LETTER
`Non-invasive analysis of acquired resistance to
`cancer therapy by sequencing of plasma DNA
`
`doi:10.1038/nature12065
`
`Muhammed Murtaza1*, Sarah-Jane Dawson1,2*, Dana W. Y. Tsui1*, Davina Gale1, Tim Forshew1, Anna M. Piskorz1,
`Christine Parkinson1,2, Suet-Feung Chin1, Zoya Kingsbury3, Alvin S. C. Wong4, Francesco Marass1, Sean Humphray3,
`James Hadfield1, David Bentley3, Tan Min Chin4,5, James D. Brenton1,2,6, Carlos Caldas1,2,6 & Nitzan Rosenfeld1
`
`demonstrated as a potential tool for detection of disease or analysis
`of tumour burden in patients with advanced cancers5,7. These studies
`established that plasma DNA contains representation of the entire
`tumour genome7, mixing together variants originating from multiple
`independent tumours5. This suggests that deeper sequencing of plasma
`DNA, applied to selected samples with high tumour burden in blood,
`may allow assessment of clonal heterogeneity and selection. In this
`study, we applied exome sequencing of ctDNA as a platform for
`non-invasive analysis of tumour evolution during systemic cancer
`treatment (Fig. 1).
`
`Plasma sample
`before treatment
`
`Plasma sample
`at progression
`
`Treatment
`
`Mutations
`
`Allele fraction
`
`Exome
`sequencing
`
`Analysis of mutations
`in plasma DNA
`
`Identification
`of mutations
`selected by
`treatment
`
`Mutations
`
`Allele fraction
`
`Figure 1 | Identification of treatment-associated mutational changes from
`exome sequencing of serial plasma samples. Overview of the study design:
`plasma was collected before treatment and at multiple time-points during
`treatment and follow-up of advanced cancer patients. Exome sequencing was
`performed on circulating DNA from plasma at selected time-points, separated
`by periods of treatment, and germline DNA. Mutations were identified across
`the plasma samples, and their abundance (allele fraction) at different time-
`points compared, generating lists of mutations that showed a significant
`increase in abundance, which may indicate underlying selection pressures
`associated with specific treatments. These lists contained mutations known to
`promote tumour growth and drug resistance, but also mutations of unknown
`significance. Accumulating such data across large cohorts could identify genes
`or pathways with recurrent mutations.
`
`Cancers acquire resistance to systemic treatment as a result of clonal
`evolution and selection1,2. Repeat biopsies to study genomic evolu-
`tion as a result of therapy are difficult, invasive and may be con-
`founded by intra-tumour heterogeneity3,4. Recent studies have
`shown that genomic alterations in solid cancers can be characterized
`by massively parallel sequencing of circulating cell-free tumour
`DNA released from cancer cells into plasma, representing a non-
`invasive liquid biopsy5–7. Here we report sequencing of cancer
`exomes in serial plasma samples to track genomic evolution of meta-
`static cancers in response to therapy. Six patients with advanced
`breast, ovarian and lung cancers were followed over 1–2 years. For
`each case, exome sequencing was performed on 2–5 plasma samples
`(19 in total) spanning multiple courses of treatment, at selected time
`points when the allele fraction of tumour mutations in plasma was
`high, allowing improved sensitivity. For two cases, synchronous
`biopsies were also analysed, confirming genome-wide represen-
`tation of the tumour genome in plasma. Quantification of allele
`fractions in plasma identified increased representation of mutant
`alleles in association with emergence of therapy resistance. These
`included an activating mutation inPIK3CA(phosphatidylinositol-4,5-
`bisphosphate 3-kinase, catalytic subunit alpha) following treatment
`with paclitaxel8; a truncating mutation in RB1 (retinoblastoma 1)
`following treatment with cisplatin9; a truncating mutation in
`MED1 (mediator complex subunit 1) following treatment with
`tamoxifen and trastuzumab10,11, and following subsequent treat-
`ment with lapatinib12,13, a splicing mutation in GAS6 (growth
`arrest-specific 6) in the same patient; and a resistance-conferring
`mutation in EGFR(epidermal growth factor receptor; T790M) follow-
`ing treatment with gefitinib14. These results establish proof of prin-
`ciple that exome-wide analysis of circulating tumour DNA could
`complement current invasive biopsy approaches to identify muta-
`tions associated with acquired drug resistance in advanced cancers.
`Serial analysis of cancer genomes in plasma constitutes a new para-
`digm for the study of clonal evolution in human cancers.
