Alkylation of the Tumor Suppressor PTEN Activates Akt and β-Catenin Signaling: A Mechanism Linking Inflammation and Oxidative Stress with Cancer

PTEN, a phosphoinositide-3-phosphatase, serves dual roles as a tumor suppressor and regulator of cellular anabolic/catabolic metabolism. Adaptation of a redox-sensitive cysteinyl thiol in PTEN for signal transduction by hydrogen peroxide may have superimposed a vulnerability to other mediators of oxidative stress and inflammation, especially reactive carbonyl species, which are commonly occurring by-products of arachidonic acid peroxidation. Using MCF7 and HEK-293 cells, we report that several reactive aldehydes and ketones, e.g. electrophilic α,β-enals (acrolein, 4-hydroxy-2-nonenal) and α,β-enones (prostaglandin A2, Δ12-prostaglandin J2 and 15-deoxy-Δ-12,14-prostaglandin J2) covalently modify and inactivate cellular PTEN, with ensuing activation of PKB/Akt kinase; phosphorylation of Akt substrates; increased cell proliferation; and increased nuclear β-catenin signaling. Alkylation of PTEN by α,β-enals/enones and interference with its restraint of cellular PKB/Akt signaling may accentuate hyperplastic and neoplastic disorders associated with chronic inflammation, oxidative stress, or aging.


Introduction
Inflammation and cancer are intricately linked [1,2]. 'Smoldering' inflammation [3], also called para-inflammation [4], occurs in many types of pre-malignant and malignant tumors, e.g. colorectal adenoma and adenocarcinoma where the content of inflammatory leukocytes and the inflammatory enzyme cyclooxygenase-2 (COX-2) influence progression, prognosis and survival [5,6]. Nonsteroidal anti-inflammatory drugs (NSAIDs) that inhibit COX-2 can prevent certain, but not all, cancers [7]; and some NSAIDs, such as sulindac sulfone, act independently of COX and prostaglandin E 2 (PGE 2 ) inhibition [8]. Other NSAIDs, e.g. celecoxib, can paradoxically enhance tumor progression in APC Min/+ mice, which model intestinal tumorigenesis [9]. While COX-2 and its metabolite PGE 2 are undoubtedly important, para-inflammation may enhance tumorigenesis by mechanisms that are incompletely understood. Innate immune mechanisms are prime candidates for investigation.
PTEN (phosphatase tensin homolog on chromosome 10) is a phosphoinositide-3-phosphatase with two physiological roles: tumor suppressor and regulator of anabolic/catabolic cell signaling. The PTEN gene is frequently mutated or inactivated in advanced cancers [26]. Using MCF7 and HEK-293 cells, we report that reactive a, ß-unsaturated carbonyls (acrolein, 4-HNE, and D12-PGJ 2 ) inactivate the PTEN protein -not the gene -by alkylation. Inactivation of PTEN by a, ß-unsaturated carbonyls leads to increased Akt signaling, enhanced nuclear b-catenin signaling, and augmented cellular proliferation. Redox signaling by PTEN may have evolved to enable cells (tissues) to stratify their response to oxidative stress. For example, transient inhibition of PTEN by reactive oxygen or carbonyl species, and the corresponding signaling through Akt/GSK3b/b-catenin/TCF4/ Lef1 might benefit the host via increasing proliferation and regeneration of tissue damaged by acute inflammation or oxidative stress. Errant and persistent PTEN inactivation by the same molecular mechanism might favor tumor progression and provide an etiological link between 'smouldering' inflammation and certain cancers, especially colorectal cancer, where both the PTEN and the APC tumor suppressors restrict nuclear b-catenin signaling [27].

