discrimination of germline egfr t790m mutations in plasma ... › content › ... · jennifer...

Biology of Human Tumors

Discrimination of Germline EGFR T790MMutations in Plasma Cell-Free DNA Allows Studyof Prevalence Across 31,414 Cancer PatientsYuebi Hu1, Ryan S. Alden1, Justin I. Odegaard2, Stephen R. Fairclough2, Ruthia Chen1,Jennifer Heng1, Nora Feeney3, Rebecca J. Nagy2, Jayshree Shah4, Bryan Ulrich3,Martin Gutierrez4, Richard B. Lanman2, Judy E. Garber5, Cloud P. Paweletz3, andGeoffrey R. Oxnard1

Abstract

Purpose:Plasma cell-freeDNA (cfDNA) analysis is increasinglyused clinically for cancer genotyping, but may lead to incidentalidentification of germline-risk alleles. We studied EGFR T790Mmutations in non–small cell lung cancer (NSCLC) toward the aimof discriminating germline and cancer-derived variants withincfDNA.

Experimental Design: Patients with EGFR-mutant NSCLC,some with known germline EGFR T790M, underwent plasmagenotyping. Separately, deidentified genomic data and buffy coatspecimens from a clinical plasma next-generation sequencing(NGS) laboratory were reviewed and tested.

Results: In patients with germline T790M mutations, theT790M allelic fraction (AF) in cfDNA approximates 50%, higherthan that of EGFR drivermutations. Reviewof plasmaNGS resultsreveals three groups of variants: a low-AF tumor group, a hetero-zygous group (�50%AF), and ahomozygous group (�100%AF).

As the EGFR driver mutation AF increases, the distribution of theheterozygous group changes, suggesting increased copy numbervariation from increased tumor content. Excluding caseswithhighcopy number variation, mutations can be differentiated intosomatic variants and incidentally identified germline variants.We then developed a bioinformatic algorithm to distinguishgermline and somatic mutations; blinded validation in 21 casesconfirmed a 100% positive predictive value for predicting germ-line T790M. Querying a database of 31,414 patients with plasmaNGS, we identified 48 with germline T790M, 43 with nonsqua-mous NSCLC (P < 0.0001).

Conclusions: With appropriate bioinformatics, plasma geno-typing can accurately predict the presence of incidentally detectedgermline risk alleles. This finding in patients indicates a need forgenetic counseling and confirmatory germline testing. Clin CancerRes; 23(23); 7351–9. �2017 AACR.

IntroductionGenomic analysis of cancer specimens has beenwidely adopted

as a tool for guiding precision cancer care. Many academic andcommercial labs now offer next-generation sequencing (NGS)panels (1, 2) to identify targetable somatic mutations. Onelimitation of tumor tissue NGS is the frequent inability to clearlydistinguish whether detected variants represent somatic events,potentially driving tumor growth, or underlying germline events,potentially representing inherited risk alleles or benign inherited

polymorphisms (3). Indeed, some have advocated for routinepaired analysis of tumor DNA and germline DNA to betterdistinguish potentially targetable somatic alterations (3). How-ever, sequencing of germline DNA requires collection of anadditional germline biospecimen, increases cost, andmay requireadditional patient consent due to ethical considerations (4).

One rapidly emerging alternative to tumor tissue genotypingis genomic analysis of plasma cell-free DNA (cfDNA), whichcan noninvasively identify targetable somatic genomic variants.Because there is variability from patient to patient in regards tohow much tumor DNA is shed into the plasma, plasma cfDNArepresents a dynamic mixture of germline cfDNA and circulat-ing tumor-derived cfDNA (ctDNA). When ctDNA levels arehigh, plasma genotyping is technically similar to tumor tissuegenotyping; when tumor DNA is absent, cfDNA genotyping istechnically more similar to germline genotyping. In practice,ctDNA levels are typically between these extremes, which couldtheoretically allow for the segregation of somatic and germlinecfDNA variants. We hypothesized that genomic analysis ofplasma cfDNA could permit simultaneous tumor and germlinegenotyping, with accurate resolution of tumor-derived variantsand germline variants.

To address this question, we chose to study somatic and germ-line variants in the EGFR gene, which include known germlinevariants and a well described group of oncogenic mutationspresent in 10%–20% of NSCLC patients. One EGFR mutation,

1Lowe Center for Thoracic Oncology, Dana-Farber Cancer Institute, Boston,Massachusetts. 2Guardant Health, Redwood City, California. 3Belfer Center forApplied Cancer Science, Dana-Farber Cancer Institute, Boston, Massachusetts.4John Theurer Cancer Center, Hackensack Meridian Health, Hackensack, NewJersey. 5Center for Cancer Genetics and Prevention, Dana-Farber Cancer Insti-tute, Boston, Massachusetts.

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

Y. Hu and R.S. Alden contributed equally to this article.

Corresponding Author: Geoffrey R. Oxnard, Dana-Farber Cancer Institute, 450Brookline Avenue, Boston, MA 02215. Phone: 617-632-6049; Fax: 617-632-5786;E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-17-1745

�2017 American Association for Cancer Research.

ClinicalCancerResearch

www.aacrjournals.org 7351

on July 8, 2020. © 2017 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst September 25, 2017; DOI: 10.1158/1078-0432.CCR-17-1745

http://crossmark.crossref.org/dialog/?doi=10.1158/1078-0432.CCR-17-1745&domain=pdf&date_stamp=2017-11-7

http://clincancerres.aacrjournals.org/

T790M, can be detected rarely as a germline variant where itspresence has been associated with familial lung cancer (5).Families harboring germline EGFR T790Mmutations are current-ly under investigation at our center as part of an ongoing mul-ticenter clinical trial (NCT01754025; ref. 6). EGFR T790M ismorecommonly seen as an acquired somatic mutation after a patientwith NSCLC develops resistance to EGFR tyrosine kinase inhibi-tors (TKI). Testing for EGFR T790M has become routine becauselung cancers harboring T790M-mediated resistance after initialtherapy exhibit exquisite sensitivity to a third-generation EGFRTKI, osimertinib (7). Because of the challenges of repeat biopsiesin patients with advanced NSCLC, plasma genotyping to test forT790M has become common in patients with resistance to first-generation EGFR TKIs, and this is an area of active investigation atour center (8). The co-existence of two research programs studyingboth germline T790M and plasma genotyping for acquiredT790M created an opportunity to study the behavior of thesegermline variants in plasma cfDNA.

