Deep sequencing of multiple regions of glial tumors reveals spatial heterogeneity for mutations in clinically relevant genes
- Akash Kumar1Email author,
- Evan A Boyle1,
- Mari Tokita2, 3,
- Andrei M Mikheev3,
- Michelle C Sanger4,
- Emily Girard4,
- John R Silber3,
- Luis F Gonzalez-Cuyar5,
- Joseph B Hiatt1,
- Andrew Adey1,
- Choli Lee1,
- Jacob O Kitzman1,
- Donald E Born5,
- Daniel L Silbergeld3,
- James M Olson4,
- Robert C Rostomily3Email author and
- Jay Shendure1Email author
© Kumar et al.; licensee BioMed Central Ltd. 2014
Received: 8 February 2014
Accepted: 4 November 2014
Published: 3 December 2014
The extent of intratumoral mutational heterogeneity remains unclear in gliomas, the most common primary brain tumors, especially with respect to point mutation. To address this, we applied single molecule molecular inversion probes targeting 33 cancer genes to assay both point mutations and gene amplifications within spatially distinct regions of 14 glial tumors.
We find evidence of regional mutational heterogeneity in multiple tumors, including mutations in TP53 and RB1 in an anaplastic oligodendroglioma and amplifications in PDGFRA and KIT in two glioblastomas (GBMs). Immunohistochemistry confirms heterogeneity of TP53 mutation and PDGFRA amplification. In all, 3 out of 14 glial tumors surveyed have evidence for heterogeneity for clinically relevant mutations.
Our results underscore the need to sample multiple regions in GBM and other glial tumors when devising personalized treatments based on genomic information, and furthermore demonstrate the importance of measuring both point mutation and copy number alteration while investigating genetic heterogeneity within cancer samples.
Regional heterogeneity of mutations has been observed in a variety of tumor types ,. This intratumoral heterogeneity has broad implications for the clinical management of cancer patients, especially in the current paradigm of personalized medicine based on genomic analysis of a single cancer biopsy. Within the context of primary brain tumors, several groups have previously identified heterogeneity of gene amplifications in genes EGFR and PDGFRA in glioblastoma multiforme (GBM) using fluorescence in situ hybridization (FISH) and array-comparative genomic hybridization on multiple regions within primary tumors ,. Despite the dropping cost of DNA sequencing, however, the extent of point mutational heterogeneity in brain tumors remains limited to a single case of GBM . This is in part because the investigation of intratumoral heterogeneity requires both sampling and deep sequencing of multiple regions in a tumor.
We recently developed a method to identify low frequency mutations across known cancer genes  using the single molecule molecular inversion probe (smMIP) assay, which combines multiplex target capture with single molecule tagging ,. Here, we extend this technique to detect gene amplifications and examine intratumoral heterogeneity by targeting 33 cancer genes across 62 spatial sections of 14 glial tumors, including 10 grade IV gliomas (all GBMs), three grade III gliomas (one each of ependymoma, astrocytoma, and anaplastic oligodendroglioma) and one grade II astrocytoma. We detected intratumoral heterogeneity in both point mutations and amplifications of genes implicated as glioma tumor drivers and therapeutic targets.
To identify high level gene amplifications in tumors, we compared read depth of smMIP-targeted regions in each tumor against that of a control tissue. As smMIP sequencing suggested that a subset of control tissues were contaminated with tumor cells, we performed analyses using either patient-matched controls (Figure S2 in Additional file 1) or a ‘universal’ control (Figure 2B). For the latter we selected control tissue from tumor BI12, as it appeared to have the least tumor contamination based on allele fraction of known pathogenic point mutations, and restricted copy number analyses to targets with >30× coverage in control tissue from BI12 as well as targets whose GC percentage ranged from 30 to 60% (n = 885 capture probes). A careful review of discrepant calls when using patient-matched versus a universal control indicated that use of the universal control was more sensitive in identifying bona fide amplification events (as confirmed with Taqman assays) secondary to the contamination of a subset of control tissues with tumor cells. After applying our filters (see Materials and methods), a total of 21 genes could be assayed in a total of 62 regions across 14 tumors (Figure 2B).