`Serial sampling of the tumour genome is required to identify the
`mutational mechanisms underlying drug resistance2. Serial tumour
`biopsies are invasive and often unattainable. Tumours are heterogen-
`eous and continuously evolve, and even if several biopsies are obtained,
`these are limited both spatially and temporally. Analysis of isolated
`circulating tumour cells (CTCs) has been proposed, but circulating
`tumour DNA (ctDNA) is more accessible and easier to process15.
`Previous studies of tumour mutations in plasma have analysed indi-
`vidual loci, genes or structural variants to quantify tumour burden and
`to detect previously-characterized resistance-conferring mutations1,6,16–18.
`Genome-wide sequencing of plasma samples is used in prenatal dia-
`gnostics, demonstrating comprehensive coverage of the genome19.
`More recently, genome-wide sequencing of plasma DNA has been
`
`1Cancer Research UK Cambridge Institute and University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE, UK. 2Addenbrooke’s Hospital, Cambridge University Hospital NHS
`Foundation Trust and NIHR Cambridge Biomedical Research Centre, Cambridge CB2 2QQ, UK. 3Illumina, Inc., Chesterford Research Park, Little Chesterford CB10 1XL, UK. 4Department of Haematology-
`Oncology, National University Cancer Institute, National University Health System, 5 Lower Kent Ridge Road, Tower block level 7, 119074 Singapore. 5Cancer Science Institute, National University of
`Singapore, Centre for Translational Medicine, 14 Medical Drive, #12-01, 117599 Singapore. 6Cambridge Experimental Cancer Medicine Centre, Cambridge CB2 0RE, UK.
`*These authors contributed equally to this work.
`
`1 0 8 | N A T U R E | V O L 4 9 7 | 2 M A Y 2 0 1 3
`
`©2013
`
`Macmillan Publishers Limited. All rights reserved
`
`00001
`
`EX1004
`
`

`

`LETTER RESEARCH
`
`c
`Case 3
`(ovarian cancer, study ID: OVO4-114)
`
`ECX
`
`Paclitaxel C-LD
`
`E1
`
`E2
`
`E3
`
`MAGED2
`SERPINB13
`
`ZIC1
`
`a
`Case 1
`(breast cancer, study ID: DETECT-37)
`
`Epirubicin
`
`Paclitaxel
`
`b
`Case 2
`(breast cancer, study ID: DETECT-52)
`Tamoxifen /
`Lapatinib /
`trastuzamab
`capecitabine
`
`E1
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0.0
`
`E2
`
`E3
`
`GAS6
`PDGFRA
`
`MED1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0.0
`
`E1
`
`E2
`
`E3
`
`SMC4
`PIK3CA
`
`FANCD2
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`AF
`
`0.2
`
`AF
`
`0.1
`
`0.0
`
`0
`
`441
`Days of follow up
`
`665
`
`0
`
`147
`Days of follow up
`
`497
`
`0
`
`215
`Days of follow up
`
`370
`
`d
`Case 4
`(ovarian cancer, study ID: OVO4-68)
`
`Paclitaxel
`
`e
`Case 5
`(ovarian cancer, study ID: OVO4-33)
`Carboplatin /
`Carboplatin /
`radiotherapy
`paclitaxel LD C-LD
`
`f
`Case 6
`(lung cancer, study ID: LUN-209)
`
`Gefitinib
`
`EGFR (exon 19 deletion)
`EGFR (T790M)
`
`E2
`
`NFKB1
`TP53
`
`EGFR (T790M)
`
`E1
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0.0
`
`Cisplatin
`
`0.6
`
`E3
`
`E3
`
`E2
`
`0.4
`
`E1
`
`E2
`
`E5
`
`E4
`
`RB1
`ZEB2
`
`MTOR
`
`PARP8
`CES4A
`
`BUB1
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0.0
`
`E1
`
`0.6
`
`0.4
`
`0.2
`
`0.0
`
`0.4
`
`0.3
`
`AF
`
`0.2
`
`AF
`
`0.1
`
`0.0
`
`0
`
`230
`Days of follow up
`
`357
`
`0
`
`189
`406 490 601
`Days of follow up
`
`0
`
`Days of follow up
`
`109
`
`‘Anchor’ mutations
`(TAm-Seq / digital PCR)
`
`‘Anchor’ mutations
`(exome sequencing)
`
`Tumour burden in plasma
`(exome-wide estimate)
`
`Examples of mutations showing
`significant changes in AF, genes
`indicated separately for each case
`
`Figure 2 | Mutations showing evidence of genomic tumour evolution. All
`panels (a–f) are made up of an upper and a lower subpanel. Upper subpanels,
`time courses for allele fractions (AF; data points) of ‘anchor’ mutations used for
`initial quantification of ctDNA levels, and the fractional concentration of
`tumour DNA (tumour burden; grey dashed lines). ‘Anchor’ mutations were
`measured using digital PCR or TAm-Seq6 for all available plasma samples, and
`using exome sequencing at selected time points indicated by E1, E2, E3 (and E4
`and E5 for case 5). Tumour burden was estimated from exome data (an
`adaptation of genome-wide aggregated allelic loss7). In a, AF was averaged over
`six mutations measured in parallel using digital PCR. In b, a single mutation in
`
`ATM (predicted amino acid change I2948F) was measured by TAm-Seq. In
`c, d and e, a single mutation in TP53 was measured by digital PCR for each case
`(R175H, K132N and R175H, respectively). In f, digital PCR was used to
`measure abundance of a deletion in exon 19 of EGFR (not quantified in exome
`sequencing data) and the EGFR T790M mutation. Lower subpanels, AF in
`exome data for selected mutations (blue, green and orange datapoints, see key)
`for each of the cases. Additional details are listed in Table 1, and a full list of
`mutations that showed a significant increase in abundance is included in
`Supplementary Tables 2–7. ECX, epirubicin, cisplatin and capecitabine; C-LD,
`carboplatin and liposomal doxorubicin; LD, liposomal doxorubicin.
