Regulation of c-Met-dependent gene expression by PTEN - Nature

Oncogene (2004) 23, 9173–9182

& 2004 Nature Publishing Group All rights reserved 0950-9232/04 $30.00 www.nature.com/onc

Regulation of c-Met-dependent gene expression by PTEN Roger Abounader*,1,2, Thomas Reznik2, Carlo Colantuoni3, Francisco Martinez-Murillo4, Eliot M Rosen5 and John Laterra*,1,2,3,6 1 Department of Neurology, School of Medicine, Johns Hopkins University, Baltimore, MD, USA; 2Kennedy Krieger Research Institute, Baltimore, MD, USA; 3Department of Neurosciences, School of Medicine, Johns Hopkins University, Baltimore, MD, USA; 4Department of Molecular Biology and Genetics, School of Medicine, Johns Hopkins University, Baltimore, MD, USA; 5 Department of Oncology, Lombardi Cancer Center, Washington, DC, USA; 6Department of Oncology, School of Medicine, Johns Hopkins University, Baltimore, MD, USA

Receptor tyrosine kinases (RTK) and the tumor suppressor PTEN co-regulate oncogenic cell signaling pathways. How these interactions influence gene transcription is inadequately understood. We used expression microarrays to investigate the effects of PTEN on gene expression changes caused by activating c-Met in human glioblastoma cells. c-Met activation by scatter factor/hepatocyte growth factor (SF/HGF) altered the expression of 27-fold more genes in PTEN-null U-373MG cells than in PTEN homozygous primary normal human astrocytes (523 vs 19 genes). Restoring wt-PTEN in U-373MG cells dramatically altered patterns of c-Met regulated gene expression. This effect was varied depending on the specific gene in question. PTEN reduced the number of c-Met regulated transcripts from 931 to 502, decreased the relative number of genes upregulated by c-Met from 46 to 25%, and increased the relative number of downregulated genes from 54 to 75%. PTEN and c-Met co-regulated many genes involved in cell growth regulation such as oncogenes, growth factors, transcription factors, and constituents of the ubiquitin pathway. c-Met activation in PTEN-null (but not PTEN reconstituted) cells led to upregulation of the EGFR agonist TGFa and subsequently to EGFR activation. Using PTEN mutants, we found that PTEN’s transcriptional effects were either lipid-phosphatase dependent, protein-phosphatase dependent, or phosphataseindependent. These results show that PTEN has critical and mechanistically complex effects on RTK-regulated gene transcription. These findings expand our understanding of tumor promoter/suppressor inter-relationships and downstream transcriptional effects of PTEN loss and c-Met overexpression in malignant gliomas. Oncogene (2004) 23, 9173–9182. doi:10.1038/sj.onc.1208146 Published online 1 November 2004 Keywords: receptor tyrosine kinase; tumor suppressor; hepatocyte growth factor; scatter factor; microarrays; epidermal growth factor receptor

*Correspondence: R Abounader and J Laterra, Kennedy Krieger Research Institute, 707 N. Broadway, Room 400, Baltimore, MD 21205, USA; E-mail: [email protected] or [email protected] Received 21 July 2004; revised 30 August 2004; accepted 30 August 2004; published online 1 November 2004

Introduction Oncogenes and tumor suppressors represent key steps in the regulatory pathways controlled by soluble growth factors. The activation of oncogenes or inactivation of tumor suppressors leads to the dysregulation of growth regulatory pathways and malignant progression. Many oncogenes are tyrosine kinases that, when activated, trigger downstream phosphorylation cascades that mediate their functions. The phosphorylation of signal transduction molecules is regulated by phosphatases that can function as tumor suppressors. The deregulation of interactions between kinases and phosphatases therefore plays an important role in tumor formation, growth and metastasis. The proto-oncogene product and tyrosine kinase receptor c-Met and the tumor suppressor phosphatase PTEN are representatives of the tyrosine kinase/ phosphatase and oncogene/tumor suppressor regulatory systems. c-Met is activated by its only known ligand scatter factor/hepatocyte growth factor (SF/HGF). cMet and SF/HGF are expressed in a large variety of normal human tissues where they mediate epithelial/ mesenchymal interactions and play important roles in morphogenesis, organogenesis and tissue regeneration (DiRenzo et al., 1991). c-Met and SF/HGF have also been linked to oncogenesis and malignancy in a wide variety of neoplasms. c-Met and SF/HGF are frequently overexpressed or mutated in human tumors including gliomas where they promote malignant progression by inducing cell cycle progression, tumor cell migration, tumor angiogenesis and cell survival (Rosen et al., 1996; Jeffers et al., 1997; Koochekpour et al., 1997; Laterra et al., 1997; Abounader et al., 1999, 2002; Bowers et al., 2000; Birchmeier et al., 2003). PTEN is a tumor suppressor protein that possesses lipid and protein phosphatase dependent as well as phosphatase-independent activities (Steck et al., 1997; Stambolic et al., 1998; Wu et al., 1998; Gu et al., 1999; Maier et al., 1999; Leslie et al., 2001). PTEN also binds to p53 and regulates its function (Stambolic et al., 2001). PTEN is mutated in various human neoplasms including high-grade gliomas (Teng et al., 1997; Wang et al., 1997; Sano et al., 1999). Reintroduction of PTEN to mutant cells can inhibit

