PTEN M-CBR3, a versatile and selective regulator of inositol 1,3,4,5,6-pentakisphosphate (Ins(1,3,4,5,6)P5). Evidence for Ins(1,3,4,5,6)P5 as a proliferative signal.

The PTEN (phosphatase and tensin homologue deleted on chromosome 10) tumor suppressor is a phosphatidylinositol 3,4,5-trisphosphate (PtdInsP3) 3-phosphatase that plays a crucial role in regulating many cellular processes by antagonizing the phosphoinositide 3-kinase signaling pathway. Although able to metabolize soluble inositol phosphates in vitro, the question of their significance as physiological substrates is unresolved. We show that inositol phosphates are not regulated by wild type PTEN, but that a synthetic mutant, PTEN M-CBR3, previously thought to be inactive toward inositides, can selectively regulate inositol 1,3,4,5,6-pentakisphosphate (Ins(1,3,4,5,6)P5). Transfection of U87-MG cells with PTEN M-CBR3 lowered Ins(1,3,4,5,6)P5 levels by 60% without detectable effect on PtdInsP3. Although PTEN M-CBR3 is a 3-phosphatase, levels of myo-inositol 1,4,5,6-tetrakisphosphate were not increased, whereas myo-inositol 1,3,4,6-tetrakisphospate levels increased by 80%. We have used PTEN M-CBR3 to study the physiological function of Ins(1,3,4,5,6)P5 and have found that Ins(1,3,4,5,6)P5 does not modulate PKB phosphorylation, nor does it regulate clathrin-mediated epidermal growth factor receptor internalization. By contrast, PTEN M-CBR3 expression, and the subsequent lowering of Ins(1,3,4,5,6)P5, are associated with reduced anchorage-independent colony formation and anchorage-dependent proliferation in U87-MG cells. Our results, together with previously published data, suggest that Ins(1,3,4,5,6)P5 has a role in proliferation.

PTEN (phosphatase and tensin homologue deleted on chromosome 10) 1 is a dual specificity phosphatase that is mutated in a wide range of human sporadic tumor types (1). The PTEN gene encodes a 403-amino acid protein, which is a member of the protein-tyrosine phosphatase family. However, there have been no good phosphoprotein substrates identified to date. The tumor suppressor function of PTEN relies on its ability to metabolize acidic nonprotein substrates (2,3). Indeed, PTEN dephosphorylates the signaling molecules, phosphatidylinositol 3-phosphate (PtdIns(3)P), phosphatidylinositol 3,4-bisphosphate (PtdIns(3,4)P 2 ), phosphatidylinositol 3,5-bisphosphate (PtdIns(3,5)P 2 ), and phosphatidylinositol 3,4,5-trisphosphate (PtdInsP 3 ) in vitro by removal of the phosphate at the 3-position of these substrates. The preferred substrate was found to be PtdInsP 3 by a factor of some 200-fold (4). In this respect, PTEN acts as a functional antagonist of phosphoinositide 3-kinase signaling pathways, promoting apoptosis and inhibiting cell-cycle progression (5)(6)(7). Evidence for this includes a naturally occurring mutation (PTEN G129E), identified in sufferers of Cowden disease, in which patients encounter multiple hamartomatous lesions, especially of the skin, mucous membranes, breast, and thyroid. The PTEN G129E mutant has comparable activity to PTEN against the synthetic phosphoprotein substrate, polyGluTyr P , but is unable to dephosphorylate inositol lipids, indicating that the lipid phosphatase activity and not protein phosphatase activity is required for tumor suppressor function (3).
