PERK Activation Promotes Medulloblastoma Tumorigenesis by Attenuating Premalignant Granule Cell Precursor Apoptosis

Evidence suggests that activation of pancreatic endoplasmic reticulum kinase (PERK) signaling in response to endoplasmic reticulum stress negatively or positively in ﬂ uences cell transformation by regulating apoptosis. Patched1 heterozygous de ﬁ cient ( Ptch1 þ / (cid:2) ) mice reproduce human Gorlin ’ s syndrome and are regarded as the best animal model to study tumorigenesis of the sonic hedgehog subgroup of medulloblastomas. It is believed that medulloblastomas in Ptch1 þ / (cid:2) mice results from the transformation of granule cell precursors (GCPs) in the developing cerebellum. Here, we determined the role of PERK signaling on medulloblastoma tumorigenesis by assessing its effects on premalignant GCPs and tumor cells. We found that PERK signaling was activated in both premalignant GCPs in young Ptch1 þ / (cid:2) mice and medulloblastoma cells in adult mice. We demonstrated that PERK haploinsuf ﬁ ciency in Ptch1 / haploinsuf of but had a minimal effect on tumor cells in adult Ptch1 þ / (cid:2) mice after malignant transformation. our data indicate that PERK haploinsuf ﬁ ciency medulloblastoma development in Ptch1 þ / (cid:2) by in during the of

Evidence suggests that activation of pancreatic endoplasmic reticulum kinase (PERK) signaling in response to endoplasmic reticulum stress negatively or positively influences cell transformation by regulating apoptosis. Patched1 heterozygous deficient (Ptch1 þ/À ) mice reproduce human Gorlin's syndrome and are regarded as the best animal model to study tumorigenesis of the sonic hedgehog subgroup of medulloblastomas. It is believed that medulloblastomas in Ptch1 þ/À mice results from the transformation of granule cell precursors (GCPs) in the developing cerebellum. Here, we determined the role of PERK signaling on medulloblastoma tumorigenesis by assessing its effects on premalignant GCPs and tumor cells. We found that PERK signaling was activated in both premalignant GCPs in young Ptch1 þ/À mice and medulloblastoma cells in adult mice. We demonstrated that PERK haploinsufficiency reduced the incidence of medulloblastomas in Ptch1 þ/À mice. Interestingly, PERK haploinsufficiency enhanced apoptosis of premalignant GCPs in young Ptch1 þ/À mice but had no significant effect on medulloblastoma cells in adult mice. Moreover, we showed that the PERK pathway was activated in medulloblastomas in humans. These results suggest that PERK signaling promotes medulloblastoma tumorigenesis by attenuating apoptosis of premalignant GCPs during the course of malignant transformation. Pancreatic endoplasmic reticulum kinase (PERK) activation in response to endoplasmic reticulum (ER) stress promotes cell survival under stressful conditions through inhibition of global protein biosynthesis and induction of certain stressinduced cytoprotective genes by phosphorylating translation initiation factor 2a (eIF2a). 1,2 Nevertheless, PERK activation also controls an apoptotic program to eliminate ERstressed cells. 3,4 Studies found that PERK activation induced by hypoxia and nutritional deficiency, hallmarks of the solid tumor microenvironment, influences tumor development by regulating tumor cell viability, tumor invasion, and angiogenesis. 5e8 However, because of the double-edged sword nature of PERK signaling, the data about its role, promoting or inhibiting, in tumor development are contradictory. 9,10 It is well documented that apoptosis serves as a natural barrier to malignant transformation. Cell transformation requires concomitant activation of unrestricted cell proliferation and suppression of apoptosis. 11,12 Interestingly, recent reports suggest that PERK signaling participates in cell transformation by regulating apoptosis. 13,14 A report found that PERK activation promotes Myc-dependent malignant transformation by attenuating cell apoptosis. 13 In contrast, another report found that PERK activation suppresses malignant transformation by enhancing cell apoptosis. 15 Medulloblastoma is the most common solid malignancy of childhood. 16 From gene expression profiling, medulloblastoma can be divided into four discrete molecular subgroups: sonic hedgehog (SHH) subgroup, WNT subgroup, subgroup 3, and subgroup 4. 17,18 Although multiple genetic mutations that drive sustained cell proliferation are documented in medulloblastoma, 17e19 the mechanisms that regulate cell apoptosis during malignant transformation remain unclear. Mice heterozygous for Patched1 (Ptch1 þ/À ), a SHH receptor, recapitulate human Gorlin's syndrome and are regarded as the best animal model to study tumorigenesis of the SHH subgroup of medulloblastomas. 20,21 Medulloblastoma in Ptch1 þ/À mice arises from granule cell precursors (GCPs) in the external granular layer (EGL) of the developing cerebellum and shows distinct steps of progression. 22e24 These mice have hyperplastic lesions that contain premalignant GCPs in the EGL as late as 6 weeks of age. Most hyperplastic lesions regress; however, a few hyperplastic lesions undergo malignant transformation and progress to medulloblastoma.
