Genomic aberrations in pediatric diffuse intrinsic pontine gliomas

Abstract

Diagnostic biopsy is not routinely performed for children with diffuse intrinsic pontine glioma (DIPG). Consequently, our understanding of DIPG biology is hindered by limited tissue availability. We performed comparative genomic hybridization (CGH) on autopsy specimens to examine the feasibility of determining DNA genomic copy number aberrations on formalin-fixed, paraffin-embedded (FFPE) blocks. Histology on FFPE blocks obtained from autopsy of pediatric patients with DIPG was reviewed. Regions were marked for processing, and DNA was extracted from the tissue core, labeled by chemical coupling with Cy5, and hybridized to 105K oligonucleotide CGH arrays. After hybridization and washing, arrays were scanned, and data segmented and processed with Nexus software. Twenty-two samples from 13 subjects were obtained. Histologic variability was noted. CGH was successfully performed on 18 of 22 samples, representing 11 of 13 subjects. All demonstrated DNA copy number abnormalities. High copy number amplification of known oncogenes and homozygous deletions of known tumor suppressor genes were observed. Additional regions of high copy number amplification and homozygous deletion and geographical variations in the CGH patterns were found. CGH performed on FFPE tissue obtained from autopsy yields satisfactory results. This sample set from patients with DIPG was highly informative, with the majority of specimens showing ≥1 abnormality related to a known cancer gene. Aberrations in candidate drug targets were observed. This study establishes the feasibility of genomic DNA analysis from DIPG autopsy material, identifies several targets for which molecular targeted therapy exists, and suggests significant heterogeneity among patients with DIPG.

Diffuse intrinsic pontine gliomas (DIPG) represent 75%–80% of pediatric brainstem tumors and account for 10%–15% of all pediatric central nervous system (CNS) tumors. Because of the tenuous tumor location, diagnostic biopsies are not routinely performed, and patients receive a diagnosis on the basis of classic MRI findings and typical clinical presentation. Most patients are treated with radiation therapy, but patients face a poor prognosis, with median survival time of <1 year from diagnosis. No chemotherapeutic agent has ever demonstrated significant efficacy in a clinical trial, but many patients are treated with chemotherapy, including investigational therapy with molecular targeted agents.

There has been no significant improvement in outcome for these patients in more than 3 decades. Consequently, the role of biopsy to obtain tissue for biologic studies has recently been debated.¹ Biopsy of these lesions is currently not routinely performed, and tissue that is available for study remains inadequate. The small number of studies on limited numbers of autopsy and biopsy samples suggests that pediatric diffuse infiltrative tumors are complex.^2–4 The majority of DIPGs are fibrillary astrocytomas and histologically resemble adult supratentorial World Health Organization (WHO) grades II-IV malignant gliomas.⁵ However, adult and pediatric high-grade gliomas (HGGs) differ in their molecular genetic aberrations and altered gene expression.^4,6–8 In addition, pediatric HGGs located supratentorially differ biologically from those located in the brainstem, specifically the pons.⁹ Recent key genomic studies have demonstrated potential therapeutic targets, such as PARP1 and PDGFRA, in subgroups of DIPGs.^4,10 A number of molecular targeted agents aimed at disrupted pathways, particularly those identified in adult HGG, have reached clinical trials for children with DIPG, but none have shown efficacy in this population.^11–13

To overcome the limited tissue resources, we sought to evaluate alternate sources of DIPG tissue on which to perform biologic studies. Archived DIPG tissue, such as formalin-fixed paraffin-embedded (FFPE) blocks, represents a largely untapped resource. We evaluated the feasibility of using comparative genomic hybridization (CGH) to obtain genomic information from FFPE blocks.

Materials and Methods

Subject Samples

FFPE blocks prepared from autopsy of children with DIPG were obtained from 4 institutions. Samples were collected as approved by the Office of Human Subjects Research. Pediatric patients with DIPG were identified by the home institution and anonymized for this analysis; therefore, limited clinical data beyond age and diagnosis were available for subjects from outside institutions. (Table 1). Sections were cut for hematoxylin and eosin (H & E) staining, and the slides were reviewed by a neuropathologist. Areas of interest were marked, and a punch biopsy was performed on the corresponding area of the FFPE block with use of a 0.6-mm tissue microarray needle. For tumor samples demonstrating significant geographic histologic heterogeneity, specimens were obtained from more than 1 area (Table 2).

Table 1.

Subject Characteristics

Subject	Age, years	Sex	Clinical history
1	14	F	Incidental finding
			no XRT; ALK
2	4	F	NA
3	13	M	XRT-induced DIPG s/p MBL
4	NA	NA	NA
5	NA	NA	NA
6	NA	NA	NA
7	NA	NA	NA
8	13	F	XRT + ALK
9	14	F	XRT, I
10	11	M	NA
11	NA	M	h/o ALL
12	4	M	XRT, I
13	7	M	NA

ALK, alkylating agent; ALL, acute lymphoblastic leukemia; I, pegylated interferon; NA, not available; XRT, radiation therapy.

