The nuclear affairs of PTEN

PTEN encodes a major tumor-suppressor protein that is a dual-specificity phosphatase. Inactivation of PTEN has been shown to be involved in heritable and sporadic cancers. Mutation or deletion of PTEN, historically the most commonly identified mechanisms of inactivation of tumor suppressors, is found only in the minority of sporadic non-cultured primary cancers, which indicates that there might be other, novel mechanisms of inactivation. Despite the absence of a classic nuclear localization signal, PTEN enters the nucleus by several mechanisms, including simple diffusion, active shuttling, cytoplasmic-localization-signal-dependent export and monoubiquitylation-dependent import. Cytoplasmic PTEN has a well-known role as a negative regulator of the PI3K/AKT pathway; however, it is becoming clear that cytosolic PTEN is not the same as nuclear PTEN. Nuclear PTEN plays a role in chromosome stability, DNA repair, cell cycle arrest and cellular stability. The balance between these functions is an important factor in determining whether a cell remains benign or becomes neoplastic.


Introduction
Phosphatase and tensin homolog (mutated in multiple advanced cancers 1) (PTEN), located on 10q23.3, encodes a 403-residue dualspecificity phosphatase that has both lipid and protein phosphatase activity. PTEN classically converts phosphatidylinositol-3,4,5trisphosphate (PIP 3 ) in the cytoplasm to phosphatidylinositol-4,5bisphosphate (PIP 2 ), thereby directly antagonizing the activity of PI3 kinase (PI3K) (Leevers et al., 1999) (Fig. 1). Its inactivation results in constitutive activation of the PI3K/AKT pathway and in subsequent increases in protein synthesis, cell cycle progression, migration and survival (Li and Sun, 1998) (Fig. 1). Recombinant PTEN has been shown to dephosphorylate serine, threonine and tyrosine residues of peptides in vitro (Myers et al., 1997) and focal adhesion kinase in vivo (Tamura et al., 1998). In addition, various studies have demonstrated that the protein phosphatase activity of PTEN inhibits activation of mitogen-activated protein kinase (MAPK) via several pathways.
Although numerous somatic mutations have been localized to the PTEN gene, these only occur in a minority of tumors, which indicates that alternative mechanisms, both genetic and non-genetic, of PTEN inactivation must exist. One such mechanism is the regulation of subcellular localization. In this Commentary, we focus on nuclear localization of PTEN, and the importance of this in normal physiology and disease. Other nontraditional mechanisms of regulation are discussed or reviewed elsewhere (Pezzolesi et al., 2006;Teresi et al., 2007;Zbuk and Eng, 2007).

Nuclear PTEN
Early studies of PTEN localization, using immunohistochemical staining techniques, found it in the cytoplasm; however, these focused primarily on overexpressed protein or tumorigenic cell lines and tissues (Furnari et al., 1997;Li and Sun, 1997;Whang et al., 1998). Subsequently, multiple reports placed PTEN in the nucleus of normal primary neurons and endothelial cells (Sano et al., 1999), myoepithelial cells of normal breast ducts (Perren et al., 1999) and normal follicular thyroid cells (Gimm et al., 2000). In pancreatic tissue, Perren and co-workers also observed nuclear PTEN in normal islet cells but predominantly cytoplasmic PTEN in sporadic endocrine pancreatic tumors (Perren et al., 2000). The loss of nuclear PTEN specifically has now been found in a variety of sporadic tumors (Brenner et al., 2002;Depowski et al., 2001;Dreher et al., 2004;Perren et al., 2000). Perhaps the first compelling evidence for a role of nuclear PTEN came from studies of melanoma. These indicated an association between a lack of nuclear PTEN and mitotic index: nuclear PTEN is apparent in quiescent cells, but in actively dividing cells PTEN localizes mostly to the cytosol (Whiteman et al., 2002). Other studies have observed similar trends in different tissue types (Gimm et al., 2000;Perren et al., 2000). This has led to the belief that, in normal, quiescent tissue, PTEN is localized primarily in the nucleus, whereas, in neoplastic tissue, cytoplasmic PTEN predominates. Thus, PTEN in the nucleus appeared to be required for tumorsuppressing ability.
At the time, however, many felt that the presence of PTEN in the nucleus was merely an artifact, because all the data had relied solely on immunohistochemical techniques. Subsequent subcellular fractionation and fluorescence microscopy in the human breast carcinoma cell line MCF-7 (Fig. 2), however, showed that localization of PTEN is cell cycle dependent, with higher nuclear PTEN levels being found in G0-G1 phase and lower levels being found in S phase (Ginn-Pease and Eng, 2003). This was consistent with the PTEN localization patterns observed in normal and tumor tissues: normal, quiescent tissues consisting mostly of cells in G0-G1 phase exhibit a predominantly nuclear localization of PTEN, whereas actively dividing cancerous tissues have a higher percentage of cells in S phase and therefore exhibit increased levels of cytoplasmic PTEN. A change in intracellular localization of PTEN thus seems to be involved in neoplastic transformation and might modulate its activity.
