INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Because defects in intracellular signaling can cause cancer, components of signal transduction pathways are common targets of antineoplastic drugs. Among these, secreted and cell surface proteins are preferred because they are easily accessible to specific agonists and antagonists. For example, many epithelial cancers show constitutive activation of the tyrosine kinase receptors for epidermal growth factor (EGF) or insulin-like growth factor (IGF) (Schlessinger 2000); this has led to the development of drugs that interfere with these receptors. Drugs like ZD1839/Iressa, which antagonizes EGF receptor activation and signaling, are presently in clinical development (Arteaga 2000).

The targeting of proteins to the secretory pathway requires a short N-terminal sequence called a "signal" or "leader" peptide (von Heijne 1985), which also determines the protein's orientation across the cellular membranes. This signal sequence is conserved in secreted and membrane-spanning proteins and has been exploited in all strategies developed to isolate or identify signal sequence genes. Thus, signal sequence traps consisting of reporter or selectable marker genes, whose expression is dependent on the acquisition of a signal sequence, or antibodies raised against signal sequence peptides, have been used with variable success (Tashiro et al. 1993; Skarnes et al. 1995; Imai et al. 1996; Klein et al. 1996; Scherer et al. 1998; Lim and Garzino-Demo 2000; Mitchell et al. 2001). To develop a strategy that would be applicable to genomewide screens for secreted and transmembrane proteins in living cells, we used a gene trap approach.

Gene traps insert a reporter gene into mostly random chromosomal sites, including transcriptionally active regions. By selecting for gene expression, recombinants are obtained in which the reporter gene is fused to the regulatory elements of endogenous genes. Transcripts generated by these fusions faithfully reflect the activity of the tagged cellular gene and thus provide an effective means to study the expression of genes in their normal chromosomal location (Friedrich and Soriano 1991; Skarnes et al. 1992; von Melchner et al. 1992). By using appropriate reporter systems, we and others have developed gene trap strategies to identify genes regulated during important biological processes or by exogenous stimuli (Reddy et al. 1992; Forrester et al. 1996; Russ et al. 1996a; Thorey et al. 1998; Medico et al. 2001).

To develop a gene trap strategy that would select for integrations into regulated genes encoding secreted and/or transmembrane proteins, we used the human CD2 cell surface antigen fused in-frame to an Escherichia coli neomycin-phosphotransferase (Ceo) as a reporter gene in the retroviral gene trap U3Ceo. Because the signal peptide sequence of the CD2 cDNA terminates in a cryptic splice acceptor site (GenBank accession no. XM_002141), we assumed that it would be removed by splicing from U3Ceo integrations into introns of expressed genes. Consequently, the cell surface expression of the CD2-neo fusion protein would rely on the acquisition of a signal sequence from an endogenous gene.

We show here that the retroviral gene trap U3Ceo effectively selects for integrations into genes encoding secreted and/or transmembrane proteins that are repressed by oncogenic transformation. This enables a genomewide screen for proteins with putative antineoplastic functions and for targets that are easily accessible to anti-cancer drugs.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Construction of a Gene Trap Retrovirus Expressing a CD2/Neomycin-Phosphotransferase Fusion Gene (U3Ceo)

To construct a gene trap that would allow selection for integrations into genes with signal sequences, we cloned the combined reporter/selectable marker gene Ceo into the U3-region of a promoter/enhancer-deleted retroviral vector based on pBabePuro (Morgenstern and Land 1990). The Ceo gene was an in-frame fusion between the genes encoding the human T-cell-specific CD2 antigen (CD2) (Seed and Aruffo 1987) and the E. coli neomycin-phosphotransferase (neo) (Colbere-Garapin et al. 1981). It was obtained by fusing an ATG-deleted and truncated CD2 cDNA to the neo gene immediately downstream of its first ATG (Fig. 1A). The truncated CD2 cDNA, which consists of nucleotides 15-782 (GenBank accession no. XM_002141), encodes the entire extracellular and transmembrane domains, but only a short segment of the intracellular domain. As has been shown in previous studies, virus replication and long terminal repeat (LTR)-mediated duplication places sequences inserted into U3 just 30 nucleotides downstream of the flanking chromosomal DNA (von Melchner and Ruley 1989). Because of a cryptic splice acceptor sequence located in the CD2 coding sequence 58 nucleotides downstream of the ATG, and a branch site consensus sequence located 21 nucleotides upstream of the splice site (nucleotides 41-47), gene trap integrations into the introns of transcribed genes were expected to express cell-provirus fusion transcripts in which the upstream exons are spliced to CD2. Because this excises CD2's signal sequence, we anticipated that U3Ceo expression would be dependent on an in-frame fusion to a signal sequence of a cellular gene (Fig. 1B). Selection for U3Ceo expression would then enrich for integrations into genes with signal sequences.

