IGFBP2 is a candidate biomarker for Ink4a-Arf status and a therapeutic target for high-grade gliomas

Lynette M. Moorea, Kristen M. Holmesa, Sarah M. Smitha, Ying Wua, Elena Tchougounovab, Lene Uhrbomb, Raymond Sawayac, Janet M. Brunera, Gregory N. Fullera,1, and Wei Zhanga,1

Departments of aPathology and cNeurosurgery, University of Texas MD Anderson Cancer Center, Houston, TX 77030; and bDepartment of Genetics and Pathology, Uppsala University, Rudbeck Laboratory, SE-751 85, Uppsala, Sweden

Edited by Webster K. Cavenee, University of California at San Diego, La Jolla, CA, and approved August 14, 2009 (received for review January 23, 2009) The levels of -like -binding 2 (IGFBP2) Genetic studies and genomic profiling of human gliomas have are elevated during progression of many human cancers. By using identified a number of common alterations, including frequent a glial-specific transgenic mouse system (RCAS/Ntv-a), we reported deletion of the INK4a-ARF locus, loss of PTEN, and overex- previously that IGFBP2 is an oncogenic factor for glioma progres- pression of IGFBP2, and these comprehensive studies have sion in combination with platelet-derived growth factor-␤ revealed valuable information regarding critical pathways used (PDGFB). Because the INK4a-ARF locus is often deleted in high- by gliomas (9, 10). IGFBP2 overexpression is correlated with grade gliomas (anaplastic oligodendroglioma and glioblastoma), high-grade gliomas and poor prognosis, and experiments from we investigated the effect of the Ink4a-Arf-null background on our laboratory have demonstrated that IGFBP2 is an oncogene IGFBP2-mediated progression of PDGFB-initiated oligodendrogli- that promotes glioma development and progression in a glial- oma. We demonstrate here that homozygous deletion of Ink4a-Arf specific mouse model (11–13). Further, IGFBP2 and PTEN ex- bypasses the requirement of exogenously introduced IGFBP2 for pression levels are negatively correlated in brain and prostate glioma progression. Instead, absence of Ink4a-Arf resulted in cancers, implicating IGFBP2 as a biomarker for PTEN status (14). elevated endogenous tumor cell IGFBP2. An inverse relationship In addition to PTEN, loss of Ink4a-Arf is also a frequent event between p16INK4a and IGFBP2 expression was also observed in in AO (3, 15). This locus encodes the p16INK4a and p14ARF human glioma tissue samples and in 90 different cancer cell lines (p19ARF in mouse) tumor-suppressor , which function as CELL BIOLOGY by using Western blotting and reverse-phase protein lysate arrays. critical regulators of the RB and p53 pathways (16). Loss of When endogenous IGFBP2 expression was attenuated by an RCAS Ink4a-Arf in a mouse model of PDGFB-initiated gliomas led to vector expressing antisense IGFBP2 in our mouse model, a de- an increase in tumor incidence compared with wild-type (17). creased incidence of anaplastic oligodendroglioma as well as Additionally, loss of Ink4a-Arf was found to cooperate with K-ras prolonged survival was observed. Thus, p16INK4a is a negative to induce GBM (18). In the present study, we examined whether regulator of the IGFBP2 oncogene. Loss of Ink4a-Arf results in loss of Ink4a-Arf is directly correlated with increased IGFBP2 increased IGFBP2, which contributes to glioma progression, expression in gliomas, and we functionally interrogated the thereby implicating IGFBP2 as a marker and potential therapeutic significance of Ink4a-Arf loss in IGFBP2-mediated glioma pro- target for Ink4a-Arf-deleted gliomas. gression by using the RCAS/Ntv-a mouse glioma model. We found that the expression of IGFBP2 was inversely liomas are the most common type of human primary brain correlated with both p16INK4a and ARF expression, as well as Gtumor. Surgical resection followed by radiation therapy with PTEN, in 90 different human cancer cell lines. We then (RT) and chemotherapy has been the most common treatment examined the relationship between Ink4a-Arf deletion and IG- regimen. In patients with anaplastic oligodendroglioma (AO), FBP2 expression in an RCAS/Ntv-a glial-specific mouse model. the standard historical treatment was surgery followed by RT, Transgenic mice expressing the avian virus receptor, tv-a, under with chemotherapy being reserved for recurrence (1). Despite the control of the Nestin promoter, which drives expression in treatment, overall survival for patients with AO is 3 to 5 years, glial progenitors, were crossed to Ink4a-Arf-null mice to generate Ϫ Ϫ whereas patients with the highest grade of glioma, glioblastoma the Ntv-a; Ink4a/ARF / mice. Somatic cell transfer with (GBM), have a survival rate of less than 1 year (1, 2). Genetic replication-competent avian leukemia virus with splice acceptor signatures have been useful for predicting prognosis and re- (RCAS) vectors allowed the specific delivery of oncogenes to the sponse to therapy. Chromosomal aberrations, especially code- glial progenitors of these mice (19). Consistent with the human letion of 1p and 19q in AO, have been associated with both data, we observed increased mouse IGFBP2 in glial progenitor chemotherapeutic response and increased overall and progres- cells from Ink4a/ARFϪ/Ϫ, Ink4aϪ/Ϫ, and ARFϪ/Ϫ mice. Delivery sion-free survival (3). The identification and characterization of and overexpression of PDGFB to the Ntv-a; Ink4a/ARFϪ/Ϫ a number of common genetic alterations in glial tumors have newborn brain led to development of AO without the cointro- incited considerable interest in the development of targeted duction of IGFBP2, which was required in the case of wild-type therapies. A number of these therapies have achieved consid- Ntv-a mice (11). Knockdown of endogenous IGFBP2 by anti- erable success in other cancers; for example, has been sense IGFBP2 reduced high-grade glioma incidence and pro- used to target the BCR-ABL fusion protein in chronic myelog- longed survival of tumor-bearing mice. These observations enous leukemia and to target KIT and PDGFR in gastrointes- tinal stromal tumors (4). Epidermal (EGFR) is commonly amplified in a wide spectrum of human Author contributions: R.S., J.M.B., G.N.F., and W.Z. designed research; L.M.M., K.M.H., S.M.S., Y.W., and E.T. performed research; L.U. contributed new reagents/analytic tools; cancers, including GBM, and is another example of targeted L.M.M., K.M.H., S.M.S., E.T., L.U., G.N.F., and W.Z. analyzed data; and L.M.M., J.M.B., G.N.F., drug development (5). However, anti-EGFR treatment has not and W.Z. wrote the paper. had the anticipated success in GBM clinical trials, despite The authors declare no conflict of interest. significant benefit in other neoplastic diseases, such as colon This article is a PNAS Direct Submission. cancer (6–8). The recent progress of targeted therapy in other 1To whom correspondence may be addressed. E-mail: [email protected] or tumor types highlights the need for identification of genetic and [email protected]. molecular alterations to serve as markers for patient stratifica- This article contains supporting information online at www.pnas.org/cgi/content/full/ tion and as targets for glioma therapeutics. 0900807106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900807106 PNAS Early Edition ͉ 1of5 Downloaded by guest on September 24, 2021 p16Ink4a/ A B WT p16Ink4a-/-ARF-/- ARF-/- p16INK4a XXXXBBBB IGFBP2 IGFBP2

