Analyses of GBM Oncogenome by GTS Identified CDKN2C Deletion
(A) Skyline profile of GBM oncogenome with ARI on y axis. AFI is represented by color scale based on percentile rank, highlighting the most focally altered regions.
(B) Two-dimensional GTS plots of log(ARI) (x axis) and log(AFI) (y axis) for amplification (left) and deletion (right). Black outer circles mark ROIs ranked among top 50 by GTS scores (Table 1). Red circles mark ROIs spanning signature events. ROIs on frequently gained chr7 are marked in blue on the amplification plot. Similarly, ROIs on chr10 are marked in blue on the deletion plot. Note separation of MET or PTEN from Chr7 or Chr10 events, respectively.
(C) Deletion ROI spanning CDKN2C is present in 4/28 glioblastoma tumor samples (two examples shown), defining a focal minimal common region of deletion of 2Mbp (dashed lines). Homozygous deletions are common in glioma cell lines (5/18) and further refine the minimal common region of deletion to p18^INK4C (solid lines). Analysis of copy-number-concordant expression by gene weight modeling (see Experimental Procedures) demonstrates highly significant coordinate loss of p18^INK4C expression in tumors and cell lines that show chromosomal deletion. Gene weights for both probe sets for p18^INK4C are significant at p < 0.01 (black circles). GW Sig, gene weight significance.

Acute Suppression of p16^INK4A Results in Compensatory Upregulation of p18^INK4C
(A) E2F transcription factors bind the p18^INK4C promoter in human astrocytes. A p18^INK4C promoter DNA fragment was immunoprecipitated with anti-E2F1, but not with the control or an RB1 antibody. (B) Human astrocytes were transiently transfected with empty vector or expression constructs for E2F1 and E2F5. INK4 and CCNE1 RNA expression was analyzed by RT-qPCR at 72 hr after transfection and normalized to GAPDH expression. Vector transfected cells were set to 1 for each transcript. E2F1, but not E2F5, specifically increased p18^INK4C (INK4C) RNA to a similar extent as the well-characterized E2F target CCNE1. Error bars represent mean ± standard deviation.
(C) Astrocytes from p53^−/− mice at passage 2 were transfected with nontargeting siRNA (siNT) or siRNA targeting p16^Ink4a (siInk4a) and assayed for p18^Ink4c protein expression at 72 hr after transfection. Loading control, vinculin.
(D) Murine p53^−/− astrocytes and normal primary human astrocytes were transfected with siNT or siInk4a and assayed for RNA expression levels of p16^Ink4a (Ink4a), p15^Ink4b (Ink4b), p18^Ink4c (Ink4c), and p19^Ink4d (Ink4d) by RT-qPCR at 72 hr posttransfection. Relative levels compared to Ink4 expression of siNT-transfected cells are shown after normalization to Gapdh. Error bars represent mean ± standard deviation.

p18^INK4C Inactivation in p16^INK4A Null GBM Cell Lines Enhances Tumorigenicity
(A) GBM cells, LN-229 (p16^INK4A/p18^INK4C null), LN-18, Hs683 (both p16^INK4A null, p18^INK4C WT), and LN-Z308 (p16^INK4A/p18^INK4C WT, CDK4/Cyclin D1-amplified) infected with retroviral expression constructs for small hairpins targeting p18^INK4C (shINK4C-2 and shINK4C-5) or a nontargeting hairpin (shNT) were scored for colony formation in soft agar at day 14 and plotted as percentage of shNT controls. Error bars represent standard deviation of duplicates. Suppression of p18^INK4C significantly increased the number of colonies in LN-18 (p = 0.01 and p = 0.006 for shINK4C-2 and shINK4C-5, respectively) and Hs683 (p = 0.07 and p = 0.004 for shINK4C-2 and shINK4C-5, respectively). Degrees of p18^INK4C knockdown were determined by RT-qPCR relative to shNT-infected cells (100%) and indicated in the bottom table. Error bars represent mean ± standard deviation.
(B) p16^INK4A/p18^INK4C null GBM lines U87MG and LN-229 as well as p16^INK4A/p18^INK4C WT, CDK4/Cyclin D1-amplified LN-Z308 stably infected with retroviral expression constructs for Flag-tagged p16^INK4A (INK4A^FLAG), Flag-tagged p18^INK4C (INK4C^FLAG), WT p18^INK4C (INK4C), or empty vector (Vector) were assayed for colony formation in soft agar. Soft agar colonies were stained and counted at day 14 and plotted as percentage ± standard deviation relative to vector-infected cells (U87MG-Vector 100%, U87MG-INK4C^FLAG 23 ± 2% [p = 0.0005], U87MG-INK4A^FLAG 18 ± 2% [p = 0.0003]; LN-229-Vector 100%, LN-229-INK4C^FLAG 23 ± 3% [p = 0.0002], LN-229-INK4A^FLAG 11 ± 2% [p=0.0001]). The scale bar indicates 2 mm.
(C) GBM cell lines LN-229 (p16^INK4A/p18^INK4C null), LN-444, LN-18 (both p16^INK4A null, p18^INK4C WT), and LN-Z308 (p16^INK4A/p18^INK4C WT, CDK4/Cyclin D1-amplified) infected with vector control (Vector), WT p18^INK4C (INK4C), or p18^INK4C variants (F37I and A61D) were assayed for soft-agar colony formation, as in (B). Both p18^INK4C variants exhibited reduced capability of repressing colony formation. Error bars represent mean ± standard deviation.
(D) Stable LN-229 cell populations were derived as in (C). Wild-type, but not mutant, p18^INK4C coimmunoprecipitated with CDK6 (p18^INK4C bands are marked by asterisks). The same effect was observed with binding to CDK4, albeit interaction between wild-type p18^INK4C and CDK4 was much weaker (data not shown).

p16^Ink4a and p18^Ink4c Coinactivation Confers Tumorigenicity to Murine Astrocytes
(A) Murine p16^Ink4a/ARF null astrocytes were infected with lentiviral shRNA expression constructs targeting GFP (shGFP) or p18^Ink4c (shInk4c). Knockdown of p18^Ink4c resulted in drastically lower p18^Ink4c protein levels (loading control, vinculin) and significantly higher colony formation in soft agar (error bars indicate mean ± standard deviation of quadruples).
(B) Stable astrocyte populations derived as in (A) were subcutaneously injected into Ncr nude mice. While p18^Ink4c-deprived cells formed tumors within 2 weeks, control cells did not generate tumor (mean ± standard deviation plotted). Injection sites of control mice were histologically confirmed to be tumor free.
(C) Tumors derived from shInk4c-infected astrocytes displayed malignant anaplastic histology and a high mitotic rate by H&E staining. Immunohistochemical analysis confirmed expression of the neural marker Nestin in a pattern similar to human astrocytomas. The scale bar indicates 30 μm.
(D) Matrigel-embedded shGFP-infected astrocytes and tumor-derived shInk4c-infected astrocytes were assayed for p18^Ink4c RNA expression relative to Gapdh expression confirming stable p18^Ink4c knockdown in tumor cells. Error bars indicate mean ± standard deviation.