ELF-MF exposure affects the robustness of epigenetic programming during granulopoiesis

Abstract

Extremely-low-frequency magnetic fields (ELF-MF) have been classified as "possibly carcinogenic" to humans on the grounds of an epidemiological association of ELF-MF exposure with an increased risk of childhood leukaemia. Yet, underlying mechanisms have remained obscure. Genome instability seems an unlikely reason as the energy transmitted by ELF-MF is too low to damage DNA and induce cancer-promoting mutations. ELF-MF, however, may perturb the epigenetic code of genomes, which is well-known to be sensitive to environmental conditions and generally deranged in cancers, including leukaemia. We examined the potential of ELF-MF to influence key epigenetic modifications in leukaemic Jurkat cells and in human CD34+ haematopoietic stem cells undergoing in vitro differentiation into the neutrophilic lineage. During granulopoiesis, sensitive genome-wide profiling of multiple replicate experiments did not reveal any statistically significant, ELF-MF-dependent alterations in the patterns of active (H3K4me2) and repressive (H3K27me3) histone marks nor in DNA methylation. However, ELF-MF exposure showed consistent effects on the reproducibility of these histone and DNA modification profiles (replicate variability), which appear to be of a stochastic nature but show preferences for the genomic context. The data indicate that ELF-MF exposure stabilizes active chromatin, particularly during the transition from a repressive to an active state during cell differentiation.

Figure 1. ELF-MF exposure does not alter global patterns of histone modifications in Jurkat cells.

Jurkat cells were ELF-MF- (50 Hz sinus, 1 mT, 5′ on/10′ off), sham-exposed or treated with 10 nM Trichostatin A (TSA) for 72 h. Profiles of H3K4me2 and H3K27me3 modification were generated by ChIP-sequencing of two replicate (representing pools of three biological replicates). (a) Principal component analysis of H3K4me2 and H3K27me3 ChIP-seq data from non-exposed (t0), ELF-MF- and sham-exposed or TSA-treated cells, comparing read counts in 500 bp genomic tiles. (b,c) Comparison of H3K4me2 and H3K27me3 read counts within 500 bp genomic tiles between TSA-treated and ELF-MF-exposed (b) or ELF-MF- and sham-exposed (c) cells. Shown are differences in relative enrichments as log2-fold change (FC) (x-axis) against the false discovery rate (FDR)-adjusted P value (likelihood ratio test) (y-axis). Statistically significant tiles (FC > ±0.6, FDR-adjusted P < 0.05) are highlighted in red (H3K4me2) or blue (H3K27me3). (d) Exemplary profiles of H3K4me2 and H3K27me3 modifications at the RPTOR and CEP170 locus with FCs between sham- and ELF-MF-exposed cells of 1.8 and 2.1, respectively, but not reaching statistical significance. (e) H3K4me2 and H3K27me3 profiles at the B3GALT6 and KMT2C locus, significantly different in ELF-MF- and TSA-treated cells.

Figure 2. Patterning of histone modifications during granulopoiesis is not affected by ELF-MF.

Human CD34+ cord blood cells were differentiated in vitro to neutrophilic progenitors under ELF-MF (50 Hz powerline signal, 1 mT, 5′ on/10′ off) or sham exposure for 5 days. (a,b) Cell cycle profiles and apoptosis were assessed by flow cytometry before and at days 4, 5 and 6 of neutrophilic differentiation and statistically analyzed by χ² test on each replicate (****P < 0.001) and pairwise comparison by Student’s t-test. Shown are the mean percentage of cells in the G1, S, and G2 phase of the cell cycle (a) and of alive, early apoptotic, late apoptotic and dead/necrotic cells (b). Error bars; SEM of n ≥ 2 and n = 3 biological replicates of cell cycle and apoptosis analysis, respectively. (c–g) H3K4me2 and H3K27me3 profiles of CD34+ cells (t0) and neutrophilic progenitors (t5) were obtained by ChIP-seq and two replicates were statistically analyzed. (c) Principal component analysis of ChIP-seq data. (d,e) Comparison of ELF-MF- and sham-exposed neutrophilic progenitors (d) or of CD34+ and combined neutrophilic progenitor cells (e). Shown are differences in relative enrichments of ChIP-seq reads within 500 and 1,000 bp genomic tiles as log2-fold change (FC) (x-axis) against the false discovery rate (FDR)-adjusted P value (likelihood ratio test with variable dispersion) (y-axis). Statistically significant (FC > ±0.6, FDR-adjusted P < 0.05) tiles differentially occupied by H3K4me2 and H3K27me3 are highlighted in red and blue, respectively. (f,g) Exemplary profiles of H3K4me2 and H3K27me3 marks at genomic loci, identified by a 500 bp tile (black box) with significant differences in neutrophilic progenitors and CD34+ cells (f) or more than 2-fold enrichment between ELF-MF- and sham-exposed progenitor cells (g).

Figure 3. The DNA methylation pattern does not change upon ELF-MF exposure.

