profiling tumor circulating cell-free dna with an enhanced

1
Profiling tumor circulang cell-free DNA with an enhanced whole-exome to enable sensive assessment of somac mutaons Simo V. Zhang*, Mengyao Tan*, Fabio C. P. Navarro*, Josee M. Northco, Shuyuan Ma, Christopher S. Nelson, Devayani P. Bhave, L. Gordon Bentley, Manju Chinnappa, Dan Norton, Gabor Bartha, Jason Harris, John Lyle, Sean M. Boyle, John West, Richard Chen Personalis, Inc. | 1330 O’Brien Dr., Menlo Park, CA 94025 * Co-first author #29 Methods Whole-Exome NeXT LB assay and Silencer To enable sensive detecon across the exome, we developed an enhanced exome assay and chemistry that augments hard to sequence genomic regions such as regions of high GC content, to enable more uniform coverage across the exome. Addionally, we achieve a high average depth of approximately 2000X for the enre exome, with addional boosted depth (5000X) for 247 clinically relevant oncogenic and tumor suppressor genes to further enhance sensivity. For analysis, we developed a computaonal algorithm, Silencer, that enables accurate somac mutaon detecon without compromising sensivity in the plasma. Silencer is empowered by an error suppression model built from a panel of normal individual plasma samples, and custom filters including a dedicated blacklist that is tailored to our NeXT LB technology. Results 100% sensivity achieved for SNVs with AF≥1% Our NeXT LB plaorm successfully and consistently detected all 25/25 (100%) known SNVs across all 2% and 1% diluon replicates, and missed only one or two events (94.6% on average) in the 0.5% diluon SeraCare® samples. We also achieved 100% sensivity in 5%, 2.5%, and 1% Horizon diluon samples. The observed AFs in SeraCare and Horizon sample are well correlated with the expected AFs. Introducon The analysis of tumors using circulang cell-free DNA (cfDNA) is beginning to transform cancer diagnosis, prognosis, response to therapy and to enable disease progression monitoring. Most cfDNA assays are centered around the idenficaon of therapeucally aconable mutaons and only cover a limited number of genes, and as such cannot inform on all genec alteraons present in the tumors. To address this, we have developed a whole-exome scale cfDNA plaorm, NeXT Liquid Biopsy™, that enables sensive detecon and tracking of mutaons in approximately 20,000 genes. The NeXT LB plaorm generates comprehensive sequencing data derived from the cfDNA (plasma), the tumor (e.g. FFPE) and the paent’s germline (e.g. blood cells), and therefore enable to closely discover and monitor clonal and subclonal evoluon of mutaons in plasma, and novel genec mutaons following surgery, and/or treatment therapies. Our NeXT LB plaorm detected 1334 somac SNVs in a cohort of 9 head and neck cancer plasma samples, 31 of which were in clinically-relevant immuno-oncology genes. The amount of detected plasma SNVs would be significantly less if intersecng with genes included across all commerically available targeted panels. With NeXT LB, we could gain addional insights on IO pathways, such as innate and adapve immune responses. To further demonstrate our sensive detecon at the whole exome scale, we developed a cell culture media system that models the shed and degraded tumor DNA fragments seen in human plasma samples. We developed cell-free media samples from two breast cancer cell lines, HCC1143 and HCC1954, from which curated sets of high-quality known somac mutaons. As demonstrated in Figure 3B and 3C, our NeXT LB plaorm achieved >95% sensivity for variants with AF>=2%, and between 85% to 92% sensivity for mutaons with AF between 1% to 2% across co-culture cell media diluons. Figure 2: NeXT LB technology sensivity with Seracare (A) and Horizon (B) cfDNA reference standards. Conclusion We have developed a whole-exome scale NeXT LB technology that enables sensive monitoring and novel detecon of so- mac SNVs from cfDNA. The NeXT LB plaorm generates a much broader view of the tumor mutaonal landscape from the plasma than typical commercially available targeted-based liquid biopsy plaorms. The plaorm enables broader monitor- ing of changes in response to cancer therapy, mechanisms of drug resistance, and intra- and inter-tumor heterogeneity. [email protected] [email protected] [email protected] Contact: We next applied NeXT LB to assess tumor-plasma concordance and cfDNA cancer driver mutaons in two independent col- orectal cancer and head and neck cancer cohorts. NeXT LB consistently detected both tumor concordant and plasma dis- nct mutaons from cancer paent samples, both of which have implicaons for non-invasive monitoring of paent re- sponse to treatment and profiling tumor heterogeneity. Further, we observed dynamic allele frequency changes in cancer driver genes in checkpoint blockade non-responders, detecng increasing allele frequency over the course of therapy, highlighng NeXT LB’s ulity in non-invasive response monitoring. 