We compared RIZ1 mutant and wild type mice on a methyl-balanced diet (diet 1), versus an imbalanced diet lacking methionine and choline (diet 2). The methyl-imbalanced diet 2 (see Supplementary Table S1
) is well known to lower hepatic SAM and cause liver cancers in rodents 
. Thus, this methyl-imbalanced diet caused liver tumors and decreased survival compared with the methyl-balanced diet (). Most of the dead or moribund animals that were suitable for autopsy analysis were found to have hepatocarcinomas. In contrast, in the absence of wild type RIZ1, there was no difference in survival regardless of diet (). These RIZ1 knockout animals developed mostly hepatocarcinomas regardless of diet. Therefore, while the balanced diet 1 conferred additional survival benefits compared to the imbalanced diet 2 in mice with wild type RIZ1, it failed to do so in mice deficient in RIZ1.
Survival of RIZ1 wild type and mutant animals on diet 1 versus diet 2.
The data also shows that, consistent with previous work 
, RIZ1+/+ mice had lower mortality and tumor incidence than RIZ1−/− mice when fed methyl-balanced diet 1 (). However, when fed imbalanced diet 2, RIZ1+/+ mice showed similar mortality as RIZ1−/− mice (). Thus the tumor suppressor function of RIZ1 is dependent on a methyl-balanced diet. The similar survival rates of RIZ1-deficient and wild type animals on diet 2 () also suggests that the capacity of diet 1 to confer additional survival benefits compared to diet 2 in RIZ1 wild type () but not in RIZ1-deficient animals () is not because of the trivial reason that the RIZ1-deficient animals may be too sick in general to respond to diet 1.
To determine the effects of diet on RIZ1, we examined RIZ1 gene expression in the liver target tissue using quantitative RT-PCR and Western blot analysis. RIZ1 mRNA level was downregulated 4.2 fold after treatment with diet 2 for 2 months (Supplementary Table S2
). The downregulation was not evident at 1 month on diet 2 but became obvious at 2, 4, and 6 months (Supplementary Table S2
). The downregulation of RIZ1 by the methyl-imbalanced diet was confirmed by western blot analysis (). In contrast to RIZ1, the shorter RIZ2 protein that lacks the PR/SET domain 
was not significantly affected by diet.
Regulation of RIZ1 protein expression by diet.
We also examined other methyltransferases and related enzymes by quantitative RT-PCR and DNA microarray analysis. A total of 25 histone methyltransferases were examined, including at least one enzyme specific for each of the amino acid residues that are known to be methylated. None of these enzymes, except RIZ1, was significantly downregulated (>2 fold) at 2 months of diet 2 treatment (Supplementary Table S2
To further determine whether RIZ1 expression is sensitive to SAM levels, we used the MATA1 knockout mice model 
. These animals have lower (3 to 4 fold) hepatic SAM and also develop liver cancers. Quantitative RT-PCR analysis of animals at 4.5–5.5 months of age showed that RIZ1 expression in wild type livers (n
6) was 2.0 fold higher than in the MATA1 knockouts (n
0.03, Student's t test, 2 tailed), comparable to the fold reduction for the same aged wild type mice on diet 2 (Supplementary Table S2
We next examined whether RIZ1 target genes were regulated by diet. DNA microarray analysis of livers of RIZ1 wild type animals at 2 months of diet treatment revealed a list of 1636 genes that were affected by diet by more than 2 fold (Supplementary Table S3
). DNA microarray analysis of livers of RIZ1 knockout animals identified 97 putative RIZ1 target genes showing more than 2 fold difference between wild type and knockout (Supplementary Table S4
). Of these, 29 were also present in the list of genes regulated by diet, indicating a significant enrichment of RIZ1 target genes in the list of diet-regulated genes (29/97 versus 1636/48000, P
<0.0001, Chi squared test, 2 tailed). The genes of interest that were upregulated by both RIZ1 knockout and methyl-imbalanced diet include c-Jun, c-Fos, and Ctgf. Some of these genes might play a direct role in liver cancers (i.e., c-Jun) 
. The effect of RIZ1 knockout or diet on the expression of these genes was confirmed by quantitative RT-PCR analysis (). The results suggest that downregulation of RIZ1 by diet 2 was associated with deregulation of RIZ1 target genes.
Quantitative RT-PCR analysis of genes that were upregulated by RIZ1 deficiency or by diet 2 treatment.
Since RIZ1 expression was not significantly altered by diet 2 at 1 month treatment (Supplementary Table S2
), any expression changes of the target genes of RIZ1 at 1 month diet treatment may reflect changes in RIZ1 activity rather than in expression level. We selected seven RIZ1 target genes and determined their expression levels at either 1 month or 2 months diet treatment. As shown in , six of the seven genes were found upregulated by diet 2 at 1 month diet treatment. The data suggest that diet 2 caused deregulation of RIZ1 target genes before significantly decreasing RIZ1 expression levels.
We next used chromatin immunoprecipitation (ChIP) assay to examine changes in histone methylation on RIZ1 target genes as a result of decreased RIZ1 activity. We used a RIZ1-specific antibody ab9710 (Abcam) that does not react with RIZ2 (see Supplementary Figure S1
for a Western blot of RIZ1 by this antibody). As shown in , the c-Jun proximal promoter was bound by RIZ1 and showed higher levels of H3 lysine 9 monomethylation in RIZ1 wild type versus knockout livers. RIZ1 did not bind to the distal region of the c-Jun promoter or the promoter of Elovl3 gene that was not repressed by RIZ1. At 1 month diet 2 treatment, H3K9me1 of c-Jun promoter was decreased even though RIZ1 binding was not changed, suggesting again that RIZ1 activity was decreased even when RIZ1 expression was at normal levels ().
Regulation of histone methylation on RIZ1 target genes.
We also examined overall DNA methylation changes in response to methyl-imbalanced diet in DNA from livers of animals at 15 months of treatment. While we found that diet 2 decreased overall DNA methylation (Supplementary Figure S2
), we did not find changes in DNA methylation (either hyper or hypo methylation) in individual random CpG rich regions (Supplementary Figure S3
) using the technique of Not1-Mse1
MS-AFLP (methylation sensitive amplified fragment length polymorphism) 
. By Western blot analysis, we did not find significant changes (>2 fold) in methylation of H3K9me1, H3K9me2, H3K9me3, H4K20me1, and H4K20me2 in the liver of animals that were on diet 2 for six months (data not shown).