Results
The a, ß-unsaturated carbonyls acrolein, 4-HNE and D12-PGJ 2 covalently modify cellular PTEN We exposed MCF-7 cells to representative a, ß-unsaturated carbonyl ( Figure 1C) or H 2 O 2 , then selectively tagged any proteins that had oxidized or carbonylated thiols using NEM-biotin ( Figure 1A). We then sequestered proteins with a biotin epitope onto NeutrAvidin (NA) beads and identified carbonylated PTEN by SDS-PAGE and immunoblotting. MCF-7 cells treated with 10 mM D12-PGJ 2 , 4-HNE, or acrolein contained carbonylated PTEN in amounts comparable to cells treated with 100 mM H 2 O 2 ( Figure 1A, NA pulldown). This method does not distinguish between oxidized and carbonylated thiols on PTEN. However, electrophoresis under non-reducing conditions, followed by western blotting, showed that PTEN migrated as a discrete isoform due to an oxidized disulfide [28,29], which occurred only in cells treated with 100 mM H 2 O 2 , but not in cells treated with a, ß-unsaturated carbonyl (D12-PGJ 2 , 4-HNE, acrolein) or 15-HpETE, a lipid hydroperoxide ( Figure 1B).
Cyclopenteneone PG-biotin analogs are model a, b-enones that alkylate PTEN The cysteinyl thiolate in the PTEN active site (-HC(X 5 )RT-) is prone to oxidation because it is a strong nucleophile, pKa ,5. This trait should also facilitate alkylation of PTEN by a, ßunsaturated carbonyls. We used CyPG-biotin analogs, which have an electrophilic b-carbon capable of nucleophilic addition (Michael reaction), as chemical models to test this hypothesis [30]. Alkylation of any cellular proteins by these analogs would introduce a biotin epitope de novo (Figure 2A). PGA 1 -biotin and D12 PGJ 2 -biotin were both taken up into MCF-7 cells and formed covalent adducts with ,20 proteins ( Figure 2B). D12 PGJ 2 -biotin (a bi-functional dienone) was more reactive than PGA 1 -biotin (mono-functional enone), agreeing with others who reported ,20-30 protein targets modified by cyPG-biotin in 3T3 cells or mitochondria [31,32]. Sequestration of de novo biotinylated cellular proteins on NA beads, followed by immunoblot with anti-PTEN antibodies, showed that PTEN formed a covalent adduct ,10-fold more readily with D12-PGJ 2 -biotin than with PGA 1 biotin ( Figure 2C The a, ß-enone, D12-PGJ 2 , interferes with PTEN suppression of Akt kinase Growth factors, insulin, and other stimuli prompt PI3-K to make PIP 3 , which recruits PKB/Akt kinase to the cell membrane where PDK1/2 phosphorylates Akt Thr 308 and Akt Ser 473 residues, respectively [33,34,35]. PTEN down-regulates PKB/Akt activation by metabolizing PIP 3 to PIP 2 [36]. a, ß-unsaturated carbonyls that alkylate cellular PTEN may interfere with its suppression of Akt kinase. A representative a, ß-enone, D12-PGJ 2 , caused a concentration and time dependent increase in phospho-(T 308 ) Akt in MCF-7 cells. As little as ,2 mM D12-PGJ 2 caused a half-maximal response ( Figure 3A). Increases in cellular phospho-(T 308 )Akt were detectable at 10 min, maximal at 30 min, and durable for .120 min ( Figure 3B). D12-PGJ 2 increased formation of phospho-(T 308 )Akt without altering formation of phospho-(S 241 )PDK1 (the kinase that phosphorylates T 308 of Akt), and without altering PTEN protein  Figure 3C). These data suggest that D12-PGJ 2 interfered with PTEN's capacity to restrain activation of Akt kinase. Consistent with this interpretation, co-treatment of cells with cyPGs plus 50 mM LY294002, lowered levels of phospho-(T 308 )Akt by inhibiting PI3-K, at the apex of the PI3-K/RPDK1/2RAkt kinase cascade ( Figure 3C, lanes 3 vs 2, and 6 vs. 5). Furthermore, 1 mM of various PGA and PGJ isomers, and other a, ß-enones directly inhibited isolated PTEN enzyme [ Table 1]. The electrophilic b carbon of cyPGs is essential for inhibition, since a structurally similar, but non-electrophilic cyPG, PGB 1 , was inactive.