The ratio of coverage of each probe was calculated relative to the control tissue (from BI12). We used DNACopy  to segment genes and obtain R, the mean ratio of coverage relative to control for each gene. We estimated the copy number for each gene by dividing R for each gene by the median value of R across all genes for each tissue. Genes with ratios above 3 were called as amplified. Genes with ratios above 6 were called as highly amplified. We did not measure deletion of genes using this method.
This process identified five tumors with gene amplifications, with three having one or more regions with a highly amplified gene (Figure 2B). Three tumors had amplification of both PDGFRA and KIT, and three tumors had amplification of EGFR. We validated copy number estimates for a subset of calls using a variety of different methods, including Taqman quantitative PCR (qPCR; across all tumors for EGFR and tumors BI05, BI06 and BI15 for PDGFRA), as well as whole genome sequencing (in tumor BI15 for EGFR). MIP copy number estimates of EGFR were highly correlated (R2 = 0.90) with delta Ct obtained by Taqman qPCR when compared across all 62 regions sequenced (Figure S3 in Additional file 1). Additionally for five tumor regions of BI15 that were subjected to light-whole genome sequencing, EGFR copy number estimates were consistent between whole genome sequencing and smMIP techniques (Supplementary methods and Figure S4 in Additional file 1).
Tumors in which only a subset of regions possess an amplification or point mutation with no other mutation shared across regions can be the result of either mutational heterogeneity within a tumor or varying levels of tumor content between different tumor regions. As an example, tumor BI15 was called as amplified for EGFR in two out of five regions with no other somatic mutations/point mutations detected across the tumor (Figures S5 and S6 in Additional file 1). Upon close inspection of histologic slides prepared from adjacent tissue, the observed difference in amplification was most likely due to lower tumor cellularity within other regions of this tumor rather than intratumoral genetic heterogeneity. This was also seen in tumor BI04, where one region without detectable PDGFRA amplification also had lower frequencies of a TP53 mutation seen across all regions. For this reason, we chose to restrict our interpretation of intratumoral heterogeneity to tumors in which all regions also shared a point mutation or gene amplification. Three tumors met these criteria and are described below.
Spatial heterogeneity of TP53 and RB1point mutations
Spatial heterogeneity of PDGFRA and KITamplifications
Additional cases of heterogeneity are potential passenger mutations
A missense mutation in KRAS was observed at moderate allele fraction (10% of reads) in region D of the glioblastoma BI12 and was not detected in at least one other region (Table S3 in Additional file 1). As this mutation does not occur within known mutation hotspots and is in a tumor with signs of hypermutation (BI12), the clinical significance of this heterogeneity remains unclear. Other somatic point mutations are heterogeneous across an individual tumor but occur within genes that have another, ubiquitously distributed mutation. BI12 has missense mutations in PTEN that are observed in regions A, B and C and not in region D (Table S3 in Additional file 1). This tumor also has another high allele fraction mutation in this gene that is present across all regions of this tumor. A similar scenario is seen in the astrocytoma BI08. Regions D and E of this tumor have low allele fraction point mutations in TP53, but all regions share another high allele fraction mutation in the same gene. These results suggest that the heterogeneously observed mutations observed are more likely secondary passenger mutations that occurred after PTEN and TP53 were inactivated in tumors BI12 and BI08, respectively (Table S3 in Additional file 1). An alternative possibility is that tumors may contain distinct subpopulations in which genes were inactivated by a different mutation.
These results demonstrate that intratumoral spatial heterogeneity with respect to clinically relevant genes occurs among multiple types of brain tumors, and spans the mutational spectrum from copy number to point mutations. Across a set of recurrently mutated cancer genes (33 genes examined for point mutations, 21 genes for amplifications), we observed heterogeneity for clinically relevant mutations in 3 of 14 (21%) glial tumors. These include point mutations in TP53 and RB1 as well as amplifications in PDGFRA/KIT. All cases of mutational heterogeneity that we detected in a tumor occur in adjacent regions, consistent with the hypothesis that spatially distinct regions represent divergent subclones of a single tumor.
