Drosophila screening model for metastasis: Semaphorin 5c is required for l(2)gl cancer phenotype

Abstract

Cancer metastasis is a complex process involving many genes and pathways. This complexity hinders the identification of molecules functionally required for this process. We have developed and used a Drosophila screening system to identify genes that are functionally important for tumorigenicity and metastasis. Deletion of Drosophila lethal giant larvae (l(2)gl) leads to highly invasive and widely metastatic tumors on transplantation into adult flies. Random homozygous P element insertions were screened for the ability to modulate the l(2)gl phenotype. Analysis of metastasis patterns of the lines containing P element insertions and lacking wild-type l(2)gl expression identified three homozygous mutations that dramatically alter tumorigenesis and/or metastasis. Semaphorin 5c (Sema 5c) is required for tumorigenicity, apontic overexpression suppresses metastasis but not tumorigenicity, and pointed up-regulation accelerates lethality of l(2)gl tumors. Furthermore, class 5 semaphorins are shown to be expressed in cancer cells and localized to the membrane. Drosophila Sema-5c and the mammalian homologs are transmembrane proteins with extracellular thrombospondin type I (TspI) repeats. TspI repeats are known in some proteins to bind and activate transforming growth factor (TGF)-β ligand. Phospho-Mad and the downstream target gene vestigial were elevated in l(2)gl tumors, thus linking Drosophila neoplasia to the Dpp (TGF-β-like) signal pathway. The activation of the Dpp pathway in l(2)gl tumors occurred only in the presence of Sema-5c. This study demonstrates that the power of Drosophila genetics can be applied to screen, identify, and characterize molecules that are functionally required for invasion and metastasis.

Metastasis is a complex process involving many molecules that function within the tumor cell and at the tumor–host interface. Our aim was to identify genes that were not essential for viability of the larva or adult but were required for the metastasis phenotype of the l(2)gl mutants. Inactivation of the Drosophila lethal giant larvae (l(2)gl) gene causes neoplastic growth of the brain and imaginal discs and lethality at the late larval/pupal stage (1). l(2)gl neoplastic tissues grow and metastasize on transplantation (2). l(2)gl tumor cells invade locally, and widely metastasize to distant organs. The tumor growth rate, metastasis patterns, and host lethality curves of l(2)gl tumors are highly reproducible and consistent in the hosts.

The l(2)gl tumor model combined with P element mutagenesis allows for screening of large numbers of mutations. The Lgl protein is expressed in the cytoplasm and at regions of cell junctions on the inner face of the cell membrane (3). The protein is required, along with the tumor suppressors discs large, and scribble (4) for basal protein targeting (5) and asymmetrical divisions of neuroblasts (6). Lgl is present in a high molecular weight protein complex, consisting primarily of Lgl homooligomers and the non-muscle myosin heavy chain (7). Homologs of l(2)gl exist in other species, including mouse (8) and human (9) and are also associated with non-muscle myosin (9), so it is likely that the role of Lgl in maintaining cytoskeletal architecture is conserved. Homologs of genes that control metastasis in Drosophila may play a similar role in higher organisms. Based on the metastatic phenotype that has been observed in tissues lacking Lgl expression, these cells exhibit many of the activities shared by mammalian metastatic cells such as proliferation as primary and secondary tumors, degradation of extracellular matrix, and motility.

The present study used the l(2)gl tumor phenotype as a functional readout to develop a discovery model for genes that are required for the metastatic process. Homozygous mutations caused by random P element insertions were generated in an l(2)gl background, and those that altered the malignant phenotype were identified. By using this model, three genes involved in metastasis were identified: the apontic gene on chromosome 2, pointed on chromosome 3, and semaphorin 5c on chromosome 3. A mutation affecting Semaphorin 5c expression blocked l(2)gl tumor growth. We examined the structural domains of this protein to determine possible functional roles of Semaphorin 5c, focusing on the potential role of the extracellular thrombospondin repeats in signaling pathways. Semaphorin 5c, a transmembrane protein of previously unknown function, was examined further by using human cell models and tumor tissue.

Materials and Methods

Drosophila Stocks. Flies were reared in shell vials on standard cornmeal, molasses, and yeast medium at 20°C. Second chromosome lethal mutations were maintained over balancers marked with y⁺ and CyO mutations in stocks that were homozygous for the y mutation on the X chromosome.