Case reportA 49-year-old never smoker with a family history of lung cancer

presented with metastatic lung adenocarcinoma; tissue genotyp-ing showed EGFR L858R and T790M mutations. Given its pre-clinical activity against both EGFR L858R and T790M mutations(9), first-line afatinib was started; however, the patient returnedwith progressive brain metastases after only 2 months of therapy.Plasma NGS then demonstrated the previously observed EGFRL858R at an allelic fraction (AF) of 5.3%, whereas the EGFRT790M allele was detected at 50.9% AF (Supplementary Fig.S1A). The patient was initiated on osimertinib, an EGFR TKIactive in the setting of EGFR T790M-mediated resistance to initialEGFR TKI (7), and had clinical benefit lasting 9 months, at whichtime scans showed early progression in the lung. Repeat plasmaNGS showed a reduced EGFR L858R at 0.6% AF but T790Mrelatively stable at 49.2% AF (Supplementary Fig. S1B). Afterfurther disease progression, repeat NGS demonstrated increasedlevels of EGFR L858R at 18% AF, T790M at 54% AF, and a thirdEGFRmutation that mediates acquired resistance to osimertinib,C797S, at 1.3% AF (10). The presence of an EGFR T790Mmutation at initial diagnosis, its high AF at cfDNA analysis, andthe family history of lung cancer raised suspicion that the EGFRT790M mutation might have represented a germline risk allele.

Materials and MethodsProtection of human subjects

Patients at DFCI were enrolled to an institutional review board(IRB) approved correlative study allowing plasma collection forgenomic analysis. Guardant Health separately received IRBapproval for analysis of deidentified data for research purposes;clinical data such as age, stage, and diagnosis was submitted attime of testing on a requisition form. An exemption from IRBreview was obtained to study deidentified plasma NGS resultsprovided to DFCI by Guardant Health.

Droplet digital PCRBlood (6–10 mL) was collected into EDTA lavender capped

vacutainer tubes and centrifuged for 10minutes at 1,200� g. Theplasma supernatant was further cleared by centrifugation for 10minutes at 3,000 � g. The second supernatant was stored incryostat tubes at �80�C until use. Cell-free DNA was isolatedusing theQIAmpCirculating Nucleic Acid Kit (Qiagen) accordingto manufacturer's protocol. Droplet digital PCR (ddPCR) wasperformed as previously described (11). Briefly, TaqMan PCRreaction mixtures were assembled from a 2 � ddPCR Mastermix(Bio-Rad) and custom 40x TaqMan probes/primers made specificfor each assay. Droplets were generated using an automateddroplet Generator (Bio-Rad). PCR was performed to endpoint.After PCR, droplets were read on either a QX-100 or QX200droplet reader (Bio-Rad). Analysis of the ddPCR data was per-formed with QuantaSoft analysis software (Bio-Rad). All ddPCRreagents were ordered from Bio-Rad. All primer and probes forwere custom-ordered from Life Technologies.

PCR conditions.* EGFR L858R. Forward primer, 50-GCAGCATGTCAAGATCAC-

AGATT-30, reverse primer, 50-CCTCCTTCTGCATGGTATTCT-TTCT-30; probe sequences: 50-VIC-AGTTTGGCCAGCCCAA-MGB-NFQ-30, 50-FAM-AGTTTGGCCCGCCCAA-MGB-NFQ-30. Cycling conditions: 95�C x 10 minutes (1 cycle), 40 cyclesof 94�C � 30 seconds and 58�C � 1 minute, and 10�C hold.

* EGFR del 19. Forward primer, 50-GTGAGAAAGTTAAAATTCCC-GTC-30, reverse primer, 50-CACACAGCAAAGCAGAAAC-30; probe sequences: 50-VIC-ATCGAGGATTTCCTTGTTG-MGB-NFQ-30, 50-FAM-AGGAATTAAGAGAAGCAACATC-MGB-NFQ-30. Cyclingconditions: 95�C� 10 minutes (1 cycle), 40 cycles of 94�C�30 seconds and 55�C � 1 minute, followed by 10�C hold.

* EGFR T790M. Forward primer, 50-GCCTGCTGGGCATCTG-30,reverse primer, 50-TCTTTGTGTTCCCGGACATAGTC-30; probesequences are: 50-VIC-ATGAGCTGCGTGATGAG-MGB-NFQ-30,50-FAM-ATGAGCTGCATGATGAG-MGB-NFQ-30. Cyclingconditions: 95�C � 10 minutes (1 cycle), 40 cycles of 94�C� 30 seconds and 58�C � 1 minute, followed by 10�C hold.

Plasma next-generation sequencingAll plasma sequencing was done using Guardant360 v2.9

(Guardant Health), as described previously (12), as part ofroutine clinical patient care or through a dedicated research study.Briefly, cfDNA was isolated from 10mL of whole blood drawn inStreck cell-free DNA tubes, enriched by hybrid capture targetingexons of 70 genes and critical introns of six genes, and sequencedto an average depth of approximately 15,000� on an IlluminaNextSeq500. Variant identification and origin assignment weredone as per standard clinical testing operations.

Translational Relevance

Although genomic analysis of plasma cell-free DNA(cfDNA) is increasingly performed to gain insight into cancerbiology, cfDNA remains a poorly understood biospecimenmade up of DNA from multiple sources. Here, we developmethods for differentiating germline and somatic variantswithin plasma cfNDA, using germline and somatic EGFRmutations as a clinical model. We then develop and validatea bioinformatic approach permitting us to identify and studyincidental germline EGFR T790Mmutationswithin a databaseof plasma next-generation sequencing results. Our approachsuggests cfDNA genomics, currently used for somatic cancergenotyping,may also offer an untapped opportunity for germ-line genomic research.