Historically, in anaplastic oligodendroglioma with intact 1p, mutations in TP53 were found to stratify outcomes, with median survival of 71 versus 16 months in patients with mutant versus wild-type TP53, respectively . While not specifically applying to our patient in whom 1p/19q is deleted, our data demonstrating discrete differences in TP53 status from different regions within an individual tumor nevertheless show the potential of genetic heterogeneity to confound the assignment of prognostication based on the detection of specific molecular markers.
In addition, decision-making regarding the use of receptor tyrosine kinase inhibitors could be influenced by the status of amplifications/mutations in PDGFRA. Our finding of regional heterogeneity of PDGFRA within tumors BI05 and BI06 confirms recent work by Sottoriva et al. and others and suggests that a single biopsy may not be sufficient to allow for informed application of targeted therapies against these presumed oncogenic drivers ,,. Clinical decision-making at recurrence will also likely be impacted by regional heterogeneity. Nickel et al.  compared mutations within a group of 10 genes from 2 regionally distinct samples of a single GBM at initial resection and 2 subsequent recurrences. No heterogeneity was detected at initial resection but heterogeneity of PIK3CA and PTEN mutation was detected at the first recurrence, and heterogeneity of PIK3CA, TP53 and EGFR mutation was detected at the second recurrence.
This study used tumor-adjacent brain to serve as a ‘control tissue’ to help identify somatic mutations. However, after extracting DNA and investigating mutations, we observed that in several cases the grossly normal brain tissue actually contained a large fraction of infiltrating tumor cells. This aspect of our study ultimately complicated the process of calling somatic point mutations (requiring filtering against databases of germline variation) and copy number (necessitating our use of a universal control). While blood would be the most appropriate material to use as a control we did not have institutional review board approval to collect blood at the time these experiments were conducted (subsequently modified to allow for blood collection in the future).
In this study we were able to identify amplification of only a subset of genes of interest, as some genes had too few probes to accurately determine copy number. Our study also did not detect genomic rearrangements and deletions such as the EGFR VIII deletion commonly found in GBM. However, one can imagine expanding this assay to consider amplifications and deletions with smMIPs by tiling probes at higher density and incorporating known SNP positions to aid in identifying cases of loss of heterozygosity. One could also capture additional glioma-relevant genes like IDH1 and IDH2 by adding probes targeting these genomic regions.
Our investigation focused on regional heterogeneity within a tumor, instead of the microscopic heterogeneity that is likely present within a given tumor biopsy. As we performed the smMIP assay on DNA extracted from tissue pieces that likely contained millions of cells, we would likely miss cases of heterogeneity where only a small population of cells within a biopsy contained a mutation (such as an amplification). Use of techniques such as immunohistochemistry, FISH and, more recently, single cell sequencing remain necessary to characterize the extent of microscopic heterogeneity in tumors.
These results validate the smMIP approach as a scalable and cost-effective platform for deep sequencing of cancer genomes to examine subclonal variation. Despite deeply sequencing multiple sections of 14 tumors, our survey required only one lane of sequencing on the Illumina HiSeq because we focused on well-known gene targets of mutation in cancer. In contrast to the technique used by a similar investigation , our method is also easily scaled and amenable to automation with samples processed in 96-well formats. This advantage in scalability enables one to easily assay many more regions (tens to hundreds) per tumor to obtain much finer scale pictures of intratumoral heterogeneity, as we are likely underestimating its extent even here by sampling of only a few regions. While our study represents an improvement over previous studies, analysis of greater number of genes in a greater number of tumors will be necessary to determine rates of regional heterogeneity in different driver mutations across GBMs.
We find multiple instances of regional heterogeneity in clinically relevant cancer genes within malignant gliomas at the time of diagnosis. We also demonstrate a scalable technique that can be used to efficiently characterize regional genetic heterogeneity for both point mutations and copy number alterations in tumors. Future challenges will include how best to interpret cases of intratumoral heterogeneity and test its impact in the context of clinical trials using targeted therapy approaches.
