Methylation of AR locus does not always reflect X chromosome inactivation state

Abstract

Clonality can be established by a lack of mosaicism in a female because of random inactivation of either the maternal or paternal X chromosome early in embryogenesis. The methylation status of CpG sites close to the trinucleotide repeats in exon 1 of the human androgen receptor (AR) X chromosome gene assay (HUMARA) has been used to determine clonality. This HUMARA at times indicated clonal hematopoiesis in healthy elderly women, thus precluding its applicability. We used a clonality assay based on quantitative expression of polymorphic X chromosome genes (qTCA) and found no evidence of clonal hematopoiesis in healthy nonanemic elderly persons. We found instances of discordance between HUMARA results and those obtained by pyrosequencing and qTCA methods, as well as by directly quantifying AR gene expression. To determine the basis of this discrepancy we examined the methylation pattern of the AR locus subject to HUMARA. Notably, we found the extent of DNA methylation to be highly variable at the AR gene in granulocytes of persons with discordant results and also in erythroid burst-forming unit colonies but not in those with clonal hematopoiesis. These data provide the molecular basis of incomplete correlation with the pattern of DNA methylation of this X chromosome AR gene locus.

Figure 1

Linear regression analysis. Linear regression analysis of the correlation between allelic expression ratios for qTCA (y-axis) and pyrosequencing (qPyro; x-axis) for 3 markers MPP1 (top left), FHL1 (top right), and IDS (bottom left). Data points were generated according to data in Table 2. For each gene, a statistically significant correlation (P ≤ .0001) was observed when expression based on qTCA was compared with that based on quantitative pyrosequencing.

Figure 2

Comparison of allelic expression ratios generated by qTCA, DNA-HUMARA, and RNA-HUMARA. (A) qTCA compared with DNA-HUMARA. The values for qTCA are the mean of all polymorphic genes for each informative subject shown in Table 2. DNA-HUMARA results are from Table 2. The Pearson correlation coefficient was used to calculate the linear relation (indicated by the line) between qTCA and DNA-HUMARA, with the shaded area above and below the line indicating the confidence intervals. (B) qTCA compared with RNA-HUMARA. The values for qTCA are the mean of all polymorphic genes for each informative subject shown in Table 2. RNA-HUMARA results are from Table 2. The Pearson correlation coefficient was used to calculate the linear relation (line) between qTCA and DNA-HUMARA, with the shaded areas above and below the line indicating the confidence intervals. (C) DNA-HUMARA compared with RNA-HUMARA. The values for DNA-HUMARA and RNA-HUMARA are from Table 2. The Pearson correlation coefficient was used to calculate the linear relation (line) between DNA-HUMARA and RNA-HUMARA, with the shaded area above and below the line indicating the confidence intervals.

Figure 3

Patterns of CpG methylation of the human androgen receptor gene HUMARA by bisulfite sequencing. (A top) Diagram of the 5′ region of the human androgen receptor gene showing the locations of core promoter region, transcription start site (+1), translation start site (ATG), and methylation-sensitive enzyme sites for HpaII. Specific CpG dinucleotides in exon1 region are highlighted in the sequence of PCR product. (down) The sequencing results of the PCR products of the exon 1 HUMARA gene in native selected granulocytes. Total 20-29 clones were analyzed from healthy controls YC25, YC17, YC3, and YC18, respectively. The white dots represent the unmethylated CpGs, whereas black dots represent the methylated CpGs. Short allele (A) and long allele (B) were determined according to the length of the AR gene exon 1 microsatellite. (B) Patterns of CpG methylation of the human androgen receptor gene HUMARA by bisulfite sequencing in individual BFU-Es of subject YC17. The sequencing results of the PCR products of the exon 1 HUMARA gene. Total 6-10 clones were analyzed for BFU-E 1, BFU-E 2, BFU-E 3, BFU-E 4, BFU-E 5, and BFU-E 6, respectively. The white dots represent the unmethylated CpGs, whereas the black dots represent the methylated CpGs. Short allele (A) and long allele (B) were determined according to the length of the AR gene exon 1 microsatellite.

Figure 4

Distribution of expected counts given 8 independent equiprobable trials, each with a probability of 0.5. This figure show the probability that somatic tissue in an adult will have a particular allele ratio, assuming stochastic inactivation of either the maternal or paternal X chromosome at the 8-cell stage of embryogenesis and no positive or negative selection over time based on X chromosome gene expression. The x-axis indicates the number of either maternal or paternal alleles that are active. For example, 8 could indicate that all of the maternal alleles are active. In that case, 8 paternal alleles would be inactive, and the allele expression ration would be 100:0. Similarly, 7 would indicate that 7 maternal and 1 paternal gene are active. In that case, the allele ration would be 87.5:12.5. The probability of each ratio is as follows: 0:8, 0.004; 1:7, 0.031; 2:6, 0.109; 3:5, 0.219; and 4:4, 0.273.