View larger version (22K): [in this window] [in a new window]   Figure 1   Mechanism of signal sequence capture by the U3Ceo gene trap. (A) U3Ceo gene trap construct containing a CD2/neomycin-phosphotransferase fusion gene (Ceo) in the U3 region of the 3'-long terminal repeat (LTR). [prom/enh ()] Promoter/enhancer deleted. (B) Mechanism of U3Ceo activation. LTR-mediated duplication places the Ceo fusion gene 30 nucleotides from the 5' chromosomal flanking regions. Proviral integrations into the introns of expressed genes result in splicing of upstream exons to the Ceo fusion gene downstream of its cryptic splice acceptor site. In-frame acquisition of a signal sequence from an endogenous gene (alternating open and filled squares) enables Ceo transport to the cell membrane and converts cells to CD2 positivity and G418 resistance.

Development of a Cell Line Susceptible to Reversible Oncogenic Transformation

To identify genes that are repressed by oncogenic transformation, we required a cell system that would allow a controlled and reversible switching from a normal to a transformed state. A cell line (N93.1/28) with such properties was derived from NIH3T3 fibroblasts after transducing a tetracycline-sensitive (tet-off system) human IGF-1 receptor (IGF-1R) gene and selecting for clones with a tight anhydrotetracycline (ATC) repressible promoter (Baasner et al. 1996). When N93.1/28 cells were exposed to IGF-1, they converted to anchorage-independent growth and formed foci in standard focus-forming assays (Fig. 2). Conversion was dose dependent and could be readily reversed by ATC, indicating that IGF-1R signaling is required for transformation (Fig. 2B) (Baserga et al. 1997).

View larger version (68K): [in this window] [in a new window]   Figure 2   Insulin-like growth factor 1 (IGF-1)-induced transformation of N93.1/28 cells. (A) Focus formation of IGF-1R overexpressing N93.1/28 cells ( ATC) in the presence of IGF-1. (B) Colony formation in soft agar. 1.5 × 103 N93.1/28 cells were incubated for 5 d in semisolid cultures ± IGF-1 or anhydrotetracycline (ATC). Colony growth was quantified by measuring light absorption at 490 nm using a spectrophotometer as (described in the Methods section).

U3Ceo Selects for Integrations into Genes Encoding Secreted and Transmembrane Proteins

To test the ability of U3Ceo to capture cellular signal sequences, we infected nontransformed N93.1/28 cells with U3Ceo virus at a simple multiplicity of infection (MOI = 1) and selected in G418. From a library of 1 × 10⁶ independent integrations, we obtained 400 G418-resistant (Neo^R) clones, indicating that only 1 in 2500 integrations are transcribed. This suggested that only a small subset of genes expressed in N93.1/28 cells are capable of activating U3Ceo, and led to the assumption that this subset represents the signal sequence genes. If this is indeed the case, the G418-resistant clones should also express CD2 as a result of in-frame fusions between an endogenous signal sequence peptide and the Ceo protein. To test this, we treated pooled Neo^R clones with the anti-CD2-specific monoclonal antibody Leu5b and analyzed by flow cytometry for CD2 expression. Figure 3 shows that >86% of the Neo^R cells were positive for CD2, indicating that the majority of U3Ceo fusion proteins have captured an endogenous signal peptide. This high frequency of CD2-expressing Neo^R cells indicated that the majority of Ceo fusion proteins expressed from gene trap integrations into genes without a signal sequence are unstable (Skarnes et al. 1995). To investigate this further, cell-provirus fusion transcripts from eight randomly selected Neo^R clones were amplified by 5'RNA-ligase mediated (RLM) RACE (Maruyama and Sugano 1994) and sequenced. In each case, cellular sequences were fused to nucleotide 69 in the CD2 cDNA, which is immediately downstream of the cryptic splice acceptor. Splicing of cellular exons to the Ceo fusion gene deleted the first 90 nucleotides of U3Ceo, including the CD2 signal sequence and the cell DNA-provirus junction. Database analysis of the fusion transcripts revealed that U3Ceo acquired signal sequences from six previously characterized proteins and from an anonymous EST (Table 1). Thus, taken together, the results indicate that U3Ceo effectively selects for integrations into genes encoding secreted and/or membrane-spanning proteins.

