inhibition in ovarian high grade serous ovarian carcinoma Iain - - PowerPoint PPT Presentation

HRD as a predictive biomarker for response to PARP inhibition in ovarian high grade serous ovarian carcinoma

Iain McNeish Professor of Gynaecological Oncology Wolfson Wohl Cancer Research Centre Institute of Cancer Sciences University of Glasgow, UK

Broad assumptions

PARP inhibitor sensitivity = f(HRD) PARP inhibitors function by blocking Base Excision Repair

Potential predictive biomarker assays to identify HRD

• Functional
• Phenotypic
• Genomic

Different tests give different results

Somatic and germline BRCA1 and 2 mutations c.20%



Platinum response 60–65%

50 100 150 200 250 10 100 1000 10000 Time (Days) CA125

TCGA Analysis 51%

TCGA, Nature (2011) 474:609

RAD51 assay 50%

Mukhopadhyay, Cancer Res (2012) 72:5675

Potential predictive biomarker assays to identify HRD

• Functional
• Phenotypic
• Genomic

High grade serous ovarian cancer

Platinum Sensitive Platinum resistant = HR defective

Platinum response

Assumption: platinum sensitivity = f(HRD) = PARPi sensitivity

Ledermann et al. Lancet Oncol (2014) 15:852

Response to platinum in relapse setting

Gelmon et al Lancet Oncol. (2011) 12:852-61

Platinum response as predictor of single agent response

Potential predictive biomarker assays to identify HRD

• Functional
• Phenotypic
• Genomic

How can genomics identify the other 30%?

Somatic and germline BRCA1 and 2 mutations c.20% TCGA Analysis 51%

TCGA, Nature (2011) 474:609

RAD51 assay 50%

Mukhopadhyay, Cancer Res (2012) 72:5675

Candidate panel gene sequencing – limited utility

Non-BRCA HR-pathway gene mutations are rare

0% 2% 4% 6% 8% 10% 12% 14% Data source: TCGA HGOC

Frequency of Tumours Cumulative frequency of non-BRCA HR gene mutations – 13%

• Determined frequencies of mutations in 28

HR genes in ≈250 women

• 16 genes with low frequency mutations

and 12 genes with no mutation found

• siRNA knockdown of 28 HR genes in

3 ovarian cancer cell lines (shown are siRNA of 10 genes in OVCAR-3)

Differential sensitivity to rucaparib (siRNA knockdown)

Controls High sensitivity Intermediate sensitivity Low sensitivity 0.2 0.4 0.6 0.8 1 1.2

IC50 Fold Change vs NT4

IC50=half maximal inhibitory concentration.

Defective HR causes widespread genomic damage

Abkevich et al, Br J Cancer (2012) 107:1776

Comparative genomic hybridization (CGH) analysis from targeted deep NGS of cancer specimens

Log2 ratio Allele frequency >3,500 genome-wide SNPs sequenced to support copy number analysis

CN=2 CN=3 CN=1 0.5 0.4 0.3

Concomitant change

BRCAmut LOH-high

Chromosome No.

LOH-low

Hypothesis 1: Ovarian cancer patients with high genomic LOH suggesting BRCA-like signature will respond to rucaparib. Hypothesis 2: Ovarian cancer patients who are “Biomarker Negative” (ie, with low genomic LOH) will not respond to rucaparib.

NGS=next-generation sequencing; mut=mutation; wt=wild type.

HRD causes genome-wide loss of heterozygosity (LOH) that can be measured by comprehensive genomic profiling based on NGS

BRCAwt

HGSC patients can be classified into three molecular subgroups: BRCAmut, BRCA-like, Biomarker Negative

15

Frequency of Tumors Extent of Genomic LOH

Genomic LOH cutoff

LOH-low LOH-high

BRCAmut BRCAwt

Initial genomic LOH cutoff derived from public data and prospectively tested in ARIEL2

TCGA and AOCS overall survival data used to develop LOH cutoff to identify HGOC patient tumors with BRCA-like signature Prospective testing of prespecified cutoff in ARIEL2

Genomic LOH Cutoff 0.05 0.10 0.15 0.20 0.25 0.30

Optimal LOH cutoff

Log-rank P Value (high vs low LOH groups)

• TCGA. Nature. (2011);474:609

Wang ZC et al. Clin Cancer Res. (2012);18:5806

Log-rank: P=0.0047 Hazard ratio=0.62

High genomic LOH (n=97) Low genomic LOH (n=212)

100 80 60 40 20 25 50 75 100 125 Overall Survival (months)

Median overall survival: 56.4 vs 38.2 months

Independent predictor from BRCAmut status Survival Probability

Myriad myChoice cutoff

Timms et al. Clin Cancer Res. (2016) 22:3764

ARIEL2 designed to assess rucaparib efficacy in three prospectively defined molecular subgroups

gBRCA=germline BRCA.

Key Eligibility (N=180)

• High-grade serous or

endometrioid OC – Known gBRCA enrollment capped at N=15

• ≥1 prior platinum

chemotherapy

• Platinum-sensitive, relapsed,

measurable disease

• Tumour tissue (screening

biopsy and archival)

BRCAmut LOH-high LOH-low Analysis of HRD Subgroups

Primary endpoint

• PFS

Secondary endpoints

• ORR

– RECIST – RECIST + CA-125

• Safety
• PK

600 mg BID rucaparib until disease progression NGS of tumour tissue allows patients to be classified

All patients (117 matched pairs)

LOHhigh archival LOHlow archival LOHhigh screening 67 17 LOHlow screening 33

Overall similar genomic LOH levels between matched archival tumors and screening biopsies

Classified into LOHhigh and LOHlow groups based on the prespecified genomic LOH cutoff.

19

Increased genomic LOH levels found in only a subset of screening biopsies compared to matched archival tumors

Final efficacy analysis Ariel2: PFS in BRCAmut and LOH-high versus LOH-low patients

Final efficacy analysis NOVA: PFS in gBRCAmut and HRD positive versus HRD negative patients

Mirza et al NEJM (2016) ePub 8th Oct 2016

gBRCAmut HRD positive HRD negative HR= 0.27 HR= 0.38 HR= 0.58 gBRCAWT

Duration of response in BRCAmut, LOH-high, and LOH-low

patients

HRD Subgroup Median duration of response, mo (95% CI) BRCAmut 9.8 (6.4, 12.9) LOH-high 10.8 (5.7, -) LOH-low 5.6 (4.6, 8.5)

Tumours with RAD51C alterations are BRCA-like (high genomic LOH) and responded to rucaparib

HR-pathway gene Genetic alteration type Germline/somatic inference HRD molecular subgroup RECIST response CA-125 response NBN Truncation Germline Biomarker Negative PR Yes RAD51C Truncation Germline BRCA-like PR Yes RAD51C Homozygous Del Somatic BRCA-like PR Yes RAD51C Splice Germline BRCA-like PR Yes RAD51C Splice Germline BRCA-like SD Yes ATM Homozygous Del Somatic Indeterminate SD Yes RAD51L3 Truncation Indeterminate BRCA-like SD Yes BRIP1 Splice Germline Biomarker Negative SD No BRIP1 Truncation Germline Biomarker Negative SD No CHEK2 Splice Indeterminate Biomarker Negative SD No CHEK2 Truncation Germline BRCA-like SD No RAD51L1 Truncation Indeterminate Biomarker Negative SD No NBN Truncation Germline Indeterminate SD NE RAD54L Truncation Somatic (subclonal) Biomarker Negative SD NE FANCA Homozygous Del Somatic BRCA-like SD NE FANCI Truncation Germline Biomarker Negative PD No ATM Truncation Somatic Indeterminate NE NE

PR=partial response; SD=stable disease; PD=progressive disease; NE=not evaluable.

LOHhigh as a predictor of response in BRCAwt tumours

24

Sensitivity (%) p value LOHhigh 78

• HRD mutation

11 <0.001 BRCA1/RAD51C methylation 48 0.02 HRD mutation or BRCA1/RAD51C methylation 59 0.13

How to assess HRD in 2016

• Germline or somatic BRCA1/2 (+/- RAD51C) mutations
• Platinum sensitivity matters – “extreme platinum responders”
• LOH analyses more sensitive than either panel sequencing or

methylation analyses

• Archival tissues can be adequate but fresh biopsy preferable

NOVA – extreme platinum responders

• HGSC = 70%
• Relapse in platinum-sensitive timeframe = 80% of

70%

• CR/Excellent PR to platinum in relapse setting = 40
• 50% of 80% of 70%

22 - 28% all HGSC patients

Ovarian high-grade serous carcinoma

• Extreme chromosomal instability

TCGA, Nature (2011) 474:609

Hoadley et al, Cell (2014) 158:929