Reactive Lipid Species from Cyclooxygenase-2 Inactivate Tumor Suppressor LKB1/STK11

LKB1, a unique serine/threonine kinase tumor suppressor, modulates anabolic and catabolic homeostasis, cell proliferation, and organ polarity. Chemically reactive lipids, e.g. cyclopentenone prostaglandins, formed a covalent adduct with LKB1 in MCF-7 and RKO cells. Site-directed mutagenesis implicated Cys210 in the LKB1 activation loop as the residue modified. Notably, ERK, JNK, and AKT serine/threonine kinases with leucine or methionine, instead of cysteine, in their activation loop did not form a covalent lipid adduct. 4-Hydroxy-2-nonenal, 4-oxo-2-nonenal, and cyclopentenone prostaglandin A and J, which all contain α,β-unsaturated carbonyls, inhibited the AMP-kinase kinase activity of cellular LKB1. In turn, this attenuated signals throughout the LKB1 → AMP kinase pathway and disrupted its restraint of ribosomal S6 kinases. The electrophilic β-carbon in these lipids appears to be critical for inhibition because unreactive lipids, e.g. PGB1, PGE2, PGF2α, and TxB2, did not inhibit LKB1 activity (p > 0.05). Ectopic expression of cyclooxygenase-2 and endogenous biosynthesis of eicosanoids also inhibited LKB1 activity in MCF-7 cells. Our results suggested a molecular mechanism whereby chronic inflammation or oxidative stress may confer risk for hypertrophic or neoplastic diseases. Moreover, chemical inactivation of LKB1 may interfere with its physiological antagonism of signals from growth factors, insulin, and oncogenes.

dent of lipids inactivating IB kinase (16,17), directed our attention to LKB1, a recently discovered tumor suppressor associated with Peutz-Jeghers hamartoma syndrome (18,19). LKB1 is a novel serine/threonine kinase (STK11) at the apex of a signaling cascade that senses cellular energy homeostasis and adjusts anabolic and catabolic processes ( Fig.  1). LKB1, an AMP-kinase kinase and tumor suppressor, is a unique link between metabolic and proliferation/polarity signaling (20 -24). We report that reactive lipid species covalently modify LKB1 at a nucleophilic Cys 210 residue in its activation loop, thereby inhibiting both the phosphorylation of AMPK␣ 3 and the downstream propagation of signals through the LKB1-AMPK␣-TSC1/2-mTOR-S6K cascade. Disruption of anabolic and catabolic homeostasis and the failure to restrain inappropriate protein translation could contribute to hamartoma formation and the heightened cancer risk in Peutz-Jeghers syndrome. Chemical inactivation of tumor suppressor proteins, like LKB1 and p53, could be an etiological factor in dysplasia and hyperplasia associated with overexpression of COX-2 or chronic inflammation that can expose cells to reactive lipid species (25,26).
Cell Culture-MCF-7 breast cancer cells (ATCC) were maintained in minimal essential medium at 37°C in a humidified incubator with 5% CO 2 . The medium was supplemented with 2 mM L-glutamine, 1.5 g/liter sodium bicarbonate, 0.1 mM nonessential amino acids, 1 mM sodium * This work was supported in part by United States Public Health Services Grants R01 AI26730 and PO1 CA73992 and the Huntsman Cancer Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  pyruvate, 0.01 mg/ml bovine insulin, 0.01 mg/ml gentamicin, and 10% fetal bovine serum. RKO colon cancer cells were maintained in minimal essential medium supplemented with 1.5 g/liter sodium bicarbonate, 0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, 0.01 mg/ml gentamicin, and 10% fetal bovine serum. In certain experiments, cells were incubated in serum-depleted medium containing 1% fetal bovine serum for 6 h prior to treatments described below.
Transfection-LKB1-HA or LKB1 (C210S)-HA was transfected into RKO cells using 1 g/l DNA, 3 l Lipofectamine2000 TM for 20 h following the manufacturer's protocol. COX-2 was transfected into MCF-7 cells using 1 g/l DNA, 3 l Lipofectamine2000 TM for 24 h following the manufacturer's protocol. Transfection efficiency was measured by immunochemical determination of COX-2 protein in cell lysates. Samples were lysed and fractionated, as above, and membranes were incubated for 12 h at 4°C with primary antibodies directed against COX-2 (1:1000).
