Genomically amplified Akt3 activates DNA repair pathway and promotes glioma progression

Kristen M. Turnera,1, Youting Suna,2, Ping Jia, Kirsi J. Granberga,b, Brady Bernardc, Limei Hua, David E. Cogdella, Xinhui Zhoua, Olli Yli-Harjab, Matti Nykterb, Ilya Shmulevichc, W. K. Alfred Yungd, Gregory N. Fullera, and Wei Zhanga,3

Departments of aPathology and dNeuro-Oncology, University of Texas MD Anderson Cancer Center, Houston, TX 77030; bDepartment of Signal Processing, Tampere University of Technology, Tampere, 33720, Finland; and cDepartment of Systems Biology, Institute for Systems Biology, Seattle, WA 98109

Edited* by Webster K. Cavenee, Ludwig Institute for Cancer Research, University of California, San Diego, La Jolla, CA, and approved February 13, 2015 (received for review July 30, 2014) Akt is a robust oncogene that plays key roles in the development disease progression and maintenance, whereas Akt1 does not play and progression of many cancers, including glioma. We evaluated a major role (15, 16). Another study in transformed mouse astro- the differential propensities of the Akt isoforms toward progression cytes indicate that Akt isoform function is potentially modulated by in the well-characterized RCAS/Ntv-a mouse model of PDGFB-driven the genetic landscape of the tumor (17). The underlying basis be- low grade glioma. A constitutively active myristoylated form of hind these differences and the major pathways that the isoforms Akt1 did not induce high-grade glioma (HGG). In stark contrast, Akt2 use remains largely unknown. To our knowledge, our study is the and Akt3 showed strong progression potential with 78% and 97% first to use a glial-specific transgenic mouse model to demonstrate of tumors diagnosed as HGG, respectively. We further revealed that Akt3 as a powerful inducer of glioma progression and to further significant variations in polarity and hydropathy values among the reveal the previously unidentified DNA repair function of Akt3. Akt isoforms in both the pleckstrin homology domain (P domain) and regulatory domain (R domain) were critical in mediating glioma Results and Discussion progression. expression profiles from representative Akt- Akt3 Tops Hierarchy in Mediating Glioma Progression. To gain a broad derived tumors indicated dominant and distinct roles for Akt3, perspective of the Akt isoforms in human cancer, we analyzed consisting primarily of DNA repair pathways. TCGA data from their copy number variation (CNV) across multiple cancer types. human GBM closely reflected the DNA repair function, as Akt3 Akt3 comprised the most frequently amplified isoform in many was significantly correlated with a 76-gene signature DNA repair cancers, including GBM, ovarian, melanoma, endometrial, and panel. Consistently, compared with Akt1 and Akt2 overexpres- breast cancer (Fig. 1A). Moreover, Akt3 was the least frequently sion models, Akt3-expressing human GBM cells had enhanced deleted isoform. In contrast, Akt1 was the least frequently ampli- activation of DNA repair proteins, leading to increased DNA fied and most deleted isoform. These CNV data suggest that Akt3 repair and subsequent resistance to radiation and temozolo- may be the most critical isoform in an array of cancer types. Indeed, mide. Given the wide range of Akt3-amplified cancers, Akt3 a prominent role for Akt3 has been well-documented in melanoma, may represent a key resistance factor. playing a critical role in tumor progression (18) and rendering cells resistant to (19). Akt3 has also been reported to be Akt | glioma | DNA repair | RCAS/tv-a mouse model the dominant isoform in hormone unresponsive breast (20) and

