Discovery and Validation of a Predictive Biomarker for Breast

Report
Discovery and Validation of a Biomarker for
Chemotherapy Sensitivity in Breast Cancer
Richard Kennedy
McClay Professor Of Medical Oncology
Queen’s University Belfast
Consultant Medical Oncologist Northern Ireland Cancer Centre
[email protected]
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
Need for A Chemotherapy Predictive Test
• Most chemotherapy used in breast cancer causes DNA damage
(alkylating agents, anthracyclines)
• Single agent response is usually around 20%
• We do not know who benefits
The Need for a Chemotherapy Test
• Best effect seen in breast cancer with known BRCA1/2 mutation or loss
of expression (“BRCAness”) – estimated 25% of breast cancers
• These tumours cannot repair their DNA after chemotherapy and are
therefore sensitive
Pre-chemotherapy
Rodger A et al. BMJ 1994;309:1431-1433
Turner N et al Nat Rev Cancer. 2004 Oct;4(10):814-9.
Kennedy RD et al. Journal of the National Cancer Institute 2004;96(22):1659-68.
Fong PC et al N Engl J Med. 2009 Jul 9;361(2):123-34
Fourquet A et al Am J Clin Oncol 2009 32:127-131
Post-chemotherapy
Imyanitov et al Hereditary Cancer in Clinical Practice 2011, 9:5
Loss of BRCA1/2 Predicts Benefit from DNA Damaging Agents
Neoadjuvant FEC Breast Cancer
Adjuvant Cisplatin NSCLC
Font et al Ann Oncol. 2011 Jan;22(1):139-44
Taron et al Hum Mol Genet. 2004 Oct 15;13(20):2443-9
Neoadjuvant Cisplatin Bladder Cancer
Palliative Cisplatin Ovarian Cancer
Quinn et al Clin Cancer Res, 2007 13; 7413
Margeli et al Plos One 2010 5(3)
The FA/BRCA Pathway
BRCA2
BRCA1
PALB2
FAB
FAP
FAM
FAL
FAJ
FAA
FAD1
RAD51c
FAC
ATM
FAI
FAD2
BRIP1
FANCO
ATR
NBS1
P
ub
I
ub
I
ub
BRIP1
PALB2
P
I
O
ub
Adapted Kennedy and D’Andrea JCO 2006:24(23):3799-808
9
• Loss of the FA/BRCA pathway in tumours
may be a good biomarker to guide the use
of DNA damaging therapies
• Problem- there is no reliable method for
detecting its loss in the clinic
ub
I
P
BRIP1 PALB2
I
I
O
Adapted Kennedy and D’Andrea JCO 2006:24(23):3799-808
Reported lost in
solid tumours
ub
BRIP1
I
I
P
PALB2
I
O
Adapted Kennedy and D’Andrea JCO 2006:24(23):3799-808
Strategies to Identify DNA Damage Response Deficient
Tumours That Respond to Chemotherapy
• Sequence DNA repair genes disrupted in cancer:
– BRCA1, BRCA2, FAA, FAB, FAC, FAD2, FAE, FAF, FAG, FAI, FAJ,
FAO, FAP, PALB2, MSH2, MSH6, etc –
– But what is a mutation or a polymorphism?
– Epigenetic silencing important
• Look for loss of expression by IHC/qPCR
– But cannot always distinguish mutant from wildtype.
• Functional assays in tumour biopsies
– H2AX/FANCD2 foci
– Impractical for routine clinical use (FFPE)
Kennedy and D’Andrea 2006 Journal of Clinical Oncology 2006:24(23):3799-808
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
Our Hypothesis
• Although there are multiple ways to lose the FA/BRCA
pathway the cell still needs to survive endogenous
DNA damage.
• There may be a common response pathway in
FA/BRCA deficient breast cancer- could be a
biomarker
• Tumours with BRCA1/2 mutations should display this
pathway.
• Fanconi anaemia patients (homozygous for mutation of
a Fanconi anaemia gene mutation) should also display
this pathway.