a, ß-unsaturated carbonyls cause a time-dependent accumulation of cellular phospho-(S 473 )Akt kinase (active) R phospho-(S 9 )GSK3b (inactive) R b-catenin and a rise in nuclear b-catenin signaling GSK3b converts b-catenin to phospho-(S 33/37 /T 41 ) b-catenin, which is rapidly eliminated by the 26S proteasome [40]. GSK3b acts in concert with the tumor suppressor APC. In cells with mutant APC, or when WNT ligands stimulate cells with wild type APC, GSK3b fails to phosphorylate b-catenin, which allows it to accumulate, associate with other nuclear transcription factors and express its target genes (e.g. c-myc, cyclin D1) [41]. PTEN can block b-catenin accumulation/signaling by favoring retention of active GSK3b and inactive PKB/Akt kinase in some [42,43], but not all experimental systems [44,45]. Accordingly, inactivated PTEN should augment b-catenin signaling by favoring retention of inactive phospho-(S 9 )GSK3b and active phospho-(S 473 ) Akt kinase. We thus hypothesized that these electrophilic mediators may affect ß-Catenin signaling through this mechanism. To investigate this further, we used HEK 293 cells which have an intact ß-Catenin signaling pathway. We found that the different a, ß-unsaturated carbonyls that alkylated PTEN (6 mM D12 PGJ 2 , 6 mM 4-HNE and 20 mM acrolein) all caused a time-dependent rise in phospho-(S 473 )Akt (i.e. active Akt kinase), with a corresponding rise in phospho-(S 9 )GSK3b (i.e. inactive GSK3b) and b-catenin ( Figure 4C). To determine if a, b-unsaturated carbonyls enhanced nuclear b-catenin signaling, we used HEK 293 cells engineered to stably express WNT3A and Super-TopFlash (STF) reporter gene [46]. These cells, called STF3A cells, thus secret WNT3A, an autocrine/paracrine stimulus for FZD receptors that slows APC-dependent degradation of bcatenin ( Figure 5A). Nuclear b-catenin signaling in STF3A cells is proportional to their luciferase expression (activity), and they are responsive to DKK1, a WNT antagonist that inhibited b-catenin signaling in STF3A cells in a concentration-dependent manner ( Figure 5B).
Several reative carbonyl metabolites, each with an electrophilic a,b enone or enal substituent, enhanced expression of the bcatenin:luciferase reporter gene in STF3A cells ( Figure 5C). Luciferase reporter activity rose by ,4-fold over baseline (p,0.01) in STF3A cells incubated with D12 PGJ 2 or 4-HNE; by ,3-fold (p,0.01) in cells with other PGJ analogs, acrolein or 4-ONE; and by ,1.5-fold (p,0.05) in cells with PGA 2 . Consistent with our mechanistic hypothesis, neither PGB 2 nor MDA had a detectable effect. PGB 2 is a cyPG, but tautomerism prevents the charge delocalization required to create an electrophilic b carbon, which is required for protein alkylation. MDA (b-hydroxy-acrolein) penetrates cell membranes poorly because it is .99% ionized at physiological pH ,7.4 used in our experiments. Neither PGE 2 nor other primary PG metabolites of COX-1 or -2 had any effect on nuclear b-catenin signaling in STF3A cells. We draw attention to the fact that ectopic over-expression of EP receptors in HEK 293 cells was required to elicit any PGE 2 mediated b-catenin signaling [47]. The weak response to PGE 2 and other PG's in Figure 5C may reflect the constitutive levels of EP, FP, DP or IP receptors in STF3A cells or rapid metabolism of PGs, or both.
Enhanced b-catenin signaling in STF3A cells was concentration dependent between 2-20 mM for acrolein, 4-HNE and D12 PGJ 2 ( Figure 6A). Depletion of cellular GSH to ,10% of baseline by treatment with 100 mM BSO potentiated b-catenin signaling, e.g. in STF3A cells treated with 2 and 6 mM D12 PGJ 2 ( Figure 6B). This is consistent with the role of reduced glutathione in the conjugation of reactive metabolites, and protection of redox sensitive proteins from alkylation [20].