`
`We performed whole exome sequencing of plasma DNA in six
`patients with advanced cancers (Supplementary Table 1): two with
`breast cancer (cases 1 and 2), three with ovarian cancer (cases 3–5),
`and one with non-small-cell lung cancer (NSCLC, case 6). Exome
`sequencing was performed on multiple plasma samples from each
`patient separated by consecutive lines of therapy, spanning up to
`665 days of clinical follow up (range 109–665 days, median 433 days).
`The ability to detect genomic events using redundant sequencing is
`dependent on the allele fraction (AF) of the mutant alleles in the
`samples analysed (ratio of mutant reads to depth of coverage at that
`locus), the sequencing depth, and the background noise rates of
`sequencing. Levels of ctDNA were previously quantified in these
`patients using digital PCR and tagged-amplicon deep sequencing6
`(TAm-Seq; Fig. 2, upper subpanels), allowing us to focus on samples
`with a high mutant AF in plasma, in which genomic changes related
`
`to the tumour could be identified even at relatively modest depth of
`sequencing. Comparison of AF measured using exome sequencing,
`digital PCR and TAm-Seq showed a high degree of concordance
`(correlation coefficient 0.8, P , 0.0001; Supplementary Fig. 1). Using
`as little as 2.3 ng of DNA (4%–20% of the DNA extracted from
`2.0–2.2 ml of plasma), and an average of 169 million reads of sequenc-
`ing per sample, we analysed the coding exons of all protein-coding
`genes at an average unique coverage depth ranging from 31-fold to
`160-fold across 19 plasma samples (Supplementary Table 2). Con-
`sistent with previous reports5,7, we observed copy number aberrations
`(CNAs, both gains and losses) in plasma samples in all patients
`across the whole genome (Supplementary Figs 2–7). These were
`strongly modulated by the fraction of tumour DNA in plasma and
`were particularly prominent in plasma samples in which mutant AF
`exceeded 50%.
`
`©2013
`
`Macmillan Publishers Limited. All rights reserved
`
`2 M A Y 2 0 1 3 | V O L 4 9 7 | N A T U R E | 1 0 9
`
`00002
`
`

`

`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`10
`9
`8
`Chromosome
`d
`
`11
`
`12 13 14 15 16 17 18 19 202122X
`
`0.7
`
`0.6
`
`0.5
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0
`
`Detected in
`primary tumour
`
`Not detected
`in primary tumour
`
`1
`
`0
`
`–1
`
`1
`
`0
`
`–1
`
`0.5
`
`0.4
`
`0.3
`
`0.2
`
`0.1
`
`0
`
`a
`
`Plasma LRR
`
`b
`
`Tumour LRR
`
`c
`
`AF in matched plasma
`
`RESEARCH LETTER
`
`0
`
`0.3
`0.25
`0.2
`0.15
`0.1
`0.05
`AF in matched tumour biopsy
`
`0.35
`
`0
`
`0.6
`0.5
`0.4
`0.3
`0.2
`0.1
`AF in matched tumour biopsy
`
`0.7
`
`Figure 3 | Genome-wide concordance between plasma DNA and tumour
`DNA. a, b, Sequencing data were used to assess CNAs in the plasma sample
`(a) and in the synchronous metastatic tumour biopsy (b) from case 4. Panels
`show log R ratio (LRR), calculated on the basis of exome data, between plasma
`DNA and normal DNA (a) and between tumour and normal DNA (b). c, AF of
`
`mutations identified in exome data from plasma or metastatic biopsy for case 1.