Site-directed Mutagenesis-A C210S LKB1 mutant was constructed using a QuikChange TM site-directed mutagenesis kit following the manufacturer's protocol. Residue 210 was converted from a Cys to a Ser by a TGC to a TCC substitution. The identity of the product was confirmed by DNA sequencing.
AMPK␣, ACC, and S6K Phosphorylation Assays by Western Blot-MCF-7 cells were treated with 0 -60 M of the designated PGs, Tx, 4-HNE, or 4-ONE for 4 h unless otherwise stated. Following this incubation, 2 mM AICAR (27), an AMP mimetic, was added to cells for 30 min. In certain experiments cells were treated with 60 M PG for 4 h, and 50 nM rapamycin for 2 h prior to AICAR treatment. The cells were lysed as described above; 15 g of protein was fractionated by SDS-PAGE, and proteins were transferred to PVDF membranes. The membranes were probed with primary antibodies directed against phospho-Thr 172 -AMPK␣ (1:1000), total AMPK␣ (1:1000), phospho-Ser 79 -ACC FIGURE 1. LKB1 signaling pathway. LKB1 (STK11) functions as an AMP-kinase kinase in cells. Activity of LKB1 is limited in cells with adequate ATP. Accumulation of AMP, or addition of its mimetic AICAR, activates the AMP-kinase kinase activity of LKB1, which converts AMPK␣ 3 phospho-Thr 172 AMPK␣. In turn, phospho-Thr 172 AMPK␣ converts ACC 3 phospho-Ser 79 acetyl-CoA carboxylase. Phospho-Thr 172 AMPK␣ also phosphorylates TSC1/2, which inhibits mTOR-mediated conversion of S6K 3 phospho-Thr 389 S6K. Cells sense changes in their AMP/ATP ratio and use the LKB13 AMPK␣ kinase cascade to maintain proper anabolic 7 catabolic homeostasis. LKB1 has a dual role as a tumor suppressor and metabolic regulator by antagonizing the phosphatidylinositol 3-kinase-AKT pathway, which propagates anabolic signals from insulin and proliferation signals from growth factors/oncogenes.
Prostaglandin E 2 and D 2 Formation-Enzyme immunoassays were performed to assess PGE 2 and PGD 2 metabolite formation. COX-2 was transfected into MCF-7 cells using 1 g/l DNA, 3 l of Lipofectamine 2000 TM for 24 h following the manufacturer's protocol. Cells were treated with 10 M COX-2 inhibitor NS-398 (Cayman Chemical) or vehicle control for 1 h followed by treatment with 100 M of arachidonic acid for 4 h. PGE2 Express EIA kit and PGD2-MOX Express kits (Cayman Chemical) were used following the manufacturer's protocol. Medium from cell culture was sampled and analyzed at 30 m. Experiments were repeated three times, and results depict the mean Ϯ S.E.
Statistics-Statistical significance was assessed by analysis of variance with Bonferroni's or Newman-Keul's post-hoc test for comparisons among groups.

Covalent Modification of Cellular LKB1 by Electrophilic Lipids-Cy-
clopentenone PG and 4-HNE react covalently with IKK (16,17). Thus, these lipids may modify other serine/threonine kinases with homology to the Cys 179 residue in IKK. LKB1, a unique serine/threonine kinase tumor suppressor, has a nucleophilic Cys 210 that aligns with the Cys 179 residue in the activation loop of IKK␣ and IKK␤. In contrast, ERK, JNK, p38, and AKT kinases have Met or Leu residues in the corresponding position ( Fig. 2A). To determine whether electrophilic PGs could react directly with LKB1 and other kinases, we used PGA 1 amidopentylbiotin (PGA 1 -APB). This PG analog has the characteristic ␣,␤-unsaturated ketone of PGA 1 but a C-1 biotin amide instead of a C-1 carboxyl group. Proteins that react with PGA 1 -APB will subsequently contain a biotin epitope, which binds to neutravidin beads, thereby enabling their isolation and identification (15). PGA 1 -amidopentylbiotin reacted covalently with cellular LKB1 and IKK␣ but not with ERK1/2, JNK2, AKT, or IKK␥ (Fig. 2B). These inert kinases lack the distinctive Cys residue corresponding to Cys 179 in IKK␣ or Cys 210 in LKB1. However, these kinases do have Cys residues at other positions. Thus, cellular serine/threonine kinases do not react indiscriminately with PGA 1 -ami-dopentylbiotin under these experimental conditions. We obtained similar results for ⌬12-PGJ 2 -amidopentylbiotin (data not shown).
