Down-regulation of adenosine monophosphate–activated protein kinase activity: A driver of cancer

Abstract

Adenosine monophosphate–activated protein kinase (AMPK), a serine/threonine protein kinase, is known as “intracellular energy sensor and regulator.” AMPK regulates multiple cellular processes including protein and lipid synthesis, cell proliferation, invasion, migration, and apoptosis. Moreover, AMPK plays a key role in the regulation of “Warburg effect” in cancer cells. AMPK activity is down-regulated in most tumor tissues compared with the corresponding adjacent paracancerous or normal tissues, indicating that the decline in AMPK activity is closely associated with the development and progression of cancer. Therefore, understanding the mechanism of AMPK deactivation during cancer progression is of pivotal importance as it may identify AMPK as a valid therapeutic target for cancer treatment. Here, we review the mechanisms by which AMPK is down-regulated in cancer.

Introduction

Cancer is one of the major diseases that causes human death and cancer-related mortality increases globally year by year. Cancer is a disease resulted from multiple genetic and epigenetic alterations that hijack normal cell physiology. The rapid growth of tumor cells depends on a sustained energy supply mainly through “Warburg effect.” The precise molecular mechanism of the deregulation of energy metabolism in cancer cells remains largely unknown; however, there is mounting evidence showing that deregulation of adenosine monophosphate–activated protein kinase (AMPK) contributes to the adaptation of cancer cells to the ever-changing environment. AMPK is a serine/threonine protein kinase and is known as “intracellular energy sensor and regulator.” AMPK plays a critical role in the regulation of cellular energy metabolism via various mechanisms. AMPK regulates the Warburg effect, protein, and lipid synthesis. In addition, through cooperation with the PI3K-Akt and mitogen-activated protein kinase (MAPK) signaling pathways, as well as interaction with a variety of transcription factors, AMPK controls a series of cellular processes including cell proliferation, invasion, migration, and apoptosis. It is well established that AMPK is a tumor-suppressor and inhibits tumorigenesis. In a variety of tumors, AMPK phosphorylation (p-AMPK) in cancer tissues with early stage (T1/T2) is significantly higher than that in advanced tumors (T3/T4),^1,2 implying that decreased activity of AMPK may be closely associated with tumor development. For example, in gastric and lung carcinomas, p-AMPK expression is significantly decreased and closely associated with cancer progression. ³ It has been widely accepted that down-regulation of AMPK activity was related to tumorigenesis. Multiple mechanisms contribute to the decline of AMPK activity during tumorigenesis, including self-inhibition, post-transcriptional, and post-translational modification. This article will review the current known mechanisms of the down-regulation of AMPK expression, protein stability, and activity.

Basic structure and function of AMPK

AMPK is a heterotrimeric complex composed of a catalytic subunit of α (α1 or α2), and regulatory subunits β (β1 or β2) and γ (γ1, γ2, or γ 3) ⁴ (Figure 1). Each subunit of AMPK has a different biological function. The N-terminal of α subunit contains the typical serine/threonine kinase activity center and thus is crucial for the catalytic activity of AMPK. The kinase activity is stimulated by Thr172 phosphorylation. While the C-terminal of α subunit interacts with β and γ subunits, in which β subunits serve as a bridge connecting α and γ via its C-terminus. γ subunit contains β subunit binding site, as well as AMP and ATP-binding sites,^5,6 thereby serving as a sensor of intracellular energy change.

Figure 1.

Upstream regulators and downstream effectors of AMPK.

As an “intracellular energy sensor,” AMPK maintains energy balance within a cell by adjusting the anabolic and catabolic pathways. In response to energy crisis, such as ischemia, hypoxia, starvation, and exercise, ATP level dropping causes an increase in the intracellular-free AMP or AMP/ATP ratio. When free AMP binds to the γ subunit, it will lead to a conformational change in the AMPK holoenzyme where the α subunit active site becomes more exposed. The exposed α subunit then undergoes autophosphorylation or phosphorylation by upstream kinases leading to the activation of AMPK. Full activation of AMPK depends on the phosphorylation of threonine at position 172 (Thr172) by upstream kinases. The currently identified upstream AMPK kinases include liver kinase B1 (LKB1), calmodulin-dependent kinase kinase (CaMKKβ), transforming growth factor-β-activating kinase 1 (TAK-1), and Ataxia telangiectasia mutated (ATM) (Figure 1).