Materials and methods
Freshly resected brain tumor specimens from adult patients were obtained with informed consent as part of the Genomics Big Idea pilot program (UW/FHCRC). Tissue, patient demographics and final diagnosis were obtained in accordance with protocols approved by the institutional review board at the University of Washington. Tumors were divided into three to five regions, depending on size. Tissue from each region was then subdivided into four pieces for use in next generation sequencing, histology, cell culture and xenotransplantation (Figure 1). In 10 cases, brain grossly uninvolved by tumor was resected to provide adequate surgical access and was utilized as a source of germline or ‘control’ DNA to identify somatic mutations (Table S1 in Additional file 1). These tissues are referred to as regions “X” in this manuscript. For all samples, DNA was isolated from snap-frozen tissue pieces using the QIAGEN DNEasy Blood and Tissue kit (Qiagen, Venlo, Netherlands).
Targeted capture and sequencing
The smMIP assay was used to genotype candidate genes. Probes were previously designed by Hiatt et al.  against 33 genes that are commonly mutated in cancer (Table S2 in Additional file 1). Targeted capture and PCR amplification were performed as previously described, except that 200 ng of genomic DNA was used for each sample instead of 500 ng . After smMIP capture, amplified products were pooled and sequenced on a single lane of the Illumina HiSeq 2000 platform with paired 100-nucleotide reads and an 8- nucleotide index read.
Primary analysis and variant calling
Initial analysis steps through to read mapping were performed as previously described , except that instead of constructing a consensus read from tagged smMIP molecules, we chose one read per unique molecular tag event at random for subsequent analysis.
Variants were called using SAMtools, and were filtered for positions with phred base quality ≥30, ≥30× coverage and the absence of a neighboring homopolymer run of four bases or more (Table S2 in Additional file 1). To remove common polymorphisms and enrich for likely somatic mutations, we imposed a number of additional requirements, including requiring variants to be observed with an allele balance of at least 5% within a sample, removing variants present within a modified database of the Exome Sequencing Project  and 1000 Genomes  pilot project that had first been stripped of all COSMIC variants, removing variants that were present at an allele balance of at least 5% in two or more control samples.
Copy number analysis
We compared read depth of smMIP-targeted regions in each tumor against that of the control tissue BI12 to identify high level gene amplifications in tumors. We restricted the copy number analysis to targets with greater than 30× coverage in control tissue and a GC content ranging from 30 to 60%. To reduce the number of potential artifacts remaining, we removed from consideration (for the purposes of copy number analysis only) 12 genes (AKT1, AKT2, CDK4, CDKN2A, FGFR3, HRAS, KRAS, MYC, NRAS, SRC, STK11, and VHL) that had fewer than 15 probes with sufficient coverage in the control tissue (BI12).
After calculating the ratio of coverage for each probe relative to control tissue from BI12, we used DNACopy  to segment genes into discrete levels of coverage and obtain R, the mean ratio of coverage relative to control for each gene. We estimated the copy number for each gene by dividing R for each gene by the median value of R across all genes for each tissue. Genes with ratios above 3 were called as amplified and genes with ratios above 6 were called as highly amplified.
DNA from five regions of tumor BI09 were subjected to Sanger sequencing (Genewiz, South Plainfield, New Jersey, USA) against positions within IDH1, TP53 and RB1.
Copy number validation
Tumors with regional heterogeneity in EGFR and PDGFRA detected using smMIP sequencing were confirmed using Taqman qPCR analysis. DNA from each region was analyzed in quadruplicate using commercially available probes against PDGFRA (assay ID: Hs02749151_cn; Life Technologies, Waltham, Massachusetts, USA) and EGFR (assay ID: Hs07526740_cn; Life Technologies). Reference primers amplified a fragment from TERT (number 4403316; Life Technologies). Finally, to compare sensitivity of the smMIP approach, all regions from all tumors were assayed in duplicate for EGFR copy number.