PTEN nucleo-cytoplasmic shuttling PTEN does not contain typical nuclear import (nuclear localization signal; NLS) or nuclear export (nuclear export signal; NES) motifs ( Fig. 3) Trotman et al., 2007). How then does it enter and exit the nucleus? Several mechanisms for the nucleocytoplasmic shuttling of PTEN have been postulated, including simple diffusion, export dependent on a putative cytoplasmic localization signal, active shuttling by the RAN GTPase or major vault protein (MVP), phosphorylation-dependent shuttling and monoubiquitylation-dependent import (see Fig. 3).
Work by Liu and co-workers suggests that PTEN is able to enter the nucleus by simple diffusion (Liu et al., 2005). Diffusion through nuclear pores is limited to proteins that have a molecular mass of less than 60,000 Da (Boulikas, 1993;Peters, 1986), which makes PTEN a candidate given its size (~ 47,000 Da). Using GFP-PTEN fusion proteins of varying sizes, Liu and co-workers found that, in HeLa cells, GFP-PTEN constructs of <60,000 Da can enter the nucleus, whereas larger fusion proteins cannot. They found that diffusion within the cytoplasm itself is reduced in the case of the larger GFP-PTEN fusion proteins and that mutation of residues 13-14 further restricts cytoplasmic movement. Indeed, these same mutations also decrease nuclear localization compared with wildtype PTEN fusion proteins. This led the authors to propose that PTEN interacts with proteins in the cytoplasm that sequester it to prevent nuclear translocation. PTEN might thus enter the nucleus by passive diffusion and its cytoplasmic sequestration could influence nucleo-cytoplasmic distribution.
Active, RAN-mediated nuclear import of PTEN has been documented in U87MG human glioblastoma cells (Gil et al., 2006). Gil and co-workers found that PTEN is excluded from the nucleus of cells expressing the GTPase-deficient, dominant-negative RANQ69L mutant. Additionally, this group demonstrated that mutations affecting phosphorylation of PTEN, which had previously been shown to affect stability of PTEN (Vazquez et al., 2000), alter its nucleo-cytoplasmic localization. In their experiment, mutation of a single phosphorylation site (S370A) caused PTEN to localize to the cytoplasm, similar to the wildtype protein, whereas the T382A mutant had a roughly equal nucleo-cytoplasmic distribution. Mutations of other single phosphorylation sites (S380A, T383A or S385A) produced strong nuclear distributions, as did mutants that have multiple defective phosphorylation sites (S370A/S385A, S380A/T382A/T383A, S370A/S385A/S380A/T382A/T383A). These findings differ from work in MCF-7 human breast cancer cells, in which expression of similar mutants did not result in altered nucleo-cytoplasmic localization of PTEN . Although these reports are contradictory, the model systems used are different. Gil and coworkers used glioblastoma cells lacking PTEN protein expression. The phosphorylation mutants were therefore expressed without a background of endogenous PTEN with which to interact. By contrast, Chung and co-workers used MCF-7 breast cancer cells already expressing PTEN protein. These contradictory data might simply be a result of different model systems (brain cancer versus breast cancer). Alternatively, endogenous PTEN could interact with the PTEN mutants and alter their localization.
A third potential mechanism of PTEN import involves MVP. PTEN binds to MVP, a protein hypothesized to be a general carrier molecule for nuclear-cytoplasmic transport (Mossink et al., 2003), in yeast two-hybrid screens (Yu et al., 2002). Further analysis confirmed that this interaction occurs in both 293T and HeLa cells and localized the binding sites to the C2 domain of PTEN and the EF-hand pair (a calcium-binding motif) of MVP. Studies in MCF-7 cells identified four non-traditional NLS-like sequences: NLS1 to NLS4 . Mutations in single individual sequences do not affect PTEN localization; however, when combinations of two sequences are targeted, specifically NLS2 and NLS3 or NLS3 and NLS4, PTEN is excluded from the nucleus. These double mutants also lack the ability to bind to MVP. Further characterization of the interactions between PTEN and MVP indicated that Ca 2+ positively regulates binding . Bipartite NLS motifs exist in other tumor-suppressor proteins, such as p53 (Efthymiadis et al., 1997) and the retinoblastoma protein pRb (Liang and Clarke, 1999), and links between the nuclear transport of proteins and the development of cancer have been demonstrated (Kau et al., 2004).
