Genetic reprogramming of tumor cells by zinc finger transcription factors

Abstract

Cancer arises by the accumulation of genetic alterations in DNA leading to aberrant gene transcription. Expression-profiling studies have correlated genomewide expression signatures with malignancy. However, functional analysis elucidating the contribution and synergy of genes in specific cancer cell phenotypes remains a formidable obstacle. Herein, we describe an alternative genetic approach for identification of genes involved in tumor progression by using a library of zinc finger artificial transcription factors (ATFs) and functional screening of tumor cells as a source of genetic plasticity and clonal selection. We isolated a six-zinc finger transcriptional activator (TF 20-VP, TF 20 containing the VP64 activator domain) that acts to reprogram a drug-sensitive, poorly invasive, and nonmetastatic cell line into a cell line with a drug-resistant, highly invasive, and metastatic phenotype. Differential expression profiles of cells expressing TF 20-VP followed by functional studies, both in vitro and in animal models, revealed that invasion and metastasis requires coregulation of multiple target genes. Significantly, the E48 antigen, associated with poor metastasis-free survival in head and neck cancer, was identified as one specific target of TF 20-VP. We have shown phenotypic modulation of tumor cell behavior by E48 expression, including enhanced cell migration in vitro and tumor cell dissemination in vivo. This study demonstrates the use of ATFs to identify the group of genes that cooperate during tumor progression. By coregulating multiple targets, ATFs can be used as master genetic switches to reprogram and modulate complex neoplastic phenotypes.

During different stages of a neoplastic disease, phenotypic diversity and clonal selection of tumor cells generate cell populations possessing traits that increase the potential for malignancy, such as increased mobility and invasiveness. Gene expression profiling of tumors has revealed that increased malignancy is associated with changes in gene expression affecting multiple loci (1–3). Because of their unique ability to orchestrate and coregulate multiple target genes, transcription factors (TFs) play a crucial role in generating phenotypic plasticity associated with cancer progression. TFs that play a role in lineage-specific differentiation, such as STATs [signal of transduction and activator of transcription (4)], and those that affect morphogenesis and embryonic development, such as Twist, Snail, and SIP1, have recently been shown to be involved in invasiveness during tumor progression (5, 6).

As a consequence of their inherent potential for altering genetic cascades, artificial TFs (ATFs) may be used to modulate cancer cell phenotypes. TFs possess several distinctive features. First, by interacting specifically with endogenous regulatory sequences, TFs can mediate the simultaneous regulation of multiple genes necessary for the control of complex phenotypes. Second, TFs have the ability either to up-regulate expression of target genes [when the DNA-binding domain (DBD) is linked to an activator of transcription, i.e., VP16, VP64, or p65] or to down-regulate target gene expression (when the same DBD is linked to a repressor domain such as the KRAB domain) (7). ATFs with zinc finger (ZF) DBDs are particularly useful because a well characterized lexicon of ZFs has been optimized for specific recognition of virtually any triplet of DNA (8, 9). Furthermore, we have recombined the existing ZF DBDs to generate multimodular libraries made of three- and six-ZF building blocks (10–12). When delivered into a tumor cell population, TF libraries provide phenotypic diversity on the order of millions of ATFs capable of “scanning” the tumor cell genome for functional, accessible regulatory sequences, each with the potential to affect transcription of multiple genes involved in tumor progression. A combination of differential DNA profiling and target site search, based on the predicted specificity of ZF units, allows selected ATFs to be used as genetic probes for the identification of genes and genetic interactions involved in malignancy.

In this article we describe the selection and characterization of a six-ZF ATF (TF 20-VP, TF 20 containing the VP64 activator domain) selected from a six-ZF combinatorial activator library capable of inducing drug resistance, cytoskeleton remodeling, matrix-dependent cell migration, and tumor cell invasion in vitro. Furthermore, TF 20-VP enhanced the number of tumor metastases in animal models. An analysis of the expression profiles of TF 20-VP-transduced cells revealed gene expression signatures involved in cancer progression. Our data support the use of ATFs as genetic switches to coregulate genes, discover new gene expression markers of cancer progression, and modulate complex phenotypes associated with malignancy.