Regulation of c-Met-dependent gene expression by PTEN R Abounader et al

9174

tumor growth by suppressing proliferation and promoting tumor cell apoptosis (Cheney et al., 1998). Besides their generally opposing effects on tumorigenicity and malignant progression, there is evidence of direct interactions between SF/HGF:c-Met and PTEN at the second messenger cell signaling level. The best documented is the lipid phosphatase-dependent antagonistic effect of PTEN on phosphatidylinositol-3-kinase (PI 3-kinase), which is a prominent component of the signaling cascades activated by c-Met (Maehama and Dixon, 1998; Wu et al., 1998). PTEN protein phosphatase is also able to dephosphorylate the p52 isoform of Shc, thus inhibiting the recruitment of the Grb2 adaptor to activated c-Met receptors and the subsequent activation of the MAP kinase cascade, another important downstream effector of c-Met (Gu et al., 1999; Xiao et al., 2001). PTEN protein phosphatase could theoretically dephosphorylate other signaling molecules downstream of c-Met. PTEN also possesses phosphatase-independent activities that, while less-well characterized, might alter other receptor tyrosine kinase-dependent actions (Maier et al., 1999; Leslie et al., 2001; Raftopoulou et al., 2004). c-Met and PTEN affect glioma malignancy either directly via associated cell signaling and/or indirectly by changing the expression of neoplasia regulating genes through changes in expression levels of transcription factors. Transcription factors known to be affected by SF/HGF include AP-1, c-Myc, Ets-1, NFk-b and HIF-1a. We and others have shown that PTEN inhibits c-myc induction, NFk-b and HIF-1 transcription factors. PTEN and receptor tyrosine kinases (RTK) can therefore coregulate the expression of numerous genes by regulating the same second messengers and transcription factors. Little is known of the detailed interactions and interdependencies between c-Met and PTEN in human brain tumors. Gene expression coregulation by c-Met and PTEN in human gliomas, indeed in any biological system, has not been adequately explored. In the present study, we used DNA microarrays to comprehensively investigate the effects of PTEN on SF/HGF-mediated gene expression changes in human glioblastoma cells. We found that SF/HGF alters the expression of substantially more genes in PTEN-null glioblastoma cells than in normal human astrocytes (NHA) that express wild-type PTEN. Furthermore, reconstitution of wild-type PTEN in PTEN-null glioblastoma cells dramatically changes SF/HGF-induced gene expression profiles in a complex manner involving phosphatasedependent as well as phosphatase-independent activities. We conclude that PTEN has critical and heterogeneous effects on receptor tyrosine kinase regulated gene expression in tumor cells. Results

U-373MG human glioblastoma cells. NHA and U373MG cells were treated with SF/HGF prior to mRNA extraction and hybridization to microarrays chips as described in the Materials and methods section. SF/ HGF altered the expression of 523 genes in U-373MG cells (314 genes were upregulated and 209 genes downregulated). In contrast, SF/HGF only changed the expression of 19 genes in primary normal human astrocytes (10 genes were upregulated and nine genes were downregulated) (Table 1a). We and others have shown that NHA express c-Met mRNA and protein and that treatment of the cells with SF/HGF leads to c-Met, MAPK and Akt phosphorylation, indicative of a functional c-Met receptor in these cells (Figure 1a) (Koochekpour et al., 1997). PTEN alters SF/HGF-regulated gene expression profiles Gene transfer experiments frequently lead to artificially high expression of the transgenic protein. This can lead to unphysiological effects that do not reflect normal functions. To exclude this possibility for our conditions of PTEN restoration used for microarray experiments, we compared transgenic PTEN protein levels achieved in U-373MG cells infected with PTEN-expressing adenoviruses to endogenous PTEN levels in NHA using immunoblotting. Levels of transgenic PTEN protein in U-373MG cells were of comparable magnitude to the levels of endogenous PTEN found in NHA cells (Figure 1b). We hypothesized that the markedly different gene expression response to SF/HGF in NHA and U-373MG cells was due, in part, to the absence of functional PTEN protein in U-373MG glioma cells. Therefore, we examined the effect of reconstituting wild-type PTEN on SF/HGF induced gene expression in U-373MG cells. PTEN was reconstituted to U-373MG cells by infection with PTEN-expressing adenoviruses (Ad-PTEN), while controls were infected with control adenoviruses (Ad-control). Cells were then treated with SF/HGF

Table 1 (a) Number of genes that are SF/HGF reponsive in normal human astrocytes vs U-373MG glioma cells based on Incyte microarray analysis. (b) Numbers and expression change directions of SF/HGF-responsive genes before and after PTEN reconstitution to U-373MG-cells as analysed on Affymetrix microarrays (a) U-373MG

Normal Human Astrocytes

Upregulated by SF/HGF: 314 Downregulated by SF/HGF: 209

Upregulated by SF/HGF: 10 Downregulated by SF/HGF: 9

(b) PTEN-null U-373MG

PTEN-reconstituted U-373MG

Upregulated by SF/HGF: 429

Unaffected: 367/429 Upregulated: 20/429 Downregulated: 42/429 Unaffected: 498/502 Upregulated: 3/502 Downregulated: 1/502 Upregulated: 103 Downregulated: 333

SF/HGF alters the expression of many more genes in glioblastoma cells than in normal astrocytes

Downregulated by SF/HGF: 502

We used Incyte gene microarrays to compare the effects of SF/HGF on gene expression profiles in NHA and

