Limits...
An integrative genomic and epigenomic approach for the study of transcriptional regulation.

Figueroa ME, Reimers M, Thompson RF, Ye K, Li Y, Selzer RR, Fridriksson J, Paietta E, Wiernik P, Green RD, Greally JM, Melnick A - PLoS ONE (2008)

Bottom Line: We predicted that integration of different genome-wide epigenetic regulatory marks along with gene expression levels would provide greater power in capturing biological differences between leukemia subtypes.We found that DNA methylation and H3K9 acetylation distinguished these leukemias of distinct cell lineage, as expected, but that an integrative analysis combining the information from each platform revealed hundreds of additional differentially expressed genes that were missed by gene expression arrays alone.Integrative epigenomic studies are thus feasible using clinical samples and provide superior detection of aberrant transcriptional programming than single-platform microarray studies.

View Article: PubMed Central - PubMed

Affiliation: Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, Bronx, New York, United States of America.

ABSTRACT
The molecular heterogeneity of acute leukemias and other tumors constitutes a major obstacle towards understanding disease pathogenesis and developing new targeted-therapies. Aberrant gene regulation is a hallmark of cancer and plays a central role in determining tumor phenotype. We predicted that integration of different genome-wide epigenetic regulatory marks along with gene expression levels would provide greater power in capturing biological differences between leukemia subtypes. Gene expression, cytosine methylation and histone H3 lysine 9 (H3K9) acetylation were measured using high-density oligonucleotide microarrays in primary human acute myeloid leukemia (AML) and acute lymphocytic leukemia (ALL) specimens. We found that DNA methylation and H3K9 acetylation distinguished these leukemias of distinct cell lineage, as expected, but that an integrative analysis combining the information from each platform revealed hundreds of additional differentially expressed genes that were missed by gene expression arrays alone. This integrated analysis also enhanced the detection and statistical significance of biological pathways dysregulated in AML and ALL. Integrative epigenomic studies are thus feasible using clinical samples and provide superior detection of aberrant transcriptional programming than single-platform microarray studies.

Show MeSH

Related in: MedlinePlus

Promoter DNA methylation shows genome-wide inverse correlation with gene expression and H3K9 acetylation: Panel A:Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and gene expression, showing a positive correlation between the two measures for the majority of genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower gene expression). Panel B: Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and H3K9 acetylation, showing a positive correlation between the two measures for many of the genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower promoter H3K9 acetylation). Panel C: Graphical representation of the data from all three platforms for one of the cases (AML.2) as custom tracks in the UCSC genome browser[40]. Four representative genes are shown here to illustrate the correlation between the three platforms. H3K9 acetylation data (in blue) is represented as the ratio of the signal between the H3K9 acetyl channel and the input channel; DNA methylation (in red) is represented as log(HpaII/MspI), so that a negative deflection corresponds to a methylated HpaII fragment while a positive one corresponds to a hypomethylated fragment; finally, gene expression data (in green) is represented as median-centered log2 of RMA-normalized intensities.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2266992&req=5

pone-0001882-g004: Promoter DNA methylation shows genome-wide inverse correlation with gene expression and H3K9 acetylation: Panel A:Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and gene expression, showing a positive correlation between the two measures for the majority of genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower gene expression). Panel B: Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and H3K9 acetylation, showing a positive correlation between the two measures for many of the genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower promoter H3K9 acetylation). Panel C: Graphical representation of the data from all three platforms for one of the cases (AML.2) as custom tracks in the UCSC genome browser[40]. Four representative genes are shown here to illustrate the correlation between the three platforms. H3K9 acetylation data (in blue) is represented as the ratio of the signal between the H3K9 acetyl channel and the input channel; DNA methylation (in red) is represented as log(HpaII/MspI), so that a negative deflection corresponds to a methylated HpaII fragment while a positive one corresponds to a hypomethylated fragment; finally, gene expression data (in green) is represented as median-centered log2 of RMA-normalized intensities.

Mentions: Promoter DNA methylation is generally believed to be associated with gene silencing, although it is not clear whether this can be generalized in a whole genome analysis. Therefore, we determined the relationship between gene expression and DNA methylation in our set of leukemia patient samples. As before, high SNR genes (>2.5) were selected for this analysis. Correlation of expression and DNA methylation among these genes revealed a bimodal distribution where two-thirds of genes displayed a strong positive correlation between expression and the log HpaII/MspI ratio (which translates biologically into an inverse correlation between gene expression and DNA methylation levels) where the peak value was r = 1.0. The remaining one-third of genes showed a weak negative correlation (peak at r = −0.5) (Figure 4A). The presence of a strong correlation between gene expression and log HpaII/MspI ratio was still clearly detected even at SNR cutoffs of 1.3. These data indicate that for a majority of promoters, DNA methylation is strongly associated with gene silencing. A similar bimodal distribution was detected for the correlations between DNA methylation and H3K9 acetylation. The largest peak of the correlations was found at r = 1.0. A second and less defined peak that represented a smaller population of genes was found with a slightly negative correlation (around r = −0.5) (Figure 4B). While the first peak is to be expected, representing silenced genes with methylated promoters lacking the H3K9ac mark associated with active chromatin, the second peak may be explained by a number of different factors (see discussion). Overall, the results demonstrate that for most genes, the presence of high levels of methylation corresponds with repression, while high levels of acetylation correspond to activation, as exemplified in Figure 4C.


An integrative genomic and epigenomic approach for the study of transcriptional regulation.