View larger version (15K): [in this window] [in a new window]   Figure 3   CD2 expression in G418-resistant (NeoR) N93.1/28 cells. 1 × 106 cells were infected with U3Ceo retrovirus at MOI = 1 and selected in G418 (800 µg/mL). 2 × 104 NeoR cells were treated with the monoclonal mouse anti-human CD2 antibody Leu-5b. CD2 expression was estimated by flow cytometry after treating the cells with an FITC-conjugated anti-mouse IgG antibody.

Selection for and against U3Ceo Expression Identifies Genes Repressed by Oncogenic Transformation

The value of secreted and cell surface proteins that are involved in transformation is presently unmatched. Therefore, we asked whether U3Ceo could be used to identify proteins whose expression is repressed in transformed cells. For this, we infected 1 × 10⁷ N93.1/28 cells with U3Ceo at a MOI = 1 and selected the integration library in G418. The G418-resistant library with U3Ceo integrations in expressed genes was then subjected to several rounds of positive and negative selection using antibody-mediated cell adhesion (panning) with the Leu-5b anti-CD2 antibody (Seed and Aruffo 1987). As shown in Figure 4A, after a first round of positive panning, most cells expressed the CD2 cell surface antigen. However, expression levels varied broadly, as one would expect from a polyclonal population with U3Ceo integrations into genes with variable promoter strength. The CD2-positive library was then transformed by IGF-1 and selected against CD2 expression to eliminate all integrations in constitutively expressed genes (Fig. 4 B,C). Finally, to ensure that only truly regulated genes were recovered, the negatively selected library was reversed to its nontransformed state by withdrawing IGF-1 and again selecting for CD2 expression. As shown in Figure 4D, the CD2 receptor returned to the cell surface in largely variable amounts, indicating that polyclonality was maintained despite multiple rounds of selection. Moreover, CD2 expression was promptly repressed following IGF-1 addition, implying that the recovered U3Ceo integrations are in tightly regulated genes (Fig. 4E).

View larger version (24K): [in this window] [in a new window]   Figure 4   Sequential enrichment for U3Ceo integrations repressed by transformation. A U3Ceo integration library consisting of ~1 × 107 N93.1/28 cells was subjected to several rounds of "panning" after selecting in G418. Following selection for CD2 expression (positive panning), the library was transformed by insulin-like growth factor 1 (IGF-1) and selected against CD2 expression (negative panning). Stepwise enrichment for U3Ceo integrations into IGF-1 regulated genes was estimated by flow cytometry as described in the legend to Figure 3. (A) Fluorescence profile after positive panning of untreated cells. (B, C) Fluorescence profile after two consecutive rounds of negative panning of IGF-1 transformed cells. (D, E) Fluorescence profile of the preselected library after a second round of positive panning ± IGF-1.

To identify some of the regulated genes trapped by U3Ceo, six clones were isolated from the preselected library by limiting dilution, and their cell-provirus fusion transcripts were amplified and sequenced. Database analysis revealed that, in each case, U3Ceo had disrupted a gene encoding a signal peptide protein involved in oncogenic transformation and metastatic spread (Table 2). Each of these genes was tightly regulated by IGF-1, indicating that gene repression was either caused directly by IGF-1 or, alternatively, occurred as a consequence of oncogenic transformation (Fig. 5, and data not shown).

View larger version (40K): [in this window] [in a new window]   Figure 5   Insulin-like growth factor 1 (IGF-1) regulated expression of the thrombin receptor gene disrupted by U3Ceo in N93.1/28 cells. (A) Northern blot analysis of the thrombin receptor transcript expressed from the undisrupted allele. N93.1/28 cells were incubated for 72 h with or without 100 ng/mL IGF-1. Total RNAs (25 µg/lane) were fractionated on formaldehyde-agarose gels, blotted onto nylon filters, and hybridized to a 32P labeled thrombin receptor-specific probe. Hybridizing transcripts were visualized by a phosphorimager after exposing for 5 h to a phosphorimager screen. (B) Analysis of U3Ceo expression from the allele disrupted by the provirus. IGF-1 regulated CD2 expression was analyzed by flow cytometry as described in the legend to Figure 3.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

In the present study we have developed a gene trap strategy enabling genomewide screening for putative tumor suppressor genes encoding secreted and/or cell surface proteins in living cells. By infecting a conditionally transformable cell line with the retroviral gene trap U3Ceo and selecting in G418, recombinants were obtained that expressed a functional CD2/neomycin-phosphotransferase fusion protein on the cell surface. In each case, transport to the cell membrane was enabled by the acquisition of an N-terminal signal peptide encoded by an endogenous gene.