Unchanged

Oncogene


9175

Figure 1 (a) Normal human astrocytes (NHA) express functional c-Met protein. Immunoblots showing c-Met protein in NHA (upper panel) and induction of MAPK (p42/44, ERK) phosphorylation in NHA cells after treatment with 10 ng/ml SF/HGF for 30 min. (b) Expression levels of PTEN in U-373MG cells after PTEN restoration are of comparable magnitude with levels found in NHA cells. Immunoblot showing PTEN protein levels in U-373MG cells infected with either PTEN expressing adenoviruses (Ad-PTEN) or control adenoviruses (Ad-control) as compared to PTEN endogenous protein levels in NHA cells

prior to RNA extraction and hybridization to Affymetrix microarray chips as described in the Materials and methods. Restoration of wild-type PTEN in U-373MG cells dramatically altered the profile of SF/HGFresponsive genes. PTEN reconstitution predominantly reduced the overall number of SF/HGF-responsive transcripts and decreased the relative number of genes that were upregulated by SF/HGF (Figure 2). In PTENnull cells, SF/HGF significantly altered the expression of 931 genes (429 were upregulated and 502 downregulated). The percent of genes responsive to SF/HGF in U-373MG cells relative to the total number of genes found on the microarrays was comparable between Affymetrix and Incyte experiments (5.8 vs 6.0%, respectively). After PTEN reconstitution, SF/HGF significantly altered the expression of only 459 genes and substantially reduced the relative number of genes upregulated by SF/HGF (123 were upregulated and 336 downregulated). The effects of PTEN reconstitution on SF/HGF-induced gene expression were in different directions (Table 1): (1) Many genes induced by SF/ HGF in PTEN-null cells were no longer SF/HGFresponsive after PTEN reconstitution; (2) A number of genes found to be unresponsive to SF/HGF in PTENnull cells were responsive to SF/HGF after PTEN reconstitution; (3) PTEN reconstitution amplified the magnitude of SF/HGF-responsiveness of some genes; and (4) PTEN reversed the direction of SF/HGFinduced expression changes in a minority of genes (i.e. genes that were upregulated by SF/HGF in PTEN-null cells became downregulated in PTEN-restored cells and vice versa). Baseline gene expression changes induced by PTEN restoration to U-373MG cells (without SF/ HGF treatment) were analysed by comparing gene expression profiles of cells infected with wtPTEN with cells infected with control adenoviruses. PTEN alone changed the expression of 297 genes (149 were

Figure 2 PTEN reconstitution blunts SF/HGF-induced gene expression changes and increases the relative number of genes downregulated by SF/HGF. Scatter plots showing all SF/HGF responsive genes in U-373MG cells before and after PTEN reconstitution. Each dot represents one gene. Genes are randomly distributed along the x-axis. The two horizontal lines delimit the area of significant change (3>Z>3 or fold change >2) as defined in the Materials and methods section. Dots lying outside the horizontal lines represent genes that are significantly changed

upregulated and 148 were downregulated). Of these 297 genes, only 33 were identical to genes found to be co-regulated by PTEN and SF/HGF as described above. Thus, the effects of PTEN reconstitution on baseline gene expression cannot account for the coregulatory effects of PTEN on SF/HGF induced gene expression regulation. PTEN has been described to promote cell death and inhibit cell proliferation under certain conditions. To determine if the gene expression changes induced by PTEN restoration could be secondary to such effects, we reconstituted PTEN to U-373MG cells under the same conditions used for the microarray experiments and analysed the cells for viability with trypan blue staining and for cell proliferation with flow cytometry and cell counting. Compared to control transfected cells, PTEN did not significantly alter cell viability (94 vs 91% viable cells in control and PTENrestored cells, respectively) nor did it inhibit cell proliferation (not shown). Oncogene


9176

Identity of gene products coregulated by SF/HGF and PTEN Many of the genes found to be coregulated by SF/HGF and PTEN are related to neoplastic processes. Examples include genes coding for various oncogenes, growth factors, cell signaling molecules, transcription factors, proteasome/ubiquitination enzymes, and other gene products that regulate cell cycle, apoptosis, cell migration and invasion (Table 2). The predominant trend was for PTEN restoration to downregulate SF/HGF-induced tumor promoting genes and to upregulate

expression of tumor inhibitory genes. In fewer cases, PTEN restoration led to the opposite. A number of genes were found to be coregulated by PTEN and SF/HGF in patterns suggesting functional co-dependencies in contributing to the glioma malignant phenotype (Table 2). Among these are the EGFR agonists TGF-alpha and diphtheria toxin receptor. These genes were induced by SF/HGF in PTEN null cells and induction was inhibited by PTEN restoration. This suggests that in PTEN-null cells, SF/HGF could also enhance malignancy by activating the EGFR pathways. Another gene induced by SF/HGF in

Table 2 Examples of genes that are co-regualted by PTEN and SF/HGF Gene name

Accession #

Description

Upregulated by SF/HGF in PTEN-null, unchanged after PTEN reconstitution Transforming growth factor-alpha (TGF-alpha) E2F transcription factor 1 (E2F1) Diphtheria toxin receptor (heparin-binding epidermal growth factor-like growth factor) Transcription factor AP-2 beta Rho GTPase activating protein 4 GDNF family receptor alpha 2 Tyrosine protein kinase (Jak3B) splice variant Mitogen-activated protein kinase kinase kinase kinase 1 CLL-12 transcript of unrearranged immunoglobulin Tumor-associated calcium signal transducer 1

NM_005225 NM_001945 NM_001945

EGFR receptor agonist Cell cycle regulation EGFR receptor agonist

NM_003221 NM_001666 NM_001495 U31601 NM_007181 X58397 NM_002354

Transcription factor Regulation of Rho GTPase cRET activation Jak family kinase JNK/SAPK activator Oncogene Tumor-associated antigen

Upregulated by SF/HGF in PTEN-null, upregulated after PTEN reconstitution Cyclin D1 Interleukin 8 C-terminal variant (IL8)

M73554 AF043337

Cell cycle regulation

Hs.25960 NM_000588 NM_017717 NM_006419 NM_012116 NM_020128 NM_006260

Oncogene Apoptosis inhibitor Cell adhesion Chemotaxis EGFR signaling regulator Overexpressed in transformed cells Inhibits antiproliferative action of interferon

