PI3K/AKT/mTOR Pathway Alterations Promote Malignant Progression and Xenograft Formation in Oligodendroglial Tumors

This article is featured in Highlights of This Issue, p. 4197

The World Health Organization (WHO) revised 2016 neuropathologic classification defines oligodendroglioma (grade 2) and anaplastic oligodendroglioma (AOD, grade 3) as a specific molecularly defined subtype of adult diffuse gliomas. Mutation of the isocitrate dehydrogenase genes (IDH1/2) represents one of the fundamental and earliest molecular events in the genesis of these tumors (1, 2), and has therefore been investigated as a potential therapeutic target (3–5). OD/AOD are also characterized by whole chromosome arm losses of 1p and 19q (6), and TERT promoter mutations are almost universal (>95%), while CIC and FUBP1 are also frequently observed, irrespective of tumor grade (1, 6–8). These molecular markers enabled improved classification of gliomas for predicting patient outcome (9, 10). Additional recurring mutations, in the genes PIK3CA, PIK3R1, and NOTCH1, have been identified in a subset of oligodendroglial tumors (11–13). In contrast, TP53 and ATRX mutations, which are commonly seen in astrocytic gliomas with IDH mutation (diffuse astrocytoma, anaplastic astrocytoma, and secondary glioblastoma), are exceedingly rare in OD/AOD.

In aggregate, IDH-mutant gliomas have better prognosis than IDH wild-type gliomas, across all WHO grades (12, 14). Within IDH-mutant gliomas, the prognosis of oligodendroglial (1p/19q codeleted) tumors is more favorable than astrocytic (TP53 and ATRX mutated) tumors. In addition, oligodendroglial tumors are highly responsive to radiation plus chemotherapy with procarbazine, lomustine (CCNU), and vincristine (known as PCV), or temozolomide (15–17). No survival difference could be distinguished between WHO grade 2 and grade 3 oligodendroglial tumors in a large clinical cohort (12). While oligodendrogliomas can undergo malignant degeneration with a fatal outcome (1, 15–17), strict diagnostic reclassification by WHO 2016 raises the possibility that the progression of IDH-mutant, 1p/19q codeleted OD/AOD may be poorly characterized; indeed, few studies have identified predictors of tumor progression within WHO2016-defined oligodendroglial tumors (11, 18, 19).

Recent studies have shown that, in OD/AOD, a subpopulation of undifferentiated cells was enriched for proliferative potential and expressed stem cell signature (20). This finding suggests that only a subset of undifferentiated cells plays a pivotal role in the maintenance of OD/AOD. Patient-derived glioma xenograft (PDX) models are derived from these undifferentiated cells, which manifest a cancer stem-like cell phenotype (21–23). Indeed, early-passage orthotopic xenografts generated with glioma tumorspheres, in most cases, recapitulate the phenotypic and genotypic characteristics of parent glioma, including critical driver gene mutations (24). Our recent efforts generating IDH-mutant xenografts indicated that the establishment and propagation of these xenografts was often dependent upon the presence of a “tertiary” mutation (genomic alterations beyond disease-defining alterations in IDH, ATRX, TP53, and chromosomes 1p/19q), which was a hallmark of clinically progressive behavior in the patient's tumor (25). Given the indolent course of primary OD/AOD, it is therefore not surprising that establishment of tumor xenograft models has been difficult, with only a few models reported (5, 26, 27). Taken together, our experience suggests that focusing efforts on clinically progressive OD/AOD may simultaneously achieve three important goals: (i) improving the chances of successful xenograft establishment; (ii) providing insights into the genomic evolutionary events that underlie malignant progression in oligodendroglial tumors; and (iii) identifying novel therapeutic targets for progressive tumors.

Herein, we describe a longitudinally monitored AOD patient, who manifested progression of two anatomically separate tumors that were resected for investigation. From a region of the tumor that subsequently rapidly progressed, we successfully established patient-derived AOD xenografts. This xenograft harbored a hotspot mutation in PIK3CA and activation of the PI3K/AKT/mTOR pathway, findings that were also observed within this region of the patient's tumor. In contrast, cells harvested from a distinct tumor region, which was stable after surgery and had wild-type PIK3CA, did not induce xenograft formation. Additional attempts at xenografting patient oligodendroglial tumors revealed a tight association between activating mutations in the PI3K/AKT/mTOR signaling pathway and successful xenograft formation, supporting the pivotal role of the PI3K/AKT/mTOR pathway in driving progression of oligodendroglial tumors.