101 63 121 70 45 74 83 27 96 187 30 23 198 29 34 Figure 3: NeXT LB technology whole-exome scale sensivity with HCC1143 (A) and HCC1954 (B) breast cancer cell media system. Figure 4: NeXT LB technology detects significantly more SNVs at whole-exome (A) or ImmunoOncology related genes (B, C) outside of footprint of targeted panels. Figure 5: The number (A) and the percentage (B) of tumor-plasma concordant mutaons idenfied from 5 late-stage colorectal cancer paents. Colorectal Cancer Paents Paent 4 Paent 5 Paent 1 Paent 2 Paent 3 Head and Neck Cancer Paents A B A B A B P aent 12: non-responder Figure 6:(A)-(B) Frequently mutated genes in HNSCC, such as TP53, GNAQ, HRAS were also detected in plasma samples. Paent 12 who is defined as non-responder by clinical measurement also demonstrated increased AF of HRAS mutaon post treatment. AF0.5% AF1% AF2% 1:115256529:T:C 1:43815009:G:T 10:43617416:T:C 12:25398284:C:T 13:28592642:C:A 14:105246551:C:T 17:7577120:C:T 17:7577538:C:T 17:7578406:C:T 19:3118942:A:T 2:209113113:G:A 20:57484420:C:T 3:138665163:G:C 3:178936091:G:A 3:178952085:A:G 3:41266124:A:G 4:1803568:C:G 4:55152093:A:T 4:55599321:A:T 5:112175639:C:T 7:140453136:A:T 7:55249071:C:T 7:55259515:T:G 9:5073770:G:T 9:80409488:T:G Diluon AF_Spectrum 0 0.01 0.02 0.03 AF0p5_50ng_1 AF0p5_50ng_2 AF0p5_50ng_3 AF1_50ng_1 AF1_50ng_2 AF1_50ng_3 AF2_50ng_1 AF2_50ng_2 AF2_50ng_3 NRAS/CSDE1 MPL RET KRAS FLT3 AKT1 TP53 TP53 TP53 GNA11 IDH1 GNAS FOXL2 PIK3CA PIK3CA CTNNB1 FGFR3 PDGFRA KIT APC BRAF EGFR EGFR JAK2 GNAQ AF Spectrum 0% 0.5% 1% 2% AF1% AF2.5% AF5% 1:115256530:G:T 1:115256536:C:T 12:25398284:C:T 3:178936091:G:A 7:55249071:C:T 7:55259515:T:G Diluon AF_Spectrum 0 0.02 0.04 0.06 HD_AF1_50ng_1 HD_AF1_50ng_2 HD_AF1_50ng_3 HD_AF2p5_50ng_1 HD_AF2p5_50ng_2 HD_AF2p5_50ng_3 HD_AF5_50ng_1 HD_AF5_50ng_2 HD_AF5_50ng_3 NRAS NRAS KRAS PIK3CA EGFR EGFR AF Spectrum 0% 1% 2.5% 5% 1334 93 0 500 1000 NeXT LB Targeted Panels Number of Plasma SNVs Plasma SNVs detected by different plaorms (Whole-Exome) A 31 16 0 10 20 30 NeXT LB Targeted Panels Number of Plasma SNVs Plasma SNVs detected by different plaorms (ImmunoOncology related genes) B Checkpoint Modulators Angen Presenng Machinery Chemokines Cytokines Innate Immune Response Adapve Immune Response 0 1 2 3 4 Number of Genes IO genes detected by NeXT LB outside of targeted panel C A B 0.75 0.80 0.85 0.90 0.95 1.00 0.5_1pct 0.5_5pct 1_2pct 1_5pct 2_3pct 2_5pct 3_4pct 3_5pct 4_5pct 5_10pct AF bins Sensivity HCC1143 0.75 0.80 0.85 0.90 0.95 1.00 0.5_1pct 0.5_5pct 1_2pct 1_5pct 2_3pct 2_5pct 3_4pct 3_5pct 4_5pct 5_10pct AF bins Sensivity HCC1954 Tumor SNVs Plasma SNVs Figure 1: The NeXT LB technology complements the ImmunoID NeXT plaorm by surveying the mutaon landscape and immune status at baseline, as well as the following mepoints. NeXT LB detected more SNVs outside of footprint of targeted panels Non-invasively profiling tumor heterogeneity and clinical relevant somac mutaons Whole-exome scale high sensivity achieved using cell media 0% 20% 40% 60% 80% Pt1 Pt2 Pt3 Pt4 Pt5 % SNVs detected in Plasma/Tumor Plasma Tumor ImmunoID NeXT Platform® T=0 T=1 T=2 T=N 0.764 0.05 0.032 HRAS:p.Gly13Arg ANKRD18B:p.Gly294Arg BTN2A2:p.Gln426Pro 0.015 0.058 0.04 0.045 NA 0.009 Tumor AF 0 0.2 0.4 0.6 0.8 Plasma AF 0 0.02 0.04 0.06 Tumor-Pre Plasma-Pre Plasma-Post 0 302 TRIM51 TMEM37 POTEE LILRA4 KRTAP10−11 EEF1A1 ARHGAP23 APOBR AGAP6 ZNF814 ZNF705G PPM1E PDE4DIP MUC12 HRAS CYP21A2 CD163L1 NBPF10 GNAQ TP53 12% 12% 12% 12% 12% 12% 12% 12% 12% 18% 18% 18% 18% 18% 18% 18% 18% 24% 24% 35% 0001_pre 0001_post 0002_pre 0002_post 0008_pre 0008_post 0009_pre 0009_post 0011_pre 0011_post 0012_pre 0012_post 0013_pre 0013_post 7140_pre 0014_pre 0014_post 0 6 Missense_Mutaon Nonsense_Mutaon Mul_Hit Mutaonal landscape detected in plasma samples