Figueroa ME, Reimers M, Thompson RF, Ye K, Li Y, Selzer RR, Fridriksson J, Paietta E, Wiernik P, Green RD, Greally JM, Melnick A - PLoS ONE (2008)

Promoter DNA methylation shows genome-wide inverse correlation with gene expression and H3K9 acetylation: Panel A:Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and gene expression, showing a positive correlation between the two measures for the majority of genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower gene expression). Panel B: Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and H3K9 acetylation, showing a positive correlation between the two measures for many of the genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower promoter H3K9 acetylation). Panel C: Graphical representation of the data from all three platforms for one of the cases (AML.2) as custom tracks in the UCSC genome browser[40]. Four representative genes are shown here to illustrate the correlation between the three platforms. H3K9 acetylation data (in blue) is represented as the ratio of the signal between the H3K9 acetyl channel and the input channel; DNA methylation (in red) is represented as log(HpaII/MspI), so that a negative deflection corresponds to a methylated HpaII fragment while a positive one corresponds to a hypomethylated fragment; finally, gene expression data (in green) is represented as median-centered log2 of RMA-normalized intensities.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2266992&req=5

pone-0001882-g004: Promoter DNA methylation shows genome-wide inverse correlation with gene expression and H3K9 acetylation: Panel A:Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and gene expression, showing a positive correlation between the two measures for the majority of genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower gene expression). Panel B: Smoothed histogram of gene-by-gene correlations between log(HpaII/MspI) values and H3K9 acetylation, showing a positive correlation between the two measures for many of the genes, which translates into a negative biological correlation (i.e. higher promoter methylation correlates with lower promoter H3K9 acetylation). Panel C: Graphical representation of the data from all three platforms for one of the cases (AML.2) as custom tracks in the UCSC genome browser[40]. Four representative genes are shown here to illustrate the correlation between the three platforms. H3K9 acetylation data (in blue) is represented as the ratio of the signal between the H3K9 acetyl channel and the input channel; DNA methylation (in red) is represented as log(HpaII/MspI), so that a negative deflection corresponds to a methylated HpaII fragment while a positive one corresponds to a hypomethylated fragment; finally, gene expression data (in green) is represented as median-centered log2 of RMA-normalized intensities.
Mentions: Promoter DNA methylation is generally believed to be associated with gene silencing, although it is not clear whether this can be generalized in a whole genome analysis. Therefore, we determined the relationship between gene expression and DNA methylation in our set of leukemia patient samples. As before, high SNR genes (>2.5) were selected for this analysis. Correlation of expression and DNA methylation among these genes revealed a bimodal distribution where two-thirds of genes displayed a strong positive correlation between expression and the log HpaII/MspI ratio (which translates biologically into an inverse correlation between gene expression and DNA methylation levels) where the peak value was r = 1.0. The remaining one-third of genes showed a weak negative correlation (peak at r = −0.5) (Figure 4A). The presence of a strong correlation between gene expression and log HpaII/MspI ratio was still clearly detected even at SNR cutoffs of 1.3. These data indicate that for a majority of promoters, DNA methylation is strongly associated with gene silencing. A similar bimodal distribution was detected for the correlations between DNA methylation and H3K9 acetylation. The largest peak of the correlations was found at r = 1.0. A second and less defined peak that represented a smaller population of genes was found with a slightly negative correlation (around r = −0.5) (Figure 4B). While the first peak is to be expected, representing silenced genes with methylated promoters lacking the H3K9ac mark associated with active chromatin, the second peak may be explained by a number of different factors (see discussion). Overall, the results demonstrate that for most genes, the presence of high levels of methylation corresponds with repression, while high levels of acetylation correspond to activation, as exemplified in Figure 4C.

Bottom Line: We predicted that integration of different genome-wide epigenetic regulatory marks along with gene expression levels would provide greater power in capturing biological differences between leukemia subtypes.We found that DNA methylation and H3K9 acetylation distinguished these leukemias of distinct cell lineage, as expected, but that an integrative analysis combining the information from each platform revealed hundreds of additional differentially expressed genes that were missed by gene expression arrays alone.Integrative epigenomic studies are thus feasible using clinical samples and provide superior detection of aberrant transcriptional programming than single-platform microarray studies.

View Article: PubMed Central - PubMed

Affiliation: Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, Bronx, New York, United States of America.

ABSTRACT
The molecular heterogeneity of acute leukemias and other tumors constitutes a major obstacle towards understanding disease pathogenesis and developing new targeted-therapies. Aberrant gene regulation is a hallmark of cancer and plays a central role in determining tumor phenotype. We predicted that integration of different genome-wide epigenetic regulatory marks along with gene expression levels would provide greater power in capturing biological differences between leukemia subtypes. Gene expression, cytosine methylation and histone H3 lysine 9 (H3K9) acetylation were measured using high-density oligonucleotide microarrays in primary human acute myeloid leukemia (AML) and acute lymphocytic leukemia (ALL) specimens. We found that DNA methylation and H3K9 acetylation distinguished these leukemias of distinct cell lineage, as expected, but that an integrative analysis combining the information from each platform revealed hundreds of additional differentially expressed genes that were missed by gene expression arrays alone. This integrated analysis also enhanced the detection and statistical significance of biological pathways dysregulated in AML and ALL. Integrative epigenomic studies are thus feasible using clinical samples and provide superior detection of aberrant transcriptional programming than single-platform microarray studies.

Show MeSH
Related in: MedlinePlus