Actin Actin 12345 678 9101112 13 14 15 16 17 1 0.9 21.3 18.7 38 22 29.3 26 C IGFBP2 vs. PTEN D IGFBP2 vs. p16INK4a 26.6% 27.8% 40 10 20

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 Fold Difference Fold Difference 0.0 27.8% 26.6% E IGFBP2 vs. p19ARF F IGFBP2 vs. IGFBP5 18.9% 10 5 3.3%

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5 3.3%

0.0 Fold Difference 0.0

Fold Difference 16.6%

Fig. 1. IGFBP2 expression is inversely related to p16INK4a expression. (A) Protein lysates from 17 human glioma samples were examined by Western blot analysis for IGFBP2 and p16INK4a protein levels. High-grade samples include grade III oligodendroglioma (lanes 4, 10, and 11) and astrocytomas (lanes 16 and 17) as well as GBM (lanes 1–3, 6–9, and 14–15). Samples 5, 12, and 13 were classified as grade II oligodendroglioma or astrocytoma and are noted in gray. (B) Western blot of Ntv-a glial progenitor cell lysates infected with empty RCAS vector (RCAS-X) or RCAS-PDGFB. IGFBP2 protein levels are elevated in cell lysates from Ink4a/ARF-deleted, Ink4a-deleted, and ARF-deleted mice. The numbers below the figure represent relative IGFBP2 levels after normalization to actin levels. (C–F) Fold difference in protein levels between IGFBP2 and PTEN (C), p16INK4a (D), p19ARF (E), and IGFBP5 (F). Protein levels in 90 cell lines were examined by protein lysate array, and fold differences were plotted. The red areas represent samples where IGFBP2 protein levels were greater than 2-fold higher than the compared protein; green areas represent samples where IGFBP2 protein levels were at least 2-fold lower. The percent of samples with greater than 2-fold differences are listed in each graph.

suggest that IGFBP2 may serve as both a marker and therapeutic colorectal carcinoma, ovarian cancer, and acute lymphoblastic target for Ink4a-Arf-deleted gliomas. leukemia (20–23). To further investigate this relationship, we mined data from Results our high-throughput reverse-phase protein lysate array analysis IGFBP2 Expression Is Inversely Correlated with p16INK4a Status. (24). This work examined 90 cell lines comprising 12 tumor types, IGFBP2 is frequently up-regulated in glioma and was identified including GBM, prostate cancer, breast cancer, and sarcoma. We recently as a potential biomarker for PTEN status and PI3K/Akt compared the relative protein levels of IGFBP2 with several pathway activation in both glioma and prostate cancer (14). We other proteins and determined the percentage of cell lines that sought to determine whether a similar relationship could be have a greater than 2-fold difference (Table 1). Among the proteins analyzed, an inverse relationship with IGFBP2 of identified between IGFBP2 and the frequently lost Ink4a/ARF INK4a locus. We examined 17 human glioma samples by Western blot greater than 2-fold was most frequent for PTEN and p16 , analysis for both IGFBP2 and p16INK4a (Fig. 1A). Fourteen confirming the results from the Western blot analysis of human glioma samples and literature reports (Fig. 1 C and D). The tumors were classified as high-grade gliomas (grade III) or GBM relationship between IGFBP2 and ARF was less frequent, (grade IV), and three were classified as low-grade (grade II) whereas IGFBP5 displayed a strong positive correlation with gliomas. We identified a strong inverse relationship between INK4a IGFBP2 (Fig. 1 E and F). elevated IGFBP2 levels and low or undetectable p16 . High The inverse correlation between p16INK4a and IGFBP2 was levels of IGFBP2 were evident in 12 of 14 high-grade tumors, as also apparent in primary glial progenitors isolated from the expected based on previous reports. Correspondingly, 11 tumors glioma mouse model. Glial progenitors were isolated from displayed no detectable p16INK4a protein. The three low-grade tumors (samples 5, 12, and 13) had low to undetectable IGFBP2 levels. Five tumors displayed detectable levels of p16INK4a, and Table 1. Protein-level relationships with IGFBP2 in four of these cases (samples 10, 12, 13, and 14), IGFBP2 levels % samples with were low or undetectable. Conversely, tumors displaying high Name Ͼ2-fold difference levels of IGFBP2 had no detectable p16INK4a. Therefore, simi- larly to published results for PTEN, IGFBP2 may serve as a PTEN 55.6 useful biomarker for p16INK4a-deleted tumors. This is of partic- p16INK4a 53.3 p19ARF 35.6 ular interest because elevated levels of IGFBP2 have been found IGFBP5 6.7 in the serum of patients with glioblastoma, prostate cancer,

2of5 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900807106 Moore et al. Downloaded by guest on September 24, 2021 Table 2. Tumor distribution in Ntv-a mice A Ntv-a WT* Ntv-a; Ink4a/ARFϪ/Ϫ PDGFB/

RCAS vectors delivered Grade II Grade III Grade II Grade III NI GFP PDGFB IGFBP2-AS