Table 2.

CGH Feasibility

Subject	Arrays	CGH performed	Quality	WHO Grade	Pathology
1	1_A	Y	P	II	Low-grade (II)
	1_B	Y	P	IV	High-grade (IV)
	1_C	Y	P	IV	Palisading necrosis
2	2_A	Y	B	III	II/III
3	3_A	Y	P	IV	Primitive cells
4	4_A	Y	F	II	Increased cellularity, minimal vascular proliferation
	4_B	Y	B	IV	Large cells with little/no cytoplasm
	4_C	Y	F	IV	Sarcomatous appearance
5	5_A	N	ND	III
	5_B	N	ND	III	Radiation necrosis
6	6_A	Y	P	IV	vacuolated
	6_B	Y	F	IV	increased vascular proliferation
7	7_A	Y	B	II	Tumor, reactive astrocytes, radiation necrosis
8	8_A	Y	P	III	gemistocytic component
	8_B	Y	P	II/III	round cells
9	9_A	Y	P	II	Radiation changes
10	10_A	Y	P	IV
11	11_A	Y	P	III	Calcifications present
12	12_A	Y	F	II	Cellular atypia, vascular, glomeruloid
	12_B	N	ND	IV	Endothelial proliferation, mitoses, pseudopalisading necrosis
13	13_A	N	ND	II	Primarily Grade II
	13_B	Y	P	IV	Small focus of Grade IV

Abbreviations: B, borderline (adequate for scoring high copy number amplification); F, fail; ND, no data (inadequate sample for CGH); P, pass.

CGH

DNA was isolated using a modified DNeasy blood and tissue kit (Qiagen). Male genomic DNA (Promega) was used as a reference and heat fragmented prior to labeling. Nonenzymatic Universal Linkage System (ULS) chemical labeling was used to directly couple fluorescent dyes with the sample and reference DNAs (1 μg each). Equal amounts of labeled sample and reference DNA were combined and hybridized to an Agilent 105k microarray for 48 h. After hybridization and washing, the microarrays were scanned in an Agilent scanner. Data were extracted using Feature Extraction software (Agilent) and analyzed in Nexus Copy Number software (Biodiscovery). The P value cutoff was .05. The microarray data has been submitted to the NCBI GEO database with the accession number (pending).

Immunohistochemistry

Immunohistochemical staining procedures were performed using a Dako autostainer and Ventana Benchmark XT (using Envision detection system and Ventana UltraView Universal DAB detection kit) on available FFPE sections using the following antibodies: EGFR (1:50; Zymed), p16 (1:200; Santa Cruz Biotechnology), and p53 (1:1000; Dako). Stains were interpreted as positive or negative (p53 and p16) or scored in a 4-tiered semiquantitative scale (EGFR: neg, 1+, 2+, 3+).

Results

Histology

All tumors were classified as glioma. Two, 4, and 7 tumors were graded as WHO grades II, III, and IV, respectively. There was significant variability among similarly graded tumors. For example, subject 10 appeared entirely grade IV, whereas subject 13 appeared primarily as grade II with a small area classified as grade IV. Tumors from subjects 1, 4, and 12 demonstrated clearly distinct areas of different tumor grade (Fig. 1).

Fig. 1.

Intrapatient tumor histologic variability on hematoxylin and eosin stain (20×) from subject 1 (A: low-grade area, B: high-grade area) and subject 12 (C: low-grade area, D: high-grade area).

CGH Feasibility

CGH was performed on 22 samples from 13 subjects (Table 2). Eighteen samples (86%) were successful, and 4 yielded insufficient DNA. Of the successful samples, 13 produced high-quality results; hybridizations from 5 samples were of suboptimal quality but were interpretable for high copy number aberrations.

Chromosomal Aberrations

All specimens demonstrated abnormal DNA copy number profiles consistent with multiple chromosome rearrangements. The cumulative data are plotted in Fig. 2, and results by sample are shown in Table 3. The most common region gained was 1q (7 of 11 subjects), and the most common region lost was 10q (7 of 11 subjects). High copy gains were observed on 1q (n = 2), 2p (n = 1), 4p (n = 2), 7p (n = 1), 10q (n = 1), and 13q (n = 1). Homozygous deletions were observed on 9p (n = 1) and 18p (n = 1).

Fig. 2.

Frequency plot indicating regions of common chromosomal aberrations. Gains are shown in green, losses in red. Gains of 1q, 7p, and 8q were most frequent; 10q was the most frequent region lost.

Table 3.

