Utility of Serial cfDNA NGS for Prospective Genomic Analysis of Patients on a Phase I Basket Study

Abstract

PURPOSE

Cell-free DNA (cfDNA) analysis offers a noninvasive means to access the tumor genome. Despite limited sensitivity of broad-panel sequencing for detecting low-frequency mutations in cfDNA, it may enable more comprehensive genomic characterization in patients with sufficiently high disease burden. We investigated the utility of large-panel cfDNA sequencing in patients enrolled to a Phase I AKT1-mutant solid tumor basket study.

METHODS

Patients had AKT1 E17K-mutant solid tumors and were treated on the multicenter basket study (ClinicalTrials.gov identifier: NCT01226316) of capivasertib, an AKT inhibitor. Serial plasma samples were prospectively collected and sequenced using exon-capture next-generation sequencing (NGS) analysis of 410 genes (Memorial Sloan Kettering [MSK]-Integrated Molecular Profiling of Actionable Cancer Target [IMPACT]) and allele-specific droplet digital polymerase chain reaction (ddPCR) for AKT1 E17K. Tumor DNA (tDNA) NGS (MSK-IMPACT) was also performed on available pretreatment tissue biopsy specimens.

RESULTS

Among 25 patients, pretreatment plasma samples were sequenced to an average coverage of 504×. Somatic mutations were called in 20/25 (80%), with mutant allele fractions highly concordant with ddPCR of AKT1 E17K (r² = 0.976). Among 17 of 20 cfDNA-positive patients with available tDNA for comparison, mutational concordance was acceptable, with 82% of recurrent mutations shared between tissue and plasma. cfDNA NGS captured additional tumor heterogeneity, identifying mutations not observed in tDNA in 38% of patients, and revealed oncogenic mutations in patients without available baseline tDNA. Longitudinal cfDNA NGS (n = 98 samples) revealed distinct patterns of clonal dynamics in response to therapy.

CONCLUSION

Large gene panel cfDNA NGS is feasible for patients with high disease burden and is concordant with single-analyte approaches, providing a robust alternative to ddPCR with greater breadth. cfDNA NGS can identify heterogeneity and potentially biologically informative and clinically relevant alterations.

INTRODUCTION

Large-scale tumor sequencing has increased our understanding of the drivers of cancer, enabling the administration of targeted therapies that would otherwise not be effective in an unselected population.^1,2 The detection of these key oncogenic events, however, has been greatly dependent on access to tumor tissue and limited by the challenges inherent to single-site tissue biopsies, including their invasiveness, limited availability, associated risk and expense, and finally their inability to capture tumor temporal and spatial heterogeneity in its entirety.^3-7 This also poses an important barrier to research and discovery in the context of clinical trials. The hope among the translational oncology community was that many of these challenges could be overcome through the advent of tumor-derived plasma cell-free DNA (cfDNA) profiling.³ It should be noted, however, that the large gene panel next-generation sequencing (NGS) approach that has been widely successful for tumor tissue sequencing has not been widely adopted for cfDNA analysis, where mutation allele frequencies often fall below the limit of detection.^8,9 Successful methods to detect mutations in the context of low circulating tumor DNA (ctDNA) levels typically incorporate molecular barcodes and require more focused panels as a tradeoff for achieving the requisite ultrahigh depth of coverage.

CONTEXT

• Key Objective

• Although large gene panel, next-generation sequencing (NGS) is routinely performed for tumor tissue sequencing, this method has not been extensively used for cell-free DNA (cfDNA) analysis, largely because of the limitations imposed by low mutation allele frequencies in plasma. This study examined whether large-panel cfDNA NGS has clinical utility in a phase I clinical trial population with advanced disease.

• Knowledge Generated

• In patients with advanced cancer enrolled to a phase I basket study, large-panel NGS of prospectively collected plasma samples was largely concordant with both tumor tissue NGS and with single-analyte testing. cfDNA NGS also revealed clinically relevant genomic heterogeneity, unappreciated by tumor NGS.