`Grey dotted line shows equality. Blue dashed line has a slope of 1.93, indicating
`the median of the AF ratio for mutations found in both samples. Key applies to
`c and d. d, As c but for case 4, blue dashed line has a slope of 0.37.
`
`For two cases, sequencing data were also available from metastatic
`tumour biopsies, collected at the same time as plasma samples (case 1
`sample E1, and case 4 sample E2), and from tumour samples collected
`at the patients’ initial presentation, 9 and 4.5 years earlier. CNAs were
`concordant between plasma and metastasis DNA in both patients
`(Fig. 3a, b, and Supplementary Fig. 7). Mutations identified in sequencing
`data20–23 from the plasma or metastatic biopsy were compared (Sup-
`plementary Information). In case 1 with breast cancer, 151 mutations
`were identified in either the plasma or the synchronous biopsy. Of
`these, 93 mutations were found in both, and mutant AFs for these were
`higher in the plasma sample compared to the metastatic biopsy. The
`correlation coefficient of mutant AFs was positive (0.71) for mutations
`that were also found in the primary tumour, but negative (20.22) for
`other mutations (Fig. 3c). In case 4 with ovarian cancer, 895 mutations
`were identified in either plasma or the tumour biopsy. For 172 muta-
`tions found in both, AFs were positively correlated (0.72) and were
`higher in the metastatic biopsy, which also contained 686 ‘private’
`mutations with AF , 0.2 that were not found in either the plasma or
`the earlier tumour sample (Fig. 3d).
`To identify changes in the mutation profiles of the tumours, we
`compared the abundance of somatic mutations found in plasma before
`and after each course of systemic treatment. For each patient, we
`examined a conservative list of mutations, including all mutations that
`were called in any of the plasma samples with a Bonferroni-corrected
`binomial probability of ,0.05 assuming a background sequencing
`error rate of 0.1%. For each mutation and course of treatment
`(spanned by a pair of plasma samples), a P-value for a possible change
`in mutant AF was calculated as the binomial probability of obtaining
`the observed number of mutant reads, given the sequencing depth and
`the observed abundance in the paired time-point, normalized by the
`fractional concentration of tumour-derived DNA in the plasma (based
`on genome-wide aggregated allelic loss5, Supplementary Table 3).
`Overall, 364 non-synonymous mutations passed with false discovery
`
`1 1 0 | N A T U R E | V O L 4 9 7 | 2 M A Y 2 0 1 3
`
`rate of ,10% for significant changes in normalized abundance, rang-
`ing from 15 to 121 for each case (median 49). These include mutations
`in well-known cancer genes, genes linked to drug resistance and drug
`metabolism, and genes not previously associated with carcinogenesis
`or therapy resistance (Supplementary Tables 4–9). Selected examples
`are shown in Table 1 and Fig. 2.
`We highlight here five examples. In case 1 with breast cancer, a
`strong increase was observed in the abundance of an activating muta-
`tion in PIK3CA following treatment with paclitaxel (Fig. 2a and
`Table 1). This mutation has been shown to promote resistance to
`paclitaxel in mammary epithelial cells8. In case 2, a patient with an
`oestrogen-receptor (ER)-positive, HER2-positive breast cancer, treat-
`ment with tamoxifen in combination with trastuzumab led to an in-
`crease in abundance of a nonsense mutation near the carboxy terminus
`of MED1, an ER co-activator that has been shown to be involved in
`tamoxifen resistance10,11. After further treatment of this patient with
`lapatinib in combination with capecitabine, we observed an increase
`in abundance of a splicing mutation in GAS6, the ligand for the tyro-
`sine kinase receptor AXL (Fig. 2b, Table 1). Activation of the AXL
`kinase pathway has been shown to cause resistance to tyrosine kinase
`inhibitors in NSCLC13 and resistance to lapatinib in ER-positive,
