Differing Src Signaling Levels Have Distinct Outcomes in Drosophila

Src is the first kinase linked to cancer and one of the original identified oncogenes (1). It is a membrane-associated kinase that regulates a broad palette of developmental and homeostatic processes, including cell proliferation, apoptosis, and adhesion. Src family kinases (SFK) are activated in most major cancer types, including breast cancer, colon cancer, melanomas, etc. (2–6).

Nevertheless, the role of SFKs in oncogenesis is not fully understood. Cell culture studies have provided important but conflicting data: for example, expression of v-Src or c-Src (viral or cellular Src, respectively) can induce differentiation, proliferation, apoptosis, or transformation depending on the cell line used (7, 8). Results such as these suggest that SFKs can act differently in different cellular contexts. Also, although v-Src is strongly oncogenic, c-Src protein is only weakly transforming, and Src activity is linked to tumorigenesis primarily through its increase late in a tumor's maturation (9). Mammals contain nine SFKs, and at least some act redundantly. c-src knock-out mice are viable and fertile, exhibiting minor bone defects (10). Redundancy is indicated by the observation that Src^−/−;Fyn^−/−;Yes^−/− triple mutant embryos die early in development (11).

v-Src's derives its strong activity in part due to a lack of the negative regulatory COOH-terminal domain, which contains a conserved tyrosine targeted for phosphorylation by COOH-terminal Src kinase (Csk) and its homologue Chk (12). Csk and Chk are cytoplasmic kinases that represent the major negative regulators of SFK activity, and the Csk/Src regulatory circuit is conserved from the early-diverging metazoan Hydra to humans (13). Knock-out of Csk activity in mice leads to strong hyperactivation of SFKs and early embryonic lethality (14). Csk and Chk have been linked to breast and brain cancer, presumably through the activation of Src (e.g., ref. 15).

Recent work has begun to characterize the sole Drosophila Csk/Chk orthologue dCsk. Reduction of dCsk activity led to increased cell proliferation, increased organ size, and organismal lethality (16, 17). By contrast, overexpression of the Drosophila Src orthologues dSrc42A or dSrc64B led to a small rough eye phenotype more reminiscent of proapoptotic genes than a reduction in dCsk activity (18–20). v-Src can induce apoptosis of cultured cells under some circumstances (21, 22), indicating that this outcome of Src overexpression can also be observed in mammalian cell lines.

Here, we use an in situ approach in Drosophila to examine more closely the nature of Src's ability to direct an overgrowth phenotype within the context of simple polarized epithelia. We present evidence supporting a model that reconciles the different phenotypes resulting from dSrc overexpression versus dCsk reduction. Our data indicate that Src signaling is biphasic: lower levels of Src pathway activity result in antiapoptotic signals, whereas strong levels potently induce apoptosis. However, when coupled with other oncogenes such as Ras, high levels of Src activity can lead to malignant overgrowth and invasion. Src activity has been reported to increase progressively in most common solid tumor types (23), and these results suggest that the role of Src changes as tumorigenesis progresses.

Fly lines and genetics. GMR-hid, UAS-dSrc42A.CA, and UAS-dSrc64B were obtained from the Bloomington Drosophila Stock Center. The following stocks were kindly provided to us: hs-dSrc64B, hs-dSrc64B^Δ540, and hs-dSrc64B^KD by N. Perrimon, sev-dSrc42^KD by K. Saigo, UAS-Ras^V12 and lats^wtsXI by G. Halder, scribble¹ by D. Bilder, and dSrc64B^D17 and dCsk^Q156Stop by A. O'Reilly and M. Simon.

Cultures and incubations were done at 25°C unless otherwise specified. The genetic suppression of GMR-hid by mildly expressed dSrc64B (hs-dSrc64B) was observed at 25°C (data not shown) and 29°C (Fig. 1).

