Research ArticleCancer

PTEN Protein Phosphatase Activity Correlates with Control of Gene Expression and Invasion, a Tumor-Suppressing Phenotype, But Not with AKT Activity

See allHide authors and affiliations

Science Signaling  28 Feb 2012:
Vol. 5, Issue 213, pp. ra18
DOI: 10.1126/scisignal.2002138


The tumor suppressor phosphatase and tensin homolog deleted on chromosome 10 (PTEN) has a well-characterized lipid phosphatase activity and a poorly characterized protein phosphatase activity. We show that both activities are required for PTEN to inhibit cellular invasion and to mediate most of its largest effects on gene expression. PTEN appears to dephosphorylate itself at threonine 366, and mutation of this site makes lipid phosphatase activity sufficient for PTEN to inhibit invasion. We propose that the dominant role for PTEN’s protein phosphatase activity is autodephosphorylation-mediated regulation of its lipid phosphatase activity. Because PTEN’s regulation of invasion and these changes in gene expression required lipid phosphatase activity, but did not correlate with the total cellular abundance of its phosphatidylinositol 3,4,5-trisphosphate (PIP3) lipid substrate or AKT activity, we propose that localized PIP3 signaling may play a role in those PTEN-mediated processes that depend on both its protein and lipid phosphatase activities. Finally, we identified a tumor-derived PTEN mutant selectively lacking protein phosphatase activity, indicating that in some circumstances the regulation of invasion and not that of AKT can correlate with PTEN-mediated tumor suppression.


Loss of function of the phosphatase PTEN (phosphatase and tensin homolog deleted on chromosome 10) is a common event in the development of human cancers (1, 2). Loss of PTEN function can occur through mutation of one or both alleles or through various nongenomic mechanisms (3). PTEN plays a role in the regulation of many physiological processes; however, its highly conserved effects on cell growth, proliferation, and survival appear to be key to its tumor suppressor functions (2, 4).

Biochemically, PTEN’s best recognized function is as a lipid phosphatase that metabolizes phosphatidylinositol 3,4,5-trisphosphate (PIP3), the primary product of the class I phosphoinositide 3-kinases (PI3Ks). Through this phosphatase activity, PTEN can directly oppose the effects of PI3K on the many PIP3-binding effector proteins that mediate PI3K-dependent signaling. These lipid-binding proteins include the well-studied AKT group of PIP3-activated serine-threonine protein kinases and several groups of regulators of the membrane-anchored adenosine diphosphate (ADP) ribosylation factor (ARF) and RAC families of small guanosine triphosphatases (GTPases) (5). Numerous biological responses, many of which are related to cell growth and proliferation, appear to be regulated by the amplitude of PI3K-dependent signals. However, there are also settings, such as cell polarization during chemotaxis (and during polarization) in some epithelia, in which the subcellular localization of PIP3 appears to be crucial (610). In these cases, localized PTEN activity appears to contribute to the establishment of PIP3 signal gradients or pools (1114).

Its evolutionarily conserved role as a core component of the PI3K signaling network provides a mechanistic explanation for many of the effects of PTEN on cell growth, proliferation, survival, metabolism, and polarity. PTEN has other potential mechanisms of action, including protein phosphatase activity and noncatalytic actions (1520); however, the importance of these alternative mechanisms is unclear. We have developed a PTEN mutant that retains wild-type lipid phosphatase activity but lacks activity against peptide substrates and used this mutant to show that both lipid and protein phosphatase activities are required within the same molecule or PTEN to inhibit glioma cell invasion (21). Here, we provide a mechanistic basis for the concerted actions of PTEN’s lipid and protein phosphatase activities and show that they are also required together to mediate many of the effects of PTEN on gene expression. Finally, we show that a mutation of PTEN, identified in a small cell lung carcinoma, selectively ablates PTEN’s protein phosphatase activity, identifying a requirement for this activity in tumor suppression.


Requirements for both the protein and the lipid phosphatase activities of PTEN in the control of glioma cell invasion and gene expression

Most of the data that imply important effects of PTEN’s protein phosphatase activity rely on a functionally selective mutant in which Glu replaces Gly129 (PTEN G129E). PTEN G129E, originally identified in two Cowden syndrome families, lacks lipid phosphatase activity but retains protein phosphatase activity (22, 23). We developed a PTEN mutant that selectively retains wild-type activity against vesicular lipid substrates but lacks substantial protein phosphatase activity, PTEN Y138L (in which Tyr138 is substituted with Leu). When expressed in near-physiological amounts in PTEN-null cells, PTEN Y138L, like the wild-type enzyme, markedly reduces total cellular PIP3 abundance and AKT phosphorylation and activity (21). Here, we extended this comparison to show indistinguishable effects of wild-type and Y138L PTEN on AKT phosphorylation in three PTEN-null glioma cell lines; the decrease in AKT phosphorylation depended on PTEN abundance but did not differ between wild-type PTEN and Y138L (Figs. 1, C and E, and 2, B and F, and fig. S1). Both PTEN proteins also suppress platelet-derived growth factor (PDGF)–stimulated AKT phosphorylation in U87MG cells [(21) and fig. S1C]. The PTEN G129E and Y138L mutants together therefore provide tools to investigate the contributions of these two catalytic activities to PTEN’s biological functions. We used lentiviruses to express these two untagged PTEN mutants in PTEN-null U87MG or DBTRG-05MG glioblastoma cells in amounts similar to those found in cells expressing endogenous PTEN and compared their effects to those expressing wild-type PTEN or a catalytically inactive mutant in which Cys124 is replaced with Ser, PTEN C124S. The invasion of glioblastoma cells into normal tissue surrounding a primary tumor is one reason for the resistance of this tumor type to therapy. Expression of wild-type PTEN suppressed the invasion of U87MG cells into a three-dimensional (3D) Matrigel matrix, whereas neither PTEN G129E nor PTEN Y138L alone, nor the two proteins expressed together, was able to do so [(21) and Fig. 1]. To ensure that both proteins were being expressed, we performed coexpression of green fluorescent protein (GFP)–PTEN G129E with untagged PTEN Y138L, still with no effect on cellular invasion (Fig. 1). This argues that both activities must exist in the same protein molecule to mediate the suppression of invasion.