PDGFB 20 0 0 38 IGFBP2 PDGFB plus IGFBP2 17 11 3 42 PDGFB plus IGFBP2-AS N/A N/A 6 27 Actin 10.80.50.2 N/A, not applicable. B *Dunlap et al. (11). 100 PDGFB (n = 86) 80 PDGFB + IGFBP2-AS (n = 62) neonatal Ntv-a mice lacking Ink4a, Arf,orInk4a/Arf, and they were infected with either an empty RCAS virus (RCAS-X) or 60 RCAS-PDGFB. Western blot analysis revealed an increase in 40 ** p = 0.002 IGFBP2 protein levels in the glial progenitors lacking Ink4a, Arf,

or both (Fig. 1B). Percent survival 20

Increased IGFBP2 in PDGFB-Induced AO in Ntv-a; Ink4a/ARF؊/؊ Mice. 0 Previous studies from our laboratory and others demonstrated 0 20 40 60 80 100 120 Days that delivery of PDGFB to wild-type Ntv-a-expressing glial progenitors induced the formation of low-grade oligodendrogli- Fig. 3. Inhibition of IGFBP2 in PDGFB-induced gliomas confers a survival omas (11, 25). Codelivery of IGFBP2 with PDGFB promoted benefit. (A) Western blot of cell lysates demonstrates the effectiveness of glioma progression, resulting in a proportion of mice (38%) IGFBP2-AS to inhibit endogenous IGFBP2. Ntv-a; wild-type and Ntv-a; Ink4a/ Ϫ Ϫ developing AO (11). We sought to determine whether increased ARF / glial progenitors were infected with PDGFB or with PDGFB and endogenous IGFBP2 levels, observed in Ink4a/ARFϪ/Ϫ murine IGFBP2-AS. The numbers below the figure represent relative IGFBP2 levels after normalization to actin levels. NI indicates no infection. (B) Survival Ϫ Ϫ glial progenitors, could induce high-grade gliomas in the absence analysis of Ntv-a; Ink4a/ARF / mice infected with PDGFB (n ϭ 86) or PDGFB CELL BIOLOGY of exogenous IGFBP2 in the Ntv-a mouse model. Neonatal and IGFBP2-AS (n ϭ 62). Inhibition of endogenous IGFBP2 by IGFBP2-AS Ϫ Ϫ Ntv-a; Ink4a/ARF / mice were infected with RCAS-PDGFB confers a survival benefit on these animals (P ϭ 0.002). and were monitored for tumor formation. Delivery of PDGFB to Ntv-a; Ink4a/ARFϪ/Ϫ mice induced tumors in 97% of infected Ϫ/Ϫ animals, and they were exclusively classified as AO (Table 2). from PDGFB alone in Ink4a/ARF mice. These groups also These tumors were histopathologically similar to the AOs gen- displayed similar tumor incidence and grade (data not shown). erated by delivery of PDGFB and IGFBP2 to wild-type Ntv-a Immunohistochemical assay for IGFBP2 revealed significant Ϫ/Ϫ mice (11). The tumors were highly aggressive and displayed up-regulation in the PDGFB-driven Ntv-a; Ink4a/ARF tu- multiple foci of pseudopalisading necrosis and microvascular mors (Fig. 2 D–F). This expression represents endogenous protein because IGFBP2 was not delivered to this set of animals. proliferation, both of which are features of AO (Fig. 2 A–C). The Ϫ Ϫ Interestingly, the highest IGFBP2 expression was seen in the delivery of IGFBP2 with PDGFB in Ink4a/ARF / mice resulted in leading, invasive edge of the tumors (Fig. 2 D and F), consistent tumors that were histopathologically identical to those generated with a known function of IGFBP2 as an invasion-enhancing protein (26). Together, these data suggest that increased endog- enous IGFBP2 levels correlate with loss of Ink4a/ARF, and the resulting tumors are of higher histopathological grade.

