|Home | About | Journals | Submit | Contact Us | Français|
Low grade astrocytomas are the most common CNS tumors in neurofibromatosis type 1(NF1) patients. While most are classic pilocytic astrocytomas (PA), some are difficult to classify, and have been termed “low grade astrocytoma subtype indeterminate” (LGSI). Some of these tumors exhibit peculiar morphologies, including plump cytoplasmic processes and macronucleoli. In the current study we performed electron microscopy, followed by gene expression, immunohistochemicai and western blot analyses in an effort to identify biological differences underlying phenotypic variation in NF1-associated low grade astrocytoma. Electron microscopy demonstrated intermediate filaments and frequent Rosenthal fiber material in both PA and LGSI. Dense core granules and/or aligned microtubules were present in the LGSI group (2 of 3 cases) and in the PA group (1 of 10 cases). Analysis of global gene expression data obtained using Affymetrix HG-U133 Plus2.0 chips (5 PA, 1 LGSI), and western blot analysis for phospho-S6 (1 LGSI, 2 PA) demonstrated a gene expression profile reflecting “neuronal differentiation” and increased phospho-S6 immunoreactivity consistent with mTOR activation in the LGSI compared with PA. These findings were confirmed by immunohistochemistry for neuronal markers, as well as combined phospho-S6/ phospho-p70S6K immunoreactivity in 4 (of 4) LGSI vs. 5 (of 13) NF1-associated PA (p=0.02), and 13 (of 39) sporadic PA. Phospho-ERK immunoreactivity was uniformly present in PA and LGSI groups, while BRAF duplication was absent by FISH in 8 NF1-associated low grade astrocytomas. In summary, differential expression of neuronal-related genes and increased mTOR activation may underlie phenotypic variations in NF1-associated low grade astrocytomas.
Neurofibromatosis type 1 (NF1) patients show a predisposition to the formation of both peripheral and central nervous system (CNS) tumors. While most NF1-associated CNS neoplasms are low grade astrocytomas with classic features of pilocytic astrocytoma (PA) (49%), a subset remain difficult to classify past the simple designation “low grade astrocytoma subtype indeterminate” (LGSI) (17%) despite detailed histologic examination . No statistically significant differences were noted between the two groups when comparing mitotic activity, MIB-1 labeling indices, p53 labeling indices, degree of microscopic infiltration, time to recurrence after surgery, radiographic progression, or overall disease-specific survival times . It is unknown what relationship LGSI tumors have to PA in this syndromic setting. A curious morphologic variation within LGSI and some PA is the presence of tumors wherein the astrocytic cells feature plump, abundant cytoplasm, thick processes, and central nuclei with prominent nucleoli. Characterization of these phenotypic variations within NF1 is lacking in the literature.
Ultrastructural studies of NF1 related intracranial astrocytomas are few with exception of ones focusing upon optic nerve gliomas [2-5]. A recent, detailed study of optic gliomas arising in a mouse model of NF1 (Nf1 +/-GFAPCKO) demonstrated axonal irregularities and glial disorganization at the ultrastructural level associated with retinal ganglion cell loss . Ultrastructurally, conventional PA are characterized by conspicuous intermediate filaments within their piloid cell element, electrondense Rosenthal fiber material, amorphous spheres (hyaline droplets), and accumulation of large cytoplasmic lysosomes (granular bodies) [2, 7]. Ultrastructural properties of the full spectrum of NF1-associated low grade astrocytomas are not well characterized to our knowledge.
NF1 deficient astrocytes exhibit increased levels of the mammalian target of rapamycin (mTOR) activation, resulting in ribosomal S6 activation in both NF1 mutant murine optic gliomas and NF1 associated pilocytic astrocytomas in humans . mTOR is a nutrient-sensor control protein that signals several translational factors in response to nutrient availability. Additionally, mTOR regulates the rate at which the translational machinery is synthesized, both at the level of transcription and translation as well as serving several other known cellular functions . Activation of mTOR leads to phosphorylation of S6 kinase and activation of ribosomal S6 , with identification of phosphorylated S6 kinase or ribosomal S6 serving as surrogates of mTOR activation.