AMPK is a tumor suppressor and plays an important role in the prevention of tumorigenesis. AMPK prevents the rapid growth of tumor cells through the inhibition of Warburg effect and by regulating the activity of acetyl-coenzyme A carboxylase (ACC), mTOR, and HMG-CoA, which in turn play a crucial role in modulating proteins and lipids synthesis.^7–10 AMPK also inhibit cell proliferation and induce apoptosis and autophagy via regulating the expression of many transcription factors such as p53, p27, and Foxo3^11,12 (Figure 1). Moreover, AMPK suppresses tumor cell invasion and migration by the negative regulation of TGF-β pathway and epithelial–mesenchymal transition (EMT).^13,14 In addition, AMPK controls the survival of CD8⁺ T cells by regulating protein phosphatase activity, thereby promoting their role in immune surveillance and anti-tumor activity. ¹⁵

AMPK inactivation by transcriptional and post-translational modification self-inhibition

Like other protein kinases, AMPK can automatically adjust its kinase activity through structural elements. Crute et al. ¹⁶ demonstrated that 313–392 fragment in the mammalian AMPK α1 catalytic subunit has almost no activity per se. However, deletion of this fragment was found to increase AMPK catalytic activity by 30-fold without any additional major effects on enzyme expression. Thus, the α catalytic subunit 313–392 domain is known to be an AMPK α subunit inhibitory region. The 313–392 sequence may inhibit AMPK activity via several different ways. First, it inhibits the binding of the C-terminus of α-subunit to the β and γ subunits, thereby reducing the exposure of the catalytic site in the N-terminal domain; secondly, in the absence of the β and γ subunits, it prevents the exposure of the α subunit activation loop or the T172 phosphorylation site or both. Another study suggests that the 313–392 sequence may inhibit AMPK function by affecting the accessibility of substrates to the catalytic cleft of AMPK kinase. ¹⁷

Through structural modeling following truncation, deletion, and site-directed mutagenesis of the human AMPK α1 catalytic subunit, Pang et al. ¹⁸ found that the fragment of 313–335 residues with the α-helix plays an important role in the inhibition of α1 catalytic subunit activity. Further study found that leucine 328 is the key residue for α1 subunit activity. Leucine residue 328 was not associated with the degradation of full-length α1 subunit, but there is a hydrophobic interaction between leucine 328 and valine 298 at the junction of the self-inhibiting and kinase catalytic domains, thereby stabilizing the self-inhibition conformation. However, deletion of 313–335 residues does not fully activate the kinase domain of α1 subunit. One possibility would be that the interaction of α1 subunit C-terminal 394–550 sequence with the kinase domain results in α1–(1–550, Δ313–335) conformational change and partial activation of α1 subunit. In addition, it was found that self-inhibiting domain might form an inactive conformation with the kinase catalytic domain, thereby interfering with substrate binding and the catalytic activity of α subunit. Moreover, binding of α1–(313–473) sequence with the regulatory AMPK β/γ subunits results in dissociation and maximum activity of full-length α catalytic subunit. Based on the inhibitory effects of 313–335 residues on AMPK activity, Pang et al. ¹⁹ have developed a new small molecule of AMPK activator PT1, which could directly activate AMPK and increase the phosphorylation of AMPK and acetyl-CoA carboxylase (ACC), providing a new opportunity to treat metabolic syndrome and cancer. PT1 can directly reduce self-inhibition and increase AMPK activity by interacting with the Glu-96 and Lys-156 residues close to the inhibitory region of the α1 subunit. In addition, in vitro experiments demonstrated that binding of PT1 with AMPK (α/β/γ trimer) further opened the AMPK catalytic subunit, resulting in increased AMPK kinase activity. Further structural modification identified a direct AMPK activator C24, which could activates AMPK not only in vitro but also in a model of diet-induced obese mice. ²⁰

Post-transcriptional regulation of AMPK by microRNAs

MicroRNAs (miRNAs) are endogenous non-coding RNAs with regulatory functions. MiRNAs are able to recognize and target messenger RNAs (mRNAs) of various genes causing mRNA degradation or preventing their translation, thus reducing the gene expression of the target protein. Mounting evidence shows that various miRNAs can reduce AMPK activity by direct and indirect targeting AMPK mRNA (Figure 2). Currently, most available studies focused on the regulation of AMPK α subunit by miRNAs. Studies have shown that miR-148b, ²¹ miR-301a, ²² and miR-144 ²³ can directly target AMPK α1. Zhao et al. ²¹ found that AMPK α1 was overexpressed and correlated with decreased expression levels of miR-148b in pancreatic cancers. In vitro experiments further showed that both mRNA and protein levels of AMPK α1 were decreased by miR-148b mimic, while increased by inhibitors of miR-148b. Zhang et al. ²² constructed luciferase reporter vectors containing AMPK α1 wild type (WT) or 3′ untranslated region (UTR) mutants and demonstrated that miR-301a significantly reduced the luciferase activity of AMPK α1 WT 3′UTR-containing reporter but not the mutated reporter, indicating that miR-301a directly targets AMPK α1. In addition, Craig et al. ²⁴ analyzed mRNA level of AMPK α with a variety of miRNAs and found 37 different miRNA families including let-7 may regulate the expression of AMPK α1, but only miR-140a can bind the 3′UTR of AMPK α2 mRNA. Kim et al. ²⁵ performed a comprehensive analysis of mRNA and miRNA expression and found AMPK α2 mRNA expression was inversely correlated with miR-19a expression, suggesting that miR-19a likely targeted the 3′UTR of AMPK α2. It is worth mentioning that miRNA-9 increases AMPK protein expression and AMPK activity. ²⁶

Figure 2.

Regulation of AMPK activity by different miRNAs.

In addition, miRNAs regulate AMPK kinase activity by targeting AMPK upstream kinases. CAB39 (MO25α) is a scaffold protein capable of forming LKB1/STRAD/CAB39 stable complex to increase LKB1 activity and positioning. Godlewski et al. ²⁷ found that miR-451 antagonist-treated cells demonstrated increased CAB39 protein level and AMPK phosphorylation. Further studies showed that miR-451 directly targets CAB39 (MO25α) and reduces LKB1-dependent phosphorylation of AMPK resulting in decreased AMPK activity and increased mTOR activity. Accordingly, overexpression of miRNA-451 reduces the protein and mRNA levels of CAB39, which in turn leads to decreased expression of LKB1, p-AMPK, p-ACC, and p-AKT. ²⁸ Moreover, Kuwabara et al. ²⁹ found that miRNA-451 directly targets CAB39, resulting in reduced activation of LKB1/AMPK signaling pathway. In addition, miRNA-195 regulates AMPK activity through targeting MO25. ³⁰

Post-translational modifications of AMPK

Dephosphorylation

AMPK activation is a reversible process, which is regulated by the AMPK α subunit T172 phosphorylation. AMPK dephosphorylation can result from either direct T172 dephosphorylation by serine/threonine protein phosphatases or AMPK α subunit S485 phosphorylation (Figure 3). Serine/threonine protein phosphatases are divided into four categories: PP1, PP2A, PP2B, and PP2C. Early studies found that PP2C is the main serine/threonine protein phosphatase for the dephosphorylation of AMPK. ³¹ A number of studies also confirmed that PP2C dephosphorylates AMPK T172 leading to AMPK inactivation.^32–34 However, recent studies have shown that AMPK can be dephosphorylated by PP1 and PP2A, as well as PP2C.^35–39 S485 phosphorylation of AMPK α subunit is also one of the mechanisms of AMPK inactivation.^40–42 Berggreen et al. ⁴⁰ showed that phosphorylation of AMPK S485 by AKT (protein kinase B (PKB)) reduced T172 phosphorylation in fat cells, resulting in a 25% decrease in active AMPK α1. Suzuki et al. ⁴³ found that there are GSK3 phosphorylation sites in the ST-stretch (aa 463–520) of the AMPK α C-terminus. In a catabolic environment such as low energy or reduced PI3K-Akt activity, the non-phosphorylated serine/threonine-rich loop (ST)-stretch may provide a unique phosphorylation-sensitive shield for AMPK T172 site and hinder the access of phosphatases to AMPK activation loop. Conversely, phosphorylation of AMPK α subunit ST-stretch at the S485 site by AKT causes a GSK3-dependent phosphorylation of T479, prompting the ST-stretch to leave AMPK kinase domain, thereby inducing the exposure and dephosphorylation of AMPK T172 site. In the anabolic environment, AMPK α subunit kinase domain is more accessible to phosphatases through other conformational change.

Figure 3.

Structure and the sites of phosphorylation (P), ubiquitination (U), and myristoylation (M) on AMPK subunits.