Metastasis Screen. A PlacW P element inserted on the X chromosome was randomly mobilized as described (10) in a heterozygous lethal giant larvae (lgl⁴) background. Autosomal insertions were mapped by standard genetic methods by crossing with a yw/yw;+/+;+/+ stock. Insertions were mapped to the third chromosome if segregation of the w⁺ eye color was independent of the segregation of the Cy marker. Insertions were mapped to the second chromosome containing the l(2)gl deletion if w⁺ segregated away from the Cy marker or mapped to the CyO balancer if w⁺ segregated with the Cy marker.

Transplantation of LacZ Marked Brain and Imaginal Disk Tissue. Brain lobes of armadillo-lacZ marked larvae were dissected and cut into halves. Each fragment was injected into the abdomen of a βgalⁿ¹ adult female by using a 33-gauge needle and cultured for 21 days at room temperature. Hosts were opened along the ventral midline and fixed in 3.7% formaldehyde in PBS, and β-galactosidase activity was detected by staining overnight at 37°C in 0.02% X-gal in 10 mM Na pyrophosphate, 0.15M NaCl, 1.0 mM MgCl₂, 5 mM ferricyanide, and 5 mM K ferrocyanide (Specialty Media, Phillipsburg, NJ).

Plasmid Rescue and RT-PCR. Genomic DNA flanking each of the P element insertions was isolated by plasmid rescue. An EcoRI genomic fragment was isolated from lines 97-2 and 115-1, and an SstI genomic fragment was isolated from line 23-2. The fragments were ligated, and phenol-chloroform extracted. One shot TOP 10 (Invitrogen) cells were transformed with the ligation mix. DNA was extracted from individual colonies and analyzed by restriction mapping by using the second polylinker sites (BamHI for lines 97-2 and 115-1 and PstI for line 23-2). Random hexamer-based reverse transcription was performed with Dnase (GenHunter, Nashville, TN) treated (2 h at 37°C) total RNA from third instar larvae. PCR conditions were: 1 cycle 94°C for 5 min, 35 cycles of 45 s at 94°C, 45 s at 58°C, 45 s 72°C, 1 cycle at 72°C for 10 min.

Western Blotting and Immunostaining. Larval brain extracts were prepared by homogenization in RIPA buffer (0.15 mM NaCl/0.05 mM Tris·HCl, pH 7.2/1% Triton X-100/1% sodium deoxycholate/0.1% SDS) containing 500 μM AEBSF hydrochloride, 150 mM aprotinin, 1 μM E-64, 0.5 mM EDTA disodium, 1 μM leuptin hemisulfate. Tris-glycine SDS sample buffer (2×) (Novex) with 4% β-mercaptoethanol was added, and extracts were boiled 5 min. Anti-peptide antibodies were generated and affinity purified against the sequences SVRIGLPKEESRN (Semaphorin 5c), CSSRRSAPPGSREPRG (SEMA5D), and KEIGPWLREFRAKNAVDC (SEMA5A). Western blot signals were detected by ECL (Amersham Pharmacia).

Membrane Preparation. A2058 cells were lysed in 10 mM Hepes, 1 mM EDTA, 250 mM sucrose, 1× protease inhibitors (Calbiochem), 1× phosphatase inhibitors I and II (Sigma). Lysate was passed through a 20-gauge needle 10 times, centrifuged, and incubated 10 min on ice. Lysate was vortexed and centrifuged at 1,000 × g for 10 min at 4°C. The pellet was washed with additional lysis buffer, vortexed, and centrifuged at 1,000 × g 10 min at 4°C. Supernatants were pooled and centrifuged at 3,000 × g for 10 min at 4°C. SDS protein sample buffer was added to sample and boiled 5 min at 95°C.