Fig. 1

PTEN requires both lipid and protein phosphatase activity to inhibit invasion. (A to D) U87MG cells were transduced with lentiviruses encoding GFP, untagged wild-type (WT) PTEN, or PTEN point mutants before (A to C) being assayed for invasion over 16 hours. One set of samples was cotransduced with both untagged PTEN Y138L and GFP–PTEN G129E. Representative of three experiments. (A) Invading cells were fixed, stained, and photographed. Each image is 800 μm wide. (B) Quantification represents the mean percentage of cells invading relative to GFP-transduced cells from 10 randomly selected fields ± SEM. ***P < 0.001, t test applied to data before normalization. All genotypes are highly significantly different from wild-type PTEN. (C) PTEN expression and AKT phosphorylation were verified by immunoblotting of adherent cell lysates, including untransduced cells and GFP-expressing cells treated with PI3K inhibitor (1 μM PI103, 30 min). HEK293T cell lysates were included to ensure that PTEN abundance was close to physiological. (D) Transduced U87MG cells grown in Matrigel for 3 days, fixed, and stained for filamentous actin (F-actin) (red) and DNA (blue). Each image is 128 μm wide. (E) Invasion assays were performed in DBTRG-05MG PTEN-null glioma cells. Quantitation of invasion is as performed in (B) and immunoblotting data are as in (C). *P < 0.05; **P < 0.01; ***P < 0.001, t test. DBTRG-05MG experiments were performed twice with similar results. (F) Table representing phosphatase activities of WT PTEN and mutants.

Fig. 2

PTEN requires both lipid and protein phosphatase activities for many of its effects on gene expression. U87MG cells were transduced with lentiviruses encoding GFP or the indicated PTEN proteins, seeded into 3D Matrigel cultures, and maintained for 16 hours. RNA was prepared and gene expression was analyzed by microarray. (A) Mean transcript changes from duplicate samples relative to sham-transduced controls. Blue denotes up-regulation; yellow, down-regulation. (B and C) Numbers of probes changed relative to “No virus” sample at P < 0.01 and fivefold change threshold. Significance levels are output from Rosetta Resolver software. Immunoblotting shows PTEN abundance and AKT phosphorylation in transduced adherent cells The significance threshold for inclusion in (A) is intermediate between (B) and (C), so that most of the contributing probes in (B) do not appear in (A), and is described in Materials and Methods. (D) Plot of fold changes for all probes in response to WT and Y138L PTEN expression. Probes are colored according to the sum probability of their expression change. Most probes respond either similarly to both WT and Y138L (clustering on diagonal) or only to PTEN WT (clustering near x axis). (E) Overlap analysis between a gene set most strongly regulated selectively by PTEN WT [169 probes with largest fold change in (C)] and publicly available microarray data from human gliomas (84% with a recognized loss of PTEN expression or gene dose). (F) PTEN abundance and AKT phosphorylation by immunoblotting from Matrigel cultures. Representative of n = 3 experiments.

To extend our analysis of PTEN’s separable activities to a more global and unbiased analysis, we performed whole-genome microarray analysis of gene expression changes occurring in U87MG cells transduced with the different forms of PTEN and grown for 16 hours in 3D Matrigel cultures. We initially analyzed the data by making pairwise comparisons with sham-transduced controls. The expression of GFP or a catalytically inactive PTEN mutant had no significant effect on gene expression in these cells (statistical analyses are described in the Supplementary Materials and Methods), which gives us confidence that our lentiviral expression system produces few artifactual effects on cell behavior. Wild-type PTEN altered the expression of a large number of transcripts (Fig. 2). PTEN G129E, which has only protein phosphatase activity, caused modest changes in the abundance of a small number of transcripts. PTEN Y138L, which has lipid phosphatase activity and can suppress cellular PIP3 levels and AKT activity, shared a large set of transcriptional responses with wild-type PTEN, but failed to affect a large group of transcripts that were the most highly responsive to the wild-type enzyme (Fig. 2, A to C). Detailed comparisons of the transcriptional responses to wild-type and Y138L PTEN showed that almost all probes either responded similarly to both enzymes or responded only to wild-type PTEN (Fig. 2D). Very few probes were affected strongly by wild-type and weakly by Y138L, consistent with a qualitatively different signaling mechanism mediated only by wild-type PTEN, rather than a quantitative difference in the ability of these proteins to influence one pathway or set of responses. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology group analysis correlated both wild-type and Y138L PTEN transcriptional responses with several recognized pathways (fig. S2), including the regulation of cell growth and cell cycle progression, consistent with the ability of both proteins to inhibit AKT phosphorylation and cell proliferation in these cells (21).

We compared the set of transcripts (table S1) most strongly induced by wild-type PTEN, but not by PTEN Y138L, with available gene expression data from human glioblastoma samples. Of the 517 tumor samples analyzed, 510 showed down-regulation of more than 25% of the top 169 PTEN-responsive genes, and more than 25% of the 517 tumor samples showed down-regulation of 81 of these 169 genes. These results show significant overlap between the genes selectively induced by wild-type PTEN and those suppressed in glioblastoma and suggest that this transcriptional control pathway may be clinically relevant (P = 0 + 0.01 and P = 0.019 ± 0.01; Fig. 2E and Supplementary Materials and Methods). We also verified that PTEN Y138L suppressed AKT phosphorylation in U87MG cells cultured in 3D Matrigel (Fig. 2F).

Evidence for PTEN autodephosphorylation of Thr366

Our analysis of PTEN-mediated effects on transcript abundance indicates that the protein phosphatase activity of PTEN is associated with modest effects acting independently, but that in the wild-type enzyme, protein phosphatase activity acts together with lipid phosphatase activity to mediate many of PTEN’s greatest effects on gene expression. One simple hypothesis to explain the requirement for both activities, that our analysis of invasion implies must exist within the same molecule, is the previous proposal that PTEN may autoregulate through the action of its protein phosphatase activity on its own C-terminal phosphorylation sites (19, 24). We suggest that this autodephosphorylation may enable PTEN to direct its lipid phosphatase activity to the appropriate cellular location rather than to regulate a noncatalytic mechanism.