Discussion
The PTEN tumor suppressor gene is frequently mutated or inactivated in advanced cancers [26,48]. PTEN is a phosphoinositide-3-phosphatase that metabolizes PIP 3 to PIP 2 [36], thereby counter-regulating PKB/Akt, a serine/threonine kinase protooncogene that controls anabolic growth and specification of cell fate [33,34,35]. PTEN, itself, is regulated post-translationally by phosphorylation [49], acetylation [50], and reversible oxidation of its catalytic cysteine 124 residue [28,29]. Oxidation of cellular PTEN can involve H 2 O 2 derived from NADPH oxidase [51], superoxide dismutase [52], or enzymatic peroxidation of arachidonic acid (AA) by COX-1, COX-2 or 5-LOX [53]. All of these enzymes are commonly over-expressed and activated by inflammation or neoplastic transformation. PTEN oxidation, and any attendant pathophysiology, varies with the degree of cellular exposure to reactive oxygen species (ROS). In these studies, we demonstrate that the chemistry which facilitates oxidation of PTEN can also facilitate its alkylation by electrophilic a, ß-enals and a, b-enones [19,20].
PTEN epitomizes the adaptation of redox-responsive thiols for cell signaling, as well as their potential vulnerability to by-products of oxidative stress and inflammation. PTEN is inactivated by two distinctive redox-mediated processes: 1) intra-or inter-molecular disulfide formation by ROS and 2) thiolate carbonylation (Michael addition) by electrophilic a, ß-unsaturated carbonyls (Figure 1-3). Hydrogen peroxide (H 2 O 2 ), a prototypical ROS, inhibits cellular PTEN by directly oxidizing its catalytic Cys 124 to a sulfenic acid intermediate, which then forms an inactive, intra-molecular Cys 124-71 disulfide [28,29]. Our data show that several representative, electrophilic carbonyl species (a,ß-enals and a,ß-enones), which can occur endogenously as byproducts of lipid peroxidation during inflammation or oxidative stress, alkylate and inactivate PTEN. Inactivation of PTEN by redox-mediated processes causes an increase in activity of the proto-oncogene Akt. Hyperactivation of Akt increases proliferation and survival of many different cancers.
Signaling by H 2 O 2 spans a wide pathophysiological continuum [54] and a comparable role for reactive electrophiles seems plausible. Reactive carbonyl species such as acrolein, 4-HNE and D12PGJ 2 represent a sub-set of electrophiles commonly produced during oxidative stress and inflammation. These findings might extrapolate to electrophilic agents lacking a carbonyl but containing other electron withdrawing groups, and we refer to them generally as ''reactive electrophiles''. First, like H 2 O 2 , reactive electrophiles occur in vivo during inflammation and oxidative stress [16,18,19,22]. Second, reactive electrophiles covalently modulate other proteins that regulate important signaling processes; i.e. LKB1/STK11 [55], NFkB [56], and IKKb. Third, H 2 O 2 and reactive electrophiles both originate from a combination of spontaneous and enzymatic processes, which often coincide in inflamed tissues [57]. H 2 O 2 derives from superoxide anion, O 2 2 , the primary metabolite of NADPH oxidases. Spontaneous and enzymatic dismutation converts O 2 2 into H 2 O 2 . Likewise, spontaneous and enzymatic lipid peroxidation generates acrolein and 4-HNE [18,57]. cyPGs originate from the lipid endoperoxide PGH 2 , the primary metabolite of COX-1 and -2. Enzymatic and spontaneous scission of endoperoxide bonds converts PGH 2 into PGE 2 and PGD 2; albumin/serum then causes their dehydration into PGA 2 , PGJ 2 , and their isomers [14,16,24].