AB055804 NM_003299 NM_003592 NM_006042 Hs.9075 NM_005534 Hs.295726 M18468

Suppressor of Myc transcriptional activity Heat shock protein (antitumorigenic) E3 ubiquitin ligase Silencing associated with cancer development Induces apoptosis Antitumor effects Angiogenesis and metastasis Growth regulation

Downregulated by SF/HGF in PTEN-null, downregulated after PTEN reconstitution SPARC-like 1 NM_004684 Inhibitor of DNA binding 3, dominant-negative helix–loop–helix NM_002167 protein

Downregulated in cancer Regulation of cell growth and differentiation

Upregulated by SF/HGF in PTEN-null, downregulated after PTEN reconstitution v-Myc avian myelocytomatosis viral related oncogene Interleukin 3 (colony-stimulating factor, multiple) (IL3) Mucin and cadherin-like Small inducible cytokine B subfamily (Cys-X-Cys motif), member 13 Cas-Br-M ectropic retroviral transforming sequence c Nuclear protein double minute 1 (MDM1) Protein-kinase, interferon-inducible double-stranded RNA dependent inhibitor Downregulated by SF/HGF in PTEN-null, unchanged after PTEN reconstitution Myc Modulator-1 Tumor rejection antigen (gp96) Cullin 1 Heparan sulfate 3-O-sulfotransferase Serine/threonine kinase 17a Interferon gamma receptor 2 Integrin, alpha V Protein kinase, cAMP-dependent, regulatory subunit, type I

Upregulated by SF/HGF after PTEN reconstitution, unchanged in PTEN-null Siva-2 Epithelial membrane protein 1

AF033111 NM_001423

Induction of apoptosis Inhibition of cell cycle

Downregulated by SF/HGF after PTEN reconstitution, unchanged in PTEN-null Transmembrane protease, serine 4 Pre-B-cell leukemia transcription factor 1 (PBX1) WNT6 precursor Wilms tumor 1 (WT1), transcript variant D Cyclin-dependent kinase-like 1 (CDC2-related kinase)

NM_016425 NM_002585 AY009401 NM_024426 NM_004196

Tumor invasion DNA binding oncoprotein Wnt signaling Transcription factor Cell cycle regulation

Oncogene


9177

PTEN-null cells is the transcription factor E2F-1 that plays a critical role in mediating cell cycle progression through the G1/S check point. This suggests that PTEN loss can enhance SF/HGF-induced cell cycle progression by upregulating E2F-1. An example of a gene that was downregulated by SF/HGF in PTEN-null cells but not in cells reconstituted with PTEN is Myc Modulator1 that can act as a tumor suppressor by inhibiting c-Myc transcriptional activity (Fujioka et al., 2001). We had previously shown that c-Myc mediates SF/HGF-induced cell cycle progression in PTEN null glioblastoma cells. SF/HGF-mediated induction of cell cycle progression could therefore be amplified by PTEN loss through c-Myc activation. Conversely, the antiapoptotic protein IL-3 was induced by SF/HGF in PTEN-null cells and downregulated by SF/HGF after PTEN restoration. Therefore, IL-3 gene expression change caused by SF/ HGF stimulation and PTEN loss could lead to amplified antiapoptotic effects in human glioblastoma. These patterns are consistent with PTEN functioning to modulate the tumor promoting gene expression effects of c-Met activation by SF/HGF. The expression changes of a selected subset of the genes identified in the Affymetrix gene chip study were also confirmed by quantitative Northern blotting. A number of genes (n ¼ 22) displayed similar changes to those seen on the microarrays, indicating a high validity of the array data analysis and results. Three representative examples of Northern verifications are shown in Figure 3. Induction of TGF-a expression by SF/HGF in PTEN-null cells leads to activation of EGFR One of the genes, the expression of which was induced by SF/HGF in PTEN-null U-373MG cells is the EGFR agonist TGF-a. To address the functional relevance of TGF-a upregulation, we studied the effects of SF/HGF and PTEN on EGFR activation. The time course of TGF-a mRNA induction by SF/HGF was first established using Northern hybridization (Figure 4a). TGF-a mRNA was induced by SF/HGF in a time-dependent manner reaching a maximum B10-fold increase at 24 h. EGFR activation was then assessed by quantifying receptor phosphorylation at tyrosine 1068 by

immunoblotting. Treatment of PTEN-null U-373MG cells with SF/HGF led to the delayed phosphorylation of EGFR at tyrosine 1068 at 24 h (Figure 4b). Restoration of PTEN to U-373MG cells prior to SF/HGF treatment led to a substantial inhibition of EGFR phosphorylation (Figure 4c). These findings show that cMet activation in PTEN-null glioma cells leads to induction of EGFR agonist expression and subsequently to EGFR activation. PTEN modulates SF/HGF-mediated gene expression changes through phosphatase-dependent and phosphatase-independent activities We used PTEN mutants and Northern analysis to determine if PTEN affects SF/HGF-regulated gene expression via lipid phosphatase-dependent, protein phosphatase-dependent, or phosphatase-independent mechanisms. PTEN, lipid phosphatase mutant (G129E), phosphatase dead mutant (C124A) or control were restored to U-373MG cells by infection with corresponding adenoviruses (Ad-PTEN, Ad-G129E, Ad-C124A or control). The cells were subsequently treated with SF/HGF and total RNA was extracted. RNA was then subjected to quantitative Northern analysis using radiolabeled cDNA probes. The results show that PTEN affects SF/HGF-induced gene expression changes through heterogeneous mechanisms. Transcriptional regulation was found to require lipid phosphatase, protein phosphatase and phosphataseindependent domains depending on the specific gene examined. In the case of some genes, for example matrix Gla, PTEN inhibited the transcriptional effect of SF/HGF on this gene while the lipid phosphataseand phosphatase-dead PTEN mutants had no effect (Figure 5). This shows that PTEN lipid phosphatase is required for modulating the SF/HGF-mediated change in matrix Gla gene expression. In the case of other genes, for example CYR61, PTEN inhibited SF/HGFmediated induction of expression, an effect that was absent in response to the phosphatase-dead mutant and accentuated in response to the lipid phosphatase-dead mutant (Figure 5). Thus, the effect of PTEN on SF/ HGF-dependent regulation of CYR61 expression involves a protein phosphatase-dependent component, the