To create whole eye clones of dCsk (EGUF, ref. 24), we generated flies with the genotype yw:ey-gal4,UAS-FLP/+;FRT82B,GMR-hid,l(3)CL-R/FRT82B,dCsk. To create eyeless-driven green fluorescent protein (GFP)–labeled clones, flies with the genotype yweyFLP1;act>y⁺>gal4,UAS-GFP;FRT82B,tub-gal80 or, alternatively, ey(3.5) FLP;act>y⁺>gal4,UAS-GFP;FRT82B,tub-gal80 were crossed with flies with the following genotypes: (a) UAS-Ras^V12;FRT82B,dCsk^j1D8/TM6b; (b) UAS-Ras^V12;FRT82B, dCsk^Q156Stop/TM6b; (c) UAS-dSrc64B,UAS-Ras^V12;FRT82B,dCsk^j1D8/TM6b; and (d) UAS-Ras^V12;FRT82B,scrib¹/TM6b.

Microscopy. Immunofluorescence was done as described previously (25). Antibodies used were rabbit anti-pY434 (activated dSrc64B, 1:100, ref. 26), mouse anti-armadillo N27A1 and rat anti–DE-cadherin (1:3 and 1:20, respectively, both from the Developmental Studies Hybridoma Bank), rabbit anti-cleaved caspase-3 (1:200, Cell Signaling Technology), and rat anti-laminin A (α-220, 1:1,000, ref. 27). Secondary antibodies were conjugated with red or green Alexa (Molecular Probes).

For scanning electron microscopy, adult flies were fixed in 95% ethanol, critical point dried, sputter coated, and viewed with a Hitachi S-2600H scanning electron microscope.

Eye size quantification. Light micrographs from adult eyes were analyzed using ImageJ software (NIH) to measure eye area in arbitrary pixel units. The eye sizes from each genotype (n = 6 in all cases) relative the control genotype were analyzed using Student's t test, and the level of statistical significance was displayed in the graphs. Error bars in the graphs correspond to SEs.

High Src signaling levels induce apoptosis whereas low levels inhibit apoptosis. Csk represents the major cytoplasmic negative regulator of SFKs, and mutations that reduce or eliminate Csk function should exhibit phenotypes that are similar to overexpressing SFKs. Strong overexpression of either of the two Drosophila SFK members, dSrc42A or dSrc64B, in the developing eye led to small rough eyes (e.g., Fig. 1A,, right). Each SFK gave rise to similar gain-of-function phenotypes, but Pedraza et al. (20) described phenotypes ranging from small rough eyes to complete eye ablation or even organism lethality due to disruption of the head capsule. This likely reflects different levels of expression achieved by different transgenic lines and culture temperatures. In contrast, others and we (16, 17) observed that mutations that reduce activity of the Drosophila Csk orthologue dCsk led to tissue overgrowth, a block in apoptotic cell death, and a resulting enlarged eye (e.g., Fig. 1A , center). Removal of a single genomic copy of the downstream Src target dSrc29 rescues this phenotype, indicating that overgrowth is due to hyperactivation of Src signaling (17). These results mirror studies in mammalian model systems indicating that Csk targets primarily SFKs. However, it leads to the question as to why overexpression of dSrc42A or dSrc64B does not phenocopy loss of dCsk function. This issue has important implications for disease, as dynamic changes in the activity of Csk and Src have each been observed in human tumors, but their precise roles in situ are unclear.

One explanation for this paradox has been suggested by mouse knock-out studies, which suggest that Csk may act independent of SFKs (28). In Drosophila, dCsk has been linked functionally to the cytoplasmic kinase Lats/Warts. Mutations in lats lead to an increase in organ size (29), a phenotype similar to dCsk. Based on its ability interact genetically with lats in vivo and to phosphorylate Lats protein in vitro, dCsk has been proposed to phosphorylate Lats in situ and in a Src-independent fashion (20). Such Src-independent effects of dCsk could explain the phenotypic differences between Src overexpression and loss of dCsk function. However, removing a genomic copy of lats also led to a significant suppression of mildly misexpressed dSrc64B^Δ540 (Supplementary Fig. S1), an activated version of dSrc64B that is insensitive to Csk activity. That is, lats shows a functional link with dSrc64B in addition to dCsk. Based on these functional relationships, we propose that most or all dCsk activity that is dependent on Lats also requires Src. These data argue against Src-independent links between dCsk and Lats.

