FGFR1: RNA, DNA and protein correlations D. Ross Camidge, MD PhD Director, Thoracic Oncology Clinical Program Associate Director for Clinical Research University of Colorado Cancer Center

TAT March 2014

Ross Camidge, MD, PhD Consulting Fees (e.g., advisory boards): Servier, Eli Lilly, Genentech/Roche, Astex, Ariad, ImmunoGen, Clarient, Excelixis, IndiPharm, Roche Research Grant: Ariad

Understanding the pathway

FGF family • 22 FGFs, of which 18 considered biologically active, 6 subfamilies – Secreted – Heparin sulfate proteoglycan bound (stabilization, protease protection) • Endocrine Factors – FGF19, 21 and 23 – Involved in bile acid, cholesterol, glucose, vit D and phosphate homeostasis – Dependent on klotho proteins (klotho or –klotho) in target tissue

• Paracrine/Autocrine Factors Dientsmann et al, Annals Onc 2013 Kelleher et al, Carcinogenesis 2013

FGFR family • 4 cell surface receptors FGFR1-4 • Ligand binding = – dimerization – transphosphorylation = Activation of FGFR substrate 2 (FRS2), then grb2/SoS, MAPK and PI3K/AKT = FRS2 independent phospholipase C-activation = context dependent activation of other pathways

• Ligand specificity affected by splice variants of FGFR1-3 (may limit autocrine effects in normal physiology) Dientsmann et al, Annals Onc 2013 Kelleher et al, Carcinogenesis 2013

Splicing: ligand specificity of FGFR is controlled by alternative splicing of the third Ig loop (IgIII) in the ligand binding domain (C-terminal half), resulting in a IIIb isoform and a IIIc isoform. Exon 8 produces the IIIb isoform and exon 9 produces the IIIc isoform. Note: FGF 3, FGF 7, FGF 10 and FGF 22 exclusively bind IIIb, whereas FGF 1 binds both and the remaining 13 bind IIIc.

Kelleher et al, Carcinogenesis 2013

Normal physiology (other than endocrine) • Angiogenesis (VEGFR, PDGFR interaction) • Musculoskeletal development

Pfeiffer syndrome

Achondroplasia Kelleher et al, Carcinogenesis 2013

Aberrations in cancer

Turner and Grose, Nat Rev Cancer 2010

FGFR1 primary aberrations in cancer While activating FGFR2 and FGFR3 mutations and translocations have been described in several cancers FGFR1 aberration primarily associated with gene amplification Breast and NSCLC examples… However, also: • FGFR1 fusions in GBM, acute myeloid and lymphoid leukemias and rhabdomyosarcoma • FGFR1 mutations in pilocytic astrocytoma

Dientsmann et al, Annals Onc 2013; Jones et al, Nat Genetic 2013 Singh et al, Science 2012; Ren et al, Leukemia 2013

TCGA: Amplification Of FGFR1 And/Or FGF (11q Amplicon With FGF3/-4/-19) Is Common (11-14%) In Breast Invasive Carcinoma*

Note the frequencies include both amplification and homozygous deletions *per TCGA methodology

TCGA: FGF Aberrations Are Equally Common In ER+ And Her-2+ Subsets – Consequently Most FGF+ BC Patients Are ER+ FGFR1 and/or 11q (FGF3/4/19) amplification according to cell phenotype* 700

Number of patients*

600 500 400

FGF-

300

FGF+

200 100 0

ER+

Her-2+

TNBC

Tumor phenotype

FGFR1 and/or 11q amplification seen in 23% of ER+, 27% of Her-2+ and 7% of TNBC *Captured in TCGA

TCGA: FGFR1 And 11q Amplification Seen Commonly In NSCLC FGFR1 amp freq

FGFR1 and/or 11q amp freq

n

Lung squamous cell carcinoma

16.9%

34.3%

178

Lung adenocarcinoma *

3.5%

14.3%

230

Tumor type

*Provisional TCGA data. Amplification determined by Genomic Identification of Significant Targets in Cancer analysis of array CGH and SNP array data in TCGA studies