Trotman and co-workers have described another potential mechanism of PTEN nuclear import . They observed that a particular PTEN mutation (K289E) in a Cowden syndrome (CS) patient results in the nuclear exclusion of PTEN. The authors determined that K289 is a major target for ubiquitylation. Because monoubiquitylation has been implicated in the transport of other proteins (Li et al., 2003;Massoumi et al., 2006;Plafker et al., 2004), they performed further analyses of this mutation, leading to the hypothesis that monoubiquitylation of PTEN at K289 results in its nuclear import, whereas polyubiquitylation causes proteasome-mediated degradation. The same group identified NEDD4-1 as the E3 ubiquitin ligase responsible, showing that it negatively regulates PTEN . They speculate that the NEDD4-1-PTEN relationship is similar to that between p53 and MDM2 -with a subtle twist. The p53 protein has a short half-life and its levels are regulated by inhibition of degradation. By contrast, PTEN is a relatively stable protein, so expression levels might be modulated by acceleration of degradation. Monoubiquitylation of PTEN by NEDD4-1 would lead to nuclear localization of PTEN, cell cycle arrest and genomic stability; however, polyubiquitylation of PTEN would lead to degradation and a loss of its tumor-suppressor activity.
Denning and co-workers recently identified a cytoplasmic localization signal (CLS) in the N-terminal domain of PTEN (Denning et al., 2007). Mutation of any residue, except residue 22, within the short N-terminal motif spanning residues 19-25 results in increased nuclear localization of PTEN. The CLS might function as either a CLS or a non-classical NES; however, the mechanism remains elusive.
The variety of mechanisms for the nuclear import and regulation of PTEN that have been identified suggests that the mechanism(s) by which PTEN enters the nucleus might be both cell-type-specific and situation-specific (Fig. 3). There appears to be a delicate interplay between the stability/degradation of the PTEN protein and its subcellular localization. Future investigations into these regulatory mechanisms should provide exciting avenues of investigation and might identify potential therapeutic targets given the distinct localization of PTEN in normal and cancerous cells.

Functions of nuclear PTEN
The lipid phosphatase activity of PTEN regulates AKT phosphorylation and p27 expression levels, whereas its protein phosphatase activity regulates MAPK phosphorylation and cyclin D1 expression (Weng et al., 2002;Weng et al., 2001). This bifurcation of effects is further complicated by the alterations in subcellular localization of PTEN that have been observed (Fig. 4).
In MCF-7 cells expressing wild-type or a nuclear-localization mutant of PTEN, levels of phosphorylated AKT (AKT-P) decrease and p27 kip1 levels increase . This indicates that these downstream targets of PTEN activity are regulated by cytoplasmic PTEN. By contrast, expression of PTEN capable of nuclear localization results in a decrease in MAPK phosphorylation and cyclin D1 levels, but expression of a PTEN nuclearlocalization-defective mutant has no effect. These results suggest that PTEN that is specifically localized in the nuclear compartment regulates these two signaling events. Nuclear PTEN also induces a G0-G1 arrest in these cells: cells expressing wild-type PTEN protein exhibit a slower growth rate and higher G 1 :S ratio than cells expressing the nuclear-localization-defective mutant. This increase in the number of cells in G0-G1 phase is probably a direct result of downregulation of cyclin D1 by nuclear PTEN. It might lead to decreased tumor growth, explaining another aspect of PTEN tumor-suppressor function (Chung et al., 2006;Denning et al., 2007;Trotman et al., 2007).
Recent work has described another exciting role for nuclear PTEN: maintaining chromosomal stability (Shen et al., 2007). Fluorescence in situ hybridization (FISH) analysis of Ptendeficient mouse embryo fibroblasts (MEFs) demonstrated spontaneous chromosomal instability in virtually all metaphase spreads examined. The chromosomal alterations included breaks, fragments, translocations and fusions, with a high level of centromeric instability. Further analysis led to the discovery of a physical association between PTEN and centromere-specific binding protein C (CENP-C) required for centromere stability. It also determined that PTEN is necessary and sufficient for induction of RAD51, which leads to double-stranded break (DSB) repair. This work was supported by patient data: PTEN protein from a CS patient with an R189X mutation (Eng, 2003) does not bind to CENP-C. Expression of the R189X mutant in wild-type MEFs results in numerous chromosomal alterations, indicating that this can induce massive chromosomal abnormalities in normal cells. These results have obvious implications for the development of potential anti-cancer therapies because many chemotherapeutic agents act through DNA-damage repair mechanisms.
PTEN might also affect apoptosis. In U87MG human glioblastoma cells, it appears to accumulate in the nucleus following the induction of apoptosis by TNF␣ or doxorubicin (Gil et al., 2006). Following the treatment of cells with TNF␣, the percentage of cells displaying condensed nuclei, a hallmark of apoptosis, is significantly higher in cells that have predominantly nuclear PTEN than those that have predominantly cytoplasmic PTEN. These effects do not depend on PTEN catalytic activity because expression of the phosphatase-dead mutants C124S and G129E does not alter the results. By contrast, studies in MCF-7 cells have determined that cytoplasmic PTEN is pro-apoptotic : expression of wild-type or a nuclearlocalization-deficient PTEN increases the activity of caspase 3/7 compared with untransfected cells. Again, the contradictory studies used different cell types and different background levels of endogenous PTEN. Nevertheless, these results suggest that the equilibrium between nuclear and cytoplasmic PTEN might be important in the regulation of apoptosis.
