|Home | About | Journals | Submit | Contact Us | Français|
Recent collaborative efforts have subclassified malignant glioblastomas into 4 clinical relevant subtypes based on their signature genetic lesions. Platelet-derived growth factor receptor α (PDGFRA) overexpression is concomitant with a loss of cyclin-dependent kinase inhibitor 2A (CDKN2A) locus (encoding P16INK4A and P14ARF) in a large number of tumors within one subtype of glioblastomas. Here we report that activation of PDGFRα conferred tumorigenicity to Ink4a/Arf-deficient mouse astrocytes and human glioma cells in the brain. Restoration of p16INK4a but not p19ARF suppressed PDGFRα-promoted glioma formation. Mechanistically, abrogation of signaling modules in PDGFRα that lost capacity to bind to SHP-2 or PI3K significantly diminished PDGFRα-promoted tumorigenesis. Furthermore, inhibition of SHP-2 by shRNAs or pharmacological inhibitors disrupted the interaction of PI3K with PDGFRα, suppressed downstream AKT/mTOR activation, and impaired tumorigenesis of Ink4a/Arf-null cells, whereas expression of an activated PI3K mutant rescued the effect of SHP-2 inhibition on tumorigenicity. PDGFRα and PDGF-A are co-expressed in clinical glioblastoma specimens, and such co-expression is linked with activation of SHP-2/AKT/mTOR signaling. Together, our data suggest that in glioblastomas with Ink4a/Arf deficiency, overexpressed PDGFRα promotes tumorigenesis through the PI3K/AKT/mTOR-mediated pathway regulated by SHP-2 activity. These findings functionally validate the genomic analysis of glioblastomas and identify SHP-2 as a potential target for treatment of glioblastomas.
Human gliomas account for the most common and malignant tumors in the CNS (1). Despite intensive treatments including maximal surgical resection, combined with radiotherapy and concurrent or adjuvant chemotherapies, median survival time of patients with high-grade glioblastoma multiforme (GBM) remains 14–16 months after diagnosis (1, 2). Recently, coordinated genomic analyses of a large cohort of clinical GBM specimens identified frequent co-alterations of genes in 3 core pathways — the P53, retinoblastoma (RB), and receptor tyrosine kinase (RTK) pathways — that are crucial in gliomagenesis (3). Notably, the gene encoding platelet-derived growth factor receptor α (PDGFRα) is amplified in approximately 13% of total GBMs analyzed, and deletion of the tumor suppressor cyclin-dependent kinase inhibitor 2A (CDKN2A) locus is frequently found in these PDGFRα-amplified GBMs (3, 4). However, to our knowledge, studies have not been conducted to determine whether these genetic aberrations act in concert to promote gliomagenesis, or to determine the underlying mechanisms of PDGFRα-stimulated tumorigenesis.
The homozygous deletion of the CDKN2A locus at chromosome 9p21, which eliminates both INK4A and ARF genes (encoding P16INK4A and P14ARF, respectively), concomitantly deregulates 2 of the major tumor suppressor pathways, the RB and P53 pathways (5). Mice lacking the Cdkn2a locus develop spontaneous tumors and are more prone to carcinogenesis (6). Whereas no spontaneous tumors are found in the brain of Ink4a/Arf-deficient mice, astrocytes and neural stem cells (NSCs) from these mice form high-grade gliomas in the brain upon EGFR activation (7). Additionally, K-Ras activation has been shown to cooperate with loss of Ink4a and Arf in mouse astrocytes (mAsts) or neural progenitor cells to generate GBMs (8).
PDGFRα is a RTK that elicits a variety of biological activities such as cell proliferation and migration via stimulation by its ligand dimers, PDGF-AA, -AB, -BB, -CC, and -DD. Activated PDGFRα associates via the autophosphorylated tyrosine (p-Y) residues at its cytoplasmic domain to various downstream SH2 domain–containing signaling molecules, including SRC family kinases (SFKs), phosphotyrosine phosphatase SHP-2, PI3K, and PLCγ (9, 10). Mice deficient in PDGFRα or PDGF-A or engineered to separately express PDGFRα with mutations of the individual p-Y sites show alterations in cellular behavior and embryo development (9, 11). In particular, PI3K has been identified as the major effector of PDGFRα signaling in vitro and in vivo (10, 12, 13). SFKs and PLCγ contribute to some but not all PDGFRα functions (10, 12–14), whereas SHP-2 is not required for cell survival during Xenopus embryogenesis (12). However, contributions of each of these signaling modules to glioma formation have not to our knowledge been demonstrated.
Activation of PDGFRα signaling has been observed in human gliomas. In clinical glioma specimens, PDGFRα and PDGF-A are overexpressed in tumor cells, while PDGFRβ is only expressed in endothelial and peri-endothelial compartments (15). PDGF-B, which binds to both PDGFRα and PDGFRβ, is an oncoprotein that causes glioma formation in the brain (16). Loss of Ink4a/Arf in mAsts enhances PDGF-B–initiated gliomagenesis and tumor progression in the brain (16). Specific activation of PDGFRα signaling by infusion of PDGF-A proteins, which only bind to PDGFRα in PDGFRα-positive type B NSCs in the subventricular zone (SVZ), leads to glioma-like growth of these cells in adult brain (17). However, how the activation of PDGFRα signaling causes glioma formation and whether co-alteration of tumor suppressor pathways is required in PDGFRα-mediated gliomagenesis have not been directly shown. In this study, we determined the synergistic impact of PDGFRA overexpression and INK4A/ARF deletion on gliomagenesis in mAsts and human glioma cells. We then used genetic and pharmacological approaches targeting individual downstream signaling enzymes to examine which specific signaling pathway(s) emanating from PDGFRα are critical in tumorigenesis. Furthermore, we confirmed our observations in clinical glioma specimens that co-overexpress PDGFRα and PDGF-A.