We used a range of phosphospecific antibodies raised against sites in the PTEN C terminus to test the hypothesis that PTEN might dephosphorylate itself. Thr366 is phosphorylated by glycogen synthase kinase 3 (GSK3); this requires a previous priming phosphorylation of Ser370 by protein kinase CK2 (25, 26). Also phosphorylated by CK2, usually to high stoichiometry, is a cluster of four sites between Ser380 and Ser385 (27, 28). We found that wild-type and catalytically inactive PTEN had similar degrees of phosphorylation of this cluster of residues (Fig. 3, A and B). We also assessed the signal obtained with an antibody that specifically recognizes the unphosphorylated 380 to 385 sites. This antibody gave a strong signal with unphosphorylated bacterially expressed PTEN, but produced only a weak signal that was unaffected by the catalytic activity of the PTEN protein with PTEN expressed in U87MG cells. However, the phosphorylation of both Thr366 and Ser370 was markedly decreased in active PTEN compared to the catalytically impaired protein (Fig. 3, A and B). This suggested that they could be sites of autodephosphorylation. When the cellular phosphorylation of Thr366 was blocked by the selective GSK3 inhibitor CT99021 (29), its phosphorylation decreased over a period of 1 to 2 hours with wild-type PTEN; however, catalytically impaired PTEN remained phosphorylated on this site for many hours (Fig. 3C and fig. S3). We tested the ability of purified PTEN to dephosphorylate Thr366 in vitro, using the purified C-terminal tail of PTEN phosphorylated with GSK3 as a substrate. In these experiments, wild-type PTEN had robust activity, that of PTEN G129E was somewhat weaker, and both PTEN C124S and Y138L had little or no activity (Fig. 3D and fig. S3C).

Fig. 3

Evidence that PTEN autodephosphorylates Thr366. U87MG cells transiently expressing PTEN WT or PTEN point mutants were lysed, total PTEN was immunoprecipitated, and phosphorylation was investigated by replicate phosphospecific immunoblotting. (A and B) Unphosphorylated bacterially expressed GST-PTEN (A) or PTEN (B) was used as a control. Both representative of three experiments. (C) U87MG cells transiently expressing PTEN or catalytically inactive PTEN C124S were treated with the GSK3 inhibitor CT99021 (5 μM) for up to 4 hours as shown. PTEN Thr366 phosphorylation was then assessed by lysis, immunoprecipitation, and immunoblotting. For direct comparison of the rate of dephosphorylation of these proteins, a longer film exposure is presented for the more weakly phosphorylated WT protein. Representative of three experiments. (D). GST-PTEN tail (351 to 403) phosphorylated with 32P on Thr366 (see fig. S3C) was used as a substrate for PTEN protein phosphatase activity. Data are presented as mean picomoles of phosphate released from triplicate samples ± SEM. Two-way comparisons were made between PTEN WT and individual PTEN mutants. ***P < 0.001 compared to WT with t test. This experiment was repeated three times with two independent substrate preparations. Quantitation of blotting data from (A) and (B) is presented in fig. S7.

Consistent with autodephosphorylation of Thr366, inhibiting PTEN’s activity by treating cells with hydrogen peroxide or hyperosmotic stress (30, 31) increased its phosphorylation on Thr366 (Fig. 4A and fig. S4). The increase in Thr366 phosphorylation stimulated by treatment with 0.5 M sorbitol was inhibited by CT99021, indicating that phosphorylation was mediated by GSK3 and is consistent with the hypothesis that phosphorylation increased as a result of reduced dephosphorylation rather than through increased activity of an alternate kinase that was activated by osmotic stress (fig. S4C).

Fig. 4

PTEN Thr366 phosphorylation inversely correlates with PTEN activity in a protein molecule–autonomous manner. (A) U87MG cells transiently expressing PTEN WT were treated for 1 hour with 0.5 M sorbitol, 1 mM hydrogen peroxide, or 100 μM zinc sulfate or washed into Hepes-buffered Krebs Ringer, also for 1 hour. (B) U87MG cells were simultaneously cotransduced to high expression levels with viruses encoding catalytically inactive GFP–PTEN C124S and untagged PTEN that was either catalytically active (WT) or catalytically inactive (C124S). Cells were lysed, total PTEN was immunoprecipitated, and phosphorylation was investigated by replicate phosphospecific immunoblotting. (C) GFP PTEN WT or catalytically inactive (C124S) was expressed in 293T cells and immunoprecipitated from cell lysates. These immune complexes were divided, and one-half was dephosphorylated with λ-phosphatase before Western blotting. All data are representative of at least three experiments. Quantitation of blotting data from (A) and (B) is presented in fig. S7.

To test the hypothesis that PTEN dephosphorylates its own C terminus intramolecularly, as predicted by the failure of coexpressed PTEN Y138L and PTEN G129E to inhibit cellular invasion, we determined the abundance of phosphorylated Thr366 (P-T366) PTEN in cells coexpressing catalytically active and inactive forms of PTEN. Although wild-type PTEN itself displayed little Thr366 phosphorylation, its coexpression failed to decrease the T366 phosphorylation of coexpressed GFP–PTEN C124S (Fig. 4B). The PI3K inhibitor PI103 failed to decrease Thr366 phosphorylation of catalytically inactive PTEN C124S (fig. S4D), indicating that the low Thr366 phosphorylation of wild-type PTEN is not simply the result of low downstream PI3K-dependent phosphorylation in cells expressing this active PIP3 phosphatase. If the PTEN C-terminal undergoes intramolecular autodephosphorylation, blocking its known interaction with the phosphatase domain (32, 33) should increase Thr366 phosphorylation as a result of its reduced association with the active site. Consistent with this prediction, we found that substituting three clustered C-terminal phosphorylation sites with alanines (S380A, T382A, T383A, also known as PTEN A3) (28), which promotes an open conformation (32, 33), markedly increased Thr366 phosphorylation (fig. S5A). Analyses of four additional PTEN mutations identified in human tumors supported an inverse correlation between PTEN activity and Thr366 phosphorylation. Similar to wild-type PTEN, two catalytically active PTEN mutants, PTEN L42R (in which Leu42 is replaced with Arg) and X404L (in which the stop codon 404 is replaced with Leu) (34, 35), displayed low Thr366 phosphorylation, whereas two mutants lacking protein phosphatase activity, G129R (in which Gly129 is replaced with Arg) (34) and Y138C (in which Tyr138 is replaced with Cys; see below), showed much more Thr366 phosphorylation (fig. S5B). λ-Phosphatase was similarly effective in mediating the in vitro dephosphorylation of immunoprecipitated wild-type and catalytically dead PTEN (Fig. 4C), suggesting that Thr366 was similarly accessible for dephosphorylation by exogenous phosphatases.