The purpose of this study was to explore phenotypic differences between different subsets of NF1-associated low grade astrocytomas, and understand what molecular differences underlie them. Since some NF1-associated LGSI tumors are morphologically reminiscent of subependymal giant cell astrocytoma (SEGA), a tumor typical of tuberous sclerosis, the molecular hallmark of which is mTOR activation through mutations of the TSC1 and TSC2 genes , we hypothesized that differential mTOR activation may underlie the phenotypic variation of other low grade astrocytomas in NF1.
Detailed clinicopathologic features of these tumors were reported in a prior study . The current study included 22 cases of NF1-associated low grade astrocytomas (16 PA, 6 LGSI, 1 WHO grade II diffuse astrocytoma). A single gan-glioglioma was also included. Six of the NF1-associated PA, as well as 39 sporadic PA, — present on a tissue microarray were also used for phospho-specific antibody immunohistochemistry. All studies were approved by the Institutional Review Board and Biospecimen committee at Mayo Clinic.
Fifteen cases with available glutaraldehyde-fixed and Epon embedded material were studied. The sections were stained with uranyl acetate and lead citrate and examined on a Tecnai 12 model transmission electron microscope (FEI Corp., Eindhoven, Netherlands). All tumors were examined for the presence of intermediate filaments, Rosenthal fibers, microtubules, dense core granules, cytoplasmic organelles, and macronucleoli by three neuropathologists (CG, FJRBWS).
RNA gene expression data were analyzed from 6 NF1-associated low grade astrocytomas obtained from a prior single hybridization experiment  using Affymetrix HG-U133 Plus 2.0 chips. Upon review of all histologic slides, one tumor had the morphologic features described above (i.e. abundant cytoplasm, prominent nucleoli, scant to absent eosinophilic granular bodies and Rosenthal fibers) and was classified as LGSI. In brief, a standard in-house MicroArray Preprocessing (MAPP) workflow was used to preprocess the data, including BG Correction, normalization and PM correction. Gene expression differences were expressed in the form of fold changes between 1 LGSI and the 5 PA.
All immunohistochemical studies were performed using a Dako autostainer and the Dual Link Envision detection system, on 5 micron thick formalin-fixed paraffin-embedded, conventional whole sections or tissue microarray sections using antibodies directed against (GFAP) (polyclonal, 1:4000; Dako, Carpinteria, Calif), S – 100 protein (polyclonal, 1:1600; Dako), chromogranin A (LK2H10, 1:500; Chemicon, Billerica, Mass), synaptophysin (clone SY38, 1:40; ICN, Costa Mesa, Calif), neurofilament protein (clone 2F11, 1:75; Dako); phospho-ERK (Rabbit polyclonal(Erkl/2) (Thr202/Tyr204; 1:250 cell signaling technology), phospho-p70S6K (Phospho-p70 S6 Kinase) (Rabbit polyclonal recognizing phospho-Thr389; 1:25; cell signaling technology), phospho-S6 (Rabbit polyclonal p-S6 (Ser235/236) 1:200, cell signaling technology), and Ki–67 (clone MIB-1, monoclonal, 1:300; Dako). Stains were scored according to a four tiered semiquantitative scale by two observers as previously described . MIB-1 (Ki-67) labeling indices were evaluated using the CAS200 imaging system (Bacus Laboratories, Lombard, III) and by examining 20 consecutive high magnification fields.