While speculative, it appears that ROS and reactive electrophiles (H 2 O 2 , acrolein, 4-HNE, D12 PGJ 2 ) may have both evolved to play disparate roles in innate immunity: 1) annihilating pathogens and 2) resolving inflammation. Analogous to inactivation of NFkB and IKKab, temporary inactivation of the PTEN tumor suppressor protein by its alkylation, and attendant activation of PKB/Akt kinase proto-oncogenes, might help normalize morphology and histology at acutely inflamed tissues by releasing their restriction on cell proliferation, anabolic growth and fate specification [34,35]. In ordinary situations repair and resolution should help terminate innate immune inflammation (Figure 7,a). However, this mechanism might also confer inescapable risks if PTEN were inactivated errantly or persistently. Furthermore, reactive electrophiles also inactivate other notable tumor suppressors, including p53 [30] and LKB1/STK11 [55]. This combined and sustained inactivation of tumor suppressors could contribute significantly to inflammation-associated tumorigenesis and subsequently prolong the cycle of tumor-associated para-inflammation (Figure 7,b). Overall, our data and model align with the observation that tumors are wounds that fail to heal [58]. In this situation, tumor progression may derive partly from mal-adaptation of a molecular mechanism that evolved to terminate and resolve innate immune inflammation.
Inflammation is a critical component of tumor progression. Many cancers arise from sites of infection, chronic irritation and inflammation. The tumor microenvironment, which is comprised largely of inflammatory cells, plays a major role in the neoplastic process, fostering proliferation, survival, and migration. We show herein that reactive carbonyl species that are commonly produced during inflammation covalently modify and inactivate PTEN tumor suppressor. Importantly, the mechanism we describe might also extrapolate to: 1) other electrophilic species generated by inflammation, oxidative or xenobiotic stress (i.e. other a, ß unsaturated aldehydes and ketones; allylic or vinyl epoxides; quinones, chlorhydrins, chloramines, vinyl sulfones; and 2) other members of the PTP superfamily that are redox sensitive. These studies extend our understanding of the mechanisms by which inflammation contributes to the initiation and progression of cancer.  This treatment selectively alkylates all reduced thiols in PTEN, but not oxidized thiols or thiols modified by Michael addition with reactive electrophiles. Samples were washed in 1 ml of O 2 -free extraction buffer then transferred to a 15-ml conical tube. After adding SDS to a final concentration of 1% v/v, the mixture was held 2 h at 25uC in the dark, and proteins were precipitated with TCA [trichloroacetic acid], 10% v/v for 1 h. The precipitate was washed twice with acetone to remove traces of TCA, NEM, and IAA. Precipitated proteins were solubilized and oxidized or modified cys residues were reduced in 0.1 ml of O 2 -free reducing buffer (50 mM Hepes-NaOH, pH 7.7, 1 mM EDTA, 2% SDS and 4 mM DTT) for 30 min at 50uC. Reduced proteins were subsequently biotinylated with 0.9 ml of a solution containing 50 mM NaHPO 4 pH 7.0, 1 mM EDTA, and 1 mM biotin conjugated to polyethylene oxide-maleimide for 30 min at 50uC. Proteins were precipitated in 10% v/v TCA for 1 h. The precipitate was isolated by centrifugation, washed with dry icechilled acetone, and solubilized in 0.3 ml of 50 mM Hepes-NaOH, pH 7.7, 1 mM EDTA, and 2% SDS. The sample was then diluted with 0.3 ml of the same solution without SDS. 15 mg protein was assayed by immunoblot for PTEN. A separate sample (200 mg protein) was added to 100 mL immobilized NA beads in 1 ml PBS, 0.4% v/v Tween 20. This suspension was rotated 16 hr at 4u, centrifuged, and beads were washed twice with PBS/0.4% v/v Tween 20. Loading buffer (50 mL) with 5% BME was added directly to beads, boiled for 10 min to release maleimidobiotinylated proteins, and 20 mL was assayed by immunoblot for total and oxidized PTEN.  with 100 mM H 2 O 2 were lysed and proteins (15 mg) in the lysates were fractionated by non-reducing SDS-10% PAGE, followed by anti-PTEN immunoblot to distinguish native PTEN from oxidized PTEN occurring as an intra-molecular Cys 124 -Cys 71 disulfide [28].