Reverse Phase Protein Microarray Analysis. Serial dilution of larval brain lysates was prepared, and a total of 50 nl (5 nl applied in a series of 10 separate applications) of lysate was arrayed with a “pin and ring” GMS 417 microarrayer (Affymetrix) by using a 500-μm pin onto nitrocellulose slides with a glass backing (Schleicher and Schuell). Staining was performed with a DAKO Immunostainer automated slide stainer by using the Catalyzed Signal Amplification (CSA) system (Dako) as described (11). Antibodies were as follows: actin 1:250 (Oncogene), phosphatidylinositol 3-kinase (PI3-K) 1:100 (Cell Signaling, Beverly, MA), T-ERK 1:500 (Cell Signaling), P-ERK 1:1000 (Cell Signaling), c-caspase 3 1:500 (Cell Signaling), SMAD1 1:100 (Santa Cruz Biotechnology), and P-SMAD1 1:250 (Cell Signaling). Specificity of each antibody was validated by detecting a single band by Western blotting.

Immunohistochemistry. Tissue microarray slides (TARP3, Tissue Array Facility, National Cancer Institute) were stained by using a DAKO Immunostainer automated slide stainer and the Envision (DAKO) system. Antigen retrieval was performed with proteinase K (DAKO) treatment. SEMA5D antibody was diluted 1:100.

Drug Treatment. Adult βgalⁿ¹ hosts transplanted with armadillo-lacZ-marked l(2)gl brain fragments were treated with 0, 0.556, 5.56, and 55.6 μg/ml PI3-K inhibitor LY294002 (Sigma) by adding drug to fly media. Flies were cultured for 21 days on drug-containing food and stained for the presence of β-galactosidase. Primary tumor size was determined by counting the cells dissociated from tumors.

Results

Generation of Homozygous Mutations That Disrupt Metastasis. The P-lacW (10) P element was mobilized in l(2)gl heterozygous males. For metastasis pattern analysis, homozygous l(2)gl larvae were isolated from the P element lines with two copies of the P element insertion (Fig. 1). In some cases, homozygosity of the P element alone or in combination with the l(2)gl deletion mutation caused embryonic or early larval lethality. We focused on those mutations that were homozygous viable until at least late third instar so that the l(2)gl tumor phenotype could be analyzed. Over 124,000 flies heterozygous for l(2)gl were screened for transposition of a single P element originally on the X chromosome. The mini-white gene was used as a marker to follow inheritance of the P element via eye color. The lacZ marker contained on the P-lacW element allowed for expression analysis by enhancer trapping (12). Nine hundred and eighty-six P element insertion events were recovered. Some of these males were eliminated because they seemed to be derived from non-independent transposition events. Eight hundred and eighty-six of these males were mated to yw;l(2)gl/Balancer females to establish independent lines. One hundred and eighty-five of the males were sterile, and forty-three of the lines lost the insertion or were weakly viable. Nine lines contained multiple P element inserts whereas four lines lost the l(2)gl lethal phenotype. The remaining lines were crossed to yw/yw females to determine the chromosome containing the P element. One hundred and twenty-two insertions were located on the y + CyO second chromosome balancer and were eliminated because they could not be analyzed in l(2)gl/l(2)gl homozygotes. Seventy-three of the insertions were on the Δ2–3transposase source third chromosome. These lines were eliminated because the insertions could not be stabilized in the presence of transposase. The remaining 450 lines were examined for double mutant viability to late third instar. We found 76 lines in which larvae homozygous for the P element insertion and lacking both copies of l(2)gl were viable until late third instar, the stage when the l(2)gl phenotype is apparent. These lines were examined for lacZ staining to identify those in which the affected genes contained lacZ expression in the larval brain. We found 16 lines that were l(2)gl/l(2)gl viable and contained lacZ expression in the brain. These lines were examined for effects on l(2)gl metastasis patterns. Larval brain lobes were cut into two fragments, and each piece was injected into the abdomen of a βgalⁿ¹ adult female. When l(2)gl brain fragments were injected into adult hosts, the injected tissue proliferated as a primary tumor and invaded adjacent tissue, and cells migrated away from the primary tumor to generate widespread metastatic colonies (Fig. 2 A and B). As described for this phenotype (2), metastatic colonies are found in the abdomen (57%), thorax (70%), head (39%), wing (35%), and leg (48%). We screened the P elements affecting genes expressed in the brain for those that disrupted the metastatic phenotype.

Fig. 1.

Functional screen for metastasis genes. P element screen to identify insertions that block l(2)gl tumorigenesis and metastasis. Adults homozygous for a P element and heterozygous for l(2)gl deletion are crossed to generate larvae that are homozygous l(2)gl^– and homozygous for the P element insertion. Brain tissue from these larvae is transplanted into adults.