Total proteins were isolated from patient frozen tissues and separated by 10% SDS-polyacrylamide gel and transferred onto a PVDF membrane. The membrane was blotted overnight at 4 °C with 1:1000 dilution of primary antibody. Antibodies used were specific for phospho-S6 (Ser235/236), total S6, and β-actin (all from Cell Signaling Technology). Next, the blot was incubated with 1:5000 dilution of secondary anti-rabbit antibody conjugated with horseradish peroxidase (Pierce). Detection was performed with the Super Signal Chemiluminescence reagent according to the manufacturer's protocol (Pierce). Relative quantification of protein levels from western blot bands was done using ImageJ software from the National Institutes of Health (http://rsb.info.nih.gov/ii/index.html).
FISH studies to evaluate for BRAF duplication, a molecular finding characteristic of the majority of sporadic PA, was performed using a custom made probe (clones RP4-592P3, RP4-726N20, RP5-839B19, RP4-813F11), targeting 7q34). The probe was labeled with SpectrumOrange™ (Abbott Molecular/Vysis.Des Plaines, III), and a CEP7 control probe was labeled with SpectrumGreen™. A total of 100 nuclei were evaluated per case. Target to control probe ratios>1.2 were considered indicative of duplication.
The 14 NF1 associated astrocytomas included 10 PA, 3 LGSI and 1 DA. One NF1 associated GG was also examined. Patients included 6 females and 9 males; mean age at time of surgery was 20 years (range 2-44). Ultrastructural findings are summarized in Table 1 and illustrated in Figure 1. Aligned, often compacted, intermediate filaments typical of astrocytic neoplasms were identified in all tumors, including glial cells of the single GG. Dense core granules were present in the GG and 3(of 3) LGSI. Granules ranged from 100-130 nm in diameter, featured cores of uniform electron density, a surrounding halo, and a sharply delimited membrane. Rosenthal fibers were identified in 4 (of 10) PA. Interestingly, Rosenthal fibers, not evident at the light microscopic level, were seen ultrastructurally in the single diffuse astrocytoma and in 2(of 3) LGSI. Aligned microtubules were identified in 2 (of 3) LGSI. Macronucleoli were identified in 4 (of 10) PA, the single GG, and 2 (of 3) LGSI. Eosinophilic granular bodies were identified in 2 (of 10) PA, the single GG, and in 2 (of 3) LGSI.
In a prior study we described a subset of low grade NF1-associated astrocytomas that demonstrated some morphologic features of PA, but in addition had unique cytology . Clinicopathologic features of the tumors evaluated in the current study are summarized in Table 2. In summary, these tumors had some features reminiscent of PA, including relative circumscription, hyalinized/glomeruloid vasculature, a loose texture, and bipolar processes in a subset of cells. However, they were characterized cy-tologically by abundant eosinophilic cytoplasm and prominent nucleoli (Figure 2). Rosenthal fibers were absent, as well as eosinophilic granular bodies except for one case. In addition, they were predominantly supratentorial (5 of 6 cases), and mostly situated in the cerebral hemispheres (4 of 6 cases), unlike NF1-associated PA, of which 5 (of 10) involved the optic pathways or brainstem (Table 1).
We compared changes in global mRNA gene expression between 1 LGSI with the cytologic traits described above and 5 PA. A total of 9035 transcripts were differentially expressed between the LGSI and the PAs. The top overexpressed genes in the LGSI were enriched for neuronal specific genes, including ones encoding proteins associated with synapses, microtubules, secretory granules and neurotransmitters (Table 3). Of note the top 3 genes with increased expression were GABA A receptor, synaptotagmin 1, and neurogranin which are all neuronal associated[12-15]. They were increased 85, 62, and 49 fold respectively in the LGSI as compared to the PA group. There were several underexpressed transcripts, including a variety of genes encoding extracellular matrix related proteins (e.g. collagens) among others.