To determine whether activation of PDGFRα signaling in Ink4a/Arf–/– mAsts leads to gliomagenesis in the brain, we overexpressed PDGFRα and/or its ligand PDGF-A chain in a retroviral vector containing an IRES-GFP in Ink4a/Arf–/– mAsts (Figure (Figure1A).1A). The impact of PDGFRα/PDGF-A overexpression on the growth and survival of these cells in vitro was moderate (Supplemental Figure 1, A and B; supplemental material available online with this article; doi: 10.1172/JCI43690DS1). However, when these cells were separately transplanted into the flanks of mice, significant s.c. tumor growth was evident in mice that separately received Ink4a/Arf–/– mAsts expressing PDGFRα, PDGF-A, or PDGFRα/PDGF-A (with the receptor and ligand co-expressed in the same cell to create an autocrine signaling), whereas minimal or no tumor formation was seen in control mice that received Ink4a/Arf–/– mAsts expressing GFP (Figure (Figure1B).1B). To determine tumorigenicity of these cells in the brain, various mAsts were separately implanted into the brain of mice. Notably, mice that received GFP mAsts did not show active tumor growth up to 42 days after implantation, while PDGFRα, PDGF-A, and PDGFRα/PDGF-A mAsts started to form tumors in the brain as early as 8–11 days and the majority of tumors reached a volume of approximately 25 to 30 mm3 in 25–35 days. Moreover, mice that separately received PDGFRα- or PDGF-A–expressing mAsts survived up to 2 months after implantation. On the other hand, all mice that received autocrine PDGFRα/PDGF-A–expressing mAsts developed large and invasive tumors by 20 days, with an average survival time of 25 days after implantation. As shown in Figure Figure1,1, C–L, significantly larger and highly invasive gliomas formed in the brains of mice that received Ink4a/Arf–/– mAsts overexpressing PDGFRα, PDGF-A, or PDGFRα/PDGF-A, whereas only small tumor lesions were found in the brain of mice that received control mAsts. Significantly, an approximately 10-fold increase in the cell proliferation index was found in gliomas derived from PDGFRα-activated mAsts compared with control tumors (Figure (Figure1,1, M–R), whereas an approximately 10-fold decrease of cell apoptosis was seen in PDGFRα/PDGF-A–expressing tumors (Figure (Figure1S).1S). Of note, in established s.c. or brain tumors, exogenous expression of PDGFRα or PDGF-A was maintained at the end of the experiments (Supplemental Figure 1, C–K, and data not shown). Taken together, these results indicate that expression of PDGFRα and/or PDGF-A confers tumorigenicity to Ink4a/Arf-deficient mAsts in the brain.
To further characterize the tumors derived from Ink4a/Arf–/– PDGFRα mAsts, we examined molecular markers of various cell lineages in the CNS development. As shown in Figure Figure2,2, tumors derived from PDGFRα-expressing mAsts were highly positive for the neural progenitor marker nestin, which was distributed along the processes of individual cells (Figure (Figure2,2, A and B). As expected, most of tumor cells showed expression of the progenitor/mature astrocyte marker GFAP (Figure (Figure2,2, C and D) but were negative for the neuronal marker class III β-tubulin (TUJ1) (Figure (Figure2,2, E and F). Significantly, NG2, an oligodendrocyte progenitor cell marker (18), was expressed in a population of tumor cells that were actively invading the surrounding brain parenchyma, whereas the core of the tumor mass showed relatively low NG2 expression (Figure (Figure2,2, G and H). In contrast, control tumors derived from Ink4a/Arf–/– GFP mAsts lacked nestin or NG2 expression (data not shown). In addition, tumors derived from Ink4a/Arf–/– PDGFRα mAsts showed negative staining for oligodendrocyte type 2 astrocyte progenitor marker A2B5 and an oligodendrocyte marker CNPase (data not shown). Notably, brain tumors derived from Ink4a/Arf–/– PDGFRα/PDGF-A mAsts (data not shown) displayed similar IHC features to tumors derived from Ink4a/Arf–/– PDGFRα mAsts, suggesting that similar dedifferentiation events also occurred in autocrine PDGFRα/PDGF-A–co-expressing tumors.
Next, we used 4 human glioma cell lines that were either INK4A/ARF null (LN444 and LN443) or WT INK4A/ARF (LN-Z308 and LN319; Figure Figure3A)3A) (19) and exogenously expressed PDGF-A in these glioma cells that express endogenous PDGFRα (Figure (Figure3B).3B). In vitro, expression of PDGF-A in INK4A/ARF-null LN444 cells significantly enhanced their capacity of anchorage-independent growth in soft agar, whereas a minimal effect was seen in WT INK4A/ARF LN-Z308 and LN319 cells (Figure (Figure3C).3C). When various glioma cells were separately implanted into the brain or the flanks of mice, PDGF-A expression markedly enhanced tumor growth and invasion of INK4A/ARF-null LN444 and LN443 glioma cells, while minimal impact of PDGF-A expression on tumorigenesis or invasion was seen in WT INK4A/ARF LN-Z308 and LN319 cells in both anatomic sites (Figure (Figure3,3, D–K, and Supplemental Figure 2). Mice received LN444/PDGF-A cells in the brain had tumor onset as early as 35 days, while mice with LN443/PDGF-A cells developed invasive intracranial tumor 25–30 days after implantation. In contrast, LN444 and LN443 parental cells only formed small tumor lesions in the brain up to 2–3 months following implantation. Mice that received LN444/PDGF-A cells survived for 75–80 days, while most of the mice that received LN443/PDGF-A cells lived up to 3 months after implantation. On the other hand, no significant brain tumor growth was found 2–3 months after implantation in mice that received LN-Z308/PDGF-A or LN319/PDGF-A cells. Thus, consistent with our findings in the Ink4a/Arf–/– mAsts, these data indicate that PDGFRα/PDGF-A signaling enhances in vivo tumor growth and invasion of human glioma cells deficient in INK4A/ARF.
