Fig 1.
Diagram comparing the workflow for target-enriched enzymatic methyl sequencing (TEEM-Seq), whole-genome enzymatic methyl sequencing (WGEM-Seq), and reduced-representation bisulfite sequencing (RRBS).
Table 1.
Targeted genomic regions for superb starling baits.
Although we targeted putative promoter regions of all of the genes (i.e., 4 kb upstream of the transcription start site where annotated), we also targeted complete exonic regions (indicated by *) or gene body regions (based on similarity with Ensembl genome sequences, including the collared flycatcher for GNRH1 and zebra finch for MC4R, indicated by **) in a subset of the genes.
Fig 2.
Comparison of mean target region coverage (after deduplication) for target-enriched enzymatic methyl sequencing (TEEM-Seq) and whole-genome enzymatic methyl sequencing (WGEM-Seq) in (a-c) putative promoter and (d-f) exonic regions. Libraries for three of the same individual superb starlings (a and d) BB-17168, (b and e) BB-17501, and (c and f) BB-14232 were sequenced using TEEM-Seq (red bars) and WGEM-Seq (blue bars).
Fig 3.
Putative promoter target region CpG site-level percent DNA methylation distribution histograms for each sample library and scatterplots with Pearson correlation coefficients between target-enriched enzymatic methyl sequencing (TEEM-Seq) and whole-genome enzymatic methyl sequencing (WGEM-Seq) libraries from the same individual superb starlings.
CpG sites shared between TEEM-Seq and WGEM-Seq libraries at 5x coverage or above were used for comparisons across all libraries (816 CpG sites shared) with the getCorrelation function in methylKit for superb starling individuals BB-17168, BB-17501, and BB-14232. Note that S4 Fig shows the same within-individual comparisons, only with higher numbers of shared CpG sites because of within-sample comparisons only. Blue regions of the scatterplots are uncorrelated, yellow regions are highly correlated, and green regions are variably correlated. The green lines represents lowess polynomial regression fits, whereas the red lines represent linear regression fits. Across samples, the correlation coefficients for TEEM-Seq versus TEEM-Seq comparisons are slightly higher (0.95 mean) than for TEEM-Seq versus WGEM-Seq comparisons (0.94 mean), which is not surprising because WGEM-Seq libraries are much lower in mean target region coverage and therefore site-level CpG percent methylation scores could be less reliable; in any case the similarity in correlation coefficient means suggests that TEEM-Seq does not flatten differences in methylation calls as compared with WGEM-Seq. As expected within each sample, the TEEM-Seq versus WGEM-Seq library correlation coefficients are higher for each sample comparison than for WGEM-Seq versus WGEM-Seq compared across samples (see also S6 Fig for WGEM-Seq versus WGEM-Seq complete coverage).
Fig 4.
Putative promoter target region CpG site-level percent DNA methylation distribution histograms for each sample library and scatterplots with Pearson correlation coefficients between target-enriched enzymatic methyl sequencing (TEEM-Seq) and reduced-representation bisulfite sequencing (RRBS) libraries from the same individual superb starlings.
CpG sites shared between TEEM-Seq and RRBS libraries at 5x coverage or above (1,390 CpG sites shared across all seven samples) were used for paired comparisons with the getCorrelation function in methylKit for superb starling individuals BB-17532, BB-17501, BB-17455, BB-17411, BB-17168, BB-14232, and B-40881. Blue regions of the scatterplots are uncorrelated, yellow regions are highly correlated, and green regions are variably correlated. The green lines represents lowess polynomial regression fits, whereas the red lines represent linear regression fits.
Fig 5.
Pairwise comparisons of three methods for quantifying DNA methylation: Target-enriched enzymatic methyl sequencing (TEEM-Seq), whole-genome enzymatic methyl sequencing (WGEM-Seq), and reduced-representation bisulfite sequencing (RRBS).
(a and d) Libraries for three of the same individual superb starlings were sequenced using both TEEM-Seq and WGEM-Seq. (b and e) Libraries for seven of the same individual superb starlings were sequenced using both TEEM-Seq and RRBS. (c and f) Libraries for three of the same individual superb starlings were sequenced using both WGEM-Seq and RRBS. WGEM-Seq and RRBS data were compared in 20 random 2.5 kb “promoter” regions that were not included in the bait set because we expected that some RRBS and WGEM-Seq reads corresponding to target ranges might have sparse coverage in comparison to the high coverage bait alignments, and as a valid parallel to check consistency between WGEM-Seq and RRBS outside of target regions. In all cases, there were no differences in DNA methylation among quantification methods, as the absolute value mean differences in DNA methylation for promoter regions and gene bodies or random 2.5 kb “promoter” regions did not differ among pairwise comparisons (a-c) In each plot, grey violins show the distribution of CpG site absolute value mean differences in DNA methylation for (a and b) promoter regions and gene bodies or (c) random 2.5 kb “promoter” regions, black lines depict the median values, and white lines show the 1st and 3rd quartiles. Additionally, (d and e) mean DNA methylation levels for each target or (f) random region did not significantly differ. Target regions are indicated by different colors according to the legends on panels d-f. See S6–S11 Tables for more detailed information.