Limits...
High-specificity bioinformatics framework for epigenomic profiling of discordant twins reveals specific and shared markers for ACPA and ACPA-positive rheumatoid arthritis

View Article: PubMed Central - PubMed

ABSTRACT

Background: Twin studies are powerful models to elucidate epigenetic modifications resulting from gene–environment interactions. Yet, commonly a limited number of clinical twin samples are available, leading to an underpowered situation afflicted with false positives and hampered by low sensitivity. We investigated genome-wide DNA methylation data from two small sets of monozygotic twins representing different phases during the progression of rheumatoid arthritis (RA) to find novel genes for further research.

Methods: We implemented a robust statistical methodology aimed at investigating a small number of samples to identify differential methylation utilizing the comprehensive CHARM platform with whole blood cell DNA from two sets of twin pairs discordant either for ACPA (antibodies to citrullinated protein antigens)-positive RA versus ACPA-negative healthy or for ACPA-positive healthy (a pre-RA stage) versus ACPA-negative healthy. To deconvolute cell type-dependent differential methylation, we assayed the methylation patterns of sorted cells and used computational algorithms to resolve the relative contributions of different cell types and used them as covariates.

Results: To identify methylation biomarkers, five healthy twin pairs discordant for ACPAs were profiled, revealing a single differentially methylated region (DMR). Seven twin pairs discordant for ACPA-positive RA revealed six significant DMRs. After deconvolution of cell type proportions, profiling of the healthy ACPA discordant twin-set revealed 17 genome-wide significant DMRs. When methylation profiles of ACPA-positive RA twin pairs were adjusted for cell type, the analysis disclosed one significant DMR, associated with the EXOSC1 gene. Additionally, the results from our methodology suggest a temporal connection of the protocadherine beta-14 gene to ACPA-positivity with clinical RA.

Conclusions: Our biostatistical methodology, optimized for a low-sample twin design, revealed non-genetically linked genes associated with two distinct phases of RA. Functional evidence is still lacking but the results reinforce further study of epigenetic modifications influencing the progression of RA. Our study design and methodology may prove generally useful in twin studies.

Electronic supplementary material: The online version of this article (doi:10.1186/s13073-016-0374-0) contains supplementary material, which is available to authorized users.

No MeSH data available.


Schematic representation of the analysis. TS1 and TS2 denote the tests comparing ACPA-positive healthy twin versus ACPA-negative healthy twin and ACPA-positive RA twin versus ACPA-negative healthy twin, respectively. Steps 1 and 2 denote the first step of the analysis, DMR identification in TS1 and TS2 without cell proportion adjustment. In step 3 we computed DMRs between each pair of cell types (neutrophils, CD4+ T cells, CD8+ T cells, and CD56+ NK cells) and observed that DMRs identified without cell proportion adjustment were associated with cell type, so likely to be associated with changes in cell proportion. For this reason in step 4 we estimated cell proportion in each sample by adapting the method of Houseman et al. [31] (see the “Cell proportion estimation” section in the “Methods”). In step 5 and 6 we used cell proportion estimations as covariates in the identification of DMRs in TS1 and TS2. A DMR is considered statistically significant if it is significant both globally (FWER <0.10) and locally (permuted p-value <0.10); details are provided in the “Methods”
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC5120506&req=5

Fig2: Schematic representation of the analysis. TS1 and TS2 denote the tests comparing ACPA-positive healthy twin versus ACPA-negative healthy twin and ACPA-positive RA twin versus ACPA-negative healthy twin, respectively. Steps 1 and 2 denote the first step of the analysis, DMR identification in TS1 and TS2 without cell proportion adjustment. In step 3 we computed DMRs between each pair of cell types (neutrophils, CD4+ T cells, CD8+ T cells, and CD56+ NK cells) and observed that DMRs identified without cell proportion adjustment were associated with cell type, so likely to be associated with changes in cell proportion. For this reason in step 4 we estimated cell proportion in each sample by adapting the method of Houseman et al. [31] (see the “Cell proportion estimation” section in the “Methods”). In step 5 and 6 we used cell proportion estimations as covariates in the identification of DMRs in TS1 and TS2. A DMR is considered statistically significant if it is significant both globally (FWER <0.10) and locally (permuted p-value <0.10); details are provided in the “Methods”

Mentions: In order to identify differentially methylated regions caused by changes in cell type proportion, we repeated the statistical analysis while considering the cell type proportion of each sample (cell type deconvolution/correction). To do this, we used the strategy depicted in Fig. 2. As a first step we analyzed the methylation profile by CHARM in physically sorted CD4+ T cells, CD8+ T cells, neutrophils, and CD56+ NK cells from five healthy individuals. Those profiles allowed us to identify DMRs characteristic of each of these cell types (see the “Sorted cell analysis” in the “Methods”). By combining those DMRs we adapted an existing and validated computational procedure [2, 31] to generate robust estimations of the cell proportions in each sample (see the “Cell proportion estimation” section in the “Methods”). As a second step we applied the same DMR-finder methodology used for the “non cell-corrected analysis” (see the “DMR candidate identification” section in the “Methods”) but this time we included as covariates the estimated cell proportions.Fig. 2