kt is among the most hyperactivated signaling pathways in Significance Ahuman cancer and is an important that regulates key cellular functions including cell growth, proliferation, angiogen- Glioblastoma is the most common and aggressive type of gli- esis, metabolism, invasion, and survival, among others oma, with a median survival of 15 mo. A major obstacle to ef- (1–3). Analysis of human glioblastomas (GBMs) from The Cancer fective treatment is de novo or acquired resistance to standard- Genome Atlas (TCGA) has revealed that the Akt signaling path- care therapies, including radiation and temozolomide. Enhanced way is one of the top altered pathways in GBM (4, 5). Additionally, DNA repair can allow damaged or mutated cells to survive, con- researchers have observed that the Akt activation status correlates tributing to resistance and tumor recurrence. We have identified with glioma grade (6). However, much less is known regarding the Akt3 as the dominant Akt isoform that robustly stimulates glioma precise roles of the Akt isoforms (Akt1, Akt2, Akt3) in this disease. progression. We also discovered key roles for Akt3 in activating Akt1, Akt2, and Akt3 share roughly 80% overall sequence DNA repair pathways, which led to enhanced survival of human identity and each contain three similar domains: a pleckstrin ho- glioblastoma cells following radiation or temozolomide treat- mology, kinase, and regulatory domain (7). Although they appear ment. Our work has potential broad application to multiple cancer to be very similar proteins, developmental studies as well as as- types in which Akt3 is expressed. Blocking this pathway may help sessment of the isoforms in cancer models have revealed that they prevent or alleviate DNA repair-mediated therapeutic resistance. possess distinct functions (1, 8). The most extensive work on iso- form differences has been performed in breast cancer. Akt2 is Author contributions: K.M.T., O.Y.-H., M.N., I.S., G.N.F., and W.Z. designed research; K.M.T., P.J., K.J.G., L.H., D.E.C., and X.Z. performed research; K.M.T., Y.S., K.J.G., B.B., O.Y.-H., M.N., reported to promote migration, invasion, and metastasis, whereas I.S., W.K.A.Y., G.N.F., and W.Z. analyzed data; and K.M.T., W.K.A.Y., G.N.F., and W.Z. wrote Akt1 is inhibitory (8, 9). Specific phosphorylation of palladin, an the paper.

actin bundling protein, by Akt1 led to decreased migration (10). The authors declare no conflict of interest. CELL BIOLOGY Conversely, Akt2 has been found to promote breast cancer epi- *This Direct Submission article had a prearranged editor. thelial to mesenchymal transition (EMT) through miR-200 mod- Data deposition: The sequences reported in this paper have been deposited in the Gene ulation (11) as well as an interaction with Snail1 on the E-cadherin Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE66177). promoter (12). Functional differences in glioma have been slightly 1Present address: Ludwig Institute for Cancer Research, University of California, San variable by using knockdown systems in transformed cells. The Diego, La Jolla, CA 92093. first studies were reported on Akt2 only, which was found to 2Present address: Sequenom Laboratories, San Diego, CA 92121. mediate invasion in rat C6 glioma cells (13), and inhibition of Akt2 3To whom correspondence should be addressed. Email: [email protected]. prolonged survival in an orthotopic rat model (14). Subsequent This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. reports have found that Akt2, and possibly Akt3, are important for 1073/pnas.1414573112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1414573112 PNAS | March 17, 2015 | vol. 112 | no. 11 | 3421–3426 Downloaded by guest on October 1, 2021 Fig. 1. Akt3 frequently amplified in cancer and tops hierarchy in glioma. (A) Copy number variation of Akt isoforms in TCGA cancers, shown in order of increasing Akt3 amplification percentage. (B) Percentage of HGG and representative tumors shown from Akt isoforms injected in conjunction with PDGFB. P values represent Fisher’s exact test. *P < 0.05, ***P < 0.001. Scale bar for images and insets, 100 and 50 μm, respectively.