Proteins targeted to the secretory pathway have generated considerable interest for two main reasons. First, most fundamental biological processes involve secretory proteins and membrane receptors. Second, secreted and membrane-bound proteins are easily accessible to exogenous drugs and are thus preferred targets for drug development. To identify such proteins, several strategies have been developed that all take advantage of a characteristic N-terminal signal peptide sequence that is required for channeling through the secretory pathway (Tashiro et al. 1993; Skarnes et al. 1995; Imai et al. 1996; Klein et al. 1996; Scherer et al. 1998; Lim and Garzino-Demo 2000; Mitchell et al. 2001). In most cases, reporter- and/or selectable marker genes encoding a transmembrane domain but lacking a signal sequence have been used to identify complementing cDNAs from cDNA libraries (Tashiro et al. 1993; Imai et al. 1996; Klein et al. 1996; Lim and Garzino-Demo 2000). Similarly, reporter genes dependent on signal sequences for expression have been used to directly screen the genome for secretory proteins. For example, a gene trap encoding a -galactosidase/neomycin-phosphotransferase (geo) fusion protein into which a CD4 transmembrane domain had been inserted was shown to enrich for integrations into signal sequence genes expressed during mouse development (Skarnes et al. 1995; Mitchell et al. 2001). However, because the efficiency of gene trap activation by signal sequence capture was <20%, this gene trap seems unlikely to be suitable for large-scale functional genomics. In contrast, U3Ceo, with its signal sequence capture frequency of >86%, seems ideally suited for this purpose. Thus, by using U3Ceo gene traps for each reading frame, it seems possible to tag all secreted and transmembrane proteins expressed in the mammalian genome.

The most attractive feature of U3Ceo is its ability to select for a subset of genes that are not only encoding secreted and/or cell surface proteins but are also regulated during a biological process. For example, by selecting for and against Ceo expression in a reversible transformation model, we expected to recover genes that are involved in oncogenesis and tumor suppression. Indeed, each of the genes disrupted by U3Ceo in this model had been previously shown to interfere with tumor growth and metastatic spread. The human homolog of the thrombin receptor (PAR-1) and the PDGF A-chain homodimer induce cell cycle arrest and apoptosis by up-regulating cyclin-dependent kinase inhibitors and caspases (Huang et al. 2000; Yu et al. 2000). The -2 type 1 collagen and the Ly-6A.2 alloantigen inhibit cellular transformation and proliferation (Travers et al. 1996; Satoh et al. 1997). Moreover, Ly-6A.2, also known as stem cell antigen-1 (Sca-1) is expressed by primitive stem cells of the hematopoietic system and mediates cell adhesion (English et al. 2000). Similarly, cell adhesion is promoted by the membrane-bound receptor-like protein tyrosine phosphatase  (R-PTP-), indicating that both proteins are likely to reduce metastatic spread (Fuchs et al. 1996). Finally, the class I major histocompatibility complex (MHC-1) is essential for the recognition of tumor antigens by the immune system. In many cancers, down-regulation of MHC-1 by tumor cells enables them to evade a specific immune response (Seliger et al. 2000). Taken together, these results underscore the ability of U3Ceo to selectively identify genes with type-2 tumor suppressor function (Lee et al. 1991).

In conclusion, the gene trap strategy described here provides a means to perform genomewide screens for secreted and/or transmembrane proteins regulated during a specific biological process. Although reversible transformation systems such as the one used here are particularly attractive for drug-target discovery in cancer research, the approach seems flexible enough to be adapted to almost any other biological system in which altered signal transduction leads to a detectable phenotype. Particularly interesting in this regard are the postgenomic large-scale mouse mutagenesis programs that are certain to benefit from the unique features of the U3Ceo gene trap (Wiles et al. 2000; Mitchell et al. 2001).

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

Plasmids and Viruses

A CD2/neomycin-phosphotransferase fusion gene (Ceo) was inserted into the NheI cloning site of a pBabeSin retroviral vector to obtain pBabeU3Ceo. pBabeSin was derived from pBabePuro (Morgenstern and Land 1990) by removing SV40puro and the U3 promoter/enhancer sequences from the 3'LTR. To this end, pBabePuro was cleaved with SalI/XhoI and NheI/BanII, respectively, and religated. To obtain Ceo, CD2 and neomycin-phosphotransferase coding sequences were amplified by PCR using the primer pairs 5'-attctagattccatgt aaatttgtagccagcttcc-3'/5'-cgcgggggatccgtagctactct gtgggctcttgtctc-3' and 5'-cgcgggggatcccattgaacaagatg gattgcacgcagg-3'/5'-cgcggggaattctctagattag-aag aactcgtcaagaaggcg-3', respectively. The CD2-amplification product was ligated as an XbaI/BamHI fragment to the 5' BamHI site of the amplified neomycin-phosphotransferase sequence and the fusion gene was cloned as an XbaI/EcoRI fragment into pBluescript KS. After verifying the in-frame fusion by sequencing (ABI310 Genetic Analyzer, Applied Biosystems), the fusion gene was ligated as an XbaI fragment into the NheI site of pBabe Sin.