This study was conducted in accordance with Declaration of Helsinki. All tumor and blood samples were collected with patient consent under protocols approved at Yokohama City University (YCU, Yokohoma, Japan) Hospital and Massachusetts General Hospital (MGH, Boston, MA) Institutional Review Boards. To create glioma tumorsphere lines, fresh tumor specimens were obtained from surgery and enzymatically dissociated with 0.1% of Trypsin and DNase. Tumorsphere lines were cultured in serum-free neural stem cell medium, as described previously (3, 25, 28). All tumorsphere lines were cryopreserved less than passage 3 to use for in vitro experiments. Temozolomide (Sigma), Nimustine (ACNU, Tokyo Chemical Industry), Lomustine (CCNU), Procarbazine (Sigma), Vincristine (Sigma), FK866 (Sigma), LY294002 (Sellek), GDC-0068 (Cayman), BYL719 (Cayman), Everolimus (Cayman), and AGI-5198 (Sigma) were used.

Cells (1–2 × 10⁵) were orthotopically implanted into the right striatum of 7- to 9-week-old female SCID Beige mice (Charles River Laboratories) within 24 hours after dissociation of tumor samples. Mice were monitored daily and sacrificed when neurologic deficits or general conditions reached the criteria for euthanasia. Brains were harvested for pathologic studies and for acutely dissociated tumor cells that were cryopreserved, cultured for in vitro experiments, or repeatedly implanted into mice brains. All mouse experiments were approved by the Institutional Animal Care and Use Committee at MGH and YCU.

Tumor tissue specimens were fixed in 10% neutral-buffered formalin and embedded in paraffin. Hematoxylin and eosin staining was performed using the standard procedures. For IHC, 5-μm thick sections were deparaffinized, treated with 0.5% H₂O₂ in methanol, rehydrated, and heated for 20 minutes for antigen retrieval. After blocking with serum, tissue sections were incubated with primary antibody [1:100 for IDH1^R132H (Dianova), Ki-67 (Dako), MSH6 (Proteintech), LKB1/STK11 (Novus Biologicals), phospho-AKT (p-AKT, Ser473, Cell Signaling Technology), phospho-4EBP1 (p-4EBP1, Cell Signaling Technology), phospho-S6K (p-S6K, Cell Signaling Technology), phospho-S6 (p-S6) ribosomal protein (Cell Signaling Technology), ATRX (Sigma), p53 (Leica), p16INK4a/CDKN2A (R&D Systems)] at 4°C overnight. The next day, sections were washed with PBS, incubated with biotinylated secondary antibody for 30 minutes at room temperature, and then incubated with ABC solution (PK-6101, PK-6102; Vector Laboratories) for 30 minutes at room temperature. Finally, sections were incubated with DAB (Dako) and counterstained with hematoxylin. Only strongly stained cells were considered positive. Positive stained cells were quantitatively analyzed and classified as strong (≧20%), moderate (5%–20%), and weak (≦5%), respectively.

Cells were lysed in RIPA buffer (Sigma-Aldrich) with a Protease Inhibitor Cocktail Tablets (Roche). Fifty micrograms of protein was separated by 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes (Millipore) by electroblotting. After blocking with 1 or 5% nonfat dry milk in TBST [25 mmol/L Tris (pH, 7.4), 137 mmol/L NaCl, 0.5% Tween20], membranes were incubated at 4°C overnight with primary antibodies. After washing and incubation with horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology), blots were washed, and signals were visualized with chemiluminescent HRP substrate (Millipore). Primary antibodies used were p-AKT, p-mTOR (Cell Signaling Technology), p-4EBP1, p-S6K, Actin (Cell Signaling Technology), and Vinculin (Cell Signaling Technology).

Genomic and bisulfite-modified DNA was extracted using DNeasy Blood & Tissue and EpiTect Bisulfite Kits (Qiagen), per the manufacturer's protocol. PCR protocol and primer sequencing for IDH1 (exon 4), IDH2 (exon 4), TERT promoter, STK11 (exon 6), and PIK3CA (exons 8, 9, and 20) are described in Supplementary Table S1. Methods for cell line fingerprinting, Sanger sequencing, pyrosequencing, FISH, MLPA, multiplex PCR technology (SNaPshot), and microsatellite instability analysis are described in Supplementary Methods.

To assess cell viability, tumorspheres were dissociated into single cells and seeded into 96-well plates at 7,000 to 8,000 cells per well. After 12 to 24 hours, chemical inhibitors were serially diluted and added to wells. Cell viability was measured by CellTiter-Glo (Promega) assay at indicated time points.

Methionine (MET)-PET CT was performed at National Center for Global Health and Medicine (Tokyo), as described previously (29). For semiquantitative interpretations, the SUV_max was determined by a standard formula. The tumor/normal ratio (T/N ratio) of MET was calculated relative to the uptake in the contralateral frontal cortex.

Statistical analysis was performed with JMP Pro12 software. For parametric analysis, two-tailed t tests were used. Two-tailed Fisher exact test was performed for analysis of frequencies of nominal data. Survival analysis was performed using Kaplan–Meier method, and log-rank test was used to compare survival differences. Data were expressed as mean ± SEM. P < 0.05 was considered statistically significant.