prostate cancer cells (21), and a recurrent MAGI3-Akt3 fusion gene independent growth in transformed astrocytes (17). Additionally, was recently discovered in triple-negative breast cancer (22). Given the cellular localization pattern of the activated form of Akt (pAkt- that Akt3 represents 50% of total Akt in the normal brain, has 473) isoforms appeared to be distinct, perhaps providing clues of been documented to be essential for normal brain development differential pathways being activated. Tumors derived from Akt1 (23), and was preferentially amplified in GBM, we hypothesized displayed mostly nuclear staining of pAkt; Akt2 tumors displayed that Akt3 is the most critical isoform involved in glioma. primarily membranous staining; whereas Akt3 tumors displayed We directly tested the ability of Akt1, Akt2, and Akt3 to promote strong nuclear and cytoplasmic staining (Fig. 1B). These findings glioma progression in the RCAS/Ntv-a glial-specific mouse system. are consistent with a recent phosphoproteomics screen in immor- As previously reported, injection of the RCAS plasmid carrying talized lung fibroblasts, which revealed more evenly distributed PDGFB transduces normal, nestin-positive glial cells that are nuclear and cytoplasmic substrates in cells expressing Akt3, com- engineered to express the RCAS receptor, tv-a, leading to forma- pared with those in Akt1- and Akt2-expressing cells (27). This tion of low-grade glioma (LGG) in more than 80% of mice (24–26). study highlighted Akt1/3-mediated phosphorylation of an RNA The introduction of Akt1 in conjunction with PDGFB led to only processing regulator, which led to alternative splicing of FGFR2. 16% incidence of high-grade glioma (HGG), which was not statis- Additional studies are reporting potential nuclear functions of tically significant from PDGFB alone. However, when Akt2 or Akt3 Akt3, including Ago2-mediated translational repressive activities was injected with PDGFB, the incidence of HGGs rose to 78% and (28), transcriptional control of IGFBP3 in lung cancer cells (29), 97%, respectively (Fig. 1B), which is significantly higher than andERBB2,ERBB3,andERα in breast tumor cells from a mouse PDGFB alone (P < 0.001, Fisher’sexacttest)andAkt1(P < 0.001, transgenic model (30). Although these effects are likely mediated by Fisher’s exact test). In addition, Akt3 produced a higher incidence multiple complex systems, Akt3 has been reported to control the of HGGs compared with Akt2 (P < 0.05, Fisher’s exact test). Akt3 stability of the major nuclear export protein, mainte- tumors were particularly striking, as they frequently exhibited bi- nance region-1 (CRM-1), leading to altered mitochondrial bio- lateral patterns and were very large, with pseudopalisading necrosis genesis and autophagy (31). It is plausible that Akt3 may regulate and marked vascular proliferation. Even the high-grade tumors other biological processes via similar mechanisms. differed among the Akt isoforms, with the Ki67 proliferation index again reflecting the Akt3 > Akt2 > Akt1 hierarchy (P < 0.001, Variabilities Among Akt Isoform Domains Reflect Functional Differences. Fisher’s exact test) (Fig. 1B and Fig. S1). The striking invasiveness, To further explore the potential basis behind the functional dif- as evidenced by the number of bilateral tumors (Fig. S1), is ferences, we compared Grantham polarity plots among the Akt consistent with a prior study that found Akt3 to promote anchorage isoforms, which reflect charge distribution across amino acids (32).

