BACKGROUND: Rewriting of the epigenome has risen as a promising alternative to gene editing for precision medicine. In nature, epigenetic silencing can result in complete attenuation of target gene expression over multiple mitotic divisions. However, persistent repression has been difficult to achieve in a predictable manner using targeted systems. RESULTS: Here, we report that persistent epigenetic memory required both a DNA methyltransferase (DNMT3A-dCas9) and a histone methyltransferase (Ezh2-dCas9 or KRAB-dCas9). We demonstrate that the histone methyltransferase requirement can be locus specific. Co-targeting Ezh2-dCas9, but not KRAB-dCas9, with DNMT3A-dCas9 and DNMT3L induced long-term HER2 repression over at least 50 days (approximately 57 cell divisions) and triggered an epigenetic switch to a heterochromatic environment. An increase in H3K27 trimethylation and DNA methylation was stably maintained and accompanied by a sustained loss of H3K27 acetylation. Interestingly, substitution of Ezh2-dCas9 with KRAB-dCas9 enabled long-term repression at some target genes (e.g., SNURF) but not at HER2, at which H3K9me3 and DNA methylation were transiently acquired and subsequently lost. Off-target DNA hypermethylation occurred at many individual CpG sites but rarely at multiple CpGs in a single promoter, consistent with no detectable effect on transcription at the off-target loci tested. Conversely, robust hypermethylation was observed at HER2. We further demonstrated that Ezh2-dCas9 required full-length DNMT3L for maximal activity and that co-targeting DNMT3L was sufficient for persistent repression by Ezh2-dCas9 or KRAB-dCas9. CONCLUSIONS: These data demonstrate that targeting different combinations of histone and DNA methyltransferases is required to achieve maximal repression at different loci. Fine-tuning of targeting tools is a necessity to engineer epigenetic memory at any given locus in any given cell type.