Ubiquitination

Ubiquitination is an important post-translational modification of proteins and plays a pivotal role in the regulation of protein stability, function, and intracellular localization. Ubiquitination of AMPK promotes its proteasome-dependent degradation. Ubiquitination of either AMPK α and β subunits or upstream kinases results in the down-regulation of AMPK activity. Pineda et al. ⁴⁴ found that MAGE-A3/6-TRIM28 ubiquitin ligase complex physically interacts and ubiquitinates AMPK α1 leading to its degradation, thereby activating mTOR signaling and inhibiting autophagy. However, under inhibition of ubiquitination process using small molecule inhibitor, AMPK α1 subunit can still be ubiquitinated. ⁴⁵ In this condition, AMPK α1 subunit ubiquitin chains are mainly composed of atypical lysine29 and lysine33 ubiquitination (Figure 3). The atypical lys29 ubiquitin chains also facilitate lysosomal protein degradation of AMPK α1 subunit. ⁴⁶ AMPK activation may also increase the expression of ubiquitin E3 ligase or MAFBx/Atrogin-1 and MuRF1 ubiquitin enzymes, which in turn mediate a negative feedback regulation. ⁴⁷ Residues 313–548 of AMPK α1 subunit are the potential targets of ubiquitination, ¹⁸ as deletion of these amino acids increases AMPK α subunit protein half-life by fourfold. ¹⁶ PEST domain between the α1 subunit residues 504 and 526 may also contribute to the rapid degradation of AMPK α1 subunit by ubiquitin-26S proteasome pathway. Moreover, an E3 ubiquitin enzyme called Cidea (cell death-inducing factor DFFA like a) targets AMPK β for ubiquitination at Lys48 and degradation, thereby resulting in decreased activity of AMPK. ⁴⁸ In addition, AMPK activity is closely related to the ubiquitination of kinases related to AMPK activity. For example, NUAK1 (AMPK-related kinase 5) and MARK4 (microtubule-affinity-regulating kinase 4) are positive regulators of AMPK kinase and can be polyubiquitinated. ⁴⁹ The ubiquitination of NUAK1 and MARK4 is mainly constituted by atypical ubiquitin chains of lysine-33 and lysine-29. ⁵⁰ In yeast two-hybrid study, AMPK was found to interact with PSMD11 proteasome, affecting its phosphorylation status and potential function, ⁵¹ but the physiology of this interaction remains elusive.

Acetylation

Acetylation is the process by which an acetyl group is transferred to protein lysine residues by acetyltransferase. Acetylation of histone proteins plays an important role in epigenetic modification and gene regulation. Protein acetylation affects protein enzyme activity, stability, and subcellular localization of proteins. ⁵² Yeast homologs of AMPK α and γ subunits are Snf1 and Snf4, and for β subunits are Sip1, Sip2, and Gal83.^53–55 Studies have shown that yeast Sip2 (AMPK β homolog) mutations can increase Snf1 activity, indicating that Sip2 regulates Snf1 activity. ⁵⁶ Lu et al. ⁵⁷ examined the half-lives of a variety of Sip2 mutants and found that Sip2 acetylation inhibits Snf1 activity, reduces downstream Sch9 (mammalian Akt/S6k homolog) phosphorylation, and ultimately leads to slow cell growth, thereby extending life span. Sip2 has K12, K16, K17, and K256 acetylation sites, which are acetylated by acetyltransferase NuA4 and deacetylated by deacetylase Rpd3. Sip2 N-terminal acetylation may change its conformation resulting in its binding to Snf1 and/or γ subunit to stabilize Sip2–Snf1 complex, thereby altering Snf1 activity. This formation of complexes may enhance the physical interaction of Sip2 with Snf1, leading to inhibition of the catalytic activity of snf1. After analysis of the acetylation sites of more than 3000 lysines of 1750 proteins by mass spectrometry, Choudhary et al. ⁵⁸ showed that PRKAA1 is an acetylated protein, suggesting that mammalian AMPK α1 may also be acetylated, but its exact mechanism remains to be determined.

Myristoylation

Myristoylation is the addition of a myristic acid to glycine residue next to N-terminal methionine residue by N-methyl transferase. Warden et al. ⁵⁹ have shown that AMPK β1 subunit contains the N-terminal myristoylation site and mutation of glycine-2 to alanine increases AMPK activity by fourfold, suggesting that AMPK β1 subunit myristoylation may inhibit AMPK activity. However, Oakhill and others showed that β subunit myristoylation is not only required for AMP-activated T172 phosphorylation but also for maximum activation of AMPK. It was reported that AMPK β subunit myristoylation can lead to phosphorylation of αT172 directly by AMP stimulation (Figure 3), while the absence of AMPK β subunit myristoylation or β-G2A mutation impairs AMP function on AMPK α1 T172 phosphorylation. ⁶⁰

Challenges and prospective

AMPK activity is lower in various tumor tissues than adjacent paratumor or normal tissues,^61,62 and further reduced with advancing tumor stages. For example, in ovarian cancer, AMPK β1 mRNA and protein expression, as well as AMPK activity, are gradually attenuated with malignant progression of cancer. ⁶³ Therefore, during the multiple steps of tumorigenesis, AMPK activity is gradually lost. Several factors such as AMPK α self-inhibition, post-transcriptional, and post-translational modification of AMPK subunits contribute to the down-regulation of AMPK activity in cancer cells, but the precise mechanism of the gradual loss of AMPK activity in different tumors remains largely unknown. It is possible that gene mutations of AMPK cause loss or lack of AMPK activity, or mutations of AMPK upstream kinases lead to decreased phosphorylation and activation of AMPK. Moreover, the reshuffling of energy metabolism of tumor cells may also change tumor cell microenvironment, resulting in the reduction of AMPK activity. These issues need further investigation. Deep understanding of the mechanisms by which AMPK is down-regulated in cancer cells may provide novel targets for cancer prevention and treatment, such as reversal of AMPK expression by miRNA antagonists.