3422 | www.pnas.org/cgi/doi/10.1073/pnas.1414573112 Turner et al. Downloaded by guest on October 1, 2021 Substantial differences were observed in two regions in the P do- inhibits Akt1/2 more selectively than Akt3, was evaluated in Phase main (amino acids 40–55 and 120–150) and the R domain (amino I and II clinical trials, with limited single agent activity (35). Our acids 445–460), but the kinase domain (K domain) was the most data suggest that future attempts to inhibit Akt should specifically consistent among the Akt isoforms (Fig. 2A and Dataset S1). target Akt2 and/or Akt3. Differences in polarity may affect protein interactions, leading to divergent cell signaling pathways. Notably, the K domain, con- Akt3-Specific Transcriptional Patterns Highlight DNA Damage/Repair taining one of the activating phosphorylation sites, Thr-308, had Processes. Given that there was a definitive hierarchy among the very little variation in polarity, indicating that this domain is likely isoforms in promoting glioma progression, we sought to analyze the not responsible for the functional differences among the isoforms. patterns of representative tumors to determine if A complementary measurement of hydropathy via Kyte–Doolittle distinct pathways were being activated. We performed microarray plots was consistent with the polarity plots, revealing similar vari- analysis on RNA isolated from high-grade tumors derived from able regions (Fig. S2 and Dataset S2). PDGFB alone and each PDGFB/Akt combination. Gene expres- We hypothesized that the variable regions in polarity and hy- sion patterns from Akt1/2/3 tumors were compared with that of dropathy could be responsible for the functional differences ob- PDGFB tumors. Consistent with the inability of Akt1 to induce served in glioma progression. To test this, we performed domain glioma progression, there were strikingly no differentially expressed swapping to generate chimeras in which the P or R domains of in Akt1 tumors compared with PDGFB. Conversely, tumors derived from Akt2 and Akt3 had 306 and 145 unique differentially Akt2 or Akt3 were swapped with Akt1. The resulting chimeras expressed genes, respectively (Fig. 3A and Dataset S3). Moreover, were referred to as 2P and 2R or 3P and 3R, respectively (Fig. 2B Akt2 and Akt3 shared only 67 differentially expressed genes, in- and Table S1). We injected each chimera with PDGFB into Ntv-a dicating that they likely activate distinct pathways. To assess the mice to determine its ability to induce progression. The data specific pathways these differentially expressed genes were associ- revealed that the P or the R domain of Akt2 or Akt3 strongly ated with, we performed (GO) enrichment analysis initiated progression. These results may have important implica- using the WebGestalt software (36). The significantly enriched tions for the development of future inhibitors. Prior attempts to (false discovery rate < 0.05) GO categories were organized as inhibit the Akt pathway in solid tumors have involved pan Akt a tree with broad/specific categories at the top and bottom. The inhibitors, most notably perifosine, a lipid-based compound that broad categories at the top of the Akt2 tree were phosphate prevents Akt translocation to the cell membrane. Phase II clinical metabolic and protein modification processes, and the most spe- trials with this inhibitor have not yielded favorable outcomes cific category at the bottom of the tree was amino acid phos- (33, 34). More recently, the allosteric inhibitor MK-2206, which phorylation (Fig. 3B). The corresponding genes from the Akt2 tree are depicted in a heatmap, and demonstrate that indeed these genes have separate patterns from Akt1 and Akt3 (Fig. 3C). The Akt3 GO analysis was very different, with broad categories in- cluding cellular macromolecule metabolic processes, response to cellular stress, and response to DNA damage stimulus. The most specific categories were DNA-related, including DNA metabolic processes, DNA replication, and DNA repair (Fig. 3D). These results reflect the strong nuclear pattern of p-Akt staining in the mouse tumors (Fig. 1B). Many DNA repair machinery genes were significantly up-regulated, including PolD1, PolE, PolE2, EXO1, PCNA,andFEN1 (Fig. 3E). Given the striking pattern and po- tential implications of up-regulated DNA repair pathways, we next analyzed the human gene expression patterns in the TCGA GBM dataset to assess whether human tumors showed similar pathways that were discovered in the mouse model. Partial correlations were calculated between the expression of each Akt isoform and a panel of 356 DNA repair genes across 166 GBM samples. In accordance with the mouse expression data, Akt3 expression was significantly associated with a panel of 76 DNA repair genes (false discovery rate < 0.05 and jpartial correlationj > 0.2) (Fig. 3F and Table S2). This pattern was not observed with either Akt1 or Akt2. These data highlight that the mouse model is a relevant and accurate repre- sentation of the human disease, and further, that Akt3 activation may contribute to therapeutic resistance (37).

Akt3 Enhances DNA Damage/Repair Process and Mediates Radiation and Temozolomide Resistance. To further investigate the potential role of Akt3 in DNA damage and repair process, we generated U87 stable cell lines that overexpress Akt1, Akt2, or Akt3 (Fig.

S3A). We first examined the cells for evidence of DNA damage CELL BIOLOGY response with γ-irradiation by assessing micronuclei (MN), a well defined readout of damaged nuclei. Strikingly, Akt3-expressing cells had the highest level of MN among them (Fig. 4A). We further characterized the role of Akt3 in DNA damage response by evaluating a key marker of DNA double strand break (DSB), Fig. 2. Variable regions in pleckstrin homology and regulatory domains con- p-γ-H2AX. We observed a significant increase in p-γ-H2AX foci tribute to Akt isoform function. (A) Grantham polarity values for each amino γ acid plotted. Enlarged portions depict variable regions. (B) Akt2/Akt1 and Akt3/ following -irradiation only in Akt3 cells, compared with empty Akt1 chimeras were generated by domain swapping and injected in combination vector (Fig. 4B)(P = 0.01). We observed that overexpression of with PDGFB. Representative H&E HGG images are shown. (Scale bar, 100 μm.) Akt3 following irradiation led to a strong phosphorylation of

Turner et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3423 Downloaded by guest on October 1, 2021 Fig. 3. Akt3 regulates DNA repair pathways. (A) Venn diagram representing number of shared and unique differentially expressed genes among mouse tumors generated from PDGFB with Akt1, Akt2, or Akt3. (B) GO enrichment analysis based on Akt2 differentially expressed genes. (C) Heatmap of the Akt2- related pathway genes from B.(D) GO enrichment analysis based on Akt3 differentially expressed genes. (E) Heatmap of Akt3-related pathway genes from D. (F) Partial correlation among Akt isoforms and Akt3 signature genes across 166 TCGA GBM samples.