Figure 3 Northern verification of gene expression changes found by microarray analysis. U-373MG cells were infected with AdPTEN or Ad-control prior to treatment with 10 ng/ml SF/HGF. Total RNA was isolated and subjected to quantitative Northern hybridization using radiolabeled cDNA probes for specific genes relative to 28S rRNA expression levels. Figure shows three representative genes with similar gene expression changes as measured on the microarrays. Results show averages of two experiments with control (PTEN null, no SF/HGF) set at 100% Oncogene


9178

PTEN (Figure 5). Other genes were regulated in more complex manners showing partial and combined involvement of the various PTEN functional domains. These results show a variable and complex involvement of PTEN domains in regulating SF/HGF-induced changes in gene expression on human glioblastoma cells. Discussion

Figure 4 SF/HGF induction of TGFa expression leads to activation of the EGFR pathway in PTEN-null cells. (a) PTENnull U-373MG cells were treated with 10 ng/ml SF/HGF for different times and TGFa mRNA levels were quantified by Northern analysis. The results show a time-dependent induction of TGFa mRNA by SF/HGF. (b) PTEN-null U-373MG cells were treated with 10 ng/ml SF/HGF (or 10 ng/ml recombinant TGFa as control) for different times and analyzed for EGFR phosphorylation at tyrosine 1068 (relative to total EGFR) by immunoblotting. SF/HGF treatment led to the delayed phosphorylation of EGFR in PTEN-null U-373MG cells. (c) U-373MG cells were infected with Ad-PTEN or Ad-control 24 h prior to treatment with 10 ng/ml SF/HGF for 24 h. Cells were analyzed for EGFR phosphorylation by immunoblotting. The results show that SF/HGF-induced EGFR phosphorylation (normalized to total EGFR measured on the same blots after stripping) was inhibited by PTEN restoration to the cells

effect of which is enhanced by the absence of PTEN lipid phosphatase. In other cases, for example, Rabkinesin 6, PTEN and its mutants modulated the effects of SF/HGF in a similar manner consistent with the involvement of phosphatase-independent functions of Oncogene

Loss of the tumor suppressor PTEN concurrent with the upregulation of receptor tyrosine kinase pathways such as those linked to c-Met and EGFR are associated with malignant glioma formation. These molecular and biochemical changes influence the tumor cell phenotype through post-translational mechanisms and downstream effects on gene expression regulation. The recent interest in developing receptor tyrosine kinase and PTEN-based cancer therapeutics requires a better understanding of how PTEN modulates the downstream transcriptional effects of receptor tyrosine kinase activation. This is the first study that analyses how PTEN and receptor tyrosine kinase-dependent pathways interact to alter gene expression profiles and gene expression regulation in a malignant neoplasm (i.e. human glioblastoma), indeed in any biological system. The results show an intricate and complex interdependency between PTEN and c-Met tyrosine kinase dependent pathways in coregulating gene expression changes. Although PTEN was found to have heterogeneous effects on SF/HGFinduced gene expression changes, it predominantly blunted SF/HGF effects on gene expression. This is consistent with our finding that SF/HGF alters the expression of many more genes in PTEN-null glioblastoma cells than in PTEN positive primary human astrocytes. The results also demonstrate that PTEN acts on receptor tyrosine kinase-dependent gene expression changes in a heterogeneous manner involving phosphatase-dependent and phosphatase-independent modes of action in a gene-specific manner. The study shows that the combined effects of PTEN loss and SF/HGF:c-Met overexpression likely contribute to malignancy by co-regulating the expression of neoplasia regulating genes. The effects of manipulating receptor tyrosine kinase dependent pathways in gliomas could vary depending on the PTEN status of the cells in question and vice versa. DNA microarray studies can yield a high frequency of false positive and/or nonspecific results. To avoid this potential problem and insure a high level of validity and reliability of the differential gene expression results, our experimental design and data analysis followed very stringent criteria. All cell treatments and microarray hybridizations were performed in duplicate and only genes that were significantly changed in both experiments were considered for further analysis. To determine significance, the stringent z-score cutoff >3 or expression ratio >2 were chosen. The expression changes of a subset of genes were also verified by quantitative Northern analysis and the


9179

Figure 5 PTEN regulates SF/HGF-mediated gene expression changes through phosphatase-dependent and phosphatase-independent mechanisms. U-373MG cells were infected with either Ad-PTEN, Ad-G129E (lipid phosphatase-dead), Ad-C124A (phosphatase-dead) or Ad-control prior to treatment with 10 ng/ml SF/HGF. Total RNA was isolated and subjected to quantitative Northern hybridization using radiolabeled cDNA probes for specific genes relative to 28S rRNA expression levels. The results show examples of genes that are regulated by PTEN lipid phosphatase (Matrix Gla), protein phosphatase (CYR61) and phosphatase-independent (Rabkinesin 6) domains of PTEN. Results are expressed as percent change of mRNA expression induced by SF/HGF relative to untreated control