Our Approach: Transcriptional profiling
Adopt a dual approach to identify this subgroup based on analysis
of both BRCA1/2 and FA mutation data sets
Tumour dataset
enriched for BRCA1/2
mutations
Fanconi Anaemia
patient dataset
Unsupervised clustering
to identify underlying
molecular subtypes
Identify molecular
processes that are
activated in FA/BRCA
deficient cells
Identify molecular
subtype representing
molecular processes
activated in FA/BRCA
deficient cells
Analysis of Fanconi Anaemia Clinical Samples
•
•
•
•
•
•
•
Grover Bagby microarray data for FA patient bone marrows
Affymetrix HG-U133A
22 FA patients with a range of mutations in FA genes:
– FANCA
– FANCC
– FANCD2
– FANCD1(BRCA2)
– FANCE
– FANCF
– FANCG
10 normal controls
Pre-processing of data using RMA
Differentially expressed probesets (FC>3, p-valueFDR <0.001)
identified using 1-way ANOVA analysis
Identified top molecular processes using Metacore Gene Ontology
Software- (p-value FDR<0.05) mostly type I interferon related
Analysis of Clinical Samples
•Collaboration with Mayo Clinic (Fergus Couch)
•Collected 146 breast cancer samples which were
enriched with 78 known BRCA1/2 mutant samples.
•Analysed transcriptional profiles using a cDNA microarray
optimised for FFPE –Breast Cancer DSA (customized
Affymetrix platform)
•Unsupervised clustering analysis applied to identify
molecular subtypes within the ER-ve and ER+ve
populations
Molecular Subtyping
• The data matrices were standardised to the
median probeset expression values.
• Hierarchical clustering (Pearson correlation
distance and Ward’s linkage) was applied to
data matrices on both probesets and samples.
• The number of sub clusters was determined
using the gap statistic
Tibshirani, R., Walther, G. and Hastie, T. (2001). Estimating the
number of data clusters via the Gap statistic. Journal of the Royal
Statistical Society B, 63, 411–423.
Analysis of Breast Cancer Dataset ER-ve Samples
Distinct group with DNA damage related biology (65% BRCA mutant)
Analysis of Breast Cancer Dataset ER+ve Samples
Distinct group with DNA damage related biology (68% BRCA mutant)
Integration of Breast Cancer and FA Patient
Results
ER-ve breast cancer patients
ER+ve breast cancer patients
BRCA1/2 mutant enriched subgroups display identical immune signaling
pathways to Fanconi anaemia patient cells (p<1x10-15)
22
Summary so Far
• We identified a novel molecular subgroup in ER
negative and ER positive breast cancer
– Defined by the same immune processes activated in
patients with Fanconi anaemia.
– Significantly enriched for BRCA1/2 mutations
• Could this subgroup represent a biomarker for
breast cancer therapy?
Generation of a Gene Signature Assay to Identify
the DDRD Subgroup Prospectively
ER-ve breast cancer patients
ER+ve breast cancer patients
Classifier
pipeline
gene signature
DDRD- DNA Damage
Response Deficient
Signature Generation Workflow
• Pre-filtering
– Variance-Intensity filtering to remove 75% of
low variance and low intensity probesets
• Classification methods
– Partial Least Squares (PLS)
– Shrinkage Discriminant Analysis (SDA)
– Diagonal SDA
• Feature selection
– Iterative feature selection (uses internal feature
ranking based on the classification algorithm
and removes the lowest ranked first)
– Maximum AUC main criteria in selection of
optimal number of features over crossvalidation
A
AUC Curves
B
44 transcript
C
D
10 repeats of 5-fold cross-validation
DDRD Signature Genes
Top genes in 44 transcript signature- Type I Interferon signalling
1 CXCL10
2 MX1
3 IDO1
4 IFI44L
5 IFI16
recruits CD8+ve T Cells through CXCR3
direct anti-viral GTPase effect
anti-viral metabolic function
interferon induced, function unclear
interferon induced, associated with DNA damage
Cut-off score for positive and negative DDRD result locked at this stage
The DDRD signature detects dysfunction
within the Fanconi Anemia/BRCA pathway
• The DDRD signature distinguishes
between the FA mutant and normal
samples with an AUC of 0.92 (CI = 0.78 1.000, P < 0.001)
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
Independent Validation of the DRDD 44
Transcript Assay as a Predictive Biomarker
1. Neoadjuvant setting
–
–
Patients receive 6 cycles of FEC/FAC chemotherapy
prior to breast cancer surgery
Used in patients with large tumours
2. Adjuvant setting
–
–
Patients receive 6 cycles of FEC chemotherapy
following breast cancer surgery
Used for majority of early breast cancer patients with
high risk disease.