Hu et al.

Clin Cancer Res; 23(23) December 1, 2017 Clinical Cancer Research7352




Germline sequencingFor selected cases, deidentified buffy coat specimens from

Guardant were provided to DFCI. With IRB approval, genomicDNA was extracted for Sanger sequencing of EGFR at the Labo-ratory for Molecular Medicine (LMM) at Harvard, using a sampleID that cannot be linked to the original participant in any way.

Statistical analysisThe relationship between AF of EGFR driver mutation and

measures of copy number variation within the heterozygousgroup of variants was analyzed using a linear regression. Theprobability density function of the distribution of the standarddeviation and mean AFs for individual cases was estimatedusing a Gaussian approximation, and outliers were identifiedusing the Tukey method (13). Wilson/Brown method was usedto determine the 95% confidence interval for EGFR T790Mprevalence in each diagnosis of interest (14). The prevalenceamong the different diagnoses was compared using a two-tailedFisher exact test.

ResultsPlasma ddPCR of EGFR T790M reveals different behaviors ofsomatic and germline variants in cfDNA

We first studied 85 patients with advanced NSCLC who hadan EGFR T790M mutation detected in plasma cfDNA using

our previously described ddPCR platform (15). Of thesepatients, three were known to carry a germline EGFRT790M mutation based on prior germline sequencing, where-as the remainder had acquired EGFR T790M after TKI treat-ment. Studying the absolute concentration of T790M alleles incopies/mL of plasma, we found that some cases with somaticT790M carried an even higher concentration of mutantT790M alleles in plasma than the three cases with germlineEGFR T790M (Fig. 1A). In contrast, when we looked at the AFof T790M, calculated as the proportion of copies of mutantT790M out of all mutant or wild-type variants at that locus,the AF of the three germline cases hovered around 50%,higher than the AF of the somatic T790M cases (Fig. 1A). Wethen studied changes in the levels of somatic versus germlineEGFR mutations in plasma cfDNA upon treatment. In patientscarrying acquired EGFR T790M after first-generation TKI resis-tance, the concentration of both the EGFR T790M mutationsand the driver mutations (e.g., L858R or exon 19 deletion)decreased dramatically in response to therapy (Fig. 1B). Incontrast, in patients with germline EGFR T790M mutations,their EGFR driver mutation responded on therapy while theEGFR T790M levels stayed relatively stable (Fig. 1B). Thesedata provided proof-of-concept that quantitation of variantlevels in plasma cfDNA can potentially be used to discrimi-nate between somatic and germline origins of tumor-associ-ated mutations.

A

B

C

1

10

100

1,000

10,000

100,000

Mut

ant c

once

ntra

tion

(cop

ies/

mL)

ND

25

50

Mut

ant a

llelic

frac

tion

(%)

Acquired T790M

Germline T790M

65

ND

n = 85n = 85

1

10

100

1,000

10,000Acquired T790M

Mut

ant c

once

ntra

tion

(cop

ies/

mL)

EGFR driver mutation

ND

EGFR T790M

On therapyBaseline

Germline T790M

On therapyBaseline1007550250

10

20

30

40

Rel

ativ

e fr

eque

ncy

(%) EGFR Q787Q

0

10075502500

10

20

30

40R

elat

ive

freq

uenc

y (%

) EGFR driver mutation

10075502500

10

20

30

40

Allelic fraction (AF)

Rel

ativ

e fr

eque

ncy

(%) EGFR T790M

Figure 1.

Germline and tumor-derived EGFRmutations within plasma cell-freeDNA. A, Across ddPCR results for85 patients, germline T790Mmutations (green) are present at asimilar concentration as somaticT790M mutations (gray), but at ahigher allelic fraction (AF).B, StudyingddPCR results for four patients ontreatment, each represented by adifferent color, the concentrations ofsomatic EGFR mutations decreasewhile the concentration of germlineEGFR T790M remains constant.C, Across plasma NGS results for 950cases, the AF distribution for EGFRT790M (green) includes a somaticpeak also seen with EGFR drivermutations (blue), as well as aheterozygous peak (arrow) also seenmore clearly with a common SNP(EGFR Q787Q, gold).

Detection of Germline Risk Alleles in Plasma cfDNA

www.aacrjournals.org Clin Cancer Res; 23(23) December 1, 2017 7353




Plasma NGS identifies three populations of variants detectedwithin cfDNA

Although quantitative plasma genotyping is feasible withddPCR, it is difficult to multiplex PCR assays across multiplegenes. In contrast, next-generation sequencing (NGS) has thepotential to capture a broad range of variants across a numberof cancer-related genes. To further investigate the behavior ofgermline and somatic EGFR mutations in plasma cfDNA, weutilized the Guardant360 version 2.9 plasma NGS assay (Guar-dant Health Inc.; Supplementary Table S1). This assaysequences exonic regions of 70 oncogenes and tumor suppres-sor genes, and intronic regions of six genes in which oncogenicrearrangements occur (12). We first queried the Guardantdatabase of clinical plasma NGS results to study the distribu-tion of somatic and germline EGFR mutations. We identifiedtest sets of 950 consecutive NSCLC samples for each of thefollowing: known somatic mutations (L858R and exon 19deletions), a common germline SNP within the EGFR tyrosinekinase domain (Q787Q; ref. 16), and T790M, and plotted theAF distribution of each (Fig. 1C). The distribution of the knownSNP comprised two discrete, normally distributed probabilitydistributions centered at AFs of 50% and 100%, compatiblewith heterozygosity and homozygosity of the Q787Q allele.The distribution of the known somatic alterations, L858R andexon 19 deletions, in contrast, demonstrated an exponentialdecay distribution beginning at the assay's limit of detectionwith the long tail extending to AFs exceeding 90%, compatiblewith somatic AFs that varied substantially but that were gen-erally low (<5%). The distribution of T790M conformedpredominantly to this same somatic distribution; however,there was a minor but discrete, normally-distributed subpop-ulation centered at 50% AF (Fig. 1C). This pattern supportsstudy of variant AF in cfDNA as a method for categorizingvariants like EGFR T790M, which may be of either germline orsomatic origin.