results corresponded well to the changes seen after microarray analysis, confirming the reliability and validity of our data analysis approach. Furthermore, the specific gene expression changes induced by PTEN reconstitution could not be the result of changes in cell growth or viability since no such effects were observed under our experimental conditions. Importantly also, the heterogeneous effects of PTEN mutants in regulating the SF/HGF-induced expression of single genes is a strong indication of a selective action of PTEN at various levels of the signaling cascade downstream of c-Met activation. We have also excluded another potential artificial effect of PTEN restoration that could originate from unphysiologically high reconstituted PTEN levels. Using Western blotting we found that transgenic PTEN protein levels achieved in U-373MG cells infected with adenoviral titers identical to those used for microarray experiments were comparable to PTEN protein levels found in normal human astrocytes. PTEN restoration changed c-Met dependent gene expression in a complex and heterogeneous manner. This response could be explained by the differential actions of PTEN’s distinct functional domains upon multiple regulatory elements of the signaling cascades and transcription factors downstream of c-Met. PTEN possesses lipid phosphatase dependent, protein phosphatase dependent and phosphatase-independent activities that potentially modulate c-Met-dependent signaling at multiple levels. It is well known that PTEN lipid phosphatase antagonizes the effect of SF/HGF on PI 3-kinase, which plays an essential role in c-Metdependent cell signaling. PTEN also binds and dephosphorylates the p52 isoform of Shc. This protein phosphatase-dependent action of PTEN inhibits the recruitment of the Grb2 adaptor and the subsequent activation of the MAP kinase cascade, another key mediator of the malignant effects induced by activation of c-Met (Xiao et al., 2001). Shc directly interacts with the c-Met substrate Gab1. Gab1 associates directly with c-Met via a c-Met-binding domain that is distinct from other known phosphotyrosine-binding domains

(Schaeper et al., 2000). The interactions of PTEN with PI 3-kinase and MAPK pathways, which mediate c-Met signal transduction, could explain the large number of genes that are co-regulated by PTEN and c-Met. It is also possible that PTEN interacts with other signaling molecules downstream of c-Met though its phosphatases or phosphatase-independent activities. PTEN’s effects on c-Met-dependent gene expression are not due to a direct inhibition of c-Met receptor activation. In fact, PTEN restoration to U-373MG cells did not affect SF/HGF-induced c-Met phosphorylation (not shown). Many neoplasia-related genes were co-regulated by PTEN and SF/HGF. In the absence of PTEN, SF/HGF frequently induced the expression of malignancypromoting genes or inhibited the expression of anticarcinogenic genes. Restoration of PTEN generally blunted or reversed these effects. This trend predicts that the tumor promoting gene expression effects of SF/HGF:c-Met overexpression would be amplified by PTEN loss. Thus, the gene expression effects of combined PTEN loss and SF/HGF:c-Met overexpression could also partly explain the association of these two events with the progression of low-grade gliomas to high-grade glioblastoma (Rosen et al., 1996; Koochekpour et al., 1997; Davies et al., 1999). In fewer cases, PTEN amplified the gene expression effects of SF/HGF indicating a more complex interaction in gene expression regulation. For some genes, PTEN reconstitution also unmasked SF/HGF-induced transcriptional changes that were not seen in PTEN null cells. Fewer neoplasia-related genes were in this category consistent with the normal cell regulatory functions of SF/HGF and c-Met in PTEN wildtype cells. A more detailed analysis of the biological significance of various potentially interesting genes that are co-regulated by PTEN and c-Met could therefore help understand the molecular and cellular changes that underlie glioblastoma and other malignant neoplasms. One such analysis was conducted for the EGFR agonist, TGFa. We found that TGFa expression was induced by SF/HGF in U-373MG cells and that this Oncogene


9180

induction is inhibited by PTEN restoration. Interestingly, SF/HGF led to the phosphorylation and therefore activation of EGFR at tyrosine 1068 in the same cells prior to (but not after) PTEN restoration. Activation of tyrosine 1068 is known to lead to activation of the MAPK and PI3K pathways via recruitment of Grb2 and Gab1. We have therefore established a transcriptionally dependent link between the c-Met and EGFR receptor tyrosine kinase pathways. EGFR could contribute to SF/HGF-induced malignancy especially in the setting of PTEN loss. The above provides an interesting example of the intricate co-regulatory effects of SF/ HGF and PTEN on the genotype and malignant phenotype of gliomas. In conclusion, this first comprehensive report gives insight into the intricate and complex role and modes of actions of PTEN in regulating the transcriptional effects of c-Met and possibly also other RTK. It shows that combined PTEN loss and SF/HGF:c-Met overactivation that occur in high-grade gliomas and other solid neoplasms can contribute to their malignant phenotype by amplifying the induction of tumor promoting genes and inhibiting tumor suppressor genes. Our findings indicate that the effects of manipulating receptor tyrosine kinase dependent pathways for therapeutic and other purposes will likely depend on the PTEN status of the target cells.

Materials and methods Adenovirus construction Wild-type PTEN and PTEN phosphatase mutants were expressed in PTEN-null U-373MG cells using replicationdefective adenoviral expression vectors. Adenoviruses with wild-type PTEN (Ad-PTEN), lipid phosphatase-dead PTEN (Ad-G129E), lipid- and protein phosphatase-dead PTEN (AdC124A), and control adenovirus (Ad-control) were constructed according to Vogelstein and co-workers (He et al., 1998). PTEN, G129E, and C124A transgenes were a kind gift from Dr Kenneth Yamada, (NIH) (Gu et al., 1999). All adenoviruses co-expressed GFP. Adenoviruses were amplified, purified and titered by the University of Pittsburgh Vector Core Facility. Viral titers as determined by plaque formation assays were B1010 pfu. Cell treatments PTEN-null U-373MG human glioblastoma cells were grown to B40% confluency in Dulbecco’s modified essential medium supplemented with 1 g/l glucose, L-glutamine, 10% fetal bovine serum (FBS) and 20 mM HEPES (pH 7.4). Primary normal human astrocytes (NHA) were purchased from Cambrex Bio Science Walkerville, Inc. Cells were grown in Astrocyte Basal Medium supplemented with 0.1% rhEGF, 0.25% insulin, 0.1% ascorbic acid, 0.1% GA-1000, 0.1% L-Glutamine and 3% FBS. For Incyte microarrays, U-373MG and NHA cells were grown to B40% confluency before being transferred to medium containing 0.1% FBS overnight. The cells were then treated with or without 10 ng/ml recombinant SF/HGF (a kind gift from Genentech Inc.) for 24 h. mRNA was subsequently isolated using Oligotex-dt mRNA kit (QIAGEN; Valencia, CA, USA) and sent to Incyte Genomics Oncogene

