Fine-tuning AKT kinase activity through direct lysine methylation

ABSTRACT

In addition to the pivotal roles for histone methylation in the transcriptional regulation, emerging evidence suggests important roles for methylation of non-histone proteins in response to extra-cellular stimulatory events, with implications in governing tumorigenesis. Among the increasing list of non-histone proteins targeted for methylation, the tri-lysine-methylation modification of AKT has been recently identified to fine-tune its kinase activity and oncogenic functions. Moreover, our results implicate the histone methyltransferase SETDB1 as the methyltransferase modifying and activating AKT in a PI3K dependent manner. As such, the oncogenic function of SETDB1 in various cancers may be attributed to tumorigenesis, at least in part, through activating AKT. Therefore, targeting SETDB1, which modulates both epigenetic marks and AKT kinase activity simultaneously, is a potential strategy for novel cancer therapeutics.

Protein lysine methylation, including mono-, di-, and tri-methylation, is one of the key protein post-translational modifications (PTMs) that govern physiological or pathological phenotypes [1]. Lysine methylation was initially studied largely in the context of histone proteins, and functions in concert with histone acetylation, ubiquitination or other PTMs to regulate the gene expression [1]. Of note, the tri-methylation of histone H3 in lysine 9 (H3K9me3) or lysine 27 (H3K27me3) by SETDB1/G9a or EZH1/2, respectively, leads to repressed transcription of targeted genes [2–4]; whereas the di- and tri-methylation of H3K4 by SET1A/B or PRDM9 leads to activated transcription of targeted genes [5,6]. Recent reports reveal that alterations in histone methylation marks play critical roles in embryonic development, especially for somatic cell nuclear transfer embryos. Briefly, methylation/demethylation of specific lysine residues on histone proteins, such as H3K27me3 or H3K9me3, could overcome the epigenetic barriers to achieve animal cloning through somatic cell nuclear transfer [7–10]. Pathologically, aberrancies in histone (de)methylation result in the active expression of oncogenes or repressive expression of tumor suppressors, which subsequently contribute to tumorigenesis [11]. To this end, specific inhibitors antagonizing histone methyltransferases/demethylases such as EZH2, DOLT1 or LSD1, have been developed and provide new and promising options for targeted cancer therapies [12,13].

Recently, accumulating evidence indicates that non-histone proteins also undergo lysine methylation, a process that is also carried out by histone methyltransferases [14,15]. Mono- and di- methylation has not been shown to alter the tertiary protein structure or charge of the lysine side chain, while in contrast, tri-methylation creates a slightly hydrophobic interface to modulate the targeted protein enzymatic activity, cellular translocation, or interaction with other proteins [15]. Thus, the further identification of non-histone methylated proteins will enhance our understanding of the biological functions of methyltransferases. As such, developing specific small molecular inhibitors targeting oncogenic methyltransferases have been considered a promising strategy for cancer therapies. To this end, many oncoproteins such as MAP3K2 [16] and AKT [17,18] have been identified to be methylated by SMYD3 and SETDB1, respectively, implying potential critical roles of these methyltransferases in tumorigenesis.

As one of the most important and frequently activated oncogenes, AKT has been extensively investigated, including its upstream regulators and downstream effectors [19]. The frequently genomic alterations such as gain-of-function mutations of EGFR in glioblastoma and lung cancer [20,21], K-Ras in lung, pancreatic and colon cancer [22–24], or PI3KCA in breast cancer [25], as well as the deletion or loss-of-function mutations of PTEN in prostate and breast cancer [26], PP2A in melanoma, lung and breast cancer [27], or VHL in renal carcinoma [28] all markedly elevate AKT kinase activity. Consequently, the activated AKT could spatially and temporally phosphorylate a plethora of downstream substrates including p21/p27, TSC2, MDM2, GSK3β, Bad and FOXOs under various conditions to govern cell cycle [29,30], survival [31,32], metabolic homeostasis [33,34], resulting in diabetes and tumorigenesis [19,35].

It is well established that the full activation of AKT is largely dependent on the phosphorylation of AKT on Thr308 by PDK1 and Ser473 by mTORC2 in a phosphatidylinositol (3,4,5)-trisphosphate (PIP₃) and membrane translocation dependent manner [36,37] (Figure 1). Therefore, the phosphorylation of AKT on T308 and S473 are recognized as the predominant modifications and the faithful indicator of AKT activation, which is antagonized by protein phosphatase 2A (PP2A) [38,39] and PH domain leucine-rich repeat protein phosphatase (PHPLH) [40], respectively. In addition to phosphorylation of AKT on T308 and S473, the constitutive phosphorylation of AKT on T450 is also catalyzed by mTORC2, a process that appears to be independent of growth factor signaling [41]. Recently, the C-terminal phosphorylation of AKT on S477 and T479 by CDK2/Cyclin A was identified to optimize the conformation of the AKT C-tail to promote AKT phosphorylation on S473, thereby enhancing AKT kinase activity in a cell cycle-dependent manner [42,43]. Meanwhile, the phosphorylation of AKT1 on S477 and T479 also respond to DNA damage and growth factor-induced mTORC2 activation [42]. Moreover, in addition to phosphorylation, other PTMs including hydroxylation, ubiquitination and acetylation have also been reported to modify AKT kinase activity directly (Figure 1). Notably, in hypoxic conditions, due to the lack of oxygen, the HIF hydroxylase EglN1 is unable to efficiently hydroxylate AKT on Pro125/Pro313, which blocks pVHL-mediated recruitment of PP2A to dephosphorylate AKT-T308, thereby resulting in AKT activation [44]. This finding indicates that the hypoxic environment frequently encountered in solid tumors, or VHL deficiencies in clear cell renal carcinomas (ccRCC), likely hyper-activates AKT to facilitate tumorigenesis and drug-resistance [44].

Figure 1.

A schematic model illustrating the major upstream regulators and the corresponding PTMs of AKT. As illustrated, oncoproteins such as TRAF6/SKP2, SETDB1, PDK1, CDK1/Cyclin A and mTORC2 (labeled in red) activate AKT by directly modifying AKT for ubiquitination, methylation and phosphorylation, respectively. EglN1 hydroxylates AKT to inhibit AKT kinase activity in part by affecting AKT phosphorylation on T308. Most PTMs such as ubiquitination, acetylation, methylation and phosphorylation could be antagonized by CYLD, SIRT1, KDM4B, PP2A and PHLPP (labeled in blue), respectively, to inhibit AKT kinase activity. Furthermore, PTMs occurring in the PH domain such as ubiquitination, acetylation and methylation could promote AKT membrane translocation, and facilitate full AKT kinase activation. PH, Pleckstrin homology domain; KD, kinase domain; PTMs, post-translational modifications.

As the PH domain in the closed conformation locks AKT in an inactive state by preventing its full activation [45], the binding of the PH domain to the membrane likely switches it to the open conformation to unlock Akt kinase activation [46,47]. Thus, PTMs occurring on the AKT PH domain may affect AKT membrane translocation to govern AKT kinase activity. Of note, the acetylation of AKT by p300/PCAF on the PH domain promotes AKT membrane translocation and kinase activity, which is reversed by the deacetylase SIRT1, likely through antagonizing AKT acetylation [48]. Similar to AKT acetylation, TRAF6 or SKP2-mediated AKT K63-linkage poly-ubiquitination can also induce AKT membrane translocation upon insulin or EGF stimulation, respectively, to facilitate AKT kinase activity [49–51], and is antagonized by CYLD-mediated de-ubiquitination [52]. Moreover, this ubiquitination process has been revealed to be mediated by SETDB1-induced AKT tri-methylation at K64, which is recognized by KDM4A that serves to bridge the connection of TRAF6 to AKT [17]. Meanwhile, additional methylated residues were identified including the tri-methylation of AKT1 on K140/K142 within the linker region, which not only affects the methylation of AKT on K64 to enhance AKT ubiquitination and membrane translocation, but also alters AKT conformation to favor the unlocked state, thereby facilitating subsequent PDK1-mediated AKT phosphorylation on T308 [18]. Thus, methylation-mediated AKT activation invokes multiple layers of regulation, leading to elevated AKT ubiquitination and conformation changes, to fully activate the AKT kinase at the plasma membrane (Figure 1).

In addition to tri-methylation of AKT, mono-methylation of AKT-K64 and K140/142, as well as di-methylation of AKT1-K140/142 was also detected by the mass spectrometry [18]. Although in this study we have focused on the tri-methylation modification of AKT [18], whether AKT undergoes mono- and di-methylation and how these methylations contribute to AKT kinase activity and oncogenic functions warrant further in-depth investigation. It is possible that methyltransferase(s) other than SETDB1 may be involved in AKT mono- or di-methylation; thus, these methyltransferases may also be considered as potential drug targets to promote AKT kinase activity and oncogenic functions.

