Joe W. Gray, Ph.D. Lawrence Berkeley National Laboratory University of California, San Francisco A systems approach to marker guided therapy in breast

Joe W. Gray, Ph.D.Joe W. Gray, Ph.D.Lawrence Berkeley National LaboratoryLawrence Berkeley National LaboratoryUniversity of California, San FranciscoUniversity of California, San Francisco

A systems approach to marker guided therapyin breast cancer


Breast cancer overview and statement of the problem

An in vitro systems approach to match treatment to “ome”

Improving and testing the model





Stage Distribution and 5-year Relative Survival by Stage at Diagnosis for 1999-2006, All Races, Females

Stage at Diagnosis Stage Distribution (%)

5-year Relative Survival (%)

Localized (confined to primary site) 60 98.0Regional (spread to regional lymphnodes)

33 83.6

Distant (cancer has metastasized) 5 23.4

Unknown (unstaged) 2 57.9

SEER RegistryWorld wide incidence - 1,150,000/yrWorldwide mortality - 410,000/yr

Improve treatment by identifying molecular subtype markers that

•predict resistance to existing therapies•predict response to experimental therapies

Overall goal

Hundreds of compounds are approved or well along in the developmental pipeline

How do we find the most effective for breast cancer?

International cancer genomics efforts are substantially increasing the number of recognizable cancer subtypes that

may respond differentially to specific therapies

State of the breast cancer genome

• Remarkable genomic and epigenomic heterogeneity between and within tumors

• Hundreds of genes and gene networks are deregulated in ways that contribute to cancer pathophysiology

• Subtypes are defined by aberrations at multiple levels:mutation, structure, copy number, chromatin modification, ncRNA, …

• Subtypes defined by recurrent aberrations are associated with outcome

• Response varies with subtype













How do we make the optimal match between drug and subtype?

Associations

???





We use a collection of 50+ breast cancer cell lines to model the molecular diversity of primary tumors

• Therapeutic approaches can be tested quickly to identify subtype specific responses

• Model can be characterized at great molecular depth to identify predictive markers

• Model can be manipulated to test predictions

To what extent do the cell lines represent what we know about breast cancer?

Cell lines model gene expression subtypes, recurrent copy number chances and mutations

We have assessed ~100 therapeutic strategies in 50 cell lines

Emphasis on signaling pathways

Establishing associations between response and molecular subtypes

UCSC Cancer Genome Browser

Molecular features Biological features

A

GI 5

0

GI 5

0

1e-05

1e-06

1e-07

1e-04

1e-05

1e-06

1e-05

1e-06

1e-07

1e-08

Re

lati

ve

ce

ll n

um

be

r a

t 3

d

ay

s

Log drug concentration

Cell line

Luminal

BasalClaudin-lowNormal

Ambiguous

CI1040 Lapatinib

AKTi

Approximately half of compounds tested show significant molecular subtype specificity

GI 5

0

Kuo, Guan, Hu, Bayani 2007Cell line

Associations

Associations

We are especially interesting in identifying genomic drivers for molecular response

We are especially interesting in identifying genomic drivers for molecular response

Most effective targeted agents are linked to genomic markers that predict response

*Except VEGFR and proteosome inhibitors

Imatinib mesylate CML BCR-ABL translocation Oncogene addiction (1982)

Imatinib mesylateSunitinibNilotinibDasatinib

GISTDermatofibrosarcoma

protuberansHypereosinophylic

syndromeMelanoma

c-KIT mutationPDFGR mutation

Oncogene addiction (1999)

TrastuzumabPertuzumabLapatinib

Breast HER2 amplification


Gefitinib, ErlotanibCetuxumab

Lung cancerBowel

EGFR mutation


PKC412, SU11248, CMT53518

AML, ALL FLT-3 mutation, tandem duplication


PARP inhibitors Breast Ovarian BRCA1/2 mutation Synthetic lethality (2005)

PLX4032 Melanoma BRAF mutation Oncogene addiction (2002)

Crizotinib Lung EML-4 ALK translocation Oncogene addiction (2007)

Tamoxifen, AIs Breast cancer ER expression Lineage (1800s)

~25% of compounds are significantly associated with genome copy number abnormalities

Spellman, Sadanandam, Kuo

Platinum, anti-metabolites and anti-mitotic apparatus protein inhibitors

effective in basal subtype cells

PI3K inhibitorPI3K inhibitorPI3K inhibitor

AURK inhibitor

PLK1 inhibitor

Kuo, Spellman, Sadanandam

LuminalBasalClaudin-low

Sensitive Resistant

Response to mitotic apparatus inhibitors is associated with transcriptional upregulation of a network of mitotic apparatus

genes

Mao, Hu et al

Why does this network exist?

Expression of mitotic apparatus genes is associated with amplification of transcription factors that target

mitotic apparatus genes

Christina Clark, Carlos Caldas

MYCZEB1

FOXM1

SOX9

All genes in the mitotic apparatus signature are targeted by these transcription factors

Mao, Curtis, 2010

Hierarchical clustering of 31 significant subtype specific drugs and BrCa cell lines.

PI3K inhibitorPI3K inhibitorPI3K inhibitor

AURK inhibitor

PLK1 inhibitor

Kuo, Spellman, SadanandamEGFR, ERBB2, PI3K

inhibitors, HDACs effective in luminal subtype cells

LuminalBasalClaudin-low

Luminal subtype preference for ERBB2 and AKT pathway inhibitors “explained” by the subtype specificity of activating genomic aberrations

GI 5

0

Lapatinib

GI 5

0

X

PIK3CA mutationPTEN mutation

ERBB2 Amplif icationPIK3CA Amplif ication

PTEN Amplif icationERBB2 Protein Expression

PIK3CA mutationPTEN mutation

ERBB2 Amplif icationPIK3CA Amplif ication

PTEN Amplif icationERBB2 Protein Expression

PIK3CA

PTEN

AKTi

Aberrations interact - AKT inhibitors synergize with lapatinib in ERBB2+, PIK3CAmt cells

Anta

goni

stic?

Syne

rgisti

cPI

K3CA

mut

ation

(blu

e is

mut

ant)

Drug combination dose

H1047R PIK3CA

H1047R PIK3CAE545K PIK3CA

E545K PIK3CAK111N PIK3CA

H1047R PIK3CA, E307K PTEN

K267fs*9 PTEN

Korkola, Cooper, et al 2010





Complicating factors

• Microenvironment• Response not durable• Response heterogeneity

The microenvironment modulates response to ERBB2 targeted drugs

2D monolayer

3D matrigel

AU565 ERBB2 amp SKBR3 ERBB2 amp

HCC1569 ERBB2 amp BT549 ERBB2 norm

Wiegelt, et al., Breast Cancer Res Treat 122:35–43, 2010

The microenvironment modulates the signaling network

HER3HER2

PI3K

PDK1

Akt

mTorC1

S6K1

S6

TSC2

Rheb

MEK

MAPK

mTorC2

PRAS40

RAF

PKCα

nucleus

COX2, CREB, cJun, NFkB, ATF2, ER, Tcf/Lef, Rb, AP1, cFos,

CXCR4, ETS, HIF1a,MYC -> CBX5

HER3, PDK1, Akt, …

cytosol-integrin

microenvironment

IRS1

Inhibition of microenv. signaling also should modulate response

Inhibition of 1-integrin signaling enhances response to ERBB2 targeted drugs in 3D but not 2D

Wiegelt, et al., Breast Cancer Res Treat 122:35–43, 2010

AIIB2

None

AU565 ERBB2 amp SKBR3 ERBB2 amp

HCC1569 ERBB2 amp BT549 ERBB2 norm

Microenvironment dependent response may explain why treatment of metastatic disease is difficult

Can we identify microenvironment independent therapies?

This motivates assessment of pathway function in situ

Britt Marie Ljung

TOF-SIMS “ome” imaging

Total Area Spectrum

Primary Ion Beam

m/z

256

256

Tag 2 map

Tag 1 Map

Sample

Immunohistochemistry or in situ hybridization with mass tag labeled reagents. Each

tag is a color.

More complications ERBB2 inhibition is not durable

Amin et al, Science TM 2010; 2: 16ra7.

HER3HER2

PI3K

PDK1

Akt

mTorC1

S6K1

S6

TSC2

Rheb

MEK

MAPK

mTorC2

PRAS40

RAF

PKCα

nucleus

COX2, CREB, cJun, NFkB, ATF2, ER, Tcf/Lef, Rb, AP1, cFos,

CXCR4, ETS, HIF1a,MYC -> CBX5

HER3, PDK1, Akt, …

cytosol-integrin

microenvironment

IRS1

Mills, Moasser et al

Understanding response dynamics

Statistical and dynamic modeling to understand long term behavior

– ODE model for short term effects (Soulaiman Itani)

– A hybrid Boolean-ODE model using to model longer term effects[Chen 2009] (Young-Hwan Chang)

Tomlin lab

Need to understand the emergent properties of complex, cross coupled systems

Need to understand the emergent properties of complex, cross coupled systems

Network behavior is context dependentNetwork behavior is context dependent

Signaling occurs in 3 dimensionsSignaling occurs in 3 dimensions

Center for Cancer Systems Biology

Digital v. analogue drug responses

Molecular responses are heterogeneous – a partial explanation for lack of durability?

Sorger et al


TCGA/ICGC projects are defining a growing number of distinct subtypes

In vitro systems suggest at least half of all therapeutic compounds show subtype specificity

Improving the model - Modeling the microenvironment, heterogeneity and long term

durability

Cell line system biologyWen-Lin KuoJim KorkolaNick WangNora BayaniBrian CooperMara JeffressAnna LapukDemetris IacovidesMina BissellMartha StampferTerry Speed (UCB)Claire Tomlin (UCB)Michael Korn (UCSF)Frank McCormick (UCSF)Gordon Mills (MDACC)Yiling Lu (MDACC)Peter Sorger (Harvard)

Collaborators

Mitotic apparatus networksZhi HuJian Hua MaoShenda GuBarbara Weber (GSK – then)Richard Wooster (GSK)Christina Clark (Cambridge)Carlos Caldas (Cambridge)

Genome biologyPaul SpellmanAnguraj SadanandamLaura HeiserShannon DortonJing HuangSteffen DurinckObi GriffithLakshmi JakkulaFrancois PepinAndy WyrobekDavid Haussler (UCSC)Josh Stuart (UCSC)

Project managementHeidi FeilerShradda Ravani

Clinical science (I SPY etc)Laura Esserman (UCSF)Laura Van’t Veer (UCSF)Rick Baehner (UCSF)Nola Hylton (UCSF)John Park (UCSF)Hope Rugo (UCSF)Britt Marie Ljung (UCSF)Hubert Stoppler (UCSF)Fred Waldman (UCSF)

NCI Center for Cancer Systems Biology, The Cancer Genome Atlas, CPTAC, Bay Area Breast Cancer SPORE, Atwater foundation, GSK, Roche, Millenium, Pfizer, Progen, Cytokinetics, Cell Biosciences, DOD Innovator, SU2C

Documents

Joe W. Gray, Ph.D. Lawrence Berkeley National Laboratory University of California, San Francisco A systems approach to marker guided therapy in breast