`HER2-positive breast cancer cell lines12. In case 4 with ovarian cancer,
`following treatment with cisplatin, we observed increase in abundance
`of a truncating mutation in the tumour-suppressor RB1 (Fig. 2d,
`Table 1), predicted to inactivate the RB1 protein (Supplementary
`Fig. 8). In the matched metastasis biopsy obtained after treatment,
`the mutation was found in 95% of sequencing reads (59 of 62), with
`apparent loss of heterozygosity at 13q containing the RB1 gene (Fig. 3a,
`b). Loss of RB1 has been linked with chemotherapy response9. Case 6
`was a NSCLC patient with an activating mutation in EGFR who was
`treated with gefitinib but progressed on treatment. Analysis by digital
`PCR detected the EGFR T790M mutation in plasma at progression,
`but not at the start of treatment. This mutation inhibits binding of
`
`©2013
`
`Macmillan Publishers Limited. All rights reserved
`
`00003
`
`

`

`Table 1 | Selected mutations whose mutant AF significantly increased following treatment
`Patient
`Cancer type
`Gene
`Effect
`Potential biological interest
`
`Associated treatment
`
`Mutant AF in plasma
`
`LETTER RESEARCH
`
`Case 1
`
`Breast
`
`PIK3CA
`
`Case 1
`
`Breast
`
`Case 1
`
`Breast
`
`BMI1
`
`SMC4
`
`E545K
`
`S324Y
`
`I1000S
`
`Case 1
`
`Breast
`
`FANCD2
`
`G56V
`
`Case 2
`
`Breast
`
`MED1
`
`S1179X
`
`Case 2
`Case 2
`
`Breast
`Breast
`
`ATM
`PDGFRA
`
`I2948F
`D714E
`
`Case 2
`
`Breast
`
`GAS6
`
`Splicing
`
`Paclitaxel
`
`Paclitaxel
`
`Paclitaxel
`
`Epirubicin
`
`Tamoxifen/trastuzumab
`
`Tamoxifen/trastuzumab
`Tamoxifen/trastuzumab
`
`Lapatinib/capecitabine
`
`Before
`
`14%
`
`3%
`
`14%
`
`3%
`
`4%
`
`6%
`0%
`
`6%
`
`4%
`
`After
`
`34%
`
`12%
`
`22%
`
`13%
`
`15%
`
`45%
`15%
`
`30%
`
`20%
`
`Case 2
`
`Breast
`
`Case 4
`
`Ovarian
`
`TP63
`
`RB1
`
`Case 4
`
`Ovarian
`
`ZEB2
`
`Case 4
`
`Ovarian
`
`MTOR
`
`Case 5
`
`Ovarian
`
`CES4A
`
`Splicing /
`S551G
`E580X
`
`Y663C
`
`P55S
`
`Retinoblastoma 1. Loss of RB1 associated with EMT and
`drug resistance9.
`Zinc finger E-box binding homeobox 2. Overexpression
`associated with cisplatin resistance in ovarian cancer28.
`K1655N Mechanistic target of rapamycin. Activating mutations in
`mTOR confers resistance to antimicrotubule agents29.
`Carboxylesterase 4A. Hydrolysis or transesterification of
`various xenobiotics.
`
`PI-3-kinase. p.E545K mutation associated with
`chemoresistance in mammary epithelial cells8.
`BMI1 polycomb ring finger oncogene. Associated with
`chemoresistance25.
`Structural maintenance of chromosomes 4.
`Downregulated in taxane resistant cell lines26.
`Fanconi anaemia complementation group D2. Chromatin
`dynamics and DNA crosslink repair27.
`Mediator complex subunit 1. Co-activator of ER with
`functional role in tamoxifen resistance10,11.
`Ataxia telangiectasia mutated.
`Platelet-derived growth factor alpha. Cell surface tyrosine
`kinase receptor.
`Growth arrest-specific 6. Ligand for AXL, overexpression
`associated with TKI resistance12,13.
`Tumour protein p63.
`
`Lapatinib/capecitabine
`
`Cisplatin
`
`Cisplatin
`
`Paclitaxel
`
`Carboplatin/paclitaxel
`Carboplatin/liposomal
`doxorubicin
`Carboplatin/paclitaxel
`Liposomal doxorubicin
`Gefitinib
`
`Gefitinib
`Gefitinib
`
`14%
`
`11%
`
`8%
`
`0%
`6%
`
`11%
`23%
`0%
`
`0%
`0%
`
`22%
`
`15%
`
`14%
`
`6%
`13%
`
`34%
`30%
`13%
`
`14%
`17%
`
`Case 5
`Case 5
`Case 6
`
`Case 6
`Case 6
`
`Ovarian
`Ovarian
`Lung
`
`Lung
`Lung
`
`BUB1
`PARP8
`EGFR
`
`TP53
`NFKB1
`
`M889K
`P81T
`T790M
`
`Y163C
`G489V
`
`Mitotic checkpoint serine/threonine-protein kinase.
`Poly [ADP-ribose] polymerase family, member 8.
`Epidermal growth factor receptor. Established to cause
`gefitinib resistance by inhibiting drug binding14.
`Tumour protein p5330.
`Nuclear factor kB30.
`
`Potential biological role and associations with drug resistance described in literature are highlighted. The ‘‘Effect’’ column lists predicted change in amino acid sequence.