High-specificity bioinformatics framework for epigenomic profiling of discordant twins reveals specific and shared markers for ACPA and ACPA-positive rheumatoid arthritis
Schematic representation of the analysis. TS1 and TS2 denote the tests comparing ACPA-positive healthy twin versus ACPA-negative healthy twin and ACPA-positive RA twin versus ACPA-negative healthy twin, respectively. Steps 1 and 2 denote the first step of the analysis, DMR identification in TS1 and TS2 without cell proportion adjustment. In step 3 we computed DMRs between each pair of cell types (neutrophils, CD4+ T cells, CD8+ T cells, and CD56+ NK cells) and observed that DMRs identified without cell proportion adjustment were associated with cell type, so likely to be associated with changes in cell proportion. For this reason in step 4 we estimated cell proportion in each sample by adapting the method of Houseman et al. [31] (see the “Cell proportion estimation” section in the “Methods”). In step 5 and 6 we used cell proportion estimations as covariates in the identification of DMRs in TS1 and TS2. A DMR is considered statistically significant if it is significant both globally (FWER <0.10) and locally (permuted p-value <0.10); details are provided in the “Methods”
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC5120506&req=5

Fig2: Schematic representation of the analysis. TS1 and TS2 denote the tests comparing ACPA-positive healthy twin versus ACPA-negative healthy twin and ACPA-positive RA twin versus ACPA-negative healthy twin, respectively. Steps 1 and 2 denote the first step of the analysis, DMR identification in TS1 and TS2 without cell proportion adjustment. In step 3 we computed DMRs between each pair of cell types (neutrophils, CD4+ T cells, CD8+ T cells, and CD56+ NK cells) and observed that DMRs identified without cell proportion adjustment were associated with cell type, so likely to be associated with changes in cell proportion. For this reason in step 4 we estimated cell proportion in each sample by adapting the method of Houseman et al. [31] (see the “Cell proportion estimation” section in the “Methods”). In step 5 and 6 we used cell proportion estimations as covariates in the identification of DMRs in TS1 and TS2. A DMR is considered statistically significant if it is significant both globally (FWER <0.10) and locally (permuted p-value <0.10); details are provided in the “Methods”
Mentions: In order to identify differentially methylated regions caused by changes in cell type proportion, we repeated the statistical analysis while considering the cell type proportion of each sample (cell type deconvolution/correction). To do this, we used the strategy depicted in Fig. 2. As a first step we analyzed the methylation profile by CHARM in physically sorted CD4+ T cells, CD8+ T cells, neutrophils, and CD56+ NK cells from five healthy individuals. Those profiles allowed us to identify DMRs characteristic of each of these cell types (see the “Sorted cell analysis” in the “Methods”). By combining those DMRs we adapted an existing and validated computational procedure [2, 31] to generate robust estimations of the cell proportions in each sample (see the “Cell proportion estimation” section in the “Methods”). As a second step we applied the same DMR-finder methodology used for the “non cell-corrected analysis” (see the “DMR candidate identification” section in the “Methods”) but this time we included as covariates the estimated cell proportions.Fig. 2

View Article: PubMed Central - PubMed

ABSTRACT

Background: Twin studies are powerful models to elucidate epigenetic modifications resulting from gene&ndash;environment interactions. Yet, commonly a limited number of clinical twin samples are available, leading to an underpowered situation afflicted with false positives and hampered by low sensitivity. We investigated genome-wide DNA methylation data from two small sets of monozygotic twins representing different phases during the progression of rheumatoid arthritis (RA) to find novel genes for further research.

Methods: We implemented a robust statistical methodology aimed at investigating a small number of samples to identify differential methylation utilizing the comprehensive CHARM platform with whole blood cell DNA from two sets of twin pairs discordant either for ACPA (antibodies to citrullinated protein antigens)-positive RA versus ACPA-negative healthy or for ACPA-positive healthy (a pre-RA stage) versus ACPA-negative healthy. To deconvolute cell type-dependent differential methylation, we assayed the methylation patterns of sorted cells and used computational algorithms to resolve the relative contributions of different cell types and used them as covariates.

Results: To identify methylation biomarkers, five healthy twin pairs discordant for ACPAs were profiled, revealing a single differentially methylated region (DMR). Seven twin pairs discordant for ACPA-positive RA revealed six significant DMRs. After deconvolution of cell type proportions, profiling of the healthy ACPA discordant twin-set revealed 17 genome-wide significant DMRs. When methylation profiles of ACPA-positive RA twin pairs were adjusted for cell type, the analysis disclosed one significant DMR, associated with the EXOSC1 gene. Additionally, the results from our methodology suggest a temporal connection of the protocadherine beta-14 gene to ACPA-positivity with clinical RA.

Conclusions: Our biostatistical methodology, optimized for a low-sample twin design, revealed non-genetically linked genes associated with two distinct phases of RA. Functional evidence is still lacking but the results reinforce further study of epigenetic modifications influencing the progression of RA. Our study design and methodology may prove generally useful in twin studies.

Electronic supplementary material: The online version of this article (doi:10.1186/s13073-016-0374-0) contains supplementary material, which is available to authorized users.

No MeSH data available.