Inc. for hybridization. For Affymetrix microarrays, U-373MG cells were grown to B40% confluency, transferred to medium containing 0.1% FBS and infected with either Ad-PTEN, Ad-129E, Ad-124A, or Ad-control adenoviruses (MOI ¼ 10). After 24 h, the cells were treated with 10 ng/ml recombinant SF/HGF or medium only as control. After an additional 24 h, total RNA was isolated using the RNeasyTM Mini Kit (QIAGEN; Valencia, CA, USA) following the manufacturer’s instructions. RNA was then used for hybridization to Affymetrix microarray chips, or for Northern analyses using radiolabeled cDNA probes as described below. For the evaluation of EGFR phosphorylation, U-373MG cells were infected with Ad-control or Ad-PTEN prior to treatment with SF/HGF for 24 h as described before, except that phenyl arsene oxide (200 nM) was added to the cells 30 min prior to collection to prevent dephosphorylation of the eventually phosphorylated receptor. Cell extracts were isolated as previously described (Abounader et al., 2001) and subjected to quantitative immunoblotting as described below. Gene microarray hybridization Incyte gene microarrays (containing B9000 genes/chip) were originally used to compare the effects of SF/HGF on gene expression profiles in NHA and U-373MG human glioblastoma cells. Microarray hybridizations were conducted by Incyte Genomics Inc., according to the company’s standardized protocols. For Affymetrix gene chips (containing B21 000 genes/chip), hybridizations were conducted according to the Affymetrix specifications. Briefly, 5 mg of total cellular RNA were used to synthesize double-stranded cDNA and biotinylated antisense cRNA was generated through in vitro transcription. In total, 10-mg of total fragmented cRNA were then hybridized to the Affymetrix human genome GeneChip array U95Av2. Ultimately, fluorescence was detected using the Hewlett-Packard G2500 GeneArray Scanner and image analysis was performed with the Micro Array Suite 5.0 software from Affymetrix, using the standard statistical default settings. For comparison between different chips, global scaling and scaling of all probesets to a user-defined target intensity (TGT) of 150 was used. To assess hybridization and GeneChip image quality we studied the following parameters: scaling factor, background, percentage of present calls, housekeeping genes (30 /50 ratios of GAPDH), and presence or absence of internal spike controls. The quality of replicate hybridizations was assessed by examining the percentage of differential calls (up- or downregulated) between pairwise comparisons. All the assessed parameters were within the average range values recommended by Affymetrix for high-quality array experiments. All experiments were performed in duplicate using two Affymetrix chips for each experimental condition. Analysis of microarray expression data Gene expression changes induced by SF/HGF in PTEN null U-373MG cells and in U-373MG cells transfected with wtPTEN as well as in NHA cells were identified using two complementary methods. The expression of single genes was considered significant if found changed by either one of two analyses: (1) All microarray expression data were analysed by SNOMAD (Standardization and Normalization of MicroArray Data), a program that identifies significantly regulated genes in microarray experiments (Colantuoni et al., 2002). This program performs a series of data conversions including global mean normalization, local background


9181 correction, logarithmic transformation of expression data in a scatter plot, local mean normalization of data, and a local variance correction to generate a list of genes that are significantly up- or downregulated. This list is provided in Z-score (standard deviation) units that reflect the position of differential gene expression relative to the distribution of all values obtained in a particular comparison. Genes showing Z-scores >3 or o3 were considered significantly changed. (2) Additionally, Affymetrix expression data were subjected to pairwise comparisons among the different experimental conditions represented by the samples. Any transcript that showed at least a two-fold change in expression level between experimental sample and control sample was considered ‘significant’. For duplicated samples, the results were filtered independently for significance on each of the four pairwise comparisons. Transcripts that were consistently significant in at least two of the four iterative comparisons were selected for the final candidate list.

designed to amplify a 400–800 bp region of the gene’s mRNA. RT–PCR was conducted using U-373MG RNA as a template and PCR products were subcloned in TOPO PCR cloning vectors (Invitrogen) and sequenced prior to use. Northern blot analysis was performed with 32P-labeled cDNA probes as previously reported (Abounader et al., 2001). Radioactivity was quantified by densitometry and by phosphorimaging using the Bio-Imaging analyser BAS 2500 (Fuji Medical Systems; Stamford, CT, USA). All blots were stripped and then rehybridized with cDNA specific for 28S rRNA. Results are expressed relative to 28S rRNA. Immunoblotting

Northern hybridization

Protein levels of phospho-EGFR and total EGFR in glioma were analysed by immunoblotting. Immunoblotting was performed as previously described (Abounader et al., 2001) using antibodies specific for phospho(1068)-EGFR and total EGFR (Cell Signaling Technologies). Signals were quantified by densitometry (Molecular Dynamics, Sunnyvale, CA, USA) and expressed as phospho-EGFR relative to total EGFR.