Interestingly, weak overexpression of dSrc64B^Δ540 led to a mildly rough eye that was, upon closer examination, more reminiscent of phenotypes observed in dCsk mutants than those with strongly expressed Src isoforms. Pupal retinas that expressed dSrc64B^Δ540 through mild induction of the heat-inducible hsp70 promoter displayed defects that included extranumerary interommatidial precursor cells (IPCs; Fig. 2A). Ectopic IPCs are a hallmark of insufficient programmed cell death (PCD) during development (30). They are also a characteristic of reduced dCsk activity, which leads to a block in Hid-dependent apoptotic removal of excess IPCs (17, 31). Furthermore, weakly expressed dSrc64B suppressed mildly but significantly the ectopic cell-death phenotype resulting from the overexpression of the proapoptotic gene hid (Fig. 1B). This antiapoptotic activity observed with low ectopic levels of Src signaling stands in contrast to the proapoptotic signals observed with high levels of ectopic Src.

These observations suggest a model that would explain the phenotypic differences between (a) a modest increase in Src activity, obtained through mild Src overexpression or partial reduction of Csk levels and (b) strong Src misexpression. In this view, low versus high levels of Src signaling differ in the pathways they invoke, with low levels favoring stable overgrowth and high levels favoring apoptotic cell death. Next, we tested predictions of this biphasic Src activity model.

Complete loss of Csk function results in high levels of Src signaling. If dCsk is the major negative regulator of Src activity, then a full loss of dCsk activity should lead to strong Src activity and an apoptosis phenotype. A previously described dCsk RNA-interference transgenic line (dCsk-IR, ref. 32) is likely a hypomorph, accounting for its overgrowth phenotype. Previously characterized dCsk mutant alleles, dCsk^flick (16), dCsk^SO30003, dCsk^SO17909 and dCsk^j1D8 (17), and dCsk^c04256 (32), are all P transposable element insertions within introns or in the carboxyl-terminus end of the gene and are likely hypomorphs as well.

Recently, a null dCsk allele was generated; dCsk^Q156Stop homozygotes show increased dSrc64B activation and defects in cell packaging in oogenesis (26) and are embryonic lethal (data not shown). Although whole eye clones of hypomorphic dCsk alleles led to an enlarged eye (Fig. 1A; refs. 17, 32), attempts to obtain adult whole eye clones (EGUFs, ref. 24) of dCsk^Q156Stop resulted in pupal lethality, presumably due to a failure to properly form the head capsule. The few dCsk^Q156Stop animals that progressed to mature pupal stages had eye regions that were replaced by melanotic scars suggestive of extensive tissue death (Fig. 2B). This pupal lethality and loss of eye tissue phenocopied the strongest phenotypes observed by strong overexpression of dSrc isoforms (Fig. 2B and data not shown). Taken together, these data indicate that whereas partial loss of dCsk activity results in mild SFK activation and tissue overgrowth, more complete loss directs strong SFK activation and tissue death.

Kinase-inactivated Src activates signaling. A direct test on whether Csk signals primarily or exclusively through Src would be to block Src signaling in the context of a Csk loss of function. Mutation of a key lysine residue in the catalytic domain of Src abolishes kinase activity (33). Such kinase-dead versions of Src (Src^KD) have been extensively used as dominant-negative tools, with the assumption that they compete with the endogenous molecules for binding to upstream activators and downstream substrates. However, previous efforts at expressing these Src^KD isoforms yielded unexpected results, including phenotypic rescue of c-src mutations, leading to the view that the intrinsic kinase activity of Src is dispensable for a significant portion of its biological roles (34, 35).

If kinase-dead Drosophila Src isoforms act to inhibit Src pathway activity, then they should suppress the phenotypes of dCsk-IR. Contrary to our expectations, however, the eye phenotype of GMR>dCsk-IR was strongly enhanced by coexpressing kinase-inactivated forms of either dSrc42A or dSrc64B (Fig. 3A and data not shown). GMR>dCsk-IR;sev-dSrc42A^KD eyes were significantly smaller and mispatterned and displayed supernumerary, clustered bristles (Fig. 3A , inset, right). Together with altered cell proliferation/survival, this phenotype was also suggestive of cell fate changes and abnormal cell adhesion/migration, both previously associated with elevated Src signaling (17, 32).

