The purpose of this paper was to identify the best library preparation and enrichment strategy for sequencing DNA from a small (mean area 5 mm2) formalin-fixed paraffin-embedded (FFPE) breast cancer specimen. The optimized strategy was then applied to compare the mutational burden among different 5 mm2 regions of ductal carcinoma in situ (DCSIS) specimens.
Mutational profiling of micro-dissected pre-malignant lesions from archived specimens.
Nachmanson D, Steward J, Yao H, Officer A, Jeong E, O'Keefe TJ, Hasteh F, Jepsen K, Hirst GL, Esserman LJ, Borowsky AD, Harismendy O
Purpose
Conclusion
DNA from a FFPE specimen was highly degraded with a DNA integrity number (DIN) of 2.4. When A-tailing libraries were prepared from ≥50 ng FFPE DNA, ≥ 20-fold coverage by whole exome sequencing was obtained for almost all targeted bases. However, when 10 ng FFPE DNA was used, the A-tailing library had a high read duplication rate (79% for whole exome sequencing and 83% for the cancer panel) and a low percentage of targeted bases with ≥20-fold coverage after removal of duplicates (17% for whole exome sequencing and 1.6% for the cancer panel). In contrast, single strand and blunt-end libraries generated from 3 ng FFPE DNA had a higher percentage of targeted bases with ≥20-fold coverage (63% and 84%, respectively for whole exome sequencing and 99.7% and approximately 45% for the cancer panel), despite having duplicate read rates of >60%. Fewer false positive somatic mutations were identified when libraries were constructed using blunt-end or single-strand priming than A-tailing.
There was a higher number of low (<5%) variant allele frequency (VAF) variants identified when DNA was isolated from the FFPE rather than frozen specimen. A bias toward C to T substitutions was observed among FFPE specimens and a bias toward C to A substitutions was observed among frozen specimens. After application of more stringent filtering based on the dataset from 200 ng DNA from FFPE specimens, the number of false positives was reduced and the variant recall increased to an expected level associated with low input. Detected copy number alteration (CNA) burden in FFPE specimens was generally consistent between low input (3 ng) blunt end and single-strand libraries and high input (200 ng) A-tailing libraries.
In the pre-malignant lesion (PML) specimens, CNA burden was generally consistent between specimens from a patient; as expected, copy number changes were identified in DCIS specimens but not normal breast epithelium. Of the 98 cancer genes evaluated, CNAs were identified in 38 genes (including ERBB2) in at least one DCIS specimen. The majority of mutations identified were Observed among multiple DCIS regions for a given patient. The mutational profile allowed the authors to build phylogenetic trees demonstrating clonal relationships.
Studies (1)
Purpose
The purpose of this study was to identify the best library preparation and enrichment strategy for sequencing DNA from a small (mean 5 mm2) FFPE breast cancer specimen. The optimized strategy was then applied to compare the mutational burden among different 5 mm2 regions of DCIS specimens. DNA was extracted from paired frozen and FFPE specimens from a HER2-positive breast invasive breast cancer specimen that had been stored for four years (details of specimen processing not included) using the QIAamp DNA FFPE tissue kit and QIAamp DNA Micro Kit. DNA integrity was assessed by capillary electrophoresis. Sequencing libraries were prepared using an A-tailing adapter ligation protocol optimized for low-input DNA and 10, 50 and 200 ng of DNA from the FFPE specimen or 200 ng of DNA from the frozen specimen. Additional blunt-end and single-strand libraries were prepared from 3 and 50 ng of DNA from the FFPE specimen. Libraries were subjected to whole exome enrichment with the Agilent SureSelect Clinical Research Exome panel or the Human All Exon V7 panel, and targeted enrichment using the Cancer All-In-One Solid Tumor panel; libraries were then pair end sequenced on a HiSeq 4000 machine. To validate the optimal sequencing method, premalignant lesions and normal breast epithelium were laser capture microdissected (average area of 5 mm2) from archived FFPE DCIS specimens. DNA was extracted after proteinase K digestion using the QIAamp DNA Micro Kit. Blunt-end libraries were prepared and sequenced on a HiSeq 4000 machine.