Fig. 2.

Metastasis patterns of l(2)gl, insertion, and excision lines. P element insertion disrupts both copies of the affected gene. Excision of the P element reverses the disruption. Tumorigenic and metastatic cells were visualized by lacZ staining after 21 days. (A) l(2)gl/l(2)gl. Primary tumor (T), metastasis (M). (B) Tissue section, invasive l(2)gl/l(2)gl tumors in host thorax muscle (M). (C) 97-2 insertion l(2)gl/97-2 insertion l(2)gl. Primary tumor (T), no metastasis. (D)97-2 excision l(2)gl/97-2 excision l(2)gl. Primary tumor (T) and metastasis (M). (E) l(2)gl/l(2)gl;23-2 insertion/23-2 insertion. Tumorigenesis and metastasis is suppressed. (F) l(2)gl/l(2)gl; 23-2 excision/23-2 excision. Metastasis phenotype is recovered (M).

Identification of Functional Mutations and Cloning of Associated Genes. Insertion 97-2 blocked metastasis in 24 of 25 hosts although it did not inhibit primary tumor growth (Fig. 2C). Excision of the P element reverted this line to the l(2)gl metastatic phenotype (Fig. 2D). P element insertion 115-1 double mutants accelerated the lethality of injected tumors. Fig. 4A shows the survival curves of hosts transplanted with 115-1 or l(2)gl tissues. The 115-1 insertion accelerates the lethality of the injected tumors (Mantel-Haenszel test, P = 0.001).

Fig. 4.

(A) Survival curves of hosts injected with 115-1 (double-mutant) tissue (circles) and l(2)gl tissue (squares). Lethality is accelerated in 115-1 double mutant, Mantel-Haenszel two-tailed P = 0.001. (B) RT-PCR analysis of apontic gene expression. Parental line cDNA, tubulin primers (lane 1); parental line cDNA, apontic primers (lane 2); 97-2 cDNA, tubulin primers (lane 3); 97-2 cDNA, apontic primers (lane 4). Apontic RT-PCR product, 174 bp; tubulin RT-PCR product, 165 bp. (C) RT-PCR analysis of pointed expression. 115-1 cDNA, tubulin primers (lane 1); 115-1 cDNA, pointed primers (lane 2); parental line cDNA, tubulin primers (lane 3); parental line cDNA, pointed primers (lane 4). Tubulin RT-PCR product, 165 bp; pointed RT-PCR product, 129 bp.

A third P element insertion, line 23-2, disrupted tumorigenesis of l(2)gl brain tissue in 28 of 29 hosts. Two copies of this P element insertion blocked proliferation of the l(2)gl primary tumor (Fig. 2E) but did not alter viability of the larva or grossly modify l(2)gl brains, which are composed of overgrown tissues with loosely adherent cells. Thus, the gene disrupted in this line is required for the l(2)gl malignant phenotype. Excision of the P element in line 23-2 resulted in reversion to the tumorigenic and metastatic phenotype (Fig. 2F).

The genomic DNA at the 3′ end of the 23-2, 97-2, and 115-1 P elements was isolated by plasmid rescue (Fig. 3) and sequenced (National Institutes of Health minicore facility). The genomic sequence flanking the 97-2 insertion was cloned in a 5.9-kb EcoRI fragment (Fig. 3). The 97-2 insertion is on 2R at 59F1, between the Pi3K59F and apontic genes. The sequence flanking the 115-1 insertion was cloned in a 5.4-kb EcoRI fragment (Fig. 3). This P element is on chromosome 3 at 94E, 445 bp from the translated region of the pointed gene. The genomic sequence flanking the 23-2 P element was cloned in a 7.1-kb SstI fragment. The 23-2 P element is inserted at 3L at 68F2, 46 bp from the translated region of the semaphorin 5c gene. Confirmation of the genomic cloning was performed by PCR amplification of genomic DNA from each line with specific primers. In each case, one primer matched the P element sequence near the 3′ end and the second primer matched a sequence in the flanking genomic DNA. PCR amplification with each P element insertion and flanking genomic DNA primer pair resulted in a product of a predicted size for that P element line but did not amplify a product in the parental line (Fig. 3 A and B).