We considered two possible explanations for the strong genetic enhancement observed between dSrc42A^KD and dCsk-IR. First, perhaps dSrc42A^KD has (kinase-independent) activity that is normally repressed by dCsk. Alternatively, kinase-dead Src isoforms may titrate away endogenous dCsk, releasing endogenous SFKs to increase pathway activity. Consistent with the latter model, sev-dSrc42A^KD by itself led to the ectopic IPC phenotype characteristic of mild Src activation (Fig. 2A,, right). Furthermore and similar to hs-dSrc64B and GMR>dCsk-IR (Fig. 3B; ref. 32), sev-dSrc42A^KD significantly suppressed the small eye phenotype resulting from hid misexpression (Fig. 3B). Finally, GMR>dCsk-IR;sev-dSrc42A^KD eyes were reduced in size in a manner similar to eyes strongly overexpressing Src (Fig. 3A , right).

These results are consistent with a role for dSrc42A^KD in activating Src signaling in vivo: activation presumably occurs by sequestering the remaining dCsk activity in GMR>dCsk-IR eyes, leading to a switch from low to high Src activity and concomitant apoptosis. Interestingly, further reducing the remaining dCsk activity in dCsk-IR eyes by enhancing RNA interference–mediated efficiency also yielded a small rough eye phenotype,⁴

Together, these experiments further support a model in which high levels of total Src activity lead to phenotypic outcomes that differ qualitatively from low Src levels.

Expression of dSrc42A can activate dSrc64B. Multiple potential mechanisms have been described to account for high Src activity in solid tumors: (a) COOH-terminal dephosphorylation by inactivation of Csk or its paralogue Chk (15) or by activation of the RPTP and PTP1B phosphatases (36); (b) loss of Src's COOH-terminal phosphorylation site through mutation (35); (c) direct activation by receptor tyrosine kinases such as epidermal growth factor receptor (EGFR) and HER2 (37); and/or (d) a strong increase in Src expression levels (35). Notably, although Src overexpression has been frequently observed in progressing tumors (9), cytoplasmic kinases are not commonly activated by simple overexpression. Csk is a common negative regulator shared by all SFKs, and our results with dSrc42A^KD further suggested that high expression levels of one SFK may activate another by titrating Csk activity.

We tested this possibility using antibodies that specifically detect activated dSrc64B (targeting pY434, ref. 26). In wild-type larvae, activated dSrc64B staining localized to the adherens junctions of epithelial cells from imaginal discs and included all cells of the eye anlage (Fig. 4), similar to reported localization of its paralogue dSrc42A (38). Such adherens junction–specific staining was abolished in dSrc64B mutant eye discs (Fig. 4), indicating that pY434 did not cross-react with any protein normally expressed in the eye. As expected, expression of ectopic dSrc64B with a sevenless promoter (sev-GAL4;UAS-dSrc64B) resulted in a strong increase in activated dScr64B staining that was specific to sevenless-expressing cells (Fig. 4; ref. 39). Strikingly, similar ectopic expression of activated dSrc42A also resulted in a modest but significant increase in the activation of dSrc64B (Fig. 4). To distinguish whether ectopic dSrc42A increased dSrc64B activity by direct phosphorylation, SFKs could undergo intermolecular autophosphorylation (40), or alternatively, by titrating away dCsk, we next expressed a kinase-dead version of dSrc42A. Expression of a kinase-dead dSrc42A isoform (19) also resulted in increased dSrc64B activation (Fig. 4). Taken together, these experiments indicate that overexpression of a SFK family member can activate other SFKs and presumably itself as well, by titration of negative regulators such as Csk.

