Limits...
Characterization of age signatures of DNA methylation in normal and cancer tissues from multiple studies.

Kim J, Kim K, Kim H, Yoon G, Lee K - BMC Genomics (2014)

Bottom Line: Genes related to the normal signature were enriched for aging-related gene ontology terms including metabolic processes, immune system processes, and cell proliferation.The related gene products of the normal signature had more than the average number of interacting partners in a protein interaction network and had a tendency not to interact directly with each other.The genomic sequences of the normal signature were well conserved and the age-associated DNAm levels could satisfactorily predict the chronological ages of tissues regardless of tissue type.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Informatics, Ajou University School of Medicine, Suwon 443-380, South Korea. kiylee@ajou.ac.kr.

ABSTRACT

Background: DNA methylation (DNAm) levels can be used to predict the chronological age of tissues; however, the characteristics of DNAm age signatures in normal and cancer tissues are not well studied using multiple studies.

Results: We studied approximately 4000 normal and cancer samples with multiple tissue types from diverse studies, and using linear and nonlinear regression models identified reliable tissue type-invariant DNAm age signatures. A normal signature comprising 127 CpG loci was highly enriched on the X chromosome. Age-hypermethylated loci were enriched for guanine-and-cytosine-rich regions in CpG islands (CGIs), whereas age-hypomethylated loci were enriched for adenine-and-thymine-rich regions in non-CGIs. However, the cancer signature comprised only 26 age-hypomethylated loci, none on the X chromosome, and with no overlap with the normal signature. Genes related to the normal signature were enriched for aging-related gene ontology terms including metabolic processes, immune system processes, and cell proliferation. The related gene products of the normal signature had more than the average number of interacting partners in a protein interaction network and had a tendency not to interact directly with each other. The genomic sequences of the normal signature were well conserved and the age-associated DNAm levels could satisfactorily predict the chronological ages of tissues regardless of tissue type. Interestingly, the age-associated DNAm increases or decreases of the normal signature were aberrantly accelerated in cancer samples.

Conclusion: These tissue type-invariant DNAm age signatures in normal and cancer can be used to address important questions in developmental biology and cancer research.

Show MeSH

Related in: MedlinePlus

Characteristics of age-associated DNA methylation signature. (A) Age prediction using the age-associated normal DNAm signature. Age was predicted with the normal signature using a multivariate linear regression after using a genetic algorithm to identify a feasible set of loci. (B) Degrees of overlap with age-associated DNAm signatures identified in previous studies. Overlap percentages were calculated by the common numbers divided by the smaller number of total loci in either study. The studies with only normal samples are orange; other studies including disease samples are gray. (C, D) The fractions of hyper- (green) or hypomethylated (blue) CpG loci in the age-associated signatures in normal (C) or cancer (D) according to genomic regions. The number on each bar indicates the count of the corresponding loci. P-value was calculated by a chi-square test. (E, F) The hyper- (E) or hypomethylation (F) patterns according to age group of normal age-associated DNA loci in CGI or non-CGI. Blue or green dotted lines show the linear regressions of median values of individual age groups using only hypo- (blue) or hypermethylated (green) loci, respectively. Numbers below are the counts of loci considered for the corresponding cases. (G, H) Nucleotide compositions of the sequences surrounding the hypo- (G) or hypermethylated loci (H) of normal age-associated DNAm signature. –log (P-value) of the y axis was calculated by random selection tests representing overrepresentation for each base at each location of the surrounding CpG.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4289351&req=5

Fig5: Characteristics of age-associated DNA methylation signature. (A) Age prediction using the age-associated normal DNAm signature. Age was predicted with the normal signature using a multivariate linear regression after using a genetic algorithm to identify a feasible set of loci. (B) Degrees of overlap with age-associated DNAm signatures identified in previous studies. Overlap percentages were calculated by the common numbers divided by the smaller number of total loci in either study. The studies with only normal samples are orange; other studies including disease samples are gray. (C, D) The fractions of hyper- (green) or hypomethylated (blue) CpG loci in the age-associated signatures in normal (C) or cancer (D) according to genomic regions. The number on each bar indicates the count of the corresponding loci. P-value was calculated by a chi-square test. (E, F) The hyper- (E) or hypomethylation (F) patterns according to age group of normal age-associated DNA loci in CGI or non-CGI. Blue or green dotted lines show the linear regressions of median values of individual age groups using only hypo- (blue) or hypermethylated (green) loci, respectively. Numbers below are the counts of loci considered for the corresponding cases. (G, H) Nucleotide compositions of the sequences surrounding the hypo- (G) or hypermethylated loci (H) of normal age-associated DNAm signature. –log (P-value) of the y axis was calculated by random selection tests representing overrepresentation for each base at each location of the surrounding CpG.