Materials and Methods

ZF Selection. TF 20-VP was selected from a six-ZF retroviral activator library (pMX-6ZFlibrary-IRES-GFP) by treating 10⁸ HeLa cells with 200 μM Taxol (Sigma) for 72 h. Surviving cells were morphologically examined for the presence of epithelial-mesenchymal transition-like phenotypes. Retroviral DNA was recovered from genomic DNA and then recloned in the retroviral vector as described (10).

Migration Assays. The cell migration assay was performed by using transwell plates (8-μm pore size) (Costar). The undersurface of the membrane was coated at 4°C overnight with 0.25 μg/ml of Laminin (Sigma L-2020) diluted in PBS and then blocked with 2% BSA. The upper compartment was seeded with 2.5 × 10⁵ transduced HeLa cells per well in 100 μl of DMEM. FBS (2%) in DMEM was added in the lower chamber. Cells were allowed to migrate 22 h, and migrated cells were stained with 0.25% Crystal violet (Sigma) in 20% methanol (wt/vol). Each experiment used triplicate wells and, within each well, counting was done in four randomly selected microscopic high-power fields (×100).

In Vitro Invasion Assays. HeLa cells transduced with different retroviral vectors expressing TF 20-VP or TF 20-SKD (TF 20 containing the SKD repressor domain) or retrovirus containing no ZF (control) were starved overnight, and then 1.25 × 10⁵ cells of each cell line were loaded in Matrigel invasion chambers (24 wells, BD Biocoat Matrigel invasion chamber, BD Biosciences) according to the manufacturer's instructions. Cells were allowed to invade for 24 h, and then invasive cells were fixed and counted with an inverted microscope (Leica, McBain Instruments, Chatsworth, CA). The experiment was done two to three times with each sample in triplicate. For cDNA expression analysis 500 ng of each transient expression vector was transfected in six-well plates by using Lipofectamin PLUS according to the manufacturer's instructions (Invitrogen). cDNA-expressing vectors were pRC-CMV-E48 and pcDNA-IL-13Rα1-HA (a kind gift of K. Kohno, University of Occupational and Environmental Health, Kita-Kyushu, Japan). The angiotensinogen (AGT) expression vector was obtained from Invitrogen (clone ID 4213559). Transfected cells overexpressing the corresponding cDNA were assessed by real-time PCR quantification.

Quantification of Tumorgenicity and Metastasis. A total of 10⁶ HeLa cells transduced with retroviral constructs expressing TF 20-VP (seven mice), TF 20-SKD (seven mice), and no ZF domains (control, seven mice) were implanted s.c. into 3- to 4-week-old nonobese diabetic (NOD) severe combined immunodeficient (SCID) mice females (The Scripps Research Institute rodent breeding colony). The weight of each animal and the primary tumor volume were monitored each week. Animals were killed at day 35 postinjection. Lungs harvested from the animals were fixed in Bouin's solution, and the number of macroscopic metastases was assessed by counting nodules at the surface of the lung under a dissecting microscope. For consistency, the upper right lobe of each lung was used for quantification. In all cases this lobe represented metastases to the whole lung. Lung metastasis generated from TF 20-VP cell injection was significantly higher than the control cell injection (P < 0.05), but no significant difference in lung metastasis was found between TF 20-SKD cell injection and the control cell injection.

Microarray Processing and Data Analysis. RNA samples (see Supporting Text, which is published as supporting information on the PNAS web site, for a detailed protocol) were processed for hybridization on Affymetrix HG-U133A microarrays following standard procedures as recommended by the manufacturer. Microarray data were retrieved as mas5 flat files and imported into genespring 6.1 (Silicon Genetics, Redwood City, CA) or excel (Microsoft).