To investigate whether re-expression p16INK4a or p19ARF is able to abrogate the enhanced tumorigenicity in Ink4a/Arf–/– mAsts, we separately expressed these 2 tumor suppressors in mAsts (Figure (Figure4A).4A). Surprisingly, re-expression of p16INK4a but not p19ARF alone in Ink4a/Arf–/– mAsts suppressed soft agar growth of PDGFRα-expressing mAsts stimulated by PDGF-A (Figure (Figure4B).4B). Consistently, enforced restoration of p16INK4a in Ink4a/Arf–/– mAsts significantly inhibited tumorigenesis of PDGFRα/PDGF-A–overexpressing mAsts in the mouse brain (Figure (Figure4C).4C). To further demonstrate the cooperative effect of PDGFRα activation and loss of Ink4a on tumorigenesis, we knocked down endogenous p16INK4a using shRNAs targeting CDKN2A in WT INK4A/ARF LN319 cells. As shown in Figure Figure4D,4D, significant inhibition of P16INK4A without affecting P14ARF expression by 2 separate shRNAs in LN319 cells resulted in a marked increase in colony formation in soft agar. Since p16INK4a inhibits CDK4/6, which inactivates RB by phosphorylation, and since p19ARF (or P14ARF) targets MDM2, which suppresses p53 (5), we thus examined responses of modulating CDK4/6 and p53 in these cells. As shown in Figure Figure4E,4E, inhibition of phosphorylation of RB, the direct downstream target of CDK4/6 by CDK4/6 inhibitor PD0332991 (20), markedly reduced PDGF-A–stimulated cell growth of Ink4a/Arf-deficient mAsts and LN444 cells in soft agar, which suggests that the tumorigenic effect of PDGFRα signaling is dependent on CDK4/6 inactivation of RB proteins (p-RB). Since we did not observe any tumor-suppressing effect of p19ARF on Ink4a/Arf–/– PDGFRα-expressing mAsts (Figure (Figure4B),4B), we asked whether the downstream p53 protein was functional in these cells. When Ink4a/Arf–/– and Ink4a/Arf–/– p19ARF mAsts were treated with cisplatin (21), p53 expression was strongly induced in Ink4a/Arf–/– PDGFRα/p19ARF-expressing mAsts but only moderately induced in Ink4a/Arf–/– PDGFRα-expressing mAsts (Supplemental Figure 3A). The mAsts expressing p19ARF were more sensitive to cisplatin inhibition of cell survival than those without p19ARF expression, indicating that p19ARF-upregulated p53 in Ink4a/Arf–/– PDGFRα-expressing mAsts rendered sensitivity to cisplatin inhibition (Supplemental Figure 3B). Collectively, these data show that p16INK4a but not p19ARF suppresses the tumorigenesis promoted by PDGFRα/PDGF-A signaling, suggesting a cooperative effect of PDGFRα activation and p16INK4a inhibition during gliomagenesis. Additionally, p19ARF loss might be required for tumor survival in certain circumstances such as in the presence of DNA-damaging agents.
Previous studies using genetic and biochemical approaches have defined the roles of signaling molecules in PDGFRα-mediated cellular functions by specific tyrosine-to-phenylalanine (Y-to-F) mutations (Figure (Figure5A)5A) (10, 12, 13). To investigate the impact of these PDGFRα mutations on tumorigenesis, we separately expressed WT PDGFRα or various PDGFRα mutants in Ink4a/Arf–/– mAsts. The PDGFRα mutant R627 (PDGFRα-R627) that harbors a lysine-to-arginine (K-to-R) mutation was included as a “receptor kinase–dead” control. As shown in Figure Figure5B,5B, stimulation of WT PDGFRα by PDGF-A resulted in autophosphorylation of the receptor and promoted phosphorylation of downstream signaling molecules Erk1/2 and Akt, whereas there was an undetectable receptor tyrosine autophosphorylation in PDGFRα-R627 (Figure (Figure5B).5B). PDGFRα Y-to-F mutations at Y572 and Y574 (PDGFRα-F572/74; for SFK binding), Y1018 (PDGFRα-F1018; for PLCγ binding), and Y988 (PDGFRα-F988) did not result in a significant decrease in p-Akt and p-Erk levels in response to PDGF-A stimulation (Figure (Figure5B).5B). Moreover, the mutation at Y731 and Y742 (PDGFRα-F731/42; for PI3K binding) led to a marked decrease in PDGF-A–stimulated p-Akt and p-Erk1/2 levels (Figure (Figure5B).5B). In agreement with a previous report (22), p-Erk1/2 was markedly reduced in PDGF-A–stimulated mAsts expressing PDGFRα-F720 (Y-to-F mutation at Y720; for SHP-2 binding), compared with WT PDGFRα mAsts (Figure (Figure5B).5B). Interestingly, p-Akt levels were also attenuated in PDGFRα-F720 cells.
