2. Unique Genes Showing Differential Methylation

*Methyltransferase-like 1* (*METTL1*) was significantly and consistently hypomethylated in vegans in both analyses of promoter methylation (62,398 CpG sites distributed among 9131 genes) and overall methylation of all genes present on the array and in our analytical set (313,161 CpG sites distributed among 18,627 genes), this with either the inclusion of SmartSVA (Tables 4 and 5) or the exclusion of surrogate variables from models (Supplementary Tables S3 and S4). Other genes that were significantly hypomethylated in the promoter regions in analyses both with and without surrogate variables included *ribosomal protein L38* (*RPL38*), *snurportin 1* (*SNUPN*), FLVCR heme transporter 2 (*FLVCR2*), and *CKLF-like MARVEL transmembrane domain containing 7* (*CMTM7*). Two genes, *METTL1* and *nei-like DNA glycosylase 2* (*NEIL2*), were significantly hypomethylated in CpG islands within the promoter, in analyses with and without SmartSVA (Table 5 and Supplementary Table S4). Besides *METTL1*, genes showing differential methylation in analyses of both promoter methylation and overall methylation of all genes included *RPL38* (SmartSVA analysis) (Tables 4 and 5), and *FLVCR2* (analysis excluding SmartSVA) (Supplementary Tables S3 and S4). Differential methylation of *D-dopachrome tautomerase-like* (*DDTL*) and *aryl hydrocarbon receptor nuclear translocator 2* (*ARNT2*) was observed in analyses of genes defined by methylation of the gene body, as well as of all genes (SmartSVA analysis, Table 4), and the same was true for *LIM homeobox protein 3* (*LHX3*) in models excluding surrogate variables (Supplementary Table S3). *Glutathione S-transferase theta pseudogene 1* (*GSTTP1*), detected in analysis of the open sea region showed the most marked hypomethylation in vegans (fold change of 0.72 and 0.76 in models with and without SmartSVA, respectively), whereas NACHT and WD repeat domain containing 2 (NWD2), identified in analysis of methylation of the south shore region, showed the greatest increase in gene methylation (fold change, 1.12) (Table 5 and Supplementary Table S4).



Number of actual differentially methylated genes determined using linear regression with SmartSVA followed by adapted Storey et al. [32] permutation approach to adjust for false discovery. 2 Individual CpGs may have been represented in more than one region when determining gene methylation of a given region. 3 Number of genes estimated to show non-null differences in methylation. 4 Fold change represents ratio of the mean methylation of vegans to that of non-vegetarians for differentially methylated (hypomethylated or hypermethylated) genes in a given region. This is averaged across significant genes.

1



hypothetical

 null data.


**Table 4.** Differential methylation of genic/intergenic regions associated with individual genes at FDR < 0.05 (based on SmartSVA method) 1.


**Table 4.** *Cont.*

<sup>1</sup> Number of differentially methylated CpG sites associated with each gene shown in brackets.

**Table 5.** Differential methylation of island-related and promoter regions associated with individual genes at FDR < 0.05 (based on SmartSVA method) 1.



#### **Table 5.** *Cont.*

<sup>1</sup> Number of differentially methylated CpG sites associated with each gene shown in parentheses.