PTEN has several structural features, including an N-terminal phosphatase domain requiring a reduced cysteine (Cys 124 ), a calcium-independent C2 domain, that has been shown to bind lipid vesicles in vitro, and a sequence shown to bind PDZ domains (see Ref. 1). In cells, PTEN exists as a phosphoprotein, with phosphorylation occurring at a region to the C terminus of the C2 domain (8,9). It has recently been shown that the C2 domain, and not the PDZ-binding sequence, plays a crucial role in membrane targeting and substrate specificity (10 -12). Functional interference with this C2 domain, exemplified by the artificially modified PTEN M-CBR3 protein first described by Lee et al. (10), causes a reduction in the ability to interact with lipid membranes but marginally increases phosphatase activity toward inositol phosphate substrates (4,10). Expression of GFP-tagged PTEN suggests that it is predominantly cytoplasmic, in agreement with most studies utilizing PTEN-selective antibodies (11,13,14). Its role as a lipid phosphatase, however, requires interaction with membranes, such that PTEN M-CBR3 is unable to regulate PtdInsP 3 levels, whereas myristoylated PTEN, which is anchored to the membrane, is more effective than PTEN in altering effects downstream of PtdInsP 3 (12,15).
It has recently been suggested that the cellular substrates of PTEN may include inositol phosphates, particularly Ins(1,3,4,5,6)P 5 (16). Therefore, the effects of PTEN could be mediated by regulation of these inositol phosphates. In this study we have clarified the effects of PTEN expression on inositol lipid and inositol phosphate levels in cells. PTEN expression lowered PtdInsP 3 and PtdIns(3,4)P 2 . We found no suggestion that PTEN could be a physiological regulator of inositol phosphates. The PTEN M-CBR3 mutant, however, selectively affected inositol phosphate levels, especially Ins(1,3,4,5,6)P 5 , without altering the levels of 3-phosphoinositide lipids. With this new insight into the effects of PTEN M-CBR3 we have re-evaluated previously published data in determining the roles played by Ins(1,3,4,5,6)P 5 and suggest that Ins(1,3,4,5,6)P 5 is a proliferative agent, because lowering its levels correlated with decreased cell growth and anchorageindependent colony formation.

EXPERIMENTAL PROCEDURES
Cell Culture-Tissue culture media and additives were provided by Invitrogen. U87-MG cells, obtained from the European Collection of Animal Cell Cultures, were maintained in minimal essential medium, plus 2 mM glutamine, 1% nonessential amino acids, 1 mM sodium pyruvate, and 10% fetal bovine serum. Expression vectors were introduced into the U87-MG cells using a previously described baculoviral delivery system adapted for mammalian expression (12). Assays were performed 24 h following DNA delivery for all experiments except FACS analysis, where cells were analyzed after 36 or 60 h. Levels of expressed protein were determined using fluorescence, as well as Western blotting of extracts from U87-MG cells expressing GFP-tagged protein. Transfection of U87-MG cells for proliferation assays was performed as described below. Vectors were prepared as previously described, except PTEN M-CBR3/G129E, which was prepared by cleaving PTEN G129E and PTEN M-CBR3 with PpuMI and BamHI in FastBacMam-EGFP, and replacing the wild type C-terminal region of the protein (from PTEN G129E) with that containing the M-CBR3 mutation. The sequence was verified, and virus was prepared as above.
Internalization of 125 I-EGF-We performed 125 I-human EGF (Amersham Biosciences) internalization assays as described by Sorkina et al. (19). Cells were grown in 24-well dishes. Unless otherwise stated, 125 I-EGF was added to cells in minimum essential medium with Earle's salts containing 0.1% bovine serum albumin at 37°C for up to 10 min. At the end of the incubation, the medium was aspirated and the monolayers were washed three times with ice-cold minimum essential medium with Earle's salts to remove unbound ligand. The cells were then incubated for 5 min with 0.2 M acetic acid (pH 2.8) containing 0.5 M NaCl at 4°C. The acid wash was combined with another short rinse in the same buffer and used to determine the amount of surface-bound 125 I-EGF. The cells were then lysed in 1 M NaOH to determine the intracellular (internalized) radioactivity. The ratio of internalized to surface radioactivity was plotted against time. Nonspecific binding was determined in the presence of 200 ng/ml unlabeled EGF.