High levels of Src signaling can cooperate oncogenically with Ras. Finally, we asked whether the importance of differing Src/Csk levels might prove relevant to models of oncogenesis. Oncogenic cooperation has been observed in most tumor paradigms: typically, two or more loci participate in aggressive tumors. For example, Src synergizes with EGFR to direct oncogenic growth in fibroblasts (5). In Drosophila, expression of the oncogenic Ras isoform dRas^V12 alone led to epithelial outgrowths described as “benign tumors,” but combining dRas^V12 with loss of Scribble (scrib) led to aggressive tumors that exhibited metastatic-like behavior (41, 42). Because c-jun-NH₂-kinase activation plays a role in this oncogenic cooperation (43, 44) and is also a common effector of dCsk-mediated phenotypes (17, 32), we asked whether dCsk/Src could cooperate oncogenically with dRas^V12.

Using the MARCM system (45), we created early eye clones of cells that expressed dRas^V12, together with mutations expected to drive differing levels of Src activation. Combining the hypomorphic allele dCsk^j1D8 together with dRas^V12 within the developing eye resulted in phenotypes similar to the ones described for dRas^V12 alone (41): most animals were viable with discrete outgrowths present in the heads of adult flies (Fig. 5B). Pairing dRas^V12 with high levels of Src activity led to a different outcome. When dRas^V12 was combined with (a) overexpression of dSrc64B plus the hypomorphic mutant dCsk^j1D8 or (b) the null allele dCsk^Q156Stop, we observed a wide palette of phenotypes that ranged from discrete outgrowths in the adult heads to larval or pupal lethality. In the most severely affected animals, strong malignant overgrowth was observed in the larva; this led to an extended larval life in which the animals continued to grow and delayed pupariation. These severely affected animals displayed an abnormally large body size, and the eye clones of cells with activated Ras and Src activities expanded dramatically, overtaking the entire cephalic complex and resulting eventually in lethality of the giant larvae (Fig. 5A and C). This developmental delay was never observed in dRas^V12;dCsk^j1D8 animals. Penetrance of each of these phenotypes is quantified in Fig. 5D.

These results indicate that (a) Src can cooperate with Ras to drive malignant overgrowth, and (b) this malignant overgrowth requires high rather than low levels of Src activation.

These experiments suggested two additional points of note. First, the strong phenotypes we observed were strikingly similar to those reported for dRas^V12 and scrib¹ oncogenic cooperation (41, 42), and further experiments were done to determine whether scrib and src act through similar mechanisms. Our results indicate that (a) dRas^V12;dCsk^Q156Stop cells displayed rounded morphology and reduced or delocalized DE-cadherin expression, suggestive of altered epithelial polarity (Fig. 6A); (b) dRas^V12;dCsk^Q156Stop cells were found occasionally embedded within the basal extracellular matrix (ECM; Fig. 6B); and (c) dRas^V12;dCsk^Q156Stop eye-antenna clones eventually extended actin-rich processes to invade distant regions of the brain (Supplementary Fig. S3). Together, these observations indicate that scribble and dCsk/Src cooperate oncogenically with Ras^V12 through similar mechanisms.

Second, in some experimental animals, we observed tumors distant from the main eye tumors primarily in the region of the gonads (e.g., Fig. 5A). This suggested that secondary tumors emerged by metastasis of cells from the primary eye tumors to distant sites as previously reported (41). However, our FLPase-mediated clones were generated with an eyeless promoter that could have weak expression in other tissues such as the optic lobes of the brain and the gonads (41, 46). Interestingly, we observed posterior tumors almost exclusively in males; closer examination indicated that early-stage tumors were localized to cells within the posterior testes (Supplementary Fig. S2). To clarify this issue, we used the more restricted eyeless-driven Flipase ey(3.5)FLP, which does not drive expression in the testes or brain.⁵

This study provides evidence that different levels of Src signaling lead to different biological outcomes. We speculate that Src activity plays two important but separable roles during tumor maturation: early low levels of Src contribute to tumor overgrowth, whereas later high levels of Src, coupled with other oncogenes such as Ras, lead to invasive migration. Our previous work emphasized the importance of Csk/Src-dependent signals present at tumor boundaries that can provoke metastatic-like behavior (32). In addition, our current data reconcile the contrasting phenotypes observed between partial reduction of Csk activity versus a strong increase in Src activity by demonstrating the importance of low versus high levels of Src pathway activation, respectively.

