Download An Expression Signature to Predict Overall and Cancer

An Expression Signature to Predict Overall and Cancer-Specific
Survival of Prostate Cancer
Zhuochun Peng, Lambert Skoog, Marie Hjelm-Eriksson, Ulrika Harmenberg, Gabriella Cohn Cedermark, Lars ÄhrlundRichter, Yudi Pawitan, Monica Nistér, Sten Nilsson, Chunde Li. Department of Clinical Oncology, Karolinska University
Hospital, Department of Oncology-Pathology, Karolinska Insitutet, Stockholm, Sweden
Background: This study is aiming at identifying biomarkers to accurately predict overall and cancer-specific
survival in prostate cancer patients. In order to explore the importance of embryonic stem cell (ESC) gene
signature, we identified 641 embryonic stem cell gene predictors (ESCGPs).
Methods: Selected ESCGPs and control genes were analyzed by multiplex quantitative PCR using prostate
fine-needle aspiration samples taken at diagnosis from a Swedish cohort of 189 prostate cancer patients
diagnosed between 1986 and 2000. Of all patients, 97.9% had overall and cancer-specific survival data and
77.9% were primarily treated only by hormone therapy. The cohort was divided into one discovery and two
validation subsets. Univariate and multivariate Cox proportional hazard ratios and Kaplan-Meier plots were
used for the survival analysis. A published dataset was used for external validation.
Results: An expression signature of F3, VGLL3 and IGFBP3, was sufficient to categorize the patients into
high-risk, intermediate-risk and low-risk subtypes. The median overall survival of the subtypes was 3.23, 4.00
and 9.85 years respectively. The difference corresponded to HRs of 5.86 (95% CI 2.91-11.78, P<0.001) for the
high-risk and 3.45 (95% CI 1.79-6.66, P<0.001) for the intermediate-risk compared to the low-risk subtype. The
obvious overall and cancer-specific survival difference between the three subtypes was still maintained within
the patients primarily treated by hormone therapy. The overall survival rate at 8 years of three subtypes was
0%, 16% and 55.6%, respectively. This obvious survival difference was independent of age, PSA level, tumor
grade and clinical stage.
Hypothesis of Embryonic Stem Cell Gene Predictors
1.
2.
3.
4.
5.
6.
Verification of Selected ESCGPs in an Swedish Prostate Cancer Cohort
with Fresh Frozen FNA and Complete Survival Follow Up
Embryonic stem cells are the origin of all tissue stem cells, tissue differentiated cells and
cancer stem cells.
Genes that are important in maintaining ES status and regulating differentiation are also
important in maintaining cancer stem cell status and abnormal differentiation
(dedifferentiation).
Genes with strong expression variations among different ES cell lines are not important in this
respect.
Genes that show consistently high or consistently low level in different ES cell lines have the
same functional importance. Different expression patterns of these genes determine different
normal or cancer tissue developments. Theses genes are named as ESCGPs (embryonic stem
cell predictors).
Although cancer stem cells comprise only a small fraction of the cancer tissue cells, the
expression of these ESCGPs can be measured by microarray, RT-PCR or qPCR.
Different expression patterns of these ESCGPs measured in the cancer tissues can reflect
cancer’s biological aggressiveness, predict treatment effect and patient survival.
Discovery set
Verification set 1
Internal
validations
Verification set 2
Verification of important ESCGPs in an
external Swedish Watchful Waiting cohort
(Sboner A et al. BMC Med Genomics 2010)
External
validation
Clinical Features of 189 Prostate Cancer Patients with FNA Analyzed by qPCR
Conclusions: These results suggest that this expression signature could be used to improve the accuracy of
the currently available clinical tools for predicting overall and cancer-specific survival and selecting patients
with potential survival benefit from hormone treatment.
Identification of 641 ESCGPs by Simple Biostatistics
641 ESCGPs
Background: The Increasing Incidence but Stable Mortality
1.
2.
3.
4.
The FNA samples were taken by a routine procedure at Karolinska University Hospital. At least one fresh cytology smear
from each patient was Giemsa stained for cytological diagnosis. The remaining duplicate smears on glass slides were kept
fresh frozen at –70°C. It was confirmed that at least 80% of all cells were cancer cells in most prostate cancer FNA samples.