Next, we determined the impact of these Y-to-F mutations on PDGFRα-promoted cell transformation of Ink4a/Arf–/– mAsts in vitro. As shown in Figure Figure5C,5C, mAsts expressing WT PDGFRα and PDGFRα-F572/74, PDGFRα-F988, and PDGFRα-F1018 PDGFRα mutants showed similar capability of anchorage-independent growth in soft agar. In contrast, expression of PDGFRα-R627, PDGFRα-F731/42, or PDGFRα-F720 instead of WT PDGFRα significantly abrogated the capacity of mAsts to form colonies in soft agar (Figure (Figure5,5, C and D), indicating that PI3K and SHP-2 signaling were critical for cell transformation in vitro. When various Ink4a/Arf–/– mAsts were separately implanted into the brains of mice, PDGFRα-R627, PDGFRα-F731/42, PDGFRα-F720, and PDGFRα-F988 significantly impaired PDGFRα-promoted tumorigenesis and invasion compared with WT PDGFRα (Figure (Figure6,6, B, C, E, and F, compared with Figure Figure6A).6A). However, mAsts expressing PDGFRα-F572/74 or PDGFRα-F1018 displayed tumorigenicity comparable to WT PDGFRα tumors (Figure (Figure6,6, D and G, compared to 6A), but showed markedly reduced tumor invasion compared with the tumors derived from mAsts expressing WT PDGFRα (Figure (Figure6,6, K and N, compared with 6H). Higher-magnification images of PDGFRα-R627, PDGFRα-F731/42, and PDGFRα-F720 mutants (Figure (Figure6,6, I, J, and L) further illustrated that these mutations significantly negated the enhanced tumorigenicity conferred by PDGFRα activation, leading to the formation of tumors similar in volume and invasiveness to those seen in GFP-expressing mAst tumors (Figure (Figure1,1, D and I). Moreover, PDGFRα-F731/42 and PDGFRα-F720 tumors showed a significantly decreased cell proliferation (Figure (Figure6,6, O and P) and increased apoptosis (Figure (Figure6,6, Q and R), whereas PDGFRα-F572/74, PDGFRα-F988, and PDGFRα-F1018 tumors exhibited moderate or minimal impacts on cell proliferation and survival compared with WT PDGFRα tumors. Conversely, retention of any 1 of the 5 signaling modules (Y731/42, Y572/74, Y720, Y988, and Y1018 PDGFRα mutants) (10) was insufficient to rescue the abolished tumorigenesis in the brain (data not shown). Taken together, these data indicate that ablation of PDGFRα association with SHP-2 or PI3K abrogates tumorigenicity of Ink4a/Arf–/– mAsts expressing PDGFRα, whereas individual association of PI3K, SHP-2, SFK, or PLCγ with the RTK was insufficient to elicit the full spectrum of tumor-promoting effects of PDGFRα.
We further investigated whether inhibition of SHP-2 and PI3K activities by pharmacological approaches suppresses tumorigenicity of PDGFRα/PDGF-A–expressing Ink4a/Arf–/– mAst and LN444 cells. To this end, we exploited several pharmacological inhibitors against PI3K (LY294002), SHP-2 (PHPS-1 and NSC87877) (23), and SFKs (SU6656, and PP2). As shown in Figure Figure7,7, LY294002 inhibited PDGF-A–induced phosphorylation of Akt at 10 μM in mAsts and 5 μM in LN444 glioma cells, whereas Erk1/2 phosphorylation was unaffected by LY294002 treatment in both types of cells (Figure (Figure7,7, A and B). Of note, a modest decrease was observed in p-Erk1/2 levels in mAsts treated with 20 μM LY294002 (Figure (Figure7A).7A). At a concentration of 100 μM (23), both PHPS-1 and NSC87877 significantly inhibited p-Erk1/2, a direct downstream target of SHP-2, in mAsts and LN444 cells (Figure (Figure7,7, A and B). Next, we examined the impact of the pharmacological interventions of these signaling enzymes on in vitro cell transformation and found that 10 μM LY294002, which had no effect on Erk1/2 activation, suppressed soft agar growth of PDGFRα-expressing mAsts stimulated by PDGF-A (Figure (Figure7A)7A) and PDGF-A–expressing LN444 cells (Figure (Figure7B).7B). Similarly, 100 μM of either PHPS-1 or NSC87877 ablated PDGFRα-stimulated, anchorage-independent growth in soft agar but only had a minimal effect on the cell survival and caspase-3 activation of both PDGFRα/PDGF-A–expressing mAsts and LN444 cells in culture (Supplemental Figure 4, A and B). In contrast, 2 SFK inhibitors, SU6656 and PP2 each at 5-μM concentration, only showed a modest impact on tumorigenesis of these cells (Figure (Figure7,7, A and B). Additionally, the MEK inhibitor PD98059 at 10 μM, which suppressed PDGF-A–induced p-Erk1/2 in mAsts (Supplemental Figure 4C), also inhibited soft agar growth of these cells (Figure (Figure5A),5A), validating Erk1/2 as a mediator of PDGFRα–SHP-2 signaling. Taken together, these data demonstrate that the PDGFRα/PDGF-A–enhanced tumorigenicity of Ink4a/Arf-deficient mAsts and human glioma cells requires intact SHP-2 and PI3K enzymatic activities.