DNA methylation of CD34+ human cord blood cells (t0) and neutrophilic progenitors (t5), in vitro differentiated under ELF-MF (50 Hz powerline signal, 1 mT, 5′ on/10′ off), sham or no exposure condition for five days, was analyzed by Illumina Infinium HumanMethylation 450 array. (a) Principal component analysis of all samples (M-values). (b,c) Differences in relative DNA methylation levels shown as log2-fold change (FC) (x-axis) are plotted against the false discovery rate (FDR)-adjusted P value (calculated by moderated t-statistics) (y-axis) for the comparison of (b) neutrophilic progenitors (combined progenitors of all exposure conditions) and CD34+ cells and (c) ELF-MF- and sham-exposed day 5 neutrophilic progenitors. Statistically significantly (FDR-adjusted P < 0.05) hypomethylated (FC < −0.6) and hypermethylated (FC > 0.6) CpGs are indicated in green and dark red, respectively.

Figure 4. ELF-MF exposure impacts the variability of the epigenetic landscape.

The squared coefficient of variation (CV²) of ChIP-seq read counts within 500 bp genomic tiles was determined based on the two replicate datasets for H3K4me2 and H3K27me3, generated from non-exposed (t0), ELF-MF-exposed (50 Hz sinus, 1 mT, 5′ on/10′ off, 72 h), sham-exposed or Trichostatin A-treated (10 nM, 72 h) Jurkat cells, and from neutrophilic progenitors after five days of in vitro differentiation under ELF-MF (50 Hz powerline, 1 mT, 5′ on/10′ off) or sham exposure. Linear comparison of CV² values of ELF-MF- (x-axis) and sham-exposed (y-axis) Jurkat cells (a) and neutrophilic progenitor cells (b) for H3K4me2 and H3K27me3 marks. Variability measures of the two replicate ChIP-seq datasets for H3K4me2 and H3K27me3 in Jurkat cells (c) and in neutrophilic progenitors (d). (e) Variance in DNA methylation levels (M value) of the three biological replicates for CD34+ human umbilical cord blood cells (t0), ELF-MF- and sham-exposed neutrophilic progenitors after five days of in vitro differentiation. (c–e) Box-and-whisker plots illustrate median (lines) and mean (black circles) CV² values with interquartile ranges (boxes), 1.5× interquartile ranges (whiskers) and outliers. P values of the Wilcoxon rank sum test and the number of genomic tiles are indicated above and below, respectively.

Figure 5. Genomic context-dependent effect of ELF-MF exposure on the robustness of epigenetic modifications.

The squared coefficient of variation (CV²) of reads in 500 bp tiles of two H3K4me2 and H3K27me3 ChIP-seq replicates and the variance of DNA methylation in the three replicates was analyzed for promoters (±1,000 bp of TSS), exons, introns, intergenic regions (UCSC hg19) or bivalent domains (H3K4me2 and H3K27me3 co-occupancy) in our data set. Box-and-whisker plots illustrate median (lines) and mean (black circles) CV² values with interquartile ranges (boxes), 1.5× interquartile ranges (whiskers) and outliers. P values of the Wilcoxon rank sum test and the number of genomic tiles are indicated above and below, respectively. (a,b) Assessment of the variability of epigenetic modifications in ELF-MF- (50 Hz sinus, 1 mT, 5′ on/10′ off, 72 h) and sham-exposed or TSA-treated (10 nM, 72 h) Jurkat cells with respect to the indicated genomic features. (c,d) As in (a,b) but with data from neutrophilic progenitors after five days of in vitro differentiation under ELF-MF (50 Hz powerline, 1 mT, 5′ on/10′ off) or sham exposure. (e,f) Comparison of the mean variabilities of ChIP-seq reads of H3K4me2 and H3K27me3 tiles (CV² values on y-axis) between ELF-MF- and sham-exposed samples of neutrophilic progenitors, plotted as a function of distance to the nearest bivalent domain or intron. (g) Variability of three replicates of DNA methylation assessed with respect to genomic features, comparing ELF-MF- and sham-exposed neutrophilic progenitors.

Figure 6. The chromatin state defines the stability of epigenetic features under ELF-MF exposure.

Variability of epigenetic marks in sham- and ELF-MF-exposed neutrophilic progenitors was intersected with tiles/sites that change significantly (FDR-adjusted P < 0.05) epigenetic modifications during differentiation of CD34+ cells to day five progenitors. (a) Median (lines) and mean (black circles) CV² values with interquartile ranges (boxes), 1.5× interquartile ranges (whiskers) and outliers, categorized according to tiles enriched in H3K4me2, H3K27me3 or both that either significantly change (up, log2-fold change >0.6; down, log2-fold change <−0.6) or remain stable (no) during differentiation. (b) As in (a) but for differential enrichments of histone modifications between ELF-MF- and sham-exposed samples. (c,d) Variability of DNA methylation of sham- and ELF-MF-exposed neutrophilic progenitors, intersected with CpGs significantly changing methylation (FDR-adjusted P < 0.05; hypo: log2-fold change <−0.6; hyper: log2-fold change >0.6) or not (no) during neutrophilic differentiation. Shown are median (lines) and mean (black circles) with interquartile ranges (boxes), 1.5× interquartile ranges (whiskers) and outliers of variance (c) and log2-fold changes between ELF-MF- and sham-exposed samples (d). (a,c) P values above indicate the statistical significance level by the Wilcoxon rank sum test and the number of included genomic tiles or CpG sites is shown below.