Ataxia telangiectasia mutated (ATM) (Fig. S3B), a major first- homologous recombination (HR), two major pathways for DSB responder of DNA damage that mediates activation of a pleth- repair. To assess NHEJ, we performed a cell-free assay in which ora of transducers and effectors of the DDR. It should be noted linearized plasmid was incubated with nuclear extracts of U87 that the functional classification of 1,077 putative ATM substrates cells stably expressing Akt isoforms. The relative amount of reli- is very similar to the pathways that were enriched in Akt3-derived gated plasmid was monitored with semiquantitative PCR. Consis- tumors, including chromatin organization, stress response, DNA tent with our hypothesis, Akt3 nuclear extracts had significantly metabolism, and responses to DNA damage (38), providing more higher levels of religated plasmid, indicating Akt3 promotes NHEJ evidence that ATM is a key player in Akt3-mediated pathways. (Fig. 4C). HR was monitored using a well characterized cell-based Next we performed experiments that directly tested the ability repair system, where the Dr-GFP cassette, which contains two of Akt3 to promote nonhomologous end joining (NHEJ) and nonfunctional GFP genes and SceI sequences were stably

3424 | www.pnas.org/cgi/doi/10.1073/pnas.1414573112 Turner et al. Downloaded by guest on October 1, 2021 Fig. 4. Specific role for Akt3 in activating DNA damage response pathways. (A, Left) Micronuclei were detected in U87 cells with Akt isoforms. (A, Right) Quantification of MN relative to control cells was assessed by counting more than 100 DAPI-stained nuclei in U87 cells with Akt isoforms exposed to 2 Gy irradiation. ***P < 0.0001 ANOVA with Dunnett’s post hoc test. (B, Left) γH2AX foci in U87 nuclei with 2 Gy irradiation, fixed and stained 30 min post- irradiation. (Scale bar, 10 μm.) (B, Right) Quantification of γH2AX foci. Data represent foci in irradiated nuclei normalized to untreated nuclei. *P = 0.01, t test, corrected by multiple testing. (C) Semiquantitative PCR measuring end joining in U87 nuclear extracts. Extent of end joining assayed with primers spanning the breakpoint. (D) Homologous recombination of U87 cells stably expressing both an Akt isoform and a nonfunctional GFP cassette. Cells were transfected with I-SceI. Percent recombination represents GFP-positive cells as measured via flow cytometry. **P < 0.01, ANOVA with Dunnett’s post hoc test. (E and F) MTT measurement of U87 cell viabilty 5 d postirradiation or TMZ treatment. *P < 0.05, ***P < 0.001, two-way ANOVA with Tukey’s Post hoc test. (G) Number of mutations in TCGA GBM samples according to Akt3 copy number. P < 0.05, one tailed t test.

transfected into U87 cells stably expressing myristoylated Akt1, Methods Akt2, or Akt3. Upon DNA double strand break via the SceI en- Copy Number Variation Analysis. Copy number data from various cancer types donuclease, cells that undergo HR produce a functional GFP allele studied by TCGA were downloaded from Synapse (https://synapse.prod.sagebase. and may be monitored via flow cytometry. The stable U87 cells org) on March 26, 2013. Only cancer types with more than 200 samples were were transfected with SceI and collected 48 h following transfection considered. The data were produced by the Affymetrix Genome-Wide Human for flow cytometry assay. Akt3 overexpression led to a significant SNP Array 6.0 platform (Affymetrix) and recurrent copy number changes were identified by the GISTIC software (40) as described (4). The copy number ampli- increase in the population of GFP cells, validating its role as fication and deletion thresholds were set to be 1 and −1, respectively. a mediator of HR (Fig. 4D). These results were further corrobo- ratedinHeLacells(Fig. S3C). Furthermore, we showed that Akt3 Statistical Analyses of Microarray Data. Microarray experiments were carried rendered cells resistant to irradiation and temozolomide (TMZ) out using the Mouse Whole Genome Oligo Microarray kit from Agilent treatments via cell viability assays (Fig. 4 E and F). To assess the Technologies, following the manufacturer’s protocol (the detailed protocol potential level of DNA repair in human GBM, we analyzed the can be found on the Agilent website: www.agilent.com). In brief, 500 ng of number of mutations according to Akt3 copy number in TCGA total RNA from each sample was used and labeled with Cy3- or Cy5-CTP. samples, and found that the Akt3 high group had significantly After 17 h of hybridization at 65 °C, the arrays were washed and scanned with Agilent’s dual-laser-based scanner. Feature extraction software GE2- fewer mutations compared with the Akt3 low group (Fig. 4G), v4_91 was used to link a feature to a design file and determine the relative indicating increased DNA repair and potential therapeutic re- fluorescence intensity between the samples. The raw data were analyzed