We then examined the impact of SHP-2 inhibitors (PHPS-1 and NSC87877) and PDGFRα-F720 on PI3K signaling. As shown in Figure Figure5B5B and Figure Figure8A,8A, PDGFRα-F720 mutant and SHP-2 inhibitors significantly decreased PDGF-A–stimulated p-Akt in PDGFRα-expressing mAsts. A nearly complete knockdown of SHP-2 by siRNAs markedly reduced the stimulated p-Akt levels in these cells, possibly through interruption of the association between PI3K (p85 subunit) and PDGFRα (Figure (Figure8B).8B). However, a previous study showed that PDGFRα-F720 mutation did not result in a decrease in PI3K association with the RTK in mouse fibroblasts (10). Similarly, when SHP-2 was knocked down in NIH3T3 cells, a minimal impact of SHP-2 inhibition on PDGF-A–induced PI3K association and Akt phosphorylation was observed in these fibroblasts (Supplemental Figure 5A), suggesting that the impact of SHP-2 on PI3K association with the RTK and p-Akt in astrocytes and glioma cells was specific. To examine whether the effect of SHP-2 knockdown on p-Akt signaling was specific to PDGFRα signaling, we knocked down SHP-2 by siRNAs in EGFRvIII-expressing Ink4a/Arf–/– mAsts (WT PTEN) and LN444 and U87MG (both PTEN mutant) cells. We observed a reduction or a modest impact of p-AKT level in LN444/EGFRvIII and U87MG/EGFRvIII cells, respectively (Supplemental Figure 5B), suggesting that SHP-2 also regulates PI3K/AKT activation in other RTK signaling. However, Akt phosphorylation was absent in Ink4a/Arf–/– EGFRvIII mAsts, possibly due to the presence of WT Pten in these cells.
We next tested whether signaling downstream to Akt was affected by SHP-2 inhibition. We found that SHP-2 inhibitors (Figure (Figure8A),8A), SHP-2 siRNA knockdown (Figure (Figure8B),8B), or PDGFRα-F731/42 and PDGFRα-F720 mutations (Figure (Figure8C)8C) significantly impaired PDGF-A–stimulated phosphorylation of S6 kinase (p-S6K) downstream to the mammalian target of rapamycin (mTOR) pathway. Thus we further determined the impact of the ablation of SHP-2 and its downstream target the mTOR pathway on cell transformation. Strikingly, stable knockdown of SHP-2 by shRNAs markedly reduced PDGFRα-promoted soft agar growth of mAsts (Figure (Figure8D).8D). Additionally, treatment with mTOR complex 1 (mTORC1) inhibitor rapamycin led to a dose-dependent decrease in soft agar colonies formed by PDGF-A–stimulated Ink4a/Arf–/– mAsts expressing PDGFRα or LN444 cells (Supplemental Figure 6, A and B), compared with their respective DMSO-treated controls. Similarly, rapamycin was able to effectively suppress the cell transformation capacity conferred by EGFRvIII expression in these cells (Supplemental Figure 6C). Previous studies suggested that the limited efficacy of rapamycin in clinical use was due to the capacity of rapamycin to potentiate PI3K/Akt signaling (24). However, we did not observe an increase of p-Akt level in both Ink4a/Arf-deficient mAsts and LN444 cells treated with rapamycin up to 72 hours (Supplemental Figure 6D). These results suggest that PDGFRα-promoted tumorigenesis necessitates an intact SHP-2 activity that regulates the PI3K/AKT/mTOR pathway. Since, in clinical glioblastomas, constitutively active PI3K mutations are mostly observed in specimens with no PDGFRA aberrations (3, 4), we introduced a constitutively active PI3K p110 subunit (p110α-CAAX) into Ink4a/Arf–/– mAsts that either expressed PDGFRα-F720 or an SHP-2 shRNA. In both cell lines, we observed a rescue effect of p110α-CAAX in soft agar colony formation (Figure (Figure8E),8E), suggesting that PI3K/AKT/mTOR acts downstream of SHP-2 in PDGFRα-overexpressing gliomas, and that activating PI3K mutations in clinical gliomas might be able to bypass SHP-2 activation to promote tumorigenesis. Taken together, our results suggest that SHP-2 regulates the PI3K/AKT/mTOR signaling emanating from PDGFRα activation.
To determine whether PDGF-A and PDGFRα are co-expressed and whether there is a link of overexpression of PDGF-A and PDGFRα and activation of the SHP-2 and the PI3K/AKT/mTOR signaling in clinical glioma specimens, we performed IHC staining on a total of 158 paraffin-embedded primary human glioma specimens using anti-PDGFRα, anti–PDGF-A, anti–p–SHP-2 Tyr542, anti–p-AKT Ser473, and anti–p-S6 Ser235/236 antibodies. PDGFRα proteins were detected at medium to high levels in 17 of 87 GBMs (WHO grade IV tumors) and in 9 of 71 grade II and III tumors. Among these PDGFRα-positive gliomas, PDGF-A was often co-expressed in the same population of tumor cells (Figure (Figure9,9, B and C, and Supplemental Table 1). Significantly, phosphorylation of SHP-2 Y542 (p–SHP-2, required for activation of SHP-2) (25), AKT (p-AKT), and ribosomal S6 subunit (p-S6) was also often detected on sister sections of the same tumor in many of the PDGF-A/PDGFRα–positive glioma specimens (Figure (Figure9,9, B–F, and Supplemental Table 1), suggesting a link between activation of PDGF-A/PDGFRα and stimulation of SHP-2, AKT, and mTOR in clinical glioblastoma specimens. To validate these data, we performed IB analyses on a separate cohort of a total of 20 snap-frozen clinical GBM specimens. As shown in Figure Figure9G,9G, PDGFRα was expressed at high levels in 7 of 20 GBM samples, 5 of which expressed PDGF-A proteins, suggesting an autocrine PDGFRα signaling in these tumors. In 4 of these 5 PDGF-A/PDGFRα–positive tumors, p-AKT, p–SHP-2, and p-S6 were also detected. Of note, expression of PDGF-A, p-AKT, p–SHP-2, and p-S6 at various levels in other tumor samples likely reflects the impact of heterogeneous gene alterations such as mutations of PTEN, TP53, and overexpression of EGFR/EGFRvIII and c-MET that affect the expression or activation (phosphorylation) of these proteins in clinical glioblastomas. Taken together, these results establish a link of PDGF-A/PDGFRα expression with activation of SHP-2 and PI3K/AKT/mTOR signaling in clinical glioblastoma samples.
