Whole Exome Sequencing analysis of Archival Formalin

Fixed Paraffin Embedded Tissue (FFPE): comparison of

different library preparation methods

Prashant K. Singh, PhD Assistant Director, Genomics Shared Resource

Roswell Park Comprehensive Cancer Center

Disclosure

I have no relevant financial or nonfinancial

relationships in the products or services described,

reviewed, evaluated or compared in this

presentation.

Outline

• Introduction

• Challenges with FFPE samples

• Different methods for NGS library prep

• Study design

• Results

Sample Procurement for Cancer research

Source: https://www.umassmed.edu/ccoe/core-services/tissue-and-tumor-bank/procurement-methods/

Formalin-Fixed, Paraffin-Embedded tissue

• FFPE procurement is the standard for tumor banking and

remains part of clinical standard of care.

• The tissue is preserved in formaldehyde, also known as formalin,

to preserve the proteins and vital structures within the tissue.

• Why FFPE?

– preserves the morphology and cellular details

– easier to cut slices of required sizes

– immunostaining and morphology

– tissue microarray

– long-term storage at ambient temperature

FFPE- Opportunities and Challenges

• These samples are an invaluable resource for clinical research.

• Characterized histological and pathological annotation combined

with extensive clinical data.

• Allow retrospective studies

– increase in number of samples

– follow up clinical data- disease outcome, treatment response and

survival

• Enables laser-capture microdissection (LCM) of specific cells or

tumor compartment to understand tumor heterogeneity and

microenvironment using NGS approaches

FFPE- Opportunities and Challenges

• Enables quantitative pathology and spatially profiling to detect multiple

biomarkers using multispectral imaging (e.g. Vectra), Mass cytometry

(e.g. Imaging Mass Cytometry) and NanoString-DSP

• Can be used with various imaging and genomic application

NanoString- DSP

FFPE- Opportunities and Challenges

• Formaldehyde, creates reversible cross-links between intracellular

macromolecules such as protein and DNA

• Crosslinks include protein–protein, protein–DNA, and DNA–formaldehyde

adducts and interstrand DNA crosslinks

• Crosslinking of DNA bases with nearby histones results in a conformational

change of DNA

• Formaldehyde-induced crosslinks of DNA reduce the stability of double-

stranded DNA

• Extensive cross-linking of proteins with nucleic acids, which leads to

fragmentation of DNA and RNA

FFPE- Opportunities and Challenges

• Major Factors contributing to DNA/RNA fragmentation:

– types of paraffin

– pre-fixation treatment

– fixation delay (i.e. perioperative ischemic time),

– the fixation process

– tissue preparation

– archival storage (age of FFPE)

FFPE- Challenges in NGS applications

• FFPE DNA has great potential for NGS analyses

• Targeted DNA sequencing to screen known actionable mutations in

FFPE samples have been successfully used

• The use of DNA extracted from FFPE for whole exome sequencing

(WES) is presently limited

– Unpredictable library prep

– Small insert sizes

– Low and/or uneven coverage

– High PCR duplicates

– Spurious variant calling

FFPE- Challenges in quality of RNA/DNA

FFPE RNA FFPE DNA

Hedegaard et al., PLoS One. 2014; 9(5):e98187

Clinical whole-genome sequencing from routine formalin-fixed, paraffin-embedded specimens: pilot study for the 100,000 Genomes Project. Genet Med. 2018 Feb 1. doi: 10.1038/gim.2017.241

FFPE- Challenges in NGS applications

FFPE samples presented a lower proportion of

regions reaching the targeted depth of 70X (0.351

versus 0.782) for the FF samples (Figure 1a)

The median standard deviation was between 1.1

and 5.8X higher in FFPE than matching FF

(p < 2.2e−16) meaning sequencing coverage was

less uniform in FFPE data.

Clinical whole-genome sequencing from routine formalin-fixed, paraffin-

embedded specimens: pilot study for the 100,000 Genomes Project.

Genet Med. 2018 Feb 1. doi: 10.1038/gim.2017.241

FFPE- Challenges in NGS applications

High FF–FFPE agreement was detected in

reliable regions whereas the agreement was

lower in regions of low complexity.

• Coverage depth was non-uniform, alignment metrics were suboptimal for FFPE data.