• Relevance

• Advanced cancer genome profiling can be accomplished by large-panel cfDNA NGS and offers potential advantages over both digital droplet polymerase chain reaction and tumor tissue sequencing approaches.

Despite these limitations, broader profiling of cfDNA with NGS does still present the opportunity to more comprehensively elucidate clinically relevant tumor genomic alterations, mutational burden, and specific mutational signatures such as microsatellite instability that could not be accurately characterized with small gene panels or digital droplet polymerase chain reaction (ddPCR) methods. Indeed, NGS panels (evaluating approximately 50 genes) have been demonstrated as feasible for prescreening Phase I trial candidates.⁸ While large cfDNA panels may not be sufficiently sensitive for all patients, they may be an informative option for early-phase clinical trial candidates who typically present with advanced and heavily pretreated disease, and therefore potentially higher ctDNA levels. To assess the feasibility of this approach, we sought to develop and evaluate an NGS panel in a heterogenous Phase I population with a shared genomic alteration to facilitate comparisons.

The AKT1 E17K mutation is a recurrent (approximately 2%), oncogenic event in human cancer and was credentialed as a therapeutic target for the AKT inhibitor, capivasertib, in a Phase I basket study for AKT1-mutant solid tumors.^10-14 We conducted a single-institution, prospective companion translational study in the genomically selected (AKT1 E17K-mutant) patient population enrolled to this global Phase I multipart basket study (ClinicalTrials.gov identifier: NCT01226316) characterizing the tumor genome and monitoring changes over time on treatment.

METHODS

All patients provided written informed consent, permitting collection and genomic analysis of blood and tumor specimens under an institutional review board–approved research protocol (ClinicalTrials.gov identifier: NCT01775072). Patients with AKT1 E17K-mutation–positive solid tumors treated at Memorial Sloan Kettering Cancer Center (MSKCC) on the Phase I, open-label, multicenter basket study of capivasertib (ClinicalTrials.gov identifier: NCT01226316)¹³ were enrolled to this translational study for plasma collection from November 2014 to April 2016. For enrollment in the therapeutic basket study, local genomic tumor testing was sufficient, and key inclusion criteria were measurable disease by RECIST v1.1 and adequate organ function; key exclusion criteria were known concurrent RAS/RAF mutations. Blood samples were collected for processing, plasma cfDNA extraction, ddPCR, and NGS analysis. Available pretreatment tumor tissue (archival or fresh) along with matched normal specimens underwent clinical NGS in the Clinical Laboratory Improvement Amendments (CLIA)-compliant Molecular Diagnostics Service at MSKCC.

Tumor DNA (tDNA) and cfDNA sequencing were performed using the MSK-Integrated Molecular Profiling of Actionable Cancer Targets (MSK-IMPACTs) assay (341- or 410-gene version) and analyzed, as previously described.^11,15,16 Our discovery pipeline required mutant allele fraction (MAF) of 0.02 for known hotspot mutations and 0.05 for all others. To increase sensitivity, mutations that were previously detected in tissue but not called independently in plasma cfDNA were genotyped directly and considered to be present if a minimum of three reads supported the variant corresponding to an MAF of at least 0.01. See the Data Supplement for further details.

Mutant Allele Quantification by ddPCR

Allele-specific assays were developed and optimized for AKT1 E17K for quantification on the Bio-Rad QX200 ddPCR system (Hercules, CA) as previously described.¹³ Variant allele fraction was determined by mutant droplet count divided by total droplet count. The lower limit of detection of one mutant droplet was used.

Statistical Analysis

Concordance of detected mutations between cfDNA and tissue for a set of samples where every patient was profiled by each method refers to the number of unique mutations detected in common for both data sets at the patient level divided by the total number of unique mutations. The linearly dependent correlation between ddPCR and NGS MAF at the AKT1 E17K locus was assessed with Pearson's product-moment coefficient. All statistical analyses were performed with the software program R version 3.3.3.