Identification of PTEN occurring as an intra-molecular disulfide
Identification of PTEN covalently modified by cyclopentenone PG-biotin analogs MCF-7 cells (MEM 1% v/v FBS) treated 1 hr with 1-10 mM of the aminopentylbiotinamide analogs of PGA 1 or D12 PGJ 2 , were lysed, sonicated 106for 1 s at 4uC, then centrifuged 10,0006g for 10 min. Supernatant with 100 mg of protein was incubated with 100 ml of NA beads in 1 ml PBS with 0.4% Tween 20 for 16 h at 4uC to sequester proteins containing a biotin epitope introduced de novo by reaction with cyPG-biotin analogs. The beads were then centrifuged at 5006 g for 5 min to isolate neutravidin-biotin complexes (NA pulldown). The beads were washed 36 with 1 ml of PBS/0.4%Tween 20 then boiled 5 min in Laemmli loading buffer with 5% BME to release bound proteins. These samples were analyzed by immunoblotting for PTEN or proteins with a biotin epitope.
Wnt/b-catenin signaling in STF3A cells STF cells are HEK-293 cells containing a stably integrated SuperTopFlash (STF) transgene with TCF binding sites upstream of a luciferase reporter gene [59]. STF cells have negligible bcatenin/TCF-LEF transactivation and luciferase expression unless they are exposed to a Wnt ligand, e.g. WNT3A. We derived a subsidiary cell line, designated STF3A, by transfecting parental STF cells with a linearized pPGK + Wnt3A plasmid and a linearized blasticidin resistance plasmid for cell selection. STF cells grown on 10-cm plates were transfected using LipofectAMINE 2000 (Invitrogen). At 24 h after transfection, cells were serially diluted, re-plated and grown in medium with blasticidin (10 mg/ ml) for 14 d. Forty colonies were screened for b-catenin/TCF-LEF activity by measuring luciferase activity normalized to total protein concentration. The selected STF3A clones, which stably expressed and secreted WNT3A, were maintained at 37uC in DMEM with 10% FBS, penicillin/streptomycin and 10 mg/ml blasticidin in a humidified incubator with 5% CO 2 . We measured wnt/b-catenin signaling in STF3A cells by quantifying their Super TopflashH luciferase reporter signal. STF3A cells (20,000/well) were plated in white, clear bottom 96-well plates coated with poly-L-lysine and grown 24 h at 37uC. Medium was removed and replaced with 200 ml fresh medium containing 0-300 ng/ml DKK1; 2-20 mM of a,b-enone containing cyPGs; 2-20 mM of a,b-enal metabolites derived from lipid peroxidation (acrolein, MDA, crotonaldehyde, 4-HNE, 4-ONE); 20 mM of primary PGs, PGH 2 , PGE 2 , PGF 2 a, PGD 2 , PGI 2 ; 20 mM BSO, a glutathione synthesis inhibitor; other inhibitors, or DMSO vehicle. STF3A cells were incubated for 24 h at 37uC, washed with 200 ml PBS at 4uC, and lysed with 20 ml lysis buffer. Luciferin (60 ml/well) was added and luciferase activity in the lysates was quantified by fluorimetry. LDH activity in the lysate, an index of cell count, was quantified by spectroscopy. The ratio of luciferase/LDH activity is proportional to nuclear bcatenin signaling in STF3A cells.

Inhibition of PTEN Enzymatic Activity
Inhibition of PTEN with 0-30 mM reactive electrophiles was quantified by using a PTEN enzyme assay kit (# 17-351, Upstate Biotechnology).

Statistical analysis
Statistical significance at p,0.05 was assessed by analysis of variance (ANOVA) with Bonferroni's post-hoc test for comparisons among groups.