Fig. 3.

Cloning of genomic regions flanking P element insertions. Genomic regions from P element insertion lines (arrowheads at insertion site). 97-2 insertion is located 15.6 kb from the Pi3K59F gene and 16.3 kb from the apontic gene; 115-1 insertion is 445 bp from the start of the translated region of the pointed gene, and 23-2 insertion is 46 bp from the start of the translated region of the semaphorin 5c gene. (A) PCR amplification of 23-2 and 97-2 insertions with primers specific for 3′ P element sequence and flanking genomic sequence. Parental genomic DNA, tubulin primers (lane 1); parental genomic DNA, 23-2 insertion primers (lane 2); parental genomic DNA, 97-2 insertion primers (lane 3); 23-2 genomic DNA, tubulin primers (lane 4); 23-2 genomic DNA, 23-2 insertion primers (lane 5); 97-2 genomic DNA, tubulin primers (lane 6); 97-2 genomic DNA, 97-2 insertion primers (lane 7). Tubulin PCR product, 655 bp; 23-2 PCR product, 241 bp; 97-2 PCR product, 261 bp. (B) PCR amplification of 115-1 insertion with specific primers. Parental genomic DNA, actin primers (lane 1); parental genomic DNA, 115-1 insertion primers (lane 2); 115-1 genomic DNA, actin primers (lane 3); 115-1 genomic DNA, 115-1 insertion primers (lane 4).

P Element Insertions Caused Up-Regulation of apontic and pointed. Expression of the apontic and pointed genes was examined by RT-PCR in lines 97-2 and 115-1. The expression of apontic is higher in the P element line 97-2 compared with the parental line (Fig. 4B). PI3K was also examined in the 97-2 line, as the insertion is between the Pi3K59F and apontic genes and could affect either or both genes. The expression level of PI3-K was also found to be up-regulated in the 97-2 insertion line. RT-PCR analysis showed that pointed was up-regulated in the 115-1 insertion line compared with the parental line (Fig. 4C).

P Element Insertion Causes Loss of Semaphorin 5c Protein Expression. Semaphorin 5c was undetectable in the 23-2 line protein extracts from dissected brain tissue. Excision of the P element resulted in recovery of protein expression (Fig. 5A) and restoration of the malignant phenotype (Fig. 2F). Expression of the CG17154 gene which is located 6 kb from the 23-2 insertion was also examined to determine whether this gene contributed to the metastasis effects in line 23-2. RT-PCR analysis showed that this gene was not disrupted by the P element insertion (data not shown). Based on sequence homology, three related mammalian semaphorins were identified with sequence domains similar to those of semaphorin 5: SEMA5A, SEMA5B, and a novel protein (KIAA1445) that we have named SEMA5D. The SEMA5D protein shares sequence similarity to SEMA5B and contains an additional 51 aa at the N terminus not shared by other class V semaphorins. All three of these proteins are class 5 semaphorins, containing thrombospondin repeats, a sema domain and a transmembrane domain (Fig. 5B). The SEMA5A and SEMA5D proteins were shown to be expressed in membrane preparations of A2058 cells (Fig. 6A) and MDA435 cells (data not shown). Furthermore, SEMA5D was shown by immunohistochemistry to be expressed in the membrane of ovarian cancer cells (Fig. 6B).

Fig. 5.

Semaphorin 5c expression is absent in 23-2 insertion line larval brain extracts. (A) Western blotting with anti-Semaphorin antibodies and Drosophila brain extracts. Parental line expresses sema-5c (lane 1), 23-2 insertion line lacks sema-5c expression (lane 2), and 23-2 excision line restores sema-5c expression (lane 3). (B) Class 5 semaphorin domains. (C) Protein microarray analysis of selected signaling proteins in l(2)gl/l(2)gl and l(2)gl/l(2)gl sema-5c/sema-5c brain tissues. Wild-type values were subtracted from l(2)gl/l(2)gl and l(2)gl/l(2)gl sema-5c/sema-5c values.

Fig. 6.

Expression of human homologs of semaphorin 5c. (A) Expression of SEMA5A and SEMA5D are detected in membrane preparations of A2058 human melanoma cells. (B) Immunohistochemistry shows membrane localization of SEMA5D in ovarian cancer cells.

