ARTICLE https://doi.org/10.1038/s41467-020-17279-1 OPEN A cytoskeleton regulator AVIL drives tumorigenesis in glioblastoma Zhongqiu Xie 1, Pawel Ł. Janczyk2,8, Ying Zhang3,8, Aiqun Liu4, Xinrui Shi1, Sandeep Singh 1, Loryn Facemire1, ✉ Kristopher Kubow5,ZiLi6, Yuemeng Jia1, Dorothy Schafer7, James W. Mandell1, Roger Abounader3 & Hui Li1,2 Glioblastoma is a deadly cancer, with no effective therapies. Better understanding and identification of selective targets are urgently needed. We found that advillin (AVIL) is 1234567890():,; overexpressed in all the glioblastomas we tested including glioblastoma stem/initiating cells, but hardly detectable in non-neoplastic astrocytes, neural stem cells or normal brain. Glioma patients with increased AVIL expression have a worse prognosis. Silencing AVIL nearly eradicated glioblastoma cells in culture, and dramatically inhibited in vivo xenografts in mice, but had no effect on normal control cells. Conversely, overexpressing AVIL promoted cell proliferation and migration, enabled fibroblasts to escape contact inhibition, and transformed immortalized astrocytes, supporting AVIL being a bona fide oncogene. We provide evidence that the tumorigenic effect of AVIL is partly mediated by FOXM1, which regulates LIN28B, whose expression also correlates with clinical prognosis. AVIL regulates the cytoskeleton through modulating F-actin, while mutants disrupting F-actin binding are defective in its tumorigenic capabilities. 1 Department of Pathology, School of Medicine, University of Virginia, Charlottesville, VA 22908, USA. 2 Department of Biochemistry and Molecular Genetics, School of Medicine, University of Virginia, Charlottesville, VA 22908, USA. 3 Department of Microbiology, Immunology, and Cancer Biology, School of Medicine, University of Virginia, Charlottesville, VA 22908, USA. 4 Tumor Hospital, Guangxi Medical University, Nanning 530021, China. 5 Department of Biology, James Madison University, Harrisonburg, VA 22807, USA. 6 Department of Orthopedics, The Second Xiangya Hospital, Central South University, Changsha 410011, China. 7 Department of Biology, University of Virginia, Charlottesville, VA 22908, USA. 8These authors contributed equally: Pawel Ł. ✉ Janczyk, Ying Zhang. email: [email protected] NATURE COMMUNICATIONS | (2020) 11:3457 | https://doi.org/10.1038/s41467-020-17279-1 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17279-1 lioblastoma (GBM) is the most common primary brain reorganization21. The expression of AVIL is more restricted, Gtumor and among the deadliest of human cancers. being low or undetectable in most tissues (Supplementary Despite advances in surgery, radiation and chemotherapy, Fig. 1c). As with many gene fusions, including IGH-MYC, and survival of patients affected by GBM remains dismal (~15 months TMPRSS2-ERG, the expression of a proto-oncogene may be after diagnosis)1–3. Clearly, better treatment options, and identi- misregulated by subjecting itself to the control elements of its fication of selective therapeutic targets are urgently needed. fusion partner. We suspected that fusion to MARS in rhabdo- Oncogene addiction describes a phenomenon according to myosarcoma is one mechanism to misregulate AVIL gene which tumor cells become reliant on the activity of a particular expression, and that AVIL may be misregulated by other oncogene and die once this activity is inhibited4,5. Many of the mechanisms in other cancers. We found that the AVIL locus is targeted cancer therapies exploit this concept6. It is perhaps best amplified in 15–18% of glioblastoma cases in The Cancer Gen- exemplified by the successful use of imatinib in the therapy of ome Atlas (TCGA) studies22,23 via cBioPortal analysis (Fig. 1a). chronic myelogenous leukemia (CML)7. In CML, the major dri- We confirmed such a copy number gain by FISH analyses, using a ver of tumorigenesis is the BCR-ABL fusion oncogene; imatinib probe covering the AVIL locus (Fig. 1b). The AVIL locus is inhibits the constitutively active BCR-ABL protein kinase, to amplified in two glioblastoma cell lines, SF767 and A172, but not which leukemic cells become addicted. Other successful examples in three other glioblastoma lines, U87, U251, T98G, or in an include trastuzumab targeting ERBB2 addiction8, and vemur- immortalized astrocyte culture. afenib targeting BRAF addiction9. The challenge is to find such 12q13-15 is frequently amplified in multiple tumor types key oncogenes. Even though large sets of genome and tran- including glioma24. AVIL is located on 12q14, 45 kb away from scriptome data are available to facilitate the identification of CDK4 toward centromere, and around 10 MB from MDM2 driver mutations in cancer, true signals are often buried in a large toward telomere. Given the importance of CDK4 and MDM2 in number of passenger events. In contrast to adult cancers, pedia- tumorigenesis, it is possible that AVIL amplification is only a tric tumors tend to have fewer point mutations and structural byproduct of a larger fragment amplification. However, AVIL changes. While studying a pediatric cancer, rhabdomyosarcoma, expression seems to be upregulated through additional mechan- we discovered a gene fusion, which results in the juxtaposition of isms: at the protein level AVIL was not detected in the normal a house-keeping gene next to the AVIL gene. Suspecting that astrocytes; in contrast, all of the tumor cell lines tested had higher other tumors may also dysregulate AVIL expression, we exam- AVIL protein expression, including the ones that do not have ined AVIL in adult cancers and found its critical role in the AVIL locus amplification (Fig. 1c). This difference was not tumorigenesis of GBM. We believe that the same approach can be obvious at the RNA level (Supplementary Fig. 2), suggesting that applied to the discovery of other oncogenes. some post-transcriptional mechanism may be at play. The cytoskeleton of the cells plays important roles in addition The overexpression of AVIL protein in GBMs was further to maintain the cell size and shape. Many critical processes confirmed by immunohistochemistry assays. Staining in normal including cell proliferation, migration, and even transcriptional human cerebral cortical gray matter was low, and largely regulations have been connected to the cytoskeleton10. Various restricted to a subpopulation of pyramidal neurons. Normal genes that modulate cytoskeleton have been associated with white matter demonstrated minimal immunoreactivity. In enhanced infiltrative and proliferative capacity11. For instance, in contrast, all the glioblastomas we tested had much stronger GBM, CTTN, an actin nucleating factor is overexpressed, and this signals (Fig. 1d, and Supplementary Fig. 3). overexpression is associated with an enhanced infiltrative capacity, We then examined the REMBRANDT database, which has and poor prognosis12,13. Here, we report an oncogene, AVIL, microarray data for 28 non-tumor brain tissues, 148 astrocytomas which encodes a protein that regulates F-actin dynamics and (WHO grade II or III), and 228 GBM (WHO grade IV) cases25. cytoskeleton. We found that AVIL is overexpressed in GBM cells Two different microarray probes showed that AVIL expression including GBM stem cells, and that AVIL overexpression is crucial correlates with tumor grade, with the highest levels in GBMs for GBM proliferation and migration. Mechanistically, AVIL (Fig. 1e). In our own collection of eight non-tumor brain tissues, functions upstream of FOXM1. FOXM1 is a member of FOX and 36 glioblastoma cases, we also confirmed the significant family. While it is silenced in differentiated cells, it is over- difference in AVIL RNA expression levels between the two groups expressed in a number of solid tumors including GBMs14. It has (Fig. 1f). To examine the expression levels of AVIL across the been also reported to mediated critical processes of tumorigenesis, major subclasses of glioma recently established and implemented such as tumor invasion, angiogenesis, and metastasis14–18.On in diagnostic neuropathology, we interrogate the RNA- the other hand, let-7 family of microRNAs functions as tumor sequencing dataset from the TCGA diffuse glioma study26. Here suppressors and inhibits glioma malignancy19. We showed mul- we subdivided the TCGA samples into GBMs with IDH-wild tiple lines of evidence supporting that AVIL regulates type, GBMs with IDH-mutant, Lower-grade glioma (LGG) with FOXM1 stability, which in turn regulates LIN28B/let-7. These IDH-wild type, LGG with mutant IDH and 1p/19q codeletion, findings support the critical role of cytoskeleton dynamics in and LGG with IDH mutant without 1p/19q codeletion. GBMs, and connect cytoskeleton regulation to the stability of Consistent with REMBRANT data, GBMs with or without IDH FOXM1 and let-7 expression. mutation have high level of AVIL expression. Within LGG, the group with IDH WT is the highest and not statically different from the GBMs. LGG IDH MT without 1p/19q codeletion is Results slightly higher than the group with codeletion (Fig. 1g). AVIL is frequently upregulated in glioblastomas. Previously, we We then examined the AVIL protein in 12 non-tumor brain identified a gene fusion in alveolar rhabdomyosarcoma, a tissues, and 14 GBMs. The AVIL protein was absent, or barely pediatric cancer20. We noticed that even though PAX3-FOXO1 is detected in the non-tumor cases, but higher levels of AVIL the most well-known fusion in this type of rhabdomyosarcoma, protein expression were seen in almost all of the GBM cases