`
`gefitinib to EGFR and has been established as the main driver of
`acquired resistance to gefitinib14. Unbiased analysis of plasma DNA
`by exome sequencing identified selection for this mutation amongst
`genomic changes that occurred following therapy (Fig. 2f, Table 1).
`In this proof of principle study, we demonstrate that exome analysis
`of plasma ctDNA represents a novel paradigm for non-invasive charac-
`terization of tumour evolution. Our data, together with recent reports5,7,
`show that CNAs and somatic mutations identified in ctDNA are
`widely representative of the tumour genome and provide an alternative
`method of tumour sampling that can overcome limitations of repeated
`biopsies. Cell-free DNA fragments from multiple lesions in the same
`individual all mix together in the peripheral blood5, therefore ctDNA is
`likely to contain a wider representation of the genomes from multiple
`metastatic sites, whereas mutations present in a single biopsy or minor
`sub-clone may be missed. This strengthens the case for the use of
`ctDNA as a biomarker for monitoring tumour burden or for the ana-
`lysis of hotspot mutation regions1,6,16,17, but also indicates that tracking
`different mutations for assessment of tumour heterogeneity and clonal
`evolution is now possible. Our data identified a subset of genes that were
`positively selected following treatment, many of which have been prev-
`iously associated with drug resistance. Other changes may represent
`‘passenger’ mutations or false-positives, but some are likely to contri-
`bute to resistance to therapy. Accumulating data across a large number of
`cases could identify new genes or pathways that are frequently mutated
`following specific treatment types, and help refine analysis algorithms.
`The approach we describe here may be broadly applicable to a large
`fraction of advanced cancers, where the median mutation burden in
`plasma (before start of treatment) is 5%–10% (refs 6, 16, 24). Analysis of
`acquired drug resistance is of particular utility in advanced or metastatic
`cancers, which is the target population for nearly all early phase clinical
`trials. Improvements in sequencing and associated technologies may
`enable similar analysis in cases with a lower tumour burden in plasma.
`At present, this non-invasive approach for characterizing cancer exomes
`in plasma is readily applicable to patients with high systemic tumour
`
`burden, enabling detailed and comprehensive evaluation of clonal
`genomic evolution associated with treatment response and resistance.
`
`METHODS SUMMARY
`Patients and samples. Cases 1–5 were recruited as part of prospective clinical
`studies at Addenbrooke’s Hospital, Cambridge, UK, approved by the local
`research ethics committee (REC reference nos 07/Q0106/63, 08/H0306/61 and
`07/Q0106/63). Case 6 was recruited as part of the ‘Hydroxychloroquine and
`gefitinib to treat lung cancer’ study (NCT00809237) at the National University
`Health System, Singapore, approved by the National Healthcare Group NHG
`IRB—DSRB 2008/00196. Written informed consent was obtained from patients,
`and serial blood samples were collected at intervals of $3 weeks.
`Extraction and sequencing of plasma DNA. DNA was extracted from plasma
`using the QIAamp circulating nucleic acid kit (Qiagen) according to the manu-
`facturer’s instructions. Barcoded sequencing libraries were prepared using a com-
`mercially available kit (ThruPLEX-FD, Rubicon Genomics). Pooled libraries were
`enriched for the exome using hybridization (TruSeq Exome Enrichment Kit,
`Illumina), quantified using quantitative PCR and pooled in 1:1 ratio for paired-
`end sequencing on a HiSeq2500 (Illumina).
`Variant calling and analysis. Sequencing data were demultiplexed and aligned
`to the hg19 genome using BWA20. Pileup files for properly paired reads with
`mapping quality $60 were generated using samtools22. AFs were calculated
`for all Q30 bases. A mutation was called if $4 mutant reads were found in plasma
`with $1 read on each strand, and no mutant reads were observed in germline DNA
`or in a prior plasma sample with $10-fold coverage. For comparison between
`consecutive plasma samples in a patient, we calculated the binomial probability of
`obtaining the observed AF (or greater) if the abundance of the mutant allele,
`normalized by tumour load in plasma (based on a modified genome-wide aggre-
`gated allelic loss method5), had remained constant between the two samples.
`
`Full Methods and any associated references are available in the online version of
`the paper.
`
`Received 5 October 2012; accepted 11 March 2013.
`Published online 7 April 2013.
`
`1. Diaz, L. A. Jr et al. The molecular evolution of acquired resistance to targeted EGFR
`blockade in colorectal cancers. Nature 486, 537–540 (2012).
`
`2 M A Y 2 0 1 3 | V O L 4 9 7 | N A T U R E | 1 1 1
`
`©2013
`
`Macmillan Publishers Limited. All rights reserved
`
`00004
`
`

`

`RESEARCH LETTER
`
`2.