Together, these data indicate that PTEN can autodephosphorylate its own C terminus at Thr366 and possibly also Ser370. Changes in the phosphorylation of Thr366 under various experimental conditions were generally accompanied by similar changes in Ser370, indicating that both sites may be substrates for PTEN autodephosphorylation (Figs. 3, A and B, and 4B and figs. S4 and S5). However, mutation of Thr366 to alanine in catalytically inactive PTEN greatly reduced phosphorylation of the nearby Ser370 (fig. S5), suggesting that Thr366 phosphorylation may itself affect the phosphorylation status of Ser370 and, therefore, that autodephosphorylation per se may be limited to Thr366.

Requirement for Thr366 dephosphorylation for PTEN to inhibit U87MG cell invasion independently of AKT

Autodephosphorylation of its C terminus could explain why the existence of both its protein and lipid phosphatase activities on the same molecule appears to be required for PTEN’s ability to inhibit U87MG cell invasion into Matrigel. If this is the case, substituting Thr366 or Ser370 with alanine to block their phosphorylation might enable a PTEN Y138L mutant (that is, a mutant that retains lipid phosphatase activity but lacks protein phosphatase activity) to suppress invasion. We found that, whereas PTEN Y138L failed to strongly suppress U87MG cell invasion in a Transwell assay, the wild-type enzyme and both PTEN Y138L T366A and PTEN Y138L S370A were able to do so (Fig. 5, A and B). Moreover, a PTEN mutant mimicking phosphorylation at this site, PTEN T366D, was unable to suppress invasion, despite having wild-type activity against both lipid and soluble substrates (26) (Fig. 5, A and B). Observation of U87MG cells expressing these different PTEN proteins also revealed differences in cellular morphology when seeded into Matrigel [as previously described for PTEN Y138L (21)], indicating that the consequences of Thr366 phosphorylation on the sequelae of PTEN’s lipid phosphatase activity extend beyond cell invasion. Cells expressing wild-type PTEN, PTEN Y138L T366A, or PTEN Y138L S370A were viable and proliferated. However, soon after seeding into Matrigel, very few of these cells had extended processes into the surrounding matrix, in contrast to cells that were invasive in Transwell assays (fig. S6 and Fig. 1D).

Fig. 5

Nonphosphorylatable mutation of Thr366 rescues the ability of PTEN Y138L to inhibit invasion. U87MG cells were transduced with recombinant lentiviruses encoding GFP, PTEN WT, or PTEN point mutants. (A) Cells were then serum-starved for 1 hour in the presence or absence of PI3K inhibitor (PI103, 1 μM) or AKT inhibitor (Akti-1/2, 1 μM) and assayed for Matrigel Transwell invasion over 16 hours. The cells were fixed, stained, and photographed in random fields. The horizontal axis of each image represents 800 μm. (B) Quantification represents the mean number of cells from 10 randomly selected fields ± SEM. ***P < 0.001, t test. Expression of PTEN and the effects of PTEN and the inhibitors on AKT phosphorylation were verified by Western blotting. (C) U87MG cells were transduced with GFP, PTEN WT, or both PTEN WT and oncogenic active E17K mutants of AKT1, AKT2, or AKT3. Transduced cells were then assayed for invasion and parallel samples were analyzed by immunoblotting. All cells transduced with PTEN displayed significantly reduced invasion compared to those transduced with GFP, but there was no significant effect of mutant AKT expression (**P < 0.01; ns, not significant; t test). The experiments in (A) and (B), and (C) are representative of three and two experiments, respectively.

To further test for any role of the AKT kinases in U87MG cell invasion, we used a selective AKT kinase inhibitor and constitutively active mutants of each AKT isoform. The AKT inhibitor Akti-1/2 had no effect on U87MG cell invasion in this assay (Fig. 5C), further supporting the hypothesis that PTEN’s ability to suppress invasion can be separated from its ability to inhibit AKT phosphorylation. When constitutively active mutant forms of AKT1, 2, and 3 were coexpressed with wild-type PTEN in U87MG cells, the phosphorylation of the AKT substrate GSK3 was increased; however, these mutants failed to promote invasion (Fig. 5C; P = 0.61, 0.84, and 0.91 for AKT1, 2, and 3, respectively, t test). The pan-PI3K inhibitor PI103, which at 1 μM suppressed AKT phosphorylation, also had no effect on invasion (Fig. 5B). We were unable to test the ability of PI103 to suppress PTEN-mediated invasion at higher concentrations because of cell death.

A tumor-derived PTEN mutant that reduces AKT phosphorylation and catalytic activity but has selectively lost protein phosphatase activity