Upload: others

Post on 25-Dec-2021

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Profiling tumor circulating cell-free DNA with an enhanced

Profiling tumor circulating cell-free DNA with an enhanced whole-exome to enable sensitive assessment of somatic mutationsSimo V. Zhang*, Mengyao Tan*, Fabio C. P. Navarro*, Josette M. Northcott, Shuyuan Ma, Christopher S. Nelson, Devayani P. Bhave, L. Gordon Bentley, Manju Chinnappa, Dan Norton, Gabor Bartha, Jason Harris, John Lyle, Sean M. Boyle, John West, Richard Chen

Personalis, Inc. | 1330 O’Brien Dr., Menlo Park, CA 94025* Co-first author

#29

MethodsWhole-Exome NeXT LB assay and Silencer To enable sensitive detection across the exome, we developed an enhanced exome assay and chemistry that augments hard to sequence genomic regions such as regions of high GC content, to enable more uniform coverage across the exome. Additionally, we achieve a high average depth of approximately 2000X for the entire exome, with additional boosted depth (5000X) for 247 clinically relevant oncogenic and tumor suppressor genes to further enhance sensitivity. For analysis, we developed a computational algorithm, Silencer, that enables accurate somatic mutation detection without compromising sensitivity in the plasma. Silencer is empowered by an error suppression model built from a panel of normal individual plasma samples, and custom filters including a dedicated blacklist that is tailored to our NeXT LB technology.

Results100% sensitivity achieved for SNVs with AF≥1%Our NeXT LB platform successfully and consistently detected all 25/25 (100%) known SNVs across all 2% and 1% dilution replicates, and missed only one or two events (94.6% on average) in the 0.5% dilution SeraCare® samples. We also achieved 100% sensitivity in 5%, 2.5%, and 1% Horizon dilution samples. The observed AFs in SeraCare and Horizon sample are well correlated with the expected AFs.

Introduction The analysis of tumors using circulating cell-free DNA (cfDNA) is beginning to transform cancer diagnosis, prognosis, response to therapy and to enable disease progression monitoring. Most cfDNA assays are centered around the identification of therapeutically actionable mutations and only cover a limited number of genes, and as such cannot inform on all genetic alterations present in the tumors. To address this, we have developed a whole-exome scale cfDNA platform, NeXT Liquid Biopsy™, that enables sensitive detection and tracking of mutations in approximately 20,000 genes. The NeXT LB platform generates comprehensive sequencing data derived from the cfDNA (plasma), the tumor (e.g. FFPE) and the patient’s germline (e.g. blood cells), and therefore enable to closely discover and monitor clonal and subclonal evolution of mutations in plasma, and novel genetic mutations following surgery, and/or treatment therapies.