`
`4.
`
`6.
`
`7.
`
`8.
`
`Aparicio, S. & Caldas, C. The implications of clonal genome evolution for cancer
`medicine. N. Engl. J. Med. 368, 842–851 (2013).
`3. Gerlinger, M. et al. Intratumor heterogeneity and branched evolution revealed by
`multiregion sequencing. N. Engl. J. Med. 366, 883–892 (2012).
`Shah, S. P. et al. The clonal and mutational evolution spectrum of primary triple-
`negative breast cancers. Nature 486, 395–399 (2012).
`5. Chan, K. C. et al. Cancer genome scanning in plasma: detection of tumor-
`associated copy number aberrations, single-nucleotide variants, and tumoral
`heterogeneity by massively parallel sequencing. Clin. Chem. 59, 211–224 (2013).
`Forshew, T. et al. Noninvasive identification and monitoring of cancer mutations by
`targeted deep sequencing of plasma DNA. Sci. Transl. Med. 4, 136ra168 (2012).
`Leary, R. J. et al. Detection of chromosomal alterations in the circulation of cancer
`patients with whole-genome sequencing. Sci. Transl. Med. 4, 162ra154 (2012).
`Isakoff, S. J. et al. Breast cancer-associated PIK3CA mutations are oncogenic in
`mammary epithelial cells. Cancer Res. 65, 10992–11000 (2005).
`9. Knudsen, E. S. & Knudsen, K. E. Tailoring to RB: tumour suppressor status and
`therapeutic response. Nature Rev. Cancer 8, 714–724 (2008).
`10. Cui, J. et al. Cross-talk between HER2 and MED1 regulates tamoxifen resistance of
`human breast cancer cells. Cancer Res. 72, 5625–5634 (2012).
`11. Nagalingam, A. et al. Med1 plays a critical role in the development of tamoxifen
`resistance. Carcinogenesis 33, 918–930 (2012).
`12. Liu, L. et al. Novel mechanism of lapatinib resistance in HER2-positive breast
`tumor cells: activation of AXL. Cancer Res. 69, 6871–6878 (2009).
`13. Zhang, Z. et al. Activation of the AXL kinase causes resistance to EGFR-targeted
`therapy in lung cancer. Nature Genet. 44, 852–860 (2012).
`14. Pao, W. et al. Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib
`is associated with a second mutation in the EGFR kinase domain. PLoS Med. 2, e73
`(2005).
`15. Punnoose, E. A. et al. Evaluation of circulating tumor cells and circulating tumor
`DNA in non-small cell lung cancer: association with clinical endpoints in a phase II
`clinical trial of pertuzumab and erlotinib. Clin. Cancer Res. 18, 2391–2401 (2012).
`16. Diehl, F. et al. Circulating mutant DNA to assess tumor dynamics. Nature Med. 14,
`985–990 (2008).
`17. McBride, D. J. et al. Use of cancer-specific genomic rearrangements to quantify
`disease burden in plasma from patients with solid tumors. Genes Chromosom.
`Cancer 49, 1062–1069 (2010).
`18. Yung, T. K. et al. Single-molecule detection of epidermal growth factor receptor
`mutations in plasma by microfluidics digital PCR in non-small cell lung cancer
`patients. Clin. Cancer Res. 15, 2076–2084 (2009).
`19. Lo, Y. M. et al. Maternal plasma DNA sequencing reveals the genome-wide genetic
`and mutational profile of the fetus. Sci. Transl. Med. 2, 61ra91 (2010).
`20. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler
`transform. Bioinformatics 25, 1754–1760 (2009).
`21. DePristo, M. A. et al. A framework for variation discovery and genotyping using
`next-generation DNA sequencing data. Nature Genet. 43, 491–498 (2011).
`22. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 25,
`2078–2079 (2009).
`23. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic
`variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164
`(2010).
`
`24. Diehl, F. et al. Detection and quantification of mutations in the plasma of patients
`with colorectal tumors. Proc. Natl Acad. Sci. USA 102, 16368–16373 (2005).
`25. Siddique, H. R. & Saleem, M. Role of BMI1, a stem cell factor, in cancer recurrence
`and chemoresistance: preclinical and clinical evidences. Stem Cells 30, 372–378
`(2012).
`26. Chang, H. et al. Identification of genes associated with chemosensitivity to SAHA/
`taxane combination treatment in taxane-resistant breast cancer cells. Breast
`Cancer Res. Treat. 125, 55–63 (2011).
`27. Sato, K. et al. Histone chaperone activity of Fanconi anemia proteins, FANCD2 and
`FANCI, is required for DNA crosslink repair. EMBO J. 31, 3524–3536 (2012).
