Histone Methylation Dynamics and Gene Regulation Occur through the Sensing of One-Carbon Metabolism

Abstract

S-adenosylmethionine (SAM) and S-adenosylhomocysteine (SAH) link one-carbon metabolism to methylation status. However, it is unknown whether regulation of SAM and SAH by nutrient availability can be directly sensed to alter the kinetics of key histone methylation marks. We provide evidence that the status of methionine metabolism is sufficient to determine levels of histone methylation by modulating SAM and SAH. This dynamic interaction led to rapid changes in H3K4me3, altered gene transcription, provided feedback regulation to one-carbon metabolism, and could be fully recovered upon restoration of methionine. Modulation of methionine in diet led to changes in metabolism and histone methylation in the liver. In humans, methionine variability in fasting serum was commensurate with concentrations needed for these dynamics and could be partly explained by diet. Together these findings demonstrate that flux through methionine metabolism and the sensing of methionine availability may allow direct communication to the chromatin state in cells.

Keywords: chromatin biology; epigenetics; metabolomics; one carbon metabolism.

Figure 1. Methionine metabolism alters histone methylation status

a.) Metabolomics profile of methionine restricted HCT116 cells. Log₂ fold change versus –log₁₀(P value) (Student’s t-test two-tailed, n = 3). b.) Effects on methionine restriction on one carbon cycle metabolism. c.) Effects of methionine restriction on histone methylation measured by immunoblotting. Band intensities are normalized to total H3 levels and relative intensity was calculated compared to control cells. d.) Effects of deprivation of other amino acids on histone methylation. Data are normalized to control media. e.) Effects of methionine restriction on histone methylation across a panel of cell lines. f.) Relative concentration of methionine cycle metabolites in cells cultured in different concentrations of methionine. g.) Concentration dependent effects of methionine restriction on histone methylation. Data are normalized to the 100 μm condition. h.) Cell proliferation of HCT116 cells for differing levels of methionine. All error bars are computed from standard error of measurement (n=3) and all quantitation is normalized to the total H3 and the fold change between experimental and control groups are reported.

Figure 2. Histone methylation is dynamically regulated by methionine metabolism

a.) Global metabolic dynamics in response to methionine restriction. b.) Pathway analysis of dynamics shown in (a). c.) Dynamics of methionine. d.) Dynamics of SAM. e.) Dynamics of SAH. f.) Dynamics of the SAM/SAH ratio. g.) Dynamics of histone methylation. h.) Quantitation of results in (g). Integrated intensities are normalized to total H3 and the fold change represents differences compared to 24 hours. All error bars are computed from standard error of measurement (n=3).

Figure 3. Histone methylation and methionine cycle dynamics are reversible

a.) Experimental setup of methionine restriction and recovery. b.) Effects of methionine restriction and recovery on global metabolism. c.) Pathway analysis of methionine recovery. d.) Effect of methionine restriction and recovery on methionine cycle metabolism. e.) Effect of methionine resctiction and recovery on methionine salvage and transsulfuration. f.) Response of histone methylation from methionine restriction and recovery. Integrated intensities are normalized to total H3 and fold change was calculated. All error bars are computed from the standard error of measurement (n=3).

Figure 4. Methionine restrictions decreases H3K4me3 and alters gene expression

a.) Change in H3K4me3 distribution around the transcription start site (TSS) after 24hr methionine restriction b.) Genome wide distribution of H3K4me3 peaks in +MET (red) and −MET (blue) conditions. c.) Enlargement of chromosome 17 d.) Changes in H3K4me3 ChIP-seq signal with corresponding changes in gene expression from RNA-seq for colon cancer genes. e.) Enzymes with decreased H3K4me3 and gene expression essential in one-carbon metabolism and related pathways are highlighted in red. f.) H3K4me3 ChIP-seq and RNA-seq signals for enzymes in (e) before and after methionine restriction.

Figure 5. Alterations in methionine metabolism can be sustained by diet in vivo

a.) Methionine restriction in mice with representative histology in methionine-restricted and control mice. Images are at 40X with Hematoxylin and Eosin staining. b.) Fold change of serum amino acids in −MET compared to +MET, measured my LC-MS (*denotes p<0.001). c.) Methionine cycle metabolites in liver of methionine-restricted mice. d.) Methionine cycle metabolites in plasma of methionine-restricted mice e.) Histone methylation in liver of methionine-restricted mice. Integrated intensities were normalized to total H3 f.) Quantitation of results histone methylation from n=8 mice in each cohort. P value is obtained from a student’s t-test equal variance. g–h) Correlation of histone methylation and methionine cycle metabolites in liver and plasma.

Figure 6. Methionine and metabolic variation in human subjects

a.) Measurement of serum methionine and clinical and dietary variables in human subjects. b.) (left) Hierarchical clustering of the distance matrix diet variables. (right) k-means clustering of subjects and diet variables (N=24). c.) Absolute concentrations of amino acids in fasting serum in 38 human subjects. d.) Coefficients of variation of amino acids in fasting serum. e. Correlation of methionine in the serum with methylated serum metabolites, N,N,N- trimethyllysine and sarcosine.f.) Correlation of methionine concentrations with dietary variables obtained from habitual diet records. g.) Correlation of methionine concentrations with fasting serum metabolite levels.

Figure 7. A computational model identifies determinants of methionine variability

a.) Overview of variable selection for the computational model. b.) Predicted versus measured methionine levels in human subjects. c.) Regression coefficients. Error bars are obtained from maximum likelihood estimates. d.) Schematic of dietary factors that contribute to each modeled variable. e.) Results from variance partitioning.