In this study, we report that PDGFRα and/or PDGF-A overexpression is able to drive gliomagenesis of Ink4a/Arf-deficient mAsts and human glioma LN444 and LN443 cells. Re-introduction of p16INK4a but not p19ARF into Ink4a/Arf-null mAsts suppresses PDGFRα-promoted tumor growth. In the absence of PI3K or SHP-2 signaling, PDGFRα fails to enhance tumorigenesis in the brain of mice. Additionally, we establish a link between activation of SHP-2 and the PI3K/AKT/mTOR signaling in PDGFRα-stimulated tumorigenesis in vitro, in mice, and in clinical glioblastoma specimens. Therefore, our data demonstrate that co-alteration of the RTK PDGFRα and tumor suppressor p16INK4a is required for gliomagenesis and that SHP-2 is a critical linker between the PI3K/AKT/mTOR pathway and PDGFRα in the formation of gliomas.
A unique feature of this study is that specific activation of PDGFRα signaling in vivo by PDGF-A, a ligand that binds to PDGFRα but not PDGFRβ (14) as an autocrine loop significantly enhanced the tumorigenesis of Ink4a/Arf-deficient mAsts and human glioma cells in the brain. Early studies of clinical glioma specimens showed that PDGF-A and PDGFRα are overexpressed in tumor cells, while PDGF-B and PDGFRβ are expressed in hyperplastic capillaries, suggesting both autocrine and paracrine loops for PDGF/PDGFR activation in gliomas (15). In neonate and adult mice, expression of PDGF-B induces de novo gliomas from GFAP-positive astrocytes and nestin-expressing glial progenitor cells through activation of PDGFRα and PDGFRβ in the brain of both WT and Ink4a/Arf-deficient animals (26, 27). Moreover, in WT Ink4a/Arf mice, infusion of PDGF-A proteins into the lateral ventricle stimulated tumor-like growth of PDGFRα-positive NSCs in the SVZ in the brain (17). Importantly, data from The Cancer Genome Atlas (TCGA) and other studies revealed that PDGFRα is overexpressed and amplified and often co-expressed with PDGF-A in clinical glioblastoma samples (3, 28, 29). Our results not only functionally validated these studies but also further demonstrated the significance of specific activation of PDGFRα signaling by PDGF-A in cooperation with loss of p16INK4a but not p19ARF in promoting gliomagenesis. We found that when PDGFRα signaling is activated in Ink4a/Arf–/– mAsts or human glioma cells, mice that received these cells developed significantly larger and highly invasive tumors in the brain. In contrast, no enhancement of tumorigenesis was found in mice that received glioma cells with PDGFRα and an intact CDKN2A locus (19). Together, our studies indicate that PDGFRα activation together with Ink4a/Arf loss results in enhanced tumor growth of both mAsts and human glioma cells in the brain.
Tumor suppressor P16INK4A is frequently mutated in clinical glioblastomas (3, 4, 29). In mice, loss of p16INK4a and p19ARF was shown to be indispensable in facilitating tumorigenesis (6). Ink4a/Arf-deficient mice were viable and developed spontaneous tumors at early ages, but without detectable tumors in the brain (6, 7). Further studies showed that Ink4a/Arf loss cooperates with oncogenic K-Ras, EGFRvIII, or PDGF-B expression in promoting gliomagenesis in the brain (7, 8, 16, 26, 27). However, compared with p16INK4a loss, which contributes to tumor initiation from mAsts, p19ARF deficiency was shown to display a more pronounced impact on cell transformation and gliomagenesis (30, 31). Our data corroborate and also differ from these studies. We showed that re-expression of p16INK4a but not p19ARF in PDGFRα-expressing Ink4a/Arf–/– mAsts inhibited tumorigenesis. Inhibition of CDK4/6, the direct target of p16INKa by a specific inhibitor (20), in Ink4a/Arf–/– mAsts and glioma cells attenuated PDGF-A stimulation of soft agar growth, suggesting that CDK4/6/p-RB signaling is required for PDGFRα-induced tumorigenesis. On the contrary, although p53 was functional in Ink4a/Arf–/– PDGFRα-expressing mAsts, p19ARF was unable to suppress soft agar growth of these cells. Since p19ARF de-represses p53 signaling while PI3K/Akt activates p53 E3 ubiquitin ligase Mdm2 (32), it is plausible that the robust PI3K/Akt activation in PDGFRα-overexpressing cells triggers Mdm2-mediated p53 degradation and thereby renders tumorigenic mAsts resistant to p19ARF inhibition (21). It is also likely that loss of p19ARF is required for survival of glioma cells under certain situations, such as treatment of DNA damage–inducing agent CDDP (Supplemental Figure 3). At least during the initiation or maintenance of cell transformation, p19ARF loss appears to be dispensable (30, 31), since knockdown of INK4A alone in WT INK4A/ARF LN319 glioma cells restored PDGFRα stimulation of anchorage-independent growth in soft agar. Taken together, these data suggest that loss of p16INK4a plays a predominant role in PDGFRα-promoted gliomagenesis.