Real-Time PCR. Changes in the expression of target genes were examined with real-time PCR. The full-length cDNA sequences for genes of interest were obtained from the National Library of Medicine (www.ncbi.nlm.nih.gov/UniGene). A detailed description of real-time PCR and other procedures are provided in Supporting Text.

Results and Discussion

TF 20 –VP Induces Phenotypic Transformations in HeLa Cells. We have developed a functional selection strategy to identify and modulate genes involved in tumor progression (Fig. 1A). The ATF is used as a tool to induce a desired phenotype by perturbing endogenous transcription and to further understand the genes that are required for neoplastic disease progression. We hypothesized that a delivery of a TF activator library into a drug-sensitive, poorly invasive, and nonmetastatic cell line would result in the generation of a phenotypically diverse cancer cell population (or cancer cell library). A complex six-ZF activator library (comprising 8.42 × 10⁷ different ATF proteins with unique DNA binding specificities) could regulate multiple genes involved in tumor progression. Indeed, natural TFs have recently been described that are able to induce epithelial-mesenchymal transitions involving dramatic remodeling of the cytoskeleton, loss of epithelial cell adhesion, and acquisition of migratory phenotypes (5). One of these natural TFs, Twist, has also been associated with resistance to taxol (13). Likewise, we hypothesized that some ATF library members would be able to regulate several genes and induce complex malignant phenotypes, including resistance to anticancer drugs and changes in cell morphology, migration, and invasion. Cells in the population displaying complex morphological transformations can be selected with phenotypic screens. The ATF responsible for the phenotypic switch can then be isolated and characterized. Expression profiles of ATF-expressing cells are compared with profiles of the original cell line to determine target genes required for the phenotype. One unique feature of this system is that the phenotype of cancer cells is transformed without changing their genetic background. Instead, ATFs mediate complex changes in gene expression profiles that can facilitate reprogramming a tumor cell phenotype.

Fig. 1.

TF 20 is able to induce complex phenotypes in HeLa cells. (A) Illustration of TF-mediated reprogramming of cancer cells. TF 20-VP induces morphological transformations and cytoskeleton remodeling in HeLa cells. (B) Aspect of a colony of control cells expressing no ZFs. (Magnification: ×100.) TF 20-SKD-transduced cells formed the same WT compact colonies (data not shown). (C and F) Control cells (C) and TF 20-VP-transduced cells (F) were stained for F-actin (Texas red-phalloidin, red) and nucleus (DAPI, blue). (Magnification: ×600.) (D and E) Morphological transformations of HeLa cells transduced with a retrovirus expressing TF 20-VP, showing cells migrating out of the colony. (Magnification: ×100.) (G) TF 20-VP enhances the migration of HeLa cells on a laminin migration assay.

We have chosen HeLa cells as a model system to study genes affecting neoplastic disease progression because this cell line is sensitive to taxol, a commonly used anticancer drug for several carcinomas, and because this cell line constitutes a prototypical model of a noninvasive, poorly metastatic cell type. HeLa cells were transduced with a six-ZF retroviral activator library [pMX-6ZF-library-VP64-IRES-GFP (11)]. Transduced cells were screened for taxol resistance, and drug-resistant clones were morphologically examined for the appearance of more complex morphological transformations, resembling epithelial-mesenchymal transition (5, 6).

Although several ATF clones were isolated that promoted drug resistance, a unique feature of one of the selected ATFs, TF 20-VP, was its ability to induce particularly aggressive phenotypic transformations. When TF 20-VP was expressed in HeLa cells by using either retroviral vectors or transient expression vectors, they acquired elongated, fibroblast-like morphologies (compare Fig. 1 B with D and E) as manifested by the orientation of the actin stress fibers (Fig. 1 C and F). HeLa cells expressing TF 20-VP also displayed enhanced migration in laminin-coated transwell assays, compared with control cells expressing no ZF or cells containing the same ZF linked to a repressor domain (TF 20-SKD) (Fig. 1G). Interestingly, TF 20-mediated effect on cell migration depended on the nature of the extracellular matrix protein used in the assay, as we observed no effect with collagen I, collagen IV, or fibronectin (data not shown). Both TF 20-VP and TF 20-SKD were expressed in HeLa cells at similar levels, as determined by semiquantitative RNA expression analysis and flow cytometry (Fig. 5, which is published as supporting information on the PNAS web site).