The clinical course and pathologic findings of a representative patient (YMG6) are shown (Fig. 1A; Supplementary Fig. S1A), and summarized as follows: A previously healthy 33-year-old man presented with headache and MRI of the brain revealed a noncontrast enhancing cystic lesion located within the left frontal lobe. This primary tumor (designated YMG6I) underwent radiographically complete resection and was diagnosed as AOD, with 20% of Ki-67 positive cells. The patient then underwent 60 Gy of radiotherapy as an adjuvant treatment. Nine years later, MRI revealed a small noncontrast enhancing lesion at the margin of the original tumor site. ¹¹C-methionine PET imaging, which has utility for evaluating recurrent gliomas (30), revealed a relatively high uptake (Fig. 1A), consistent with tumor recurrence. The recurrent tumor was resected and the pathological diagnosis was WHO grade 2 glioma (YMG6R1) consisting of reactive astrocytes admixed with atypical cells. Temozolomide was administered for 24 cycles. Four years later, a contrast-enhanced recurrent lesion emerged, that was again resected (YMG6R2) and the diagnosis was AOD (WHO grade 3). Temozolomide treatment resumed for additional 20 cycles. However, two distant contrast-enhancing tumors with higher uptake of ¹¹C-methionine developed: one at the left frontal base (YMG6R3F) and a second within the left medial temporal lobe (YMG6R3T), while the initial site of tumor remained unchanged. Gross total resection was performed for both distant tumors. Both tumors were diagnosed as WHO grade 3 glioma with Ki-67 index approximately 40%. Intriguingly, YMG6R3T (left temporal tumor) subsequently developed a rapid local recurrence in 4 months, whereas the left frontal tumor (YMG6R3F) has remained stable for more than 24 months. Finally, the left temporal tumor was again resected (YMG6R4) with the histopathology exhibiting similar findings (WHO grade 3).

Genomic analysis identified mutations in IDH1 (R132H) and the TERT promoter (c.-124C>T, C228T) in all the tumors throughout the course including the initial tumor (YMG6I), consistent with these mutations being early events in gliomagenesis (Fig. 1B; Supplementary Fig. S1B; refs. 1, 11, 12). FISH and Multiplex Ligation-dependent Probe Amplification (MLPA) assay demonstrated whole chromosome arm losses of 1p and 19q (Fig. 1B; Supplementary Fig. S1B and S1C). Altogether, the WHO 2016 integrated classification was AOD, IDH1 mutant and 1p/19q codeleted.

We confirmed matched DNA identification in YMG6R3F and YMG6R3T cells (Supplementary Fig. S2). To evaluate genomic evolution, we profiled each tumor using an NGS panel that interrogates the exons of 94 cancer-related genes, direct DNA sequencing, and IHC analysis (Fig. 1C; Supplementary Fig. S1B; Supplementary Table S2). We found sequence variants in MSH6 and STK11 in normal blood DNA, not previously reported as pathogenic mutations, likely indicating germline polymorphisms (Supplementary Table S2; Supplementary Fig. S3A). MSH6 and STK11 was positively expressed in YMG6I (Supplementary Fig. S3B and S3C). Germline mutations in the MSH6 and STK11 are associated with Lynch syndrome and Peutz–Jeghers syndrome, respectively. However, upper gastrointestinal endoscopy, endoscopy colonoscopy, and whole-body FDG-PET imaging did not find any findings to suggest these diagnoses were present in this patient (Supplementary Fig. S3E–S3G). In addition, MSI was not found in all tumor samples, including non-hypermutant YMG6R2 and hypermutant YMG6R3T (Supplementary Fig. S4; Supplementary Table S3), further evidence against these germline mutations directly promoting AOD growth or treatment resistance in this case.

Five mutations were detected in YMG6R2 (IDH1, TERT, MSH6, STK11, and SMARCA4), while tumor specimens from subsequent recurrences harbored a much greater number of mutations: YMG6R3F (29 mutations), YMG6R3T (55), and YMG6R4T (27) cells, respectively (Fig. 1C). Twenty-seven of 29 mutations (93.1%) in YMG6R3F were shared by YMG6R3T, whereas 28 of 55 YMG6R3T mutations (50.9%) were private. YMG6R4T had all the 27 mutations that YMG6R3F and YMG6R3T shared but did not have the 2 private mutations of YMG6R3F (Fig. 1C; Supplementary Table S2). These findings imply that YMG6R3T and YMG6R4T arose from a subclone of YMG6R3F, and subsequently acquired additional mutations.