Next, we performed immunohistochemistry using commercially available antibodies for routine diagnostic use. Both PA and LGSI demonstrated strong expression of glial markers, including GFAP and S100 protein. In addition, LGSI showed more frequent expression of neuronal-associated markers compared to PA, including synaptophysin (2 of 4 vs. 2 of 6, respectively), chromogranin A (1 of 4 vs. 0 of 5, respectively), and neurofilament protein (2 of 4 vs. 0 of 4, respectively) (Figure 3 and Table 4). The two LGSI expressing neurofilament protein in tumor cells also expressed synaptophysin. In the 2 PA expressing synaptophysin, it was limited to rare cells. The GG showed strong expression of synaptophysin and chromogranin in the neuronal component. Partial infiltration of underlying brain parenchyma was noted in 5 (of 6) PA and 3 (of 4) LGSI. These findings support a predominantly glial and partial neuronal phenotype for NF1-associated LGSI compared with PA.
Given the increased cytoplasmic size and the presence of macronucleoli in the LGSI group, we hypothesized that increased mTOR signaling may partially explain this phenotype. We performed western blot analysis using an antibody recognizing phospho-S6, a marker reflecting mTOR activation. Increased levels of phospho-S6 were present in 3 NF1-associated low grade astrocytomas (1 LGSI and 2 PA) compared to cerebral cortex and cerebellum, levels being highest in the LGSI, when adjusting for beta actin or total S6 (Figure 4).
Next we performed immunohistochemistry using antibodies recognizing phospho-S6 and phospho-p70S6K in formalin-fixed paraffin-embedded tissue. Immunopositivity for both was noted in 7 (of 12) and 5 (of 13) PA and 4 (of 4) of LGSI each respectively (Figure 5). Combined immunoreactivity for phospho-S6 and phospho-p70S6K was noted in 5(of 13) of PA and 4 (of 4) LGSI, the difference being statistically significant (p=0.02) (Fisher exact test). To assess the status of mTOR activation in additional non-NFl associated tumors, we also performed immunohistochemistry using two tissue microarrays including sporadic tumors. Combined immunopositivity for phospho-S6 and phospho-p70S6K was demonstrated in 13 (of 39)(33%) sporadic PA.
In contrast there were no differences with respect to phospho-ERK immunoreactivity, which was uniformly expressed in all NF1-associated PA and LGSI, or in MIB1 labeling (mean MIB1 3.2 and 5 respectively, p=0.2). BRAF duplication was absent in 8 NF1-associated low grade astrocytomas, including 7 PA and 1 LGSI. When reviewing gene expression data focusing on mTOR target genes, the results were variable and difficult to interpret, with some genes showing increased expression and some decreased expression in the single LGSI compared to the NF1-associated PA. Genes overexpressed 2-3 fold in the LGSI included those encoding for eukaryotic translation initiation factor 4E, eukaryotic translation initiation factor 3, subunit B, and RHEB. These findings suggest that increased mTOR signaling, but not MAPK/ERK activation or proliferation, may underlie the phenotypic variations noted in NF1-associated low grade astrocytomas.
NF1 associated low grade astrocytomas are predominantly pilocytic in type. Prior studies have indicated that a subset of low grade astrocytomas are difficult to classify, but offer no explanation of phenotypic variations in terms of possible underlying molecular mechanisms. As would be expected, the most conspicuous ultrastructural finding within all tumors in this study was the presence of numerous intermediate filaments, typical of astrocytic neoplasms. In our study, the majority of tumors available for ultrastructural review were PA. Most demonstrated typical features one would expect, such as Rosenthal fibers and intermediate filaments. In the GG intermediate filaments were present only within phenotypically glial cells. As expected, dense core granules were abundantly present in the GG, but were also sparsely represented in all three LGSI examined. Rosenthal fibers not appreciated on light microscopy were present ultrastructurally in LGSI as well as the single diffuse astrocytoma, thus suggesting morphologic similarities, at least at the ultrastructural level, among all low grade astrocy-tomas in NF1. Instead, the presence of ultrastructural features usually associated with a neuronal phenotype, including aligned micro-tubules and dense core granules, was typical, albeit an unexpected feature, of LGSI. These findings within low grade astrocytomas may represent phenotypic divergent differentiation in the setting of NF1.