TF 20 –VP Induces Cell Invasion in Vitro and Enhances the Number of Distal Metastases in NOD SCID Mice. The enhanced migratory/epithelial-mesenchymal transition type phenotype of cells expressing TF 20-VP suggested altered invasive behavior. To test TF 20-VP invasiveness in vitro, we performed Matrigel invasion assays. HeLa cells transduced with a control retrovirus without ZFs and cells transduced with TF20-SKD were poorly invasive in this assay. However, TF 20-VP was able to enhance cell invasion in vitro 20-fold relative to cells transduced with a control retrovirus (Fig. 2A). Invading cells conserved the morphological signatures of TF 20-VP-expressing cells seen in Fig. 1 D and E.

Fig. 2.

TF 20-VP enhances cell invasion and metastasis. (A) Invasion assays with control and TF-transduced cells were performed in vitro by using Matrigel chambers. (B) TF 20-VP increased the number of lung micrometastases in a NOD SCID mouse model of spontaneous metastasis. Groups represent HeLa cells expressing TF 20 VP64 activator domain (TF 20-VP), the same protein but linked to a repressor domain (TF 20-SKD), and cells expressing no ZF domains (control). P values between TF 20-VP and control and between TF 20-VP and TF 20-SKD were <0.05; P values between TF 20-SKD and control were >0.4. (Magnifications: ×200.)

Because invasive phenotypes are thought to be critical to a cell's ability to metastasize we investigated whether or not TF 20-VP was able to promote metastasis in a mouse model. Approximately 10⁶ HeLa cells transduced with a control virus, TF 20-VP, or TF 20-SKD were implanted s.c. into NOD SCID mice (n = 7). Semiquantitative RT-PCR analysis using TF-specific primers and GFP analysis by flow cytometry showed that tumors transduced with either TF 20-VP or TF 20-SKD expressed TFs with similar expression levels and that strong TF expression persisted through the time course of the experiment (Fig. 5A). At day 35 postinjection mice were killed and the number of lung micrometastases was determined (Fig. 2B). Mice implanted with TF 20-VP-transduced cells developed more lung metastases than mice implanted with control cells (transduced with retrovirus without ZFs) or cells transduced with TF 20-SKD, suggesting that TF 20-VP modulated the expression of genes enhancing the ability of tumor cells to produce metastasis in the lungs. TF 20–VP-transduced cells had a lower proliferation index in vitro than control cells or TF 20–SKD-transduced cells. In addition, TF 20-VP did not promote primary tumor growth compared with control tumors, whereas TF 20–SKD inhibited tumor growth in vivo (Fig. 6, which is published as supporting information on the PNAS web site). This finding is consistent with recent reports showing that metastatic potential is not always correlated to the number of cells in the primary tumor or tumor size (1).

TF 20 –VP Regulates Specific Gene Expression Signatures. We next evaluated the altered transcriptional profile mediated by TF 20-VP by determining which genes are differentially regulated by this ATF. We prepared duplicate independent transductions of HeLa cells with TF 20-VP, TF 20-SKD, and control retrovirus. Untransduced cells were also evaluated. RNA expression profiles were analyzed by using a HG-U133A array from Affymetrix with ≈18,500 genes. Eight genes, listed in Table 1, which is published as supporting information on the PNAS web site, appeared to be differentially expressed in TF 20-VP-transduced cells compared with control and TF 20-SKD groups. Quantitative real-time expression analysis was used to verify that five of these eight genes were differentially regulated in TF 20-VP-expressing cells only (Table 2, which is published as supporting information on the PNAS web site). Three of these genes were highly regulated by TF 20-VP: E48 antigen (E48), AGT, and IL-13 receptor α1 (IL-13Rα1).