These experiments indicate that the key role for PTEN’s protein phosphatase activity in inhibiting glioma cell invasion is autodephosphorylation of Thr366. However, these analyses provide only indirect evidence for the relevance of PTEN’s protein phosphatase activity to its role in tumor suppression. The PTEN Y138L mutant we used in this study was developed from a systematic screen informed by studies of the PTEN-related proteins, transmembrane phosphatase with tensin homology (TPTE) and TPTE and PTEN homologous inositol lipid phosphatase (TPIP) (21). To test for a more direct correlation between PTEN’s protein phosphatase activity and its ability to act as a tumor suppressor, we analyzed a PTEN Tyr138Cys (Y138C) mutant identified in a metastatic small cell lung cancer cell line, NCI-H196 (36). When expressed, purified, and assayed in vitro, PTEN Y138C had catalytic properties like those of PTEN Y138L, displaying lipid phosphatase activity, but not protein phosphatase activity (Fig. 6). Similarly, expression of physiological amounts of PTEN Y138C in PTEN-null U87MG cells decreased cellular PIP3 abundance and AKT phosphorylation as efficiently as either wild-type or Y138L PTEN, yet (like Y138L) failed to suppress cellular invasion into Matrigel (Fig. 6). We therefore investigated the function of endogenous PTEN Y138C in NCI-H196 cells. Reverse transcription–polymerase chain reaction (RT-PCR) validation confirmed the expression of this mutant allele but failed to detect wild-type PTEN or other secondary mutations (Fig. 7, A and B). Accordingly, phosphatase assays of immunoprecipitated PTEN showed that these cells express PTEN with robust lipid phosphatase activity but little or no protein phosphatase activity (Fig. 7C). As would be predicted, we also found that the mutant PTEN Y138C and C124S proteins expressed endogenously in NCI-H196 and U343MG cells, respectively, both displayed increased phosphorylation of Thr366 relative to that of wild-type protein in human embryonic kidney (HEK) 293T cells (Fig. 7D). Notably, PTEN knockdown in NCI-H196 cells with lentiviral short hairpin RNA (shRNA) led to an increase in cellular AKT phosphorylation (Fig. 7E). This shows that the PTEN Y138C mutant is stably expressed and is active in the suppression of the AKT branch of PI3K-dependent signaling in these cells and implicates a distinct signaling mechanism as driving selection for the mutation of PTEN in this tumor.

Fig. 6

PTEN Y138C selectively retains lipid phosphatase activity and regulates PIP3 and AKT, but not invasion. (A) Purified PTEN WT, Y138L, and Y138C were electrophoresed and Coomassie-stained. (B and C) Time course of the activity of the indicated proteins assayed with (B) radiolabeled PIP3 in phosphatidylcholine vesicles and (C) radiolabeled phosphorylated poly(GluTyr) substrates. Data are mean activity ± range/2 from duplicate assays. Representative of n = 5. (D to G) U87MG cells were transduced with lentiviruses encoding GFP, PTEN, or PTEN mutants or treated with the PI3K inhibitor wortmannin (100 nM, 20 min). (D) PTEN abundance and AKT signaling analyzed by immunoblotting with PTEN, P-T308 AKT, P-S473 AKT, P-S21/9 GSK3, and AKT-substrate phosphorylation consensus antibodies. Representative of n = 4 experiments. (E) Total cellular PIP3 measured and presented as mean of triplicate samples ± SEM. Wort, wortmannin-treated control. Experiments performed once with duplicate analyses of three cell samples for each condition. (F) Transduced cells were assayed for invasion over 16 hours. The cells were fixed, stained, photographed, and (G) counted. Quantification represents the mean cell number from 10 randomly selected fields ± SEM. ***P < 0.001, t test. Representative of three experiments. Quantitation and statistical analysis of blotting data from (D) is presented in fig. S8.

Fig. 7

PTEN Y138C endogenously expressed in NCI-H196 small cell lung carcinoma cells suppresses AKT activity. (A and B) Total RNA was isolated from HEK293T, U343MG, and NCI-H196 cells and reverse-transcribed, and the expressed full-length PTEN cDNA population was sequenced. (C) Endogenous PTEN was immunoprecipitated from HEK293T, U343MG, and NCI-H196 cells. The phosphatase activity of immune complexes was assayed against radiolabeled PIP3 vesicles and phosphorylated poly(GluTyr). The activity is shown as mean activity ± range/2 from duplicate assays normalized to the activity of HEK293T cells. Representative of three experiments. (D) PTEN was immunoprecipitated from HEK293T, U343MG, and NCI-H196 cells, and phosphorylation of PTEN was investigated by replicate phosphospecific immunoblotting. Representative of three experiments. (E) PTEN was knocked down with lentiviruses encoding PTEN shRNA in HEK293T cells expressing PTEN WT, U343MG cells expressing the catalytically inactive PTEN C124S mutant, and NCI-H196 cells expressing the PTEN Y138C mutant lacking protein phosphatase activity. Effects on AKT phosphorylation were determined by Western blotting. Representative of four experiments. Quantitation and statistical analysis of blotting data from (E) is presented in fig. S8.


We show that PTEN’s ability to decrease total cellular PIP3 abundance and AKT phosphorylation is not sufficient to mediate many of its greatest effects on gene expression or its ability to limit cellular invasion by U87MG cells grown in 3D cultures. We also show that a mutation identified in a small cell lung cancer, PTEN Y138C, selectively lacks protein phosphatase activity but retains the ability to decrease both cellular PIP3 abundance and the phosphorylation of AKT. This implicates other PTEN-dependent signaling mechanisms in the regulation of cellular invasion, gene expression, and tumor suppression. Our data also indicate that PTEN autodephosphorylates Thr366 in its C-terminal tail and that this autodephosphorylation is required together with lipid phosphatase activity for PTEN to control invasion, cell morphology, and probably many gene expression responses.

Kinases in the AKT family have received a great deal of attention as likely mediators of pro-oncogenic signaling downstream of PTEN and PI3K. In Pten heterozygous mice that develop tumors in multiple tissues, deletion of Akt1 blocks tumorigenesis (37) and selectively reducing Akt activity (caused by mutating the Pdk1 pleckstrin homology domain) leads to delayed and slower-growing tumors (38). These data indicate that AKT activity is necessary for tumor formation in many tissues and are consistent with the hypothesis that the increased AKT activity driven by PTEN loss promotes tumor formation. Accordingly, cancer drug discovery programs to develop AKT inhibitors are in progress and show some promise (39, 40). However, the importance in tumors of the other signaling pathways activated downstream of PTEN and PI3K is poorly understood.

Our data indicate that PTEN affects invasion and gene expression in U87MG cells through mechanisms that require its lipid phosphatase activity, but do not correlate with the effects of PTEN on the total cellular PIP3 pool or AKT phosphorylation. We speculate that this is through activity against a specific pool of PIP3 or by generating a gradient of PIP3 and thereby regulating PI3K-dependent signaling events that do not correlate with P-AKT. We propose a model in which phosphorylation of Thr366 interferes with a specific mechanism downstream of spatially localized PTEN activity, probably through controlling the interaction of PTEN with a particular protein or proteins. In contrast to other phosphorylation sites in the PTEN C terminus, Thr366 appears not to affect the ability of PTEN to access membranes through electrostatic interactions, metabolize lipid substrates, or regulate AKT (26). Only the minimal catalytic core of PTEN (its phosphatase and C2 domains) is required to interact with the plasma membrane and reduce the total cellular abundance of PIP3 and inhibit AKT activity (41). However, more complex targeting of PTEN is required for the control of other more spatially regulated PI3K-dependent processes, such as membrane ruffling and cell polarity (35, 42). We suggest that this previously observed distinction between untargeted lipid phosphatase activity that controls total cellular abundance of PIP3 and more complex regulated PTEN activity is closely related to our observation that P-T366–inhibited targeting is required for PTEN-dependent regulation of cellular morphology and invasion.