The third important aspect in this study is that we functionally assigned signal modules of PDGFRα in promoting gliomagenesis and tumor invasion in the brain of mice. Previous studies by uncoupling individual signaling pathways from PDGFRα using a series of F-to-Y mutants revealed unequal contributions of each signaling pathway emanating from PDGFRα activation (11, 12, 14). Our in vivo and biochemical data corroborate these reports. When compared with WT PDGFRα overexpression, loss of intrinsic tyrosine kinase activity (R627) of the RTK or binding capacity to PI3K (F731/42) or SHP-2 (F720) abrogated PDGFRα-promoted gliomagenesis, thus signifying the central roles of PI3K and SHP-2 signaling in PDGFRα function. However, while disruption of PDGFRα association with SFKs (F572/74) or PLCγ (F1018) only had a moderate impact on tumor formation in the brain, significant inhibition of tumor cell infiltration in the brain was seen in these tumors, as compared with the WT PDGFRα tumors, thus validating the role of SFKs and PLCγ in mediating PDGFRα-stimulated cell invasion (10). Moreover, a separate set of experiments with individual “add-back panel” mutants in which 1 of 5 signal modules was individually retained as single or double Y residues (Y731/42, Y572/74, Y720, Y988, and Y1018) could not fully restore PDGFRα-promoted tumorigenesis in the brain (data not shown), consistent with the previous findings using an identical set of PDGFRα add-back panel mutants (10). It is plausible that, similar to this previous in vitro study (10), in which retention of a combination of 2 or 3 add-back signaling modules with comparable levels of total RTK protein in WT PDGFRα–expressing cells was able to restore PDGFRα-mediated biological responses, co-activation of PI3K (Y731/42)– and SHP-2 (Y720)–mediated signaling may be required for the full spectrum of WT PDGFRα–promoted gliomagenesis in the brain.
The last (and what we believe is also the most novel) finding in this study is the emergence of SHP-2 as an essential mediator in PDGFRα-promoted gliomagenesis. SHP-2 (encoded by PTPN11 gene) is a protein tyrosine phosphatase (PTP) identified as a bona fide proto-oncogene that activates Ras/MAPK signaling through a yet-to-be-defined mechanism (33). Additionally, the role of SHP-2 in mediating PDGFRα signaling has not been clear (9, 11, 14). In human cancers including GBMs, mutations of SHP-2 or its binding partners have been reported, leading to sustained Ras/MAPK signaling (34, 35). Recent genomic analysis of TCGA data has designated PTPN11 as one of the 6 “linker” genes, which are statistically enriched for connections to various GBM altered genes, thus suggesting a critical role for SHP-2/PTPN11 in modulation of downstream biological signaling in gliomagenesis (36). Our data not only establish the critical role of SHP-2 in mediating PDGFRα activation for glioma formation but also functionally validate this hypothesis. We showed that inhibition of SHP-2 function by removal of its binding module in PDGFRα (F720 mutant), gene knockdown or pharmacological inhibitors significantly impaired PDGFRα stimulation of tumorigenesis in vivo and in vitro and its downstream signaling effectors, Erk1/2 and PI3K/Akt/mTOR in Ink4a/Arf-deficient mAsts and glioma cells. Significantly, re-expression of a constitutively active p110, the catalytic subunit of PI3K, rescued the inhibition of SHP-2 in Ink4a/Arf-deficient mAsts. In EGFRvIII-expressing U87MG glioma cells, SHP-2 regulates activities of ERK/2 and CDC2 that modulate cell cycle progression but has a minimal effect on AKT activation (23). On the contrary, we found that knockdown of SHP-2 not only attenuated PDGFRα stimulation of Erk1/2 activity but also impaired PI3K/Akt/mTOR activity in Ink4a/Arf-deficient mAsts and human glioma cells. The impact of inhibition of SHP-2 on PI3K/Akt signaling appears to be specific in astrocytic tumors, since we did not observe inhibited PDGF-A stimulation of p-Akt or association of PI3K with PDGFRα in SHP-2–knockdown NIH3T3 fibroblasts. On the other hand, the MEK inhibitor PD98059 also reduced the tumorigenicity of Ink4a/Arf-deficient mAsts, indicating that SHP-2–mediated PDGFRα signaling requires not only PI3K/Akt but also Erk1/2 activation in gliomagenesis. Additionally, SHP-2 was found to either positively or negatively regulate PI3K/AKT activity (34). However, in our model systems, SHP-2 is required for full activation of PI3K/Akt/mTOR, and inhibition of mTOR by rapamycin markedly suppressed PDGFRα-promoted tumorigenesis. These data are significant since a recent proteomic study revealed that mTOR signaling is predominantly activated in “PDGFRA co-cluster” glioblastomas (37), thus corroborating our observations. However, since phosphorylation of Akt Ser473 occurs both upstream and downstream of mTORC2 signaling, our data do not rule out the role of mTORC2 in PDGFRα-activated signaling. Lastly, our data also demonstrated the importance of activation of PI3K/mTOR signaling by SHP-2 in PDGFRα- or EGFRvIII-promoted tumorigenesis. Taken together, our findings suggest SHP-2 as a critical modulator that regulates PDGFRα-mediated PI3K/AKT/mTOR activities in the development of malignant glioblastomas.