E48 is a glycosylphosphatidylinositol-anchored molecule that plays a role in cell–cell adhesion. E48 (LY-6D) belongs to the gene family of Ly-6 antigens (14, 15). E48 is highly expressed in squamous cell carcinomas of the head and neck and constitutes a marker of disseminated tumor cells, particularly in lymph nodes and bone marrow (16, 17). These tumors are characterized by local invasion resulting in poor prognosis. AGT is a precursor of AgtII, a cell-signaling molecule associated with a variety of disorders, such as cardiovascular remodeling and cancer (18–20). IL-13Rα1 is a receptor for IL-13 and IL-4 and has been shown to mediate signaling processes that result in activation of the Jak1–STAT pathway. Interestingly, strong expression of this target has been detected in 16 cell lines for squamous cell carcinomas of the head and neck (21–23).

A mAb that detects human E48 antigen was used to confirm a 500- to 1,000-fold induction of E48 antigen in HeLa cells transfected with TF 20-VP. In contrast, untransduced HeLa cells and cells transduced with TF 20-SKD did not express significant E48 antigen. Immunofluorescence of TF 20-VP-transduced HeLa cells confirmed the E48 up-regulation, particularly in cell–cell junctions (Fig. 3C). We detected 8- to 10-fold up-regulation of IL-13Rα1 both by DNA arrays and real-time PCR. In addition to these three strongly up-regulated markers, we observed a 3.1-fold up–regulation of the gene ABCC5, a multidrug resistance ABC transporter and a 4.7-fold up–regulation of a cDNA of unknown function (cDNA: FLJ22642 fis, clone HS106970). To study the specificity of gene regulation mediated by TF 20-VP, we analyzed expression levels of these five genes in cells expressing four other unrelated TFs. As shown in Fig. 7, which is published as supporting information on the PNAS web site, these five genes were specifically up-regulated in only TF 20-VP-expressing cells.

Fig. 3.

TF 20-VP regulates the endogenous E48 gene, a marker of disseminated tumor cells of squamous cell carcinomas of the head and neck. (A) HeLa cells were analyzed for E48 expression 72 h posttransduction. (B) HeLa cells recovered from mouse primary tumors 35 days postinjection. Flow cytometry analyses were performed with a mAb that detects human E48. E48 is up-regulated in HeLa cells transduced with a TF 20-VP retrovirus (red); HeLa cells expressing TF 20-SKD (light blue); control cells expressing no ZFs (green); untransduced cells (filled blue). (C) Immunofluorescence analysis of E48 expression in HeLa cells expressing TF 20-VP, control cells expressing no ZFs (control), and cells overexpressing an E48 cDNA (E48). Expression of E48 was induced in the cell–cell junctions (arrow). (Magnifications: ×400.) (D) Ectopic expression of E48 induces cell migration in Laminin-coated transwells. HeLa cells were transduced with EGFP, TF 20-VP, and E48 cDNA (E48). Untransduced (HeLa cells) were also evaluated. (E) HeLa cells transduced with TF 20-VP, TF 20-SKD, and E48 retroviruses were injected i.v. in chicken embryos; disseminated tumor cells were detected in distal organs (lung and lower chorioallantoic membrane, CAM) by real-time alu-PCR as described (27). HeLa cells expressing both TF 20-VP and E48 cDNA enhance 10- to 20-fold and 5-fold, respectively, the number of experimental metastasis in a chicken embryo model of organ colonization.