RESULTS

Between November 2014 and April 2016, 25 patients with AKT1 E17K–mutant advanced solid tumors were enrolled at our institution to the global capivasertib Phase I basket trial and underwent longitudinal plasma collection (Appendix Fig A1, online only). Most patients (56%) had breast cancer, followed by 24% with gynecologic cancers. Patients were heavily pretreated with a median number of 7 [1-14] prior lines of therapy for metastatic disease (Table 1). Pretreatment cfDNA samples (n = 25) underwent NGS and ddPCR analysis. Across all sequenced pretreatment plasma samples (n = 25), median cfDNA yield (from 3 mL extracted plasma) was 24 ng (range, 8-158).

TABLE 1.

Baseline Patient Characteristics, n = 25

AKT1 E17K Detection in Pretreatment cfDNA (n = 25)

Pretreatment plasma cfDNA samples were sequenced using MSK-IMPACT to an average coverage of 540×. We detected AKT1 E17K and additional somatic mutations in 80% of patients (20/25 cfDNA-positive). We also performed ddPCR genotyping of AKT1 E17K in cfDNA, detecting the alteration in 92% (23/25) of patients. For those patients in whom AKT1 E17K was detected by both assays, we observed a very high correlation in MAFs from both approaches (r² = 0.976, Fig 1).

FIG 1.

AKT1-mutant allele fraction (MAF) concordance in cell-free DNA between Droplet Digital polymerase chain reaction (ddPCR) and next-generation sequencing (NGS).

Among the five patients in whom the AKT1 E17K mutation was not detectable by NGS, four patients (1 lung, 1 uterine, 1 cervical, and 1 ER plus breast cancer) had low cfDNA concentrations (median, 4 [3-5] ng/mL v 11 [3-53] ng/mL in the 20 patients with detectable AKT1 by NGS) and had a median of 0 (0-2) tumor mutations detected in their cfDNA. Interestingly, three of these four patients had clinical benefit from capivasertib (objective response or stable disease for > 24 weeks). In contrast, the fifth patient with an undetectable AKT1 mutation by NGS, who had breast cancer with rapid disease progression on capivasertib, had a relatively high cfDNA concentration (44 ng/mL) with nine other mutations detected in cfDNA (Appendix Table A1). Taken together, this suggests that the AKT1 mutation detected on prior tumor sequencing in this patient and subsequently used as the basis for enrollment on the capivasertib study was likely a subclonal event, although more recent tissue for confirmation with tumor NGS was unavailable for this study.

Finally, among the two patients with an undetectable AKT1 mutation by ddPCR (one uterine, and one ovarian granulosa cell tumor), the mutation was detectable by NGS in the patient with granulosa cell tumor at allele frequency 0.014 and was the only mutation observed in the cfDNA of this patient (Appendix Table A1). This was notable, given that the AKT1 E17K mutation was observed to be subclonal in the prior tumor sample sequenced from this patient, whereas two hotspot mutations inferred to be clonal in the tumor (FOXL2 C134W and telomerase reverse transcriptase [TERT] promoter) were both not detected in her cfDNA. Intriguingly, this patient derived significant clinical benefit from capivasertib.

Concordance Between Pretreatment cfDNA and tDNA by NGS

For the 20 cfDNA samples with detectable AKT E17K on NGS, pretreatment tDNA sequencing data were available for comparison in 17 of 20 patients. Considering mutations across all target genes, concordance between tissue and plasma NGS was acceptable. Seventy-three percent (56/77) of all tDNA mutations were also detected in plasma, and 77% (56/73) of all cfDNA mutations were also detected in tumor tissue. Concordance among recurrent or hotspot mutations (defined here as those present in the COSMIC database) was even higher: 94% (33/35) of all tDNA mutations were detected in plasma, and 87% (33/38) of all cfDNA mutations were detected in tumor tissue. Altogether, 60% of all mutations and 82% of recurrent mutations were shared between tissue and plasma (Table 2). This reaffirms that the plasma NGS analysis faithfully represents the genomic profile of tumor tissue. Moreover, in eight of 17 cases with both pretreatment tumor and plasma NGS, 17 new mutations not present in archival tumors were detected in cfDNA, including hotspot mutations in ESR1 (four cases), which have been shown to confer resistance to endocrine therapy¹⁷ (Appendix Table A2). These results evince the potential of cfDNA to capture genetic heterogeneity that may be missed when examining DNA obtained from a single tumor site. We notably did not find a correlation between the extent of cfDNA and tDNA genomic discordance and the length of time from tissue biopsy to plasma collection (Appendix Fig A2).