Inhibiting IGFBP2 Expression in Ntv-a; Ink4a/ARF؊/؊ Mice Enhances Survival. After the experiments that clearly demonstrated that IGFBP2 overexpression was correlated with major tumor- suppressor gene loss and tumor progression, we sought to assess the effects of IGFBP2 inhibition on tumors that arise in Ntv-a; Ink4a/ARFϪ/Ϫ mice. We cloned an RCAS-IGFBP2 antisense construct (IGFBP2-AS) and validated it by infecting Ntv-a; Ink4a/ARFϪ/Ϫ primary glial progenitors. Protein lysates from the infected cells were examined by Western blot analysis for IGFBP2 protein level (Fig. 3A). IGFBP2-AS reduced the level Fig. 2. IGFBP2 is up-regulated in PDGFB-induced anaplastic oligodendrogli- of endogenous IGFBP2 protein in these cells by 60–80% com- omas in Ntv-a; Ink4a/ARF Ϫ/Ϫ mice. (A–C) H&E staining of representative PDGFB-induced tumors from Ntv-a; Ink4a/ARF Ϫ/Ϫ mice displaying features of pared with control infections. We next codelivered IGFBP2-AS Ϫ/Ϫ anaplastic oligodendroglioma. (A) Oligodendroglioma with regular, round and PDGFB to Ntv-a; Ink4a/ARF mice and monitored tumor nuclei surrounded by perinuclear halos. (B) Representative tumor displaying formation. The animals were killed upon visual observation, in pseudopalisading necrosis. (C) Microvascular proliferation in a representative a blinded fashion, of disease or discomfort. The tumor incidence tumor. (D–F) Tumors were examined for IGFBP2 by immunohistochemistry. (D) in this injection set was 100%. The tumors were histopatholog- Representative tumor demonstrating up-regulation of endogenous IGFBP2, ically analyzed and graded as described above. The majority of indicated by dark brown nuclear staining. Boxes represent regions photo- tumors (82%) that arose from the delivery of PDGFB and graphed at higher power in E and F.(E) Higher magnification of tumor bulk IGFBP2-AS were AO, similar to the tumors induced by PDGFB from the region surrounded by the left box in D. The tumor bulk shows sparse and weak IGFBP2 staining. (F) Higher magnification of leading edge of tumor alone, both of which featured brisk mitotic activity, foci of from the region surrounded by the right box in D, demonstrating high levels pseudopalisading necrosis, and microvascular proliferation to of IGFBP2 at the invasive edge of the tumor. (Magnification: A, E, and F, 400ϫ; varying degrees. A smaller fraction of the tumors were classified B and C, 100ϫ; and D,40ϫ.) as low-grade oligodendroglioma (18%). Kaplan–Meier analysis