Invasiveness and Tumor Cell Dissemination Requires Regulation of Multiple Targets. Our functional analysis focused on the three genes most strongly and specifically up–regulated by TF 20–VP: E48, AGT, and IL-13Rα1. However, additional experiments are warranted to evaluate the specific contribution of ABCC5 and the gene of unknown function. To extend our expression analysis to the in vivo model, we first determined whether or not E48, AGT, and IL-13Rα1 were differentially regulated in the primary tumors from NOD SCID mice implanted with HeLa cells expressing TF 20-VP. Cells were recovered from tumors 2 and 6 weeks postinjection. Cells expressing GFP were sorted by flow cytometry, and gene expression was analyzed by real-time PCR. E48 and AGT were strongly expressed in the tumors at 2 and 6 weeks (Table 2). However, we did not detect significant up-regulation of IL-13Rα1 in tumors at either time point. One possibility is that IL-13Rα1 plays a role in the initial steps of tumor progression and is silenced in tumors by 2 weeks postinjection. In contrast, E48 expression was increased by three orders of magnitude in cells derived from TF 20-VP-containing tumors at both 2 and 6 weeks compared with control or TF 20-SKD tumors, as assessed by real-time RNA quantification and flow cytometry (Table 2 and Fig. 3B).

TF 20 VP-mediated induction of E48 expression in vitro and its increased expression in late stages of tumor development in vivo suggested a role of E48 in tumor progression. Several glycosylphosphatidylinositol-anchored proteins have been shown to promote cytoskeleton reorganization, changes in cell shape, cell attachment and extracellular matrix-specific migration in neutrophil cells (24), pre-B lymphocytes (25), and breast carcinomas (26). These effects depend on a cross-talk between the glycosylphosphatidylinositol-containing protein and specific integrins. In light of these observations, we first investigated the possibility that E48 expression could influence cell migration. As indicated in Fig. 3D, ectopic expression of E48 cDNA in HeLa cells induced extracellular matrix-dependent cell migration. As in TF 20-VP-expressing cells, E48 induced migration in laminin-coated transwells (Fig. 3D), but not in collagen or fibronectin-coated wells, suggesting an involvement of specific integrin signaling. We also observed that, like TF 20-VP transduced cells, cells that overexpressed E48 had elongated fibroblast-like cell morphologies (data not shown), suggesting that E48 participates in the cytoskeleton-remodeling characteristic of TF 20-VP-expressing cells.

To better understand the function of E48 and its potential involvement in promoting tumor progression in vivo, we studied experimental metastasis formation in chicken embryos. The chicken embryo system has two important advantages for studying the role of potential TF-targeted cDNAs: metastases can be generated and analyzed much more quickly than in mouse models (7 days in chickens versus 35 days in mouse) and methodology exists to allow precise time course-dependent quantification of disseminated tumor cells by using real-time PCR to detect alu-human sequences in chicken organs (27). In the experimental metastasis system described in Fig. 3E, control HeLa cells and cells transduced with TF 20-VP, TF 20-SKD, or a retroviral construct expressing human E48 cDNA (E48) were injected i.v. into the allantoic vein, and disseminated cells were detected by quantitative PCR. In this model, TF 20-VP was able to enhance the process of organ colonization in the lungs and the lower chorioallantoic membrane by 10- and 20-fold, respectively, compared with control and TF 20-SKD cells. Furthermore, E48-expressing cells were able to increase by 5-fold the colonization of HeLa cells in the lower chorioallantoic membrane relative to the control. Thus, these data confirmed the role of TF 20-VP in promoting organ colonization and also suggested a role of E48 antigen in enhancing tumor cell dissemination in specific organs. Nevertheless, E48 by itself was not able to fully recapitulate the TF 20-VP effect in tumor cell dissemination, suggesting that the fully metastatic phenotype conferred by TF 20-VP requires concerted and synergistic action of several targets.

Subsequently, we analyzed the possibility that coexpression of TF 20-VP-targeted genes could cooperate in generating the invasive phenotype of TF 20-VP-expressing cells. We introduced the cDNAs coding for E48, AGT, or IL-13Rα1 into HeLa cells by using transient expression vectors, and transfected cells were evaluated by using in vitro invasion assays. Expression of these proteins was efficiently up-regulated, as confirmed by real-time quantification. As shown in Fig. 4A, transfection of individual targets had little impact on cell invasion. However, the simultaneous expression of AGT, E48, and IL-13Rα1 resulted in a cooperative effect that enhanced the ability of HeLa cells to invade the matrix. Interestingly, cotransfection of E48 and AGT recapitulated the elongated phenotype and the cytoskeleton-remodeling characteristic of TF 20-VP expression (Fig. 4B). Cotransfection of IL-13Rα1 did not enhance cell shape and affected only the efficiency of invasion. Nevertheless, cells transiently transfected with TF 20-VP were 2.6-fold more efficient in promoting cell invasion than E48, AGT, and IL-13Rα1 cDNAs combined. This result further confirms that TF 20-VP regulates multiple gene targets that are cooperatively necessary for cell invasion. The failure of cDNA transfection to fully recapitulate the phenotype provided by TF 20-VP might also be attributed to contributions provided by splice variants that are not produced under cDNA delivery.