Site-directed mutagenesis demonstrated that Cys 210 in the activation loop of LKB1 is required for lipid adduct formation. RKO cells express negligible amounts of LKB1; therefore, we transfected them with plasmids encoding LKB1 or the corresponding C210S mutant. PGA 1 -amidopentylbiotin formed a covalent adduct with wild type LKB1 protein but not with the LKB1-C210S mutant protein (Fig. 3A). Adduct formation with LKB1 was not dependent on ectopic overexpression of the LKB1 protein. PGA 1 -amidopentylbiotin formed adducts with wild type LKB1 expressed physiologically in MCF-7 cells (Fig. 3B). Additionally, adduct formation is not unique to PGA 1 as both PGA 1 -amidopentylbiotin and ⌬12-PGJ 2-amidopentylbiotin formed adducts with LKB1 in MCF7 cells (Fig. 3B).
Inhibition of the LKB1-AMPK␣-mTOR-S6 Kinase Signaling Pathway by Reactive Lipid Species-Inhibition of LKB1 by reactive lipids affected proximal and distal components in its signaling pathway downstream from AMPK␣. For instance, PGA 1 , ⌬12-PGJ 2 , and 4-HNE inhibited the phosphorylation of ACC, a substrate for phospho-Thr 172 -AMPK␣, whereas PGB 1 did not (Fig. 5A). Likewise, signaling through the S6 kinase pathway was affected. The basal level of phospho-Thr 389 -S6 kinase was readily detectable in MCF-7 cells, which must be grown in media containing 0.01 mg/ml insulin (Fig. 5, B and C, lane 1). Activation of the cellular LKB1-AMPK␣ pathway with 2 mM AICAR reduced the level of phospho-Thr 389 -S6K, the active form of S6K (Fig. 5, B and C,  lane 3). ⌬12-PGJ 2 , alone, enhanced S6K activity in MCF-7 cells (Fig. 5, B and C, lane 2 versus lane 1). ⌬12-PGJ 2 antagonized the effect of AICAR on cellular LKB1 activity and restored the content of phospho-Thr 389 -S6K to the control, tonic level (Fig. 5, B and C, lane 4 versus lane 3 and lane 1). Finally, rapamycin, which interacts directly with mTOR, overrode the inactivation of LKB1 by ⌬12-PGJ 2 and blocked phosphorylation of S6K (Fig. 5, B and C, lanes 5 and 6). Thus, by inactivating cellular LKB1 at the apex of the AMPK␣ pathway, ⌬12-PGJ 2 enhanced the distal formation of anabolically active phospho-Thr 389 -S6K, consistent with the scheme shown in Fig. 1. (25,26). This exposes cells to autocoid mediators and related, reactive lipid species. To extend our investigation, we tested whether or not transfection of MCF7 cells with a plasmid encoding COX-2 would inhibit LKB1 kinase activity by enabling biosynthesis of endogenous eicosanoids. Ectopic expression of COX-2 in the presence of its substrate, arachidonic acid, inhibited the kinase activity of LKB1. LKB1 activity, measured as phospho-Thr 172 -AMPK␣ formation, was designated 100% in mock-transfected MCF-7 cells stimulated with 2 mM AICAR (Fig. 6A). LKB1 activity remained comparable with control in MCF-7 cells transfected with COX-2 and stimulated with 2 mM AICAR. The phospho-Thr 172 -AMPK␣ content fell (p Ͻ 0.01) in MCF-7 cells transfected with COX-2 and supplemented with 100 M arachidonic acid (Fig. 6B, lane 5). In mock-transfected MCF-7 cells incubated with 100 M arachidonic acid, the phospho-Thr 172 -AMPK␣ levels were indistinguishable from the control (p Ͼ 0.05) (Fig. 6A).