We further studied AF distribution by performing plasma NGSon pretreatment and on-treatment plasma specimens from threecases with germline EGFR T790M known to harbor EGFR drivermutations in their cancer (two with L858R, one with L861Q).Studying the AFs of all coding and noncoding variants identifiedon plasma NGS, we were able to clearly visualize three groups ofvariants (Fig. 2A). The lowest AF group of variants includes theEGFR driver and TP53 mutations, representing cancer-derivedvariants. The highest AF group of variants was centered around100% AF, representing homozygous germline variants. Finally, amiddle group of variants centered around 50%, includes theknown germline EGFR T790M mutations and represents hetero-zygous germline variants. On treatment with a third-generationEGFR TKI (two with osimertinib, one with ASP7283), the low AFcancer-derived variants decreased or became undetectable(24%!0.2%, 3.7%!ND, 1.1%!ND). In contrast, the middlegroup of heterozygous germline variants changed only modestlyand remained centered around50%AF (56%!49%, 52%!49%,49%!50%). Interestingly, some of these heterozygous variantsappeared to have an increase in AF as the cancer responded totherapy whereas others had a decrease in AF. We hypothesizedthat these changes in the heterozygous group on treatment couldrepresent a reduction in tumor-derived copy number variation,leading to a change in the variant AF in cfDNA.

We then similarly studied all coding and noncoding variantsfrom plasma NGS of the initial case presented above (Supple-

mentary Fig. S1A). This revealed a pattern similar to the germlineEGFR T790M cases studied, where the patient's EGFR T790Mmutation fell within the heterozygous group of variants, and theAF changed minimally on therapy as compared to the EGFRL858R mutation (Supplementary Fig. S1B). This pattern seemedsuspicious for an incidentally detected germline EGFR T790Mmutation detected with plasma NGS.

Increased AF of driver mutation in cfDNA is associated withincreased copy number variation in heterozygous variants

To further study the relationship between tumor content incfDNA and heterozygous copy number variation, we queried theGuardant database for an additional 63 plasma NGS cases whichwere positive for EGFR T790M and 39 plasma NGS cases positivefor anEGFRdrivermutationwithout T790M. These 105 cases eachhad a median of 107 coding and noncoding variants detected.Looking at the AF distribution of all 10,702 variants in total(Fig. 2B), one can clearly identify a trimodal distribution withthree AF peaks at approximately 0%, 49%, and 100%. Comparedwith noncoding exonic and intronic variants, coding missenseand nonsense variants were enriched in the low AF group ofvariants (Fig. 2C), consistent with this representing a group ofcancer-derived variants.

To study the relationship between potential germline andsomatic variants, we plotted each plasma NGS case individuallyin order of low to high AF of the EGFR driver mutations (Fig. 3A).Although AF of the driver mutation is not a perfect measure oftumor content in cfDNA (due to the presence or absence of EGFRgene amplification in some cases) it can serve as a rough estimateof tumor content in cfDNA across a cohort (17). Studying thedistribution of variant AFs in the heterozygous group, which wedesignated as all variants with an AF between 25% and 75%, wefound that the distribution changed as the AF of the EGFR drivermutation increased. An increase in EGFR driver AF was associatedwith an increased standard deviation of the heterozygous group(Fig. 3B) as well as an increase in the absolute difference betweenthe case and the population mean, both of which suggest thepresence of cancer-derived copy number variation. Studying thestandard deviation of the AF of the variants in the heterozygousgroup, we fit a normal distribution to 94 cases whereas 11 caseshad outlier characteristics (Supplementary Fig. S2A). Similarly,studying the median of the AF for the variants in the heterozygousgroup, we fit a normal distribution to 94 cases whereas 11 caseshad outlier characteristics (Supplementary Fig. S2B). Because theseoutlier populations overlapped, 16 cases in total exhibited one ofthese two outlier characteristics with evidence of high copy num-ber variation in cfDNA, thought to be due to high levels of tumorDNA causing variability in the AF of germline variants.

Because the high copy number variation can result in sub-stantial deviation of the AF of germline variants from theexpected 50%, we hypothesized that germline-somatic discrim-ination would be impaired in these outlier cases. We thussegregated these 16 outlier cases from the 89 cases withoutoutlier characteristics (Fig. 4). Visual review of the codingvariants of outlier cases (Fig. 4A), it is challenging to distinguisha clear separation between germline heterozygous variants andsomatic cancer-derived variants. In contrast, visual review of thecoding variants of cases without these characteristics of highcopy number variation (Fig. 4B), one can clearly distinguish agroup of heterozygous variants with AFs in the range of 35% to60%, which are non-overlapping with a group of cancer-

Hu et al.





derived variants with AFs below 30%. We thus propose that byexcluding plasma NGS cases with high copy number variation(and thus high tumor content), plasma NGS results can beaccurately differentiated into somatic variants within the can-cer-derived group and incidentally identified germline riskalleles within the heterozygous group.