TABLE 2.

Mutation Concordance Between cfDNA-Positive Samples and Available Pretreatment Tumor (tDNA), n = 17

cfDNA NGS Revealed Clinically Relevant Mutations for Each Patient Who Lacked Baseline tDNA

Of the four patients for whom tissue was not available, pretreatment cfDNA NGS revealed the molecular profile of their cancer, identifying 20 variants in total, 65% (13/20) of which were known to be recurrent based on COSMIC. Multiple co-occurring oncogenic mutations were detected in all four patients (Appendix Table A3). In two of the four patients, we detected coincident mutations in ARAF and KRAS that notably would have precluded these patients’ participation in the capivasertib clinical trial if known prior to enrollment, given the fact that any concurrent RAF/RAS mutations were an exclusion criterion for this study. Moreover, both patients did not respond to the study drug. Interestingly, the KRAS A146T mutation was identified in the plasma from a patient with colorectal cancer previously genotyped as KRAS wild-type from prior mass spectrometry-based testing but who subsequently received epidermal growth factor receptor (EGFR)-targeted therapy with cetuximab prior to enrollment on the capivasertib basket study. This raises the possibility that NGS of pretreatment cfDNA in this patient may have identified a resistance mutation that emerged in response to prior EGFR-directed therapy or captured heterogeneity of this patient’s cancer.¹⁸ Taken together, these results demonstrate the potential added value of plasma analysis in this context.

Serial cfDNA NGS Revealed Clonal Dynamics

We reasoned that longitudinal profiling could reveal clonal evolution and the emergence of mutations associated with AKT inhibitor resistance. Longitudinal (ie, pretreatment and at disease progression) plasma samples were available for cfDNA NGS in most (92%, 23/25) patients. In the 18 patients with sufficiently high cfDNA tumor content for accurate quantitation of MAF, parallel tracking of multiple co-occurring mutations appeared to demonstrate two distinct clonal patterns (Figs 2A and 2B). A concordant or congruent pattern was seen in the majority (83%, 15/18) of patients, observed as the MAF of identified mutations rising or declining together in concert. By contrast, a discordant or incongruent pattern was seen in 17% (3/18) of patients with the MAF of identified mutations rising or declining in opposition or divergently. An example of the latter was observed in the patient with KRAS-mutant colorectal cancer alluded to previously (Fig 2B). In the case of this patient, longitudinal cfDNA samples and tracking the MAF values of each mutation (KRAS A146T, AKT1 E17K, and SMAD4 D537G) revealed two divergent trajectories: the MAFs of AKT1 and SMAD4 decreased significantly and in parallel under the pressure of AKT inhibition, whereas the MAF of KRAS increased nearly four-fold coinciding with disease progression, suggesting the selection of a KRAS-mutant clone as a basis for drug resistance.

FIG 2.

Patterns of clonal dynamics in longitudinal cell-free DNA samples. (A) Concordant: mutant allele fraction (MAF) of each variant rises together in this patient with ovarian cancer. (B) Discordant: as KRAS MAF rises, MAFs of the other variants are declining in this patient with colorectal cancer.