Flow Cytometric Analysis of Cell Cycle Distribution-Adherent cells were harvested by trypsinization, washed once in PBS, and re-suspended in ice-cold 70% (v/v) ethanol in water. Cells were washed twice in PBS plus 1% (w/v) bovine serum albumin and stained for 20 min in PBS plus 0.1% (v/v) Triton X-100 containing 50 g/ml propidium iodide and 50 g/ml RNase A. The DNA content of cells was determined using a FACSCalibur flow cytometer (BD Biosciences) and CellQuest software. Red fluorescence (585 Ϯ 42 nm) was acquired on a linear scale, and pulse width analysis was used to exclude doublets. Cell cycle distribution was determined using FlowJo software (Tree Star Inc.).
Proliferation Assays-Anchorage-independent colony assays were adapted from those described previously (20). Briefly, U87-MG cells were transiently transfected (FuGENE 6, Roche Applied Science) with pCDNA3.1ϩ alone, or PTEN expression constructs. 24 h after transfection, cells were suspended in 15% serum-containing media with 0.5 mg/ml G418 and 0.3% agar and layered in triplicate onto 0.6% agar medium in 6-well plates. Plates were then incubated for 3 weeks, with the addition of 0.5 ml of fresh medium after 10 days. To test anchoragedependent growth, a similar method was employed to that used by Furnari et al. (21). U87-MG cells were transfected and changed to fresh medium with 1 mg/ml G418 24 h post-transfection. Five days after transfection, nontransfected controls had very little viability. Cell numbers were determined at days 5, 7, and 9 using CellTitre96 reagent (Promega) according to the manufacturer's instructions.

RESULTS
As previously described, expression of PTEN in U87-MG cells caused a decrease in PtdInsP 3 levels. Greater expression of PTEN (far above the levels of endogenous expression in PTEN-positive cells) caused a significant decline in PtdIns(3,4)P 2 but not of PtdIns(3)P and PtdIns(3,5)P 2 (Fig. 1). We have attempted to express lower levels of wild type and mutant PTEN to study the cellular consequences of such protein expression.
A comparison with extracts obtained from cells with normal PTEN status showed that levels of expression of GFP-PTEN were comparable with the range of PTEN expression found in a variety of cell types (Fig. 2). We were surprised to see that 1321N1 astrocytoma cells were either PTEN-null or expressed levels of protein below the detection limit of these assays, because this had not previously been reported and these cells have normally low basal levels of PtdInsP 3 and PKB phosphorylation. Under the conditions used, levels of PtdIns(3)P, PtdIns(3,4)P 2 , and PtdIns(3,5)P 2 were not significantly decreased (Table I). Expression of similar levels of the catalytically dead mutant, PTEN C124S, and the lipid phosphatasedead mutant, PTEN G129E (which retains phosphoprotein phosphatase activity), slightly raised PtdInsP 3 levels. It has been suggested that this is due to "substrate trapping," resulting in a stable lipid-enzyme complex, protecting the PtdInsP 3 from metabolism by other phosphatases (3). PTEN M-CBR3, PTEN G129E/M-CBR3, and EGFP did not alter inositol lipid levels (Table I).
Expression of PTEN M-CBR3 in U87-MG cells was unable to alter the phosphorylation state of PKB. Expression of similar levels of wild type PTEN, which lowers PtdInsP 3 levels, caused a marked decrease in phospho-T 308 and S 473 -PKB, whereas PDGF receptor activation increased phosphorylation at both sites (Fig. 4). These results suggest that Ins(1,3,4,5,6)P 5 does not affect PKB phosphorylation directly or interfere with Pt-dInsP 3 dependent phosphorylation of PKB. In this respect, PKB appears to have a higher selectivity toward inositol lipids than previously suggested (22,23). These results reiterate our recent findings with the substrate specificity of PTEN, in that short-chain inositol lipid analogues, or deacylated lipids can yield misleading data with respect to ligand affinities, because such studies fail to replicate binding in the context of a complete biological membrane surface.
Inositol phosphates, including Ins(1,3,4,5,6)P 5 , have been implicated in inhibition of clathrin-mediated internalization, by means of preventing triskelion formation (24,25). EGFR internalization was monitored by means of incubating 125 I-EGF with U87-MG cells. Rates of internalization were identical whether PTEN M-CBR3 was expressed or not. Incubating U87-MG cells with 125 I-EGF at 4°C prevented any internalization (Fig. 5). Similarly, PD158780, an EGFR kinase inhibitor previously shown to block receptor phosphorylation, and hence internalization, was found to behave as expected, significantly inhibiting internalization (data not shown).