Semaphorin 5c Mutation Disrupts Dpp Signaling in l(2)gl. The structural domains of Semaphorin 5c were studied to determine a potential functional role in l(2)gl tumors. The Semaphorin 5c protein is a transmembrane protein containing extracellular thrombospondin type I (TspI) repeats, which are known to regulate TGF-β signaling in mammals through activation of the ligand (13, 14). The short gastrulation gene (sog) is a negative regulator of a Drosophila TGF molecule, decapentaplegic (Dpp), and is proposed to function by sequestering Dpp through binding of the ligand to repeated von Willebrand motif sequences that are homologous to sequences found in thrombospondin (15). We examined the hypothesis that Sema-5c might play a role in the regulation of the Dpp signal transduction pathway via its extracellular thrombospondin repeats. Signaling via the Dpp pathway was studied in l(2)gl/l(2)gl and in l(2)gl/l(2)gl;semaphorin 5c/semaphorin 5c tissues by reverse phase protein microarray analysis (linearity r = 0.99, SD < 10% of the mean; ref. 11; Fig. 5C). The levels of activated (phosphorylated) and inactive Mothers against Dpp (Mad), the primary effector in the Dpp pathway (16), homologous to human Smad1 (17), were measured. Furthermore, we examined the activation of other signaling pathways that have been shown to play a role in mammalian cancer. PI3-K (18), ERK (19), Akt (11), and cleaved caspase 3 (20) were studied in brain extracts from l(2)gl, l(2)gl;sema-5c double mutants, and wild-type larvae (Fig. 5C). A striking difference in phospho-Mad levels was observed. Phospho-Mad was significantly overexpressed in l(2)gl/l(2)gl compared with wild-type tissues. After loss of Sema-5c, P-Mad levels were reduced to levels similar to wild-type.

The expression of genes downstream of phospho-Mad was examined by RT-PCR to determine targets of Dpp that may play a role in the l(2)gl phenotype. The principal genes regulated through Dpp signaling that have been identified in the wing imaginal disk model are spalt (21), optomotor blind (22), and vestigial (23). The spalt and optomotor blind genes were unchanged in l(2)gl compared with wild type. The expression of the Dpp target gene vestigial, however, was increased in l(2)gl tissue compared with wild-type or l(2)gl/l(2)gl;sema-5c/sema-5c mutants (Fig. 7 A and B).

Fig. 7.

(A and B) Expression of the Dpp target gene vestigial is increased in l(2)gl brain tissue compared with wild-type. RT-PCR analysis demonstrated elevated vestigial levels (quantitated in proportion to tubulin) (n = 3) in l(2)gl tissues compared with wild-type or lgl/lgl;sema-5c/sema-5c (23-2). (C) Model for role of TSP-1 repeats in Semaphorin 5c activation of Dpp pathway.

PI3-K was also reduced in l(2)gl/;sema-5c double mutants compared with l(2)gl. To further study the role of PI3-K in l(2)gl tumors, the PI3-K inhibitor LY294002 was orally administered to Drosophila adults injected with l(2)gl tissue. LY294002 treatment reduced the primary tumor size to 7% of untreated host tumors, without adverse effects to the hosts (data not shown).

Discussion

We used a Drosophila model to identify three genes that functionally alter tumorigenicity or metastasis. These genes were identified by the alteration of their expression via insertion of a P element and the subsequent effects on l(2)gl tumorigenesis and/or metastasis. Reversion of the phenotype by excision of the P element showed that the effects were due to the P element. Two of the genes, pointed and apontic, act at the level of regulation of gene transcription/translation and may influence multiple downstream events. In contrast, the third gene, semaphorin 5c, is a transmembrane protein with a large extracellular domain that contains seven thrombospondin type I (Tsp I) repeats. Such repeats are well established (14, 15) to bind and activate TGF-β. Consequently, we explored the state of the signal pathway activated by the TGF-like Drosophila Dpp ligand. Our results demonstrate that l(2)gl tumorigenicity requires Sema-5c, which is in turn linked to activation of the Dpp/Mothers against Dpp (Mad) signal pathway and up-regulation of the downstream target gene, vestigial.