The elevated number of mutations in YMG6R3F, YMG6R3T, and YMG6R4T were suggestive of the emergence of a DNA hypermutation phenotype. G:C>A:T transition mutations were overwhelmingly dominant, as observed in 28 of 29 (96.6%) in YMG6R3F, 51 of 55 (92.7%) in YMG6R3T, and 26 of 27 (96.3%) in YMG6R4T, a pattern consistent with temozolomide-associated DNA hypermutation (Fig. 1D; refs. 1, 31). We then focused on mutations other than the 5 detected in the low-mutant load YMG6R2. Considering the strand on which C>T change occurred, the majority of C>T changes were observed at non-CpG sites, a signature of temozolomide-associated mutation rather than spontaneous deamination, as mutations at CpG sites were only 16.0%, 13.3%, and 19.0% in YMG6R3F, YMG6R3T, and YMG6R4T, respectively (Fig. 1E). Somatic mutation of MSH6^Gly409Glu was identified in YMG6R4T in which MSH6 expression was negative (Supplementary Fig. S3B and S3D). Thus, the DNA hypermutation phenotype of these later recurrences was consistent with temozolomide treatment–induced hypermutation, manifest in this case after at least 24 cycles of temozolomide treatment. Recurrent glioma with temozolomide-associated hypermutation has been shown to have increased MGMT methylation compared with initial untreated tumors (32). In this case, both nonhypermutated and hypermutated tumors after temozolomide treatment had high MGMT promoter methylation status (Supplementary Table S4).

Of note, among the 28 private mutations of YMG6R3T, only PIK3CA (c.1624G>A, Glu542Lys) was retained in YMG6R4T (Fig. 1C and F; Supplementary Table S2), an observation suggesting PIK3CA^E542K, and not the other private mutations, may be a driver genetic event of YMG6R3T and YMG6R4T. Interpreting the functional consequences of individual mutations in cases with hypermutation can be difficult, as the vast majority of alterations are nonselected passenger events (1). Nonetheless, PIK3CA mutations at highly recurrent sites (E542K, E545K, and H1047R) have been shown to activate oncogenic PIK3/AKT/mTOR pathways in cancer (33, 34). In contrast, mutations in ATRX and TP53 were identified in YMG6R3T, not YMG6R4T, and IHC analysis demonstrated retained expression of ATRX and low expression of p53 in YMG6R3T (Supplementary Fig. S5), indicating these may be nonfunctional passenger mutations. Intriguingly, pyrosequencing identified PIK3CA^E542K mutation in a small subset of tumor cells in YMG6I, which during subsequent treatment then fell below the detection limit in YMG6R1, YMG6R2, and YMG6R3F (Supplementary Fig. S6A). No other PI3K/AKT/mTOR pathway gene mutation was identified in YMG6R2 and YMG6R3F (Fig. 1C and F; Supplementary Table S2). To assess the functional consequences of PIK3CA^E542K mutation, we performed IHC for downstream effectors in the PI3K/AKT/mTOR pathway. Results demonstrated strong expression of p-AKT as well as p-4EBP1 and p-S6K, in YMG6R3T and YMG6R4T, but not in YMG6R1, YMG6R2, and YMG6R3F (Fig. 1G; Supplementary Fig. S6B). We did note scattered expression of p-4EBP1 and p-S6K in YMG6I, consistent with PI3K/AKT/mTOR pathway activation in a subset of cells with PIK3CA^E542K.

Glioma sphere cell lines were established from YMG6R3F and YMG6R3T and implanted into mice brains to assess the potential for engraftment. YMG6R3F implanted mice did not develop neurologic symptoms during a 250-day follow-up period and autopsy revealed only a small tumor (YMG6R3F secondary cells, derived from first-generation xenograft, designated YMG6R3Fsc; Fig. 2A). In contrast, YMG6R3T implanted mice rapidly developed neurologic deterioration (124 days), and large nodular tumors expressing IDH1^R132H were found (YMG6R3Tsc; Fig. 2B). Histologically, glial tumor cells with atypical nuclei that resembling oligodendrocytes exhibited high proliferative activity (Ki-67 index of 40%), consistent with the diagnosis of AOD (Fig. 2C). For genomic analysis, we first confirmed matched DNA identification between YMG6RT3 and the first-generation xenograft YMG6R3Tsc1 (Supplementary Fig. S2). YMG6R3Tsc1 harbored mutations in IDH1 (R132H) and TERT promoter (c.-124C>T, C228T), and whole chromosome arm loss of 1p and 19q (Fig. 2D and E; Supplementary Fig. S7A), fulfilling the WHO 2016 diagnostic criteria of oligodendroglioma. YMG6R4T cells were also tumorigenic (YMG6R4Tsc; Fig. 2F and G), and recapitulated the genotype of AOD (Supplementary Fig. S7B). After serial transplantation in mouse brains, PDXs generated from YMG6R3T and YMG6R4T cells maintained the phenotypic and genotypic characteristics of AOD (Supplementary Fig. S8A–S8C). YMG6R3T and YMG6R4T xenografts were similarly and consistently lethal within 130 days, whereas YMG6R3F xenografts did not cause animal death (Fig. 2H).

