Discovery of structural alterations in solid tumor oligodendroglioma by single molecule analysis

Abstract

Background: Solid tumors present a panoply of genomic alterations, from single base changes to the gain or loss of entire chromosomes. Although aberrations at the two extremes of this spectrum are readily defined, comprehensive discernment of the complex and disperse mutational spectrum of cancer genomes remains a significant challenge for current genome analysis platforms. In this context, high throughput, single molecule platforms like Optical Mapping offer a unique perspective.

Results: Using measurements from large ensembles of individual DNA molecules, we have discovered genomic structural alterations in the solid tumor oligodendroglioma. Over a thousand structural variants were identified in each tumor sample, without any prior hypotheses, and often in genomic regions deemed intractable by other technologies. These findings were then validated by comprehensive comparisons to variants reported in external and internal databases, and by selected experimental corroborations. Alterations range in size from under 5 kb to hundreds of kilobases, and comprise insertions, deletions, inversions and compound events. Candidate mutations were scored at sub-genic resolution and unambiguously reveal structural details at aberrant loci.

Conclusions: The Optical Mapping system provides a rich description of the complex genomes of solid tumors, including sequence level aberrations, structural alterations and copy number variants that power generation of functional hypotheses for oligodendroglioma genetics.

Figure 1

An overview of the Optical Mapping system. A: Single cells, obtained from a slice of the tumor biopsy, are purified by a percoll density gradient, then mixed with agarose and allowed to solidify in a mold, forming rectangular inserts. Prior to mapping, cells are lysed within the insert, the DNA electrophoretically extracted, elongated and immobilized on an Optical Mapping surface by means of capillary flow through a microchannel device. The lower half of panel A is a representative image of properly elongated DNA (long, white horizontal lines) after surface digestion, stained with YOYO-1. Microchannels are 100 μm wide as indicated by the scale bar (grey bar). B: Enlarged image of surface-bound genomic DNA, digested with SwaI, showing discrete restriction fragments separated by gaps. C: Automated machine vision detects DNA molecules (pseudocolored green), and calculates the mass of each fragment (white numbers), creating ordered restriction maps (Rmaps) from single DNA molecules (yellow bars). D and E: Strategy for constructing a genome-wide optical map starting from single molecule Rmaps. Rmaps are first clustered on a restriction map generated in-silico from the reference sequence of the human genome by pairwise alignment. Then consensus optical map contigs are constructed by de novo assembly of the Rmaps from a given window. Finally, the consensus map contigs are aligned back to the reference map, and differences are identified.

Figure 2

Intra-tumor heterogeneity. A: Copy number profiles, inferred from analysis of optical map coverage of tumor HF087 and HF1551. For each panel, the x-axis is co-ordinates of the human genome (chromosome numbers are indicated at the top), and y-axis is counts of Rmaps that align to a particular genomic interval. The grey curve plots the observed number of counts in an interval, and the red line indicates the sequence of underlying copy number states (also called the Viterbi path). B: Copy number analysis of tumor HF1551 by slice. Slice 1 (green) has LOH of chromosomes 1p, 19q, 14 and 21, while slice 2 (blue) has losses of chromosomes 1p and 19 only.

Figure 3

Genome-wide distribution of optical structural alterations (OSAs) detected in oligodendroglioma. A: Horizontal yellow bars, numbered on the left, represent human chromosomes (heterochromatic regions are in grey). Tick marks depict locations of structural variants from HF087 (red) and HF1551 (blue). B: The total number of events detected in each tumor sample, also broken down by category. C: An example of each class of variant is shown in the inset figure.

Figure 4

Spectrum of genomic alterations in oligodendroglioma. A: ~800 single base alterations were found in each tumor, like the G > A transition in the STMN2 gene shown. B: A ~8 kb insertion from tumor HF1551. 179 such indels were detected per sample. C: A 352 kb inversion that disrupts pseudogene INTS4L1 and encompasses the zinc-finger transcription factor ZNF92. D: Loss of one copy of chromosome 1p, a hallmark of oligodendroglioma.

Figure 5

Size distribution of indels found by Optical Mapping. Histogram of indel sizes detected by Optical Mapping. The x-axis is size of indel in kilobase pairs, and the y-axis is number of events. For comparison, a similar graph is shown (inset) of the distribution of indel sizes from the Database of Genomic Variants.

Figure 6

Experimental validation of PARK2 mutation by PCR-sequencing. A: EC in PARK2 gene on chromosome 6 of HF1551. The enlarged figure shows the position of the PCR amplicon. B: Restriction digest of the PCR amplicon. The undigested amplicon is 848 bp. Digestion with SwaI restriction enzyme is expected to yield two fragments of 577 bp and 271 bp (based on the location of the EC in the optical map). An addition digestion was performed with NheI enzyme to ensure the correct amplicon was being analyzed. The expected sizes of the NheI fragments are 700 bp and 148 bp. C: Sequence of the PCR amplicon showing the G > T transversion that creates a new SwaI cut site.

Figure 7

Experimental validation of an EC in the STMN2 gene by PCR-sequencing. A: EC in STMN2 gene on chromosome 8 of HF087. The enlarged figure shows the position of the PCR amplicon. B: Restriction digest of the PCR amplicon. The undigested amplicon is 1003 bp. Digestion with SwaI restriction enzyme is expected to yield two fragments of 519 bp and 484 bp (based on the location of the EC in the optical map). An addition digestion was performed with NheI enzyme to ensure the correct amplicon was being analyzed. The expected sizes of the NheI fragments are 616 bp and 387 bp. C: Sequence of the PCR amplicon showing the G > A transition that creates a new SwaI cut site (blue triangle).

Figure 8

Candidate genes harboring OSAs in both HF087 and HF1551. A: NPAS3 which bears a complex alteration in HF087 and an EC in HF1551. B: OSBPL3 which harbors cut differences in both tumor samples.

Figure 9

Strategy for assembling the ~500 kb inversion in the Williams-Beuren region in HF1551. A: Construction of the modified, ‘hypothesis’ reference map for directed assembly. The map has a ~500 kb inversion in the center, flanked on either side by 500 kb of sequence that agrees with the reference map. B: After 8 iterations of map assembly, an optical consensus map is obtained that spans the hypothesized reference, and has multiple Rmaps that bridges across both left and right breakpoints.