Mutation Profiling of Colorectal Cancer ctDNA using AmpliSeq CHP2 Cancer Panel Weihua Liu, Zhenyu Yan, Paul Labrousse, Peng Fang, Jennifer Biroschak, Cindy Spittle, Chad Galderisi and Jin Li. MolecularMD, Cambridge, MA and Portland, OR.

Introduction

Analysis of ctDNA in CRC has potential applications in early detection, monitoring response to therapy, tracking resistance and assessment of minimal disease. Emergence of resistance mutations in KRAS, NRAS, BRAF, EGFR and other genes after treatment with anti-EGFR mAb therapy can be monitored non-invasively by serial “Liquid Biopsy” to allow detection of tumor recurrence months earlier than conventional imaging. To this end multiple approaches have been developed including BEAMing, ddPCR and next generation sequencing. Here we use an amplicon based target enrichment method, AmpliSeq v2 (CHP2), to study a set of tumor/plasma pairs (n=13) collected from stage IV colon cancer patients to determine whether amplicon based NGS can be used to profile mutations in ctDNA from CRC patients. A data analysis pipeline to detect de novo ctDNA mutations was developed and trained using this dataset.

Methods13 matched tumor and plasma samples were purchased from Indivumed. Mutations were assessed using AmpliSeq v2 (CHP2) cancer panel, designed to survey 2800 mutations in 50 cancer-related genes. For mutation detection in tumor samples, we performed library prep in parallel for Ion Torrent and Illumina MiSeq using 10 ng DNA as shown in Figure 1. Plasma DNA sequencing was only performed using Ion Torrent PGM. Plasma DNA was extracted from 3 or 5 mL plasma using the QIAamp DSP circulating NA kit and eluted in 10 µL buffer per mL plasma. The sample input used in the CHP2 reactions was between 3.8 – 55 ng ctDNA. The background noise level of hotspot nt positions was calculated from 10 healthy donor plasma samples. A custom data analysis program was developed to detect variants clearly different from inter-run and intra-run background noise.

ResultsDetermination of ctDNA concentration: The ctDNA samples can be divided into two groups based on the DNA concentration measured by Qubit (range of 0.16-13 ng/ul): 6 with >1 ng/ul and 7 with <1 ng/ul (Figure. 2).

Somatic mutations found in paired tumor and plasma samples: All tumor samples had mutations in one or several genes e.g. KRAS, NRAS, PIK3CA, TP53, BRAF, APC, SMAD4 and FBXW7 (Table 1). The mutations found in the tumor samples by AmpliSeq v2 cancer panel were also confirmed by the Illumina MiSeq platform. Several germline SNPs were detected in PIK3CA, KDR KIT, JAK3, MET and CDKN2A gene and excluded from the concordance calculation between tumor and plasma. Eleven out of 13 ctDNA samples showed expected somatic mutations found in the matched tumors. Twenty eight out of 35 somatic mutations including 8 truncating mutations, mostly in APC and a 3bp insertion in SMAD4, were detected in 13 tumor/plasma pairs. Two ctDNA samples (highlighted in red) which failed to show the expected mutations were from patients who have poorly differentiated mucinous adenocarcinoma and received neo-adjuvant therapies. The sequencing read depth of these mutations was in the range of 1528-17364. The lowest Allele Frequency detected in the plasma was 0.06% with a 7130 read depth.

Background noise level: SMAD4 P356H in a ctDNA sample had AF 0.06% and 0.08% in two independent library prep and sequencing runs. The background noise level at this mutation site from wild type ctDNA samples is zero as shown by red arrows in Figure 2 below.

Conclusions AmpliSeq CHP2 cancer panel achieved 0.06-0.2% allele frequency sensitivity and 85% (11/13) tumor/plasma concordance, which are comparable with previous reports using alternative NGS panels.

We also the demonstrated the efficiency of target enrichment by AmpliSeq based multiplex PCR, which can detect as low as several mutant copies from wild type background.

For Further InformationPlease contact info@molecularmd.com or visit www.molecularmd.com.

Figure 2. ctDNA concentration in CRC plasma samples.

Table 1. Somatic mutations found in paired tumor and plasma samples

Figure 2. IGV screenshot for mutation SMAD4 P356H with AF 0.08%.

Table 3. ctDNA mutation pattern.

Reproducibility for mutations with Allele Frequency <0.2%: Four mutations (NRAS Q61L, KRAS G12C, PIK3CA Q546K and SMAD4 P356H) with Allele Frequencies (AF) between 0.06-0.2% were reproduced in a second PCR, library prep and sequencing run (Table 2).