Unlike AKT1, AKT2 and AKT3 do not undergo significant tri-methylation on K64 or K140/142 [17,18]. Structurally, among the three human AKT members there is ~90% amino acid homology between the kinase domains, ~80% homology between the PH and C-terminal regulatory domains, but only ~50% homology between their linker regions [53], indicating that the linker region possibly has important roles in conferring distinct substrates or biological functions of the three AKT isoforms during tumorigenesis. In support of this notion, genetic studies have revealed that different Akt isoforms in mice have specific functional roles [54]. Briefly, disruption of Akt1 leads to a significant decrease in body size and adipogenesis in mice. Whereas deletion of Akt2 is associated with severe insulin resistance and diabetes. In contrast, loss of Akt3 decreases mouse brain size up to 20%. Mechanically, different AKT members play distinct roles in tumor progression, especially in mammary tumors [55], and the antagonistic roles of AKT1 and AKT2 have been identified in cell motility and invasion of breast cancer cells, but the underlying molecular mechanism remains largely elusive [56,57].

Although multiple lysine residues on AKT1 including K140/142 and K64 were identified by different groups to undergo methylation modifications, SETDB1 was characterized as the major methyltransferase to methylate these residues and activate AKT in a complicated manner [17,18]. Since oncogenic roles of SETDB1 have been demonstrated in multiple types of human cancers, great effort has been employed to identify SETDB1-specific inhibitors [27]. However, as of now, there are no specific small molecular inhibitors of SETDB1. The antibiotic mithramycin A and its derivatives were recently discovered to transcriptionally suppress SETDB1 [58], which has been validated in cells and xenograft mouse models [18]. We recently also demonstrated that treatment with mithramycin A could significantly decrease tumor growth in part by retarding AKT kinase activity. As bacterially purified recombinant SETDB1 proteins failed to methylate core or purified histones [59], this limitation may stall the development of the specific SETDB1 inhibitors when using an in vitro approach. Thus, more effort is necessary to develop a cell-based approach to screen for specific SETDB1 inhibitors in the future.

Notably, we identified that demethylation of AKT-K140/K142 was carried out by histone demethylase KDM4B (also termed JMJD2B), but not other members of the KDM4 family [18]. Consistently, we found that depletion or inhibition of KDM4B by shRNA [18] or specific inhibitors [60] could significantly increase AKT phosphorylation due to the reduced levels of AKT methylation (unpublished data). In contrast, depletion of KDM4A moderately decreased AKT phosphorylation although KDM4A could not demethylate AKT. Consistent with our findings, another group independently identified that KDM4A could bridge the methylated AKT with TRAF6 for AKT ubiquitination, membrane translocation and activation in a catalytic independent manner [17]. It has been reported that KDM4A and KDM4B share similar structure and functions in demethylating H3K9me3 [61]. However, whether KDM4A and KDM4B share other non-histone substrates and play redundant cellular functions is not well defined to date. In our experimental setting, KDM4A and KDM4B display opposing roles in regulating AKT kinase activity through distinct molecular mechanisms. Similar opposing phenotypes have been identified in other protein families, such as AKT1 and AKT2, which play opposing roles in regulating cell migration [56], or HIF1α and HIF2α, which display contrasting properties in clear cell renal carcinoma [62]. Thus, it remains to be further investigated whether KDM4A and KDM4B govern tumorigenesis by impacting H3K9 epigenetic modification or other non-histone substrates such as AKT.

Although KDM4 inhibitors have been developed and under preclinical study for anti-cancer treatment [60], our study demonstrates that inhibitors specifically targeting KDM4B could potentially activate AKT by increasing AKT methylation (unpublished data), which will promote the anti-apoptotic response potentially suppressing the efficacy of KDM4B inhibitors for tumor treatment. However, although KDM4A displays oncogenic functions by enhancing TRAF6-mediated AKT ubiquitination and activation, KDM4A demethylase activity is not necessary for this function. Hence, our data indicate that inhibitors targeting KDM4A would not restrain AKT kinase activity and will, therefore, suppress tumorigenesis likely via modulating histone or other protein methylations in addition to repressing AKT.

In summary, given the aberrant regulation and critical roles of the PI3K/AKT oncogenic signaling pathway in tumorigenesis, targeting the PI3K/Akt pathway with PI3K and/or AKT inhibitors have been developed for anti-cancer therapies. However, cellular toxicity of these inhibitors has restricted their potential clinical use [63,64]. Thus, recent reports have contributed to identify the upstream regulators of the AKT oncogenic kinase, including the identification of TRAF6/Skp2, CDK2/Cyclin A, p300/PCAF and most recently, SETDB1. These enzymes directly modify AKT kinase activity by promoting distinct types of PTMs occurring on AKT, and are considered as potential cancer therapeutic targets. As such, the development of specific inhibitors of these critical upstream positive regulators of AKT may serve as novel strategies for targeting hyperactive AKT-driven human cancers.

Acknowledgments

We apologize to many investigators whose important works were not cited here due to space limitations. This work is support in part by the NIH grants to W.W. (GM094777 and CA177910).

Disclosure statement

No potential conflict of interest was reported by the authors.