DISCUSSION
Recently, LKB1 was identified as the gene responsible for Peutz-Jeghers syndrome (18,19), which predisposes to tumors of the digestive tract, reproductive organs, and breast. Cancer incidence in Peutz-Jeghers syndrome is 5-18-fold greater than in the general population (29), and it is the only familial cancer syndrome attributed to loss-of-function mutations in a serine/threonine kinase. The cellular substrates and biological roles of LKB1 were unknown until 2003-2004, when investigators discovered that it phosphorylated AMPK and functioned as a unique AMP-kinase kinase (20 -24). LKB1 has two vital cellular roles as follows: tumor suppressor and regulator of anabolic/catabolic homeostasis. LKB1 achieves its purpose by negatively regulating the phosphatidylinositol 3-kinase-AKT pathway (Fig. 1). When activated by their respective stimuli, LKB1 and AKT send opposing signals that determine the ratio of the active/inactive forms of TSC1/2, mTOR, and ribosomal S6K. This, in turn, balances anabolic and catabolic processes in order to maintain cellular energy homeostasis (ATP levels) (23). Our data show that exogenous electrophilic lipids, as well as endogenous catalysis by cellular COX-2, can inhibit LKB1 activity and shift the equilibrium in the pathway toward phosphorylation and activation of ribosomal S6K. If this should occur when cells have too little ATP to support proper translation of RNA into proteins, it might facilitate tumor progression. Dysregulation of protein translation is important in several cancers (30 -32). Low expression of LKB1 in breast tumors is associated with poor prognosis and survival (33) and with the transition from pre-malignant to malignant tumor growth in lung cancer (34). Most interestingly, several studies show that hamartomas overexpress COX-2 (35,36), and its expression facilitates tumorigenesis via a hamartoma-adenoma-carcinoma sequence (37,38). Although speculative at this point, our results suggest that inhibition of LKB1 because of its chemical inactivation by reactive lipid species or overexpression of COX-2, such as in Fig. 6, may have consequences similar to loss of expression of LKB1 or mutational inactivation.
LKB1 is a tumor suppressor that often retains one wild type allele, deviating from Knudson's two-hit hypothesis. This suggests that LKB1 is a potential candidate for inactivating processes that are distinct from genetic or epigenetic lesions. Furthermore, we reasoned that LKB1 would be covalently modified and inactivated by cyclopentenone PG  activity. Cells expressing COX-2 and treated with 100 M arachidonic acid showed a decrease in LKB1 activity as measured by phospho-AMPK␣ content relative to control. Cells expressing COX-2 and exposed to 100 M (R)-ibuprofen showed a significant recovery in LKB1 activity as measured by phospho-AMPK␣ compared with cells not treated with ibuprofen. B, representative immunoblot showing inhibition of the phosphorylation of AMPK␣ because of COX-2 transfection and arachidonic acid treatment and recovery of the phosphorylation of AMPK␣ because of preincubation with ibuprofen. The intensity of the AMPK␣ and phospho-Thr 172 AMPK␣ bands was measured and depicted as a percentage of the control (i.e. 100% ϭ AICAR alone, lane 1). Values represent mean Ϯ S.E., n ϭ 3.