Fig. 4.

In vitro induction of cell invasion requires coexpression of multiple targets. (A) HeLa cells were transiently transfected with individual cDNAs encoding E48, AGT, IL-13Rα1, and TF 20-VP. Transfected cells were loaded into Matrigel chambers, and invading cells were fixed and counted. Values represent averages of two wells, and experiments were done in triplicate. (B) E48 and AGT coexpression in HeLa cells suffices to recapitulate the cell morphology changes mediated by TF 20-VP. (Magnification: ×400.)

It is possible that other gene products up-regulated by TF 20-VP but not detected in this study are necessary to fully recapitulate the TF 20-VP phenotype. By binding to multiple targets, a TF can orchestrate and regulate the expression of each target gene, and its splice variants, at the proper level. Unlike heterologous cDNAs that express a single splice variant, TFs use the genomic scaffold that provides endogenous elements and spatiotemporal cues necessary for target gene regulation. A biochemical analysis of the E48, AGT, and IL-13Rα1 gene promoters has revealed functional TF 20 binding sites for the AGT and IL-13Rα1 proximal promoters, suggesting direct regulation. The E48 reporter gene was not regulated by a TF 20 site found upstream of the proximal promoter and could be regulated indirectly by another gene or a more distal site (Fig. 8, which is published as supporting information on the PNAS web site). Although our experiments analyzed the effect of a limited number of targets on the process of cell invasion in vitro, in vivo these genes may contribute to any one of many different steps of the metastatic cascade such as mobility, invasion of surrounding tissues, intravasation or extravasation and survival in the vasculature, or colonization and survival in distal organs.

Overall, the above experiments provide evidence that TF 20-VP regulates expression of E48, AGT, and IL-13Rα1. Real-time PCR experiments on HeLa cells showed that overexpression of each given target individually did not affect transcription of the other markers (Table 2). This finding indicates that TF 20 affects transcription of these three genes autonomously.

Several reports have suggested an association of E48 antigen expression with increased tumor progression (28–31). The E48 antigen is highly expressed in locally invasive squamous cell carcinomas of the head and neck and constitutes a marker of disseminated tumor cells both in lymph nodes and bone marrow (32). Recently, the presence of micrometastatic cells in bone marrow of patients with squamous cell carcinomas of the head and neck with two or more lymph node metastases has been correlated with a poor metastasis-free survival (33). It has been hypothesized that E48 is a signal transduction protein that plays a role in cell–cell communication, and recent reports suggest involvement in mediating selectin-dependent binding of premetastatic tumor cells to the endothelium (28). Nevertheless, the signal transduction pathway and the adhesion system activated by E48 was unknown. Our data demonstrate a role of E48 in promoting matrix-dependent migration of tumor cells, suggesting integrin-dependent signaling. Additionally, we found that E48 is able to facilitate organ colonization in in vivo models, thus confirming the role of this antigen in tumor dissemination.

Expression of AGT, a precursor of the signaling molecule AgtII, is regulated in a tissue-specific manner (34–37). In heart tissue, AgtII activates Jak1, Jak2, and Tyk2, resulting in activation of STATs (38). In addition, AgtII can influence important cellular processes, such as cytoskeleton remodeling, migration, and survival (19).

