Advances in Translational

Research

Yoon-La Choi

Department of Pathology

Samsung Medical Center

GBCC 2013. 12, Seoul

Agenda

• Molecular classification, Multigene prognostic predictor

• Genomic landscape, Genomic sequencing to search “actionable” driver

• Patient-derived xenograft (PDX) model for pre-clinical assessment

Molecular classification, Multigene prognostic predictor

Commercially Available

Multigene Signature

Reis-Filho and Pusztal, 2011

Expression Signatures: Current Status

• Which signature(s)?

- Now commercially available: MammaPrint, OncotypeDx, Genomic

grade index, Intrinsic subtype(PAM50), others

- Coming soon: Others

• Which patients?

- All signatures more useful for assessing prognosis in ER+

cancers (i.e., luminal) than in ER- cancers (i.e., HER2 and basal

molecular types)

• PAM50 Risk of Recurrence (ROR) Score provided more prognostic information in endocrine-treated, ER+ node-, patients than OncotypeDx recurrence score, especially in HER2-group

- More patients scored as high risk and fewer as intermediate risk

• OncotypeDx Recurrence Score commonly used to identify patients with ER+ breast cancer who could possibly be spared cytotoxic therapy(i.e., those with low recurrence score)

• None has yet been validated in prospective clinical trials TALORx (OncotypeDx)

MINDACT (MammaPrint)

RxPONDER (OncotypeDx)

JCO 2013

• The genes are rarely overlapped among multi-gene signatures even with the same purpose of prognostic prediction of early breast cancer.

• Similar prognostic power even with different gene collection.

• The power of prognostic prediction does not increase with summation or combination.

• They use similar biologic characteristic of genes.

• Proliferation genes are the common driving force in all prognostic signatures.

Large-size discovery data set (>1,000)

Need of analytic tool, combining of biologic and statistical significance

Characteristics of Multi-gene signatures

Breast cancer prognostic marker development process

Data Set Collection and Analysis Simulation with candidate

genes

Roche Light Cycler 480 based Kit Retrospective Study Product and Central Lab

Ensel Oh, Breast Cancer Res Treat 2011

imm

unity

pro

life

ration

Proliferation

p1

p2

p3

i1

i2

i3

p1 p2 p3

Symbol Name Location

p-genes (proliferation related)

UBE2C Ubiquitin-conjugating enzyme E2C 20q13.12

PRC1 Protein regulator of cytokinesis1 15q26.1

CCNB2 Cyclin B2 15q21.3

CDC20 Cell division cycle 20 homologur 1p34.1

i-genes (immunity related)

CD3D CD3d molecule, delta (CD3-TCR complex) 11q23

CD52 CD52 molecule 1p36

CCL19 Chemokine (C-C motif) ligand 19 9p13

TRBV20-1 T-cell receptor beta variable 20-1 7q34

Poor

Good

Test Optimization

GenesWell TM CDx-BCT

□ Oncotype DX®

□ Survival Probability obtained from GenesWell™CDx

□ GenesWell™CDx

Linear Predictor(LP) = -aaa*p.mean + bbb*i.mean + ccc

Survival

Probability

If Recurrence Score is 30,

-> Chance Rate of Recurrence

within 10 year = 20%

: Chemotherapy

If LP is 4.9 (Patient 5)

5/10 Years Survival Probability =

98%,95%

: No Chemotherapy!

Comparative Advantages

1) Patent Position

- No Conflict of Prognostic

Biomarkers with those of

Oncotype DX

2) Automatic Quantification

System

- Short Analysis Time

- Safe Handling for Samples

- Higher Reproducibility

- Low Cost of Goods

- Reduce Labor Cost (1/10)

Genomic landscape, Genomic sequencing to search “actionable” driver

Nature 2013

Genomic Analysis

The New Frontier

• Identification of genomic alterations in tumors, particularly “driver alterations”

- mutations

- Copy number variations

- Translocations

• Variety of techniques (WGS, WES, WTS, Targeted NGS)

• Goal: To identify “actionable” targets in order to personalized therapy

Whole Genome Sequencing

Advantages Disadvantages

-Point mutations -CNVs -Small indels -Rearrangements -Somatic mutations in non-coding regions (promoters, enhances, and non-coding RNAs)

-Point mutations and indels: >30-fold haploid coverage -Rearrangements: >10-fold physical coverage -High cost -Complicated analysis

DNA sample

Randomly fragment genomic DNA

DNA fragment with adapters

DNA cluster

Sequencing

Fragmentation

Adapter Ligation

Amplification

Whole Exome Sequencing

Advantages Disadvantages

-High sequence coverage and less raw sequence and cost than WGS. -Point mutations -Small indels -CNAs (need further validation)

-Incomplete exome capturing -Limit to mutations in coding regions -Lack efficient methodology to detect structural variations

Whole Transcriptome Sequencing

Advantages Disadvantages

-Higher sequence coverage and less raw sequence and cost than WGS -Calling somatic point mutations -Identifying differently expressed genes -Detecting fusion genes

-Mutations in genes expressed at low levels is hampered owing to lack of statistical power -Failing to detect truncating mutations because of nonsense mutations

Targeted sequencing (Panel)

• The advent of NGS technology has enabled the development of cost-effective multi-gene sequencing panels.

Advantages Disadvantages

-Higher sequence coverage with clinical diagnostic level -Calling somatic point mutations -CAN, fusion gene -Custom-required design

-Incomplete target capturing -Lack efficient methodology to detect structural variations

Catalog cancer panel

Life Technologies Agilent Illumina

Item Ion AmpliSeq Cancer HotSpot Panel v2

HaloPlex Cancer Research Panel TruSeq Amplicon Cancer Panel

Input DNA 10 ng 225 ng 225 ng 250 ng

Gene 50 ea 47 ea 47 ea 48 ea

Target size ~10 Kb 9.937 Kb 9.937 Kb 35 Kb

Sequenceable size 94.607 Kb 83.204 Kb 70 Kb

Total amplicon 207 ea 3,092 ea 2,121 ea 212 ea

Amplicon size 154 bp 150 bp 150 bp 170~190 bp

Minimal order 8 rxn 16 rxn 96 rxn

Sequencer PGM

(314chip, 4plex) MiSeq

(300Mb, 6plex) PGM

(316chip, 4plex) MiSeq

(300Mb, 8plex)

Coverage 500 x 500 x 500x 500 x

Total cost /sample ~500 $ ~400 $ ~550$ ~350$

Clinical level NGS Cancer Panel

Clinical level NGS Cancer Panel

Title Foundation Medicine (USA)

PGDx (USA)

RainDance (USA)

Samsung Genome Institute, SMC

Gene 236 120 54 80 (500)

Sample amount >=50 ng gDNA ? >=250 ng gDNA >=50 ng gDNA

Sample type Fresh, FFPE Fresh, FFPE Fresh, FFPE Fresh, FFPE

Matched normal Not needed Needed Not needed Not needed

Mean read depth >250X >250X ? >500X

Turn around time 3 weeks 3 weeks ? 1 week

Price $5,800 ? ? <$1,000

Insurance Acquired Not acquired Not acquired Not acquired

Variant type SNP, INDEL, CNV, Chromosomal rearrangement

SNP, INDEL, CNV, Chromosomal rearrangement

SNP, INDEL SNP, INDEL, CNV, Chromosomal rearrangement

Sensitivity >99% at >=10% allele frequency

>=1% allele frequency

? >99% at>=1% allele frequency

Specificity >99% PPV ? ? >99% PPV

Comprehensive genome analysis

Bioinformatics Personal targets

• DB (in-house and public) • Tumor, normal, blood e.g. lung, stomach and colon

• Prediction of clinical outcomes

• Integration of clinical and genome information

• Targeted re-sequencing • WES/WTS • Single cell analysis on multiple serial samples

• Automated genome analysis

• Knowledge-based, customized analysis

High-precision genome analysis for N-of-1 trial

Genomic analysis of chemotherapy-refractory alveolar rhabdomyosarcoma

27-year-old male with abdominal wall mass

Multiple mass with

pleural seeding

VDC + IE for

3weeks

Near complete remission

(sample #1)

3month, chest wall

mass, Salvage VIP

Pleural biopsy

(sample #2)

Pleural effusion

(sample #3)

Malignant acites

(sample #4)

CDK4

Gene Name

Gene Description Exome sequencing (#2) Sanger sequencing

Allele (%) Variant Tumor(#1) Ascites(#3) Primary cell culture

(from #2)

SCN1B sodium channel, voltage-gated, type I, beta 52.4 c.218A>G:p.Y73C 218A>G 218A>G 218A>G

PPP1R3A protein phosphatase 1, regulatory subunit 3A 46.5 c.1941T>A:p.D647E 1941T>A 1941T>A 1941T>A

GRID2 glutamate receptor, ionotropic, delta 2 46.3 c.1944C>A:p.Y648X WT 1944C>A 1944C>A

APBA2 amyloid beta (A4) precursor protein-binding, family A, me

mber 2 43.4 c.266G>A:p.G89D WT 266G>A 266G>A

ZNF142 zinc finger protein 142 42.9 c.1138G>A:p.A380T WT 1138G>A 1138G>A

ZYG11A zyg-11 homolog A

(C. elegans) 41.9 c.1762G>T:p.E588X 1762G>T 1762G>T 1762G>T

RBFOX1 RNA binding protein, fox-1 homolog (C. elegans) 1 39.8 c.241C>A:p.H81N WT 241C>A 241C>A

TCF7L1 transcription factor 7-like 1 (T-cell specific, HMG-box)

31.6 c.1006GA:p.V336M WT WT WT

TEX13B testis expressed 13B 22.2 c.406C>T:p.L136F WT WT WT

DSCAML1 Down syndrome cell adhesion molecule like 1 13.5 c.3110T>G:p.L1037R WT WT WT

Whole exome sequencing

SCN1B PPP1R3A ZYG11A

SCN1B PPP1R3A ZYG11A GRID2 APBA2 ZNF142 RBFOX1

SCN1B PPP1R3A ZYG11A GRID2 APBA2 ZNF142 RBFOX1 TCF7L1 TEX13B

DSCAML1

SCN1B PPP1R3A ZYG11A GRID2 APBA2 ZNF142 RBFOX1 TCF7L1 TEX13B

DSCAML1

CDK inhibitor

Genomic Analysis

The New Frontier

• Caveats

- “Actionable” = Clinically useful

- Many tumors have multiple genetic alterations

- Intratumoral heterogeneity

- Genetic alterations frequently change during the course of tumor progression

Patient-derived xenograft model (AVATAR) for pre-

clinical assays

Tentler, J. J. et al. (2012) Nat. Rev. Clin. Oncol.

Establishment and Testing of PDX models

“Avatars” : mice or other animals with human tissue implanted onto them. (xenografts)

Need of AVATAR

• Which drugs are the most effective for the patient is difficult, and testing different drugs puts the patient at risk.

• Personalized xenograft model : a large number of cancer drugs can be tested for their effectiveness on a particular patient.

• Xenograft model often resembles the tumor in the patient better than cultured cancer.

Challenges and limitations

• Difficult to make xenografts. Success rates of tumor implantation depend according to cancer types, and it can be as low as <1% in prostate cancer and ~10% in breast cancer.

(~70% in GBM, ~60% in Lung cancer in SMC IRCR)

• Creation of xenograft models and drug screening take about 6-8 months

• Long-term collaboration and joint effort among clinicians, biologists, scientists, and patients.

Cancer Sample QC Genomic Data

Generation Analysis & Validation

Agilent 8x60K platform

Agilent 8x60K platform

1. Array CGH

2. Gene Expression

3. Mutation Profile : Cancer panel

STR Genotyping

Human-Mouse ALB Test

DNA, RNA Extraction

Pathology QC

0

20

40

60

80

100

Fail

Intermediate

Accept

Histologic QC

Histologic and Protein Characteristics of AVATAR

Case %

Acceptable DNA quality 198 95.0

No human DNA in xenograft mouse DNA 8 3.9

Mouse DNA contamination in human DNA 2 1.1

Total 208 100

Human Human+Mouse Mouse

Human – Mouse Test

STR Genotyping

TP53

KRAS PIK3CA

MLH1

FBXW7

CTNNB1

HRAS

APC

CDKN2A

STK11

BRAF

ERBB2 NRAS RB1

SMAD4 VHL

Ion Torrent Ampliseq Cancer Panel

Human

PDX

AVATAR mouse Passage 2

Patient

AVATAR mouse Passage 5

AVATAR mouse Passage 2

Patient

AVATAR mouse Passage 5

Xenograft specific

380 g

enes

133 samples aCGH Analysis

Removal of Xenograft-specific genes

breast gastric colon bladder pancreas

ovary lung brain meta

133 samples aCGH Analysis

Candidate target genes in breast cancer are maintained in PDX

Human

PDX

New drug therapeutic efficacy validation and lead optimization

• Whole-genome comparisons

• Structural and copy number aberrations were found to be retained with high fidelity.

• Variable numbers of PDX-specific somatic events were documented, although they were only rarely functionally significant.

• PDX models are an important resource for the search for genome-forward treatment options and capture endocrine-drug resistance etiologies that are not observed in standard cell lines.

Thank you..

Thanks to our patients and Samsung Medical Center Team

Samsung Medical Center Seok Jin Nam Jeong Eon Lee Young-Hyuck Im Yoen Hee Park Yong Bum Cho Jeeyun Lee Samsung Genome Institute Woogyang Park Institute for Refractory Cancer Research, SMC Do-Hyun Nam Kyeung Min Joo

Lab of CGMP, SMC Ensel Oh Yu Jin Kim Ji-Young Song Min Jung Sung Suzie Ahn Seoul National University Young-Kee Shin Ryong Nam Kim Gencurix Sang-Rae Cho