Our NeXT LB platform detected 1334 somatic SNVs in a cohort of 9 head and neck cancer plasma samples, 31 of which were in clinically-relevant immuno-oncology genes. The amount of detected plasma SNVs would be significantly less if intersecting with genes included across all commerically available targeted panels. With NeXT LB, we could gain additional insights on IO pathways, such as innate and adaptive immune responses.

To further demonstrate our sensitive detection at the whole exome scale, we developed a cell culture media system that models the shed and degraded tumor DNA fragments seen in human plasma samples. We developed cell-free media samples from two breast cancer cell lines, HCC1143 and HCC1954, from which curated sets of high-quality known somatic mutations. As demonstrated in Figure 3B and 3C, our NeXT LB platform achieved >95% sensitivity for variants with AF>=2%, and between 85% to 92% sensitivity for mutations with AF between 1% to 2% across co-culture cell media dilutions.

Figure 2: NeXT LB technology sensitivity with Seracare (A) and Horizon (B) cfDNA reference standards.

ConclusionWe have developed a whole-exome scale NeXT LB technology that enables sensitive monitoring and novel detection of so-matic SNVs from cfDNA. The NeXT LB platform generates a much broader view of the tumor mutational landscape from the plasma than typical commercially available targeted-based liquid biopsy platforms. The platform enables broader monitor-ing of changes in response to cancer therapy, mechanisms of drug resistance, and intra- and inter-tumor heterogeneity.

[email protected]@[email protected]

Contact:

We next applied NeXT LB to assess tumor-plasma concordance and cfDNA cancer driver mutations in two independent col-orectal cancer and head and neck cancer cohorts. NeXT LB consistently detected both tumor concordant and plasma dis-tinct mutations from cancer patient samples, both of which have implications for non-invasive monitoring of patient re-sponse to treatment and profiling tumor heterogeneity. Further, we observed dynamic allele frequency changes in cancer driver genes in checkpoint blockade non-responders, detecting increasing allele frequency over the course of therapy, highlighting NeXT LB’s utility in non-invasive response monitoring.

101 63121 70 4574 83 2796

187 3023 198 2934

Figure 3: NeXT LB technology whole-exome scale sensitivity with HCC1143 (A) and HCC1954 (B) breast cancer cell media system.

Figure 4: NeXT LB technology detects significantly more SNVs at whole-exome (A) or ImmunoOncology related genes (B, C) outside of footprint of targeted panels.

Figure 5: The number (A) and the percentage (B) of tumor-plasma concordant mutations identified from 5 late-stage colorectal cancer patients.

Colorectal Cancer Patients

Patient 4 Patient 5

Patient 1 Patient 2 Patient 3

Head and Neck Cancer Patients

A B

A B

A

BPatient 12: non-responder

Figure 6:(A)-(B) Frequently mutated genes in HNSCC, such as TP53, GNAQ, HRAS were also detected in plasma samples. Patient 12 who is defined as non-responder by clinical measurement also demonstrated increased AF of HRAS mutation post treatment.

AF0.5% AF1% AF2%

1:115256529:T:C1:43815009:G:T10:43617416:T:C12:25398284:C:T13:28592642:C:A14:105246551:C:T17:7577120:C:T17:7577538:C:T17:7578406:C:T19:3118942:A:T2:209113113:G:A20:57484420:C:T3:138665163:G:C3:178936091:G:A3:178952085:A:G3:41266124:A:G4:1803568:C:G4:55152093:A:T4:55599321:A:T5:112175639:C:T7:140453136:A:T7:55249071:C:T7:55259515:T:G9:5073770:G:T9:80409488:T:G