Nothern hybridization was used to verify microarray data for a subset of transcripts and to examine the influence of lipidphosphatase-dead and total phosphatase-dead PTEN mutants on SF/HGF-induced gene expression changes. To synthesize cDNA’s for specific genes, oligonucleotide primers were

Acknowledgements Supported by NIH RO1 Grant NS045209 (Roger Abounader) and NIH ROI Grants NS032148 and NS43987 (John Laterra).

References Abounader R, Lal B, Luddy C, Koe G, Davidson B, Rosen EM and Laterra J. (2002). FASEB J., 16, 108–110. Abounader R, Ranganathan S, Kim BY, Nichols C and Laterra J. (2001). J. Neurochem., 76, 1497–1508. Abounader R, Ranganathan S, Lal B, Fielding K, Book A, Dietz H, Burger P and Laterra J. (1999). J. Natl. Cancer Inst., 91, 1548–1556. Birchmeier C, Birchmeier W, Gherardi E and Vande Woude GF. (2003). Nat. Rev. Mol. Cell Biol., 4, 915–925. Bowers DC, Fan S, Walter K, Abounader R, Williams JA, Rosen EM and Laterra J. (2000). Cancer Res., 60, 4277–4283. Cheney IW, Johnson DE, Vaillancourt MT, Avanzini J, Morimoto A, Demers GW, Wills KN, Shabram PW, Bolen JB, Tavtigian SV and Bookstein R. (1998). Cancer Res., 58, 2331–2334. Colantuoni C, Henry G, Zeger S and Pevsner J. (2002). Biotechniques, 32, 1316–1320. Davies MP, Gibbs FE, Halliwell N, Joyce KA, Roebuck MM, Rossi ML, Salisbury J, Sibson DR, Tacconi L and Walker C. (1999). Br. J. Cancer, 79, 1542–1548. DiRenzo MF, Narshimhan RP, Olivero M, Bretti S, Giordano S, Medico E, Gaglia P, Zara P and Comoglio PM. (1991). Oncogene, 6, 1997–2003. Fujioka Y, Taira T, Maeda Y, Tanaka S, Nishihara H, IguchiAriga SM, Nagashima K and Ariga H. (2001). J. Biol. Chem., 276, 45137–45144. Gu J, Tamura M, Pankov R, Danen EH, Takino T, Matsumoto K and Yamada KM. (1999). J. Cell Biol., 146, 389–403. He T-C, Zhou S, DaCosta LT, Yu J, Kinzler KW and Vogelstein B. (1998). Proc. Natl. Acad. Sci. USA, 95, 2509–2514.

Jeffers M, Schmidt L, Nakaigawa N, Webb CP, Weirich G, Kishida T, Zbar B and VandeWoude GF. (1997). Proc. Natl. Acad. Sci. USA, 21, 11445–11450. Koochekpour S, Jeffers M, Rulong S, Taylor G, Klineberg E, Hudson EA, Resau JH and Vande Woude GF. (1997). Cancer Res., 57, 5391–5398. Laterra J, Nam M, Rosen E, Rao JS and Johnston P. (1997). Lab. Invest., 76, 565–577. Leslie NR, Bennett D, Gray A, Pass I, Hoang-Xuan K and Downes CP. (2001). Biochem. J., 357, 427–435. Maehama T and Dixon JE. (1998). J. Biol. Chem., 273, 13375–13378. Maier D, Jones G, Li X, Schonthal AH, Gratzl O, Van Meir EG and Merlo A. (1999). Cancer Res., 59, 5479–5482. Raftopoulou M, Etienne-Manneville S, Self A, Nicholls S and Hall A. (2004). Science, 303, 1179–1181. Rosen EM, Laterra J, Joseph A, Jin L, Fuchs A, Witte M, Weinarnd M and Goldberg ID. (1996). Intl. J. Cancer, 67, 248–255. Sano T, Lin H, Chen X, Langford LA, Koul D, Bondy ML, Hess KR, Myers JN, Hong YK, Yung WK and Steck PA. (1999). Cancer Res., 59, 1820–1824. Schaeper U, Gehring NH, Fuchs KP, Sachs M, Kempkes B and Birchmeier W. (2000). J. Cell Biol., 149, 1419–1432. Stambolic V, MacPherson D, Sas D, Lin Y, Snow B, Jang Y, Benchimol S and Mak TW. (2001). Mol. Cell, 8, 317–325. Stambolic V, Suzuki A, de la Pompa JL, Brothers GM, Mirtsos C, Sasaki T, Ruland J, Penninger JM, Siderovski DP and Mak TW. (1998). Cell, 95, 29–39. Steck PA, Pershouse MA, Jasser SA, Yung WK, Lin H, Ligon AH, Langford LA, Baumgard ML, Hattier T, Oncogene


9182 Davis T, Frye C, Hu R, Swedlund B, Teng DH and Tavtigian SV. (1997). Nat. Genet., 15, 356–362. Teng DH, Hu R, Lin H, Davis T, Iliev D, Frye C, Swedlund B, Hansen KL, Vinson VL, Gumpper KL, Ellis L, El-Naggar A, Frazier M, Jasser S, Langford LA, Lee J, Mills GB, Pershouse MA, Pollack RE, Tornos C, Troncoso P, Yung WK, Fujii G, Berson A, Bookstein B, Bolen JB, Tavtigian SV and Steck PA. (1997). Cancer Res., 57, 5221–5225.

Oncogene

Wang SI, Puc J, Li J, Bruce JN, Cairns P, Sidransky D and Parsons R. (1997). Cancer Res., 57, 4183–4186. Wu X, Senechal K, Neshat MS, Whang YE and Sawyers CL. (1998). Proc. Natl. Acad. Sci. USA, 95, 15587–15591. Xiao GH, Jeffers M, Bellacosa A, Mitsuuchi Y, Vande Woude GF and Testa JR. (2001). Proc. Natl. Acad. Sci. USA, 98, 247–252.