AMP 2019: Correlation Between Mutations Found in FFPE Tumor Tissue and Paired cfDNA Samples

Liquid biopsies represent a promising area of facilitating cancer research as blood collection is less invasive than tumor biopsies. Cell free DNA (cfDNA) consists of small (150 – 500 bp) DNA fragments that circulate in the blood. cfDNA levels tend to be low in healthy, non-pregnant patients, and increase in patients with cancer, pregnancy, or extensive damage to tissue. cfDNA is believed to be derive mostly from apoptotic cells for which biomarkers for a variety of diseases have been found in cfDNA.

• Apoptotic or necrotic cell death results in near-complete digestion of native chromatin from normal cell, tumor or fetus.
• Each 160-175 bp DNA is wrapped ~1.67 times around one nucleosome. These protein-bound DNA fragments preferentially survive digestion and are released into the circulation, and can be recovered from peripheral blood plasma as cfDNA.
• Typical cfDNA peaks characterized by Agilent 2100 Bioanalyzer, with a main peak at 175 bp, second and third peaks at 350 and 525 bp.

FFPE tissue is often used to look for cancer-associated mutations despite invasiveness; however it does not always correlate with the mutations seen in cfDNA. In this poster we present a comparison of matched FFPE and plasma samples to determine how many mutation are seen in both tissues. We also look at where the mutational mismatches appear in the chromosome. Different chromosomal regions can have different mismatch rates, and we use this to draw conclusions about the best chromosomal locations for biomarkers. We automated from extraction through sequencing in collaboration with Swift biosciences.

As cfDNA is extracted from blood, it is a non-invasive way to detect disease; however, there is some concern that cfDNA does not contain the same biomarkers as tumor tissue. Tumor tissue is typically removed and stored as formalin-fixed, paraffin-embedded tissue, a process that preserves the morphological structures well but chemically modifies and degrades the nucleic acids.

Methods

Sample Preparation

DNA was extracted from the FFPE tissues using FormaPure XL Total automated on a Biomek i7 hybrid. DNA concentrations of 4 10 uM curls were estimated using the Quant-iT Picogreen; yields varied between blocks. Some blocks had very low yields, most likely due to tissue distribution in the block. The extraction was repeated with 7 10 uM curls. DNA from FFPE was sheared on a Covaris S220 following the 200bp shear protocol. cfDNA was extracted from 1 mL of plasma using the Apostle MiniMax™ High Efficiency cfDNA kit. cfDNA yield varied between the samples as expected. Concentration was estimated using the Kapa hg-Quant kit. 

Library Construction

cfDNA and sheared FFPE DNA was processed with Swift Accel-NGS 2S Hyb DNA Library Kit. DNA input was 100ng. Following the library construction, the Swift Pan-Cancer Hyb Panel was used for gene enrichment. The hybridized libraries were sequenced on the Illumina NextSeq 550.

Analysis

The sequencing data was analyzed using the BWA enrichment tool on Illumina BaseSpace mapped to the genes in the panel. 

The two extractions, library construction and hybridization panel have all been automated on a Biomek Workstation.
The approximate time for an automated workflows are compared to manual work flows in
the table below.

Extraction Results and Sequencing Coverage

DNA extracted from FFPE was done both manually and on a Biomek i5 multichannel workstation. Both the manual user and the liquid handler extracted similar amounts of DNA (Figure 1 ); they are not significantly different as determined by a student t-test (P=0.79). Three of the blocks had very low yields and were extracted again. The initial low yield were most likely due to in block tissue distribution. In figure 2, the first 12 curls had much lower yield of DNA per curl than the second 14 curls did.

Below is a table of the sequencing coverage for all the samples. All samples had sequencing coverage of 50x at least 92% of the bases sequenced.

Majority of Variants are identified in both sample types

We identified both single nucleotide variants (SNV) and insertion and deletion (indel) events in the sequencing results from all the samples. First we compared the SNVs found in both sample types. The majority of the SNVs and indels were identified in both sampling methods and all tissue types (Figure 1 and Figure 2). In all of the samples unique SNVs and indels were identified in both cfDNA and FFPE DNA. More SNVs were identified when sequencing DNA from FFPE than cfDNA, however this difference wasn’t statistically significant. Conversely, more indels were identified when sequencing cfDNA and this difference is statistically significant.

We also wanted to verify that the variants that were found in both sample types were found at equal frequency. This was determined by comparing the allele frequency identified by sequencing cfDNA and DNA from FFPE tissue. We compared the ratio of the two allele frequencies. For all 8 pairs the ratio of cfDNA allele frequency to FFPE allele frequency was 1.1; the allele frequency for variants identified by cfDNA and FFPE were not significantly different than each other. Interestingly though. When looking at variants with allele frequency 2 standard deviations from the mean. cfDNA showed allele frequencies higher than FFPE for more variants.

Variant location

To determine if mutations were evenly distributed throughout chromosomes, each chromosome was divided into 10 bins; the first 10% of each of the 23 chromosomes were treated as one bin. The mutations were pooled into the 10 bins for each chromosome. This was done separately for mutations found in only DNA from only FFPE or cfDNA and found in both FFPE DNA and cfDNA.

The graph to the right shows how the mutations mapped across the 10 bins. A Pearson’s correlation was done to test for how differently the mutations mapped across the 10 bins. Mutations found only in FFPE tissue did not correlate (P=0.5) as well as mutations found only in cfDNA (P=0.8) to the position of mutations found in both FFPE DNA and cfDNA.

Variants Found at ClinVar Pathogenic Locations

We wanted to determine if we found any pathogenically relevant variants. To do this we compared against a list of variants from the NCBI ClinVar database. We only used variants that had the clinical significance of conflicting interpretations, uncertain significance, likely pathogenic, pathogenic and risk factor. We did this for all four of the tumor types. We were able to identify 2 variants in the ClinVar database only found using the DNA from FFPE tissue.

In most of the samples however, we were unable to find any variants identified in the ClinVar database. We were however, able to find variants at the same position as variants in the ClinVar database. These could also alter the encoded protein and would be good candidates for further study into phenotype. Interestingly all of these variants were found in the FFPE DNA sequences; one of them was only found in FFPE DNA sequences; higher sequencing coverage of the cfDNA could help to identify these variants.

Conclusions

Here we show:

• Sequencing of cfDNA captures the majority of variants that found in sequenced FFPE DNA
• More indels are identified using cfDNA than with FFPE DNA, especially with breast tissue
• Distribution of variants across the genome differs when sequencing FFPE DNA
• More previously identified clinically relevant variants, as identified by the ClinVar database were found when sequencing FFPE DNA

This study is small and further work should be done using larger data sets to gain more conclusive information.

Beckman Coulter makes no warranties of any kind whatsoever express or implied, with respect to these protocols, including but not limited to warranties of fitness for a particular purpose or  merchantability or that the protocol is non-infringing. All warranties are expressly disclaimed. Your use of the method is solely at your own risk, without recourse to Beckman Coulter. Not intended or validated for use in the diagnosis of disease or other conditions. These protocols are for demonstration only, and are not validated by Beckman Coulter.