Prediction consistency and clinical presentations of breast cancer molecular subtypes for Han Chinese population

Core promoter in TNBC is highly mutated with rich ethnic signature

The core promoter plays an essential role in regulating transcription initiation by controlling the interaction between transcriptional factors and sequence motifs in the core promoter. Although mutation in core promoter sequences is expected to cause abnormal gene expression leading to pathogenic consequences, limited supporting evidence showed the involvement of core promoter mutation in diseases. Our previous study showed that the core promoter is highly polymorphic in worldwide human ethnic populations in reflecting human history and adaptation. Our recent characterization of the core promoter in triple-negative breast cancer (TNBC), a subtype of breast cancer, in a Chinese TNBC cohort revealed the wide presence of core promoter mutation in TNBC. In the current study, we analyzed the core promoter in a Thai TNBC cohort. We also observed rich core promoter mutation in the Thai TNBC patients. We compared the core promoter mutations between Chinese and Thai TNBC cohorts. We observed substantial differences of core promoter mutation in TNBC between the two cohorts, as reflected by the mutation spectrum, mutation-effected gene and functional category, and altered gene expression. Our study confirmed that the core promoter in TNBC is highly mutable, and is highly ethnic-specific.

Etiological roles of core promoter variation in triple-negative breast cancer

Abnormal gene expression plays key role in cancer development. A core promoter is located around the transcriptional start site. Through interaction between core promoter sequences and transcriptional factors, core promoter controls transcriptional initiation. We hypothesized that in cancer, core promoter sequences could be mutated to interfere the interaction with transcriptional factors, resulting in altered transcriptional initiation and abnormal gene expression and cancer development. We used triple-negative breast cancer (TNBC) as a model to test our hypothesis. We collected genome-wide core promoter variants from 279 TNBC genomes. After extensive filtering of normal genomic polymorphism, we identified 19,427 recurrent somatic variants in 1,238 core promoters of 1,274 genes and 1,694 recurrent germline variants in 272 core promoters of 294 genes. Many of the affected genes were oncogenes and tumor suppressors. Analysis of RNA-seq data from the same patient cohort identified increased or decreased gene expression in 439 somatic and 85 germline variants-affected genes, and the results were validated by luciferase reporter assay. By comparing with the core promoter variation data from 610 unclassified breast cancer, we observed that core promoter variants in TNBC were highly TNBC-specific. We further identified the drugs targeting the genes with core promoter variation. Our study demonstrates that core promoter is highly mutable in cancer, and can play etiological roles in TNBC and other types of cancer through influencing transcriptional initiation.

Molecular subtyping of breast cancer intrinsic taxonomy with oligonucleotide microarray and NanoString nCounter

Breast cancer intrinsic subtypes have been identified based on the transcription of a predefined gene expression (GE) profiles and algorithm (PAM50). This study compared molecular subtyping with oligonucleotide microarray and NanoString nCounter assay. A total of 109 Taiwanese breast cancers (24 with adjacent normal breast tissues) were assayed with Affymetrix Human Genome U133 plus 2.0 microarrays and 144 were assayed with the NanoString nCounter while 64 patients were assayed for both platforms. Subtyping with the nearest centroid (single sample prediction) was performed, and 16 out of 24 (67%) matched normal breasts were categorized as the normal breast-like subtype. For 64 breast cancers assayed for both platforms, 41 (65%, one unclassified by microarray) were predicted with an identical subtype, resulting in a fair Kappa statistic of 0.60. Taking nCounter subtyping as the gold standard, prediction accuracy was 43% (3/7), 81% (13/16), 25% (5/20), and 100% (20/20) for basal-like, HER2-enriched, luminal A and luminal B subtype predicted from microarray GE profiles. Microarray identified more luminal B cases from luminal A subtype predicted by nCounter. It's not uncommon to use microarray for breast cancer molecular subtyping for research. Our study showed that fundamental discrepancy existed between distinct GE assays, and cross platform equivalence should be carefully appraised when molecular subtyping was conducted with oligonucleotide microarray.

Residual risk stratification of Taiwanese breast cancers following curative therapies with the extended concurrent genes signature

INTRODUCTION

The aim of the study was to perform digital RNA counting to validate a gene expression signature for operable breast cancers initially treated with curative intention, and the risk of recurrence, distant metastasis, and mortality was predicted.

METHODS

Candidate genes were initially discovered from the coherent genomic and transcriptional alternations from microarrays, and the extended concurrent genes were used to build a risk stratification model from archived formalin-fixed paraffin-embedded (FFPE) tissues with the NanoString nCounter.

RESULTS

The extended concurrent genes signature was prognostic in 144 Taiwanese breast cancers (5-year relapse-free survival: 89.8 and 69.4% for low- and high-risk group, log-rank test: P = 0.004). Cross-platform comparability was evidenced from significant and positive correlations for most genes as well as equal covariance matrix across 64 patients assayed for both microarray and digital RNA counting.

DISCUSSION

Archived FFPE samples could be successfully assayed by the NanoString nCounter. The purposed signature was prognostic stratifying breast cancer patients into groups with distinct survival patterns, and clinical applicability of the residual risk model was proved.

Activity of distinct growth factor receptor network components in breast tumors uncovers two biologically relevant subtypes

BACKGROUND

The growth factor receptor network (GFRN) plays a significant role in driving key oncogenic processes. However, assessment of global GFRN activity is challenging due to complex crosstalk among GFRN components, or pathways, and the inability to study complex signaling networks in patient tumors. Here, pathway-specific genomic signatures were used to interrogate GFRN activity in breast tumors and the consequent phenotypic impact of GRFN activity patterns.

METHODS

Novel pathway signatures were generated in human primary mammary epithelial cells by overexpressing key genes from GFRN pathways (HER2, IGF1R, AKT1, EGFR, KRAS (G12V), RAF1, BAD). The pathway analysis toolkit Adaptive Signature Selection and InteGratioN (ASSIGN) was used to estimate pathway activity for GFRN components in 1119 breast tumors from The Cancer Genome Atlas (TCGA) and across 55 breast cancer cell lines from the Integrative Cancer Biology Program (ICBP43). These signatures were investigated for their relationship to pro- and anti-apoptotic protein expression and drug response in breast cancer cell lines.

RESULTS

Application of these signatures to breast tumor gene expression data identified two novel discrete phenotypes characterized by concordant, aberrant activation of either the HER2, IGF1R, and AKT pathways ("the survival phenotype") or the EGFR, KRAS (G12V), RAF1, and BAD pathways ("the growth phenotype"). These phenotypes described a significant amount of the variability in the total expression data across breast cancer tumors and characterized distinctive patterns in apoptosis evasion and drug response. The growth phenotype expressed lower levels of BIM and higher levels of MCL-1 proteins. Further, the growth phenotype was more sensitive to common chemotherapies and targeted therapies directed at EGFR and MEK. Alternatively, the survival phenotype was more sensitive to drugs inhibiting HER2, PI3K, AKT, and mTOR, but more resistant to chemotherapies.

CONCLUSIONS

Gene expression profiling revealed a bifurcation pattern in GFRN activity represented by two discrete phenotypes. These phenotypes correlate to unique mechanisms of apoptosis and drug response and have the potential of pinpointing targetable aberration(s) for more effective breast cancer treatments.

Molecular portrait of breast cancer in China reveals comprehensive transcriptomic likeness to Caucasian breast cancer and low prevalence of luminal A subtype

The recent dramatic increase in breast cancer incidence across China with progressive urbanization and economic development has signaled the urgent need for molecular and clinical detailing of breast cancer in the Chinese population. Our analyses of a unique transethnic collection of breast cancer frozen specimens from Shanghai Fudan Cancer Center (Chinese Han) profiled simultaneously with an analogous Caucasian Italian series revealed consistent transcriptomic data lacking in batch effects. The prevalence of Luminal A subtype was significantly lower in Chinese series, impacting the overall prevalence of estrogen receptor (ER)-positive disease in a large cohort of Chinese/Caucasian patients. Unsupervised and supervised comparison of gene and microRNA (miRNA) profiles of Chinese and Caucasian samples revealed extensive similarity in the comprehensive taxonomy of transcriptional elements regulating breast cancer biology. Partition of gene expression data using gene lists relevant to breast cancer as "intrinsic" and "extracellular matrix" genes identified Chinese and Caucasian subgroups with equivalent global gene and miRNA profiles. These findings indicate that in the Chinese and Caucasian groups, breast neoplasia and the surrounding stromal characteristics undergo the same differentiation and molecular processes. Transcriptional similarity across transethnic cohorts may simplify translational medicine approaches and clinical management of breast cancer patients worldwide.

Multiclass prediction with partial least square regression for gene expression data: applications in breast cancer intrinsic taxonomy

Multiclass prediction remains an obstacle for high-throughput data analysis such as microarray gene expression profiles. Despite recent advancements in machine learning and bioinformatics, most classification tools were limited to the applications of binary responses. Our aim was to apply partial least square (PLS) regression for breast cancer intrinsic taxonomy, of which five distinct molecular subtypes were identified. The PAM50 signature genes were used as predictive variables in PLS analysis, and the latent gene component scores were used in binary logistic regression for each molecular subtype. The 139 prototypical arrays for PAM50 development were used as training dataset, and three independent microarray studies with Han Chinese origin were used for independent validation (n = 535). The agreement between PAM50 centroid-based single sample prediction (SSP) and PLS-regression was excellent (weighted Kappa: 0.988) within the training samples, but deteriorated substantially in independent samples, which could attribute to much more unclassified samples by PLS-regression. If these unclassified samples were removed, the agreement between PAM50 SSP and PLS-regression improved enormously (weighted Kappa: 0.829 as opposed to 0.541 when unclassified samples were analyzed). Our study ascertained the feasibility of PLS-regression in multi-class prediction, and distinct clinical presentations and prognostic discrepancies were observed across breast cancer molecular subtypes.

Concurrent gene signatures for han chinese breast cancers

The interplay between copy number variation (CNV) and differential gene expression may be able to shed light on molecular process underlying breast cancer and lead to the discovery of cancer-related genes. In the current study, genes concurrently identified in array comparative genomic hybridization (CGH) and gene expression microarrays were used to derive gene signatures for Han Chinese breast cancers.

We performed 23 array CGHs and 81 gene expression microarrays in breast cancer samples from Taiwanese women. Genes with coherent patterns of both CNV and differential gene expression were identified from the 21 samples assayed using both platforms. We used these genes to derive signatures associated with clinical ER and HER2 status and disease-free survival.

Distributions of signature genes were strongly associated with chromosomal location: chromosome 16 for ER and 17 for HER2. A breast cancer risk predictive model was built based on the first supervised principal component from 16 genes (RCAN3, MCOLN2, DENND2D, RWDD3, ZMYM6, CAPZA1, GPR18, WARS2, TRIM45, SCRN1, CSNK1E, HBXIP, CSDE1, MRPL20, IKZF1, and COL20A1), and distinct survival patterns were observed between the high- and low-risk groups from the combined dataset of 408 microarrays. The risk score was significantly higher in breast cancer patients with recurrence, metastasis, or mortality than in relapse-free individuals (0.241 versus 0, P<0.001). The concurrent gene risk predictive model remained discriminative across distinct clinical ER and HER2 statuses in subgroup analysis. Prognostic comparisons with published gene expression signatures showed a better discerning ability of concurrent genes, many of which were rarely identifiable if expression data were pre-selected by phenotype correlations or variability of individual genes.

We conclude that parallel analysis of CGH and microarray data, in conjunction with known gene expression patterns, can be used to identify biomarkers with prognostic values in breast cancer.