Independent Validation of the DRDD 44
Transcript Assay as a Predictive Biomarker
1. Neoadjuvant setting
–
–
Patients receive 6 cycles of FEC/FAC chemotherapy
prior to breast cancer surgery
Used in patients with large tumours
2. Adjuvant setting
–
–
Patients receive 6 cycles of FEC chemotherapy
following breast cancer surgery
Used for majority of early breast cancer patients with
high risk disease.
DDRD Test Validation in
Neoadjuvant Setting
• 203 neoadjuvant treated breast cancer patients
• Endpoint: Pathologic complete response
• Allowed predictive utility to be assessed
Data set
FAC1
FAC2
FEC
Sample
Number
87
50
66
Array
ER status
Platform
ER-Positive & ER-negative Affy U133A array
ER-Positive & ER-negative Affy U133A array
ER-negative
Affy X3P array
Reference
Tabchy et. al. Clin Can Res 2012
Iwamto et. al. JNCI 2011
Bonnefoi et. al. Lancet Oncol. 2007
The DDRD Assay Predicts Response to
Chemotherapy
203 patients treated with FEC/FAC neoadjuvant chemotherapy
AUC
Sens
Spec
NPV
PPV
Acc
0.78
0.82
0.59
0.90
0.44
0.76
Lower C.I.Upper C.I.
0.70
0.85
0.71
0.92
0.53
0.63
0.83
0.95
0.37
0.48
0.66
0.83
Independent Validation of the DRDD 44
Transcript Assay as a Predictive Biomarker
1. Neoadjuvant setting
–
–
Patients receive 6 cycles of FEC/FAC chemotherapy
prior to breast cancer surgery
Used in patients with large tumours
2. Adjuvant setting
–
–
Patients receive 6 cycles of FEC chemotherapy
following breast cancer surgery
Used for majority of early breast cancer patients with
high risk disease.
DDRD Test Validation in The
Adjuvant Setting
• Prospective-retrospective independent validation
• 191 adjuvant FEC treated breast cancer patients treated
in NI Cancer Centre 2000-2007
• At least 5 year follow-up available
• Applied the DDRD assay to macrodissected 10 uM
FFPE sections
QC Approach
DDRD Positive Patients Have Greater RFS
Following Chemotherapy
Performance of the DDRD Breast Dx Assay to 191 breast cancers treated with
adjuvant DNA-damaging FEC chemotherapy
Table 3 Multivariate analysis of the predictive value of the DDRD assay for relapse
free survival 5 years following FEC chemotherapy
Assay Performance in the Adjuvant Setting
Parameter
Multivariable Analysis Odds Ratio with
95% CI*†
DDRD-positive
0.37 (0.15, 0.88)
ER-negative
0.46 (0.20, 1.06)
HER2-positive
1.72 (0.79, 3.73)
Age
0.59 (0.32, 1.08)
Grade (3 vs 1&2)‡
1.14 (0.48, 2.71)
T Stage
T1
1.00
T2
0.63 (0.29, 1.40)
T3
1.13 (0.35, 3.65)
Nodal Status
N0
1.00
N1
1.02 (0.47, 2.21)
DDRD = DNA-Damage Response Deficiency, ER = Estrogen Receptor
lymphocytic infiltration within validation datasets
Association between DDRD positivity
and other pathological factors
Total
% DDRD-
Association p-
number
positive
value*
ER-positive
112
25.00
ER-negative
77
51.95
HER2-positive
46
41.30
HER2-negative
128
32.03
Triple negative
44
54.55
Non-triple negative
135
28.15
Lymphocytic infiltrate positive
35
74.29
Lymphocytic infiltrate negative
155
27.74
Total
% DDRD-
Association p-
number
positive
value*
ER-positive
70
24.29
ER-negative
134
67.91
Adjuvant Dataset
Neoadjuvant Dataset
0.0001
0.2564
0.0014
<0.0001
0.0001
Analytical Validation
Reagent effects
Operator effects
Operator 1
Operator 2
Centre effects
Analytical
•
•
•
•
Clinical and Laboratory Standards Institute guidelines
3 samples spanning the signature score range
processed in duplicate over 9 days
Factors included
–
–
–
–
operator (3 in total),
cDNA amplification (3 NuGEN WT-Systems used)
biotin labelling (3 NuGEN Encore Biotin Modules used)
hybridisation, washing and staining (3 Affymetrix GeneChip® Hybridization,
Wash, and Stain Kits used)
– Pathology lab
• Total standard deviation estimates with 95% confidence intervals
and variance was estimated for each sample profiled.
• The total standard deviation <5% of the DDRD assay’s reportable
range
• The variance was determined to be constant across the signature
score range.
The DDRD Assay is Only Prognostic in the
Context of DNA Damaging Chemotherapy
DDRD test predicts
benefit to DNA
damage based
therapy and is not
a prognostic test unlike Oncotype
Dx, MammaPrint
664 patients
No chemotherapy
Desmedt C, Piette F, Loi S et al. Clin Cancer Res. 2007;13:3207-3214.
Wang Y, Klijn JG, Zhang Y et al. Lancet. 2005;365:671-679.
Sorlie T, Perou CM, Fan C et al. Mol Cancer Ther. 2006;5:2914-2918.
42
Current Clinical Practice
43
DDRD Impact on Clinical Practice
44
Clinical application of DDRD assay
• Aim is to launch 2014
• Main applications
– DNA damaging Chemotherapy versus no
chemotherapy in intermediate risk patients
– Anthracycline versus taxane based chemotherapy
45
Application of DDRD Assay to
Other Cancer Types
Epithelial Ovarian Cancer
• Collaboration with Charlie Gourley University of
Edinburgh
• 157 high grade serous ovarian cancers treated with 6
cycles of cisplatin
• Sample Type: FFPE 10 μm section
• Profiled on Ovarian DSA Affymetrix platform and DDRD
assay signature applied
• Endpoint defined as either:
– Response measured using RECIST or CA125
DDRD Assay Predicts for Response in
Platinum Treated Ovarian Cancer
 OR for response= 3.46 (2.10-4.76)
DDRD Predicts for overall Survival in
Serous Ovarian Cancer
HR-0.56 (0.46-0.68)
DDRD in Early Stage Oesophageal
Adenocarcinoma
• n=43, 11% DDRD +ve
• Median RFS DDRD +ve:
Undefined
• Median RFS DDRD –ve: 36.59
months
• Median OS DDRD +ve:
92.31 months
• Median OS DDRD –ve:
36.39 months
Richard Turkington ACL QUB
DDRD in Oesophageal Adenocarcinoma
Multivariable analysis of recurrence-free survival (n=43, 20 non-events and 23 events)
Multivariable analysis of overall survival (n=43, 17 non-events and 26 events)
Richard Turkington
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
BRCA1 mutant breast and ovarian cancers are
associated with lymphocytic Infiltrate
• Poorly differentiated and highly proliferative.
• Demonstrate a CXCR3+ T cell lymphocytic infiltrate.
Sporadic
FA/BRCA pathway deficient
Lakhani et al Breast Cancer Res. 1999;1(1):31-5
Fujiwara et al Am J Surg Pathol. 2012 Aug;36(8):1170-7
Lymphocytic Infiltrate Is Associated with
Chemotherapy Response in Breast Cancer
•
•
•
•
•
Denkert et al JCO 2010 105-133
1058 patients studied
OR of 1.21 (1.08-1.36) for pCR with lymphocyte infiltrate
Tumour expresses CXCL9 and 10 (mRNA level)
CXCR3 positive T cell infiltrate (receptor for CXCL9/10)
Does loss of the FA/BRCA Pathway
Activate the DDRD Immune Response?
Cell line data
72Hr Response Versus Signature Score
MDA-MB468
Signature Score
More DNA Damage Response Deficiency
Test Score Correlates With Cisplatin Sensitivity in Cell Lines
HCC1937
T47D
ZR75-1
MDA-MB231
HS578T
-log10 IC50
More Sensitivity to Cisplatin
BRCA1 Deficient HCC1937 Cell Line Has a High DDRD Score
0.70
MDA-MB468
0.60
Signature Score
More DNA repair Deficiency
72Hr Response Versus Signature Score
HCC1937
0.50
0.40
T47D
0.30
0.20
MDA-MB231
0.10
0.00
3.00
3.50
4.00
4.50
HS578T
5.00
5.50
6.00
-log10 IC50
More Sensitivity to Cisplatin
Quinn, Kennedy et al. Cancer Research 2003;63:6221-8
Correction of BRCA1 Deficiency Reduces DDRD Signature Score
0.70
0.60
Signature Score
More DNA repair Deficiency
72Hr Response Versus Signature Score
HCC1937
0.50
0.40
T47D
0.30
0.20
MDA-MB231
HCC-BR
0.10
0.00
3.00
3.50
4.00
HS578T
4.50
5.00
5.50
6.00
-log10 IC50
More Sensitivity to Cisplatin
HCC 1937
E
V
BR
BRCA1
 γ Tubulin
Quinn, Kennedy et al. Cancer Research 2003;63:6221-8
Why does loss of FA/BRCA
Pathway (BRCAness) cause an
Immune Response?
CXCL10 is Markedly Induced by 5-HU in FA/BRCA
Deficient cells
HU=Hydroxurea
(S phase DNA replication)
Tax-Paclitaxel
(M phase mitosis)
HU and Taxol selected as IC30 dose at 72hrs
Cisplatin Induces CXCL10 Expression in S Phase
Liu et al Mol Cancer Res 2009 (7) 1099-1109
Hypothesis: Replication Fork Stall Activates
DDRD Immune Response
Endogenous damage
Replication Fork Stall
FA pathway:
>TLS
>HR
>NHEJ
Replication
Re-established
Hypothesis: Replication Fork Stall Activates
DDRD Immune Response
Endogenous damage
Replication Fork Stall
FA pathway:
>TLS
>HR
>NHEJ
???
IFN
genes
Replication
Re-established
Outline
• Chemotherapy for Breast Cancer
• Identification of a clinical biomarker for loss of the
FA/BRCA pathway
• Independent validation of the biomarker
• Biology underlying the biomarker
• Conclusions and future directions
Conclusions
• We have developed an assay that can predict benefit from DNA
damaging chemotherapy in breast cancer
• Appears to represent a specific immune response in S phase of the cell
cycle –possibly due to abnormal DNA structures
• DDRD positive patients
– 4 times more likely to gain a complete response neoadjuvant setting
– 3 times less likely to develop recurrence adjuvant setting
• Aim to enter clinic by end 2014
Future Directions
1.
Basic Science
•
•
2.
How endogenous DNA damage activates immune response
Requirement of immune response for survival of FA/BRCA pathway
deficient tumours
Clinical- Application to other cancers
•
•
•
3.
Prostate cancer- University of Surrey- Cyclophosphamide
Does it predict radiation sensitivity?- need dataset
Does it predict response to PARP inhibitors?- need dataset
Neoadjuvant chemotherapy in breast cancer study
•
FEC versus paclitaxel sequencing study based on biomarker
Acknowledgements
Acknowledgments
CCRCB
Almac UK
Eileen Parkes Steven
Walker
Mary Harte
Nuala McCabe
Paul Mullan
Gareth Irwin
Jackie James
Manuel Salto-Tellez
David Boyle
Jennifer Quinn
Richard Turkington
Jude O’Donnell
Laura Hill
Steve Deharo
Timothy Davison
Karen Keating
Fionnuala McDyer
Paul Harkin
Mayo Clinic
Fergus Couch
Harvard Med
School
Alan D’Andrea
The Patients

similar documents