A partial neuronal phenotype was also demonstrated by gene expression analysis (in a single tumor) and by increased expression of neuronal markers at the protein level. However, the predominant line of differentiation of these tumors remains glial, as reflected by strong consistent GFAP immunoreactivity and intermediate filament accumulation in all tumors. Classic neuronal and glioneuronal tumors have also been described in the setting of NF1, including gan-gliogliomas [1, 16], DNET like lesions  and even desmoplastic infantile gangliogliomas . Unlike these lesions, our LGSI subset exhibits a predominant glial phenotype with only partial divergent neuronal differentiation, as described in other astrocytomas, including subependymal giant cell astrocytoma and pleomorphic xan-thoastrocytoma [17, 18].
The morphology of this low grade astrocytoma subset is somewhat reminiscent of another syndrome-associated tumor, i.e. subependymal giant cell astrocytoma. The features include nucleolar prominence, abundance of cytoplasm, expression of neuronal markers, and favorable prognosis . Subependymal giant cell astrocytoma is nearly restricted to patients with tuberous sclerosis, a syndrome arising as a consequence of TSC1 or TSC2 germline mutations. The molecular consequence is activation of the mTOR pathway . mTOR activation may also be an important biological event in epilepsy associated glioneuronal tumors .
Increased mTOR activation has been shown to occur in NF1 deficient astrocytes and in 4 of 6 NF1-associated PA in a prior study . Curiously, in our group of NF1-associated low grade astrocytomas mTOR activation, as reflected by phospho-S6 and phospho-p70S6K levels, was frequent, but seemed to be higher in the low grade astrocytomas with peculiar cytologic features. Phospho-S6 levels are increased as a result of activation of mTOR which leads to phosphorylation of p70S6 kinase and thus activation of ribosomal S6, which leads to increased protein synthesis. The result may be increased cell size. These findings support the notion that some low grade astrocytomas represent a distinct morphologic subset of NF1-associated tumors. Although no obvious clinical differences are noted with respect to NF1-associated PA, our findings could be clinically relevant since the mTOR pathway may be pharmacologically targeted for therapeutic benefit .
MAPK/ERK activation, as demonstrated by phospho-ERK immunoreactivity, was uniform in all tumors studied. This is expected, given that NF1 loss results in persistent activated RAS and downstream signaling through the MAPK/ERK pathway. BRAF rearrangements and mutations represent an alternative mechanism of MAPK/ERK activation in sporadic PA . It is of interest that prior studies have found BRAF alterations to be absent in NF1-associated astrocytomas [22, 23], including 7 PA and 1 LGSI used in the current study. Therefore, BRAF testing is not useful in distinguishing these subtypes of low grade astrocytoma in the setting of NF1.
One caveat is that not all LGSI demonstrate this unique morphology, and some may in fact represent other, inadequately sampled histologies, such as PA . Indeed, the possibility of under-sampling resulting in an erroneous diagnosis is a consideration whenever tumors with overlapping morphologic and ultrastructural features are encountered.
Our study suggests that NF1-associated PA and LGSI share some phenotypic features at the ultrastructural level, particularly Rosenthal fiber formation. However, the presence of occasional dense core granules and microtubules in addition to increased expression of neuronal associated genes, may represent divergent differentiation of some low grade astrocytomas in comparison with classic PA, the most prototypic NF1 associated astrocytoma. Furthermore, increased mTOR activation may be responsible for larger cell size despite similar proliferative indices and clinical behavior. Our findings, perhaps underlying the unconventional morphology of these tumors, require further studies to shed light upon the biology and molecular mechanisms underlying NF1-associated glial tumorigenesis.
The authors thank the microarray, cytogenetics, tissue and cell molecular analysis of the Mayo Clinic for excellent technical assistance.
This work was funded in part by grant P50 CA108961 from the Mayo SPORE in Brain Cancer (CG, FJR, JNS) and Mayo Clinic CTSA (FJR) through grant number UL1 RR024150 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH).