Limits...
Extensive Transcriptomic and Genomic Analysis Provides New Insights about Luminal Breast Cancers.

Tishchenko I, Milioli HH, Riveros C, Moscato P - PLoS ONE (2016)

Bottom Line: Our results indicate that the separation between the molecular luminal A and B subtypes-per definition-is not associated with intrinsic characteristics evident in the differentiation between other subtypes.The comparison with the current separation into luminal A and B subtypes revealed a substantially improved survival stratification.A proposition for a revisited delineation is provided in this study.

View Article: PubMed Central - PubMed

Affiliation: Information-based Medicine Program, Hunter Medical Research Institute, New Lambton Heights, NSW, Australia.

ABSTRACT
Despite constituting approximately two thirds of all breast cancers, the luminal A and B tumours are poorly classified at both clinical and molecular levels. There are contradictory reports on the nature of these subtypes: some define them as intrinsic entities, others as a continuum. With the aim of addressing these uncertainties and identifying molecular signatures of patients at risk, we conducted a comprehensive transcriptomic and genomic analysis of 2,425 luminal breast cancer samples. Our results indicate that the separation between the molecular luminal A and B subtypes-per definition-is not associated with intrinsic characteristics evident in the differentiation between other subtypes. Moreover, t-SNE and MST-kNN clustering approaches based on 10,000 probes, associated with luminal tumour initiation and/or development, revealed the close connections between luminal A and B tumours, with no evidence of a clear boundary between them. Thus, we considered all luminal tumours as a single heterogeneous group for analysis purposes. We first stratified luminal tumours into two distinct groups by their HER2 gene cluster co-expression: HER2-amplified luminal and ordinary-luminal. The former group is associated with distinct transcriptomic and genomic profiles, and poor prognosis; it comprises approximately 8% of all luminal cases. For the remaining ordinary-luminal tumours we further identified the molecular signature correlated with disease outcomes, exhibiting an approximately continuous gene expression range from low to high risk. Thus, we employed four virtual quantiles to segregate the groups of patients. The clinico-pathological characteristics and ratios of genomic aberrations are concordant with the variations in gene expression profiles, hinting at a progressive staging. The comparison with the current separation into luminal A and B subtypes revealed a substantially improved survival stratification. Concluding, we suggest a review of the definition of luminal A and B subtypes. A proposition for a revisited delineation is provided in this study.

No MeSH data available.


Related in: MedlinePlus

Separation features between luminal A and luminal B, basal-like, and HER2-enriched subtypes defined by PAM50 assay.The heat maps are generated from Illumina probe profiles, normalised using mean expression levels of control samples (black), where an over-expression relative to controls is shown in red, and an under-expression in green. Samples in each heat map are ordered by expression levels of the probe mostly differentiating between the corresponding pair of subtypes. (a) Luminal A (n = 451) and B (n = 229) samples are ordered by CEP55. There is no clear boundary between these subtypes evident, and the global distribution functions of the top two genes mostly differentiating between these subtypes, are unimodal. (b) Luminal subtypes ordered by expression levels of MKI67. (c) Luminal A (n = 451) and basal-like (n = 125) samples are ordered by absolute expression values of MLPH. These subtypes exhibit varying expression levels relative to controls (under- and over-expression), and the global density distribution functions of the top two genes are multi-modal with two peaks—one corresponding to luminal A and the other one exclusively to basal-like. (d) Luminal A (n = 451) and HER2-enriched (n = 91) samples are ordered by ESR1. These subtypes also show varying expression levels relative to controls, and the global density distribution functions of the top two probes are multi-modal.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4920434&req=5

pone.0158259.g001: Separation features between luminal A and luminal B, basal-like, and HER2-enriched subtypes defined by PAM50 assay.The heat maps are generated from Illumina probe profiles, normalised using mean expression levels of control samples (black), where an over-expression relative to controls is shown in red, and an under-expression in green. Samples in each heat map are ordered by expression levels of the probe mostly differentiating between the corresponding pair of subtypes. (a) Luminal A (n = 451) and B (n = 229) samples are ordered by CEP55. There is no clear boundary between these subtypes evident, and the global distribution functions of the top two genes mostly differentiating between these subtypes, are unimodal. (b) Luminal subtypes ordered by expression levels of MKI67. (c) Luminal A (n = 451) and basal-like (n = 125) samples are ordered by absolute expression values of MLPH. These subtypes exhibit varying expression levels relative to controls (under- and over-expression), and the global density distribution functions of the top two genes are multi-modal with two peaks—one corresponding to luminal A and the other one exclusively to basal-like. (d) Luminal A (n = 451) and HER2-enriched (n = 91) samples are ordered by ESR1. These subtypes also show varying expression levels relative to controls, and the global density distribution functions of the top two probes are multi-modal.

Mentions: The differentiation between luminal A and B tumours, where all samples are ordered by CEP55—the Illumina probe with the most separation power between these two groups—is shown in Fig 1a. This image reveals a rather gradual change of expression levels of the PAM50 genes mainly correlated to cell cycle and cell proliferation, such as BIRC5, CCNB1, CDC20, CEP55, KIF2C, MELK, MKI67 and UBE2C. There is no clear boundary evident in this data set with regards to the PAM50 list, and all expression levels seem to diverge from the mean values of controls in the same direction (luminal A and B are both either under- or over-expressed for each probe). To further support this observation, we plotted density distribution functions and ordered mRNA expression values of two most representative probes of the segregation between luminal A and B subtypes (CEP55 and PTTG1), also shown in Fig 1a; corresponding plots for all PAM50 genes are provided in S1 and S2 Figs. These two global density distributions are unimodal functions with only one local maximum, hinting that any separation based on thresholds, such as absolute expression levels corresponding to proliferation states, may be ambiguous. The almost linear gradient of the function representing ordered expression values, at the region where luminal A and B samples meet, also suggests an underlying unavoidable uncertainty in these groups separation based on the given set of genes.


Extensive Transcriptomic and Genomic Analysis Provides New Insights about Luminal Breast Cancers.

Tishchenko I, Milioli HH, Riveros C, Moscato P - PLoS ONE (2016)

Separation features between luminal A and luminal B, basal-like, and HER2-enriched subtypes defined by PAM50 assay.The heat maps are generated from Illumina probe profiles, normalised using mean expression levels of control samples (black), where an over-expression relative to controls is shown in red, and an under-expression in green. Samples in each heat map are ordered by expression levels of the probe mostly differentiating between the corresponding pair of subtypes. (a) Luminal A (n = 451) and B (n = 229) samples are ordered by CEP55. There is no clear boundary between these subtypes evident, and the global distribution functions of the top two genes mostly differentiating between these subtypes, are unimodal. (b) Luminal subtypes ordered by expression levels of MKI67. (c) Luminal A (n = 451) and basal-like (n = 125) samples are ordered by absolute expression values of MLPH. These subtypes exhibit varying expression levels relative to controls (under- and over-expression), and the global density distribution functions of the top two genes are multi-modal with two peaks—one corresponding to luminal A and the other one exclusively to basal-like. (d) Luminal A (n = 451) and HER2-enriched (n = 91) samples are ordered by ESR1. These subtypes also show varying expression levels relative to controls, and the global density distribution functions of the top two probes are multi-modal.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4920434&req=5