14

Sq NSCLC: Weiss et al • 155 Sq NSCLC. SNP array analysis. GISTIC algorithm. • 9.7% (15/155) 133Kb region 8p12 (at least 4 copies = amplification). Region includes FGFR1. • 1% (6/581) 8p12 amp non-squamous NSCLC • 8p12 probe FISH 22% (34/153) Sq NSCLC (at least 9 copies = amplification)

Weiss et al. Sci Transl Med. 2010 Dec 15;2(62):62ra93.

Sq NSCLC: Dutt et al • 628 NSCLC tumors. 104 NSCLC cell lines SNP array analysis. • log2 ratio >0.7 or 3.25 normalized DNA copies 8p11-12 region • GISTIC algorithm identified 170Kb region 8p11. Region includes FGFR1. • 21% (12/57) of squamous cell carcinomas • 3.4% (20/588) of adenocarcinomas

Dutt et al. PLoS One 2011

Incidence of FGFR1 gene amplification in Chinese NSCLC patient samples NSCLC tumor histology

FGFR1 amplification incidence [no. of samples (%)]

All (n = 127)

SCC (n = 48)

AC (n = 76)

AC/SCC (n = 3)

11 (8.7)

6 (12.5)

5 (7.0)

0 (0)

127 resection NSCLC FISH analysis. FGFR1 amplification (FISH 6) was defined as an FGFR1/CEP8 gene probe ratio of ≥2 or cluster signals in ≥10% of tumor cells.

Zhang J et al. Clin Cancer Res 2012;18:6658-6667

Differing definitions of ‘amplification’ in Genomic and CISH or FISH analyses

Ratio to chromosome 8 centromere, total copy number (includes high polysomy), cytogenetic appearance (‘clusters’) – TCGA – e.g. ‘2 standard deviations from random expectation’ Dientsmann et al, Annals Onc 2013

copy number is a continuous variable – and cutpoint depends on correlate looked for – not just thresholding as for mutations or gene rearrangements (cf TCGA)

qRT-PCR assay: exon 15 and 19 of FGFR1 relative to reference genes 347 NSCLC resection specimens correlation with OS (ten portions) CNV >3.5 = optimal Separation

Gadgeel et al, PLoS One 2013

262 resected squamous NSCLC: Following Weiss et al. Amp = at least 9 copies FGFR1

Kim et al, JCO 2013

Resection based series: • OS/DFS correlation could reflect: – Potential for hidden metastatic disease – General therapeutic responsiveness at relapse – Comorbidities associated with risk factors for amp

• Does not equal predictive correlation for specific FGFR1-directed inhibition

83 lung cancer cell lines Cellular ATP assay IC504μmol/L)

©2012 by American Association for Cancer Research

Zhang J et al. Clin Cancer Res 2012;18:6658-6667

FGFR1 amplification in breast, lung, and osteosarcoma cancer cells is associated with response to NVP-BGJ398.

• FGFR1 copy number gain defined as log2 ratio ≥1 (equal to ≥4 normalized DNA copies) • IC50 less than 500 nmol/L were classified as sensitive • 2/7 lung ca AMP sensitive (28%) • 2/5 breast ca AMP sensitive (40%) • 1/1 osteosarcoma AMP sensitive (100%)

©2012 by American Association for Cancer Research

Guagnano V et al. Cancer Discovery 2012;2:1118-1133

‘Sensitive’ does not equal ORR. ALK example Objective response details (all evaluable patients)

% Decrease or increase from baseline

100

N=116

80

ORR (95% CI)

60

Median response duration

48 weeks

Median time to response

8 weeks

40

61% (52, 70)

Disease control rate at 8, 16 weeks

79%, 67%

20 0 –20 –40 –60 –80

Progressive disease Stable disease Partial response

–100

Complete response

Camidge et al, ASCO 2011

Why? Protein vs gene as driver: Sq NSCLC PDTX?

4/5 AMP PDTX models sensitive

Zhang et al, CCR 2012

Copy number associates with driver state but is it the driver?

58 lung cancer cell lines

Wynes et al, submitted

Message vs gene?

Cutpoints from cell line TMA and clinically achievable IC50 for FGFR inhibitor Wynes et al, submitted

Ongoing Ponatinib IITs in FGFR NSCLC A Phase II Study of Ponatinib in Cohorts of Patients With Lung Cancer Preselected Using Different Candidate Predictive Biomarkers (RET and FGFR) Clinical Trial Schema:

Acquire tumor block/slides – confirm histology, confirm not EGFR Mt or ALK+ (if adenocarcinoma), confirm adequate for molecular testing Prospectively screen FGFR1 mRNA ISH and FGFR1 copy number SISH all patients

NB FGFR cohorts All histologies

FGFR1 ISH+ve, SISH+ve

FGFR1 ISH+ve, SISH-ve

FGFR1 ISH-ve, SISH+ve

FGFR1 ISH-ve, SISH–ve ‘double negative’

• Trial uses initial FGFR1 entry cutpoints set by cell line TMA, IC50 and predicted achievable Cmin with ponatinib at 45mg • Iterative design, adding new cohorts with modified entry cutpoints as data emerge

FGFR1 gene amplification more closely associated with high mRNA expression in breast cancer, as compared to sqNSCLC 825 invasive breast carcinomas

178 lung squamous carcinomas

The degree of correlation between FGFR1 amplification and FGFR1 mRNA expression varies across disease states • •

Breast cancer: median z-score of 1.77 vs -0.06 (p-value: 1.3-10) sqNSCLC: median z-score of 0.58 vs -0.32 (p-value: 0.023) Raw data for TCGA lung invasive breast and lung squamous carcinomas published in Nature 2012. TCGA lung squamous study used RNA-Seq for expression analysis

Necessary but not sufficient?

Cell line sensitivity correlates with FGFR1 and c-myc mRNA levels

3T3 transfection/co-transfection

C-Myc IHC +++ 40% Sq NSCLC Independent of FGFR1 amp status Malchers et al, Cancer Discovery 2013

MYC mRNA is not associated with ponatinib sensitivity in CU cell line panel (p=0.98)

Ponatinib IC50 values from Wynes et al, Clin Cancer Res, 2014 MYC mRNA levels from the CCLE

Ligand binding is essential for growth of FGFR1 amp xenografts.

(using ectopic FGFR1 ligand trap)

©2014 by American Association for Cancer Research

Malchers F et al. Cancer Discovery 2014;4:246-257

Wynes et al, submitted

FGFR1 summary 1 • FGFR1 activation mostly associated with FGFR1 amplification • (Rare FGFR1 mutations and fusions exist) • Definition of ‘amplification’ depends on correlated endpoint: – Presence above background – OS or DFS in resection series – Prediction of inhibitor sensitivity in preclinical models (for clinical use) • definition of ‘sensitive’ IC50 • X% of sensitive cell lines are amplified vs Y% of amplified cell lines are sensitive

FGFR1 summary 2 • FGFR1 amp predicts for preclinical ‘sensitivity’ as low as 28% in Sq NSCLC and 40% breast Ca. • Even for accurate predictive biomarker preclinical sensitivity rate will be higher than ORR (due to IC50 and RECIST details) • However, FGFR1 gene amplification does not guarantee message or protein over-expression. And mRNA/protein can be independent of amp. • Correlation of amp and protein/mRNA varies by tumor type or histology within tumor type (eg Sq vs non-Sq NSCLC) • Co-factors (to be explored clinically) may influence inhibitor sensitivity – FGFR1 ligands? – Myc expression?

Questions?