In summary, this study provides molecular insights into the mechanisms by which PDGFRA amplification together with loss of INK4A/ARF promotes gliomagenesis in the brain. Our data identified SHP-2, as well as PI3K, as a pivotal mediator of PDGFRα signaling in glioma formation. These results have direct clinical relevance, since we not only establish a model system to demonstrate the co-operative role of PDGFRA overexpression and INK4A/ARF loss in clinical glioblastomas, but also provide functional evidence to validate genomic analyses demonstrating that SHP-2/PTPN11 is an essential “linker” among glioma altered genes (36). Secondly, activated mTOR signaling is predominantly found within the subclass of glioblastomas with abnormal PDGFRα signaling (37). Finally, constitutively active PI3K mutations (PIK3CA and PIK3R1) occur mostly in clinical glioblastomas without PDGFRA aberrations (4). Consequently, our results strongly suggest SHP-2/PTPN11 as a potential target for treatments of glioblastomas with PDGFRA overexpression.
Primary Ink4a/Arf–/– mAsts were derived and propagated as previously described (7). Human glioma cell lines LN444, LN443, LN-Z308, and LN319 were obtained from ATCC or from our own collection (38). Unless otherwise mentioned, all glioma cell lines and primary mAsts were routinely maintained in 5% CO2 at 37°C, in DMEM (Invitrogen) containing 10% FBS (Hyclone), 100 U/ml penicillin, and 100 μg/ml streptomycin (Invitrogen).
Studies using human tissues were reviewed and approved by the Institutional Review Board involving Human Subjects of the University of Pittsburgh. The specimens were de-identified human tissues, and thus no informed consent was required. A total of 158 paraffin-embedded thin sections of primary human glioma specimens were used, including 87 WHO grade IV GBMs, 34 grade III anaplastic astrocytomas, anaplastic oligodendrogliomas, or anaplastic oligodendroastrocytomas, and 37 grade II oligodendrogliomas and diffuse astrocytomas. Thin sections of human glioma specimens and mouse brains with various tumors were analyzed by IHC using indicated (e.g., Figure Figure88 legend and Results) antibodies or a TUNEL staining kit as previously described (38, 39). Briefly, the 5-μm human tissue sections were deparaffinized in xylene, followed by rehydration in graded ethanol. After washing with TBS, the antigen was retrieved by boiling the sections in a citrate buffer (pH 6.0) twice for 5 minutes. For mouse brain tissues with various gliomas, 5-μm frozen sections were fixed in pre-chilled acetone at –20°C for 5 minutes, rinsed with PBS, and blocked by AquaBlock (East Coast Biologics Inc.) for 1 hour at room temperature. Afterward, various tissue sections were then incubated with a primary antibody overnight at 4°C and blocked by Peroxidase Blocking Reagent (DAKO) for 10 minutes, followed by incubation with a biotinylated secondary antibody for 30 minutes at room temperature. After washing in PBS, stained tissue sections were visualized by diaminobenzidine chromophore and H2O2, followed by hematoxylin counterstaining. Sections were then dehydrated by graded ethanol and mounted with Permount Solution (Fisher).
Colony formation assay in soft agar was performed as previously described (40). Briefly, approximately 5,000 cells were seeded in a 0.5% Noble Agar top layer with a bottom layer of 0.8% Noble Agar in each of the triplicate wells of a 24-well plate. Growth factor–reduced Matrigel (1 mg/ml) was added into the top layer with or without inhibitors as indicated in figure legends (e.g., Figure Figure77 legend) and described in Results. PDGF-A–expressing cells in 10% of total cells were included as a source of PDGF-A. DMEM containing 10% FBS was added 3 days after plating and changed every 3 days thereafter. Colonies were scored after 2–3 weeks using Olympus SZX12 stereomicroscope, and data were analyzed using GraphPad Software.
Experiments of mouse glioma xenografts, IP, IB, tissue imaging and analyses, and siRNA knockdown were performed as previously described (39). For details, see Supplemental Experimental Procedures.
One-way ANOVA or an unpaired, 2-tailed Student’s t test followed by Newman-Keuls post-test was performed using GraphPad Prism software. A P value of 0.05 or less was considered statistically significant.
We would like to thank Carl-Henrik Heldin at Uppsala University for PDGFRα and PDGF-A cDNAs, Russell O. Pieper at University of California San Francisco for pMXI-gfp retroviral vector, Lynda Chin at Dana-Farber Cancer Institute for the cDNA of p16INK4a, Yi Zheng at Cincinnati Children’s Hospital Medical Center for cDNA of p19ARF, Tao Cheng at University of Pittsburgh for pMSCV retroviral vectors, and N. Balass for proofreading the manuscript. This work was supported by grants from the NIH (CA102011 and CA130966 to S.-Y. Cheng; EY012509 to A. Kazlauskas; CA87375 to K. Symes), American Cancer Society (RSG CSM-107144), and Pennsylvania Department of Health and Innovative Research Scholar Awards of the Hillman Foundation (to S.-Y. Cheng and B. Hu).
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2011;121(3):905–917. doi:10.1172/JCI43690.