High-titer U3Ceo retroviral supernatants were generated by transient transfection of pBabeU3Ceo into BOSC23 packaging cells and used for infections as described previously (Russ et al. 1996b).

Cell Cultures and Colony Assays

Cells were grown in Dulbecco's Modified Eagles Medium supplemented with 10% newborn calf serum (NCS). NIH3T3 cells overexpressing the human IGF-1 receptor under control of the tetracycline-regulated promoter were generated by stable transfection as previously described (Baasner et al. 1996). Clones isolated by limiting dilution were tested for tetracycline-regulated IGF1-R expression by Northern blotting and by reversible tumorigenicity in vitro as described previously (Andreu et al. 1998). The cell line N93.1/28 with tightly regulated IGF-1R expression was selected for all experiments.

Anchorage-independent growth was analyzed in soft agar colony assays by plating 1.5 × 10³ cells into 96 well plates containing 0.5% w/v DMEM/agar supplemented with 10% (v/v) NCS. After incubating for 7 d, cell proliferation was estimated using the XTT-Cell Proliferation Kit (Roche Diagnostics, Mannheim) according to the manufacturer`s instructions. Light absorption reflecting colony proliferation was measured at 490 nm using a Victor 1420 multilabel counter (Wallac, Turku).

Panning and Antibody Treatment

The positive and negative panning procedures with the human CD2-specific mouse monoclonal antibody Leu-5b (Becton Dickinson, Heidelberg) were performed as follows: 100-mm bacterial plates were treated with 10 mL of a 20 µg/mL solution of anti-mouse-IgG-antibody (Cappel, Durham). After incubating for 1.5 h, plates were washed three times with 0.15 M NaCl and overlaid with 10 mL of 1% (w/v) solution of bovine serum albumin in PBS. After incubating overnight at room temperature, the bovine serum albumin was removed and plates were stored at -20°C until use. To identify genes repressed by IGF-1, we infected 1 × 10⁷ N93.1/28 cells with U3Ceo at MOI = 1 and first selected for 10 d in G418 (800µg/mL). For panning, cells were suspended at a concentration of 5 × 10⁶/mL and incubated for 30 min at room temperature in a 300 ng/mL Leu-5b antibody solution. After washing in PBS, the antibody-treated cells were placed on top of the anti-mouse-IgG antibody coated plates. After incubating for 1 h at room temperature, plates were gently washed with 10 mL PBS to discard the nonadherent cells. Adherent cells were subsequently harvested in DMEM medium supplemented with 10% NCS and exposed to 100 ng/mL recombinant human IGF-1 (Sigma, Deisenhofen) for 3 d. Transformed cells were subjected to negative panning essentially as described earlier except that this time the nonadherent cells were harvested and the adherent cells discarded.

Flow Cytometry

1 × 10⁶ cells were suspended in 50 µL FACS-Wash solution (Becton Dickinson, Heidelberg) and incubated for 30 min at 4°C with 2.5 µg/mL of CD2-specific mouse monoclonal antibody Leu-5b. Following washings, the cells were incubated for 30 min at 4°C in 50 µL FACS-Wash solution containing 10% (v/v) FITC-conjugated anti-mouse-IgG antibody (Roche Diagnostics, Mannheim). Finally, 2 × 10⁴ cells were suspended in 500 µL FACS-Wash solution and analyzed with a FACSCALIBUR (Becton Dickinson) flow cytometer using FITC-specific settings.

5'-RACE

5'-RACE was performed on 5 µg of total RNA using the FirstChoice RLM-RACE-Kit (Ambion) according to the manufacturer's instructions. CD2 specific primers were 5'-caagttgatgtc ctgacccaag-3' and 5'-ggtttccaaggcattcgtaatctc-3' (nested). The first amplification was for 35 cycles, at 96°C for 30 sec, 60°C for 30 sec, and 72°C for 2 min, whereas the second (nested) amplification was for 35 cycles at 96°C for 30 sec, 62°C for 30 sec, and 72°C for 2 min. RACE products were cloned into the pGEM-Teasy vector (Promega) and inserts of at least two independent E. coli clones were sequenced using a T7-specific primer (Promega).