Loss of PTEN Binding Adapter Protein NHERF1 from Plasma Membrane in Glioblastoma Contributes to PTEN Inactivation

Glioblastoma multiforme (GBM), the most aggressive and frequent glioma form, may arise de novo, or secondary to the progression of lower grade astrocytomas (1). GBM is refractory to conventional treatment and has a 12- to 15-month median survival. The pathways amenable to targeted therapy are being uncovered in GBM, and one such pathway is the phosphatidylinositol-3-OH kinase (PI3K)/Akt. The PTEN tumor suppressor, the major inhibitor of the pathway, directly antagonizes PI3K by dephosphorylating phosphatidylinositol 3,4,5-trisphosphate (2), which then recruits Akt and unmasks its kinase domain through a conformational change (3). This allows Akt activation by Thr³⁰⁸ and Ser⁴⁷³ phosphorylation (4, 5). PTEN contains a phosphatase domain, a phospholipid-binding C2 domain, and a tail-region ending in a PDZ (PSD-95/Disc-large/ZO-1)–binding motif (6, 7). Through this motif, PTEN could bind and be recruited at the plasma membrane (PM) by PDZ domain proteins, including NHERF1/EBP50 (Na⁺/H⁺ exchanger regulatory factor/ezrin-radixin-moesin–binding phosphoprotein 50), Par-3, and membrane-associated guanylate kinase with inverted orientation proteins (8–10). NHERF1, an adaptor protein localized at the PM in physiologic conditions, consists of two tandem PDZ domains, of which PDZ1 binds to PTEN PDZ motif, and a COOH-terminal ezrin-radixin-moesin–binding region.

PTEN inactivation has been shown to occur by mutation and loss of heterozygosity in ∼20% to 40% of GBM tumors (11, 12) and in 77% of GBM cell lines (13). Recently, NHERF1 was reported to be overexpressed in GBM as well as being involved in GBM invasiveness (14). However, no mechanistic explanation was presented regarding the role of NHERF1 in GBM. We show that NHERF1, which is normally expressed at the PM in astrocytes, had shifted cytoplasmic expression in the majority of GBM tumors and cell lines. Mechanistically, this shift impairs PTEN recruitment at the PM and results in Akt activation. We thus reveal PTEN regulation in GBM by membrane-localized NHERF1, which cancer cells frequently evade by loss of membrane NHERF1.

The protocol for immunofluorescence was described (15). The GBM tissue microarray processing is described in the Supplementary material.

Frozen GBM tumor samples of ∼3 cm diameter were dissected in regions that were pretriturated with an 18-gauge needle in hypotonic buffer [10 mmol/L HEPES (pH 7.5), 10 mmol/L KCl, and 3 mmol/L MgCl₂] and clarified at 600 × g for 30 seconds. The material from the supernatant was fractionated as described (16).

GBM cells, grown in DMEM supplemented with 10% fetal bovine serum, were obtained as follows: LN18, LN229, LN308, LN428 (Dr. Erwin van Meir, Department of Neurosurgery, Emory University, Atlanta, GA; 2004), U87-MG (American Type Culture Collection, 1998), U251-MG, U373-MG, D54 (Dr. T.J. Liu, Department of Neuro-Oncology, University of Texas M.D. Anderson Cancer Center, Houston, TX; 2003), and A172 (Dr. Brian Varnum, Inflammation Department, Amgen, Inc., Thousand Oaks, CA; 1998). They were tested and authenticated in October 2009 by the Cell Line Fingerprinting Core of the Brain Tumor Center at M.D. Anderson Cancer Center using short tandem repeat profiling with GenomeLab Human STR Primer Set (Beckman Coulter). The protocols for retroviral infection, cell lysis, Western blotting, and cellular fractionation were described (7, 8). The proliferation assay, antibodies, plasmids, and short hairpin RNAs are described in the Supplementary material.