Blinded validation of a bioinformatic algorithm fordiscriminating between germline and somatic alterations inplasma NGS data

Following the logic of these proof-of-concept studies, wedeveloped and evaluated an integrated bioinformatics algorithmto segregate germline and somatic alterations across the 70 genes

assayed using plasma NGS. This algorithm first assigns variants apresumed germline or somatic origin using a priori knowledge,including both internal and external databases of knowngermlineand somatic variants (pathogenic and benign). For example, theEGFR Q787Q alteration is a benign polymorphism present in�52% of germline exomes in the ExAC database (http://exac.broadinstitute.org/), allowing this to be designated as of pre-sumed germline origin regardless of allelic fraction (16). Con-versely, the EGFR L858R alteration is a relatively common onco-genic mutation in NSCLC but does not appear in germlinedatabases, allowing this to be designated as of presumed somaticorigin. Such a priori binning usually results in a median of 78variants per case being assigned as germline, which allows

A B

C

All Non-coding

Synon-ymous

Mis-sense

Non-sense

100

90

80

70

60

50

40

30

20

10

0

Alle

licfr

actio

n

Per ce ntag eof var i an ts

ND 0

5

10

15

20

25

30

35

40

0 25 50 75 1000.01

0.1

1

10

100

Allelic fraction (AF)

Rel

ativ

efr

eque

ncy

(%)

Base-line

Ontherapy

pt #3

Homozygousgroup

Heterozygousgroup

Tumorgroup

Base-line

Ontherapy

pt #2

EGFR T790MEGFR driver mutationTP53 mutationOther alterations

Base-line

Ontherapy

25

50

75

100

pt #1

AFof

alte

ratio

nsby

plas

ma

NGS

( %)

ND

Figure 2.

Discriminating germline and somatic variants in plasma cell-free DNA. A, Pretreatment and on-treatment plasma specimens from three initial cases, allpositive for germline EGFR T790M, were studied using plasma NGS. Studying all coding and noncoding variants detected, three groups of variantsare evident, corresponding to the expected AF of homozygous, heterozygous, and tumor-derived variants. Variants in the tumor-derived group respondon therapy while variants in the homozygous and heterozygous groups remain at a relatively constant AF. B, An additional 102 cases, for a total of 105 cases,were then studied using plasma NGS. Studying all coding and noncoding variants detected across 105 cases, a trimodal distribution is seen with peaks near0% (likely tumor derived), 49% (likely heterozygous), and 100% (likely homozygous). C, Formissense and nonsense variants, there is enrichment at lowAF (arrows),where tumor-derived variants would be expected to be found. In contrast, synonymous variants, likely reflecting benign germline polymorphisms, areenriched around 50% and 100% AF.





http://exac.broadinstitute.org/

http://exac.broadinstitute.org/


construction of a heterozygous probability distribution by variantAF as described in the studies above. If all presumed somaticmutations (which are generally fewer in number) are presentbelow the lower limit of this heterozygous germline distribution,germline-somatic discrimination of the remaining unassignedvariants proceeds according to their AF relative to the germlinedistribution described by a priori variant classification.However, ifthe AF of presumed somatic variants exceed the AF of the lowerlimit of the heterozygous germline distribution, or if extremechromosomal instability is detected (as assessed by the apparentdiploid fractionof the genome), germline/somatic discriminationis deemed uncertain for the variants remaining within that regionof overlap, and variants are presumed to be somatic in origin andreported as such. This approach is intended to allow identificationof suspected germline variants with a high positive predictivevalue, understanding that sensitivity for variants of germlineorigin will be reduced in settings of high tumor DNA content.

This algorithm was then applied to 21 prospectively collectedclinical samples with high AF (30–75%) EGFR T790Mmutationsdetected on plasma NGS (Supplementary Fig. S3). Cases were

segregated into two cohorts based upon the germline-somaticsegregation of EGFR T790M described above. Cohort A included11 cases in which the distribution of somatic versus germlinederived variants led to the prediction of a germline T790Mmutation being present. Cohort B included 10 cases in whichdetermination of germline vs. somatic was complicated by highcopy number variation and a broad heterozygous group. Thegenomic DNA-containing cellular fractions of each sample werethen irreversibly deidentified and, with IRB approval, submittedto a CLIA-certified clinical laboratory for EGFR sequencing in adouble-blinded manner such that no germline result was trace-able to any individual patient. All 11 cases in cohort A wereconfirmed to harbor a germline EGFR T790M (positive predictivevalue 100%, 11/11). Of 10 cases in cohort B, one was found to begermline, resulting in a sensitivity of 92% (11/12) and an overallaccuracy of 95% (20/21). The presence of a germline sample inCohort B is suspected to be a case with high tumor content suchthat the AF of presumed somatic mutations overlapped withheterozygous germline distribution, making it impossible todiscriminate germline variants with certainty.

A B

25

50

75

100

Subject

Alle

licfra

ctio

nof

alte

ratio

nsby

plas

ma

NGS

(%)

EGFR driver mutation

Other coding alterationsNoncoding alterations

Mean of heterozygous bandEGFR Q787Q (known SNP)

ND

20 40 60 800

5

10

15

20

AF of EGFR driver mutation

Stan

dard

devi

atio

nof

AF

P < 0.0001R2 = 0.5581

ND

20 40 60 800

2

4

6

AF of EGFR driver mutation

Abs

olut

edi

ffere

nce

betw

een

c ase

and

popu

l atio

nm

ean

(%)

P < 0.0001R2 = 0.2327

ND

Figure 3.

Increased tumor content is associated with increased copy number variation within a heterozygous group of variants. A, AF of all variants found on plasmaNGS of 105 cases positive for EGFR mutations, in increasing order of EGFR driver mutation AF (blue), with a common EGFR SNP shown (gold). B, Studyingthe variant AFs between 25% and 75%, the SD and absolute difference between case and population mean were calculated; both increased with an increasein EGFR driver AF.

Hu et al.





Using a plasma NGS database to study the epidemiology ofgermline EGFR T790M

Having validated a method for identifying plasma NGS caseswhich carry germline EGFR T790M, we hypothesized that wecould use existing plasma NGS data to learn about the asso-ciation of germline variants with specific cancer types. We againqueried the Guardant clinical testing database of 31,414 con-secutive unique patients representing a wide variety of adult

solid tumor types to identify 911 cases positive for EGFRT790M, of which 48 were of germline origin as adjudicatedby the above methodology. Though nonsquamous NSCLCwas the cancer diagnosis in a minority of the overall patientcohort (41%), this was the cancer diagnosis in 43 of the 48patients with germline EGFR T790M (90%, Fig. 5A). Of theremaining five patients with germline EGFR T790M, three had arelated diagnosis (squamous NSCLC, small cell lung cancer,

A B

25

50

75

100

Subject

Alle

licfra

cti o

nof

alte

ratio

nsby

plas

ma

NGS

(%)

EGFR T790MEGFR driver mutationOther coding alterations

ND

25

50

75

100

Subject

Alle

licfra

ctio

nof

alte

ratio

nsby

plas

ma

NGS

(%)

ND

Figure 4.