Several complex cellular processes have been identified that rely to some degree on phosphoinositide 3-kinase signaling and can be inhibited by PTEN, including colony formation in soft agar (10,26) and cell spreading (12). The ability of PTEN M-CBR3 to inhibit anchorage-independent colony formation was compared with that of wild type PTEN. U87-MG cells were transfected with expression vectors for untagged PTEN proteins carrying a Neomycin/G418 resistance gene and after 24 h seeded into soft agar with G418 selection. Cells transfected with vector alone or phosphatase-dead PTEN formed large numbers of colonies within 3 weeks (Fig. 6a). PTEN greatly inhibited colony formation, but this was not mimicked by the PTEN C124S, PTEN G129E, or PTEN G129E/M-CBR3 mutants. PTEN M-CBR3 had a small, but significant effect in reducing colony number (Fig. 6a), confirming its effects on anchorage-independent growth.
The effects of PTEN on anchorage-independent growth and cellular proliferation appear to be mediated by its ability to cause G 1 arrest. The effects of PTEN M-CBR3 on cell-cycle distribution were also monitored. Cells treated with baculoviral expression vectors for 36 or 60 h were analyzed by FACS. As Profiles of all other mutants tested showed no difference to GFP controls (Fig. 7), suggesting that Ins(1,3,4,5,6)P 5 is not required to pass through any particular part of the cell cycle. DISCUSSION The biological roles of PTEN have generally been attributed to its ability to metabolize the lipid second messenger, Pt-dInsP 3 . More recently, it has been shown that PtdIns(3,5)P 2 , PtdIns(3,4)P 2 , and PtdIns(3)P are also substrates in vitro (28,29). We show clearly that the primary substrate for PTEN is PtdInsP 3 and that the inositol lipid bisphosphates are not lowered following PTEN expression at close to physiological levels in U87-MG cells. Higher levels of expression can regulate PtdIns(3,4)P 2 but not PtdIns(3,5)P 2 . This may reflect the differences in rate of hydrolysis observed previously (4) or the relative rates of turnover of these molecules in vivo. Alternatively, this may merely reflect the product-precursor relationship existing between PtdInsP 3 and PtdIns(3,4)P 2 via 5-phosphatases, such as SHIP and SHIP2 (30), whereas PtdIns(3,5)P 2 synthesis is likely to be independent of PtdInsP 3 .
It has been shown that PTEN can also dephosphorylate inositol phosphates in vitro, but their significance as physiological substrates has not been fully resolved. Ins(1,3,4,5)P 4 was shown to be a weaker substrate than PtdInsP 3 by a factor of between 10 2 and 10 4 depending upon the conditions of assay (4). Ins(1,3,4,5,6)P 5 and InsP 6 are considered to be the substrates of a distinct phosphatase, MIPP. Indeed, studies involving brain and liver extracts from MIPP knockout mice were unable to detect any significant Ins(1,3,4,5,6)P 5 phosphatase activity in preparations where PTEN should have been present (31), although these assays were performed in the absence of a reducing agent, required for optimal activity of PTEN. PTEN overexpression, however, did lower cellular Ins(1,3,4,5,6)P 5 levels (16). In agreement with this study, we found that PTEN expression lowered Ins(1,3,4,5,6)P 5 levels (Table II and Fig. 3), but because these effects were also observed following expression of the catalytically inactive PTEN C124S, and the lipid-phosphatase inactive PTEN G129E mutants, we conclude that this effect is not mediated by the phosphatase activity of PTEN. They are, however, related to expression of PTEN-like proteins, because the same decline in Ins(1,3,4,5,6)P 5 is not observed when EGFP alone is expressed. We have previously found that inositol lipid metabolism by PTEN requires the C2 domain and that interfering with this domain (as with the PTEN M-CBR3 mutant) severely impedes lipid phosphatase activity but enhances the activity observed using a soluble substrate (4,10). We now show that this artificial mutant can lower Ins(1,3,4,5,6)P 5 levels without similar effects on inositol lipids or other inositol phosphates. We also note that levels of Ins(1,3,4,5)P 4 , another substrate of PTEN M-CBR3, are not affected. This observation can be explained by the relative rates of turnover of each