TABLE 1 Prostaglandin formation by MCF-7 cells
Values are mean Ϯ S.E., n ϭ 3 experiments. MCF-7 cells were transfected with a COX-2 or a mock construct. 24 hours after transfection, cells were incubated for 1 h with culture medium containing vehicle or 10 M NS-398. Cells were then incubated with 100 M arachidonic acid (AA) for 30 min, and levels of PGE 2 and PGD 2 in the medium were quantified by immunoassay. and 4-HNE as follows. PGA 1 and 15-deoxy-PGJ 2 inhibit some serine/ threonine kinases but not others. For example, they inhibit IKK but not ERK1, ERK2, p38, or JNK (16,39). Based on site-directed mutagenesis and pharmacological experiments with inert and reactive eicosanoids, Rossi et al. (16) proposed a Michael addition at the Cys 179 residue in the activation loop as the molecular basis for inhibition of IKK␣ and IKK␤. Because LKB1 is a rare example of a tumor suppressor, which functions as a serine/threonine kinase (18,19), we compared the amino acid sequence of LKB1 flanking its activation loop with those for IKK, ERK, JNK, AKT, and p38. This sequence alignment revealed a Cys 210 residue in LKB1 that corresponded to the Cys 179 residue in IKK. These Cys residues are distinct to LKB1, IKK␣, and IKK␤ but not ERK, JNK, and p38 kinases, which have Met or Leu in comparable positions (Fig. 2).
Our results show that cyclopentenone PGs of the A-and J-series covalently modify LKB1 and IKK␣ but not ERK1/2, JNK2, IKK␥, AKT, or an LKB1 mutant with Cys 210 replaced by Ser 210 . Thus, these cellular serine/threonine kinases do not react indiscriminately with PGA 1 -amidopentylbiotin under our experimental conditions. The results in Figs. 2 and 3 are compatible with the following: (i) the pharmacological data mentioned above (16,39); (ii) the known chemical reactivity between thiols (ϪSH) and ␣,␤-unsaturated carbonyls (Michael reaction); and (iii) the influence of other amino acids on steric accessibility of the thiol and the chemical reactivity of cysteine. Interactions between cysteine and other amino acids determine the pK a and nucleophilic character of its thiol (RSH % RS Ϫ ). Virtually all lipids bind to cellular proteins by reversible, low affinity interactions (K d ϳ 10 Ϫ6 liter/mol). For some lipids with an ␣,␤-unsaturated carbonyl, this low affinity interaction with proteins seems to position their electrophilic ␤-carbon near a nucleophilic substituent that is favorably disposed for Michael addition, i.e. Cys 179 in IKK (16), Cys 210 in LKB1, Cys 38 in the p65 subunit of NFB (40), Cys 62 in the p50 unit of NFB (41), Cys 29 in cathepsin (42), and Cys 35 in thioredoxin (43), and results in a covalent lipid adduct. Thus, pK a and steric accessibility of cysteinyl thiols strongly influence, and limit, which particular cellular proteins can form Michael adducts with A-or J-series PGs (15,16,40,41,43,44) or 4-HNE (15,17,42,45). The chemistry of nucleophilic Michael addition reactions argues against any individual ␣,␤-unsaturated carbonyl acting as a regulatory molecule via specific, covalent reaction with any individual protein receptor. If these reactive lipids modulate biological processes, they probably do so with less specificity than classical autocoids. For example, multiple lipids with ␣,␤-unsaturated ketones can react with a particular protein, e.g. thioredoxin reductase (15), IKK (16), and LKB1. Likewise, a particular electrophilic lipid, e.g. 4-HNE, can react with more than one protein (15,17,42,45). During oxidative stress, inflammation, or overexpression of COX-2, the biological milieu may contain multiple electrophilic species. Consequently, each species with an ␣,␤-unsaturated ketone will contribute, stoichiometrically, to any pathophysiological effect. 4-HNE, cyclopentenone PGs, and 8-iso-cyclopentenone PGs, have been detected in vivo within cellular proteins at sites of inflammation (46 -49). Investigators have even hypothesized that these reactive lipid species help resolve inflammation by covalently modifying IB kinases and repressing NFB (16,50). Such an unconventional mechanism of action for resolving inflammation may confer inevitable risks, such as the impairment of other proteins and tumors suppressors. Disruption of tumor suppressors by inflammatory mediators seems especially relevant to cancers associated with chronic inflammation or overexpression of cyclooxygenase and lipooxygenase (25,26).