Distinguishing heterozygous andtumor-derived coding variants incases with low copy number variation.In outlier cases with high copy numbervariation (A), it is difficult todistinguish germline and somaticvariants. When there is lower copynumber variation (B), it is possible tovisually distinguish which cases ofgermline EGFR T790M (green) arelikely germline.

Germline T790M detected

No germline T790M detected

Total 48 31,366Nonsquamous

NSCLC 43 (90%) 12,731 (41%)Other cancers 5 (10%) 18,635 (59%)

A B

NonsquamousNSCLC

Othercancers

Overall population

0.001

0.01

0.1

1

Prev

alen

ce(%

)

**

Figure 5.

Prevalence of germline T790M. A, Querying a database of 31,414 unique cancer patients with plasma NGS results, 48 (0.15%) were found to carry a germlineEGFR T790M mutation. Nonsquamous non-small cell lung cancer (NSCLC) was the dominant diagnosis in these patients. B, As compared to thepopulation prevalence of germline EGFR T790M in a reference cohort (0.008%), there is a higher prevalence in subjects with nonsquamous NSCLC (0.34%)but not in subjects with other cancers (0.03%, P ¼ 0.06), suggesting germline EGFR T790M is a risk variant for lung cancer. � , P < 0.05.






carcinoma of unknown primary). The population frequency ofgermline EGFR T790M in patients with nonsquamous NSCLC(43/12774, 0.34%) was substantially higher than was seen inpatients with another cancer diagnosis (5/18640, 0.03%, Fig.5B), the latter being only moderately higher than that reportedby general population sequencing efforts (e.g., ExAC's medianallele frequency of 0.0082%) (16). These calculations areimperfect given the sub-perfect accuracy of determining germ-line status described above, but these observations are none-theless congruent with existing hypotheses that patients withgermline T790M are at increased risk specifically for NSCLC,and suggest that this allele does not confer substantiallyincreased risk of other cancers aside from lung cancer.

DiscussionThis analysis for thefirst timedemonstrates the power of cfDNA

genomics as a tool for investigation of germline cancer risk alleles.We developed and validated a bioinformatic algorithm to distin-guish germline variants from cancer-derived somatic variantswithin cfDNA NGS profiles, suggesting that a single assay canboth offer insight into tumor genotype for therapy selection aswell as screen for hereditary risk alleles. We then queried a largeclinical testing database to explore the rare germline allele, EGFRT790M, and observed enrichment for this mutation in patientswith nonsquamous NSCLC. We believe that the application ofcfDNA sequencing to study association of germline alleles withspecific cancer mutations represents a novel approach deservingfurther investigation.

As is the case with tumor NGS, it is expected that plasma NGSwill at times identify incidental germline mutations in cancerpatients that have potential clinical implications for patients andtheir families. Here, we focused on germline EGFRmutations, butthis approach can be applied to other germline mutations withpotential implications both for therapy and inherited risk, such asmutations in TP53 (18), BRCA1/2 (19), and mismatch repairgenes (20). Reporting clinically significant germline findingsincidental to somatic genomic testing in cfDNA may provideclinical benefits to advanced cancer patients. In a large tissue-based pan-cancer NGS study, 2.3%of patients were found to havepreviously unrecognized pathogenic germline variants (21). Inaddition, numerous reports in non-lung cancers have foundsignificant rates of potentially targetable germline alterations inpatientswhowould not be screened on the basis of family history,particularly in DNA repair genes (22–25).

In our validation, we found that presence of a suspected germ-line mutation using our bioinformatic algorithm predicted for aconfirmed germline mutation with a high positive predictivevalue, a critical diagnostic performance characteristic for a rule-in test. However, in the setting of high tumor content and highcopy number variation, germline variants will be extremely dif-ficulty to accurately distinguish from somatic variants. Impor-tantly, genotyping of cfDNA does not replace genetic testing ofgermline DNA—patients with incidentally detected germlinemutations on plasma NGS should be referred for formal geneticcounseling and confirmatory germline testing, as per guidelinerecommendation (26–28). Given the widespread use of plasmagenotyping for lung cancer care, patients with high levels of EGFRT790M on plasma genotyping are now eligible for germlinetesting on our ongoing study of families with germline EGFRT790M mutations (NCT01754025; ref. 6).

Discernment of germline and somatic variants in plasmacfDNA has potential impact on the understanding of cancerbiology. With tumor NGS, it is difficult to determine if a variantof unknown significance in an oncogene represents a potentialdriver mutation or a germline polymorphism. In plasma NGScases without high copy number variation, it now appears pos-sible todifferentiate these two types of genomic alterations using asingle assay, reducing the risk of a germline polymorphism beingtherapeutically targeted in error. In addition, in instances whereserial plasma genotyping is being used over time to monitorresponse and resistance to therapy, the ability to distinguishgermline and somatic variants in plasma cfDNA can make iteasier to accurately track tumor DNA levels. Finally, our approachcould aid accurate calculation of tumor mutation burden (TMB),an emerging biomarker for understanding sensitivity and resis-tance to immune checkpoint inhibitors (29), by reducing thechance of germline polymorphisms being mistaken for poten-tially antigenic somatic mutations.

One limitation of our analysis is the focus on a single raregermline variant, EGFR T790M; studies of other germlinecancer risk alleles are planned. A second limitation is our useof peripheral white cells as our source of germline DNA, anapproach which could be vulnerable to somatic mosaicismsuch as clonal hematopoiesis of indeterminate potential(CHIP; refs. 30, 31). However, the clear majority of mutationsfound in patients with clonal hematopoiesis are present at anAF in the range of 0% to 30%, and are unlikely to be mistakenfor germline variants. As large studies have found no somaticEGFR mutations within peripheral white cells, clonal hemato-poiesis would not interfere the analysis performed in thisreport. Additional validation studies could consider using analternate source for germline DNA.