Signaling mediated by IL-13R involves binding of cytokines IL-13 and IL-4 to the receptor and triggers activation of two pathways: the Jak–STAT pathway and the phosphatidylinositol 3-kinase pathway (39). One intriguing possibility is that some AgtII- and IL-13R-signaling cascades are necessary for the TF 20-VP-mediated metastatic phenotype. In this regard, TF 20-VP-expressing cells offer an experimental framework to study the effect of activators and inhibitors of signaling pathways in the process of cell invasion and metastasis. Our preliminary experiments indicate that TF-20-VP-mediated cell invasion in vitro is stimulated by AgtII and dramatically abolished by a Jak2-specific inhibitor (Fig. 9, which is published as supporting information on the PNAS web site). This finding supports the possibility that Jak2 signaling is important for the TF 20-VP-mediated invasive phenotype. Recently, activation of the Jak2–STAT pathway has been shown to influence survival, invasion, and metastasis in lymphomas (40). We expect that TF-mediated dysregulation of critical signaling molecules will allow investigators to transform the phenotype of tumor cells or “reprogram” tumor cells.

In this study we have described an ATF, TF 20-VP, selected from a combinatorial six-ZF TF library that is able to alter phenotypic properties of HeLa cells. HeLa cells are taxol-sensitive, noninvasive, and nonmetastatic. However, upon delivery of TF 20-VP these cells became drug-resistant, invasive, and metastatic. We demonstrated, using in vivo models, that the ability to enhance metastasis and dissemination of tumor cells depended on linkage to an activator domain. The fact that the same selected six-ZF DBD domain linked to a repressor domain or cells not expressing any ZF did not manifest the phenotype, allowed us to perform an analysis of differentially expressed targets. We validated, by real-time PCR, a group of five genes whose expression was altered only in TF 20-VP-expressing cells. Recently, an increasing number of genetic profiling experiments have been reported with the ultimate goal of identifying genetic markers involved in cancer progression and metastasis (1–3). These studies often compared tumors from different genetic backgrounds and the resulting heterogeneity of the samples complicated the functional analysis of the targets. The use of ATFs to artificially modify cancer cell phenotypes is appealing because the TF introduces a transcriptional perturbation of gene expression but the genetic background of the cell line remains the same. TFs selected from combinatorial libraries can be used to identify generic biomarkers of tumor progression. Of the five differentially expressed genes reported here, three were cell surface proteins, illustrating the power of this approach in the identification of cell surface antigens involved in tumor progression.

Functional studies suggest that three of the up-regulated markers, E48, AGT, and IL-13Rα1, contributed synergistically in recapitulating some aspects of the TF 20-VP-induced phenotype. This phenomenon is consistent with the fact that increased malignancy of human tumors seems to require the action of several genes operating in concert. Mechanistically a TF can achieve simultaneous regulation of several genes by binding multiple promoters. This ability to regulate multiple defined genes is a distinctive feature of TFs, compared with other strategies to modulate gene expression, such as ribozymes, antisense, and RNA interference. Whereas the latter strategies typically target a specific RNA sequence from a single gene, ATFs can bind similar or identical DNA sequences located in several regulatory regions, facilitating targeting of multiple genes. Additionally, ATFs can induce either gain-of-function phenotypes or knockdown phenotypes, whereas RNA-derived strategies can achieve only knockdowns.

This work demonstrates that ATFs selected from combinatorial libraries can be used to transcriptionally reprogram cancer cells to modify complex phenotypes and identify genes involved in cancer progression. ATF library screens could also be used to interfere with the regulation of genetic or signaling cascades to modify or revert certain aspects of a malignant phenotype. In recent experiments, we observed that TF 20-SKD is able to efficiently reduce cell invasion in a highly metastatic melanoma cell line (Fig. 10 which is published as supporting information on the PNAS web site). Together with the xenograph model (Fig. 6), these experiments demonstrate the use of ATFs to negatively effect cell invasion and tumor growth. In summary, we have shown that ATFs can be used as tools to dissect the function of genes in tumor progression. This work also suggests potential applications of ATFs in cancer therapy.