Immunohistochemistry and FISH
Immunohistochemistry for IDH1 and p53 was performed on 4-micron paraffin sections using mouse anti-human p53 clone (1:2,000 dilution; DAKO, Glostrup, Denmark) and mouse anti-human IDH1 R132H (1:200 dilution; Dianova, Hamburg, Germany). All tumors were investigated for IDH1 mutation by neuropathology, while only a subset of tumors was investigated for p53 expression by immunohistochemistry (Table S1 in Additional file 1). Immunohistochemistry for EGFR and PDGFRA was performed on 5- to 6-micron paraffin sections using mouse anti-human EGFR, clone 2-18C9 (pharmDx kit, DAKO) and rabbit anti-human PDGFRα, clone D1E1E (1:500 dilution; Cell Signaling, Danvers, Massachusetts, USA). Dual-color EGFR FISH was performed using commercially available probes (LSI EGFR SpectrumOrange/CEP 7 SpectrumGreen, number 32-191053; Abbott Molecular, Chicago, Illinois, USA) with DAPI counterstain using standard methods. Slides were imaged using an Olympus DP72 digital camera mounted on a Nikon E400 microscope. Fifty nuclei were scored for each region. EGFR amplification was called if more than 10% of nuclei either contained many EGFR signals or exhibited a EGFR:CEP7 ratio greater than 2. 1p19q deletion FISH was performed using commercially available probes (number 04 N60-020, Abbot Molecular) using standard methods.
All sequence data from smMIP capture experiments have been deposited in the NCBI Sequence Read Archive (SRA) under accession number SRP049298.
fluorescence in situ hybridization
molecular inversion probe
polymerase chain reaction
single molecule molecular inversion probe
We would like to thank Stephen Salipante and Aaron McKenna for helpful discussions. We would also like to thank the patients and families who have contributed tissue for this project. This work was supported by institutional funds from the University of Washington and the Fred Hutchinson Cancer Research Center the Brain Tumor Research Fund of the Department of Neurological Surgery, and grants from the National Cancer Institute (CA160080 to JS; RO1CA136808 and CA136808 to RCR; R21NS082542 to AMM; CA155360 and CA114567-07 to JO) and the National Institute of Aging (AG039173 to JBH).
- Yachida S, Jones S, Bozic I, Antal T, Leary R, Fu B, Kamiyama M, Hruban RH, Eshleman JR, Nowak MA, Velculescu VE, Kinzler KW, Vogelstein B, Iacobuzio-Donahue CA: Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature. 2010, 467: 1114-1117. 10.1038/nature09515.PubMedPubMed CentralView ArticleGoogle Scholar
- Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E, Martinez P, Matthews N, Stewart A, Tarpey P, Varela I, Phillimore B, Begum S, McDonald NQ, Butler A, Jones D, Raine K, Latimer C, Santos CR, Nohadani M, Eklund AC, Spencer-Dene B, Clark G, Pickering L, Stamp G, Gore M, Szallasi Z, Downward J, Futreal PA, Swanton C: Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med. 2012, 366: 883-892. 10.1056/NEJMoa1113205.PubMedView ArticleGoogle Scholar
- Sottoriva A, Spiteri I, Piccirillo SG, Touloumis A, Collins VP, Marioni JC, Curtis C, Watts C, Tavare S: Intratumor heterogeneity in human glioblastoma reflects cancer evolutionary dynamics. Proc Natl Acad Sci U S A. 2013, 110: 4009-4014. 10.1073/pnas.1219747110.PubMedPubMed CentralView ArticleGoogle Scholar