molecule and their product-precursor relationships. The turnover of Ins(1,3,4,5)P 4 by endogenous phosphatases is far more rapid than that of Ins(1,3,4,5,6)P 5 . The concomitant rise in Ins(1,3,4,6)P 4 levels is likely indicative of a compensatory increase in Ins(1,3,4,5,6)P 5 biosynthesis, because the former is considered to be the metabolic precursor of Ins(1,3,4,5,6)P 5 . These cellular effects are mediated by the lipid phosphatase-like activity of PTEN M-CBR3, because PTEN G129E/M-CBR3, which should retain its protein phosphatase activity while losing activity toward inositol phosphates, was without effect. This suggests that Ins(1,3,4,5,6)P 5 is a key factor in mammalian cells and that, under normal circumstances, its level is under tight control. These effects are present without any detectable effect on any known inositol lipid. PTEN M-CBR3 thus provides a valuable tool whereby Ins(1,3,4,5,6)P 5 can be selectively regulated. This has enabled us to evaluate some of the physiological roles that have been previously ascribed to Ins(1,3,4,5,6)P 5 .
The pleckstrin homology (PH) domains of many proteins have been associated with the ability of these proteins to associate with membrane surfaces. Although the ability of a small number of these proteins to bind inositol phosphates has been studied (see Refs. 22 and 32), it is only recently that these interactions have been considered to be physiologically rele-  vant. Competition between inositol phosphates and inositol lipids for PH domains is clearly observed in the case of phospholipase C␦ 1 (PLC␦ 1 ).
We were able to address directly whether Ins(1,3,4,5,6)P 5 has a role to play in PKB regulation. We showed that PDGF was able to further raise, and that PTEN expression was able to reduce, the phosphorylation status of PKB in these cells. Reduction of Ins(1,3,4,5,6)P 5 levels following expression of PTEN M-CBR3 was without significant effect on PKB phosphorylation, suggesting it is not a physiological regulator and is incapable either of activating directly or competing effectively with the lipid activators of this protein kinase.
We have also studied the effect of PTEN M-CBR3 expression on protein trafficking, because Ins(1,3,4,5,6)P 5 and InsP 6 have been proposed to attenuate the desensitization of substance P receptors (25). The ability of Ins(1,3,4,5,6)P 5 to inhibit triskelion formation in vitro has been questioned, due to a particularly low affinity for AP-3 (24). The internalization of low levels of EGF is mediated by clathrin-coated pits (see Ref. 19). Low levels of EGF caused rapid internalization that was sensitive to the EGF receptor kinase inhibitor, PD158780, and reduced temperature. The effects were not altered by lowering Ins(1,3,4,5,6)P 5 , suggesting that this inositol phosphate has no role to play in trafficking of tyrosine kinase-coupled receptors. Our approach has been somewhat different to that of other studies, for example, in which InsP 6 is injected into oocytes or other cells and the consequences monitored (25). We have determined the consequences of PTEN M-CBR3 expression on endogenous inositol phosphate levels, whereas the injection of a particular inositol phosphate does not guarantee that it is not metabolized to generate other compounds with biological activity. It also remains possible that Ins(1,3,4,5,6)P 5 -mediated trafficking is strictly limited to serpentine receptors coupled to heterotrimeric G-proteins.
The higher inositol phosphates, Ins(1,3,4,5,6)P 5 and InsP 6 , have also been implicated in cell proliferation. Overexpression of cytosolic MIPP, achieved by removal of the N-terminal endoplasmic reticulum-targetting sequence and the C-terminal endoplasmic reticulum-recycling signal (Ser Asp Glu Leu), has been shown previously to lower Ins(1,3,4,5,6)P 5 by 60%, and InsP 6 levels by 40%, and to cause a decrease in the rate of cell proliferation (31). It has also been reported that the transition from proliferation to hypertrophy in chicken chondrocyte maturation is accompanied by the up-regulation of Band 17, subsequently identified as the chicken homologue of MIPP (33). These results suggest that a reduction in cell growth correlates with a reduction in the levels of Ins(1,3,4,5,6)P 5 .