Summary of Findings
DNA from the FFPE specimen was highly degraded with a DIN of 2.4. A-tailing libraries prepared from ≥ 50 ng of DNA from the DCIS FFPE specimen had a coverage of 20X or greater for most targeted bases (>80%) after whole exome sequencing. However, an input amount of 10 ng DNA resulted in a 79% read duplication rate and, after duplicates were removed, only 17% of targeted bases had a coverage ≥20X. Conversely, whole exome sequencing of single strand and blunt-end libraries prepared from 3 ng of DNA (from the DCIS FFPE specimen) generated ≥20X coverage of at 63% and 84% of targeted bases, respectively, despite having duplicate read rates of approximately 70% and 60%, respectively. When the A-tailing library prepared from 10 ng FFPE DNA was enriched using the cancer panel prior to sequencing, duplication was high (83%) and only 1.6% of targeted bases were sequenced at a coverage ≥20X. Blunt-end libraries prepared from 3 ng FFPE DNA and enriched with the cancel panel had a slightly lower percentage duplication (73% versus 83%) and a much higher percentage of bases with ≥20X coverage (99.7% versus 1.6%) than A-tailing libraries prepared from 10 ng FFPE DNA. Despite having a high rate of duplication (approximately 80%), single-strand sequencing libraries prepared from 3 ng FFPE DNA had a high percentage of bases with ≥20X (approximately 45%), greater than observed in A-tailing libraries enriched using the cancer panel (1.6%).
The number of low VAF (<5%) variants was higher when DNA from the FFPE was sequenced compared to the matched frozen specimen (422,322 versus 175,364). DNA from the FFPE specimen had a bias toward C to T substitutions (53%), but DNA from frozen specimens had a bias toward C to A substitutions (52%). Fewer false positive somatic mutations were identified when libraries were constructed using blunt-end or single-strand priming than the A-tailing method (1,967 and 3,137, respectively, versus 7,475). After more stringent filtering (based on the dataset from 200 ng DNA from FFPE specimens) was applied, the number of false positives was reduced and variant recall increased such that it reached the expected level associated with low input. Detected CNA burden in FFPE specimens were consistent between low input (3 ng), blunt end and single-strand libraries and high input (200 ng) A-tailing libraries (7%, 8.1%, and 8.4%, respectively). Importantly, consistent CNAs were identified by whole exome sequencing and enrichment with a cancer panel that included the documented ERBB2 amplification, but in a few cases the breakpoint was missed using the cancer panel.
In the PML specimens, the percentage of duplicate reads ranged between 32-82% and was inversely related to DNA input. CNA burden was generally consistent between specimens from a patient, but copy number changes were identified in DCIS specimens as expected. Of the 98 cancer genes evaluated, CNAs were observed in 38 genes (including ERBB2) in at least one DCIS specimen. The majority of mutations occurred among multiple DCIS regions from a patient. The mutational profile allowed the authors to build phylogenetic trees demonstrating clonal relationships.
Biospecimens
TissueBreastPreservation Types
FrozenFormalinDiagnoses
NeoplasticCarcinomaPlatforms
DNANext generation sequencingDNAAutomated electrophoresis/BioanalyzerPre-analytical Factors
Classification Pre-Analytical Factor Value(s) Biospecimen Aliquots and Components Biospecimen heterogeneity Intratumoral sampling (exact positions not specified)Biospecimen Preservation Type of fixation/preservation Formalin (buffered)FrozenNext generation sequencing Specific Priming method A-TailingBlunt-EndSingle-StrandNext generation sequencing Specific Targeted nucleic acid Cancer gene panelWhole exomeNext generation sequencing Specific Template modification A-TailingBlunt-EndSingle-StrandNext generation sequencing Specific Template/input amount 10 ng200 ng3 ng50 ng