Supplementary Information for:
Independent genomewide screens identify the tumor suppressor
VTRNA2-1 as a human epiallele responsive to periconceptional
environment
Matt J. Silver1, Noah J. Kessler1, Branwen J. Hennig, Paula Dominguez-Salas, Eleonora
Laritsky, Maria S. Baker, Cristian Coarfa, Hector Hernandez-Vargas, Jovita M. Castelino,
Michael N. Routledge, Yun Yun Gong, Zdenko Herceg, Yong Sun Lee, Kwanbok Lee, Sophie
E. Moore, Anthony J. Fulford, Andrew M. Prentice2, Robert A. Waterland2
1
2
These authors contributed equally.
Correspondence to: waterland@bcm.edu or Andrew.Prentice@lshtm.ac.uk
This PDF file includes:
Figures S1 to S11
1
Supplementary Figure 1: Density plot of average methylation (by 200 bp bin) in hair follicle
(HF) in individual 2 (C02) vs. individual 1 (C01), for all 4.5M 200 bp bins that were informative in
both samples. Genomewide, DNA methylation in HF is highly correlated across the two
individuals (R2=0.930).
2
Supplementary Figure 2: Lymphoblastoid cell lines from 132 HapMap individuals display
interindividual variation at VTRNA2-1 similar to what is observed in primary tissues. (Analysis of
Illumina 450k data of Zhang et al (Hum Mol Genet 2014.) (a) At each of the 10 probes
corresponding to the VTRNA2-1 DMR, a wide range of individual methylation values is
observed. (b) Distribution of individual average methylation across all 10 probes confirms
regional interindividual variation, with a strong bimodal distribution.
3
Supplementary Figure 3: Validation of 450k beta values and optimization of analysis pipeline.
Comparison of % methylation estimates by pyrosequencing vs. 450k array beta values at five
ME loci, for six different QC pipelines (see Supplementary Methods for details). Correlations are
generated from 554 data points for each method, corresponding to loci measured in multiple
individuals for which comparable 450k and pyrosequencing data are available. (a) Pipeline 1A,
Pearson R (P)=0.89; Spearman R (S)=0.91. (b) Pipeline 1B, P=0.89; S=0.91. (c) Pipeline 2,
P=0.86; S=0.84. (d) Pipeline 3, P=0.86; S=0.85. (e) Pipeline 4, P=0.92; S=0.94. (f) Pipeline 5,
P=0.92; S=0.94.
4
0.5
estimated wbc proportion
0.4
0.3
SEASON
rainy
dry
0.2
0.1
0.0
CD8T
CD4T
NK
Bcell
cell type
Mono
Gran
Supplementary Figure 4: Estimated white blood cell (WBC) fraction vs. WBC type. Estimates
were derived from our 450k methylation data using the method of Jaffe & Irizarry (Genome Biol
2014). The results suggest subtle SoC effects on WBC composition, particularly for CD4T cells
and NK cells. (CD8T: CD8+ T cells, CD4T: CD4+ T cells, NK: natural killer cells, Bcell: B cells,
Mono: monocytes, Gran: granulocytes.)
5
Supplementary Figure 5: Plots of methylation estimates (unadjusted beta values) of other top
ranking DMRs associated with SoC in the 450k analysis. (a) PAX8 (ranked 2nd). (b) PRDM9
(ranked 3rd). (c) ZFP57 (ranked 6th).
6
Supplementary Figure 6: Illumina 450k data on peripheral blood leukocytes and colonic
mucosa (Harris et al., Epigenetics 2013) confirms systemic interindividual variation in DNA
methylation across the entire VTRNA2-1 imprinted DMR. (a) Among all 10 individuals studied,
at the 15 probes in the VTRNA2-1 SoC DMR, methylation in colonic mucosa is strongly
correlated with that in peripheral blood leukocytes. (b) In leukocytes (GSE32148), two
individuals (subjects 3 and 8) are hypomethylated across the entire VTRNA2-1 imprinted DMR
(box). (c) The same two individuals are hypomethylated in colonic mucosa (GSE32146).
7
Supplementary Figure 7: Standard curve of the VTRNA2-1 pyrosequencing assay.
