using tcga data to inform on precision medicine in late ... · pdf fileusing tcga data to...

$: Using TCGA Data to Inform on Precision Medicine in Late ... · PDF fileUsing TCGA data to inform on precision medicine in late-stage ... of the pre-auricular tumor: \ ... Data to Inform$
Using TCGA data to inform on precision medicine in late-stage cancer settings

Andrew J. Mungall, Ph.D. British Columbia Cancer Agency, Genome Sciences Centre [email protected]

3rd TCGA Symposium, Natcher Conference Center, NIH Campus, Bethesda, MD 12th May 2014

Presenter

Presentation Notes

Thank the organizers and abstract selection cmte. I will move rather quickly through my slides but I’ll be happy to talk at my poster #10

Personalized Oncogenomics

• BC Cancer Agency – Cancer Care and Research • Provincial population-based cancer control program

– Prevention, Screening, Diagnosis and Treatment

• Scope: – POG aims to bridge divide between genomics research and

clinical practice – identify tumour-specific therapeutic targets in cancer

patients with late stage disease

2

Consult & Consent

Specimens: • Tumour Biopsy • Archival tumour • Peripheral blood

Libraries: • PCR-free genomes (T/N) • FFPE genome • Strand-specific RNA-seq

Illumina Sequencing

and Data Analysis

Therapeutic recommendation

to Clinician

Presenter

Presentation Notes

The BCCA has a mandate to provide cancer care and research to the population of BC. 4.5M people living in an area twice the size of California. The Personalized Oncogenomics (POG) project is a collaboration between the GSC and clinicians aiming to identify tumour-specific therapy. Following consultation with the patient a biopsy is scheduled to obtain tumour tissue and peripheral blood. The diagnostic archival tissue block is also requested. Tumour and normal genome libraries and tumour transcriptome libraries are constructed for Illumina HiSeq sequencing. Following data analysis a report is provided to the referring clinician and discussed at a weekly meeting.

Gene coverage + QC

Assembly

Trans-Abyss

Assembly

Trans-Abyss

Paired calling

Concatenate fastqs

Tumour-Normal Fold change

Genome validator

CNV/LOH

Comparison to expression database

Pathway Analysis

Mining of drug databases and literature

Merge bams

Preliminary Report

Clonal Heterogeneity

Alignment + Symlinking

Probing for known events

Split bams

Tumour + Normal Genome Archival Genome Tumour Transcriptome

Microbial detection

Variant calling

Repositioning

Filtering Post-processing and

tumour content adjustment

Final Report

Using TCGA data

3

Presenter

Presentation Notes

Emphasize the structured complexity and that TCGA data is used at multiple timepoints e.g. in populating variant databases and as a reference for gene expression levels.

Patients enrolled and data generated

0

5

10

15

20

25

30

Num

ber o

f pat

ient

s

Patients enrolled since July 2012 (pediatric) 83 (8)

Biopsies performed 69

Tumour types 28

Metastasis tumour genome coverage 93x

Matched normal genome coverage 46x

Archival tumour genome coverage 46x

Average tumour RNA-seq reads 306M

Analysis reported (in progress) 50 (19)

Average time from biopsy to report 38 days

4

Presenter

Presentation Notes

To date 83 patients have been enrolled into the POG project and 69 biopsies performed. The bar chart depicts the frequency of the 28 different primary tumour types. On average we have sequenced fresh frozen tumour genomes to 93x coverage and 46x coverage for the normal and archival coverage. Transcriptomes are sequenced at 1 or two per HiSeq lane providing 306M reads on average. Reports are out for 50 of the cases with a TAT of 38 days. If asked, BRCA numbers may appear disproportionately high for two main reasons: We have 3 highly motivated breast cancer clinicians It is technically easier to obtain biopsies from breast than other sites.