DISCUSSION

In the current paradigm of genome-driven precision oncology, patients found to harbor tumor-specific driver alterations may receive genomically guided targeted therapy either as standard of care (if proven for that indication) or as part of a multiple-histology molecularly driven study, often described as a basket trial.^2,19-21 This latter method of evaluating targeted therapeutics has necessitated incorporation of comprehensive translational studies on collected biospecimens to enable the identification of predictive biomarkers of response and resistance, both to identify patients most likely to benefit from the intervention and to suggest potential combination strategies that could prevent or delay the emergence of drug-resistant clones.²² This single-center translational study of serial cfDNA analysis in patients enrolled to an AKT1-mutant therapeutic basket study (ClinicalTrials.gov identifier: NCT01226316) is one of the first reports of large-panel cfDNA sequencing conducted in a molecularly selected, diverse cancer-type patient population receiving therapy directed at a specific, recurrent oncogenic molecular alteration, in this case the AKT1 E17K mutation.

We report AKT1 E17K mutation detection in 80% of patients by cfDNA NGS, despite a relatively low median cfDNA concentration (8 ng/mL) overall among study patients. To validate our findings, we performed ddPCR genotyping of AKT1 E17K in cfDNA and observed very high correlation in MAF between both approaches, consistent with prior studies indicating that both sequencing technologies can be used for longitudinal monitoring of MAF.²³ We demonstrated a good concordance between cfDNA and tDNA NGS and found that large-panel cfDNA NGS can capture the spectrum of disease-associated mutations broadly, detecting many additional alterations not identified by tumor NGS, results with at least prognostic relevance in four patients found to have ESR1 mutations.²⁴ Furthermore, in all four of the patients in this study without available tumor tissue, cfDNA NGS revealed multiple co-occurring oncogenic mutations, results with potentially predictive relevance in two patients in whom concurrent RAS/RAF mutations were detected,^24,25 and which, if known a priori, would have precluded their participation on the therapeutic trial. As both of these patients had a poor response to the AKT inhibitor capivasertib, our results suggest that routine use of cfDNA to screen patients for targeted therapy clinical trials could improve the ability of oncologists to identify those patients most likely to benefit clinically. Finally, in longitudinal samples, we observed two distinct patterns of tumor clonal dynamics in response to targeted therapy, a phenomenon also noted by other groups.^26,27

Our study had several important limitations including a relatively low number of patients evaluated, the inclusion of a diverse group of cancers, and the absence of corresponding tumor tissue in a subset. Our results also highlight the limitations of cfDNA NGS in certain tumor types; indeed, in 8% of patients in our study, cfDNA NGS did not detect any tumor mutations even in the context of patients with advanced disease, potentially reflecting low tumor shedding in these patients, given that the median cfDNA concentration among them was relatively low at 4.2 ng/mL.

While cfDNA NGS offers a minimally invasive approach to the identification of somatic genomic alterations that are predictive of drug sensitivity and resistance, it will likely not entirely replace the need for tumor biopsies as the gold standard for diagnosis and genotyping but rather provides a complementary and sometimes necessary substitution for tissue analysis where an open biopsy is not feasible. Although large-panel NGS molecular profiling of cfDNA has been hindered somewhat by both a relative paucity of total cfDNA and a low ctDNA fraction (fraction of cfDNA derived from tumor) when compared with analogous profiling from tissue biopsies, the development of methods to increase the sensitivity and specificity of cfDNA sequencing with error suppression that leverages duplex sequencing with unique molecular indexing (UMI) has been underway.²⁸ Since this analysis was conducted, MSK has validated and implemented a more sensitive ultra-deep coverage cfDNA panel, MSK-ACCESS, to enable higher sensitivity to detect ctDNA mutations following UMI-based background error suppression.²⁹ As a tradeoff, however, this assay captures only 13% the territory of MSK-IMPACT and therefore would, by design, fail to detect certain mutations and mutation signatures compared with a larger panel.

This study demonstrates that cfDNA provides an additional and, in some instances, vital source of tumor genomic material to aid characterization of the molecular landscape of a rare genomic subgroup of cancers and reaffirms that plasma can often faithfully represent the genomic profile of tumor tissue. Moreover, cfDNA NGS captured the genomic changes occurring over time in these tumors in response to targeted therapy with the AKT inhibitor capivasertib. In conclusion, in a genomically directed targeted therapy study, NGS of cfDNA with a large gene panel was feasible and highly concordant with ddPCR approaches, captured disease heterogeneity, and permitted serial monitoring of tumor clone dynamics that highlight distinct patterns of genomic or molecular response to targeted therapy. This technology ultimately provides a useful tool to aid early-phase drug development in molecularly selected studies.