`28. Haslehurst, A. M. et al. EMT transcription factors snail and slug directly contribute
`to cisplatin resistance in ovarian cancer. BMC Cancer 12, 91 (2012).
`29. VanderWeele, D. J., Zhou, R. & Rudin, C. M. Akt up-regulation increases resistance
`to microtubule-directed chemotherapeutic agents through mammalian target of
`rapamycin. Mol. Cancer Ther. 3, 1605–1613 (2004).
`30. Wu, C. C., Yu, C. T., Chang, G. C., Lai, J. M. & Hsu, S. L. Aurora-A promotes gefitinib
`resistance via a NF-kB signaling pathway in p53 knockdown lung cancer cells.
`Biochem. Biophys. Res. Commun. 405, 168–172 (2011).
`
`Supplementary Information is available in the online version of the paper.
`
`Acknowledgements We thank J. Langmore and K. Solomon (Rubicon Genomics) for
`early access to library preparation products. We thank L. Jones, S. Richardson,
`C. Hodgkin and H. Biggs for recruiting patients into the DETECT and CTCR-OVO4
`studies, all medical and ancillary staff in the breast and gynaecological cancer clinic
`and patients for consenting to participate. We thank the Human Research Tissue Bank
`at Addenbrooke’s Hospital which is supported by the NIHR Cambridge Biomedical
`Research Centre. We thank the Cancer Science Institute, National University of
`Singapore, and the Hematology-Oncology Research Group, National University Health
`System, Singapore for their support. We acknowledge the support of Cancer Research
`UK, the University of Cambridge, National Institute for Health Research Cambridge
`Biomedical Research Centre, Cambridge Experimental Cancer Medicine Centre,
`Hutchison Whampoa Limited, and the National Medical Research Council, Singapore.
`S.-J.D. is supported by an Australian NHMRC/RG Menzies Early Career Fellowship that
`is administered through the Peter MacCallum Cancer Centre, Victoria, Australia.
`
`Author Contributions M.M., S.-J.D., T.F., D.W.Y.T., D.G., J.D.B., C.C. and N.R. designed the
`study. M.M., D.W.Y.T. and T.F. developed methods. S.-J.D., C.P., A.S.C.W., T.M.C., J.D.B.
`and C.C. designed and conducted the prospective clinical studies. M.M., S.-J.D.,
`D.W.Y.T., D.G., T.F. and A.M.P. generated data. Z.K., S.H. and D.B. contributed sequencing
`data. M.M., F.M. and N.R. analysed sequencing data. S.-F.C. and J.H. contributed to
`experiments and data analysis. M.M., S.-J.D., D.W.Y.T., T.M.C., J.D.B., C.C. and N.R.
`interpreted data. M.M. and N.R. wrote the paper with assistance from S.-J.D., D.W.Y.T.,
`C.C., J.D.B. and other authors. All authors approved the final manuscript. J.D.B., C.C. and
`N.R. are the project co-leaders and joint senior authors.
`
`Author Information Reprints and permissions information is available at
`www.nature.com/reprints. The authors declare competing financial interests: details
`are available in the online version of the paper. Readers are welcome to comment on
`the online version of the paper. Correspondence and requests for materials should be
`addressed to J.D.B. (james.brenton@cruk.cam.ac.uk), C.C.
`(carlos.caldas@cruk.cam.ac.uk) or N.R. (nitzan.rosenfeld@cruk.cam.ac.uk).
`
`1 1 2 | N A T U R E | V O L 4 9 7 | 2 M A Y 2 0 1 3
`
`©2013
`
`Macmillan Publishers Limited. All rights reserved
`
`00005
`
`

`

`METHODS
`Sample collection. Cases 1–5: patients were recruited as part of prospective clin-
`ical studies at Addenbrooke’s Hospital, Cambridge, UK, approved by local
`research ethics committee (REC reference nos 07/Q0106/63, 08/H0306/61 and
`07/Q0106/63). Written informed consent was obtained from the patients. Serial
`blood samples were collected in EDTA tubes at intervals of $3 weeks, and cen-
`trifuged within 1 h at 820g for 10 min to separate the plasma from the peripheral
`blood cells. The plasma was then further centrifuged at 20,000g for 10 min to pellet
`any remaining cells. The plasma was then stored at 280uC until DNA extraction.
`Case 6: this patient was recruited as part of the ‘Hydroxychloroquine and
`gefitinib to treat lung cancer’ study (NCT00809237) at the National University
`Health System, Singapore, approved by the National Healthcare Group NHG IRB-
`DSRB 2008/00196. Blood was collected in CPT tubes (BD Vacutai