We found that neither lipid nor protein phosphatase activity independently blocked invasion or could mediate the largest effects of wild-type PTEN on gene expression. However, a PTEN mutant with only lipid phosphatase activity, PTEN Y138L, acquired the ability to suppress cellular invasion when combined with mutation of Thr366, an inhibitory C-terminal phosphorylation site that appears to be dephosphorylated by the protein phosphatase activity of PTEN. This provides a mechanistic understanding of how the two activities could act in concert within the same molecule. Phosphorylation of Thr366 regulates the interaction of PTEN with myosin V (43) and directly mediates binding to 58-kD microspherule protein (MSP58) (18). Potentially controlled by such interactions, the role of spatially targeted PTEN activity and resultant PIP3 gradients in cell regulation is well appreciated (44, 45). Our data argue against important independent roles for PTEN’s protein phosphatase activity, at least in the regulation of invasion and large gene expression responses in U87MG cells.

Our data suggest that the principal role for PTEN’s protein phosphatase activity may be to directly dephosphorylate its own C terminus. An alternative explanation for some of our Thr366 phosphorylation data is that catalytically inactive mutants such as PTEN C124S favor a closed conformation that protects P-T366 from dephosphorylation by another phosphatase. However, the increased phosphorylation of the open conformation mutant PTEN A3 and the similar dephosphorylation of wild-type and C124S PTEN by λ-phosphatase in vitro argue against this. Our proposal that Thr366 is a substrate for autodephosphorylation implies that this phosphorylated residue would, at least transiently, occupy the PTEN active site. One speculative interpretation of the relatively slow apparent rate of autodephosphorylation (estimated cellular half-life measured in tens of minutes; Fig. 3C and fig. S3) would imply that this threonine residue may be a poor substrate for PTEN and, through prolonged residence, occlude access to the active site.

We were unable to show in vitro that a phosphorylated full-length PTEN protein would autodephosphorylate, which raises the possibility that dephosphorylation may be mediated by another PTEN-associated phosphatase. However, because PTEN Thr366 dephosphorylation appears to be slow and because autodephosphorylation-based control of PTEN function would likely involve additional factors that influence the rate of this reaction, such as unidentified binding partners that promote or inhibit dephosphorylation, we favor a regulated direct mechanism. Some modest support for the inhibitory regulation of PTEN autodephosphorylation by factors such as oxidation or ubiquitination is provided by the observed increases in PTEN Thr366 phosphorylation observed in cells treated with hydrogen peroxide or hyperosmotic stress, which lead to PTEN oxidation or ubiquitination, respectively [Fig. 4A, fig. S4, and (31)]. Our data also provide only some explanation for the increased phosphorylation of Thr366 seen in the protein phosphatase active mutant PTEN G129E. Although we observed somewhat reduced dephosphorylation of a C-terminal tail substrate in vitro by PTEN G129E, it appears to us that the degree of increased cellular Thr366 phosphorylation observed requires an additional mechanism, for example, that this mutation favors a conformation in which the C-terminal tail is less closely associated with the phosphatase domain.

The strength of the comparison of the PTEN-responsive gene expression profiles we observed in vitro with the clinical glioblastoma samples supports the conclusion that the Matrigel cultures used here are a reasonable model of the in vivo conditions. In particular, it confirms that the set of genes that we observe to be highly responsive to PTEN, without correlating with PTEN’s effects on AKT, accurately represent at least a subset of those genes that correlate with PTEN loss in glioblastoma.

It will be important to determine the signaling mechanisms responsible for those effects of PTEN on invasion and gene expression that do not correlate with its inhibition of AKT phosphorylation. A deeper understanding of the downstream pathways activated in tumors with complete or selective loss of PTEN function should have important implications regarding the use of therapies targeting downstream events.

Materials and Methods

Cell culture and cell-based assays

U87MG glioblastoma and HEK293T cells were cultured as previously described (21). NCI-H196 cells were purchased from the American Type Culture Collection and cultured in RPMI-1640 medium and 10% fetal bovine serum (FBS). U343MG cells were purchased from Cell Lines Service and cultured in minimum essential medium, 10% FBS, and nonessential amino acid mix (Invitrogen Gibco). DBTRG-05MG cells were cultured in RPMI medium supplemented with 10% FBS and nonessential amino acids. Lentiviral particles were prepared as described elsewhere (21). U87MG cells were transduced with lentiviruses in medium supplemented with polybrene (16 μg/ml; hexadimethrine bromide, Sigma). The medium was changed 24 hours after transduction, and the cells were processed 48 hours after transduction.

3D Matrigel cultures of U87MG cells were prepared as described elsewhere (21). For these experiments, cells were transiently transduced in adherent culture 48 hours before seeding into Matrigel. For the analysis of cell morphology, after 8 hours, the cultures were fixed with 2% paraformaldehyde in phosphate-buffered saline (PBS) for 30 min and then photographed by phase-contrast microscopy. For each sample, 10 images were taken in random fields, and the cells were counted by means of the National Institutes of Health (NIH) ImageJ cell counter software.

For Transwell Matrigel invasion assays, U87MG cells were transduced with lentiviral particles encoding wild-type or mutant forms of PTEN. Forty-eight hours after transduction, the cells were serum-starved in migration buffer [Dulbecco’s modified Eagle’s medium (DMEM), 1% bovine serum albumin (BSA), and 0.5% FBS] for an hour. For inhibitor treatments, 1 μM PI103 or 1 μM Akti-1/2 was added to the migration buffer. The cells were then scraped off in migration buffer and 400,000 cells in 500 μl were plated onto the upper chamber of Matrigel Transwell invasion chambers (BD Biosciences). Migration buffer (600 μl) was added to the lower chambers. The cells were allowed to invade for 16 hours and fixed and stained with a Quick-Diff staining kit (Reagena). The chambers were dried and imaged at 10× magnification, and the cells were counted with the NIH ImageJ cell counter software.