Run1 Run2

Sample Name

Gene name

(HGNC) CDS Amino_acids Allele Freq (%)Read

Depth Allele Freq (%)Read

DepthA3575 NRAS c.182A>T p.Gln61Leu(61Q>L) 0.17 5299 0.10 4804A3608 KRAS c.34G>T p.Gly12Cys(12G>C) 0.08 6596 0.24 3680A3608 PIK3CA c.1636C>A p.Gln546Lys(546Q>K) 0.16 5044 0.14 2820A3669 SMAD4 c.1067C>A p.Pro356His(356P>H) 0.06 7130 0.08 15361

Table 2. Reproducibility for mutations with Allele Frequency <0.2%.

Gene Name Amino_acids A3107 A3214 A3575 A3605 A3608 A3615 A3650 A3669 A3687 A3721 A3822 A3972 A4573

APC

p.Arg1450X(1450R>*) 0.2 p.Arg876X(876R>*) 0.3 22.0

p.Gln1429X(1429Q>*) 1.2 1.4 BRAF p.Val600Glu(600V>E) 2.1

FBXW7 p.Arg387Cys(387R>C) 2.7

KRAS

p.Gly12Asp(12G>D) 1.4 19.3 30.1p.Gly12Cys(12G>C) 0.1 p.Gly12Val(12G>V) 1.7 p.Gly13Asp(13G>D) 31.6

NRAS p.Gln61Leu(61Q>L) 0.2

PIK3CA

p.Gln546Lys(546Q>K) 0.2 p.Glu545Gly(545E>G) 15.7 0.9 p.Glu545Lys(545E>K) 18.2 23.1

SMAD4

p.Glu330Gln(330E>Q) 0.9 p.Leu529dup(529L>LL) 0.8 p.Pro356His(356P>H) 0.1

TP53

p.Arg175His(175R>H) 2.6 p.Arg342X(342R>*) 0.2

p.Cys238Tyr(238C>Y) 0.4 p.Gln167X(167Q>*) 3.0

p.Gly245Ser(245G>S) 1.2 p.Gly266Glu(266G>E) 2.8

Results

Number of mutations grouped by AF in tumor vs plasma: ctDNA mutations can be grouped into 3 levels: allele frequencies >10% (N = 7), < 1% (N = 11) and between 1-10% (N = 10).

ctDNA mutation pattern: Co-existing mutations in the same tumor tend to be detected simultaneously in the matched plasma at the same level of allele frequency (Table 3). Two exceptions of this were that only driver mutations (NRAS Q60L and PIK3CA E545G) were detected in two ctDNA samples (circled in red), while mutations in other genes TP53, APC and GNAQ were found only in the matched tumors, probably due to differential roles of these genes in cancer progression and metastasis.

WT

WT

Gene AA change HDx ddPCR AF (%) NGS AF (%) NGS coverageAKT1 p.E17K 0.57 0.67 2555BRAF p.V600E 0.55 0.42 2358

CTNNB1 p.S33Y Not tested 0.78 2441EGFR p.G719S 1.15 1.17 7749KRAS p.G13D 0.44 0.42 4992NRAS p.A59T 0.42 0.50 10384

PIK3CA p.E542K 0.45 0.41 4858PIK3CA p.H1047R 0.50 ND ND

SMARCB1 p.? (c.1119-41G>A) Not tested 0.71 2387SMO p.V404M Not tested 0.26 6050

Evaluation of mutation standard for ct DNA: As mutation standard is needed for clinical assay, we evaluated Horizon Dx ctDNA 0.5% reference standard, which is fragmented DNA from cell line mixtures to mimic the characteristics of plasma DNA. 20 ng of Horizon Dx ctDNA 0.5% reference standard was analyzed using AmpliSeq CHP2 cancer panel. NGS results were compared to ddPCR results from HorizonDx. All mutations except PIK3CA H1047R mutation were detected at the expected AF (Table 4). The drop-off of this mutation from the HDx standard is under investigation. However, our previous data in breast cancer showed that AmpliSeq v2 was able to detect PIK3CA H1047R mutation at 0.6% AF. Several mutations were not tested by ddPCR but detected by AmpliSeq v2. These mutations were determined to be true because they were reported by Cosmic mutation database in the cell line used for preparing the standard. Table 4. Mutations detected in Horizon Dx 0.5% ctDNA reference standard using AmpliSeq CHP2.

Figure 1. In parallel library prep work flow for Ion Torrent and Illumina ready library based on AmpliSeq.

ResultsResults

AF> 10% AF 1-10% AF <1%