Detecting Oncogenic Mutations in Whole Blood

WHITE PAPER Detecting Oncogenic Mutations in Whole Blood Analytical validation of Cynvenio Biosystems LiquidBiopsy circulating tumor cell (CTC) capture and next-generation sequencing (NGS) September 2013

Overview With the advent of next-generation sequencing (NGS), the pace of discovery in oncology has quickened and the possibilities of precision medicine are rapidly becoming reality. As a more detailed picture of the mutations that drive oncogenesis emerges, new avenues for oncology researchers, clinicians, and drug developers have opened up. However, practical and technical challenges remain that must be overcome to leverage the full potential of NGS and realize the promise of precision medicine. This white paper describes the validation of Cynvenio s LiquidBiopsy system, a precision medicine platform for capture and genetic testing of circulating tumor cells (CTCs) from whole blood. LiquidBiopsy Allows molecular analysis of solid tumors directly from blood Offers the opportunity to capture serial patient samples over time and track disease Delivers sequence data on over 4500 single-nucleotide variants (SNVs) in 50 oncogenes commonly mutated in breast, colon, prostate, and lung tumors Identifies mutations present in as few as 1% of CTCs in a heterogeneous tumor population Can be used to support clinical oncology decision-making, research, and trial management Numerous technical hurdles have up to now prevented routine recovery and sequencing of CTC genomic DNA (see Background, below). These barriers have been overcome with the LiquidBiopsy CLIA-certified services. Here we present data that validates the enumeration and capture of CTCs and resequencing of CTC genomic DNA without whole genome amplification and discuss relevant applications of the technology. 2

Background A critical barrier to routine use of NGS in oncology, particularly for solid tumors, is the difficulty of obtaining adequate high-quality DNA (or RNA) template for molecular analysis. Tissue biopsy has several limitations as a potential source of template for sequencing: It is obviously invasive, not always practicable, captures only a small sample of a complex tumor cell population at a single time point, and has little utility for monitoring changes in tumor properties over time or in response to treatment. Preservation of tissue through standard methods, such as formalin-fixed, paraffin-embedded (FFPE) samples, can also create problems for later analysis, including DNA loss and fragmentation (1). Monitoring tumor cells in blood offers an alternative or supplement to tissue biopsy. Although the existence of CTCs and their suspected role in metastasis dates back over a century, only recently has technology advanced to allow routine CTC isolation based on expression of specific cell surface markers for epithelial-derived tumors (EpCAM, CK, and absence of CD45) (2). With the advent of technology able to enumerate CTCs, the prognostic value of CTCs has been validated in several cancers, including breast, colon, and prostate (3-5), demonstrating that CTCs can be used as a relatively noninvasive method to monitor cancer, although studies comparing CTC and primary tumor mutations suggest the degree of concordance may vary with disease and stage (6, e.g.). While CTCs offer a window into tumor status in vivo, they are extremely rare cells, making them difficult to isolate in the quantity and purity required for molecular analysis. The potential clinical importance of rare clones within a tumor cell population for development of drug resistance and metastasis has long been recognized, and there is a growing appreciation of the heterogeneity of tumors in space and time. In this context, in order for genetic analysis to be meaningful, it must be exquisitely sensitive. Robust sequencing data from CTCs requires that they be isolated from non-target cells in sufficient quantity and purity to generate a signal greater than the background of non-target cells that are inevitably carried over into the purified cell fraction. While whole genome amplification can increase the quantity of isolated DNA template, this further degrades the signal-to-noise ratio, and this, until now, has precluded a robust analysis of genetic mutations in isolated CTCs (7). 3

Cynvenio Approach In response to the opportunity and challenges described above, Cynvenio has created a patented system called LiquidBiopsy to characterize genetic mutations in CTCs from patient blood samples. As diagrammed in Figure 1, the LiquidBiopsy workflow is composed of three main steps, each supported by custom technology developed at Cynvenio: 1. Blood sample collection - a blood draw is prepared according to package instructions using tubes provided in the LiquidBiopsy Patient Sample kit. The fixed sample is stable for 96 hours without refrigeration, allowing ample time for shipping to Cynvenio s CLIAcertified laboratory. 2. CTC isolation - a proprietary microfluidic platform is used to enumerate and isolate CTCs at high purity using magnetic capture of affinitylabeled cells. 3. Next-generation sequencing - targeted amplification and resequencing to a read depth of 2000-fold is applied to assess over 4500 single-nucleotide variants (SNVs) in 50 oncogenes. As detailed in the Validation section below, LiquidBiopsy is a patented, CLIA-certified service suitable for applications in patient care, research, and clinical trial management and has been shown to: Efficiently capture CTCs at high purity from whole blood Accurately and sensitively detect targeted oncogenic mutations, even those present in just 1% of CTCs Blood sample collection LiquidBiopsy Patient Sample Kit CLIA-certified laboratory service CTC isolation Keep sample stable 96h Patented platform with high yield and high purity Genomic sequencing CLIA-certified 50 genes, >4500 SNVs 1% sensitivity Read depth 2000 0.0014% False pos. Clinically actionable data Figure 1. LiquidBiopsy workflow 4

LiquidBiopsy Validation METHODS Sample collection. Whole blood is drawn and fixed with the LiquidBiopsy fixative (Cynvenio Biosystems, Inc., Westlake Village, CA) within 4 hours of collection according to package instructions. After fixation, blood is shipped to the processing lab using the pre-addressed FedEx shipping material provided in the kit. Samples are stable up to 96 hours after fixation and no dry ice is required for shipping. CTC labeling, enumeration, and recovery. Once the fixed blood sample arrives at Cynvenio s CLIA-certified laboratory, cells are spun out and CTC labeled with a ferrofluid-anti-epcam antibody conjugate. The anti-epcam-labeled CTC are isolated from the mixed population of cells in the total sample using a custom-built LiquidBiopsy microfluidics chip (Figure 2). Additional antibody conjugates can be used for affinity labeling when desired. In the cell spiking experiments, tumor cell lines, e.g. MCF7, were added to normal donor blood prior to fixing. The LiquidBiopsy system employs a three-layer laminar flow chip with PBS in the top compartment, labeled blood sample in the middle, and densityadjusted buffer on the bottom. Following the loading of the sample, cells are permeabilized and labeled in flow conditions with DAPI, pancytokeratin (anti-ck), and anti-cd45 antibodies. The EpCAM +, CK +, CD45 -, DAPI + cells are recovered in a single tube for further molecular analysis using a custom Cynvenio SpinElute tube. Following quality control steps, CTC are enumerated and recovered cells can be returned to the originator or analyzed in the Cynvenio laboratory by NGS, PCR, FISH, or other methods. Figure 2. CTC LiquidBiopsy capture. LiquidBiopsy microfluidic chip captures ferrofluid labeled CTCs (blue) by applying ultra-high magnetic gradients as the sample passes through a custom sheath flow cell. Purified CTCs, free of all but ~0.001% of normal blood cells (red), are readily recovered for further molecular analysis. 5

Table 1: CT-Seq 50 oncogene panel (AmpliSeq Panel v2) CT-Seq AmpliSeq Panel v2 ABL1 EZH2 JAK2 PTEN AKT1 FBXW7 JAK3 PTPN11 ALK FGFR1 KDR RB1 APC FGFR2 KIT RET ATM FGFR3 KRAS SMAD4 BRAF FLT3 MET SMARCB1 CDH1 GNA11 MLH1 SMO CDKN2A GNAQ MPL SRC CSF1R GNAS NOTCH1 STK11 CTNNB1 HNF1A NPM1 TP53 EGFR HRAS NRAS VHL ERBB2 IDH1 PDGFRA ERBB4 IDH2 PIK3CA Next-generation sequencing. CT-Seq is a CLIAcertified next-generation sequencing service that allows characterization of SNVs in CTC recovered on the LiquidBiopsy platform without whole genome amplification. DNA is released from the pellet using Cynvenio Digest buffer. Target-specific amplification of genomic DNA is conducted using targeted amplification (Ampliseq v.2 Cancer hot spot panel, Life Technologies) of >4500 mutations in 50 known oncogenes commonly mutated in tumor cells (Table 1). Following library construction, template DNA is sequenced on the Ion Torrent PGM platform (Life/Ion Torrent) to a read depth of 2,000, i.e. 20-fold coverage of a single nucleotide mutation present in 1% of cells. (LiquidBiopsy samples are also compatible with Illumina platforms.) Sensitivity. In validation experiments, tumor cell lines A549, MCF7, HCC1419, and H1975 were added in a defined ratio to non-target cells to determine accuracy and sensitivity of detection of SNVs using CT-Seq. The model tumor cell ratios were based on the published ploidies of eight different mutations contained within the tumor cell genomes: PIK3CA, EGFR, CDKN2A, KRAS, TP53, and STK11. 6

% Purity Cell Density (cells/ml) Cell Density (cells/ml) Non-Target cell (c/ml) DETECTING ONCOGENIC MUTATIONS IN WHOLE BLOOD RESULTS Standardized sample collection. Blood samples collected and fixed using the Cynvenio Patient Sample Kit consistently yielded high-quality DNA template for later analysis. This is in contrast FFPE samples, which are not prepared according to a single standardized method, resulting in variable template quality (see discussion in 8, e.g.). CTC Isolation: Efficient cell recovery with high purity. To validate cell recovery using the LiquidBiopsy system under controlled conditions, blood from normal subjects was spiked with the MCF7 tumor cell line, as described in Methods. Recovery of the eluted cells following labeling and separation on the LiquidBiopsy platform was linear at tumor cell densities from 10 to 1,000 cells/ml (Figure 3, top left). Purity of recovered cells pellets was 1-80% and depended on the concentration of tumor cells in the sample (bottom left). On average, fewer than 65 non-target cells/ml were carried over into the eluted cell fraction (top right). This low background allows detection and enumeration of as few as 10 CTC at 1% purity. Interassay variability of target cell recovery was shown to be minimal regardless of instrument or operator (bottom right). CT-Seq: Sensitive and accurate nextgeneration sequencing. Given the speed and low cost per sequencing cycle, we have embraced the Ion Torrent PGM platform. There are some challenges associated with ion detector sequencing. For instance, significant sequencing Linearity of Recovery Non-Target Capture Average < 65 cells/ml 1000.0 700 100.0 525 10.0 350 1.0 0.1 0.1 1.0 10.0 100.0 1000.0 Spiked Cell Density (cells/ml) 175 0 Separate Clinical Samples (N=195) + 2SD - 2SD 100 Linearity of Purity 100 Intra-Assay Variability 75 75 50 50 25 25 0 1 10 100 1000 Spiked Cell Density (cells/ml) 0 0 25 50 75 100 Spiked Cell Density (cells/ml) Figure 3. LiquidBiopsy CLIA validation summary. Incremental numbers of tumor cells (MCF7) were added to normal donor blood and 7.5mL aliquots were processed using the Cynvenio protocol. Recovery (top left) and purity (bottom left) were assessed using DAPI, cytokeratin, and CD45 staining to distinguish target (CD45 -, CK +, DAPI + ) from non-target (CD45 +, CK -, DAPI + ) populations. (Top right) Non-target cell capture (EpCAM +,CK -, CD45 +, DAPI + ). (Lower right) Interassay variability across different instruments and operators. 7

DETECTING ONCOGENIC MUTATIONS IN WHOLE BLOOD Table 2:! Validation summary Accuracy Precision Linear Range Sensitivity 100% over reportable range 100% over reportable range 0.12% - 50% allele frequency R2 = 0.9925 lower limit of detection 1.0% allele frequency noise is associated with PGM ion detector sequencing contributing high InDel error rate, and sequencing errors associated with strings of homopolymers. In addition, the platform supported Torrent Suite is designed to provide value for common templates either for calling SNPs based upon Mendelian models or for calling SNVs in somatic cancer models. In each case the templates are from samples with abundant amounts of tissue and known tumor representations. For somatic tumor modeling the Torrent Suite does not realistically permit calling below 5% mutation frequency. The Cynvenio CT-Seq test model does not fit in to these descriptors. First and foremost our sample type are cell pellets derived from whole blood. The cellular composition of these pellets is quite heterogenous. We do not use cellular or tumor composition as an predictor of CT-seq mutation frequency. Furthermore the noise associated with PGM sequencing affects our sequencing quality. Consequently we have developed proprietary software solutions to assist us in filtering our sequencing noise including the well known artifacts associated with homologous repeats. By virtue of these solutions and a case controlled approach to sequencing, we have validated our sequencing platform sensitivity to 1% with a false positive rate of zero over our Reportable Range. As summarized in Table 2, experiments with engineered samples spiked with a mixture of wellcharacterized tumor cell lines isolated by LiquidBiopsy (see Methods) demonstrated that CTSeq produces accurate, precise, and robust sequencing results with a sensitivity in the range required for detection of mutations in the small number of CTCs typically recovered from a single blood draw (~120 cells). The test proved exquisitely sensitive, with a lower limit of detection of 0.12% allelic frequency and an average threshold sensitivity of 1%. Resequencing of targeted mutations was 100% accurate and linear across multiplechromosomes, with a regression correlation of R2= 0.9925 over the reportable range from 1% to 50% allele frequency in the engineered cell population (Figure 4). Mutant frequency (%) 15 12 Figure 4. Linearity of mutant calls across all chromosomes. Sequencing performance and linearity were indexed using a model tumor cell mixture, as described in Methods. Each point (+) is the average mutant frequency detected in triplicate runs for eight different loci. 9 6 3 0 0 10 20 30 40 50 Tumor Cell Contribution (%) 8

Although sequencing sensitivity was in excess of 0.1%, actual sequencing sensitivity is affected by the clinical sample purity. Routinely, clinical samples have a cell purity that averages 20%. The actual limit of detection of SNVs is determined both by the absolute number of informative cells as well as their purity. Thus the lower limit of detection is approximately 1%. Patient sample results. Results to date with patient samples demonstrate the utility of LiquidBiopsy for applications in precision medicine. Experiments with blood samples from multiple patients with solid tumors of the breast, colon, prostate, or lung have show good recovery of CTC and the ability to detect mutations also found in the corresponding tumor biopsy. Across 236 experiments with four tumor types (prostate, lung, kidney, breast), a standard blood draw of 7.5 ml yielded an average of 120 cells with the expected CTC markers (EpCAM+, CK+, CD45-). The average purity of the recovered cells was 20% CK + cells (8). In one pilot study, CTC isolation and sequencing was performed on patient blood samples taken at different time points to monitor breast cancer markers. Twelve years after a masectomy to remove an ER +, Her2 - tumor, the patient presented with elevated tumor biomarkers but showed no evidence of tumor with a whole body CT scan. Baseline LiquidBiopsy detected two mutations in CTC, including an activating PIK3ca mutation that may predict sensitivity to PIK3ca inhibitors such as everolimus. Two weeks after initiating fulvestrant treatment, CTC number had decreased almost threefold. However, the frequency of PIK3ca mutation had increased by a similar amount (Table 3). These data illustrate how genomic analysis of the CTC population can be applied to detect, characterize, and monitor disease. Table 3: Longitudinal monitoring in breast cancer. Twelve years post-masectomy, a patient presented with elevated tumor biomarkers, but no detectable tumor on body scan. Baseline LiquidBiopsy identified two mutations in circulating CTC. Following treatment with fulvestrant the number of CTC declined, while the frequency of the PIK3ca mutation increased. Sample time CTC count CTC Purity (%) Mutation Baseline 4094 88% PIK3ca, E542K 12.4% PTPN11, L74F 1.1% 2 weeks post-treatment 1492 49% PIK3ca, E542K 31.0% Mutant Frequency (%) 9

Discussion LiquidBiopsy sequencing of CTC mutations directly from whole blood samples has significant potential to advance oncology on multiple fronts, from patient care, to clinical research, to drug development and clinical trial management. CLINICAL ONCOLOGY AND PATIENT CARE For patient care, the ability to readily obtain tumor sequence from whole blood opens new opportunities to support clinical decision-making with LiquidBiopsy testing. Knowing the mutations present in a patient s tumor can inform initial treatment selection, aid in monitoring the response to treatment, and, should resistance emerge, support selection of a next line of therapy. In cases where no tumor sample is available through biopsy or the tissue quantity is insufficient, LiquidBiopsy offers the potential to recover tumor cells for analysis. This noninvasive method for genomic testing also makes regular high-precision monitoring for recurrence possible, e.g. in breast cancer, where the ability to detect a newly emerging tumor clone early in the metastatic process could allow pre-emptive clinical intervention that otherwise would not be supported. For individuals with a history of breast cancer, LiquidBiopsy is now being offered as a recurrence monitoring service called ClearID Breast Cancer. ClearID Breast Cancer is a two-component testing service that identifies elevated CTC levels in breast cancer survivors (ClearID SCORE). Upon identification of elevated CTC levels, patient samples are sequenced for SNVs associated with breast cancer (ClearID SEQ). ClearID Breast Cancer offers clinicians monitoring breast cancer survivors a simple, non-invasive blood test where disease-linked mutations can trigger more vigilant monitoring than existing standard of care. RESEARCH AND CLINICAL TRIAL MANAGEMENT LiquidBiopsy services also have many potential applications in oncology research, drug development, and clinical trial management. Blood samples from subjects can be drawn and the CTC immediately analyzed or banked for later analysis. The simplicity of sample collection and the ability to bank samples makes the technology readily adaptable to multi-country clinical trials. Because of the sensitivity of the genetic sequencing test, resistant or responsive clones within a tumor are accessible even if they occur at a frequency of 1% in the blood. Therefore, LiquidBiopsy sequencing results could offer early surrogate readout of treatment response, help identify clinically relevant patient sub-populations, and enable adaptive clinical trials, and ultimately, the development of highly tumor-specific therapies. 10

Summary Cynvenio received CLIA certification for its LiquidBiopsy services after successfully validating the test s ability to: 1. Recover CTC with high efficiency and purity from a simple blood draw 2. Accurately identify over 4500 mutations (SNVs) in 50 oncogenes with sensitivity down to 1% allelic frequency LiquidBiopsy is the first technology to demonstrate accurate and sensitive detection of rare mutations in CTC isolated from whole blood without whole genome amplification. This technology will be a valuable tool in the clinic, laboratory, and industry in on-going efforts to advance precision medicine. References 1. Pinard R, et al. Assessment of whole genome amplification-induced bias through high-throughput, massively parallel whole genome sequencing. BMC Genomics 2006; 7:216. 2. Allard WJ, et al. Tumor cells circulate in the peripheral blood of all major carcinomas but not in healthy subjects or patients with nonmalignant diseases. Clin Cancer Res 2004; 10:6897-6904. 3. Cristofanilli M, et al. Circulating tumor cells, disease progression, and survival in metastatic breast cancer. N Engl J Med 2004; 351:781-791. 4. Danila, DC, et al. Circulating tumor cell number and prognosis in progressive castration-resistant prostate cancer. Clin Cancer Res 2007: 13:7053-7058. 5. Cohen, SJ, et al. Relationship of circulating tumor cells to tumor response, progression-free survival, and overall survival in patients with metastatic colorectal cancer. J Clin Oncology 2008; 26:3213-3221. 6. Powell AA, et al. Single cell profiling of circulating tumor cells: transcriptional heterogeneity and diversity from breast cancer cell lines. PLoS One 2012; 7(5):e33788. 7. Strauss WM, et al. Rare cell analysis without whole genome amplification by massively parallel sequencing. Cynvenio Biosystems, Inc. white paper. 8. Kotoula V, et al. PloS One 2009; 4(11):e7746 9. Data on file,, Westlake Village, CA, USA 11 2013 Cynvenio Cynvenio Biosystems, Biosystems, Inc. Inc. www.cynvenio.com (805)-777-0017