• High discrepancy between these data sets was dependent on the analytical tool set employed.

Clinical whole-genome sequencing from routine formalin-fixed, paraffin-embedded specimens: pilot study for the 100,000 Genomes Project. Genet Med. 2018 Feb 1. doi: 10.1038/gim.2017.241

Scope of Present Study

• Utility of FFPE samples for Whole Exome Sequencing (WES).

• The pre-capture PCR step within the hybridization-based target

enrichment protocol is a critical and can significantly affect the quality

of sequencing data and downstream analysis.

• To increase efficacy of template production we tested several different

methods to add adaptors (pre-capture PCR) within the WES protocol.

Study Design

• Samples: 2 patients derived FFPE (4 samples total)

• Pre-Capture PCR using 6 different library prep methods

• Target enrichment with Agilent SureSelect Human All Exon V6.

Basic workflow for NGS library preparation

Biotechniques. 2014 Feb 1;56(2):61-4, 66, 68, passim. doi: 10.2144/000114133. eCollection 2014.

Different adapters for NGS library prep

Results- Sample QC

• DNA extraction: Covaris truXTRAC FFPE DNA

• DNA QC

– Qubit

– TapeStation

– qPCR QC

Results- Sample QC

Sample ID Original (ng/uL)

ddCq

77-old 18.54 2.08

77-New 16.28 0.23

64-old 23.15 2.32

64-New 10.28 0.57

• New FFPE samples were of high quality compared to Old FFPE samples

• DNA was sheared with a Covaris S220 Focused-ultrasonicator

• Pre-capture libraries were prepared using Six different library preparation kits

• 100ng DNA for each sample used as input for library prep

Results- Pre-capture Library Yields

• Pre-capture PCR showed high yields for New FFPE samples for all the kits compared to old FFPE.

• SureSelect XTHS produced the highest yield.

• Swift kit showed most consistent yield for both Old and New FFPE samples.

• DNA restoration from NEB resulted in better yield compared to non-restored sample

ddCq Agilent XTHS

KAPA HyperPrep

Rubicon ThruPLEX

Swift Accel-NGS 2S Plus

NEBNext Ultra II

NEBNext Ultra II

restored

PCR cycle 12 10 12 12 10 10

77-New(FFPE1) 0.23 2580 1052 398 476 316 588

64-New(FFPE2) 0.57 1605 633 358 540 251 386

77-old(FFPE3) 2.08 762 325 161 464 189 296

64-old(FFPE4) 2.32 2100 355 139 378 278 355

Pre-capture library prep yield comparison (nanograms of DNA)

Results- % On-target

• The percentage of reads in the exome regions is a useful metric to evaluate target enrichment.

• We typically achieve a over 85% on target rate with V6 kit

• New FFPE samples showed better on target than old

FFPE samples

• SureSelect XTHS achieved 75-85% on target

• NEB kit showed higher on target after restoration.

• The Swift kit generated consistent on-target rates

from 57% to 59.4%

Average depth (X)

Results- Average Read Depth

Average read depth comparison at 100X sequencing depth (60 million reads), except where indicated

• To calculate Avg read depth all the samples were normalized to 60 million PE reads (100X)

• Significant viability between samples and kits

• New FFPE samples performed better than Old samples

• None of kits achieved expect 100X avg depth

• NEB and Agilent kits performed better.