Mentions: We built a feasible age-prediction model to see whether the normal 127-site signature could be used as a tissue-invariant age predictor. We applied a multiple linear regression model after identifying a feasible subset of the signature using a genetic algorithm (Methods). The selected age-prediction model was composed of 20 CpG loci of the signature (see “Predicted age” column in Additional file 1: Table S2). The correlation between the actual ages of the combined normal samples and their predicted ages using the model was highly significant (R = 0.91, P = 0.002 from 10,000 random selection tests using all loci in the platform; Figure 5A), which indicates that the DNAm levels of the age-associated signature sites can be used to predict the age of tissues, regardless of tissue type. We next compared the age-associated normal signature with those identified in previous studies (Additional file 1: Table S3). Most previous studies identified age-associated loci using a FDR threshold in a linear model. Thus, we compared the loci resulting from only linear regression (FDR < 0.01) and found that 430 age-associated CpG loci were age-associated in the integrated normal samples. For instance, a recent study using the Illumina Human 450 K platform and a linear regression model identified 137993 CpG loci associated with age in blood cells of 421 healthy subjects aged from 14 to 94 years [11]. Of these 137993 loci, the 6696 CpG loci present on the Illumina 27 K overlapped 73% with our 430 age-associated loci. Another study by Day et al. [13] found that 4747 CpG loci correlated with age in four tissue types, including brain samples, using a linear regression method, and the degree of overlap with our loci was 47%. Notably, we observed higher degrees of overlap of CpGs with previous studies that used only normal samples than with other studies that included diseased samples (Figure 5B and Additional file 1: Table S3). Sixteen of our 127 age-associated loci were not identified in the previous studies. Interestingly, 13 of these 16 loci were located on the X chromosome (see “Unique CpG” column in Additional file 1: Table S2).Figure 5


Characterization of age signatures of DNA methylation in normal and cancer tissues from multiple studies.

Kim J, Kim K, Kim H, Yoon G, Lee K - BMC Genomics (2014)

Characteristics of age-associated DNA methylation signature. (A) Age prediction using the age-associated normal DNAm signature. Age was predicted with the normal signature using a multivariate linear regression after using a genetic algorithm to identify a feasible set of loci. (B) Degrees of overlap with age-associated DNAm signatures identified in previous studies. Overlap percentages were calculated by the common numbers divided by the smaller number of total loci in either study. The studies with only normal samples are orange; other studies including disease samples are gray. (C, D) The fractions of hyper- (green) or hypomethylated (blue) CpG loci in the age-associated signatures in normal (C) or cancer (D) according to genomic regions. The number on each bar indicates the count of the corresponding loci. P-value was calculated by a chi-square test. (E, F) The hyper- (E) or hypomethylation (F) patterns according to age group of normal age-associated DNA loci in CGI or non-CGI. Blue or green dotted lines show the linear regressions of median values of individual age groups using only hypo- (blue) or hypermethylated (green) loci, respectively. Numbers below are the counts of loci considered for the corresponding cases. (G, H) Nucleotide compositions of the sequences surrounding the hypo- (G) or hypermethylated loci (H) of normal age-associated DNAm signature. –log (P-value) of the y axis was calculated by random selection tests representing overrepresentation for each base at each location of the surrounding CpG.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4289351&req=5