Moore et al. PNAS Early Edition ͉ 3of5 Downloaded by guest on September 24, 2021 of the PDGFB- and IGFBP2-AS-infected animals compared emerges as a promising target for cancer therapeutics. As a with animals infected with PDGFB alone revealed that IG- proof-of-principle experiment, we introduced an antisense FBP2-AS conferred a survival advantage to these animals. IGFBP2, in combination with PDGFB, into the brain of Ink4a/ Survival analysis was performed on two independent cohorts of ARFϪ/Ϫ mice. The antisense construct used in these studies was animals, with similar results (Fig. S1). The combined results originally designed for the full-length human IGFBP2 cDNA. demonstrate that animals infected with IGFBP2-AS survived Despite this imperfect design, we still observed a modest de- significantly longer than those infected with PDGFB alone (P ϭ crease in IGFBP2 expression with this vector. Additionally, we 0.002; Fig. 3B). observed the beneficial effect of significantly prolonged survival. Glioma patients suffer the devastating effects of short survival Discussion after diagnosis, and little improvement has occurred during four Normal cellular processes are controlled by a complex network decades of research. Any improvement of survival constitutes an of positive and negative regulator interactions. During cancer advance and provides significant hope. The relationship between development, this network becomes deregulated, often revealing IGFBP2 and both p16INK4a and PTEN suggests that IGFBP2 the coordinated inactivation of tumor-suppressor together may be a critical point of convergence, and elevated IGFBP2 with activation of oncogenes that are critical for the transfor- levels provide insight into the status of two important tumor- mation process. Genetic approaches using animal models have suppressor genes. This information may provide useful information proven to be valuable in delineating the mechanistic relation- for new prognostic studies. Additionally, optimized design of anti- ships among these observed events. In human high-grade glio- sense constructs, siRNA molecules, or small-molecule inhibitors mas, deletion and inactivation of the tumor-suppressor genes could yield efficacious therapeutic agents directed against onco- PTEN and Ink4a/ARF are common events, although deletion of genic IGFBP2. The RCAS/Ntv-a mouse model is an ideal preclin- these genes alone does not cause development of glioma in ical model to test the effectiveness of such a strategy. transgenic mouse models. Activation and overexpression of oncogenes, such as PDGFR and IGFBP2, are common onco- Materials and Methods genic events in high-grade gliomas. Previous studies suggest that RCAS Constructs and DF-1 Cells. RCAS viral vectors encoding PDGFB and IGFBP2 PTEN deletion and overexpression of IGFBP2 are coordinated have been described (11, 25). The antisense IGFBP2 RCAS construct encodes events in high-grade glioma and prostate cancers (14). In this the full-length human cDNA in the reverse orientation. To produce viral study, by using a large cohort of glioma tissues and systemic particles, RCAS constructs were transfected into avian DF-1 fibroblasts by cancer cell lines, we demonstrated that IGFBP2 is also inversely calcium phosphate transfection. DF-1 cells were maintained in DMEM with INK4a correlated with p16 and ARF. One notable exception 10% FBS at 39 °C with 10% CO2. apparent in the human glioma tissues is sample 2. This GBM sample displays both high levels of IGFBP2 and detectable Mice. The Ntv-a and Ink4a/ARFϪ/Ϫ mouse lines have been described previously p16INK4a. However, p16INK4a/ARF is likely not the only regulator (29, 30). We established an Ntv-a; Ink4a/ARFϪ/Ϫ mouse model by crossing Ntv-a Ϫ Ϫ of IGFBP2, as suggested by the similar relationship between transgenic mice with the Ink4a/ARF / mouse line. For in vivo infections, DF-1 IGFBP2 and PTEN (14). Additionally, inactivating mutations of cells producing each RCAS virus were resuspended in sterile PBS. To infect pups INK4a with a combination of RCAS viruses, DF-1 cells producing each virus were p16 have also been identified in human glioma (27). By ϫ 4 ϫ 4 INK4a mixed before injection. For each pup, 2 