How POG guides treatment decision making

1) Providing directed cytotoxic chemotherapy choices & targeted therapeutic options

2) Complemented/corrected clinical tests 3) Changed diagnosis 4) Identifying primary tumour sites when

previously unknown

5

Presenter

Presentation Notes

Areas in which POG has helped to guide treatment decision making include: The provision of directed cytotoxic chemotherapy choices and/or targeted therapeutic options The complementation or correction of approved clinical tests The change of a cancer diagnosis And the identification of primary tumour when previously ambiguous Time does not permit to provide examples for each of these areas but I will present case studies demonstrating the first two uses. Therapeutic utility (e.g. POG03) Limitations of existing clinical assays (e.g. POG30) Altered diagnoses (e.g. POG05) Tumour evolution (e.g. POG03)

1) Case POG 003 – SCC Provided targeted therapeutic options

6

2007 2008 2009 2010 2011 2012 2013 2014

“red rash” on chest Developed bleeding ulcerations Diagnosed with Squamous cell carcinoma

Three lines of chemotherapy and multiple rounds of radiotherapy started in 2007

• Jan 2012 – new preauricular node • Jun – Aug 2012 node and chest mass growing (pain and hearing loss) • Sep 2012, preauricular and chest lesions biopsied for POG

Presenter

Presentation Notes

The first of the case studies is a male patient presenting with bleeding ulcerations following a history of “red rash” on his chest for a few years but otherwise felt well. In 2007 he was diagnosed with Squamous Cell Carcinoma of the skin and multiple rounds of chemotherapy and radiotherapy treatment began. In Jan 2012 a new lesion near his right ear (preauricular node) developed and as the tumour grew hearing loss ensued. In Sep 2012, both preauricular and chest lesions were biopsied for POG.

POG 003: Squamous cell carcinoma of skin

7

2224 (134)

1848 genes

4075 (222)

3153 genes

28 (0)

18 (13) 17

genes Chest

Preauricular node

Preauricular node

Single nucleotide mutations (truncating mutations)

Small indels (frameshift mutations)

Chest

22 (19) 22

genes

548 genes in common

Presenter

Presentation Notes

Sequence analysis revealed very few of the thousands of somatic mutations in common between the two tumours Note different experimental design: which was primary is unclear, so sequence both biopsies

POG 003 – Copy number variants

Gain Loss

Chest Preauricular node

8

Chest Preauricular node chr16 chr16

• Samples almost completely distinct • Very few common breakpoints

Presenter

Presentation Notes

Equally, copy number analysis revealed that the tumour samples were almost completely distinct, sharing very few common breakpoints as shown here for chr16.

Pathway summary

9

SHC1

GRB2

SOS1

TGFBR2

SMAD2

SMAD3

SMAD4

CDKN2A CDKN1C

CCND1

CDK4/6

CCNA1

CDK2

cell proliferation

overexpressed compared to normal skin & compendium

X activating mutation

X mutation copy number gain copy number loss

TP53

activation interaction inhibition interaction indirect activation

X inactivating mutation

GAB1

PI3K

AKT1

mTOR BAD

p70S6Kb BCL2

evading apoptosis

protein synthesis

MMPs

MMP1 MMP9

Chest tumour

Laminins LAMC2 LAMB3

PTEN

TGFB1

TGFBR3 EGFR PDGFR

TGFA

ATM

SRC

FAK

endothelial cell mitogenesis migration

angiogenesis

Integrins ITGA3 ITGA6

TYMP o

VEGFs MMPs IL8

BCR

CDC42 RAC3

PAK7

AXIN1

microtubule stabilization

& organization

PDGFA

X

cell cycle progression

CHK2 CHK1 CRKL

RAS

CDKN2C

Double strand DNA repair

PARP1

Single strand DNA repair

GPCR signaling

X

NF1 X

X X X

X X X

X

X

X ATR

FANCM

cell cycle checkpoints

X X X

DNA repair defects, PTEN loss, AKT1 gain

Under-expressed compared to normal skin & compendium

Presenter

Presentation Notes

Integrating genome and transcriptome data in pathway analysis highlights multiple pathway perturbations consistent with advanced stage disease and DNA repair defects. Including under-expression of PTEN and increased AKT1 expression. showed different drivers in these two tumours Based on the analysis of the in the chest tumor: PTEN homozygous loss and AKT gain and overexpression suggesting everolimus PARP1 is a protein that is important for repairing single-strand breaks ('nicks' in the DNA). If such nicks persist unrepaired until DNA is replicated (which must precede cell division), then the replication itself can cause double strand breaks to form.[7] http://www.nature.com.ezproxy.library.ubc.ca/nrd/journal/v10/n1/pdf/nrd3320.pdf GPCR pathway

SHC1

GRB2

SOS1

TGFBR2

SMAD2

SMAD3

SMAD4

CDKN2A CDKN1C

CCND1

CDK4/6

CCNA1

CDK2

cell proliferation

TP53

GAB1

PI3K

AKT1

mTOR BAD

p70S6Kb BCL2

evading apoptosis

protein synthesis

MMPs

MMP1 MMP9

Pre-auricular tumour

Laminins LAMC2 LAMB3

PTEN

TGFB1

TGFBR3 EGFR PDGFR

TGFA

ATM

SRC

FAK

endothelial cell mitogenesis migration

angiogenesis

Integrins ITGA3 ITGA6

TYMP o

VEGFs MMPs IL8

BCR

CDC42 RAC3

PAK7

AXIN1

microtubule stabilization

& organization

PDGFA

cell cycle progression

ATR

CHK2 CHK1 CRKL

RAS

CDKN2C

Double strand DNA repair

PARP1

Single strand DNA repair

GPCR signaling

X

X

NF1 X

X

X

X X

FANCM

cell cycle checkpoints

X X

DNA repair defects, EGFR amplicon, MMPs Case 3

Pathway summary

10

Under-expressed compared to normal skin & compendium

overexpressed compared to normal skin & compendium activating mutation

mutation copy number gain copy number loss

inactivating mutation X

X X

activation interaction inhibition interaction indirect activation

Presenter

Presentation Notes

Based on the analysis of the pre-auricular tumor: high-level amplification and over-expression of EGFR suggesting erlotinib

Therapeutic options, treatment and response

• Treatment: – High-level amplification and over-expression of EGFR

in the preauricular tumour suggested erlotinib – PTEN homozygous loss and AKT gain and over-

expression in the chest lesion suggested everolimus

• Response: – Dramatic reduction in the size and extent of his

tumours – Hearing returned to his right ear – Dramatically reduced use of pain medications – After a few months, the pre-auricular tumour

progressed, so we re-biopsied and sequenced this tumour

Dec 6, 2012

Jan 15, 2013 11

Presenter

Presentation Notes

December 2012 initiated erlotinib - within 3 weeks the preauricular tumour had completely involuted. Dramatically reduced use of pain medications (from 8 pills/day to 1 pill/night) However, the chest lesion did not respond. After 3 weeks everolimus was added to the treatment regimen and the chest mass flattened considerably. Sustained response for 9 months. Progression of the preauricular tumour has been re-biopsied.

Progressed preauricular

tumour • Further EGFR

amplification and over-expression observed – Higher copy number (55)

than in the 2012 biopsy (32)

– Highest expression level in all TCGA samples

12

Presenter

Presentation Notes

After 9 months, the pre-auricular tumor progressed, so we re-biopsied and sequenced this tumour Further EGFR amplification and overexpression observed Higher copy number (55) than in the 2012 biopsy (32) And highest expression level observed in all TCGA data EGFR gene expression relative to TCGA tumour (yellow) and normal (green) datasets using RSEM

2) Case POG 030 - NSCLC Complemented/corrected clinical tests

• 68 yo male lifelong never smoker diagnosed with non-small-cell lung adenocarcinoma

Jan Feb 2013

Mar Apr May Jun Jul Aug

Confirmed Non-small cell lung adenocarcinoma, metastasis in lymph node

FNA of right supraclavicular node - EGFR and ALK tests negative

Radiation and three lines of chemo between March and July 2013 but rapid progression

Biopsied same node for POG

13

• Transcriptome and genome sequencing revealed chr2 inversion fusing EML4-ALK genes

• Sequence analysis at chr2 breakpoints identified a further inversion and insertion into chr12 that appears to prevent Vysis dual-colour break-apart probe from hybridizing

EML4 exons 1-13 ALK exons 20-29

coiled-coil domain - dimerization, - kinase activation

tyrosine kinase domain

EML4-ALK fusion found in this patient

14

inversion

inversion & insertion into

chr12

chr2

chr12

FISH probes

Presenter

Presentation Notes

ALK rearrangement was scored negative by the original clinical FISH probe test – only 3% cells positive

POG30

POG21 POG9

POG25

POG24

POG30 POG9

POG25 POG21 POG24

ALK ROS1

TCGA lung adeno (126 samples)

TCGA lung adeno (126 samples)

99th centile

94th centile

ALK and ROS1 are highly expressed in POG 030 (compared with TCGA lung adenocarcinoma and other POG lung cases)

15

Presenter

Presentation Notes

EML4-ALK fusion, with high overall expression of ALK �(FC=4.1, 99th centile vs TCGA) ROS1 overexpression �(FC=5.9, 94th centile vs TCGA)

Response to ALK inhibition (Crizotinib)

• EML4-ALK fusion, with high overall expression of ALK together with ROS1 over-expression

• TKI Crizotinib was immediately administered

• The tumour responded dramatically

16

Sept 4 2013 – before Crizotinib Crizotinib started Sept 25 Scan from Dec 12, 2013

Presenter

Presentation Notes

Response to ALK Inhibition. A: Computed tomography (CT) scan of the liver metastasis at baseline on May 22, 2013, prior to the completion of the genomic study and administration of crizotinib. B: CT scan image of the same metastatic mass obtained on December 12, 2013 after treatment with crizotinib.

17

• For each patient, sequenced ≥3 genomes (normal, archival, tumour) and 1 transcriptome (tumour)

• 82 consented patients with advanced cancer - 74 biopsies attempted

- 3 biopsies failed, 2 patients withdrew consent

- Full data available for 50 patients - 9 in progress, 6 on hold, 4 patients died during analysis phase

- Clinically evaluated in 38 cases

• POG informative or actionable for treatment: 33/38 (87%) • Treatment available and offered: 18/33 (55%)

– ≥ 6 (18%) patients died during or shortly after analysis, precluding treatment

Evaluation of POG Results

Presenter

Presentation Notes

When the genomic data was felt to aid in diagnosis and/or identify treatment options it was labeled as "informative or actionable". This was determined for each patient's data by the clinician(s) looking after that specific patient. Limits are:�� 1. the statement of informative and/or actionable is subjective and is a judgment made by the clinician(s) that are looking after the patient. Other clinicians might draw different conclusions.� 2. there were some patients that we have done - 9 of them - that I wasn't able to track down the clinicians to ask their opinions on informative or actionable so I haven't counted these patients - obviously I need to track them down and fill this data in.� 3. there are some patients whose data was cancelled (they died in the middle) or were put on hold - I didn't count them� 4. there was one patient who we sequenced who didn't have enough tissue - but I counted that as a non-informative as opposed to omitting it� 5. the data on what was actually acted on in part looks low because there are many reasons it wasn't acted on - some patients died or were too ill, other patients didn't need treatment yet, some patients didn't have treatments available to them�

Personalized Oncogenomics - Phase II • REB approval for 5,000 cases (in 5yrs) • Move from ~1 pt / week to >1 pt / day • Emphasis on genome + transcriptome sequencing • Include the “oncopanel” for a rapid TAT “first look” • Emphasis will expand beyond end stage patients • Increase speed and accuracy of sequence analysis &

report generation • Verifying “actionable” results in clinical lab. prior to

treatment

18

Acknowledgements Janessa Laskin and Marco Marra

Co-Project Leaders Care Research

BCCA Francois Benard Kim Chi Stephan Chia Karen Gelmon Cheryl Ho Farzad Jamshidi Aly Karsan Christian

Kollmannsberger Howie Lim Caroline Lohrisch Martin Monty Torsten Nielsen Dan Renouf Tamara Shenkier Sophie Sun Anna Tinker Youwen Zhou

Yussanne Ma Richard Ma Simon Haile Merhu Helen McDonald Michelle Moksa Richard Moore Andy Mungall Karen Mungall Pawan Pandoh Erin Pleasance Robyn Roscoe Jacquie Schein Yaoqing Shen Young Song Nina Thiessen

VGH David Schaeffer Tony Ng

CTAG David Huntsman Julie Ho Julie Lorette Amy Lum Sarah Padilla Peggy Tsang Stephen Yip

Molecular Oncology

Sam Aparicio Sohrab Shah Peter Eirew

BCCH Rod Rassekh Rebecca Deyell Anna Lee Coleen Jantzen Coleen

Fitzgerald

GSC Steven Jones Jianghong An Carolyn Ch’ng Robin Coope Richard Corbett Katty Cruz Nisa Dar Alexandra Fok Jasleen Grewal An He Katayoon Kasaian Heather Kirk Sreeja Leelakumari Yvonne Li William Long

SFU Peter Chow-White

We gratefully acknowledge the participation of our patients and families and the generous support of the BC Cancer Foundation.

Tina Wong Natasja Wye Yongjun Zhao Kelsey Zhu

19

Poster #10

THANK YOU!

using tcga data to inform on precision medicine in late ... · pdf fileusing tcga data to...

Documents