Our data highlights the need for laboratories offering plasmagenotyping to be vigilant for germline variants and prepared toreport to providers when suspicion for a pathogenic germlinemutation is high. Moreover, clinicians ordering plasma genotyp-ingmust be prepared for the possibility of incidentally identifyinggermline risk alleles, and will need referral systems in place tocare for these patients should such results be identified. Prospec-tive studies are needed to better understand the frequency of suchincidental germline findings in routine oncology care.

Disclosure of Potential Conflicts of InterestJ.I. Odegaard, S.R. Fairclough, R.J. Nagy, and R.B. Lanman are employees of

Guardant Health; R.J. Nagy and R.B. Lanman have ownership interest inGuardant Health. J.E. Garber receives research funding or has research colla-borations with Myriad Genetics, AstraZeneca, and Ambry and receives consult-ing fees from Helix. C.P. Paweletz received consulting fees from Digital Bioa-nalytics and honoraria fromBio-Rad. G.R.Oxnard received consulting fees fromAstraZeneca and Inivata and honoraria from Bio-Rad and Sysmex. All otherauthors have no potential conflicts of interest to report.

Authors' ContributionsConception and design: Y. Hu, R.S. Alden, J.I. Odegaard, R.J. Nagy,R.B. Lanman, G.R. OxnardDevelopment of methodology: Y. Hu, R.S. Alden, J.I. Odegaard, R.B. Lanman,G.R. OxnardAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): Y. Hu, J.I. Odegaard, R. Chen, N. Feeney, R.J. Nagy,J. Shah, B. Ulrich, M. Gutierrez, R.B. Lanman, G.R. OxnardAnalysis and interpretation of data (e.g., statistical analysis, biostatistics, com-putational analysis): Y. Hu, R.S. Alden, J.I. Odegaard, S.R. Fairclough, J. Heng,N. Feeney, R.J. Nagy, M. Gutierrez, R.B. Lanman, C.P. Paweletz, G.R. Oxnard

Hu et al.





Writing, review, and/or revision of the manuscript: Y. Hu, R.S. Alden,J.I. Odegaard, S.R. Fairclough, R. Chen, N. Feeney, R.J. Nagy, J. Shah,M. Gutierrez, R.B. Lanman, J.E. Garber, C.P. Paweletz, G.R. OxnardAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): Y. Hu, J.I. Odegaard, R. Chen, J. Heng, R.J. Nagy,C.P. Paweletz, G.R. OxnardStudy supervision: G.R. Oxnard

Grant SupportThis research was supported in part by the Damon Runyon Cancer

Research Foundation, Anna Fuller Fund, Conquer Cancer Foundation of

ASCO, Bonnie J. Addario Lung Cancer Foundation, National Cancer Institute(NCI) of the NIH (R01 CA114465), NCI Cancer Clinical InvestigatorTeam Leadership Award (P30 CA006516 supplement), and Stading-YoungerCancer Research Foundation Thoracic Oncology Fund.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received June 19, 2017; revisedAugust 8, 2017; accepted September 18, 2017;published OnlineFirst September 25, 2017.

References1. Wagle N, Berger MF, Davis MJ, Blumenstiel B, Defelice M, Pochanard P,

et al. High-throughput detection of actionable genomic alterations inclinical tumor samples by targeted, massively parallel sequencing. CancerDiscov 2012;2:82–93.

2. Hagemann IS, Devarakonda S, Lockwood CM, Spencer DH, Guebert K,Bredemeyer AJ, et al. Clinical next-generation sequencing in patients withnon-small cell lung cancer. Cancer 2015;121:631–9.

3. Jones S, Anagnostou V, Lytle K, Parpart-Li S, Nesselbush M, Riley DR, et al.Personalized genomic analyses for cancer mutation discovery and inter-pretation. Sci Transl Med 2015;7:283ra53.

4. Schrijver I, Aziz N, Farkas DH, Furtado M, Gonzalez AF, Greiner TC, et al.Opportunities and challenges associated with clinical diagnostic genomesequencing: a report of the Association for Molecular Pathology. J MolDiagn 2012;14:525–40.

5. Bell DW, Gore I, Okimoto RA, Godin-Heymann N, Sordella R, Mulloy R,et al. Inherited susceptibility to lung cancer may be associated with theT790M drug resistance mutation in EGFR. Nat Genet 2005;37:1315–6.

6. Oxnard GR, Heng JC, Root EJ, Rainville IR, Sable-Hunt AL, Shane-CarsonKP, et al. Initial results of a prospective, multicenter trial to study inheritedlung cancer risk associated with germline EGFR T790M: INHERIT EGFR.J Clin Oncol 2015;33:1505.

7. Janne PA, Yang JC, Kim DW, Planchard D, Ohe Y, Ramalingam SS, et al.AZD9291 in EGFR inhibitor-resistant non-small-cell lung cancer. N Engl JMed 2015;372:1689–99.

8. SacherAG, PaweletzC,Dahlberg SE, AldenRS,O'Connell A, FeeneyN, et al.Prospective validation of rapid plasma genotyping for the detection ofEGFR and KRAS mutations in advanced lung cancer. JAMA Oncol2016;2:1014–22.

9. Regales L, Gong Y, Shen R, de Stanchina E, Vivanco I, Goel A, et al. Dualtargeting of EGFR canovercome amajor drug resistancemutation inmousemodels of EGFR mutant lung cancer. J Clin Invest 2009;119:3000–10.

10. Thress KS, Paweletz CP, Felip E, Cho BC, Stetson D, Dougherty B, et al.Acquired EGFR C797S mutation mediates resistance to AZD9291 in non-small cell lung cancer harboring EGFR T790M. Nat Med 2015;21:560–2.