10 to 5 10 cells in a volume no using genetic models, we have shown that p16 /ARF regu- more than 4 ␮L were injected intracranially in the right hemisphere of post- lates IGFBP2 by demonstrating that deletion of Ink4a/ARF, natal day 1 mice by using a gas-tight Hamilton syringe. Injected animals were Ink4a,orArf results in increased glial cell expression of IGFBP2. monitored daily for signs of tumor development and killed upon outward The regulatory relationship between loss of Ink4a/ARF and signs of disease or distress. All animal experiments were performed in accor- up-regulation of IGFBP2 has phenotypic consequences in tu- dance with the University of Texas MD Anderson Cancer Center Institutional mors. In the wild-type Ntv-a mouse model, PDGFB alone results Animal Care and Use Committee. in development of oligodendroglioma, and addition of IGFBP2 promotes progression to AO, which is characterized by activation Histopathology and Statistical Analysis. Mouse brains were formalin-fixed, of the Akt pathway and increased cellular proliferation (11). paraffin-embedded, and stained with H&E as described previously (11). Sec- However, in Ntv-a; Ink4a/ARFϪ/Ϫ mice, introduction of PDGFB tions were analyzed for the presence or absence of tumor and graded accord- alone is sufficient to produce AO. Addition of IGFBP2 did not ing to the World Health Organization 2007 standards for human gliomas. contribute to tumor phenotype, because endogenous IGFBP2 Survival analysis was performed by applying the log-rank test to the Kaplan– was already elevated secondary to loss of Ink4a/ARF. Therefore, Meier graphs generated in GraphPad Prism 5. this study demonstrates that a regulatory relationship exists Reverse-Phase Protein Lysate Arrays. Reverse-phase protein lysate arrays were between a tumor-suppressor gene and the oncogene IGFBP2. performed and analyzed as described previously (24). The 90 human cancer This previously uncharacterized relationship is likely more ubiq- cell lines include GBM, prostate, lung, and breast cancer lines. After normal- uitous because we reproducibly observed the same inverse ization to actin, the ratios of IGFBP2 protein levels to PTEN, p16INK4a, p19ARF, relationship in 90 human cancer cell lines derived from several and IGFBP5 were calculated and plotted. The percentage of samples with a major human cancer types. This is not surprising, because loss of greater than 2-fold difference in protein levels was calculated. Ink4a/ARF and overexpression of IGFBP2 are two of the most frequent events observed during cancer progression of almost all Immunohistochemistry. IGFBP2 immunohistochemistry was performed as de- cancer types (28). Additional investigation is needed to elucidate scribed previously (11). Briefly, antigen retrieval from deparaffinized slides the specific mechanism of IGFBP2 expression suppression by was achieved by heat denaturation (microwave steaming) with a citrate Ink4a/ARF. buffer target retrieval solution (DAKO Cytomation). IGFBP2 antibody was Correlative and genetic studies both show that IGFBP2 over- applied (1:250; C-10; Santa Cruz Biotechnology) and visualized by Dako En- ϩ expression represents a surrogate marker for Ink4a/ARF dele- Vision System-HRP. Slides were counterstained with hematoxylin. IGFBP2- positive cells were identified by dark brown nuclear staining. tion and inactivation in cancers. Whereas replenishing a missing tumor-suppressor gene via gene therapy has proven to be Murine Primary Brain Cell Culture. Primary murine glial progenitor cultures challenging, inhibition of an activated oncogenic signal appears were prepared as described previously (11). Conditioned media from the to be a more promising and tractable therapeutic strategy. respective RCAS-producing DF-1 cells were collected after 24 h, sterile-filtered IGFBP2 is commonly overexpressed in human cancers, and through 0.45-␮m filters, and added to the primary Ntv-a cell cultures. The overexpression of IGFBP2 has been linked to short survival in infection was repeated with conditioned media every day for 1 week. After many cancer types, including glioblastoma; therefore, IGFBP2 infection, cell lysates were collected for Western blot analysis.