sistance. The proposed model mechanism by which Akt3 promotes using R 2.15.2 with the Limma package (3). Background correction, within- CELL BIOLOGY DNA repair and glioma progression is depicted in Fig. S4. and between-array normalization were performed with parameters: GBMs are notoriously resistant to therapies and are among the bc.method = “normexp”, method = “loess”, and method = “Aquantile”,re- most difficult cancers to treat (39). Identifying and blocking key spectively. The normalized data were then log2 transformed. Selection of pathways which mediate resistance could positively impact the differentially expressed probes between conditions was carried out by the successful treatment of this disease. We have demonstrated that lmFit and eBayes functions included in the Limma package (41), which perform a linear model analysis and an empirical Bayes estimation of the Akt3 strongly promotes glioma progression and mediates activa- variance, taking into account the small sample size of microarray experi- tion of DNA repair pathways. We believe that this work has po- ments. P values were adjusted to obtain the q value, i.e., False Discovery Rate tential to be highly impactful, because Akt3 is operative in multiple (FDR), by the Benjamini–Hochberg multiple testing correction (42). FDR is cancer types and may be a key factor in therapeutic resistance. controlled at a level of 0.1. The probes were then mapped to genes (when

Turner et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3425 Downloaded by guest on October 1, 2021 multiple probes mapped to the same gene, the probe with the best q value Mutation Copy Number Variation Analysis. The mutation data (Illumina Ge- was selected to represent the gene). Heatmap plots were generated using nome Analyzer DNA Sequencing, Illumina) of 291 TCGA GBM samples was the R software package to visualize the normalized gene expression across downloaded from the Broad Institute Firehose data standardization run on samples. Genes were grouped by hierarchical clustering. March 9, 2013. Total number of mutations was calculated for each sample. The samples are then divided into two groups according to Akt3′s copy Functional Categories and Pathway Analysis. Differentially expressed genes number. Specifically, the Akt3 low (high) copy number group consists of from each experimental condition were used for the pathway analysis. We samples which are two MAD (Median Absolute Deviation) below (above) the performed gene ontology (GO) enrichment analyses using the WebGestalt median copy number value of Akt3. One-tailed t test is used to test the software (36) to identify the top enriched GO categories, which were or- hypothesis that Akt3 high copy number group has lower mutation counts ganized as Directed Acyclic Graphs (DAG, which is a tree-like structure). The than the other group. most broad categories at the top of the DAG are biology process, molecular function and cellular component, respectively. Going down the DAG, the Statistical Analyses. Differences in HGG and tumor proliferation index were categories become more specific. The top 10 enriched categories of each ’ tree were highlighted in red (if q value < 0.05) or maroon (otherwise). analyzed with the one-sided Fisher s exact test; H2AX foci and HR assay were analyzed via the one-sided t test with multiple testing correction. Please Partial Correlation Between the Expression of Akt Isoforms and DNA Repair refer to SI Methods for detailed descriptions of bioinformatics analyses. Genes. DNA repair genes were downloaded from the Gene Ontology (GO) category “DNA repair” (GO:0006281), including 356 human DNA repair Animal Studies. All animal experiments were performed in accordance with genes. RNA sequencing data (Illumina HiSEq. 2000, Illumina) of 166 TCGA The University of Texas MD Anderson Cancer Center Institutional Animal Care GBM samples was downloaded from the Broad Institute Firehose data and Use Committee guidelines. standardization run on March 9, 2013. Partial correlation measures the de- gree of association between two variables, with influences from a third set ACKNOWLEDGMENTS. We thank Peter Vogt (The Scripps Research Institute) of variables removed. To specify, the partial correlation between X and Y for discussions and for the RCAS-Akt plasmids. We also thank Bethany Alicie given a third set of variables Z is the correlation between the residuals and Sarah Dunlap for technical assistance. This work was partially supported obtained from the linear regression of X on Z and of Y on Z, respectively. We by NIH Grants CA141432, CA09850305, and U24CA143835; an MD Anderson Cancer Center core grant from the National Cancer Institute (CA16672); calculated the partial correlation between the expression of an Akt isoform a grant from Goldhirsh Foundation; the Academy of Finland (Projects and a DNA repair gene, with effects from the other Akt isoforms removed. 132877); Sigrid Juselius Foundation; and the Finnish Funding Agency for The expression of 76 DNA repair genes were significantly associated with Technology and Innovation Finland Distinguished Professor program (O.Y.-H. Akt3, but not with Akt1 or Akt2 (Benjamini–Hochberg adjusted false dis- and M.N.). Y.S. was supported by an Odyssey Fellowship. K.J.G. was supported covery rate < 0.05 and jpartial correlationj >0.2). by a grant from the Academy of Finland (259038).