8
9007225117 R06C01
0.004
Sample Failure Rate
failRate
0.003
0.002
0.001
0.000
6164621132_R01C01
6164621132_R01C02
6164621132_R02C01
6164621132_R02C02
6164621132_R03C01
6164621132_R03C02
6164621132_R04C01
6164621132_R04C02
6164621132_R05C01
6164621132_R05C02
6164621132_R06C01
6164621132_R06C02
6164621142_R01C01
6164621142_R01C02
6164621142_R02C01
6164621142_R02C02
6164621142_R03C01
6164621142_R03C02
6164621142_R04C01
6164621142_R04C02
6164621142_R05C01
6164621142_R05C02
6164621142_R06C01
6164621142_R06C02
7471147005_R01C01
7471147005_R01C02
7471147005_R02C01
7471147005_R02C02
7471147005_R03C01
7471147005_R03C02
7471147005_R04C01
7471147005_R04C02
7471147005_R05C01
7471147005_R05C02
7471147005_R06C01
7471147005_R06C02
7507875165_R03C02
7507875165_R04C02
7507875165_R05C02
7973219166_R01C01
7973219166_R01C02
7973219166_R02C01
7973219166_R02C02
7973219166_R03C01
7973219166_R03C02
7973219166_R04C01
7973219166_R04C02
7973219166_R05C01
7973219166_R05C02
7973219166_R06C01
7973219166_R06C02
7973219167_R01C01
7973219167_R01C02
7973219167_R02C01
7973219167_R02C02
7973219167_R03C01
7973219167_R03C02
7973219167_R04C01
7973219167_R04C02
7973219167_R05C01
7973219167_R05C02
7973219167_R06C01
7973219167_R06C02
7990895010_R01C01
7990895010_R01C02
7990895010_R02C01
7990895010_R03C01
7990895010_R04C01
7990895010_R05C01
7990895010_R06C01
7990895016_R01C02
7990895016_R02C02
7990895016_R03C02
7990895016_R04C02
7990895016_R05C02
7990895016_R06C01
7990895016_R06C02
7990895030_R01C01
7990895030_R01C02
7990895030_R02C01
7990895030_R02C02
7990895030_R03C01
7990895030_R03C02
7990895030_R04C01
7990895030_R04C02
7990895030_R05C01
7990895030_R05C02
7990895030_R06C01
7990895030_R06C02
7990895042_R01C01
7990895042_R01C02
7990895042_R02C01
7990895042_R02C02
7990895042_R03C01
7990895042_R04C01
7990895042_R05C01
7990895042_R06C01
9007225089_R01C02
9007225089_R02C01
9007225089_R02C02
9007225089_R03C01
9007225089_R03C02
9007225089_R04C01
9007225089_R04C02
9007225089_R05C01
9007225089_R05C02
9007225089_R06C01
9007225089_R06C02
9007225117_R01C01
9007225117_R01C02
9007225117_R02C01
9007225117_R02C02
9007225117_R03C01
9007225117_R03C02
9007225117_R04C01
9007225117_R04C02
9007225117_R05C01
9007225117_R05C02
9007225117_R06C0
9007225117_R06C
9007225164_R01C
9007225164_R02
9007225164_R0
9007225164_R
sampleID
Supplementary Figure 8: Proportion of probes with detection p-value > 0.01 for each sample.
9
9007225117 R06C01
500
PC2
250
0
−250
−1000
−500
0
PC1
9007225117 R06C01
PC3
250
0
−250
−250
0
PC2
250
500
Supplementary Figure 9: PCA of array-wide methylation. Potential outlier is also sample with
highest number of probes with detection p-value > 0.01 (see Supplementary Fig. 7). This probe
was removed from subsequent analyses.
10
Supplementary Figure 10: Bland-Altman plots illustrate cross-replicate differences in
methylation across the range of possible beta estimates, for each of the 3 replicates in the
study. In these plots, probe-wise mean estimated beta values (across both replicates, x-axis)
are plotted against differences in estimated betas (y-axis). 95% confidence intervals for the
mean difference across all probes (mean +/- 2SD) are represented by the horizontal dotted
lines.
11
Supplementary Figure 11: Comparison of WBC estimates across two technical replicates.
Spearman R=0.99, p<10-8.
12