4of5 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900807106 Moore et al. Downloaded by guest on September 24, 2021 Western Blot Analysis. Western blotting was performed as described previ- ACKNOWLEDGMENTS. We thank Hua Wang for previously constructing and ously (31). Antibodies against IGFBP2 (C-18; Santa Cruz Biotechnology) and characterizing the antisense IGFBP2 RCAS vector; David Cogdell and Cristian actin (Ab5; Thermo Scientific) were purchased and used to detect both human Mircean for help with lysate array data analysis; and Monica Linan for tech- and mouse proteins. p16INK4a (JC8; Santa Cruz Biotechnology) was purchased nical assistance. This work was partially supported by a National Institutes of Health (NIH) R01 grant (to W.Z. and G.N.F.); an Advanced Research Program and used to detect p16INK4a in human glioma samples. All human glioma Grant from the Texas Higher Education Coordinating Board; the Anthony samples were obtained from the University of Texas MD Anderson Cancer Bullock Brain Tumor Research Fund; the Goldhirsh Foundation; the Oreffice Center Brain Tumor Center tissue bank and were collected under an institu- Foundation; and James S. McDonnel Foundation; an NIH Roadmap Pharma- tional review board-approved protocol. Quantification of Western blots was coinformatics Training Grant (to K.M.H.); and an NIH Pharmacoinformatics performed by using Kodak ID imaging software. Training Fellowship and an American Legion Fellowship (to S.M.S.).