We also considered what precisely is being modeled with the use of altered Src isoforms, a common approach in the study of signaling pathways including Src. Unexpectedly, we observed that kinase-dead versions of the two Drosophila SFKs did not behave as expected for dominant-negative isoforms. Furthermore, ectopic expression of dSrc42A, including as a kinase-dead isoform, led to the activation of its sole paralogue, dSrc64B. Our data suggest an alternative explanation to paradoxical and controversial observations reported for vertebrate Src. The primary phenotype of src^−/− knock-out mice, osteopetrosis caused by defective osteoclasts (10), was rescued by introducing a kinase-dead Src isoform (Src^KD). In addition, Src^KD rescued the reduced levels of phosphorylated tyrosine observed in src^−/− osteoclasts (34). This led to the suggestion that the essential activities of Src are mediated exclusively through protein-protein interactions and not through kinase activity. Our work suggests an alternative model in which endogenous SFKs are ectopically activated in Src^KD-rescued osteoclasts through the titration of Csk activity; this ectopic activation can then compensate for the loss of any individual SFK. In this view, the ability of Src^KD to rescue Src still invokes kinase activity, albeit indirectly.

Our results indicate that high levels of Src signaling can show oncogenic cooperation with Ras to direct tumoral growth and organismal lethality. Similarly to Ras/Scrib, Ras/dCsk cells displayed overgrowth, loss of epithelial polarity, migration, and invasion into the ECM. Our data indicate that, although cells did not metastasize into distant tissues, they were capable of invading nearby tissues within the cephalic complex such as the brain. In mammals, the ability to invade a local blood or lymph vessel could provide a malignant cell the ability to reach and colonize most tissues of the organism. Therefore, current Drosophila models could model some (albeit not all) fundamental aspects of the metastatic process.

Increasing levels of Src activation correlate with advancing tumorigenic stages in a variety of human cancers, including lung, breast, pancreatic, ovarian, and colorectal cancer (e.g., ref. 23). Strikingly, we observed a similar correlation with Src-driven oncogenesis in Drosophila: high (but not low) levels of Src signaling cooperated with oncogenic Ras to produce invasive overgrowth and organismal lethality. One question in the cancer field is whether ectopic expression studies truly model aspects of human tumors. Our results indicate that they can; however, attention must be paid to levels of activity and the specific aspects of tumorigenesis that each level emulates. Emerging models for tumorigenesis should consider not only the activation of multiple and specific pathways, but should also calibrate these levels to explore, e.g., benign overgrowth versus metastasis versus cell death.

Although Src activity can affect cell proliferation and survival (refs. 16, 17, this work), a growing consensus in the field is that a major role for Src is to promote metastasis by stimulating migratory behavior in transformed cells (reviewed in ref. 35). Our study suggests that the proliferative and survival signals from low Src activation contributes to tumor growth in early stages. In line with this, a recent study shows that mice with keratinocyte-restricted deletion of Csk have mild SFK activation and develop epidermal hyperplasia, but without malignant transformation (47).

In advanced stages, increased Src signaling, together with the acquisition of strong antiapoptotic signals that protect cells from Src-induced apoptosis such as oncogenic Ras, may reveal the ability of Src activation to drive invasive migration and, potentially, metastasis. For example, we note that pancreatic ductal adenocarcinoma involves the invasion of exocrine cells through intrapancreatic nerves, leading to severe damage and pain, promoting cancer spread, and precluding resection. Src, expressed at high levels in the majority of these tumors, is likely required for tumor progression (48, 49), and cell culture studies suggest that invasion requires Ras signaling for pancreatic tumor targeting (50). Finally, Src-specific kinase inhibitors have entered clinical trials (35). Our results with dominant-negative Src isoforms suggest caution, however, as the potential of a Src:drug complex to titrate Csk/Chk activity may yield unpredictable results in the context of Src-overexpressing tumors.

Grant support: NIH grants R01CA109730 and R01CA84309.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank A. O'Reilly, M. Simon, M. Uhlirova, D. Bohmann, G. Halder, D. Montell, N. Perrimon, T. Miura, and D. Van Vactor for generously providing reagents that were critical for this study. We also thank current and former Cagan lab members for their advice and support and M. Encinas and G. Halder for helpful discussions.