To explore the function of NHERF1 in GBM, we examined NHERF1 and PTEN expression in tissue microarrays containing normal adjacent brain, grade 3 anaplastic astrocytoma (AA), and grade 4 GBM samples (Fig. 1A). In normal brain, NHERF1 staining was mainly associated with the membrane of cells. We confirmed that these cells are astrocytes by double staining for NHERF1 and glial fibrillary acidic protein, a marker for astrocytes (Fig. 1B). NHERF1 localized especially to the membrane in astrocytic processes (Fig. 1A and B). In tumors, we observed two patterns of NHERF1 staining: reactive astrocyte–like, in which cells presented both cytoplasmic and membrane staining, and cytoplasmic-like, in which NHERF1 localized to the cytoplasm of tumor cells (Fig. 1A). The reactive astrocyte–like pattern was rare, being slightly more elevated in AA (18%) than in GBM (8.1%; Fig. 1A, top graph), indicating loss of membrane NHERF1 in the vast majority of GBM. PTEN costaining appeared mainly in the cytoplasm of cells in 35.3% AA and 59.5% GBM (Fig. 1A, bottom graph), perhaps indicating a tendency to upregulate PTEN in GBM. To correlate NHERF1 subcellular distribution with Akt activation, we fractionated various regions from frozen GBM tumors (Fig. 1C). NHERF1 was either lost or had predominant cytoplasmic distribution in the majority of the regions (Fig. 1D). Most importantly, the phosphorylated Akt levels inversely correlated with NHERF1 membrane levels (r = −0.917), suggesting that loss of membrane NHERF1 contributes to Akt activation.

We investigated the subcellular distribution of NHERF1 in a panel of GBM cell lines and observed that depending on PTEN status, distinct patterns of NHERF1 intracellular localization were apparent (Fig. 2A). In PTEN-negative cell lines, NHERF1 was localized in both cytoplasmic and membrane compartments. In PTEN-positive cell lines, NHERF1 was either predominantly expressed in the cytoplasm, in LN18 and LN428 cells, or in the membrane, in LN229 cells.

Similarly to GBM tumors, in PTEN-positive cells, NHERF1 membrane levels were inversely correlated with phosphorylated Akt levels (Fig. 2B). In cells with cytoplasmic NHERF1, Akt activation was also translated to downstream effectors, such as p70^S6K. Because NHERF1 associates with PTEN (8), these results suggested that PTEN is fully active at the PM to suppress Akt only in LN229 cells that express membrane NHERF1. Analysis of endogenous PTEN revealed cytoplasmic staining reaching the PM in LN229 cells and restricted perinuclear staining in LN18 and LN428 cells (Fig. 2C), supporting defective PTEN membrane recruitment in cells with cytoplasmic NHERF1. Indeed, the expression of membrane-targeted Myr-PTEN in LN18 cells completely suppressed Akt activation and strongly inhibited proliferation in comparison with wild-type PTEN, indicating the presence of a PTEN membrane recruitment deficit in these cells (Supplementary Figure).

To show that NHERF1 membrane expression has suppressive effects on Akt activation and cell growth, we reconstituted LN18 and LN428 cells that express cytoplasmic NHERF1 with a membrane-targeted Myr-NHERF1 form. We confirmed that Myr-NHERF1 displayed PM localization, whereas control overexpressed NHERF1 had diffuse localization in all cell compartments (Fig. 3A). Coexpression of Myr-NHERF1 with wild-type PTEN redistributed PTEN to the PM (Fig. 3B), indicating PTEN membrane recruitment by membrane-localized NHERF1. We next examined whether the observed PTEN membrane recruitment is followed by Akt suppression. In both cell lines, Myr-NHERF1 reduced phosphorylated Akt levels (Fig. 3C). Overexpressed control NHERF1 determined a less pronounced effect, most likely due to partial membrane localization. The variable Akt suppression following Myr-NHERF1 expression might result from differences in endogenous PTEN recruitment due to PTEN tail-region phosphorylation that inhibits NHERF1 binding (8). Consistent with the effects on Akt, Myr-NHERF1 specifically decreased proliferation in LN18 and LN428 cells, whereas it had no effect in LN229 cells that expressed endogenous membrane-localized NHERF1 (Fig. 3D). These data suggested a growth suppressor function for NHERF1 at the membrane through Akt suppression.

In reciprocal experiments, we depleted NHERF1 from LN229 cells that express membrane-localized NHERF1 and from control LN18 cells that express cytoplasmic NHERF1 (Fig. 4A). Cell proliferation was significantly increased by NHERF1 depletion from LN229 but not from LN18 cells, suggesting that NHERF1 acts as a tumor suppressor at the membrane and its cytoplasmic shift might have the same effect on cell proliferation as its depletion. NHERF1 depletion in LN229 cells led to a 2-fold Akt activation in both basal and stimulated conditions (Fig. 4B), confirming that membrane NHERF1 depletion activates Akt in GBM cells. However, Akt activation following NHERF1 depletion seemed moderate. To assess how well endogenous PTEN suppresses Akt, we also depleted PTEN in the same cells (Fig. 4C). Efficient PTEN knockdown resulted in a similar 2-fold Akt activation (Fig. 4C), indicating comparable patterns of Akt activation in cells lacking either PTEN or membrane NHERF1.

The PI3K/Akt pathway has become an important target for anticancer therapy because of its constitutive activation in a wide variety of cancers. PTEN is the pathway suppressor most frequently altered in cancers. In particular, PTEN-inactivating mutations occur at a high rate in GBM (11, 12), leading to PI3K/Akt pathway activation. We show that PTEN may be additionally inactivated in GBM by mislocalization of NHERF1, a PTEN-binding adapter protein (8). NHERF1 has PM localization in normal cells (17), including astrocytes. Remarkably, NHERF1 is shifted to the cytoplasm in the majority of GBM tumors and cell lines, and we show that the oncogenic effect of NHERF1 membrane loss is due to PTEN displacement from the membrane with consecutive Akt activation (Fig. 4D). Interestingly, in overexpression experiments, the PDZ motif of PTEN is dispensable for GBM growth suppression (7). Thus, the interaction between PTEN and NHERF1 might not be the only factor involved in PTEN activity. Another mechanism may arise from stochastic association-dissociation events between PTEN molecules localized in the cytoplasmic pool and the PM, depending on PTEN conformation states determined by phosphorylation (ref. 18; Fig. 4D). In this case, the presence of membrane-localized NHERF1 might act synergistically to suppress Akt by increasing PTEN association with the PM.

The NHERF1 gene locus at 17q25.1 suffers frequent allelic loss in breast and ovarian cancers, and NHERF1 mutations with loss of heterozygosity have been described in 3% of invasive breast carcinomas (19). NHERF1 loss or cytoplasmic overexpression has been observed in breast cancer cells (17), raising the possibility that NHERF1 intracellular shift might represent a more general oncogenic mechanism. Using a combination of NHERF1 membrane-targeting and membrane depletion experiments in GBM cells with cytoplasmic or membrane endogenous NHERF1, we unequivocally showed that NHERF1 behaves as a tumor suppressor when localized at the PM. Aside from the effects on the PI3K/Akt pathway, other oncogenic influences determined by NHERF1 membrane loss might arise in tumor cells from either disruption of membrane complexes or formation of aberrant cytoplasmic or nuclear complexes between mislocalized NHERF1 and other proteins, such as β-catenin (20). Indeed, β-catenin transcriptional activity is enhanced either by NHERF1 loss (16) or by NHERF1 intracellular shift (20), and other oncogenic mechanisms, perhaps related to ezrin-radixin-moesin proteins, might also be activated during NHERF1 intracellular shift.

In summary, we propose a model in which membrane-localized NHERF1 in normal cells contributes to PTEN membrane recruitment, whereas NHERF1 cytoplasmic redistribution in cancer cells alters the regulation of PTEN at the PM. This novel mechanism attributes to NHERF1 a key role in regulating the PI3K/Akt pathway and opens the avenue for establishing prospective markers of tumor progression and PI3K-targeted therapy.

No potential conflicts of interest were disclosed.

Grant Support: NCI-CA107201 (M-M. Georgescu) and NCI-CA16672 for DNA sequencing.

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