Deep sequencing of multiple regions of glial tumors reveals spatial heterogeneity for mutations in clinically relevant genes

Background 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. Results 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. Conclusions 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. Electronic supplementary material The online version of this article (doi:10.1186/s13059-014-0530-z) contains supplementary material, which is available to authorized users.


Supp.
: a) Copy number calls from MIP captures using matched control tissue for samples BI04, BI06, BI07, BI08, BI09, BI10, BI11, BI12, BI14 and BI15, and a universal control (BI12) for samples BI01, BI02, BI05 and BI13 (as for the latter, matched control tissue was not available). Amplification indicates genes with coverage three-fold higher than median coverage across a sample. High Amplification indicates genes with coverage six-fold higher than median coverage across a sample. Notably, this analysis does not detect EGFR amplification in regions A, B and C of BI10 and region D of tumor BI15, events that were detected with the use of the universal control (BI12; see Figure 2 of main text). These EGFR amplifications were also detected using Taqman qPCR (Supp. Figure 3). Careful review indicates the reason for discrepancy is likely due to increased tumor contamination within the respective matched control tissues (region "X" in tumors BI04, BI07, BI08 and BI10 in Supp. Table 3). In those cases, the control tissue exhibited a high allele fraction of mutation in known cancer genes. For this reason, we chose to rely on the results of analysis where all tumors were matched with BI12 as a universal control. Panels b) and c) show the raw data used to call copy number using universal and matched controls, respectively. Tumor regions are shown on the xaxis with vertical lines separating regions from different tumors. Probes (grouped by gene) are shown on the y-axis. The color represents the read depth at each probe normalized against the median read depth across all other probes from the same tumor sample. Use of a "universal control" enables better detection of high-level EGFR amplifications in multiple regions of both tumor BI10 and BI15. We used high thresholds to call a gene as amplified (CN estimate>3). Blocks of higher signal may correspond to aneuploidy; however thresholds were not set for this sensitivity.

Supp. Figure 3: Validation of EGFR gene estimates
Correlation with of copy number estimates from smMIP vs. Taqman for EGFR. Taqman experiments were performed in duplicate for EGFR across all 62 regions investigated in this study. smMIP and Taqman copy number estimate were highly correlated with an R 2 of .90. Importantly, all highlevel amplifications of EGFR (delta Ct ≤ -2) were identified by the smMIP assay.
Supp. Figure 4: Validation of EGFR copy number by low-pass whole genome sequencing. DNA isolated from regions A-E in BI15 were subjected to light genome sequencing on the Illumina Miseq. Read depth within 1 Mb intervals across Chromosome 7 is normalized with respect to mean read depth across all chromosomes within each sample (see Supplementary Methods). Normalized read depth from whole genome sequencing within the 1 Mb region containing EGFR is highlighted in red within CN plots. Copy number of EGFR (WG_CN in table) from low-pass whole genome sequencing was compared with those estimates obtained using the MIP assay (MIP_CN). Regions A and B contain high-level amplification in the region containing EGFR while a similar amplification is not seen within regions C, D and E.
Supp. Figure 5: Measured EGFR amplification heterogeneity a result of varying levels of stromal contamination in BI15. a) GBM tumor used in dissection. b) Copy number estimates based on smMIP probe data. EGFR amplification (labeled) was called in regions A and B with only mild amplification detected in region C, D and E. Histologic examination and whole genome sequencing (Supp. Figure 6) suggested a marked decrease in tumor cellularity in regions C, D and E which likely accounted for the difference in copy number. c) and d) show representative FISH detection of EGFR amplification in region A (left images) and its absence in region E, respectively. Unprocessed images were obtained using a dual pass filter for spectrum orange and spectrum green and spectrum blue (DAPI). e) Validation of EGFR amplification in region A using single cell sequencing. Single cells from regions A and E were flow sorted, amplified and sequenced on the Illumina Miseq, resulting in 100,000 reads per sample. Copy number profiles were created by plotting read depth across the genome in 1 Mb intervals, with color of each genomic region corresponding to the number of mapping reads per interval. Four of seven cells from region A (15_A_2, 15_A_4, 15_A_6 and 15_A_7) have high level EGFR amplification while zero of seven cells in region E have similar amplification.

Supp Figure 6: Whole genome copy number profiles of regions A-E in BI15.
To identify other possible copy number alterations that may be shared across all tumor sections, DNA isolated from regions A-E from BI15 (shown as 15_a -15_e), the corresponding control region X (15_x) and two unrelated cell lines (NA12878 and HeLa) were subjected to light genome sequencing on the Illumina Miseq. 500,000 reads per sample were aligned to the hg19 reference and copy number is shown across the genome in 1 Mb intervals. Regions A and B of BI15 share gain in chromosome 7 loss of chromosome 10. However, no gross chromosomal aberration was shared across all tumor regions. Black line corresponds to the mean coverage across all 1 Mb windows in autosomes. Shaded regions correspond to the region 1 S.D. below and above mean coverage for each sample. Two cell lines (12878 and HeLa) derived from female individuals are shown for comparison. Chromosome X appears as lost in all regions from tumor BI15 (including control) of the tumor as it was derived from a male patient.

Supp. Figure 7: Sanger validation of TP53 and RB1 heterogeneity in tumor BI09.
We performed Sanger sequencing across three different loci from 5 regions of tumor BI09. All five regions (A-E) share mutations in IDH1. Tumor regions A and B have detectable mutations in TP53, while regions D and E have detectable mutations in RB1.

Supp. Figure 8: H&E and immunohistochemical (IHC) staining of p53 and IDH1 in tumor BI09
. IHC staining of p53 and IDH1 is shown across tumor regions A-E and corresponding control region X from tumor BI09. The pattern of staining differs across each of the five regions (A-E) and is consistent with the intratumoral heterogeneity identified with by sequencing. Partitioning of IDH1 photographs D and E illustrates that IDH1 heterogeneity was also present within these sections.
Supp. Figure 9: H&E and PDGFRα IHC staining of regions A and E in tumor BI05. IHC of regions A and E reveals differential staining of PDGFRα, with staining prominent in region A and not in region E. This is consistent with genomic findings: with amplification of the PDGFRA gene observed in regions A and B but not C, D or E. Original magnification 40x. Scale bar indicates 30 microns. EGFR IHC revealed robust expression across all regions (not shown). Supp. Table 3: Protein-altering candidate somatic mutations. Allele balance of protein-altering candidate somatic mutations across all tumor regions are shown. Tumors were divided into multiple regions (A-E) with a corresponding "control tissue" (X) available from 10 tumors. Candidate mutations were not previously observed in a database derived from >5,000 exomes from the Exome Sequencing Project (ESP) that had been modified to remove positions also found in COSMIC. In several cases (e.g. BI04, BI07, BI08 and BI10) the "control tissue" (X) exhibited a high allele fraction of mutation in known cancer genes (e.g. TP53). We concluded that this was the result of contamination of tumor cells within control tissue and subsequently used a "universal control" for calling copy number alterations (BI12). For several tumors, we did not sequence all five regions plus a control tissue. For these tumors, relevant regions are marked as "NA" or not available.