1. Collins VP (2007) Mechanisms of disease: Genetic predictors of response to treatment 18. Uhrbom L, et al. (2002) Ink4a-Arf loss cooperates with kras activation in astrocytes and in brain tumors. Nat Clin Prac Oncol 4:362–374. neural progenitors to generate glioblastomas of various morphologies depending on 2. Quon H, Abdulkarim B (2008) Adjuvant treatment of anaplastic oligodendrogliomas activated akt. Cancer Res 62:5551–5558. and oligoastrocytomas. Cochrane Database Syst Rev (2):CD007104. 19. Pao W, Klimstra DS, Fisher GH, Varmus HE (2003) Use of avian retroviral vectors to 3. Cairncross JG, et al. (1998) Specific genetic predictors of chemotherapeutic response introduce transcriptional regulators into mammalian cells for analyses of tumor main- and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst tenance. Proc Natl Acad Sci USA 100:8764–8769. 90:1473–1479. 20. Baron-Hay S, Boyle F, Ferrier A, Scott C (2004) Elevated serum insulin-like growth factor 4. Norden AD, Wen PY (2006) Glioma therapy in adults. Neurologist 12:279–292. binding protein-2 as a prognostic marker in patients with ovarian cancer. Clin Cancer 5. Grandis JR, Sok JC (2004) Signaling through the receptor Res 10:1796–1806. during the development of malignancy. Pharmacol Ther 102:37–46. 21. Cohen P, et al. (1993) Elevated levels of insulin-like growth factor-binding protein-2 in 6. Cunningham D, et al. (2004) monotherapy and cetuximab plus irinotecan the serum of prostate cancer patients. J Clin Endocrinol Metab 76:1031–1035. in irinotecan-refractory metastatic colorectal cancer. N Engl J Med 351:337–345. 7. Rich JN, et al. (2004) Phase II trial of gefitinib in recurrent glioblastoma. J Clin Oncol 22. Inman BA, et al. (2005) Insulin-like growth factor binding protein 2: An androgen- 22:133–142. dependent predictor of prostate cancer survival. Eur Urol 47:695–702. 8. Vogelbaum MA, et al. (2009) Phase II trial of preirradiation and concurrent temozo- 23. Vorwerk P, et al. (2005) Insulin-like growth factor binding protein-2 at diagnosis of lomide in patients with newly diagnosed anaplastic oligodendrogliomas and mixed childhood acute lymphoblastic leukemia and the prediction of relapse risk. J Clin anaplastic oligoastrocytomas: RTOG BR0131. Neuro Oncol 11:167–175. Endocrinol Metab 90:3022–3027. 9. Ohgaki H, Kleihues P (2007) Genetic pathways to primary and secondary glioblastoma. 24. Mendes KN, et al. (2007) Analysis of signaling pathways in 90 cancer cell lines by protein Am J Pathol 170:1445–1453. lysate array. J Proteome Res 6:2753–2767. 10. The Cancer Genome Atlas Research Network (2008) Comprehensive genomic charac- 25. Dai C, et al. (2001) PDGF autocrine stimulation dedifferentiates cultured astrocytes and terization defines human glioblastoma genes and core pathways. Nature 455:1061– induces oligodendrogliomas and oligoastrocytomas from neural progenitors and 1068. astrocytes in vivo. Genes Dev 15:1913–1925. CELL BIOLOGY 11. Dunlap SM, et al. (2007) Insulin-like growth factor binding protein 2 promotes glioma 26. Wang H, et al. (2003) Insulin-like growth factor binding protein 2 enhances glioblas- development and progression. Proc Natl Acad Sci USA 104:11736–11741. toma invasion by activating invasion-enhancing genes. Cancer Res 63:4315–4321. 12. Fuller GN, et al. (1999) Reactivation of insulin-like growth factor binding protein 2 27. Arap W, Knudsen ES, Wang JY, Cavenee WK, Huang HJ (1997) Point mutations can expression in glioblastoma multiforme: A revelation by parallel inactivate in vitro and in vivo activities of p16(Ink4a)/cdkn2a in human glioma. Onco- profiling. Cancer Res 59:4228–4232. gene 14:603–609. 13. Sallinen SL, et al. (2000) Identification of differentially expressed genes in human 28. Sharpless NE (2005) Ink4a/arf: A multifunctional tumor suppressor locus. Mutat Res gliomas by DNA microarray and tissue chip techniques. Cancer Res 60:6617–6622. 576:22–38. 14. Mehrian-Shai R, et al. (2007) Insulin growth factor-binding protein 2 is a candidate 29. Holland EC, Hively WP, DePinho RA, Varmus HE (1998) A constitutively active epidermal biomarker for PTEN status and PI3K/AKT pathway activation in glioblastoma and growth factor receptor cooperates with disruption of G1 cell-cycle arrest pathways to prostate cancer. Proc Natl Acad Sci USA 104:5563–5568. 15. Bigner SH, Rasheed BKA, Wiltshire R, McLendon RE (1999) Morphologic and molecular induce glioma-like lesions in mice. Genes Dev 12:3675–3685. genetic aspects of oligodendroglial neoplasms. Neuro Oncol 1:52–60. 30. Serrano M, et al. (1996) Role of the ink4a locus in tumor suppression and cell mortality. 16. Serrano M (2000) The Ink4a/arf locus in murine tumorigenesis. Carcinogenesis 21:865– Cell 85:27–37. 869. 31. Jabbur JR, et al. (2002) Mdm-2 binding and taf(II)31 recruitment is regulated by 17. Tchougounova E, et al. (2007) Loss of Arf causes tumor progression of PDGFB-induced hydrogen bond disruption between the p53 residues thr18 and asp21. Oncogene oligodendroglioma. Oncogene 26:6289–6296. 21:7100–7113.

Moore et al. PNAS Early Edition ͉ 5of5 Downloaded by guest on September 24, 2021