Dilution

AF_Spectrum

0

0.01

0.02

0.03

AF0p5_50ng_

1

AF0p5_50ng_

2

AF0p5_50ng_

3

AF1_50ng_

1

AF1_50ng_

2

AF1_50ng_

3

AF2_50ng_

1

AF2_50ng_

2

AF2_50ng_

3

NRAS/CSDE1

MPL

RET

KRAS

FLT3

AKT1

TP53

TP53

TP53

GNA11

IDH1

GNAS

FOXL2

PIK3CA

PIK3CA

CTNNB1

FGFR3

PDGFRA

KIT

APC

BRAF

EGFR

EGFR

JAK2

GNAQ

AF S

pect

rum

0%

0.5%

1%

2%

AF1% AF2.5% AF5%

1:115256530:G:T

1:115256536:C:T

12:25398284:C:T

3:178936091:G:A

7:55249071:C:T

7:55259515:T:G

Dilution

AF_Spectrum

0

0.02

0.04

0.06

HD_AF1_50ng_

1

HD_AF1_50ng_

2

HD_AF1_50ng_

3

HD_AF2p5_50ng_

1

HD_AF2p5_50ng_

2

HD_AF2p5_50ng_

3

HD_AF5_50ng_

1

HD_AF5_50ng_

2

HD_AF5_50ng_

3

NRAS

NRAS

KRAS

PIK3CA

EGFR

EGFR

AF S

pect

rum

0%

1%

2.5%

5%

1334

930

500

1000

NeXT LB Targeted Panels

Nu

mb

er o

f P

lasm

a SN

Vs

Plasma SNVs detected by different platforms(Whole-Exome)

A

31

16

0

10

20

30

NeXT LB Targeted Panels

Nu

mb

er

of

Pla

sma

SNV

s

Plasma SNVs detected by different platforms(ImmunoOncology related genes)

B

Checkpoint Modulators

Antigen Presenting Machinery

Chemokines

Cytokines

Innate Immune Response

Adaptive Immune Response

0 1 2 3 4Number of Genes

IO genes detected by NeXT LBoutside of targeted panel

C

A

B

●●

● ● ●

● ●● ●

0.75

0.80

0.85

0.90

0.95

1.00

0.5_1pct

0.5_5pct

1_2pct

1_5pct

2_3pct

2_5pct

3_4pct

3_5pct

4_5pct

5_10pct

AF bins

Sens

itivi

ty

HCC1143

● ●●

● ●

● ●

●●

0.75

0.80

0.85

0.90

0.95

1.00

0.5_1pct

0.5_5pct

1_2pct

1_5pct

2_3pct

2_5pct

3_4pct

3_5pct

4_5pct

5_10pct

AF bins

Sens

itivi

ty

HCC1954

Tumor SNVsPlasma SNVs

Figure 1: The NeXT LB technology complements the ImmunoID NeXT platform by surveying the mutation landscape and immune status at baseline, as well as the following timepoints.

NeXT LB detected more SNVs outside of footprint of targeted panels

Non-invasively profiling tumor heterogeneity and clinical relevant somatic mutations

Whole-exome scale high sensitivity achieved using cell media

0%

20%

40%

60%

80%

Pt1 Pt2 Pt3 Pt4 Pt5

% S

NVs

det

ecte

d in

Plas

ma/

Tum

or

PlasmaTumor

ImmunoID NeXT Platform®

T=0 T=1 T=2 T=N

0.764

0.05

0.032

HRAS:p.Gly13Arg

ANKRD18B:p.Gly294Arg

BTN2A2:p.Gln426Pro

0.015 0.058

0.04 0.045

NA 0.009

Tumor AF

00.20.40.60.8

Plasma AF

00.020.040.06

Tumor-Pre Plasma-Pre Plasma-Post

0

302

NA

NA

TRIM51TMEM37

POTEELILRA4

KRTAP10−11EEF1A1

ARHGAP23APOBRAGAP6

ZNF814ZNF705G

PPM1EPDE4DIP

MUC12HRAS

CYP21A2CD163L1NBPF10

GNAQTP53

12%12%12%12%12%12%12%12%12%18%18%18%18%18%18%18%18%24%24%35%

0001

_pre

0001

_pos

t

0002

_pre

0002

_pos

t

0008

_pre

0008

_pos

t

0009

_pre

0009

_pos

t

0011

_pre

0011

_pos

t

0012

_pre

0012

_pos

t

0013

_pre

0013

_pos

t

7140

_pre

0014

_pre

0014

_pos

t

0 6

Missense_MutationNonsense_Mutation

Multi_Hit

Mutational landscape detected in plasma samples