- Snuderl M, Fazlollahi L, Le LP, Nitta M, Zhelyazkova BH, Davidson CJ, Akhavanfard S, Cahill DP, Aldape KD, Betensky RA, Louis DN, Iafrate AJ: Mosaic amplification of multiple receptor tyrosine kinase genes in glioblastoma. Cancer Cell. 2011, 20: 810-817. 10.1016/j.ccr.2011.11.005.PubMedView ArticleGoogle Scholar
- Nickel GC, Barnholtz-Sloan J, Gould MP, McMahon S, Cohen A, Adams MD, Guda K, Cohen M, Sloan AE, LaFramboise T: Characterizing mutational heterogeneity in a glioblastoma patient with double recurrence. PLoS One. 2012, 7: e35262-10.1371/journal.pone.0035262.PubMedPubMed CentralView ArticleGoogle Scholar
- Hiatt JB, Pritchard CC, Salipante SJ, O'Roak BJ, Shendure J: Single molecule molecular inversion probes for targeted, high-accuracy detection of low-frequency variation. Genome Res. 2013, 23: 843-854. 10.1101/gr.147686.112.PubMedPubMed CentralView ArticleGoogle Scholar
- Turner EH, Lee C, Ng SB, Nickerson DA, Shendure J: Massively parallel exon capture and library-free resequencing across 16 genomes. Nat Methods. 2009, 6: 315-316. 10.1038/nmeth.f.248.PubMedPubMed CentralView ArticleGoogle Scholar
- Venkatraman ES, Olshen AB: A faster circular binary segmentation algorithm for the analysis of array CGH data. Bioinformatics. 2007, 23: 657-663. 10.1093/bioinformatics/btl646.PubMedView ArticleGoogle Scholar
- Watanabe T, Nobusawa S, Kleihues P, Ohgaki H: IDH1 mutations are early events in the development of astrocytomas and oligodendrogliomas. Am J Pathol. 2009, 174: 1149-1153. 10.2353/ajpath.2009.080958.PubMedPubMed CentralView ArticleGoogle Scholar
- Lass U, Numann A, von Eckardstein K, Kiwit J, Stockhammer F, Horaczek JA, Veelken J, Herold-Mende C, Jeuken J, von Deimling A, Mueller W: Clonal analysis in recurrent astrocytic, oligoastrocytic and oligodendroglial tumors implicates IDH1- mutation as common tumor initiating event. PLoS One. 2012, 7: e41298-10.1371/journal.pone.0041298.PubMedPubMed CentralView ArticleGoogle Scholar
- Szerlip NJ, Pedraza A, Chakravarty D, Azim M, McGuire J, Fang Y, Ozawa T, Holland EC, Huse JT, Jhanwar S, Leversha MA, Mikkelsen T, Brennan CW: Intratumoral heterogeneity of receptor tyrosine kinases EGFR and PDGFRA amplification in glioblastoma defines subpopulations with distinct growth factor response. Proc Natl Acad Sci U S A. 2012, 109: 3041-3046. 10.1073/pnas.1114033109.PubMedPubMed CentralView ArticleGoogle Scholar
- Ino Y, Betensky RA, Zlatescu MC, Sasaki H, Macdonald DR, Stemmer-Rachamimov AO, Ramsay DA, Cairncross JG, Louis DN: Molecular subtypes of anaplastic oligodendroglioma: implications for patient management at diagnosis. Clin Cancer Res. 2001, 7: 839-845.PubMedGoogle Scholar
- Okada Y, Hurwitz EE, Esposito JM, Brower MA, Nutt CL, Louis DN: Selection pressures of TP53 mutation and microenvironmental location influence epidermal growth factor receptor gene amplification in human glioblastomas. Cancer Res. 2003, 63: 413-416.PubMedGoogle Scholar
- Fu W, O'Connor TD, Jun G, Kang HM, Abecasis G, Leal SM, Gabriel S, Rieder MJ, Altshuler D, Shendure J, Nickerson DA, Akey JM, Bamshad MJ, NHLBI Exome Sequencing Project: Analysis of 6,515 exomes reveals the recent origin of most human protein-coding variants. Nature. 2013, 493: 216-220. 10.1038/nature11690.PubMedPubMed CentralView ArticleGoogle Scholar
- Abecasis GR, Altshuler D, Auton A, Brooks LD, Durbin RM, Gibbs RA, Hurles ME, McVean GA: A map of human genome variation from population-scale sequencing. Nature. 2010, 467: 1061-1073. 10.1038/nature09534.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.