CGH Results by Sample

Subject	Arrays	10q loss	1q gain	PDGFRA	KIT	EGFR	MYCN	IRS2	MDM4	CDKN2A	PTEN
1	1_A
	1_B	X	X			G			G		L
	1_C	X	X			G			G		L
2	2_A	X (distal)	X			FHCG	FHCG		G
3	3_A			FHCG					FHCG	FHCL
4	4_B	X									FHCL
6	6_A	X	X						G		L
7	7_A	X		FHCG	FHCG						L
8	8_A		X			G		FHCG	G
	8_B		X			G		FHCG	G
9	9_A	X									L
10	10_A	X	X				G		G		L
11	11_A		X						G
13	13_B		X

Note: Empty cells indicate no change.

Abbreviations: FHCG, focal high copy gain; FHCL, focal high copy loss; G, gain; L, loss; X, present.

Genes in areas of aberrations included known oncogenes and tumor suppressor genes. Of note, focal high copy number amplification was observed in segments, including the known amplification targets PDGFRA, EGFR, CCND1, IRS2, MDM4, and MYCN. One subject (subject 8) had extremely high copy number gain on 13q and a second piece of amplification of the terminus containing the loci for IRS-2 and SOX-1. Homozygous deletions were observed in areas of known tumor suppressor genes, including PTEN and CDKN2A (Fig. 3). Subject 3 was notable for very high copy gain of loci, including MDM4, gain of 4q (PDGFRA), and deletion of 9p containing the p16 locus.

Fig. 3.

Representative areas of high copy amplification and homozygous deletions of known cancer genes, including (A) deletion of PTEN in subject 4, (B) amplification of EGFR in subject 2, (C) deletion of CDKN2A in subject 3, and (D) amplification of PDGFRA in subject 3.

Geographic variation in chromosomal aberrations was noted in subject 1, derived from a patient who did not receive radiation therapy. Histological review of the tumor revealed a clear demarcation between low-grade and high-grade areas. CGH was performed from both areas. The high-grade region contained several aberrations absent in the low-grade region, but shared a gain on chromosome 3q. The boundary of this segment of gain appeared to be identical within the resolution of the arrays used, suggesting a clonal relationship, given the similar break points (Fig. 4). Immunohistochemistry results are shown in Table 4.

Fig. 4.

Whole genome plot illustrating geographic variation in histologically low-grade and high-grade areas from subject 1. Note the shared aberration on 3q suggesting a clonal relationship between these regions, and additional aberrations observed on 1, 7, 10, and 14.

Table 4.

Immunohistochemistry

Subject	Area	EGFR	P53	P16Ink4a
1	Low grade	Neg	Neg	Neg
	High grade	1+; <5%
2		ND	Pos	ND
3		ND	Neg	ND
4		ND	ND	ND
5		ND	ND	ND
6		ND	ND	ND
7		ND	ND	ND
8		2+; 70%	Pos	Pos
9		1+; 50%	Pos (focal)	Rare Pos
10		Neg	Pos (focal)	Rare Pos
11		2–3+; 70%	Pos	Neg
12		1+; 40%	Neg	Pos
13		Neg	Neg	Rare Pos

Abbreviations: ND, not done.

Discussion

This report demonstrates that CGH analysis of FFPE blocks obtained from autopsy of children with DIPG is both feasible and informative. CGH studies on FFPE samples have historically been limited, because the DNA isolated from routinely processed FFPE blocks is frequently degraded into small fragments.¹⁴ Recently, the ULS has been shown to be a reliable labeling method for performing CGH on DNA derived from FFPE blocks.

In our study, DNA structural aberrations were frequent, occurring in areas of known oncogenes and tumor suppressor genes and additional areas. It is not surprising that those tumors appearing the most aggressive histologically (e.g., subject 10) also showed the highest number of chromosomal aberrations. This is consistent with studies performed in adults with supratentorial malignant gliomas that demonstrated an increased number of genetic alterations with increasing tumor grade.¹⁵ Copy number aberrations were identified in all samples but varied considerably between samples. Remarkably, a number of aberrations were identified in genes that are directly related to signal transduction pathways and that could potentially be targeted therapeutically.

The more frequent DNA structural aberrations included gain of 1q, gain of 7p, gain of 7q, and loss of 10q, consistent with prior studies in HGG.^16,17 Of significance, loss of 10q was noted in 7 of 11 subjects in our study. This locus is associated with the tumor suppressor gene, PTEN, which encodes a phosphatase that, among other effects, negatively controls activation of the PI3K pathway.¹⁸ In our sample set, 5 of 7 samples stained positive for p16Ink4a. The INK4A/ARF proteins are important in the control of growth arrest and senescence,¹⁹ suggesting that the PI3K pathway may be a significant target in DIPG. Gain of 4q, associated with the PDGFRA locus, was noted in only 2 subjects, in contrast to prior studies that demonstrated its presence in up to 17% of pediatric HGGs and up to 36% of DIPG.^4,9 PDGFRA, which is known to be expressed in malignant gliomas,^20,21 is found in distinct neural precursor cells during embryonic development;²² this is intriguing given that most DIPGs occur in children 5–10 years of age, suggesting an embryologic origin. Of note, focal amplification of PDGFRA was noted in subject 3 without coamplification of the closely linked KIT gene (in contrast to that in subject 7), again suggesting that PDGFRA may be an important player in DIPG.

