viability to 94% (p = 0.01). The concomitant expression of PTEN-IR aggravated the effect of Ras V12 , increasing the percentage of non-metamorphosing pupae from 83% to 100% (p = 0.002). Here too, co-expression of E4orf4 resulted in a dramatic increase in adult eclosion to 85% in this background, very similar to a rescue to 88% eclosion in flies expressing ey > Ras V12 without PTEN-IR.
To investigate the effect of E4orf4 on tumor tissues, adult eyes were examined. As shown in Fig. 4c, the phenotype of Ras V12 , E4orf4-expressing eyes was milder than that of eyes expressing Ras V12 alone, which were visualized in 0.6% adults that have eclosed. The ability of E4orf4 to rescue the eye phenotype of Ras V12 , chico expressing flies could not be evaluated due to pupallethality of the Ras V12 , chico genotype. As previously reported, E4orf4 expression caused a minor rough eye phenotype in a WT or chico backgrounds.
The results summarized in Fig. 4 confirm that E4orf4 is a potent inhibitor of tumorigenesis.

E4orf4 partners that contribute to cancer elimination by E4orf4
We previously reported that two E4orf4 partners, PP2A and Src, contributed in an additive manner to the mild Eye-antennal discs were dissected from these larvae and analyzed by confocal microscopy. GFP-expressing clones were visualized by GFP fluorescence and clones expressing E4orf4 were stained with E4orf4-specific antibodies (blue). Active (cleaved) Dcp1 staining is shown in red. All images represent projections of multiple sections. A representative eye disc is shown in each picture with a 50 μm scale bar. An enlarged inset of E4orf4-expressing Ras V12 , scrib − clones surrounded by concentrated active Dcp1 staining is displayed in d E4orf4-induced cell-death in normal Drosophila tissues 12 . We now tested whether these partners contributed similarly to the E4orf4-induced elimination of cancer tissues. The flies used in this experiment carried a UAS-Ras V12 transgene on the 2nd chromosome and an additional transgene integrated in the attP2 site on the 3rd chromosome containing a UAS-driven cDNA of RFP, WT E4orf4, or E4orf4 mutants that cannot bind Src ("Src − ") or PP2A ("PP2A − "), or both ("Src − , PP2A − ") 12 . Integration of all transgenes at the same chromosomal site minimizes position effects on their expression. Flies of these genotypes were crossed to ey > GAL4 flies at 24 o C. As shown in Fig. 5a, ey-GAL4-driven Ras V12 expression in the presence of a control RFP gene reduced fly viability and only 2% of the flies reached adulthood. Despite the relatively low expression levels of WT E4orf4 integrated at the attP2 site (Fig. 1c), E4orf4 was still able to increase viability of ey > Ras V12 -expressing flies to 60%. The Src − E4orf4 mutant that is unable to bind Src kinase was only slightly less efficient than WT E4orf4 in rescuing the viability of ey > Ras V12 -expressing flies, leading to 50% adult eclosion. In contrast, the PP2A − E4orf4 mutant, which could not bind PP2A, lost completely the ability to counteract Ras V12 activity, and fly viability remained 0.7%. Although the WT E4orf4 protein accumulated at higher levels than the E4orf4 mutants, the dissimilarity in function of the two mutants cannot be explained by differences in protein levels, as the levels of the PP2A − mutant were consistently higher than the levels of the Src − mutant (Fig. 5b). The double mutant, PP2A − Src − , which lost the ability to bind both Src and PP2A was also unable to increase fly viability, resulting in 0% adult eclosion.
In addition to examining the ability of E4orf4 mutants to rescue the viability of ey > Ras V12 -expressing flies, we   -attP2). b Protein extracts were prepared from ten 3rd instar larvae expressing RFP, WT E4orf4, or the E4orf4 mutants described in a, using the ey driver at 24 o C. A representative western blot is shown, stained with antibodies to E4orf4 and to Tubulin, which served as a loading control. c Eye-antennal imaginal discs obtained from crosses as described in a were dissected from age-matched 3rd-instar larvae grown at 24 o C. The discs were stained with antibodies to ELAV and were analyzed by confocal microscopy. All images were taken at the same magnification (x40) and represent projections of multiple sections. A representative eye disc is shown in each picture with a 50 μm scale bar. . It should be noted that even though flies with the tws mutation did not express a control GFP transgene to match the E4orf4 transgene whereas the other flies contained this control, potential competition for the GAL4 driver by UAS-E4orf4 did not lead to rescue of eye morphology in the mutant background evaluated their ability to rescue eye disc differentiation. Eye discs of 3rd instar larvae that expressed Ras V12 together with RFP or E4orf4 under ey > GAL4 regulation were stained with anti-ELAV antibodies. As seen before (Fig. 1a), expression of Ras V12 in the absence of E4orf4 resulted in loss of normal differentiation marked by irregular ELAV staining (Fig. 5c). WT E4orf4 expression (the "weak1" transgene) enhanced normal differentiation, as did expression of the Src − E4orf4 mutant. In contrast, the PP2A − E4orf4 mutant was much less efficient in recovering normal ommatidia patterns, further confirming its reduced ability to counteract Ras V12 effects.
To validate the contribution of PP2A to the ability of E4orf4 to counteract Ras V12 activities, the influence of E4orf4 on the Ras V12 induced eye phenotype was compared in WT eyes and in eyes lacking tws, the fly PP2A-B55 subunit that mediates the E4orf4-PP2A interaction 13,14 . Figure 5d demonstrates that E4orf4 caused a minor rough eye phenotype in both WT and tws backgrounds, possibly through Src binding. Ras V12 expression at 24 o C was lethal and no adult progeny eclosed. Concomitant expression of E4orf4 rescued fly viability and produced eyes with some defects, which retained regular differentiation in large areas. In contrast, the tws mutation reduced Ras V12 -induced lethality and adult eyes with a typical "raspberry-like" appearance were observed. However, E4orf4 expression did not improve eye morphology in this background. To overcome the Ras V12 -induced lethality, flies were additionally reared at 18 o C. At this temperature, ey > Ras V12 flies could eclose and they presented rough, overproliferating eye tissues and some small growths, whereas E4orf4 expression reverted eye morphology to near normal. In the tws background, the ey > Ras V12 eyes remained abnormal upon E4orf4 expression.
The findings shown in Fig. 5 demonstrate that the interaction with PP2A is crucial for the ability of E4orf4 to rescue flies from Ras V12 -induced tumorigenesis and lethality, but surprisingly, the interaction of E4orf4 with Src has only a minor contribution to counteracting Ras V12 -induced lethality. This differs from the situation in normal fly tissues, in which PP2A and Src contribute additively to E4orf4 function 12 .
E4orf4 rescues effects of early scrib elimination even when expressed at a late stage In the previous experiments, the concomitant expression of Ras V12 and E4orf4 was induced by the ey driver (Figs. 1, 4, and 5), or was induced following ey > FLP activation together with scrib deletion (Fig. 2). Thus, the results of these experiments show that E4orf4 inhibits tumorigenesis when expressed simultaneously with cancer induction. To examine the consequences of late E4orf4 expression, scrib − GFP-marked homozygous clones were induced under the regulation of ey > FLP, which is expressed in the eye disc as early as the end of the 1st instar larval stage 33 . In contrast, the expression of E4orf4 or a GFP control gene was driven by GMR-GAL4, which is activated in the eye disc only at the 3rd instar larval stage 34 . As seen in Fig. 6a, the eyes of flies with mitotic clones lacking scrib and expressing the control GFP gene were abnormal, with small growths and necroses. However, GMR > E4orf4 expression resulted in a significant suppression of this adult eye phenotype. We have previously shown that, in WT background, GMR > E4orf4 on its own induced only a mild rough eye phenotype 12 .
To evaluate the size and state of differentiation of scribdeficient clones, the eye discs of 3rd instar larvae containing GFP-marked mitotic clones were co-stained for ELAV and E4orf4. As seen in Fig. 6b, GFP-marked scribdeficient clones were relatively large and ELAV staining in these clones was irregular compared with WT clones, indicating abnormal differentiation. In contrast, scribdeficient clones that expressed E4orf4 were smaller and the pattern of ELAV staining resembled more closely the staining in WT clones, although E4orf4 had some effect on photoreceptor differentiation as the spacing between the ommatidia appeared increased. It should be noted that because the expression of E4orf4 and GFP was driven by GMR-GAL4, it was seen only in the ey > FLP-generated clones, in which the GAL4 system inhibitor Gal80 was removed by recombination. Figure 6c demonstrates that whereas E4orf4 reduced the size of WT clones to 72%, it reduced the size of scrib-deficient clones to 43% compared to the GFP-expressing control. Statistical analysis (Summary Independent-Samples Test) further demonstrated that the decrease in cancer clone sizes induced by E4orf4 was significantly larger than the E4orf4-induced decrease in WT clone sizes (p = 0.012). These findings indicate that E4orf4 can specifically eliminate scrib − clones and rescue adult eye morphology even when expressed at a much later stage than scrib elimination, and that its toxic effects on WT cells are much milder than its toxic effects in a cancer tissue.

E4orf4 effects in various types of tumors
Previous work in tissue culture revealed that E4orf4induced cell-death was more efficient in mammalian cancer cells than in normal cells 13 . By utilizing the power of fly genetics, the present work demonstrates that E4orf4 can also strongly suppress the tumorigenic process in a whole organism with minimal damage to normal tissues.
Various types of tumors were used in this study: Hyperplastic Ras V12 -induced tumors, which overproliferate and cause imaginal disc overgrowth without metastasizing; Neoplastic scrib − tumors, which lose the monolayered organization and correct apical-basal polarity of wild-type epithelia 35 and are deficient in normal differentiation as well as prone to metastasis; and highly aggressive and metastatic Ras V12 , scrib − tumors. In addition, we examined the effects of E4orf4 on tumors induced by combined activation of the Ras and insulin-PI3K pathways. The tumor types investigated here are highly relevant to human tumors. Activation of Ras signaling occurs in~30% of human cancers 36 and deregulation of scrib, as well as of many other cell polarityregulating genes contributes to many types of human tumors 37 . Disruption of insulin-PI3K signaling is also quite common in human cancers 38 .  E4orf4). The 50 μm scale bar in the bottom right applies to both pictures. b Eye-antennal discs were dissected from 3rd-instar larvae grown at 24 o C and analyzed by confocal microscopy. Flies with scrib − clones were as described in a. Flies with WT clones were generated by a cross between yw,eyFLP;19A,UAS-med8GFP2A;FRT82B,tubGAL80/TM3 and GMR-GAL4; FRT82B,UAS-GFP (WT, GFP) or GMR-GAL4; FRT82B, UAS-E4orf4-17.22 (WT, E4orf4). All images were taken at the same magnification (x40) and represent projections of multiple sections. WT and scrib − clones, as indicated in the figure, are marked by GFP fluorescence (green) or stained with antibodies to E4orf4 (blue) (left column). The discs were also stained with ELAV-specific antibodies (red)(middle column) to visualize their state of differentiation. A merge of all stains is seen in the right column. A representative eye disc is shown in each picture. The 50 μm scale bar in the bottom right applies to all the pictures. c The sizes of GFP-marked clones from eye-antennal discs with WT or scrib − mutant backgrounds were detected by GFP fluorescence (present also in E4orf4-expressing clones, but not shown in b) and quantified using ImageJ software. Average pixel numbers are shown. All the discs used here were age-matched. N > 8 from > 3 independent experiments, error bars represent the standard deviation. Statistical significance was determined by a student t-test. *p < 0.05, NS: not significant This work shows that E4orf4 was efficient in inhibiting tumorigenesis in representatives of all types of tumor tissues described above. Especially exciting was the ability of E4orf4 to prevent formation of the massive metastatic tumors induced by the combination of Ras V12 overexpression and scrib deletion (Fig. 2). E4orf4 succeeded in preventing both overgrowth and the ability of cancer cells to totally colonize the tissue and invade the VNC. The mild effects of E4orf4 on normal tissues were reaffirmed.
E4orf4 induction of cell-death is responsible, at least in part, for its anti-tumorigenic effect. Although technical difficulties prevented the visualization of E4orf4-induced cell-death in early 2nd instar larvae shortly after initiation of E4orf4 expression in the highly aggressive Ras V12 , scrib − tumors, Dcp1 activation was evident around clones with strong E4orf4 expression in 3rd instar larvae (Fig. 3). This observation suggests that induction of cell-death played a role in the elimination of tumorous clones. E4orf4 is also known to kill normal cells, but to a lesser degree 12 , and thus the elimination of aggressive Ras V12 , scrib − cancer clones as well as of scrib − clones was more efficient than removal of WT clones (Figs. 2a and 6c). Effects on cell proliferation could also potentially contribute to cancer elimination by E4orf4 but we found no evidence supporting this possibility.
The ability of E4orf4 to revert the effects of scrib deletion when expressed at a later stage (Fig. 6) is also interesting. scrib − clones were shown to be eliminated by their WT neighbors and by JNK-dependent apoptosis, causing eye deformities 30,39,40 . It is therefore possible that E4orf4, which inhibits classical apoptosis in normal cells 12 , could also inhibit apoptosis in the scrib − clones, thus rescuing normal eye morphology. It is also possible, however, that E4orf4 reverted the neoplastic identity of these clones through other mechanisms.

The contribution of E4orf4 partners
Two E4orf4 partners, PP2A and Src kinase contribute additively to E4orf4-induced cell-death both in mammalian transformed cell lines 32 and in normal Drosophila tissues 12 . The major surprising finding here was that PP2A binding was crucial for the ability of E4orf4 to inhibit Ras carcinogenesis in vivo, whereas Src binding played a marginal role (Fig. 5). Several types of PP2A holoenzymes act as tumor suppressors, and mutations in the scaffolding A subunit, as well as in some regulatory B subunits, were found in various types of human cancer 41 . Moreover, PP2A activation by small molecules was shown to inhibit growth of some tumors 42,43 . In contrast, in some cellular contexts the PP2A B55 subunit was reported to promote tumorigenesis 44 and certain PP2A inhibitors are tested as chemotherapeutics 43 . Thus, PP2A activity must be carefully balanced in the cellular context. The mechanisms by which PP2A supports E4orf4-induced cell-death are still debated. Although one report suggested that inhibition of PP2A activity contributed to this process 45 , several findings indicate that the E4orf4-PP2A complex possesses PP2A phosphatase activity 14 and that PP2A contributes positively to E4orf4-induced celldeath 13,15 . Furthermore, PP2A is recruited by E4orf4 to several substrates involved in its functions, including induction of cell-death 2,46 . Thus, recruitment of PP2A to novel substrates or stabilization of existing interactions may shift the balance in several E4orf4-targeted cellular pathways. Cancer cells may be more sensitive to these changes than normal cells, leading to cell-death. As one example, E4orf4 inhibits some branches of the DDR in a PP2A-dependent manner 3 . Many cancer cells have deficiencies in DNA damage signaling and may rely more heavily on their remaining intact DDR pathways. Targeting the remaining pathways by a DDR inhibitor such as E4orf4 may therefore be selectively toxic to these cancer cells.

The potential of E4orf4 for cancer therapy
This work demonstrated the remarkable activity of E4orf4 as an anticancer agent in vivo, providing a strong incentive for development of E4orf4-based cancer treatments. The potential advantages of E4orf4-based cancer therapy include the following: (1) E4orf4 kills several types of tumor cell lines but not normal cells 6,8,13,47,48 . (2) E4orf4 induces p53-independent cell-death 8,49 , which could be effective in p53-deficient tumors. (3) E4orf4 induces caspase-independent cell-death 5 , which could be useful in cancers with deficiencies in classical apoptosis. (4) As E4orf4 induces a unique mode of programmed celldeath by targeting a combination of several cellular pathways 2,3 , it may provide alternative solutions in cases of therapy-resistant disease. Therefore, design of E4orf4based cancer therapies would be highly beneficial. Cell polarity genes as well as the PI3K pathway are associated with the formation of diverse types of cancer 23,24,38 , and providing appropriate therapy would be beneficial. Moreover, despite more than three decades of research, no anti-Ras therapies have become available 50 . Therefore, finding effective treatment of cancers with Ras activation 36 would in itself be a satisfactory result of using E4orf4-based cancer therapy.

Adult eclosion test
The number of empty pupal cases from which adult flies have eclosed and of pupae that died before or after metamorphosis were counted in each vial for up to 3 weeks after pupa formation. Eclosed adults were removed daily. The ratio between empty pupal cases and the total number of pupae was calculated.

Phenotypic characterization of adult eyes
Adult flies were anesthetized with CO2. Equal numbers of female eyes were examined for each genotype (n > 20 unless otherwise stated). The eyes were photographed using AxioCam MRc and Zeiss Discovery V8 binocular. To represent the depth of field in full, a series of images with different foci were taken, and the images were stacked using the Helicon Focus software.

Clone size quantification
The size of GFP-positive clones was measured in pixels using the ImageJ software (NIH). Briefly, images of all samples were captured using the same magnification and acquisition conditions in a confocal microscope. A Zstack was formed from all eye-antennal discs and a maximum intensity projection was created to include the entire depth of the discs. Eye disc size was quantified by increasing the brightness/contrast level marking the disc margins and measuring its area. The area of GFP-marked clones with intensities above a defined threshold were measured and summarized.

Data analysis
The statistical significance of differences in eclosion rates or clone sizes between the various fly groups were calculated by an unpaired t-test. Summary Independent-Samples Test was used to assess whether the differences between E4orf4-induced changes in WT and scrib − clone sizes were significant. This statistical analysis was carried out using the SPSS software package (Release 23.0.0.0, SPSS Inc., 2014).