pone.0158259.g001: Separation features between luminal A and luminal B, basal-like, and HER2-enriched subtypes defined by PAM50 assay.The heat maps are generated from Illumina probe profiles, normalised using mean expression levels of control samples (black), where an over-expression relative to controls is shown in red, and an under-expression in green. Samples in each heat map are ordered by expression levels of the probe mostly differentiating between the corresponding pair of subtypes. (a) Luminal A (n = 451) and B (n = 229) samples are ordered by CEP55. There is no clear boundary between these subtypes evident, and the global distribution functions of the top two genes mostly differentiating between these subtypes, are unimodal. (b) Luminal subtypes ordered by expression levels of MKI67. (c) Luminal A (n = 451) and basal-like (n = 125) samples are ordered by absolute expression values of MLPH. These subtypes exhibit varying expression levels relative to controls (under- and over-expression), and the global density distribution functions of the top two genes are multi-modal with two peaks—one corresponding to luminal A and the other one exclusively to basal-like. (d) Luminal A (n = 451) and HER2-enriched (n = 91) samples are ordered by ESR1. These subtypes also show varying expression levels relative to controls, and the global density distribution functions of the top two probes are multi-modal.
Mentions: The differentiation between luminal A and B tumours, where all samples are ordered by CEP55—the Illumina probe with the most separation power between these two groups—is shown in Fig 1a. This image reveals a rather gradual change of expression levels of the PAM50 genes mainly correlated to cell cycle and cell proliferation, such as BIRC5, CCNB1, CDC20, CEP55, KIF2C, MELK, MKI67 and UBE2C. There is no clear boundary evident in this data set with regards to the PAM50 list, and all expression levels seem to diverge from the mean values of controls in the same direction (luminal A and B are both either under- or over-expressed for each probe). To further support this observation, we plotted density distribution functions and ordered mRNA expression values of two most representative probes of the segregation between luminal A and B subtypes (CEP55 and PTTG1), also shown in Fig 1a; corresponding plots for all PAM50 genes are provided in S1 and S2 Figs. These two global density distributions are unimodal functions with only one local maximum, hinting that any separation based on thresholds, such as absolute expression levels corresponding to proliferation states, may be ambiguous. The almost linear gradient of the function representing ordered expression values, at the region where luminal A and B samples meet, also suggests an underlying unavoidable uncertainty in these groups separation based on the given set of genes.

Bottom Line: Our results indicate that the separation between the molecular luminal A and B subtypes-per definition-is not associated with intrinsic characteristics evident in the differentiation between other subtypes.The comparison with the current separation into luminal A and B subtypes revealed a substantially improved survival stratification.A proposition for a revisited delineation is provided in this study.

View Article: PubMed Central - PubMed

Affiliation: Information-based Medicine Program, Hunter Medical Research Institute, New Lambton Heights, NSW, Australia.

ABSTRACT
Despite constituting approximately two thirds of all breast cancers, the luminal A and B tumours are poorly classified at both clinical and molecular levels. There are contradictory reports on the nature of these subtypes: some define them as intrinsic entities, others as a continuum. With the aim of addressing these uncertainties and identifying molecular signatures of patients at risk, we conducted a comprehensive transcriptomic and genomic analysis of 2,425 luminal breast cancer samples. Our results indicate that the separation between the molecular luminal A and B subtypes-per definition-is not associated with intrinsic characteristics evident in the differentiation between other subtypes. Moreover, t-SNE and MST-kNN clustering approaches based on 10,000 probes, associated with luminal tumour initiation and/or development, revealed the close connections between luminal A and B tumours, with no evidence of a clear boundary between them. Thus, we considered all luminal tumours as a single heterogeneous group for analysis purposes. We first stratified luminal tumours into two distinct groups by their HER2 gene cluster co-expression: HER2-amplified luminal and ordinary-luminal. The former group is associated with distinct transcriptomic and genomic profiles, and poor prognosis; it comprises approximately 8% of all luminal cases. For the remaining ordinary-luminal tumours we further identified the molecular signature correlated with disease outcomes, exhibiting an approximately continuous gene expression range from low to high risk. Thus, we employed four virtual quantiles to segregate the groups of patients. The clinico-pathological characteristics and ratios of genomic aberrations are concordant with the variations in gene expression profiles, hinting at a progressive staging. The comparison with the current separation into luminal A and B subtypes revealed a substantially improved survival stratification. Concluding, we suggest a review of the definition of luminal A and B subtypes. A proposition for a revisited delineation is provided in this study.

No MeSH data available.


Related in: MedlinePlus