Highly amplified genes in our sample set include EGFR, PDGFRA, and IRS2, all of which are related to the PI3 kinase pathway. Losses in the region of PTEN, including one instance consistent with PTEN homozygous deletion, further highlight the importance of this pathway in DIPG. Changes in tumor suppressor genes (e.g., deletion of CDKN2A) and gain of oncogenes (e.g., MYCN and MDM4) may significantly affect the responsiveness of tumors to therapeutic intervention. MDM4 plays a role in apoptosis and is known to inhibit p53/TP53-mediated cell-cycle arrest. The IRS proteins are required for the transforming ability of several other oncogenes and may be oncogenic themselves.²³ High copy amplification was observed in our sample set at the locus for IRS2.

There are essentially 4 articles in the literature evaluating the molecular genetics of DIPGs. Comparison regarding the frequency of aberrations may or may not be appropriate, because the studies differ in several ways. For example, the study by Louis et al.² included 7 brainstem tumors but did not subclassify these as DIPG, and 2 samples were from biopsy rather than autopsy. Zargooni et al.⁴ also included 2 pretreatment biopsies; although this study included 9 postmortem specimens notably obtained within 40 h of death, the time to tissue acquisition in our study and other published studies is not always known. The study by Barrow et al.,¹⁰ which used FFPE specimens from supratentorial HGG and DIPG, was limited to grade III and IV gliomas, and 10 of the 13 DIPG samples were obtained from pretreatment biopsies. The paper by Paugh et al.⁹ did not distinguish results from patients with DIPG and those with supratentorial HGG in each instance. However, despite these differences and despite the facts that these studies have a limited number of samples and extensive heterogeneity among patients, striking consistent findings include aberrations in regions containing TP53, PDGFRA, PTEN, and AKT3/pI3K, suggesting that these may be appropriate targets for therapy.

One of the most intriguing observations in this study is from subject 1, which showed geographic variation in a patient who did not receive radiation therapy. It is well known that progression of malignancy is a multistep process involving genomic DNA alterations. In this case, histologic review revealed very distinct areas of tumor grade with clear demarcation. With immunohistochemistry, the low-grade area demonstrated a low MIB-1 index and was negative for p53 and EGFR expression. The high-grade area showed a very high MIB-1 index, p53 was negative, and very focal areas of EGFR positivity were noted (Table 4). There were few genomic alterations in the low-grade glioma area. These, along with additional alterations, were present in areas of HGG, giving potential insight into the mechanisms underlying tumor development and progression. The primary genomic aberration observed in the low-grade area was gain at chromosome 3q. Although not found in all studies, gain of chromosome 3q, along with changes in chromosomes 7, 10, 9p, 19, and 13q, have previously been associated with glioblastoma progression.²⁴ This area of chromosome 3 encodes the PAK2 gene, a serine/threonine kinase associated with regulation of apoptosis. Additional aberrations in the high-grade area include gain of 1q, which contains the area encoding MDM4, another gene involved in apoptosis. Areas encoding the DNA repair proteins MGMT and PARP1 were also lost in the high-grade area but not the low-grade area.

We chose to evaluate CGH on FFPE DIPG samples because of the paucity of tissue available for study on these tumors, which, presumably, contributes to the lack of progress in gaining a better understanding of the biology of this disease. Our results support the impression that these lesions are complex and that significant inter- and intrapatient variability exists. The heterogeneity revealed by this study strongly suggests that clinical trials with targeted agents should be interpreted in light of the genetic aberrations present in the individual tumor, emphasizing the potential benefits of tumor biopsy. How to identify the ideal site to biopsy given the intratumoral variability remains to be determined.

In conclusion, this study demonstrates that genomic studies on FFPE tissue blocks obtained from autopsy of children with DIPG are feasible. Our observations also provide insights into the biology of this disease, revealing histologic heterogeneity among tissue samples and a striking diversity of copy number aberrations. Of importance, it reveals a number of potentially druggable targets. Although retrospective studies of autopsy samples represent an alternative to biopsy for exploring DIPG biology, the significant interpatient variability suggests that it would ultimately be necessary to evaluate individual tumors to accurately guide the choice of specific molecular targeted therapy. The DNA obtained from these specimens has the potential to be used in future DNA sequencing studies, and a larger scale study is underway.

Funding

This research was supported in part by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research.