Appendix

APPENDIX 1. SUPPLEMENTAL METHODS

In our prior report of an international basket study evaluating AZD5363 in patients with AKT1-mutant solid tumors (ClinicalTrials.gov identifier: NCT01226316), baseline cell-free DNA (cfDNA) samples were evaluable from 43 patients enrolled across multiple centres. All 43 baseline samples were analysed at Memorial Sloan Kettering Cancer Center (MSKCC) using the same previously described digital droplet polymerase chain reaction (ddPCR) methods (Data supplement, Hyman & Smyth et al, JCO 2017).

Of these 43 patients, 25 were enrolled at MSKCC and were additionally enrolled in an institutional translational biospecimen study (ClinicalTrials.gov identifier: NCT01775072), where we collected further baseline and longitudinal samples in these 25 patients, of whom the longitudinal ddPCR data of 23 patients were available for description in the prior report.

The purpose of this study is to describe the findings of the next-generation sequencing (NGS) analyses of those samples, placed in the context of all available genomic data from these 25 patients, specifically from collected tumor samples (both archival and fresh) and prior ddPCR data.

Sample Collection

Blood collection and plasma isolation.

Three hundred thirteen serial venous blood samples (10 mL each) were collected in cfDNA blood collection tube (Streck) tubes from 25 patients during routine phlebotomy pre- and on-treatment (at the time of every treatment study visit, occurring every 21 days).

Whole blood was centrifuged in two steps to isolate cell-free plasma; red blood cells were separated from plasma by centrifugation at 800 × g for 10 minutes (ambient temperature), and separated plasma was then centrifuged in a high-speed microcentrifuge at 18,000 × g for 10 minutes (ambient temperature), as previously described.¹³ Cell-free plasma was then stored at −80°C until ready for cfDNA extraction.

Tumor sample collection.

Formalin-fixed, paraffin-embedded (FFPE) blocks or fresh tumor tissue were collected either prospectively from on-study pretreatment biopsies or from archival specimens.

A matched normal sample of blood (3 mL) in an EDTA tube was also collected from each patient, typically at the screening visit for the therapeutic study.

DNA Extraction and Quantification

cfDNA extraction from plasma.

cfDNA was extracted from 3 mL of plasma with the QIAsymphony diagnostic sample preparation Virus/Pathogen Midi Kit, following the manufacturer’s instructions and using a QIAGEN custom 3-mL protocol (3.5-mL plasma aliquots were used to account for the 0.5-mL dead volume required by the QIAsymphony).

cfDNA quantitation.

DNA isolated from plasma was quantified by the TapeStation system (Agilent 2200) following the manufacturer’s instructions. From November 2015, the Fragment Analyzer system (Advanced Analytical Technologies, Santa Clara, CA) was used to quantify isolated DNA following the manufacturer’s instructions and as previously described.¹³ Correlation between TapeStation and Fragment Analyzer was verified following MSKCC’s stringent clinical method validation process and in compliance with Clinical Laboratory Improvement Amendments (CLIA) regulations.

DNA extraction from FFPE and patient-matched normal samples.

DNA from FFPE sections and patient-matched normal samples were extracted with the Qiagen DNeasy Tissue kit and the EZ1 Advanced XL system (Qiagen, Valencia, CA), respectively, and according to the manufacturer’s instructions and as previously described.^11,13

Targeted Massively Parallel NGS Analysis of cfDNA and Tumor DNA

In addition to pre- and end-of-treatment cfDNA samples, some further on-treatment cfDNA samples also underwent NGS.

Tumor DNA and cfDNA sequencing were performed using the MSK-Integrated Molecular Profiling of Actionable Cancer Targets assay, as previously described.^15,16 In brief, this hybridization assay captures all protein-coding exons and selected intronic and regulatory regions of key cancer-associated genes for targeted deep sequencing (either with a 341-gene or 410-gene version of the assay, n = 10 and 11 tumor samples, respectively). All cfDNA samples were captured using the 410-gene version of the assay. Illumina libraries were constructed with KAPA Hyper Prep Kit followed by ligation with a 5-μM adaptor concentration (catalog #KK8504; Kapa Biosystems, Wilmington, MA). Libraries of this targeted capture were pooled in equimolar concentrations and sequenced on an Illumina HiSeq 2500 system (Illumina, San Diego, CA) as paired end 100–base pair reads. Data were analyzed as previously described.^11,15

Our discovery pipeline required mutant allele fraction of 0.02 for known hotspot mutations and 0.05 for all others. To increase sensitivity, mutations that were previously detected in tissue but not called independently in plasma cfDNA were genotyped directly and considered to be present if a minimum of three reads supported the variant corresponding to a mutant allele fraction of at least 0.01. Germline variants were eliminated through the use of patient-matched blood DNA. Each alteration identified by the pipeline was manually reviewed to ensure that visible artifacts were removed, and complex events identified separately were merged and represented correctly.

Concordance Analysis Between Pretreatment cfDNA and tDNA by NGS

The purpose of the concordance analysis was to compare the co-occurring alterations between plasma and tissue for those patients with detectable ctDNA and explore the basis for any observed heterogeneity. We chose to exclude the five AKT1-negative cfDNA samples from the plasma-tissue concordance analysis as we reasoned that these patients had little-to-no ctDNA shedding and would not adequately capture the landscape of somatic alterations in these patients. These five cases are included in the calculation of the rate of ctDNA detection, where we established that they had low overall cfDNA concentrations.

FIG A1.

Consort diagram.

FIG A2.

Time between tumor tissue and cfDNA acquisition. Time between pretreatment tumor tissue biopsies and cfDNA collection did not correlate with discordance between mutation calls in tumor tissue versus cfDNA. cfDNA, cell-free DNA.

TABLE A1.

Discordant AKT1 E17K detection, n = 6

TABLE A2.

Mutations Identified in Pretreatment cfDNA and Absent in Pretreatment tDNA, n = 17

TABLE A3.

Mutations Identified in Pretreatment cfDNA of Patients Without tDNA, n = 4

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

Lillian M. Smyth

Employment: Loxo Oncology at Lilly

Stock and Other Ownership Interests: Lilly

Honoraria: AstraZeneca, Pfizer, Roche/Genentech

Consulting or Advisory Role: AstraZeneca, Roche/Genentech, Loxo Oncology at Lilly, Pfizer, Novartis

Research Funding: AstraZeneca, Roche/Genentech, Puma Biotechnology

Travel, Accommodations, Expenses: Pfizer, Roche/Genentech, Puma Biotechnology

Juber Ahamad A. Patel

Research Funding: GRAIL

Patents, Royalties, Other Intellectual Property: I am an inventor on a pending patent application “Systems and Methods for Detecting Cancer via cfDNA Screening”

Brian Houck-Loomis

Patents, Royalties, Other Intellectual Property: BioLegend

Gaia Schiavon

Employment: AstraZeneca

Stock and Other Ownership Interests: AstraZeneca

Bob T. Li

Consulting or Advisory Role: Guardant Health, Hengrui Therapeutics

Research Funding: Roche/Genentech, AstraZeneca, GRAIL, Daiichi Sankyo, Hengrui Therapeutics, Guardant Health, Amgen, Lilly, MORE Health

Patents, Royalties, Other Intellectual Property: US62/514,661, US62/685,057, Karger Publishers—Book royalty

Travel, Accommodations, Expenses: MORE Health, Jiangsu Hengrui Medicine

(Optional) Uncompensated Relationships: Amgen, AstraZeneca, Genentech, Lilly, Boehringer Ingelheim

Pedram Razavi

Honoraria: Epic Sciences, Inivata

Consulting or Advisory Role: Novartis, AstraZeneca, Foundation Medicine

Research Funding: GRAIL, Illumina, Novartis

Travel, Accommodations, Expenses: Epic Sciences, Guardant Health

Jorge S. Reis-Filho

Leadership: Grupo Oncoclinicas

Stock and Other Ownership Interests: Repare Therapeutics, PAIGE.AI

Consulting or Advisory Role: Genentech/Roche, Invicro, Ventana Medical Systems, Volition RX, Paige.AI, Goldman Sachs, Novartis, Repare Therapeutics

Barry S. Taylor

Consulting or Advisory Role: Boehringer Ingelheim, Loxo Oncology at Lilly

Research Funding: Genentech

José Baselga

Employment: AstraZeneca

Leadership: Infinity Pharmaceuticals, Varian Medical Systems, Bristol-Myers Squibb, Foghorn

Stock and Other Ownership Interests: Juno Therapeutics, GRAIL, Tango, Venthera, Northern Biologics, ApoGen Biotechnologies, Aura Biosciences

Consulting or Advisory Role: Novartis

Patents, Royalties, Other Intellectual Property: Combination Therapy Using PDK1 and PI3K Inhibitors, Use of Phosphoinositide 3-Kinase Inhibitors for Treatment of Vascular Malformations, Inhibition of KMT2D for the Treatment of Cancer

Travel, Accommodations, Expenses: Roche, Daiichi Sankyo

David B. Solit

Stock and Other Ownership Interests: Loxo

Consulting or Advisory Role: Pfizer, Loxo, Illumina, Vividion Therapeutics, Lilly, QED Therapeutics, BridgeBio Pharma

David M. Hyman

Employment: Lilly, Loxo

Stock and Other Ownership Interests: Fount

Consulting or Advisory Role: Chugai Pharma, Boehringer Ingelheim, AstraZeneca, Pfizer, Bayer, Genentech, Fount, Lilly

Research Funding: AstraZeneca, Puma Biotechnology, Loxo, Bayer

Travel, Accommodations, Expenses: Genentech, Chugai Pharma

Michael F. Berger

Consulting or Advisory Role: Roche

Research Funding: Grail

Patents, Royalties, Other Intellectual Property: Provisional patent pending for “Systems and Methods for Detecting Cancer via cfDNA Screening”

Sarat Chandarlapaty

Consulting or Advisory Role: Sermonix Pharmaceuticals, Novartis, Context Therapeutics, Lilly, Revolutions Medicine, Bristol-Myers Squibb, Paige.ai

Research Funding: Novartis, Daiichi Sankyo, Sanofi, Lilly, Genentech

Patents, Royalties, Other Intellectual Property: Patents for (1) targeting mutant ER with ER PROTACS and (2) detecting genomic and histologic alterations in breast cancer using machine learning algorithms

Travel, Accommodations, Expenses: Bristol-Myers Squibb

No other potential conflicts of interest were reported.

AUTHOR CONTRIBUTIONS

Conception and design: Lillian M. Smyth, José Baselga, David B. Solit, David M. Hyman

Financial support: David B. Solit, David M. Hyman

Administrative support: Duygu S. Selcuklu, David B. Solit, David M. Hyman

Collection and assembly of data: Lillian M. Smyth, Jonathan B. Reichel, Jiabin Tang, Fanli Meng, Duygu S. Selcuklu, Brian Houck-Loomis, Daoqi You, Aliaksandra Samoila, David B. Solit, David M. Hyman, Michael F. Berger, Sarat Chandarlapaty

Data analysis and interpretation: Lillian M. Smyth, Jonathan B. Reichel, Jiabin Tang, Juber Ahamad A. Patel, Daoqi You, Gaia Schiavon, Bob T. Li, Pedram Razavi, Salvatore Piscuoglio, Jorge S. Reis-Filho, Barry S. Taylor, José Baselga, David B. Solit, David M. Hyman, Michael F. Berger, Sarat Chandarlapaty

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors