Circ-AKT3 and AKT3-174aa are expressed at low levels in GBM, which could more easily induce AKT-thr308 phosphorylation and sequential activation by p-PDK1. circ-AKT3 plasmid or AKT3-174aa ORF. Circ-AKT3 level was detected by PCR. Error bars represent three independent experiments, **, alternative splicing variants. Methods High-throughput RNA sequencing and quantitative reverse transcription-PCR were used NB-598 hydrochloride to identify the differentially expressed circRNAs in GBM samples and in paired normal tissues. High throughput RNA sequencing was used to identify circ-AKT3 regulated signaling pathways. Mass spectrometry, western blotting and immunofluorescence staining analyses were used to validate AKT3-174aa expression. The tumor suppressive role of AKT3-174aa was validated in vitro and in vivo. The competing conversation between AKT3-174aa and p-PDK1 was investigated by mass spectrometry and immunoprecipitation analyses. Results Circ-AKT3 is usually a previously uncharacterized transcript variant. Circ-AKT3 is usually expressed at low levels in GBM tissues compared with the expression in paired adjacent normal brain tissues. Circ-AKT3 encodes a 174 amino acid (aa) novel protein, which we named AKT3-174aa, by utilizing overlapping start-stop codons. AKT3-174aa overexpression decreased the cell proliferation, radiation resistance and in vivo tumorigenicity of GBM cells, while the knockdown of circ-AKT3 enhanced the malignant phenotypes of astrocytoma cells. AKT3-174aa competitively interacts with phosphorylated PDK1, reduces AKT-thr308 phosphorylation, and plays a negative regulatory role in modulating the PI3K/AKT signal intensity. Conclusions Our data indicate that this impaired circRNA expression of the gene contributes to GBM tumorigenesis, and our data corroborate the hypothesis that restoring AKT3-174aa while inhibiting activated AKT may provide more benefits for certain GBM patients. Electronic supplementary material The online version of this article (10.1186/s12943-019-1056-5) contains supplementary material, which is available to authorized users. transcript variant and that in addition to PTEN, AKT-174aa is usually a newly identified unfavorable regulator of the RTK/PI3K pathway. Methods Human cancer and normal tissues All GBM (gene (hsa_circ_0017250, hsa_circ_0112784 and hsa_circ_0112781) that were detected in our RNA-seq data, and all of them had a lower expression in GBM compared with that in normal tissues (Fig. ?(Fig.1a,1a, lower panel). Another circRNA generated from (hsa_circ_0000199) was recently reported to be upregulated in gastric cancer [20]. We did not find it in our RNA-seq data, which could be explained by organ-specificity. Of these three circRNAs, hsa_circ_0017250 and hsa_circ_0112781 were more abundantly expressed in normal brain tissues. Moreover, only hsa_circ_0017250 contains a complete ORF as shown in Fig. ?Fig.1b.1b. Due to the significant lower level of hsa_circ_0017250 (termed as circ-AKT3) in tumor group compare with that in normal group (11 folds), we next explored its endogenous expression. Open in a separate window Fig. 1 Identification of circ-AKT3 as a novel gene alternative splicing transcript. a Upper, Volcano plot of circRNA expression. X-axis: log2 ratio of circRNA expression levels between normal NB-598 hydrochloride and tumor tissues. Y-axis: the FDR value (?log10 transformed) of circRNAs. The green dot indicates hsa_circ_0017250 (circ-AKT3). Lower, identified three circRNAs from gene in our RNA-seq. b Illustration of the complete ORF in circ-AKT3. Circ-AKT3 used an overlap start-stop codon UAAUGA. c Illustration of the annotated genomic region of AKT3, the putative different mRNA splicing forms (linear splicing and head-to-tail splicing) and the validation strategy for the circular exon 3C7 (circ-AKT3). NB-598 hydrochloride Sanger sequencing following PCR conducted using the indicated divergent flanking primers showed the head-to-tail splicing of circ-AKT3 in HEK293T cells. d Left, relative RNA level of circ-AKT3 and linear AKT3 in different time point. Right, relative RNA level of circ-AKT3 and linear AKT3 treated with RNase R. Error bars represent three independent experiments, **, gene with a full length of 524?nt. Divergent primers spanning the circ-AKT3 junction amplified the predicted PCR products, and Sanger sequencing confirmed the circular junction (Fig. ?(Fig.1c).1c). Compared with linear form of mRNA, circ-AKT3 is resistant to RNase R digestion [21] and has a longer half-life (Fig. ?(Fig.1d).1d). We designed a circ-AKT3 overexpression plasmid and two circ-AKT3 junction shRNAs (Fig. ?(Fig.1e,1e, upper panel). These plasmids or shRNAs NB-598 hydrochloride could specifically upregulate or knockdown the circ-AKT3 level in 293?T cells, without affecting the linear AKT3 mRNA (Fig. ?(Fig.1e,1e, lower panel). Immunofluorescence was performed using a junction-specific probe and showed that circ-AKT3 mostly localized in the cytoplasm of NHA cells and that the overexpression or knockdown of circ-AKT3 could enhance AKAP12 or attenuate the fluorescence signal, suggested the specificity of the probes (Fig. ?(Fig.1f,1f, left panel). Cell fraction qPCR using circ-AKT3.