<p>Supplementary Table S5. RPPA analysis of phosphorylated and total protein levels in PC9 and NCI-H1975 AZD9291 resistant populations compared to respective parental cells.</p>

Supplementary Figure Legend. Legends for Supplementary Figures S1-S6.

Supplementary Tables S1-S4. Generation of resistant cell populations (S1); Small molecule inhibitors (S2); IC50 (µM) values from cell growth inhibition assays comparing compound sensitivity between parental and resistant cell populations (S3); Genetic analysis of resistant cell populations (S4).

Supplementary Methods and References. Description of additional methods and procedures used in the study. Also includes Supplementary References.

<p>Supplementary Figures S1-S6. Comparison of genetic alterations across multiple populations resistant to AZD9291 and other EGFR TKIs (S1); Treatment of resistant populations with AZD9291 (S2); Detection and Validation of a novel NRAS E63K mutation (S3); Lysates were prepared from parental PC9 and resistant populations analysed for levels of total and phosphorylated ERK, NRAS and KRAS by western blot (S4); The novel NRAS E63K mutation is an activating mutation that when expressed enhances resistance to cell growth inhibition by gefitinib or AZD9291 in EGFRm cell lines (S5); In vitro combination of AZD9291 with selumetinib induces more profound phenotype inhibition (S6).</p>

Supplementary Figure from Identification of a Molecularly-Defined Subset of Breast and Ovarian Cancer Models that Respond to WEE1 or ATR Inhibition, Overcoming PARP Inhibitor Resistance

Abstract Noninvasive circulating tumor DNA (ctDNA) assays have the potential to revolutionize clinical trial design; however, challenges still remain in using ctDNA for the molecular profiling of cancer with next-generation sequencing (NGS). Our previous work has defined some of the technical and biologic factors that contribute to these challenges, in particular for variants below 1% allele frequency (AF). To continue to examine these factors, we have designed a study to investigate tumor-plasma discordant variants reported from commercial vendors with an orthogonal ultrasensitive NGS assay focusing primarily on false positive variant calls. The study used 24 commercially purchased matched FFPE tumor-plasma samples from breast, prostate, lung, and ovarian cancer. The FFPE tumor-normal pairs were sent to Foundation Medicine (FMI) for sequencing, while replicate sets of plasma were shipped to three commercial vendors for ctDNA sequencing using the currently available panel at the time of shipment. An additional set of samples, which included tumor, normal, and plasma samples, was sent to a fourth vendor for ultrasensitive NGS assay design and sequencing. The returned variants from three commercial vendors were compared to the tissue variants to define truth. True positive variants were defined as an alteration in the plasma that matches either the tissue or another vendors plasma result, while a false positive is a reported variant identified by only one vendor. The comparative analysis identified that Vendor A had the highest sensitivity at 87%, while Vendor B had the highest PPV at 91%. It should be noted that Vendor B had the smallest targeted panel. Vendors A and C contributed the majority of reported false positive variants with Vendor B only having two false positive variants. A total of 54 false positive variant calls were identified, collated along with several driver mutations and germline SNPs, and sent to Vendor D in a blinded fashion for assay design and sequencing. The orthogonal NGS data returned from Vendor D found that two false positive variants had read support to be tissue false negatives. The tissue AF of both variants was found to be just below the standard cutoff for FMI of 5% AF. Both variants were from Vendor A, which improved the sensitivity of Vendor A to 88%. The additional false positive calls from Vendor A were found in the plasma only, with no supporting evidence in the tissue. Vendor B’s false positive calls had no support from the orthogonal assay in either plasma or tissue. Vendor C had variants that had no support in either the tissue or plasma and variants that have support in the plasma. Overall, the false positive variants tested with the ultrasensitive NGS assay could be grouped into three general categories: variant was not found in either the tissue or plasma, variant was found in only the plasma, and variant had support in both the tissue and plasma. The false positive variants found in plasma or in both tissue and plasma may represent true positive variants and need to be investigated further. Citation Format: Daniel Stetson, Ambar Ahmed, Barrett Nuttall, Tristan Lubinski, J. Carl Barrett, Brian Dougherty. Examination of ctDNA false positive variants reported from commercial vendors by ultrasensitive orthogonal testing [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr A63.

PURPOSE:Discordance between plasma and tumor variant calling has been attributed primarily to tumor heterogeneity, whereas technical variables remain largely unexplored.MATERIALS AND METHODS:To measure these variables, we tested four next-generation sequencing (NGS) gene panel assays for mutations in circulating tumor DNA (ctDNA) using replicate sets of 24 plasma samples and compared the results with matched tumor-normal tissue pairs.RESULTS:Our orthogonal approach identified false-negative (FN) and false-positive (FP) variants with high confidence and revealed substantial variability among the ctDNA assays, with a range of sensitivity (38% to 89%) and positive predictive value (36% to 80%). Most discordance in our cross-vendor study was observed below 1% variant allele frequency. FP variants displayed mutational biases and tended to be novel variants not found in somatic databases. Of the 56 unique variants called by all four ctDNA assays, 41 (68%) resulted from technical discordance.CONCLUSION:These findings suggest that most NGS assay discordance is a result of technical variations and, to a lesser extent, biologic factors such as clonal hematopoiesis of indeterminate potential and tumor heterogeneity.

The PI3Kα signaling pathway is frequently hyper-activated in breast cancer (BrCa), as a result of mutations/amplifications in oncogenes (e.g. HER2), decreased function in tumor suppressors (e.g. PTEN) or activating mutations in key components of the pathway. In particular, activating mutations of PIK3CA (~45%) are frequently found in luminal A BrCa samples. Genomic studies have uncovered inactivating mutations in MAP3K1 (13-20%) and MAP2K4 (~8%), two upstream kinases of the JNK apoptotic pathway in luminal A BrCa samples. Further, simultaneous mutation of PIK3CA and MAP3K1 are found in ~11% of mutant PIK3CA tumors. How these two alterations may cooperate to elicit tumorigenesis and impact the sensitivity to PI3K and AKT inhibitors is currently unknown. Using CRISPR gene editing we have genetically disrupted MAP3K1 expression in mutant PIK3CA cell lines to specifically create in vitro models reflecting the mutational status of PIK3CA and MAP3K1 in BrCa patients. MAP3K1 deficient cell lines exhibited ~2.4-fold increased proliferation rate and decreased sensitivity to PI3Kα/δ(AZD8835) and AKT (AZD5363) inhibitors (~2.61 and ~5.23-fold IC50 increases, respectively) compared with parental control cell lines. In addition, mechanistic analysis revealed that MAP3K1 disruption enhances AKT phosphorylation and downstream signaling and reduces sensitivity to AZD5363-mediated pathway inhibition. This appears to be a consequence of deficient MAP3K1-JNK signaling increasing IRS1 stability and therefore promoting IRS1 binding to p85, resulting in enhanced PI3Kα activity. Using 3D-MCF10A-PI3KαH1047R models, we found that MAP3K1 depletion increased overall acinar volume and counteracted AZD5363-mediated reduction of acinar growth due to enhanced proliferation and reduced apoptosis. Furthermore, in vivo efficacy studies revealed that MAP3K1-deficient MCF7 tumors were less sensitive to AKT inhibitor treatment, compared with parental MCF7 tumors. Our study provides mechanistic and in vivo evidence indicating a role for MAP3K1 as a tumor suppressor gene at least in the context of PIK3CA-mutant backgrounds. Further, our work predicts that MAP3K1 mutational status may be considered as a predictive biomarker for efficacy in PI3K pathway inhibitor trials.

Abstract The increased reliance on liquid biopsy next generation sequencing assays for oncology clinical decisions has highlighted the importance of understanding the factors that impact the analytical validation and clinical utility of these tests. Although recent studies have been published examining the concordance between tumor and plasma sequencing data from the same patient, there has been a lack of insight into the intra-assay variability within these tests. We have previously performed a replicate study across multiple ctDNA sequencing vendors and found a high degree of variability among assays particularly at low allele fractions (AF<1%). To further extend our analysis, we examined the variability of ctDNA sequencing results from replicate patient plasma samples with the same vendor assay. The cohort consisted of 25 total samples: seven cancer patient plasma samples, representing triplicate 2 ml baseline plasma samples, plus four healthy normal samples, representing two different pools of six subjects in duplicate. The plasma names were blinded and shipped to Guardant Health for Guardant360 panel testing and both the raw data and reported variant calls were returned to AstraZeneca for comparative analysis. A total of 17 unique variants were called in this study, of which four variants were concordant for all three replicate plasmas and 13 variants were reported in only one or two replicates. Examining the raw data for all samples, the majority of the discordant variants were able to be identified at low levels. Plasmas had low amounts of ctDNA, and two patients had low unique molecule counts, with allele fractions ranging from 0.10-1.13%. Interestingly, the reported yield of the extracted DNA did not correlate with the ability to detect low frequency somatic variants. Data will be presented showing the unblinded comparisons between the replicate samples for both the cancer patient plasma data and the pooled normal plasmas. This study highlights the importance of reproducibility studies for ctDNA samples with variants at low allele fractions to discern between false negative and false positive finding. Citation Format: Daniel Stetson, Brian Dougherty, Ambar Ahmed, Barrett Nuttall, Tristan J. Lubinski, Fakhera Afshar, Amelia Raymond, Carl Barrett. Examination of intra-assay variability in a commercial ctDNA NGS assay with triplicate clinical plasma and pooled normal samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-223.

Abstract The increased usage of circulating tumor DNA (ctDNA) sequencing for oncology clinical research demonstrates a critical need for sensitive and specific testing. While we have observed a high degree of concordance between single tumor mutations in tumor and plasma, several recent studies have highlighted a lack of concordance between plasma and tumor panel NGS gene panel testing due to biological and technical factors. To explore further these factors and benchmark ctDNA NGS testing services, a set of matched plasma, tumor, and normal samples from 24 subjects were acquired from three biobanking companies. Replicate 2 ml-plasma samples were tested by four ctDNA sequencing companies, and matching tumor/normal samples were tested by two tumor sequencing companies. Concordance was measured by comparing plasma mutations to tumor mutations as well as comparing mutations among the same plasma tested by the ctDNA companies. While our experience with NGS of matched samples from clinical trials typically identifies ~30% of patients with no detectable mutation and therefore likely not shedding tumor DNA, with the retrospectively collected commercial samples ~60% lacked detectable high confidence mutations, likely due to quality control issues with sample collection. We also found variation in the concordance of ctDNA mutation detection rates among the four vendors, due to significant differences in DNA yield and assay sensitivity. While factors such as tumor heterogeneity and timing of plasma-tumor collection can lower concordance rates, the majority of discordance in our study was due to technical rather than biological variation. Assay analytical variance and the impact of reporting false positive variants are key factors that need to be addressed as plasma-based NGS testing is more widely incorporated into translational and clinical research. Examples illustrating the complexity of the analyses and giving support for confidence in ctDNA testing results will be given. Citation Format: Daniel Stetson, Brian Dougherty, Ambar Ahmed, Tristan Lubinski, Aleksandra Markovets, Kenneth Thress, Robert McEwen, Gaia Schiavon, David Whitston, Barrett Nuttall, J. Carl Barrett. Examination of analytical factors impacting concordance of plasma-tumor testing by next-generation sequencing (NGS) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr LB-249. doi:10.1158/1538-7445.AM2017-LB-249

Current understanding of the mutation spectrum of relapsed/refractory (RR) tumors is limited. We performed whole exome sequencing (WES) on 47 diffuse large B cell lymphoma (DLBCL) tumors that persisted after R-CHOP treatment, 8 matched to primary biopsies. We compared genomic alterations from the RR cohort against two treatment-naïve DLBCL cohorts (n=112). While the overall number and types of mutations did not differ significantly, we identified frequency changes in DLBCL driver genes. The overall frequency of MYD88 mutant samples increased (12% to 19%), but we noted a decrease in p.L265P (8% to 4%) and increase in p.S219C mutations (2% to 6%). CARD11 p.D230N, PIM1 p.K115N and CD79B p.Y196C mutations were not observed in the RR cohort, although these mutations were prominent in the primary DLBCL samples. We observed an increase in BCL2 mutations (21% to 38% of samples), BCL2 amplifications (3% to 6% of samples) and CREBBP mutations (31% to 42% of samples) in the RR cohort, supported by acquisition of mutations in these genes in relapsed compared to diagnostic biopsies from the same patient. These increases may reflect the genetic characteristics of R-CHOP RR tumors expected to be enriched for during clinical trial enrollment. These findings hold significance for a number of emerging targeted therapies aligned to genetic targets and biomarkers in DLBCL, reinforcing the importance of time-of-treatment biomarker screening during DLBCL therapy selection.

Efforts to develop effective cancer therapeutics have been hindered by a lack of clinically predictive preclinical models which recapitulate this complex disease. Patient derived xenograft (PDX) models have emerged as valuable tools for translational research but have several practical limitations including lack of sustained growth in vitro. In this study, we utilized Conditional Reprogramming (CR) cell technology- a novel cell culture system facilitating the generation of stable cultures from patient biopsies-to establish PDX-derived cell lines which maintain the characteristics of the parental PDX tumor. Human lung and ovarian PDX tumors were successfully propagated using CR technology to create stable explant cell lines (CR-PDX). These CR-PDX cell lines maintained parental driver mutations and allele frequency without clonal drift. Purified CR-PDX cell lines were amenable to high throughput chemosensitivity screening and in vitro genetic knockdown studies. Additionally, re-implanted CR-PDX cells proliferated to form tumors that retained the growth kinetics, histology, and drug responses of the parental PDX tumor. CR technology can be used to generate and expand stable cell lines from PDX tumors without compromising fundamental biological properties of the model. It offers the ability to expand PDX cells in vitro for subsequent 2D screening assays as well as for use in vivo to reduce variability, animal usage and study costs. The methods and data detailed here provide a platform to generate physiologically relevant and predictive preclinical models to enhance drug discovery efforts.

4041 Background: A Phase II trial in 124 patients (pts) with AGC (NCT01063517; Study 39) showed that, as 2nd-line therapy, the PARP inhibitor olaparib (Lynparza; tablets) plus paclitaxel, followed by maintenance olaparib (O/P arm), significantly improved overall survival (OS) vs paclitaxel alone (P arm) both in the overall population and in pts with low ATM protein levels (ATMlow), for whom the trial was enriched (Bang et al JCO 2015). The current study explored the relationship of other candidate biomarkers with ATM levels and clinical outcomes. Methods: Next-generation DNA sequencing (Foundation Medicine) of tumor samples from Study 39 pts was performed, followed by comprehensive genomic analysis. Candidate biomarkers included microsatellite instability (MSI)-like status (overall mutation load), ATM mutations, mutations in other genes involved in homologous recombination repair (HRRm), and TP53 and ARID1A mutations. ATM levels were determined by IHC at study entry. Associations between genetic markers were assessed by contingency table analysis, while associations between individual genetic markers and clinical outcome were assessed by Cox proportional hazards modeling. Results: 55/124 pts (ATMlow , n = 28; ATM positive [+ve], n = 27) had ≥ 1 sample evaluable for tumor genetics (EFTG): 6 pts (11%) had loss-of-function ATM mutations (all in ATMlow pts) and 35 pts (64%) had a TP53 mutation (TP53m). HRRm (including in ATM), ARID1A mutations and MSI-like profiles were found in 11, 19 and 7 pts, respectively, most of whom were ATMlow . OS HRs (95% CI) for O/P vs P in the EFTG population, ATMlow pts, ATM +ve pts and TP53m pts were 0.34 (0.16–0.74), 0.17 (0.05–0.65), 0.65 (0.24–1.73) and 0.515 (0.21–1.29), respectively. Pt numbers were small, but these exploratory analyses suggest that 1) ATMlow pts who receive O/P have the longest OS, regardless of TP53m status; 2) the OS HR in ATMlow pts is lower than for subsets of pts with other candidate biomarkers; and 3) an OS benefit is seen with O/P for pts without ATM mutations. Conclusions: For pts with AGC, ATM level remains the leading predictive biomarker for determining response to olaparib treatment. Clinical trial information: NCT01063517.

Abstract Patient-derived xenografts (PDX) are widely recognized as a more physiologically relevant preclinical model to standard cell line xenografts. PDX models faithfully recapitulate the original patient genetic profile, gene expression patterns and tissue histology. Despite their benefits, PDX models are limited by their inherent variability, lower throughput and lack of growth in vitro. The ability to generate cell lines from PDX models would enable high throughput chemosensitivity screens, ex vivo genetic manipulation and the development of novel orthotopic models. Development of stable PDX cell lines remains a challenge due to murine stromal outgrowth, lineage commitment and limited differentiation potential. Conditional reprogramming (CR) cell technology is a novel cell culture system facilitating the generation of stable cultures without genetic manipulation (Liu, Am J Pathol, 2012). The success of CR cell technology is dependent upon the combination of feeder cell-derived factors and Rho Kinase (ROCK) inhibitor. CR cells, therefore, represent a new class of progenitor-like cells, distinct from the phenotype of embryonic stem (ES) cells and induced pluripotent stem (iPS) cells. The purpose of this study was to identify the advantages, limitations and potential applications of CR technology for derivation of PDX explant cell lines. Early passage human lung and ovarian PDX tumors were cultured in CR conditions to create stable explant cell lines. Cell lines were established from 5/8 (63%) PDX tumors and were expanded over 6 months in culture with varying morphologies and growth kinetics. Due to normal outgrowth of murine stromal cells, early CR cultures contained mixed populations and required murine depletion to enrich for human cells. Key oncogenic mutations in a model of ovarian papillary serous adenocarcinoma were preserved in the enriched tumor cell population. While purified CR PDX cell lines were amenable to high throughput chemosensitivity screens, in vitro chemosensitivity did not consistently predict response in in vivo murine models. The CR PDX cell lines were additionally assessed for genetic manipulation and ability to form tumors in vivo. Collectively, these results demonstrate the applications of CR technology for the generation of stable explant cell lines from PDX models for preclinical studies. Citation Format: Alexandra Borodovsky, Travis J. McQuiston, Brian Dougherty, Ambar Ahmed, David Whitston, Daniel Stetson, Gretchen K. Hubbard, Sharon S. Challberg, Brian A. Pollok, Celina M. D’Cruz. Use of conditional reprogramming to develop and characterize cell cultures from patient-derived xenograft (PDX) models of human lung and ovarian cancer. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 641.

Abstract Our current understanding of the mutation spectrum of relapse/refractory patients is limited. Several published studies describing the mutational landscape of Diffuse Large B cell Lymphoma (DLBCL) have focused, by design, on diagnostic (pre-treatment) biopsies alone, while re-biopsy of patients who are refractory to first line therapy or who relapse on treatment is not standard of care in DLBCL. We have performed whole exome sequencing on 47 post treatment DLBCL core needle biopsies, 8 with matched diagnostic biopsies. Samples were obtained during a phase II trial prior to the start of treatment with the BCR targeted agent fostamatinib (1,2). Patients had progressed following therapy with an anthracycline-based regimen such as R-CHOP and had a median of 3 prior therapies (range 1-8). We compared the whole exome somatic variant and copy number calls from the post treatment and matched pairs to 2 cohorts of primary DLBCL (n = 112) analyzed with the same mutation caller (3,4). The average mutation rate between the pre and post treatment samples (n = 250, 282 respectively) and the paired diagnostic and post treatment biopsies were similar. However, we found that the mutation spectrum between the paired biopsies differed. Known DLBCL hotspot mutations such as MYD88 L265P remained consistent between pre and post treatment biopsies, however novel mutations in known DLBCL targets were found to emerge in the post treatment biopsies. 1. Veldman-Jones, M. et al. Reproducible, quantitative and flexible molecular sub-typing of clinical DLBCL samples using the NanoString nCounter system. Clin Cancer Res (2014) 2. Flinn I, B.N., Blum KA, et al. A Phase II Trial to Evaluate the Efficacy of Fostamatinib in Patients with Relapsed or Refractory Diffuse Large B-Cell Lymphoma (DLBCL). American Society of Hematology (San Francisco, 2014). 3. Zhang, J. et al. Genetic heterogeneity of diffuse large B-cell lymphoma. Proc Natl Acad Sci U S A 110, 1398-403 (2013). 4. Pasqualucci, L. et al. Analysis of the coding genome of diffuse large B-cell lymphoma. Nat Genet 43, 830-7 (2011). Citation Format: Danielle Greenawalt, Kate Byth, Zhongwu Lai, Justin Johnson, Ambar Ahmed, Brian Dougherty, Kenneth Thress, Michael Zinda, Winnie S. Liang, John Carpten, Stephen Fawell, J. Carl Barrett. Whole exome sequencing of pre and post treatment diffuse large B cell lymphoma reveals the mutation spectrum of the relapse/refractory patient population. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1706. doi:10.1158/1538-7445.AM2015-1706

AbstractResistance to targeted EGFR inhibitors is likely to develop in EGFR-mutant lung cancers. Early identification of innate or acquired resistance mechanisms to these agents is essential to direct development of future therapies. We describe the detection of heterogeneous mechanisms of resistance within populations of EGFR-mutant cells (PC9 and/or NCI-H1975) with acquired resistance to current and newly developed EGFR tyrosine kinase inhibitors, including AZD9291. We report the detection of NRAS mutations, including a novel E63K mutation, and a gain of copy number of WT NRAS or WT KRAS in cell populations resistant to gefitinib, afatinib, WZ4002, or AZD9291. Compared with parental cells, a number of resistant cell populations were more sensitive to inhibition by the MEK inhibitor selumetinib (AZD6244; ARRY-142886) when treated in combination with the originating EGFR inhibitor. In vitro, a combination of AZD9291 with selumetinib prevented emergence of resistance in PC9 cells and delayed resistance in NCI-H1975 cells. In vivo, concomitant dosing of AZD9291 with selumetinib caused regression of AZD9291-resistant tumors in an EGFRm/T790M transgenic model. Our data support the use of a combination of AZD9291 with a MEK inhibitor to delay or prevent resistance to AZD9291 in EGFRm and/or EGFRm/T790M tumors. Furthermore, these findings suggest that NRAS modifications in tumor samples from patients who have progressed on current or EGFR inhibitors in development may support subsequent treatment with a combination of EGFR and MEK inhibition. Cancer Res; 75(12); 2489–500. ©2015 AACR.