Expression of PTEN M-CBR3 has been shown to inhibit proliferation of U87-MG (10,15) and LNCaP cells (15). We show that these effects correlate with a decline in Ins(1,3,4,5,6)P 5 without affecting PtdInsP 3 levels (Fig. 6, this  study). These results would suggest that Ins(1,3,4,5,6)P 5 alone FIG. 4. Only wild type PTEN alters PKB phosphorylation levels. U87-MG cells were transfected using virus as described in the legend to Table II. Supernatants were prepared from lysates, and these (20 g of total protein) were analyzed by SDS-PAGE (Novex) and probed with anti-phospho-T 308 , phospho-S 473 , and total PKB antibodies.  6. a, inhibition of anchorage-independent colony formation by PTEN proteins. U87-MG cells were transfected with expression vectors for untagged PTEN proteins also carrying antibiotic resistance genes. Cells were seeded in soft agar and selected in antibiotic for 3 weeks. Colonies formed were counted, and numbers directly compared with vector-transfected controls. Data points are presented as % control Ϯ S.E. calculated from six dishes plated from three independent transfections. b, inhibition of cellular proliferation by PTEN proteins. U87-MG cells were transfected with expression vectors for untagged PTEN proteins. Cells were selected in G418 and plated into 96-well multiplates and assayed for viable cells after 5, 7, or 9 days. Data (representative of n ϭ 2) are presented as mean Ϯ S.E. absorbance at 490 nm per well from 6 wells.
can alter growth rates of cells and act as a proliferative agent itself. PKB phosphorylation is not altered by PTEN M-CBR3 overexpression in any of the studies described. Because PKB phosphorylation is highly sensitive to small changes in PtdInsP 3 concentration (34), this suggests that PtdInsP 3 pools are not affected by PTEN M-CBR3, strengthening the argument that Ins(1,3,4,5,6)P 5 is the causative agent in altering proliferation.
The studies of anchorage-independent and anchoragedependent proliferation (this study) yield much the same data as those described above. Overexpression of PTEN M-CBR3, which lowers Ins(1,3,4,5,6)P 5 levels, decreased colony number and proliferation rate, although not to the same extent as wild type PTEN. Cell cycle analysis further strengthens our argument that the effects observed using PTEN M-CBR3 are mediated by Ins(1,3,4,5,6)P 5 and not by PtdInsP 3 . PTEN M-CBR3expressing cells, while growing more slowly than control cells, showed no change in their cell cycle profile (Fig. 7). In contrast, PTEN-expressing cells showed evidence of G 1 arrest.
In summary, we have characterized the effects of PTEN expression on inositol phosphate and inositol lipid levels and shown they are limited to inositol lipids, primarily PtdInsP 3 . Endogenous PTEN probably does not play a significant role in metabolism of inositol phosphates in vivo. We have determined that the PTEN M-CBR3 mutant specifically lowers Ins(1,3,4,5,6)P 5 without affecting PtdInsP 3 . Using PTEN M-CBR3 as a selective tool, we find that Ins(1,3,4,5,6)P 5 plays no significant role in PKB phosphorylation or receptor trafficking, but plays a positive role in proliferation, albeit not as strongly as PtdInsP 3 . PTEN M-CBR3 is a versatile and specific regulator of Ins(1,3,4,5,6)P 5 and can be used to determine the physiological roles played by this relatively abundant and widespread inositol phosphate. FIG. 7. Cell cycle effects of PTEN proteins. Flow cytometric analysis of U87-MG cells transfected using virus for 60 h. DNA was stained with propidium iodide, and cellular content was analyzed. The percentage of cells in G 1 , S, or G 2 /M phases were calculated using CellQuest software.