1. Altomare DA, Testa JR (2005) Perturbations of the AKT signaling pathway in human 22. Banerji S, et al. (2012) Sequence analysis of mutations and translocations across breast cancer. Oncogene 24(50):7455–7464. cancer subtypes. Nature 486(7403):405–409. 2. Gonzalez E, McGraw TE (2009) The Akt : Isoform specificity in metabolism and 23. Easton RM, et al. (2005) Role for Akt3/protein kinase Bgamma in attainment of cancer. Cell Cycle 8(16):2502–2508. normal brain size. Mol Cell Biol 25(5):1869–1878. 3. Khan KH, Yap TA, Yan L, Cunningham D (2013) Targeting the PI3K-AKT-mTOR sig- 24. Holmes KM, et al. (2012) -like -binding protein 2-driven glioma naling network in cancer. Chin J Cancer 32(5):253–265. progression is prevented by blocking a clinically significant integrin, integrin-linked 4. Cancer Genome Atlas Research Network (2008) Comprehensive genomic character- kinase, and NF-κB network. Proc Natl Acad Sci USA 109(9):3475–3480. ization defines human glioblastoma genes and core pathways. Nature 455(7216): 25. Dunlap SM, et al. (2007) Insulin-like growth factor binding protein 2 promotes glioma 1061–1068. development and progression. Proc Natl Acad Sci USA 104(28):11736–11741. 5. Brennan CW, et al.; TCGA Research Network (2013) The somatic genomic landscape of 26. Moore LM, et al. (2009) IGFBP2 is a candidate biomarker for Ink4a-Arf status and glioblastoma. Cell 155(2):462–477. a therapeutic target for high-grade gliomas. Proc Natl Acad Sci USA 106(39): 6. Wang H, et al. (2004) Analysis of the activation status of Akt, NFkappaB, and Stat3 in 16675–16679. human diffuse gliomas. Lab Invest 84(8):941–951. 27. Sanidas I, et al. (2014) Phosphoproteomics screen reveals akt isoform-specific signals 7. Cheng GZ, et al. (2008) Advances of AKT pathway in human oncogenesis and as linking RNA processing to lung cancer. Mol Cell 53(4):577–590. a target for anti-cancer drug discovery. Curr Cancer Drug Targets 8(1):2–6. 28. Horman SR, et al. (2013) Akt-mediated phosphorylation of argonaute 2 down- 8. Dillon RL, Muller WJ (2010) Distinct biological roles for the akt family in mammary regulates cleavage and upregulates translational repression of MicroRNA targets. tumor progression. Cancer Res 70(11):4260–4264. Mol Cell 50(3):356–367. 9. Maroulakou IG, Oemler W, Naber SP, Tsichlis PN (2007) Akt1 ablation inhibits, 29. Jin Q, et al. (2014) Transcriptional and posttranslational regulation of insulin-like whereas Akt2 ablation accelerates, the development of mammary adenocarcinomas growth factor binding protein-3 by Akt3. 35(10):2232–2243. in mouse mammary tumor virus (MMTV)-ErbB2/neu and MMTV-polyoma middle T 30. Grabinski N, et al. (2014) AKT3 regulates ErbB2, ErbB3 and estrogen receptor α ex- transgenic mice. Cancer Res 67(1):167–177. pression and contributes to endocrine therapy resistance of ErbB2(+) breast tumor 10. Chin YR, Toker A (2010) The actin-bundling protein palladin is an Akt1-specific sub- cells from Balb-neuT mice. Cell Signal 26(5):1021–1029. strate that regulates breast cancer cell migration. Mol Cell 38(3):333–344. 31. Corum DG, Tsichlis PN, Muise-Helmericks RC (2014) AKT3 controls mitochondrial 11. Iliopoulos D, et al. (2009) MicroRNAs differentially regulated by Akt isoforms control biogenesis and autophagy via regulation of the major nuclear export protein CRM-1. EMT and stem cell renewal in cancer cells. Sci Signal 2(92):ra62. FASEB J 28(1):395–407. 