For protein analysis, adherent cells were washed twice in ice-cold PBS and lysed in ice-cold lysis buffer [25 mM tris-HCl (pH 7.4), 150 mM NaCl, 1% NP-40, 1 mM EGTA, 1 mM EDTA, 5 mM sodium pyrophosphate, 10 mM β-glycerophosphate, and 50 mM sodium fluoride] containing freshly added 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM benzamidine, aprotinin (10 μg/ml), leupeptin (10 μg/ml), 1 mM sodium orthovanadate, and 0.1% 2-mercaptoethanol. Microcystin (1 μM) was added when used for P-T366 analysis. Equal amounts of proteins were separated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) with precast 4 to 12% gradient gels (Invitrogen) and blotted onto polyvinylidene difluoride (PVDF) membranes (Polyscreen; NEN/PerkinElmer). Most reagents for electrophoresis and blotting were purchased from Invitrogen, and standard manufacturers’ protocols were followed. Western blotting analysis directly of lysates from cells cultured in Matrigel was limited to antibodies with the highest target selectivity by the high concentrations of sample protein derived from this matrix. Quantitation of blotting data was performed with AIDA densitometry software to analyze images obtained either directly from blots with a charge-coupled device camera or from processed film with a transmission scanner. Details regarding the specific antibodies used in this study are provided in the Supplementary Materials and Methods. Quantitation of immunoblot data in Figs. 3 and 4 is provided in fig. S7, and quantitation and analysis of immunoblot data from Figs. 6 and 7 are provided in fig. S8.

Plasmids and lentiviruses

Plasmid and lentiviral expression vectors for untagged and glutathione S-transferase (GST)–tagged human PTEN (21, 31) and GFP-PTEN (46) have been previously described. The expression vectors pHR-SIN PTEN Y138C and pGEX6P1 PTEN Y138C were prepared by site-directed mutagenesis of the corresponding PTEN wild-type constructs with the following primers. F-GTGTAATGATATGTGCATGTTTATTACATCGGGGC and R-GCCCCGATGTAATAAACATGCACATATCATTACAC. PTEN G129R was prepared by site-directed mutagenesis of the pHR-SIN PTEN wild-type constructs with the following primers: F-GATTTCCTGCAGAAAGACGTGAAGGCGTATACAGG and R-CCTGTATACGCCTTCACGTCTTTCTGCAGGAAATC. PTEN L42R was prepared by site-directed mutagenesis of the pHR-SIN PTEN wild-type constructs with the following primers: F-CAATTCACTGTAAAGCTGGAAAGAGACGAACTGGTGTAATG and R-CATTACACCAGTTCGTCTCTTTCCAGCTTTACAGTGAATTG. Plasmids encoding AKT1, AKT2, and AKT3 were gifts from D. Alessi (University of Dundee). The E17K mutagenesis of AKT1 used the primers F-CTGCACAAACGAGGGAAGTACATCAAGACCTGG and R-CCAGGTCTTGATGTACTTCCCTCGTTTGTGCAG. The E17K mutagenesis of AKT2 used the following primers: F-CACAAGCGTGGTAAA TACATCAAGACC and R-GGTCTTGATGTATTTACCACGCTTGTG. A FLAG tag and Bam HI site were added upstream of the Akt2-E17K with the primers F-GTGGGATCCACCGCCATGGATTACAAGGATGACGACGATAAAAATGAGGTGTCTGTCATC and R-CACGGATCCTCACTCGCGGATGCTGGC, and it was cloned into the Bam HI site of pHR-SIN vector. The E17K mutagenesis of AKT3 was done with the following primers: F-CAGAAGAGGGGAAAATATATAAAAAAC and R-GTTTTTTATATATTTTCCCCTCTTCTG. A FLAG tag and Bgl II site were added upstream of the Akt3-E17K with the primers F-GTGAGATCTACCGCCATGGATTACAAGGATGACGACGATAAAAGCGATGTTACCATTGTG and R-CACAGATCTCTTATTCTCGTCCACTTG, and it was cloned into the Bam HI site of pHR-SIN vector. An expression vector for GST-PTEN tail (351 to 403) was produced from a GFP-PTEN tail (351 to 403) vector (47) with the restriction enzymes Eco RI and Not I. For PTEN knockdown, pLKO.1-Puro lentiviral shRNA vectors were purchased from Sigma/TRC. The puromycin resistance gene was replaced with a GFP complementary DNA (cDNA) through a Bam HI–Kpn I cloning. All PCR clonings and mutagenesis were performed with the KOD Hot Start Polymerase (Novagen). Dpn I enzyme was purchased from Promega.

Microarray analysis of gene expression in 3D cultures

Twelve-well plates were coated with 300 μl of Geltrex (Invitrogen) and allowed to gel in a 37°C incubator for 60 min. U87MG cells transiently expressing GFP or PTEN proteins were trypsinized, and 1 × 105 cells were suspended in 500 μl of medium and placed on top of the Geltrex gel. After 4 hours, the medium was aspirated from the culture and an additional 500-μl layer of Geltrex was overlaid and allowed to gel for 60 min before 0.5 ml of medium was added. These 3D cultures were incubated for 16 hours. The top medium was then removed and total RNA was isolated with RNeasy kit from Qiagen following standard manufacturers’ protocol. Expression microarray analysis was performed by CXR Biosciences. RNA integrity was verified with an Agilent Bioanalyzer. One-color hybridization of Cy3-labeled duplicate samples was to Agilent 4x44K Whole Human Genome Oligo Microarray slides (G4112F). Initial data analysis used Rosetta Resolver 6 software to identify probes significantly different between samples in pairwise comparisons. A minimum probability threshold of P < 0.01 was applied in these comparisons. For RT-PCR, 2 to 5 μg of total RNA were converted to cDNA with the Sprint RT Random Hexamer Kit (Clontech). Quantitative PCR used the TaqMan assay system (Applied Biosystems) and a StepOnePlus Real-Time PCR System (Applied Biosystems). The expressed PTEN coding sequences from NCI-H196, U343MG, and HEK293T cells were sequenced by preparing total RNA from each cell type and converting to cDNA as described above. Full-length PTEN cDNA was PCR-amplified, and the product was sequenced with internal primers (; University of Dundee).