Mutations of Perk (also known as Eif2ak3) were observed in human medulloblastoma. A report found that 7 of 35 human medulloblastomas exhibit Perk mutations. 25 Our previous in vitro study found that PERK activation facilitates human medulloblastoma cell migration and invasion through induction of vascular endothelial growth factor A (VEGF-A). 26 The report also found that PERK inactivation suppresses human medulloblastoma cell migration and invasion. 26 Moreover, a study found that ER stress responsive genes are up-regulated in premalignant GCPs in young Ptch1 þ/À mice. 22 Thus, we sought to elucidate the impact of the PERK branch of the ER stress response on the development of medulloblastoma. In this study, we confirmed PERK activation in human medulloblastoma and reported that PERK activation facilitated medulloblastoma development in Ptch1 þ/À mice by attenuating premalignant GCP apoptosis. PERK haploinsufficiency reduced the incidence of medulloblastomas in Ptch1 þ/À mice, which was associated with increased apoptosis of premalignant GCPs in young mice. Surprisingly, PERK haploinsufficiency had a minimal effect on medulloblastoma cells in adult Ptch1 þ/À mice.
Mice were monitored daily to detect medulloblastoma phenotypes, including ataxia, decreased movement, poor grooming, and doming of the skull, until the age of 8 months. All animal procedures were conducted in complete compliance with the NIH's Guide for the Care and Use of Laboratory Animals 28 and were approved by the Institutional Animal Care and Use Committee of the University of Minnesota.
Terminal deoxynucleotidyl transferaseemediated biotinylated UTP nick end labeling (TUNEL) staining was performed on paraffin sections with the use of the ApopTag kit (Millipore) according to the manufacturer's instructions. ajp.amjpathol.org -The American Journal of Pathology IHC for BrdU (dilution 1:1000; Sigma-Aldrich) was performed on paraffin sections as described previously. 29,30 We quantified positive cells for BrdU, TUNEL, and VEGF-A and for CD31 þ blood vessels in the center of medulloblastoma or hyperplastic lesions in the cerebellum as described previously. 29,30 Western Blot Analysis Cerebellar tissues were removed from adult control wild-type (WT) mice and adult symptomatic Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice, which displayed typical medulloblastoma clinical phenotypes and enlarged cerebellum. These tissues were homogenized in 5 volumes of Triton X-100 buffer with the use of a motorized homogenizer as previously described. 31e33 After incubation on ice for 15 minutes, the extracts were cleared by centrifugation at 18,000 Â g for 30 minutes twice. The protein concentration of each extract was determined by DC Protein Assay (Bio-Rad Laboratories, Hercules, CA). The extracts (120 mg) were separated by SDS-PAGE and transferred to nitrocellulose membranes. The blots were then incubated with primary antibodies to p-PERK (dilution 1:1000; Santa Cruz Biotechnology), p-eIF2a (dilution 1:1000; Cell Signaling Technology), eIF2a (dilution 1:1000; Santa Cruz Biotechnology), and actin (dilution 1:5000; Sigma-Aldrich), followed by the horseradish peroxidaseconjugated secondary antibody (Vector Laboratories). The chemiluminescent signal was detected by the ECL Detection Reagents (GE Healthcare Biosciences, Piscataway, NJ). The intensity of the recorded chemiluminescence signal was quantified with the ImageQuantTL LAS4000 software version 1.2 from GE Healthcare Life Sciences.