Results- Percentage of Target Bases at 20X # Reads 1X 10X 20X 50X

FFPE1 60M 97.7% 94.6% 87.0% 51.7%

FFPE2 60M 97.3% 92.8% 85.1% 56.7%

FFPE3 60M 97.4% 92.8% 81.2% 29.5%

FFPE4 60M 97.4% 91.6% 80.3% 45.7%

FFPE1 60M 97.8% 95.3% 86.2% 40.3%

FFPE2 46.6M* 97.7% 94.2% 82.1% 28.1%

FFPE3 60M 97.4% 93.8% 82.6% 30.2%

FFPE4 60M 97.4% 87.2% 62.1% 19.7%

FFPE1 60M 97.7% 94.3% 82.4% 37.1%

FFPE2 60M 97.6% 94.3% 83.3% 33.4%

FFPE3 60M 97.2% 88.5% 65.5% 9.9%

FFPE4 60M 97.2% 83.3% 54.7% 10.6%

FFPE1 60M 97.8% 95.2% 84.9% 39.2%

FFPE2 51M* 97.7% 94.5% 82.3% 31.1%

FFPE3 45M** 97.3% 89.9% 70.5% 17.7%

FFPE4 60M 97.5% 91.1% 74.0% 28.7%

FFPE1 60M 97.2% 82.8% 62.7% 34.2%

FFPE2 60M 97.3% 88.0% 72.2% 41.4%

FFPE3 60M 96.7% 83.6% 65.4% 34.2%

FFPE4 60M 95.8% 72.2% 50.6% 27.1%

FFPE1 60M 97.1% 82.3% 63.5% 37.3%

FFPE2 60M 97.0% 83.9% 65.1% 35.5%

FFPE3 60M 96.5% 81.6% 62.3% 30.7%

FFPE4 60M 96.3% 77.3% 56.4% 31.1%

NEB-Restor

Agilent XT HS

KAPA

Rubicon

Swift

NEB

• Read depth of 20X is considered sufficient to result in a high-confidence call of the variant

• New samples performed better than old samples

• SureSelect achieved over 80% targets at 20X for all samples

• KAPA , Rubicon and Swift performed equally for New FFPE samples but Swift performed better with low quality samples (FFPE 3 and 4)

Summary

ddCq Pre-Cap

yield (ng) # Reads Avg Depth % on

Target 20X

Agilent

FFPE1 0.23 2580 60M 62 76.6 87.00%

FFPE2 0.57 1605 60M 71 84.1 85.10%

FFPE3 20.8 762 60M 39 76.1 81.20%

FFPE4 2.32 2100 60M 60 81.7 80.30%

KAPA

FFPE1 0.23 1052 60M 51 63.2 86.20%

FFPE2 0.57 633 46.6M* 41 63.2 82.10%

FFPE3 20.8 325 60M 42 61.7 82.60%

FFPE4 2.32 355 60M 34 46.1 62.10%

Rubicon

FFPE1 0.23 398 60M 49 63.3 82.40%

FFPE2 0.57 358 60M 44 54.9 83.30%

FFPE3 20.8 161 60M 28 52.2 65.50%

FFPE4 2.32 139 60M 26 47.5 54.70%

Swift

FFPE1 0.23 476 60M 51 58.9 84.90%

FFPE2 0.57 540 51M* 44 59.4 82.30%

FFPE3 20.8 464 45M** 33 58.5 70.50%

FFPE4 2.32 378 60M 42 57.0 74.00%

NEB

FFPE1 0.23 316 60M 52 63.6 62.70%

FFPE2 0.57 251 60M 65 74.2 72.20%

FFPE3 20.8 189 60M 52 69.7 65.40%

FFPE4 2.32 278 60M 43 56.9 50.60%

NEB restored

FFPE1 0.23 588 60M 60 71.2 63.50%

FFPE2 0.57 386 60M 56 67.7 65.10%

FFPE3 20.8 296 60M 46 65.1 62.30%

FFPE4 2.32 355 60M 52 66.8 56.40%

• There is significant variability among samples.

• Overall SureSelectXTHS performed better for most QC metrics.

• Swift kits produces consistent results for both new and old samples.

• DNA Restoration improved performance with NEB kit.

• Overall New FFPE samples performed better than old samples with all of the kits.

WES of Clinical sample

• Exceptional responder to Immunotherapy. • To identify neoantigens • WES on different histology from same

tumor

DNA from two different tissue blocks from same patient performed differently

Conclusions

• DNA quality of FFPE samples is critical for WES performance.

• Processing of tissue and age of FFPE affect performance.

• All 6 kits tested performed better with new samples compared to old

samples

• Overall SureSelectXTHS performed better for most QC metrics.

• Swift kits produced consistent results for both new and old samples.

• DNA Restoration improved performance with NEB kit.

Future directions

• Optimized FFPE processing

• Improve DNA/RNA extraction

• Improved library prep and hybridization

Acknowledgement

Team GSR Roswell Park Genomics Shared Resource Roswell Park Bioinfromatics Shared Resource

Agilent New England Biolabs‎