We identified 15 mutations in YMG6R3Tsc1 xenografts (Fig. 2I; Supplementary Table S2). Among them, 8 mutations, including IDH1^R132H, TERT promoter, PIK3CA^E542K, were retained from the parental primary tumor (YMG6R3T). Interestingly, we identified only 5 mutations in YMG6R4Tsc1, derived from the tumor that progressed after YMG6R3T, including IDH1^R132H, TERT promoter, and PIK3CA^E542K (Fig. 2I; Supplementary Table S2), whereas the other mutations (associated with hypermutation phenotype) that were lost were passengers. Importantly, we confirmed the maintenance of PIK3CA^E542K in serially passaged xenografts by Sanger sequencing (Fig. 2J; Supplementary Fig. S8C). IHC analysis revealed strong expression of p-AKT, p-4EBP1, and p-S6K in YMG6R3T and YMG6R4T xenografts, as compared to nonlethal, slow-growing YMG6R3F xenografts (Fig. 2K; Supplementary Fig. S8D). We therefore hypothesized that PIK3CA^E542K is the driver mutation that promotes oligodendroglioma progression in the patient and in orthotopic xenografts.

To test the hypothesis that genetic activation of the PI3K/AKT/mTOR pathway promotes xenograft formation in oligodendroglial tumors, we collected 12 additional oligodendroglioma specimens that harbored mutations in IDH1^R132H and the TERT promoter, and 1p/19q co-deletion, and performed orthotopic transplantation assays (Table 1). This cohort included 4 AODs with NGS-detected genetic alteration in the PI3K/AKT/mTOR pathway (Supplementary Table S5): YMG23 (newly diagnosed AOD, PIK3CA^His1047Arg and CDKN2A-CDKN2B homozygous deletion; Fig. 3A; Supplementary Fig. S9A and S9B); YMG5 (recurrent AOD, PIK3R1^{Ile571AsnfsTer31}; Fig. 3B; Supplementary Fig. S9C); MGG60 (newly diagnosed AOD, PIK3CA^His1047Leu; ref. 25); MGG137 (newly diagnosed AOD, MTOR^Ser2215Phe; Fig. 3C). We observed xenograft formation from YMG5, YMG23, MGG60, and MGG137, which all harbored genetic alteration in the PI3K/AKT/mTOR pathway (Fig. 3D–F). Mutations of IDH1, TERT, and PIK3CA as well as CDKN2A-CDKN2B homozygous deletion and p16INK4a/CDKN2A inactivation were retained in YMG23sc1 xenografts (Supplementary Fig. S9D–S9F). In contrast, we did not observe xenograft formation from YMG46, which harbored alteration of PIK3CA^Glu469Gly, not reported in the COSMIC database (Fig. 3G and H; Supplementary Fig. S10A; Supplementary Table S5). Also, we did not detect PI3K/AKT/mTOR pathway mutation in seven ODs/AODs: YMG28, YMG53, MGG78, MGG82, MGG109, MGG126, and MGG130 (Table 1; Fig. 3I and J; Supplementary Fig. S10B and S10C; ref. 25). These ODs/AODs without activating PI3K/AKT/mTOR pathway alteration did not form orthotopic xenografts (Table 1; Fig. 3I and J; Supplementary Fig. S10D). Presence of PI3K/AKT/mTOR pathway gene activating mutation was therefore tightly associated with xenograft lethality (6/6 PI3K/AKT/mTOR pathway mutant vs. 0/9 PI3K/AKT/mTOR pathway wild type; P < 0.0001; Table 1; Fig. 3K). There was no significant difference in Ki-67 index between xenograft generating and nongenerating patient specimens (P = 0.13; Supplementary Table S6).

IHC of patient tumor specimens demonstrated detectable expression of p-AKT, p-4EBP1, and p-S6K (or p-S6) in PI3K/AKT/mTOR pathway–mutated YMG23, YMG5, and MGG137 (Fig. 4A–C; Supplementary Table S6), as opposed to weak expression in PI3K/AKT/mTOR pathway wild-type YMG28, YMG46, and YMG53 (Fig. 4D–F). The xenografts derived from AODs carrying activating alterations in the PI3K/AKT/mTOR pathway similarly recapitulated increased expression of p-4EBP1 and p-S6K or p-S6 (Fig. 4A–C; Supplementary Table S7). Phosphorylation of AKT was also preserved in PIK3CA-mutant xenografts, YMG23sc1 and YMG5sc1. Serially passaged xenografts of YMG23 retained mutations of IDH1, TERT, and PIK3CA as well as PI3K/AKT/mTOR pathway activation found in the patient (Supplementary Fig. S11). Taken together, our results support a proposed role for activating PIK3/AKT/mTOR pathway alterations in promoting tumor progression and xenograft formation in oligodendroglial tumors.

The YMG6 patient experienced tumor progression on prolonged temozolomide therapy, suggestive of emergence of acquired resistance (Fig. 1A). Our PDX models offered an opportunity to test therapeutic vulnerabilities in vitro and in vivo. We first used a cell viability assay to test temozolomide response of tumor neurosphere cultures (tumorspheres) derived from the patient tumor in vitro. Despite methylated MGMT promoter, MSH6 inactivated YMG6R4T cells were highly resistant to temozolomide, in line with reported effects of mismatch repair deficiency on temozolomide response (Fig. 5A; Supplementary Fig. S3D; Supplementary Table S2; refs. 35, 36). YMG23 (AOD, MGMT unmethylated) and YMG14 (GBM, IDH wild type, MGMT unmethylated) tumorspheres were also resistant to temozolomide, while YMG12 (GBM, IDH wild type, MGMT methylated, MMR intact) and MGG152 (GBM, IDH1 mutant, MGMT methylated, MMR intact) tumorspheres were responsive. Indeed, YMG23 and YMG14 patient tumors progressed on temozolomide (Supplementary Fig. S12A and S12B). These results support the proposal that either inactivated MSH6 or unmethylated MGMT promoter can increase temozolomide resistance in glioma.

Intriguingly, in patient YMG6, a progressive contrast-enhancing lesion that emerged after removal of YMG6R4T showed complete response to 4 cycles of procarbazine, nimustine (ACNU), and vincristine, the regimen called “PAV,” which is analogous to PCV (Fig. 5B). Consistent with this clinical observation, YMG6R4T tumorspheres also responded to ACNU, CCNU, procarbazine, and vincristine (monotherapy) in vitro (Fig. 5C). YMG23 cells also responded to ACNU, CCNU, and procarbazine, but resisted vincristine. In contrast, YMG36 (GBM, IDH-wild type) responded rather poorly to ACNU, CCNU, and vincristine (Fig. 5C). Clinically, PIK3/AKT/mTOR gene mutant tumors, YMG5 and MGG137, showed durable response to chemotherapeutic agents (PAV or temozolomide regimen; Supplementary Fig. S12C and S12D). These observations indicate that alkylating agents can still retain effectiveness in PIK3/AKT/mTOR pathway mutant oligodendrogliomas, suggesting instead that alkylator sensitivity is dependent on MMR status or MGMT promoter status (37). Indeed, we did not find significant differences between the overall survival of patients with PIK3CA/PIK3R1–mutant oligodendrogliomas as compared with patients with PIK3CA/PIK3R1 wild-type oligodendrogliomas (Supplementary Fig. S13; Supplementary Table S8).

We recently reported that metabolic depletion of NAD⁺ using NAMPT inhibitors alone or in combination with temozolomide, which also lowers intracellular NAD⁺ levels, represents a potential therapeutic strategy for IDH1-mutant gliomas (3, 35, 38). As expected, OD/AOD tumorspheres were sensitive to NAMPT inhibitor, and combination with temozolomide increased the NAMPT inhibitor effects (Fig. 5D; Supplementary Fig. S14A). On the other hand, OD/AOD tumorspheres were resistant to direct inhibition of the mutant IDH1 enzyme using AGI-5198 (Fig. 5E; Supplementary Fig. S14B), consistent with previous observations in gliomaspheres (3).

We next tested whether inhibition of PI3K/AKT/mTOR signaling could inhibit xenograft formation and induce selective cytotoxicity in PI3K pathway altered AOD. To test whether inhibitors of the PI3K pathway could suppress the PI3K/AKT/mTOR pathway, YMG6R4T xenograft cells were treated with DMSO, 50 μmol/L of LY294002, or 5 μmol/L of GDC-0068 for 12 hours. As expected, PI3K inhibitor LY294002 and AKT inhibitor GDC-0068 potently blocked phosphorylation of AKT, mTOR, 4EBP1, and S6K in YMG6R4T cells (Fig. 5F), indicating on-target effects. LY294002 and GDC-0068 induced cytotoxic effects in PIK3CA-mutant YMG6R4T and YMG23 cells, while they did not in PIK3CA wild-type YMG28 tumorsphere cells (Fig. 5G). A similar profile was observed when YMG6R4T and YMG23 (PIK3/AKT/mTOR pathway mutant; sensitive) and YMG28 (resistant) were tested with another PI3K inhibitor (BYL719) and an mTOR inhibitor (everolimus; Fig. 5H). In contrast, YMG46 (nonactivating mutation of PIK3CA (E469K) was not sensitive to these compounds (Fig. 5I). Finally, we treated YMG6R4T cells with DMSO or LY294002 for 24 hours in vitro and implanted them orthotopically (Fig. 5J). Together, these results suggest that activating mutations in the PIK3/AKT/mTOR pathway may predict response to targeted inhibition of this signaling pathway.

PI3K pathway genes are frequently mutated in oligodendroglial tumors, including PIK3CA (20%) and PIK3R1 (9%; ref. 13). These mutations are predominantly found in recurrent tumors when paired primary and progressive tumors are compared (11), implying acquisition of PI3K pathway mutation is associated with progression, and therefore may be a driver for malignant transformation. However, beyond these temporal associations, no evidence exists supporting these mutations as promoters of tumor progression in oligodendroglioma. We here provide functional evidence that PIK3/AKT/mTOR pathway activating mutation promotes malignant progression using a series of patient-intracranial oligodendroglial xenografts and a patient whose two evolutionarily divergent and genetically distinct tumors followed contrasting clinical and laboratory courses. In the patient case, the PIK3CA^E542K mutant tumor rapidly recurred and formed lethal xenografts, while the PIK3CA wild-type tumor did not progress nor form progressive xenografts. In our oligodendroglioma series, we found that hotspot mutation and activation of the PI3K/AKT/mTOR pathway was tightly associated with progressive AOD xenografts, and found that PI3K/AKT/mTOR pathway activated mutant tumor models were selectively sensitive to PI3K/AKT/mTOR pathway inhibitors.

PI3K/AKT/mTOR pathway is a canonical growth regulatory pathway in cancer. A pan-cancer proteo-genomic atlas demonstrated that PI3K/AKT/mTOR pathway activation was correlated with distribution of mutations in key PI3K/AKT/mTOR pathway genes (33). Highly recurrent mutations at E542, E545, and H1047 in PIK3CA, nonsense/frameshift/indel mutations involving PIK3R1, and hotspot mutation (e.g., S2215F) in MTOR all potently induced expression of p-AKT (Ser473). In contrast, tumors with mutations not predicted to be functional demonstrated weak or no significant effect on p-AKT (Ser473; ref. 33). These prior reports, alongside our current work, collectively support the model that alterations in PI3K/AKT/mTOR pathway genes activate the PI3K pathway and promote malignant progression in oligodendroglial tumors.

Previous reports indicate that recurrence of oligodendroglioma is not accompanied by dramatic changes in the number of mutation and genome-wide methylation status (39). Interestingly, our case study displayed DNA hypermutation after tumor progression on temozolomide, and PIK3CA mutation emerged in one of the two recurrent lesions. Notably however, both PIK3CA-mutant (YMG6R3T) and PIK3CA wild-type (YMG6R3F) tumors harbored a DNA hypermutation phenotype, yet the former subsequently progressed and the latter did not, indicating the DNA hypermutation phenotype itself was not a direct promoter of malignant progression. In addition, the fact that the DNA hypermutation phenotype was lost in the subsequent progressive tumor (YMG6R4T) suggests the majority of mutations found in YMG6R3T might be passengers, whereas serially retained mutations, such as PIK3CA^E542K, represent the functional drivers.

As few oligodendroglioma xenograft models have been established to date, complete knowledge of the molecular drivers of oligodendroglioma progression remains unclear (5, 26, 27). A previous study demonstrated that ectopic expression of IDH1^R132H in the subventricular zone of the mouse brain induced early IDH-mutant gliomagenesis, including 2-hydroxyglutarate production and a global DNA methylation phenotype, however, was unable to sustain tumor growth (40). We also reported that secondary mutations, including TP53 and ATRX in astrocytic tumors or TERT promoter mutation and 1p/19q codeletion in oligodendroglioma, were not sufficient to form IDH1-mutant xenografts in immunocompromised mice (25). Recently, Philip and colleagues demonstrated mutant IDH1 cooperated with PDGFRA and loss of CDKN2A, ATRX, and PTEN to promote tumor development in vivo (41). These findings suggest that a tertiary genomic event may be required for xenograft formation in oligodendrogliomas as well.

All of our progressive oligodendroglioma xenograft models that successfully generated orthotopic xenograft carried functional genetic alterations in PIK3CA, PIK3R1, or MTOR. In contrast, none of PI3K pathway gene wild-type tumors or tumor with PIK3CA nonfunctional mutation induced xenograft formation. Clinically, we observed rapid tumor progression after resection of YMG6R3T and YMG6R4T (PIK3CA mutant), whereas YMG6R3F (PIK3CA wild-type) remained stable postoperatively, mirroring the inability to generate xenograft ex vivo. These findings are consistent with previous studies that associated genomic alterations in the PI3K pathway, such as PIK3CA and PIK3R1 mutations, with progression in IDH1-mutant gliomas (11, 25). A recent study also demonstrated that hotspot mutations in PIK3CA potentiated PI3K signaling and promoted tumorigenesis of immortalized normal human astrocytes in vivo (42). In support of these findings, our study also demonstrated delayed xenograft growth by PI3K pathway inhibition. These results support the crucial role of PI3K/AKT/mTOR pathway gene alteration in the formation and progression of oligodendroglioma xenografts.

Our oligodendroglioma xenograft models should facilitate research exploring therapeutic targets in progressive oligodendroglial tumors. Using these models, we provide experimental evidence that oligodendroglial tumors with PI3K/AKT/mTOR mutation are susceptible to selective inhibition of the PI3K/AKT/mTOR pathway. Interestingly, alkylating chemotherapeutics such as ACNU and procarbazine remained effective in the patients described and our oligodendroglioma models after developing resistance to temozolomide and developing temozolomide-induced hypermutation phenotype. Indeed, we did not find a survival difference of PIK3CA/PIK3R1–mutant oligodendrogliomas, when compared with wild-type oligodendrogliomas. Also, we found in our index case, that although the primary tumor (YMG6I) harbored PIK3CA^E542K, this subpopulation of cells fell below the limit of detection during radiation and temozolomide treatment. These findings imply that PI3K pathway gene alteration in oligodendrogliomas may promote malignant progression, but not necessarily resistance to effective cytotoxic chemotherapies. Early detection of subclonal PI3K pathway activating mutations could therefore potentially inform the timing of initiating such treatments, although further translational clinical studies will be needed to guide treatment decisions for patients with oligodendroglioma.

In conclusion, we identified the critical role of PI3K/AKT/mTOR gene activating mutation in promoting malignant transformation and generating xenografts in oligodendroglial tumors. These xenograft models provide insights into potential therapeutic strategies, including PI3K/AKT/mTOR pathway inhibitors, in a subset of oligodendroglial tumors.

T.T. Batchelor reports receiving commercial research grants from Pfizer and Oncoceutics, is a consultant/advisory board member for Genomicare, Champions Biotechnology, NXDC, Merck, Amgen, and Proximagen, and reports receiving other remuneration from CRICO, UpToDate, Oakstone, and Jiahui Health. A.S. Chi is an employee of Neon Therapeutics, and is a consultant/advisory board member for Cota Healthcare. A.J. Iafrate reports receiving commercial research grants from Sanofi, holds ownership interest (including patents) in ArcherDx, and is a consultant/advisory board member for Chugai, Debiopharm, and Roche. D.P. Cahill is a consultant/advisory board member for Lilly and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Tateishi, T. Nakamura, J.J. Miller, J. Sasame, T.T. Batchelor, H. Wakimoto, D.P. Cahill

Development of methodology: K. Tateishi, A.J. Iafrate, H. Wakimoto, D.P. Cahill

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Tateishi, T. Nakamura, T.A. Juratli, E.A. Williams, S. Miyake, A.L. Fink, N. Lelic, M.V.A. Koerner, Y. Miyake, T. Tanaka, R. Minamimoto, S. Matsunaga, S. Mukaihara, T. Shuto, H. Taguchi, N. Udaka, W.T. Curry, D. Dias-Santagata, T. Yamamoto, K. Ichimura, T.T. Batchelor, A.J. Iafrate, D.P. Cahill

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Tateishi, T.A. Juratli, Y. Matsushita, J.J. Miller, K. Fujimoto, T. Tanaka, S. Yamanaka, D. Dias-Santagata, K. Ichimura, A.S. Chi, A.J. Iafrate, H. Wakimoto, D.P. Cahill

Writing, review, and/or revision of the manuscript: K. Tateishi, J.J. Miller, H. Murata, W.T. Curry, T. Yamamoto, K. Ichimura, T.T. Batchelor, A.S. Chi, A.J. Iafrate, H. Wakimoto, D.P. Cahill

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Nishi, S.S. Tummala, N. Lelic, A. Ryo, S. Yamanaka

Study supervision: K. Tateishi, H. Murata, H. Wakimoto, D.P. Cahill

The authors thank Mrs. Emi Hirata and Yasuko Tanaka for technical assistance.

This work was supported by Grant-Aid for Scientific Research C (16K10765 to K. Tateishi), Princess Takamatsu Cancer Research Fund (to K. Tateishi), Takeda Science Foundation (to K. Tateishi), The Yasuda Medical Foundation (to K. Tateishi), Japanese Foundation for Multidisciplinary Treatment of Cancer (to K. Tateishi), Yokohama Foundation for Advancement of Medical Science (to K. Tateishi), SGH Foundation for Cancer Research (to K. Tateishi), a Bristol-Myers Squibb research grant (to K. Tateishi), OligoNation (to D.P. Cahill), and NIH R01CA227821 (to D.P. Cahill and H. Wakimoto).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.