In vitro PTEN enzyme assays and cellular PIP3 measurements

The preparation of 3-33P–labeled phosphoinositide substrates, the purification of GST-tagged PTEN, the removal of the GST tag, and further purification of untagged PTEN have been described previously (48). The phosphorylation of the peptide polymer 4:1 poly(GluTyr) (Sigma) with insulin receptor kinase (Upstate) has also been previously described (47, 49). PIP3 assays were conducted with substrate vesicles prepared by sonication of 100 μM phosphatidylcholine, 1 μM unlabeled PIP3, and 100,000 dpm [3-33P]PIP3. These were incubated in 10 mM Hepes (pH 7.4), 125 mM NaCl, 1 mM EGTA, and 10 mM dithiothreitol (DTT) with 100 ng of enzyme at 37°C for the indicated time. Poly(Glu-Tyr)P phosphatase assays were conducted in 25 mM Hepes (pH 7.4), 1 mM EGTA, and 10 mM DTT with 2 μg of enzyme and 100,000 dpm (about 1 μg) of phosphorylated substrate per assay, also at 37°C for the indicated time. Reactions were terminated directly by the addition of 500 μl of ice-cold 1 M perchloric acid and BSA (100 μg/ml), left on ice for 30 min, and spun at 14,000 rpm at 4°C for 10 min. The supernatant was removed, and ammonium molybdate was added to a final concentration of 10 mg/ml. After extraction with two volumes of toluene/isobutanol [1:1 (v/v)], the upper phase was removed and radioactivity was determined by scintillation counting. Cellular PIP3 measurements were made as described (50).

PTEN dephosphorylation in vitro

A PTEN C-terminal tail substrate was purified by expressing in bacteria a fusion protein consisting of this region of PTEN (351 to 403) with an N-terminal GST tag, following expression and purification conditions described for GST-PTEN unless specified. Expression was induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at 30°C for 4 hours. Glutathione Sepharose beads (Amersham) with bound GST-PTEN tail were washed with wash buffer [50 mM tris (pH 7.4), 400 mM NaCl, 1 mM EDTA, and protease inhibitors]. About 1 mg of immobilized GST-PTEN tail protein was phosphorylated with 1 U of CK2 using 2 mM unlabeled adenosine triphosphate (ATP) for 90 min in buffer [20 mM Hepes (pH 7.4), 10 mM MgCl2, and 1 mM DTT] to phosphorylate Ser370 along with the Ser380-Ser385 cluster. The kinase was then extensively washed away with wash buffer before 1 U of purified GSK3β and 20 mM unlabeled ATP and radiolabeled [γ-32P]ATP (adjusted to 20,000 counts per picomole of phosphate incorporated) were added for 90 min to incorporate radiolabeled phosphate specifically onto Thr366. The phosphorylated GST-PTEN tail protein was then extensively washed again in wash buffer (above) before the protein was eluted with 20 mM reduced glutathione. This purified substrate was then stored at −80°C in 25% glycerol and used in PTEN phosphatase assays as described above for the poly(GluTyr) substrate. Experiments were performed with two independent preparations of phosphorylated substrate. For experiments with λ-phosphatase, GFP-PTEN was expressed in 293T cells, immunoprecipitated with GFP-Trap beads (ChromoTek), and washed three times in lysis buffer and four times in assay buffer. The beads were incubated with 800 to 1000 U of GST–λ-phosphatase [Division of Signal Transduction Therapy (DSTT), University of Dundee] in a buffer containing 50 mM tris-HCl, 0.1 mM Na2EDTA, 5 mM DTT, 2 mM MnCl2, and 0.01% Brij-35 (pH 7.5 at 25°C) for 1 hour at 30°C.

Statistical analyses

The significance of differences between sample groups was tested either with the Student’s t test or by one-way analysis of variance (ANOVA) and Tukey test as detailed in each figure legend. Further statistical tests applied during the analysis of microarray data are described in the Supplementary Materials and Methods.

Supplementary Materials

Materials and Methods

Fig. S1. Regulation of AKT phosphorylation by wild-type PTEN and PTEN Y138L.

Fig. S2. Assigned gene sets associated with transcript changes induced by wild-type PTEN expression in U87MG cells.

Fig. S3. Autodephosphorylation analysis of PTEN mutant proteins.

Fig. S4. Effects of kinase inhibition and hyperosmotic stress on PTEN Thr366 phosphorylation.

Fig. S5. Further analysis of PTEN Thr366 phosphorylation.

Fig. S6. Effects of PTEN on U87MG cell morphology in 3D Matrigel cultures.

Fig. S7. Quantitation of blotting data.

Fig. S8. Quantitation and statistical analysis of blotting data.

Table S1. Wild-type PTEN–responsive gene list used for clinical data comparison.


References and Notes

Acknowledgments: We thank H. McLauchlan, J. Hastie, and the staff in the DSTT (University of Dundee) for provision of purified antibodies and purified λ-phosphatase. We also thank D. Alessi for providing cDNAs encoding AKT1, 2, and 3 and P. Downes for helpful discussions and critical reading of the manuscript. Funding: P.T. was supported by a Biotechnology and Biological Sciences Research Council Dorothy Hodgkin studentship. G.Z. was funded by a Wellcome Trust Clinical Ph.D. Fellowship. This work was funded by grants to N.R.L. from the Medical Research Council (G0801865) and the Association for International Cancer Research (04029) and by the pharmaceutical companies of the DSTT consortium (AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Merck Serono, and Pfizer). Author contributions: P.T., G.Z., L.S., H.M., A.G., N.M.P., L.D., and N.R.L. designed and performed the experiments. P.T., G.Z., L.S., N.S., G.J.B., and N.R.L. analyzed the data. P.T., N.S., and N.R.L. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The microarray data have been submitted to the ArrayExpress Archive with accession number E-MTAB-989.
View Abstract

Navigate This Article