11. Paweletz CP, Sacher AG, Raymond CK, Alden RS, O'Connell A, Mach SL,et al. Bias-corrected targeted next-generation sequencing for rapid, multi-plexed detection of actionable alterations in cell-free DNA from advancedlung cancer patients. Clin Cancer Res 2016;22:915–22.

12. Mack PC, Banks KC, Zill OA, Mortimer SA, Chudova DI, Odegaard J, et al.O.02: plasma next generation sequencing of over 5,000 advanced non-small cell lung cancer patients with clinical correlations. J Thorac Oncol2016;11:S168–S9.

13. Tukey JW. Exploratory data analysis. Reading, MA: Addison-WesleyPublishing Company; 1977.

14. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomialproportion. Stat Sci 2001;16:101–17.

15. Oxnard GR, Paweletz CP, Kuang Y, Mach SL, O'Connell A, Messineo MM,et al. Noninvasive detection of response and resistance in EGFR-mutantlung cancer using quantitative next-generation genotyping of cell-freeplasma DNA. Clin Cancer Res 2014;20:1698–705.

16. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al.Analysis of protein-coding genetic variation in 60,706 humans. Nature2016;536:285–91.

17. Oxnard GR, Thress KS, Alden RS, Lawrance R, Paweletz CP, Cantarini M,et al. Association between plasma genotyping and outcomes of treatmentwith osimertinib (AZD9291) in advanced non-small-cell lung cancer.J Clin Oncol 2016;34:3375–82.

18. VanderLaan PA, RangachariD,Mockus SM, SpotlowV, ReddiHV,MalcolmJ, et al.Mutations in TP53, PIK3CA, PTENandother genes in EGFRmutatedlung cancers: correlation with clinical outcomes. Lung Cancer 2017;106:17–21.

19. Somlo G, Frankel P, Arun B, Ma CX, Garcia A, Cigler T, et al. Efficacy of thePARP inhibitor veliparib with carboplatin or as a single agent in patientswith germline BRCA1- or BRCA2-associated metastatic breast cancer. ClinCancer Res 2017;23:4066–76.

20. Le DT, Uram JN,WangH, Bartlett BR, Kemberling H, Eyring AD, et al. PD-1blockade in tumors with mismatch-repair deficiency. N Engl J Med2015;372:2509–20.

21. Meric-Bernstam F, Brusco L, Daniels M, Wathoo C, Bailey AM, Strong L,et al. Incidental germline variants in 1000 advanced cancers on a prospec-tive somatic genomic profiling protocol. Ann Oncol 2016;27:795–800.

22. Shindo K, Yu J, Suenaga M, Fesharakizadeh S, Cho C, Macgregor-Das A,et al. Deleterious germline mutations in patients with apparently sporadicpancreatic adenocarcinoma. J Clin Oncol. In press.

23. Hampel H, Panescu J, Lockman J, Sotamaa K, Fix D, Comeras I, et al.Comment on: screening for lynch syndrome (Hereditary NonpolyposisColorectal Cancer) among endometrial cancer patients. Cancer Res2007;67:9603.

24. Abida W, Armenia J, Gopalan A, Brennan R, Walsh M, Barron D, et al.Prospective genomic profiling of prostate cancer across disease statesreveals germline and somatic alterations that may affect clinical decisionmaking. JCO Precision Oncol 2017:1–16.

25. Zhang S, Royer R, Li S,McLaughlin JR, Rosen B, RischHA, et al. Frequenciesof BRCA1 and BRCA2 mutations among 1,342 unselected patients withinvasive ovarian cancer. Gynecol Oncol 2011;121:353–7.

26. Daly MB, Pilarski R, Berry M, Buys SS, Farmer M, Friedman S, et al. NCCNguidelines insights: genetic/familial high-risk assessment: breast and ovar-ian, version 2.2017. J Natl Compr Canc Netw 2017;15:9–20.

27. Robson ME, Bradbury AR, Arun B, Domchek SM, Ford JM, Hampel HL,et al. American Society of Clinical Oncology Policy Statement Update:genetic and genomic testing for cancer susceptibility. J Clin Oncol2015;33:3660–7.

28. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, et al. Standardsand guidelines for the interpretationof sequence variants: a joint consensusrecommendation of the American College of Medical Genetics and Geno-mics and the Association for Molecular Pathology. Genet Med 2015;17:405–24.

29. Yarchoan M, Johnson BA III, Lutz ER, Laheru DA, Jaffee EM. Targetingneoantigens to augment antitumour immunity. Nat Rev Cancer 2017;17:209–22.

30. Jaiswal S, Fontanillas P, Flannick J, Manning A, Grauman PV,Mar BG, et al.Age-related clonal hematopoiesis associated with adverse outcomes.N Engl J Med 2014;371:2488–98.

31. Genovese G, Kahler AK, Handsaker RE, Lindberg J, Rose SA, Bakhoum SF,et al. Clonal hematopoiesis and blood-cancer risk inferred from bloodDNA sequence. N Engl J Med 2014;371:2477–87.






2017;23:7351-7359. Published OnlineFirst September 25, 2017.Clin Cancer Res Yuebi Hu, Ryan S. Alden, Justin I. Odegaard, et al. PatientsCell-Free DNA Allows Study of Prevalence Across 31,414 Cancer

T790M Mutations in PlasmaEGFRDiscrimination of Germline

Updated version

10.1158/1078-0432.CCR-17-1745doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2017/09/23/1078-0432.CCR-17-1745.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/23/23/7351.full#ref-list-1

This article cites 28 articles, 10 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/23/23/7351.full#related-urls

This article has been cited by 8 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/23/23/7351To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-17-1745

http://clincancerres.aacrjournals.org/content/suppl/2017/09/23/1078-0432.CCR-17-1745.DC1

http://clincancerres.aacrjournals.org/content/23/23/7351.full#ref-list-1

http://clincancerres.aacrjournals.org/content/23/23/7351.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/23/23/7351


discrimination of germline egfr t790m mutations in plasma ... › content › ... · jennifer...

Documents