Fig5: Characteristics of age-associated DNA methylation signature. (A) Age prediction using the age-associated normal DNAm signature. Age was predicted with the normal signature using a multivariate linear regression after using a genetic algorithm to identify a feasible set of loci. (B) Degrees of overlap with age-associated DNAm signatures identified in previous studies. Overlap percentages were calculated by the common numbers divided by the smaller number of total loci in either study. The studies with only normal samples are orange; other studies including disease samples are gray. (C, D) The fractions of hyper- (green) or hypomethylated (blue) CpG loci in the age-associated signatures in normal (C) or cancer (D) according to genomic regions. The number on each bar indicates the count of the corresponding loci. P-value was calculated by a chi-square test. (E, F) The hyper- (E) or hypomethylation (F) patterns according to age group of normal age-associated DNA loci in CGI or non-CGI. Blue or green dotted lines show the linear regressions of median values of individual age groups using only hypo- (blue) or hypermethylated (green) loci, respectively. Numbers below are the counts of loci considered for the corresponding cases. (G, H) Nucleotide compositions of the sequences surrounding the hypo- (G) or hypermethylated loci (H) of normal age-associated DNAm signature. –log (P-value) of the y axis was calculated by random selection tests representing overrepresentation for each base at each location of the surrounding CpG.
Mentions: We built a feasible age-prediction model to see whether the normal 127-site signature could be used as a tissue-invariant age predictor. We applied a multiple linear regression model after identifying a feasible subset of the signature using a genetic algorithm (Methods). The selected age-prediction model was composed of 20 CpG loci of the signature (see “Predicted age” column in Additional file 1: Table S2). The correlation between the actual ages of the combined normal samples and their predicted ages using the model was highly significant (R = 0.91, P = 0.002 from 10,000 random selection tests using all loci in the platform; Figure 5A), which indicates that the DNAm levels of the age-associated signature sites can be used to predict the age of tissues, regardless of tissue type. We next compared the age-associated normal signature with those identified in previous studies (Additional file 1: Table S3). Most previous studies identified age-associated loci using a FDR threshold in a linear model. Thus, we compared the loci resulting from only linear regression (FDR < 0.01) and found that 430 age-associated CpG loci were age-associated in the integrated normal samples. For instance, a recent study using the Illumina Human 450 K platform and a linear regression model identified 137993 CpG loci associated with age in blood cells of 421 healthy subjects aged from 14 to 94 years [11]. Of these 137993 loci, the 6696 CpG loci present on the Illumina 27 K overlapped 73% with our 430 age-associated loci. Another study by Day et al. [13] found that 4747 CpG loci correlated with age in four tissue types, including brain samples, using a linear regression method, and the degree of overlap with our loci was 47%. Notably, we observed higher degrees of overlap of CpGs with previous studies that used only normal samples than with other studies that included diseased samples (Figure 5B and Additional file 1: Table S3). Sixteen of our 127 age-associated loci were not identified in the previous studies. Interestingly, 13 of these 16 loci were located on the X chromosome (see “Unique CpG” column in Additional file 1: Table S2).Figure 5

Bottom Line: Genes related to the normal signature were enriched for aging-related gene ontology terms including metabolic processes, immune system processes, and cell proliferation.The related gene products of the normal signature had more than the average number of interacting partners in a protein interaction network and had a tendency not to interact directly with each other.The genomic sequences of the normal signature were well conserved and the age-associated DNAm levels could satisfactorily predict the chronological ages of tissues regardless of tissue type.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Informatics, Ajou University School of Medicine, Suwon 443-380, South Korea. kiylee@ajou.ac.kr.

ABSTRACT

Background: DNA methylation (DNAm) levels can be used to predict the chronological age of tissues; however, the characteristics of DNAm age signatures in normal and cancer tissues are not well studied using multiple studies.

Results: We studied approximately 4000 normal and cancer samples with multiple tissue types from diverse studies, and using linear and nonlinear regression models identified reliable tissue type-invariant DNAm age signatures. A normal signature comprising 127 CpG loci was highly enriched on the X chromosome. Age-hypermethylated loci were enriched for guanine-and-cytosine-rich regions in CpG islands (CGIs), whereas age-hypomethylated loci were enriched for adenine-and-thymine-rich regions in non-CGIs. However, the cancer signature comprised only 26 age-hypomethylated loci, none on the X chromosome, and with no overlap with the normal signature. Genes related to the normal signature were enriched for aging-related gene ontology terms including metabolic processes, immune system processes, and cell proliferation. The related gene products of the normal signature had more than the average number of interacting partners in a protein interaction network and had a tendency not to interact directly with each other. The genomic sequences of the normal signature were well conserved and the age-associated DNAm levels could satisfactorily predict the chronological ages of tissues regardless of tissue type. Interestingly, the age-associated DNAm increases or decreases of the normal signature were aberrantly accelerated in cancer samples.

Conclusion: These tissue type-invariant DNAm age signatures in normal and cancer can be used to address important questions in developmental biology and cancer research.

Show MeSH
Related in: MedlinePlus