The 189 prostate cancer patients were diagnosed in 1986-2001.
In this cohort, very few patients had an indolent cancer, over 50% of the patients had a high grade and advanced cancer, and
castration therapy was the only primary treatment for 77.9% of the patients.
PSA value at diagnosis, age at diagnosis, WHO tumor grade, and clinical stage clinical data were included for analysis.
Cox Proportional Hazards by Univariate Analysis in the Complete Set
Prostate Cancer Clinical Outcome Flow Chart
Endpoint
Diagnosis
Insignificant cancer
Diagnosis
Low or
Intermediate
risk cancer
Diagnosis
High or
Intermediate
risk cancer
Radical
Surgery
Or
Curative
Radiation
Radical
Surgery
Or
Curative
Radiation
+
Hormone
therapy
Of the 25 gene markers,
10 (F3, WNT5B, VGLL3,
CTGF, IGFBP3, c-MAF-a,
c-MAF-b, AMACR, MUC1
and EZH2) showed a
significant correlation with
either overall or cancerspecific survival.
No progression
No treatment
Death due to:
Other diseases
Other causes
No recurrence
Salvage
or Postop
Radiation
Relapse
PSA
Death due to:
Prostate cancer
Hormone
therapy
PSA
Castration
Resistance
PSA
Metastases
Palliation
Chemotherapy
Radiation
New therapies
Surrogate
Endpoint ≠ Endpoint
Surrogate
Endpoint ≠ Endpoint
These 641 genes showed a strong ability to classify different subtypes of prostate
tissues and other cancers. A. Original by Lapointe et al., PNAS, 2004. B. By ESCGPs
External Validation of ESCGPs in Swedish Watchful Waiting Cohort
Clear Survival Difference between Tumor Subtypes Classified by ESCGP
Signature 1 (VGLL3, IGFBP3 and F3) in the Complete Data Set
95 had data of the ESCGP signature 1 (VGLL3,
IGFBP3 and F3). FNA samples from these 95
patients were classified into three tumor
subtypes by the ESCGP signature 1 with an
unsupervised hierarchical clustering method
using gene-median centered delta Ct values.
This was visualized by Treeview software.
F3 (negative) and EZH2 (positive), showed a significant association
with early deaths, whereas AMACR and MUC1 were border line
significant. However, there was no significant difference between
the genes and survival in the indolent group.
Survival difference between three tumor subtypes in
patients primarily treated with castration therapy
Discussions
Sampling methods and heterogeneity
•
Present: micro-dissection of representative tumor focuses in core
biopsies
•
Future: MR-guided fine needle aspiration
Overall Survival
Survival difference between three tumor subtypes
in complete set
ESCGPs may reflect not only tumor risk factors but also patient risk factors
•
F3 and IGFBP3 have functional importance in both cancer and other
diseases such as cardiovascular diseases and diabetes
Cancer Specific Survival
Time since diagnosis (Years)
The weight of the ESCGPs in the survival prediction may increase
significantly in early and small prostate cancers for which the power and
accuracy of PSA and Gleason score are decreased
Expected Effect of Clinical Application
Time since diagnosis (Years)
Overall
Survival
Without ESCGP
AUC: 0.7551249
With ESCGP
•
Accuracy of prediction of overall and cancer-specific survival can be
improved to over 80% by the ESCGPs together with clinical parameters.
Watchful Waiting
Or
HormoneTreatment
Radical
Treatments
AUC: 0.8146962
Current
90%
Cancer
Survival
Cancer
With Recurrence
Cancer
No Recurrence
Normal
Two of these genes (F3
and WNT5B) presented a
more significant P value
than did PSA when they
were used as continuous
variables, and this level
of significance remained
after a stringent
Bonferroni correction was
performed for the multiple
testing of 30 variables
(P<β=0.0016667)
10%
Watchful Waiting
Extensive Radical Modest
Or
And Adjuvant
Localized
Treatments
Treatment HormoneTreatment
AUC: 0.7260167
AUC: 0.7930384
Future
30%
ROC for the Prediction of 5 Years Survival
40%
30%
PRESENTED BY: Zhuochun Peng,
[email protected]