12. Villagrasa P, et al. (2012) Akt2 interacts with Snail1 in the E-cadherin promoter. On- 32. Grantham R (1974) Amino acid difference formula to help explain protein evolution. cogene 31(36):4022–4033. Science 185(4154):862–864. 13. Pu P, Kang C, Li J, Jiang H (2004) Antisense and dominant-negative AKT2 cDNA in- 33. Ernst DS, et al. (2005) Phase II study of perifosine in previously untreated patients with hibits glioma cell invasion. Tumour Biol 25(4):172–178. metastatic melanoma. Invest New Drugs 23(6):569–575. 14. Pu P, Kang C, Li J, Jiang H, Cheng J (2006) The effects of antisense AKT2 RNA on the 34. Chee KG, et al. (2007) The AKT inhibitor perifosine in biochemically recurrent prostate inhibition of malignant glioma cell growth in vitro and in vivo. J Neurooncol 76(1): cancer: A phase II California/Pittsburgh cancer consortium trial. Clin Genitourin Can- 1–11. cer 5(7):433–437. 15. Michiue H, Eguchi A, Scadeng M, Dowdy SF (2009) Induction of in vivo synthetic lethal 35. Yap TA, et al. (2011) First-in-man clinical trial of the oral pan-AKT inhibitor MK-2206 RNAi responses to treat glioblastoma. Cancer Biol Ther 8(23):2306–2313. in patients with advanced solid tumors. J Clin Oncol 29(35):4688–4695. 16. Mure H, et al. (2010) Akt2 and Akt3 play a pivotal role in malignant gliomas. Neuro- 36. Zhang B, Kirov S, Snoddy J (2005) WebGestalt: An integrated system for exploring oncol 12(3):221–232. gene sets in various biological contexts. Nucleic Acids Res 33(web server issue): 17. Endersby R, Zhu X, Hay N, Ellison DW, Baker SJ (2011) Nonredundant functions for W741–748. Akt isoforms in astrocyte growth and gliomagenesis in an orthotopic transplantation 37. Liu Q, Turner KM, Alfred Yung WK, Chen K, Zhang W (2014) Role of AKT signaling in model. Cancer Res 71(12):4106–4116. DNA repair and clinical response to cancer therapy. Neuro-oncol 16(10):1313–1323. 18. Cheung M, Sharma A, Madhunapantula SV, Robertson GP (2008) Akt3 and mutant 38. Shiloh Y, Ziv Y (2012) The ATM protein: The importance of being active. J Cell Biol V600E B-Raf cooperate to promote early melanoma development. Cancer Res 68(9): 198(3):273–275. 3429–3439. 39. Mirimanoff RO (2014) High-grade gliomas: Reality and hopes. Chin J Cancer 33(1):1–3. 19. Shao Y, Aplin AE (2010) Akt3-mediated resistance to apoptosis in B-RAF-targeted 40. Beroukhim R, et al. (2007) Assessing the significance of chromosomal aberrations in melanoma cells. Cancer Res 70(16):6670–6681. cancer: Methodology and application to glioma. Proc Natl Acad Sci USA 104(50): 20. O’Hurley G, et al. (2014) Investigation of molecular alterations of AKT-3 in triple- 20007–20012. negative breast cancer. Histopathology 64(5):660–670. 41. Smyth GK, Yang YH, Speed T (2003) Statistical issues in cDNA microarray data analysis. 21. Nakatani K, et al. (1999) Up-regulation of Akt3 in estrogen receptor-deficient breast Methods Mol Biol 224:111–136. cancers and androgen-independent prostate cancer lines. J Biol Chem 274(31): 42. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical and 21528–21532. powerful approach to multiple testing. J R Stat Soc, B 57(1):289–300.

3426 | www.pnas.org/cgi/doi/10.1073/pnas.1414573112 Turner et al. Downloaded by guest on October 1, 2021