Undoubtedly, future studies will shed more light on the roles of both cytoplasmic and nuclear PTEN in the regulation of carcinogenesis. It is interesting to note that these preliminary studies seem to indicate that the lipid phosphatase activity of PTEN predominates in the cytoplasmic compartment, whereas the protein phosphatase activity is generally nuclear. Whether this is a general Fig. 3. PTEN motifs involved in nucleo-cytoplasmic shuttling. PTEN consists primarily of two domains: the phosphatase and C-terminal domains. The phosphatase domain consists of residues 1-185, with its core motif (green) spanning residues 124-130. The C-terminal domain, residues 186-403, contains the lipid-binding C2 domain (residues 186-351), two PEST sequences involved in protein stability (residues 350-375 and 379-396) and a PDZ domain (shown in red), which is involved in protein-protein interactions. A putative cytoplasmic localization signal (CLS) spans residues 19-25. Four separate non-classical NLS motifs (NLS1 to NLS4) involved in binding to and the import by MVP have been localized to residues 10-14 (RNKRR), 160-164 (RTRDKK), 233-237 (RREDK) and 265-269 (KKDK). A mono-ubiquitylation site is present at K289 and regulates nuclear localization, as does a stretch of serine and threonine phosphorylation sites: S370, S380, T382, T383 and S385 (asterisks).

Perspectives
Investigations of the role of nuclear PTEN are still in their infancy; yet we have begun to reveal how subcellular localization of PTEN might modulate its activity. Accumulating evidence suggests that there are multiple mechanisms of PTEN nuclear import that depend on the cell type and cellular environment. It is clear that the role of nuclear PTEN is not the same as that of cytoplasmic PTEN. Nuclear PTEN plays a strong role in regulating cellular homeostasis and stability. It contributes to chromosome stability, DNA repair and cell cycle arrest. Loss of nuclear PTEN results in chromosomal alterations and unregulated cell cycling (Fig. 4). Tipping the localization balance from nuclear to cytoplasmic PTEN clearly affects its tumor-suppressor function, which can, at worst, lead to the development of neoplasia.
Whereas somatic PTEN mutations in tumors are relatively rare, germline PTEN mutations are found in ~85% of CS patients (Eng, 2003;Marsh et al., 1998). Although these mutations have been documented in the exonic, intronic and promoter regions of the protein (Waite and Eng, 2002), to date, no mutation has been shown to result in mis-localization of the PTEN protein among 2200 research subjects (C.E., K.A.W. and Tammy M. Sadler, unpublished). Such a mutation might be embryonic lethal given the important role of nuclear PTEN in cell proliferation; however, this remains to be seen. Interestingly, changes in nuclear PTEN protein are observed in tumors that have somatic mutations, which supports the idea that germline mutations that result in altered PTEN localization are lethal.
It is interesting to note that, in a subset of patients, no coding, promoter or deletion mutations in PTEN are apparent. Alternative mechanisms of PTEN dysregulation, such as alterations in subcellular localization, that result in clinical manifestations must therefore exist. Continued examination of samples from CS patients will undoubtedly uncover novel mutations that reveal additional mechanisms of PTEN localization. How these mechanisms interact dynamically will be essential to determine. Additional functions will certainly be uncovered and these should help us further elucidate the importance of this protein in human health and disease.
The authors would like to thank Kevin Zbuk for critical review of this manuscript. C.E. is a recipient of the Doris Duke Distinguished Clinical Scientist Award. The identification and characterization of PTEN in the authors' laboratory was partially funded by the American Cancer Society (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) and the investigation into the nuclearcytoplasmic trafficking of PTEN in the authors' laboratory was partially funded by the National Cancer Institute (1R01CA118980). Fig. 4. Nuclear and cytoplasmic PTEN signaling. PTEN localizes to both the cytoplasm and the nucleus and shuttles between each by a variety of mechanisms. PTEN function is, at least in part, determined by its subcellular localization. The 'classic' PTEN function is cytoplasmic and includes downregulation of AKT, which increases p27 levels and thereby leads to apoptosis. By contrast, nuclear PTEN has a variety of functions: downregulation of MAPK (ERK), leading to a decrease in cyclin D1 levels and G0-G1 arrest; upregulation of RAD51 levels and double-stranded-break repair; an interaction with CENP-C, which enhances centromere stability specifically and overall genomic stability; and apoptosis.