TaqMan Real-Time PCR
Cerebellar tissues were removed from adult control WT mice and adult symptomatic Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice, which displayed typical medulloblastoma clinical phenotypes and enlarged cerebellum. RNA was isolated from the cerebellum with the use of TRIzol reagent (Invitrogen, Carlsbad, CA) and was treated with DNase I (Invitrogen) to eliminate genomic DNA. Reverse transcription was performed with the iScript cDNA Synthesis Kit (Bio-Rad Laboratories). TaqMan real-time PCR was performed with iQ Supermix (Bio-Rad Laboratories) on the LightCycler 480 System (Roche Diagnostics Corporation, Indianapolis, IN) as described previously. 26,29,31 Human Medulloblastoma Samples and IHC Paraffin-embedded human medulloblastoma sections (5-mm thickness) and brain sections from individuals without brain tumor were provided by Brain Tumor Tissue Bank (London, ON, Canada), including six cases of pediatric medulloblastoma, three cases of adult medulloblastoma, and three nontumor brain tissue samples. IHC for p-PERK and CAATT enhancer binding protein homologous protein (CHOP; dilution 1:100; Thermo Scientific, Grand Island, NY) was performed as described in Histology, Immunohistochemistry, and TUNEL Staining. The degree of positive staining for p-PERK was determined by using a semiquantitative staining index Z intensity Â distribution. The intensity of cytoplasmic staining detected by IHC was scored on a scale of 0 to 4: 0 Z negative staining, 1 Z weak staining, 2 Z modest staining, 3 Z moderate staining, and 4 Z strong staining. The distribution was scored on a scale of 1 to 4 from the percentage of cells with any level of staining: 1 Z 1% to 25% positive cells, 2 Z 26% to 50% positive cells, 3 Z 51% to 75% positive cells, and 4 Z 76% to 100% positive cells.

Statistical Analysis
Data are expressed as means AE SD. Comparisons between two groups were statistically evaluated by Student's t-test with the use of GraphPad Prism software version 6 (GraphPad Software Inc., La Jolla, CA). The incidence of symptomatic medulloblastoma between Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice was statistically evaluated by Kaplan-Meier analysis and the incidence of total medulloblastoma between these two groups was statistically evaluated by c 2 test with the use of Prism 6 (GraphPad). P < 0.05 was considered significant.

PERK Activation in Human Medulloblastoma
During ER stress, PERK is activated through oligomerization and autophosphorylation, which coordinates an adaptive program by phosphorylating eIF2a at serine 51. 1 To assess the potential relevance of previous observations of the enhanced expression of ER stress responsive genes in mouse models of medulloblastoma, 22,29 we estimated the level of PERK activity in human medulloblastoma by p-PERK IHC. As expected, immunoreactivity of p-PERK was undetectable in nontumor brain tissues ( Figure 1A). Interestingly, a number of tumor cells were positive for p-PERK in clinical samples of pediatric medulloblastoma ( Figure 1B, Table 1), and the p-PERK IHC signal was also markedly increased in the tumor cells in adult medulloblastoma ( Figure 1C, Table 1). Moreover, semiquantitative analysis confirmed that PERK was activated in medulloblastoma in all six cases of pediatric patients and three cases of adult patients (Table 1). However, the intensity and distribution of p-PERK staining in human medulloblastoma were variable, which likely reflects the heterogeneity of human tumor. In addition, IHC for CHOP, a PERK-responsive gene, found that immunoreactivity of CHOP was undetectable in nontumor brain tissues ( Figure 1D), but it became detectable in a number of tumor cells in both pediatric and adult medulloblastomas ( Figure 1, E and F). Taken together, our data suggest that PERK signaling is enhanced in human medulloblastoma, prompting an enquiry into its potential pathophysiologic significance.

PERK in Medulloblastoma Tumorigenesis
The American Journal of Pathologyajp.amjpathol.org 1941

PERK Activation in Medulloblastoma Cells in Adult Ptch1 þ/À Mice and in Premalignant GCPs in Young Mice
Medulloblastomas in Ptch1 þ/À mice arises from the cerebellar GCPs in the EGL. 34,35 During early postnatal development, GCPs proliferate, differentiate, and migrate to their final destination, the internal granule layer. In normal mice, by postnatal day 21, this process is complete, and there are no GCPs in the EGL. In contrast, Ptch1 þ/À mice have abnormal patches of premalignant GCPs in the EGL of the cerebellum at the age of 3 weeks. 22,36 By the age of 6 weeks, these mice have focal hyperplastic lesions or diffuse hyperplastic lesions that contain premalignant GCPs in the cerebellum. These hyperplastic lesions either undergo malignant transformation and progress to medulloblastoma or regress over time. 22,24,37 H&E staining revealed typical medulloblastoma in the cerebellum of adult Ptch1 þ/À mice that displayed medulloblastoma clinical phenotypes ( Figure 2B) and typical hyperplastic lesions in the EGL in the cerebellum of 6-week-old Ptch1 þ/À mice ( Figure 2D) compared with the normal cerebellum of age-matched WT mice ( Figure 2, A and C).
To investigate the role of PERK signaling in medulloblastoma tumorigenesis, we determined the activity of PERK signaling in medulloblastoma cells and premalignant GCPs in Ptch1 þ/À mice. Similar to our finding in human medulloblastoma samples (see Perk Activation in Human Medulloblastoma), p-PERK IHC found that the immunoreactivity of p-PERK was undetectable in cells in the cerebellum of adult WT mice ( Figure 2E), but it became detectable in a number of medulloblastoma cells in adult Ptch1 þ/À mice ( Figure 2F). Western blot analysis also found that the levels of p-PERK and p-eIF2a were significantly increased in the cerebellum of adult Ptch1 þ/À mice with medulloblastomas compared with adult WT mice (Figure 2, I and J). Although real-time PCR analysis found comparable levels of PERK mRNA in the cerebellum of adult WT mice and Ptch1 þ/À mice with medulloblastomas, the mRNA levels of CHOP and growth arrest and DNA damage 34 (GADD34), target genes of the PERK-eIF2a pathway, were significantly increased in the cerebellum of adult Ptch1 þ/À mice with medulloblastomas compared with adult WT mice ( Figure 2K). Moreover, consistent with a previous report indicating that the levels of ER stress responsive genes were elevated in premalignant GCPs in hyperplastic lesions in 6-week-old Ptch1 þ/À mice, 22 p-PERK IHC found that the immunoreactivity of p-PERK was undetectable in the cerebellar cells of 6-week-old WT mice ( Figure 2G) but became detectable in a number of premalignant GCPs in hyperplastic lesions in Ptch1 þ/À mice ( Figure 2H). Taken together, these data indicate that PERK signaling is activated in both premalignant GCPs and medulloblastoma cells in Ptch1 þ/À mice, suggesting its potential role in tumorigenesis.
PERK Haploinsufficiency Reduces the Incidence of Medulloblastomas in Ptch1 þ/À Mice Ptch1 þ/À mice develop symptomatic medulloblastomas typically between the ages of 8 and 35 weeks. 21,38 Perk Figure 1 Activation of PERK signaling in human medulloblastoma. A: p-PERK IHC found that immunoreactivity of p-PERK was undetectable in nontumor brain tissues. B: A representative image from pediatric medulloblastoma 5 in Table 1 shows that a number of tumor cells are positive for p-PERK in pediatric medulloblastoma. C: A representative image from adult medulloblastoma 1 in Table 1 shows that the level of p-PERK is markedly increased in the tumor cells in adult medulloblastoma. D: Immunoreactivity of CHOP was undetectable in nontumor brain tissues. E: Many tumor cells were positive for CHOP in pediatric medulloblastoma. F: Many tumor cells were positive for CHOP in adult medulloblastoma. n Z 3 patients (A, C, D, F); n Z 6 patients (B and E). Scale bar Z 50 mm. CHOP, CAATT enhancer binding protein homologous protein; IHC, immunohistochemistry; p-, phosphorylated; PERK, pancreatic endoplasmic reticulum kinase.  , and a hyperplastic lesion (arrow) in the cerebellar EGL of 6-week-old Ptch1 þ/À mice (D). E and F: p-PERK IHC found that p-PERK is undetectable in cells in the cerebellum of adult WT mice but became detectable in a number of cells in medulloblastomas in adult Ptch1 þ/À mice. G and H: p-PERK IHC found that p-PERK was undetectable in cells in the cerebellum of 6-week-old WT mice but became detectable in a number of cells in hyperplastic lesions in the cerebellar EGL of 6-week-old Ptch1 þ/À mice. I and J: Western blot analysis found significantly increased levels of p-PERK and p-eIF2a in the cerebellum of adult Ptch1 þ/À mice with medulloblastomas compared with adult WT mice. K: Real-time PCR analysis found that the mRNA level of PERK in the cerebellum of adult Ptch1 þ/À mice with medulloblastomas was comparable with adult WT mice and that the mRNA levels of CHOP and GADD34 were significantly increased in the cerebellum of adult Ptch1 þ/À mice with medulloblastomas compared with adult WT mice. Data are expressed as means AE SD. n Z 3 animals. *P < 0.05. Scale bars: 1000 mm (AeD); 100 mm (EeH). CHOP, CAATT enhancer binding protein homologous protein; EGL, external granular layer; eIF2a, translation initiation factor 2a; GADD34, growth arrest and DNA damage 34; H&E, hematoxylin and eosin; IHC, immunohistochemistry; p-, phosphorylated; PERK, pancreatic endoplasmic reticulum kinase; WT, wild-type.
We further determined whether PERK heterozygous deficiency impaired the activity of PERK signaling in medulloblastoma cells in Ptch1 þ/À mice. p-PERK IHC found that the immunoreactivity of p-PERK was noticeably decreased in medulloblastomas in Ptch1 þ/À ; Perk þ/À mice compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 4, C and D). Moreover, Western blot analysis found that the levels of p-PERK and p-eIF2a were significantly decreased in the cerebellum of Ptch1 þ/À ; Perk þ/À mice with medulloblastomas compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 4, E and F). As expected, real-time PCR analysis found that the level of PERK mRNA was reduced by approximately one-half in the cerebellum of Ptch1 þ/À ; Perk þ/À mice with medulloblastomas compared with Ptch1 þ/À ; Perk þ/þ mice ( Figure 4G). Moreover, real-time PCR analysis found that PERK heterozygous deficiency significantly reduced the expression of CHOP and GADD34 in the cerebellum of Ptch1 þ/À mice with medulloblastomas ( Figure 4G). Taken together, these results indicate that PERK haploinsufficiency reduces the incidence of medulloblastomas in Ptch1 þ/À mice, likely through reduced PERK-mediated eIF2a phosphorylation.
Next, we determined the effects of PERK haploinsufficiency on cell proliferation, cell apoptosis, and angiogenesis in medulloblastomas in Ptch1 þ/À mice. H&E staining indicated that the structural characteristics of medulloblastomas in Ptch1 þ/À ; Perk þ/À mice was comparable with Ptch1 þ/À ; Perk þ/þ mice (Figure 4, A and B). IHC for glial fibrillary acidic protein and synaptophysin found that the expression pattern of both glial fibrillary acidic protein and synaptophysin in medulloblastomas in Ptch1 þ/À ; Perk þ/À mice was comparable with Ptch1 þ/À ; Perk þ/þ mice ( Figure 5, AeD). BrdU labeling ( Figure 5, E, F, and M) and PCNA IHC (data not shown) found that PERK heterozygous deficiency did not significantly change the number of proliferating cells in medulloblastomas in Ptch1 þ/À mice. Moreover, TUNEL staining found that PERK heterozygous deficiency had no significant effect on cell apoptosis in medulloblastomas in Ptch1 þ/À mice ( Figure 5, G, H, and M). A number of studies suggest that activation of PERK signaling enhances angiogenesis in tumors, including medulloblastoma, by induction of VEGF-A. 29,42 Surprisingly, VEGF-A IHC found that PERK heterozygous deficiency did not significantly alter VEGF-A expression in medulloblastomas in Ptch1 þ/À mice ( Figure 5, I, J, and M), and CD31 IHC found that PERK heterozygous deficiency did not significantly change angiogenesis in this tumor ( Figure 5, K, L, and N). Collectively, these data suggest that PERK haploinsufficiency has a minimal effect on cell proliferation, cell apoptosis, and angiogenesis in medulloblastomas in adult Ptch1 þ/À mice. Thus, it is unlikely that PERK haploinsufficiency impairs medulloblastoma formation in Ptch1 þ/À mice through its effects on tumor cells.

PERK Haploinsufficiency Increases Premalignant GCPs
Apoptosis in the Cerebellum of Young Ptch1 þ/À Mice Evidence is emerging that PERK signaling influences tumorigenesis by regulating cell apoptosis during the course reduced the incidence of symptomatic medulloblastomas in Ptch1 þ/À mice. The percentage of mice that developed symptomatic medulloblastoma. n Z 57 animals. PERK, pancreatic endoplasmic reticulum kinase. ajp.amjpathol.org -The American Journal of Pathology of malignant transformation. 13,15,43 Medulloblastoma tumorigenesis in the cerebellum of Ptch1 þ/À mice found distinct steps of progression; first, focal hyperplastic lesions develop in the EGL; second, diffuse hyperplastic lesions form in the EGL and molecular layer; finally, medulloblastoma arises. 23,24,37 Previous reports found that both focal hyperplastic lesions and diffuse hyperplastic lesions can be observed in the cerebellum of Ptch1 þ/À mice as late as 6 weeks of age. 22,24 The diffuse hyperplastic lesions are reversible and represent the final step in the process of malignant transformation, in which premalignant GCPs either gain additional genetic mutations or epigenetic changes to become irreversible tumor cells, or undergo regression. 24,37,44 Therefore, we determined the effects of PERK haploinsufficiency on the proliferation and apoptosis of premalignant GCPs in diffuse hyperplastic lesions in the cerebellum of 6-week-old Ptch1 þ/À mice.
We serially sectioned whole paraffin-embedded halfcerebellum of 6-week-old mice, each half-cerebellum yielded approximately 210 sections 5-mm thick. Every 10th section in the series was either stained by H&E or immunostained by the NeuN antibody or PCNA antibody. Focal hyperplastic lesions, which contained >10 premalignant GCPs that were negative for NeuN (a marker of mature neurons) and positive for PCNA, were observed in the cerebellum of Ptch1 þ/À ; Perk þ/þ mice (Supplemental Figure S1, A, C, and E) and Ptch1 þ/À ; Perk þ/À mice (Supplemental Figure S1, B, D, and F). Moreover, both Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice had diffuse hyperplastic lesions in the cerebellum (Supplemental Figure S1, GeL), which contained a thick layer (>15 cell layers) of NeuN-negative and PCNA-positive premalignant GCPs diffusing in the EGL and molecular layer. We found, however, that the number of total hyperplastic lesions in the cerebellum of 6-week-old Ptch1 þ/À ; Perk þ/À mice was comparable with Ptch1 þ/À ; Perk þ/þ mice ( Figure 6A). Similarly, we found that PERK heterozygous deficiency did not reduce the number of diffuse hyperplastic lesions in the cerebellum of 6-week-old Ptch1 þ/À mice ( Figure 6B and Figure 7, A and B).

PERK in Medulloblastoma Tumorigenesis
The American Journal of Pathologyajp.amjpathol.org 6-week-old Ptch1 þ/À ; Perk þ/À mice compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 7, C and D). BrdU labeling (Figure 7, E, F, and K) and PCNA IHC (data not shown) found that PERK heterozygous deficiency did not significantly alter the rate of cell proliferation in diffuse hyperplastic lesions in 6-week-old Ptch1 þ/À mice. Importantly, TUNEL labeling found that PERK heterozygous deficiency significantly increased the number of apoptotic cells in diffuse hyperplastic lesions in 6-week-old Ptch1 þ/À mice (Figure 7, G, H, and L). Active-caspase 3 IHC also found that the number of apoptotic cells were significantly increased in diffuse hyperplastic lesions in 6-week-old Ptch1 þ/À ; Perk þ/À mice compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 7, I, J, and M). However, there were noticeably fewer active-caspase 3-positive cells than TUNEL-positive cells in corresponding lesions. This discrepant result likely reflects that TUNEL labeling is a more sensitive method to detect cell apoptosis in diffuse hyperplastic lesions than active-caspase 3 IHC. Collectively, these data suggest that PERK haploinsufficiency enhances premalignant GCP apoptosis in diffuse hyperplastic lesions in the cerebellum of Ptch1 þ/À mice during GFAP IHC found that PERK heterozygous deficiency did not change the expression pattern of GFAP in medulloblastomas in adult Ptch1 þ/À mice. C and D: Synaptophysin IHC found that PERK heterozygous deficiency did not change the expression pattern of synaptophysin in medulloblastomas in adult Ptch1 þ/À mice. E, F, and M: BrdU labeling revealed comparable number of proliferating cells in medulloblastomas in adult Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice. G, H, and M: TUNEL staining revealed comparable number of apoptotic cells in medulloblastomas in adult Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice. I, J, and M: VEGF-A IHC revealed comparable number of VEGF-Aepositive cells in medulloblastomas in adult Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice. K, L, and N: CD31 IHC revealed comparable number of blood vessels in medulloblastomas in adult Ptch1 þ/À ; Perk þ/þ mice and Ptch1 þ/À ; Perk þ/À mice. Data are expressed as means AE SD. n Z 4 animals. Scale bars: 50 mm (AeJ); 100 mm (K and L). BrdU, bromodeoxyuridine; CD31, cluster of differentiation 31; GFAP, glial fibrillary acidic protein; IHC, immunohistochemistry; PERK, pancreatic endoplasmic reticulum kinase; TUNEL, terminal deoxynucleotidyl transferasee mediated biotinylated UTP nick end labeling; VEGF-A, vascular endothelial growth factor A. ajp.amjpathol.org -The American Journal of Pathology malignant transformation of medulloblastoma. As described in PERK Haploinsufficiency Reduces the Incidence of Medulloblastomas in Ptch1 þ/À Mice, we found that PERK haploinsufficiency reduced the incidence of medulloblastoma but had a minimal effect on tumor cells in adult Ptch1 þ/À mice after malignant transformation. Taken together, our data indicate that PERK haploinsufficiency suppresses medulloblastoma development in Ptch1 þ/À mice by enhancing premalignant GCP apoptosis in diffuse hyperplastic lesions during the course of malignant transformation.
It was found that PERK activation attenuates the activity of the p53 pathway, the master regulator of apoptosis during malignant transformation. 45,46 A previous report found that p53 deficiency increases the incidence and accelerates the development of medulloblastomas in Ptch1 þ/À mice. 47 Therefore, we were interested in the potential role of p53 in the proapoptotic effects of PERK haploinsufficiency on premalignant GCPs. p53 Immunostaining found immunoreactivity of p53 was noticeably increased in diffuse hyperplastic lesions in 6-week-old Ptch1 þ/À ; Perk þ/À mice compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 7, N and O). PUMA, a Bcl-2 homology 3-only Bcl-2 family member, is a p53-responsive gene and a critical mediator of p53-induced apoptosis. 48,49 Interestingly, PUMA immunostaining found that immunoreactivity of PUMA was markedly elevated in diffuse hyperplastic lesions in 6-week-old Ptch1 þ/À ; Perk þ/À mice compared with Ptch1 þ/À ; Perk þ/þ mice (Figure 7, P and Q). Collectively, these data suggest the possibility that PERK haploinsufficiency induces premalignant GCP apoptosis in diffuse hyperplastic lesions during the course of malignant transformation by increasing the activity of the p53-PUMA pathway.

Discussion
With the use of in vitro and xenograft models, recent studies suggest that PERK activation might promote malignancy by attenuating apoptosis during the course of cell transformation. 13,43 Although two previous papers indicate the inhibiting effects of PERK deficiency on tumor growth in genetically engineered mouse models, 50,51 the promoting role of PERK signaling in malignant transformation has not been verified in animal models that faithfully resemble human tumors. Ptch1 þ/À mice recapitulate human Gorlin's syndrome and are regarded as the best mouse model to study the mechanisms of malignant transformation of the SHH subgroup of medulloblastomas. 21,23,24,44 With the use of this mouse model, herein, we provided compelling evidence that PERK activation promoted medulloblastoma development by attenuating premalignant GCP apoptosis during the course of malignant transformation. We showed that PERK signaling was activated in premalignant GCP in focal hyperplastic lesions and diffuse hyperplastic lesions in the cerebellum of young Ptch1 þ/À mice, in medulloblastoma cells in adult Ptch1 þ/À mice, and in medulloblastoma cells in human patients. We demonstrated that PERK haploinsufficiency reduced the incidence of medulloblastomas in Ptch1 þ/À mice. Importantly, we found that PERK haploinsufficiency had a minimal effect on medulloblastoma cells in adult Ptch1 þ/À mice after malignant transformation but significantly enhanced premalignant GCP apoptosis in diffuse hyperplastic lesions in young Ptch1 þ/À mice during malignant transformation.
