A novel tumor suppressor protein encoded by circular AKT3 RNA inhibits glioblastoma tumorigenicity by competing with active phosphoinositide-dependent Kinase-1

GBM is one of the most lethal human cancers. With multimodal treatments available, including surgery, radiation and chemotherapy, the median survival time of patients with GBM is only 12–14 months [1]. Although many novel strategies have been tested for years, no effective target therapy has been validated in GBM to function better than temozolomide [1, 2]. The disruption of RTK/PI3K signaling is considered to be one of the major pathways during GBM tumorigenesis and progression [3]. The aberrant activation of RTK/PI3K signaling in GBM has multiple causes: the frequent amplification/mutation of EGFR (~ 45%), the mutation/homozygous deletion of PTEN (~ 36%) and the mutation of PI3K (~ 15%); these mutations collaboratively activate this pathway in ~ 88% GBM patients [4]. AKT, a serine-threonine kinase, is a key downstream effector of the RTK/PI3K pathway. Activated AKT, which is observed in ~ 70% GBM patients, especially in those with PTEN loss [5], mediates tumor cells proliferation, survival and malignant transformation through multiple downstream targets, including GSK-3, MDM2, and mTOR et al. [6]. The above evidence suggests that AKT is a reasonable therapeutic target for some GBM patients.

The human AKT family has three isoforms, AKT1, AKT2 and AKT3. Although they share with very similar sequences, isoform-specific functions have been reported. In knockout mice, AKT1 deprivation showed a growth retardation phenotype, and mice with AKT2 deletion had type II diabetes [7, 8]. In contrast, AKT3 facilitated postnatal brain development instead of homeostasis, indicating the critical role of AKT3 in the central nervous system [9]. AKT3, but not AKT1 or AKT2, was required for the anchorage-independent growth of transformed astrocytes and human glioma cells [10]. Although AKT3 expression was reported to be downregulated in GBM, it has shown a stronger kinase activity than that of AKT2 [11]. However, AKT3 was also shown to delay GBM progression in some GBM cases, raising the controversial role of AKT3 in GBM tumorigenesis [12].

We have previously reported that circular RNA encodes functional peptides or proteins in GBM [13‐15]. By using high-throughput RNA sequencing from GBM clinical samples and paired normal brain tissues, we identified that circ-AKT3 has low expression levels in GBM. Circ-AKT3 encodes a 174 amino acid novel protein, which we named AKT3-174aa. Functionally inverse to AKT3, AKT3-174aa blocks AKT thr-308 phosphorylation by competing with activated PDK1 and inhibits the proliferation, radiation resistance and tumorigenicity of GBM cells. Our results show that circ-AKT3 is a novel functional AKT transcript variant and that in addition to PTEN, AKT-174aa is a newly identified negative regulator of the RTK/PI3K pathway.

Of the multiple oncogenic signaling pathways that drive GBM tumorigenesis and progression, the RTK/PI3K/AKT pathway plays a central role. Evidence has shown that at least one RTK status was found altered in 67% of GBM cases overall, including EGFR, PDGFRA, MET and FGFR2/3. Meanwhile, the PI3K mutation was found in 25% of GBM cases. Although the PI3K and PTEN mutations were mutually exclusive, 59% of GBM cases showed one or the other. After adding up these genetic alterations, nearly 90% of GBM cases had at least one alteration in RTK/PI3K/AKT signaling and 39% had two or more [3, 27]. AKT represents a nodal point in this pathway. After PI3K activation, AKT was recruited to the plasma through the PH-domain and was them phosphorylated sequentially at thr308 and ser473 to become fully activated [28, 29]. Specifically, PDK1 (another PH domain-containing kinase) directly phosphorylated AKT at thr308, which is the most critical initial step for AKT activation [30]. AKT activation promoted cancer progression through the following three major biological functions: survival, proliferation and growth [31]. So far, three AKT isoforms have been identified (AKT1, AKT 2 and AKT3), which, in general, are broadly expressed, although there are some isoform-specific features. In breast cancer, AKT1 had been demonstrated to suppress while AKT2 promotes migration and invasion [32, 33]. In PTEN-deficient prostate cancer, AKT2, but not AKT1, mediates cellular survival and proliferation [5]. AKT3, but not AKT1 or AKT2, mediated the resistance to apoptosis in BRAF-targeted melanoma cells; promoted anchorage-independent growth in triple negative breast cancer; and controlled the VEGF-induced angiogenesis in ovarian cancer [34‐36]. However, the functions of AKT3 in GBM remain controversial: although reports have shown that AKT3 had lower expression levels, its kinase activity was higher than that of AKT2 [11]. Additionally, AKT3 was reported to exert a tumor suppressive function in GBM [12]. In current study, we showed that AKT3 was downregulated in GBM compare with normal brain tissue, which was consistent with the results of previous studies. But the consequences of the AKT3 downregulation in GBM may not be only due to its higher kinase activity; as AKT3-174aa may also play its part.

We and other groups have shown that circRNAs could encode functional peptides or proteins [13, 14, 37, 38]. Compared with their linear host gene products, circRNA encoded proteins or peptides are usually exerted independently of the biological functions. For example, we found that circPINTexon2 encoded PINT87aa, which is involved in translation elongation, while LINC-PINT showed a functional dependence with PRC2 [14, 39]. The most recent study also supported that circRNAs and linear mRNAs may have individual functions during prostate cancer carcinogenesis [40]. On the other hand, some translational products of the circRNAs involved in their host gene functions, such as SHPRH-146aa protects SHPRH from degradation [14] and FBXW7-185aa induce c-Myc degradation [13]. Interestingly, two recent reports have shown that AKT hyperactivation required SETDB1-induced methylation during tumorigenesis [41, 42]. The reported K64, K140 and K142 methylation sites were all located inside AKT3-174aa, implying that AKT3-174aa also may protect AKT from being methylated; however, further evidence is needed. The potential multiple ‘safe-guard’ role of AKT3-174aa showed that it may be a critical negative regulator of PI3K/AKT signaling expect for PTEN. Vo et al. recently reported that although circRNAs were globally expressed in cancers at low levels, circ-AKT3 seemed to be one of the exclusions [43]. However, their study did not include GBM samples, which were reported to have lower linear AKT3 levels compare with those of the other malignancies including breast cancer, ovarian cancer and melanoma [11, 34‐36, 44].

Except for inducing glioma invasion and anchorage-independent growth, AKT3 was reported to significantly activate DNA repair and resistance to radiation and chemotherapy in GBM cases [45]. However, AKT3 did not show any prognostic correlations with GBM patients in the TCGA database. We inferred that the lower expression of AKT3 in some GBM cases may confound the oncogenic characters of AKT3 and may induce the nonsignificant prognostic results. Instead, AKT3-174aa expression showed as a positive correlation with the patients’ total survival with GBM in our study. Considering that AKT3-174aa is one of the few negative regulators of PI3K/AKT signaling (others include PTEN, PHLPP [46], and CTMP [47]), its expression level should be more intensively tested in larger cohort to confirm whether it is an independent biomarker for GBM or other types of human cancers.

Given the central role of PI3K/AKT signaling in GBM, targeting PI3K or AKT is a logical rationale in developing novel therapeutic strategies. However, PI3K inhibitors, such as LY294002/wortmannin, and AKT inhibitors, such as perifosine, were only applied in experimental studies instead of in clinical trials. Basically, the unsatisfactory results of these novel drugs were attributed to the difficulties to penetrate the blood brain barrier and the fast restoration of p-AKT [44]. Although the tumor-suppressive role of AKT3-174aa prevents its druggable potential, we think that the low expression of AKT3-174aa may allow for AKT be easily exposed to p-PDK1 or SETDB1 in GBM, and this makes AKT more sensitive to activation cascades. Thus, effectively restore AKT3-174aa expression may benefit certain GBM patients to PI3K/Akt signaling target therapy, although the optimized delivery system for BBB penetration is required. Our results provide some evidence that the internal balance of the abovementioned genes’ alternative splicing products needs to be restored during GBM target-therapy, which could enhance or maintain the drug efficiency.