Disruption of apontic specifically blocked metastasis but not tumorigenicity of l(2)gl tumors. The apontic gene is described as a transcription factor affecting genes necessary for migration (24) or homeotic targets (25). We propose that apontic may act via downstream targets to control migration, and invasion of l(2)gl tumor cells. The pointed gene may also contribute to l(2)gl metastasis through regulation of downstream genes. Disruption of pointed caused an acceleration of host lethality from transplanted l(2)gl tumors. pointed is a member of the ets-like transcription factor family (26), conserved between vertebrates and Drosophila (27, 28). The pointed gene has been shown to have developmental roles in the eye, neurogenesis, tracheal cell migration, and oogenesis (26). pointed has been found to be involved in negative feedback regulation of EGF-R signaling, as well as to control dpp expression during embryogenesis (29). The human homolog of pointed, the c-Ets1 proto-oncogene, has been shown to regulate the expression of genes important in extracellular matrix remodeling and invasion, including stromelysin–1 (29), collagenase-1 (30), and urokinase-type plasminogen activator (31).

The semaphorin 5c gene is shown here to be required for proliferation of l(2)gl tumors and provides a functional link to the Dpp signal pathway. The semaphorin 5c gene belongs to the class 5 group of semaphorins, which are transmembrane proteins with short cytoplasmic (C-terminal) tails and extracellular domains containing seven thrombospondin type I repeats, a plexin domain, and a semaphorin domain sequences (32). The semaphorin 5c gene was identified in Drosophila by scanning the genome for sema sequences (32). In keeping with our observations here, null mutations of sema5c are fully viable (33). Class 3 semaphorins, previously linked to cancer (34), are structurally different from class 5, lacking the thrombospondin repeats present in the transmembrane class 5 semaphorins.

We examined the activation state of Mothers against dpp (Mad) a proximal effector in the Drosophila TGF-β signal pathway (16). Our results indicate that Sema-5c may function through the Dpp/TGF growth signaling pathway for the following reasons: (i) l(2)gl null mutant overgrown brain tissue exhibited significant elevation of phospho-Mad compared with wild-type tissue; (ii) in the absence of Sema-5c expression, the l(2)gl null mutant tissue contained phospho-Mad level equivalent to or less than wild-type tissue; and (iii) we could stimulate the phosphorylation of Mad by addition of Dpp to wild-type tissues. The Drosophila Mad protein has been shown to function through DNA binding to mediate activation of specific target genes (23), including those identified in the imaginal wing disk model: spalt (21), optomotor blind (22), and vestigial (23). Ectopic expression of the Dpp protein or constitutively active Dpp receptor has been shown to cause large overgrowth in the wing imaginal disk, leading to outgrowths or duplication of the wing (35–37). The excessive proliferation caused by constitutively active Dpp receptor (Tkv^Q2530) required normal activity of PI3-K (35). Furthermore, of the target genes tested, only vestigial was found to be required for the Dpp-induced overgrowth (35). We found that l(2)gl tissue had increased levels of the Dpp target gene vestigial but not spalt or optomotor blind, suggesting that the expression of vestigial may play a role in producing the overgrowth phenotype in l(2)gl tissues.

It is somewhat surprising that l(2)gl tumors had elevated pMad because Arquier et al. (38) reported that l(2)gl causes loss of Dpp signaling. The conclusions of Arquier et al. (38) were in regard to the vesicular secretion of the active form of Dpp. l(2)gl tumors may be dependent on activation of the Dpp ligand. Our working hypothesis is analogous to the findings of significant ectopic expression of Dpp in overgrown Drosophila tissues of several tumor suppressor mutants (39) and to the findings of Martin-Castellanos and Edgar (35) that Dpp can stimulate wing overgrowth through vestigial regulation. l(2)gl mutant cells that express Sema-5c may have the capacity to activate, or possibly present, the Dpp ligand to the receptor complex (Fig. 7), resulting in a neoplastic augmentation of the Dpp pathway.

For human cancer, it has been observed that aggressive tumors of many histologic types can be stimulated by TGF-β-induced signals (40). Furthermore, the homolog of Mad, human Smad1 (17), is known to be activated in human cancer (41). Human class 5 Semaphorins Sema5A, Sema5B, and Sema5D (KIAA1445) (Figs. 5 and 6) have not previously been associated with the TGF-β pathway and cancer.