Apoptosis is believed to serve as a natural barrier to malignancy. 11,12 The transforming effects of oncogenes, which drive unrestrained cell proliferation, are countered by intrinsic apoptosis mechanisms, which are triggered in response to various physiologic stresses that cells experience during the course of malignant transformation. Recent studies report that ER stress is activated in cells during malignant transformation and that PERK activation in response to ER stress either facilitates the transformation by suppressing apoptosis or impairs the transformation by promoting apoptosis. 13,15,43 In this study, we showed that PERK signaling was activated in premalignant GCPs in reversible hyperplastic lesions in the cerebellum of young Ptch1 þ/À mice and that PERK haploinsufficiency enhanced premalignant GCP apoptosis and reduced tumor incidence in these mice. Thus, our data suggest that PERK signaling promotes medulloblastoma tumorigenesis by circumventing apoptosis during the transformation of GCPs. It is generally believed that activation of PERK signaling preserves cell viability under stressful conditions through inhibition of global protein biosynthesis and induction of certain stressinduced cytoprotective genes. 2,3 Several studies report that PERK activation attenuates the activity of the p53 pathway, the master regulator of apoptosis during malignant transformation. 45,46 Loss-of-function mutations of p53 contribute Figure 6 PERK heterozygous deficiency did not reduce the number of the cerebellar hyperplastic lesions in young Ptch1 þ/À mice. A: The scatter plot shows that PERK heterozygous deficiency did not significantly change the number of total hyperplastic lesions in the cerebellum of 6-week-old Ptch1 þ/À mice. B: The scatter plot shows that PERK heterozygous deficiency did not reduce the number of diffuse hyperplastic lesions in the cerebellum of 6-week-old Ptch1 þ/À mice. Data are expressed as means AE SD. n Z 10 mice. PERK, pancreatic endoplasmic reticulum kinase.
The PERK-eIF2a pathway is a major regulator of cell viability during ER stress under various physiologic and pathologic conditions. 2,3 As expected, we found that PERK heterozygous deficiency significantly attenuated the activity of the PERK-eIF2a pathway in medulloblastoma cells in adult Ptch1 þ/À mice. Surprisingly, TUNEL labeling found that PERK heterozygous deficiency had no significant effect on cell apoptosis in the tumor. Moreover, BrdU labeling and PCNA immunostaining found that PERK heterozygous deficiency did not alter cell proliferation in the tumor. Multiple cooperative mutations of oncogenes and oncosuppressor genes are necessary for tumorigenesis. It is generally believed that additional mutations of oncogenes or oncosuppressor genes, besides Ptch1 heterozygous mutation, are required for medulloblastoma tumorigenesis in Ptch1 þ/À mice. 23,44,53 The diminished impact of PERK haploinsufficiency on medulloblastoma cell viability, compared with its proapoptotic effect on premalignant GCPs in diffuse hyperplastic lesions, suggests that the tumor cells in Ptch1 þ/À ; Perk þ/À mice gain additional mutations or epigenetic changes that compensate or overcome the deficiency of the PERK-eIF2a pathway.
Data from in vitro studies and xenograft models suggest that PERK activation enhances tumor angiogenesis by inducing VEGF-A expression. 42,54 In vitro studies from our laboratory and other groups found that PERK activation enhances VEGF-A expression in human medulloblastoma cells. 26,55 Our previous study also suggests that inactivation of GADD34, a regulatory subunit of a phosphatase complex that dephosphorylates eIF2a, enhances the activity of the PERK-eIF2a pathway and may facilitate medulloblastoma formation in the cerebellum of transgenic mice that express the immune cytokine interferon-g in the central nervous system by promoting angiogenesis through induction of VEGF-A. 29 Moreover, a report indicates that PERK deficiency attenuates tumor cell proliferation and angiogenesis in b-cell insulinomas in transgenic mice that express the SV40 large T-antigen specifically in b cells. 50 In contrast, another report indicates that PERK deficiency delays tumor development but has no important effect on angiogenesis in mammary adenocarcinoma in transgenic mice that express oncogene Neu specifically in mammary epithelium. 51 We showed here that PERK haploinsufficiency reduced the incidence of medulloblastomas in Ptch1 þ/À mice but did not affect VEGF-A expression or angiogenesis in the tumor. Collectively, these data likely suggest that the promoting effects of PERK signaling on VEGF-A expression and/or tumor angiogenesis are not ubiquitous, possibly determined by genetic mutations or epigenetic changes in individual molecular subtypes of tumors.

Conclusions
With the use of Ptch1 þ/À mice, the results presented in this study suggest that PERK activation contributes to medulloblastoma development through inhibition of premalignant GCP apoptosis during the course of malignant transformation. This work represents the first experimental demonstration of a link between PERK activation and malignant transformation in an animal model that faithfully resembles human tumor. Moreover, we demonstrated PERK activation in human medulloblastoma. Unfortunately, human samples do not afford an opportunity to study the early stages of tumor development, but the persistence of PERK signaling is also consistent with a pervasive role in medulloblastoma tumorigenesis.