BackgroundEpigenome-wide association studies (EWAS) based on human brain samples allow a deep and direct understanding of epigenetic dysregulation in Alzheimer’s disease (AD). However, strong variation of cell-type proportions across brain tissue samples represents a significant source of data noise. Here, we report the first EWAS based on sorted neuronal and non-neuronal (mostly glia) nuclei from postmortem human brain tissues.ResultsWe show that cell sorting strongly enhances the robust detection of disease-related DNA methylation changes even in a relatively small cohort. We identify numerous genes with cell-type-specific methylation signatures and document differential methylation dynamics associated with aging specifically in neurons such as CLU, SYNJ2 and NCOR2 or in glia RAI1,CXXC5 and INPP5A. Further, we found neuron or glia-specific associations with AD Braak stage progression at genes such as MCF2L, ANK1, MAP2, LRRC8B, STK32C and S100B. A comparison of our study with previous tissue-based EWAS validates multiple AD-associated DNA methylation signals and additionally specifies their origin to neuron, e.g., HOXA3 or glia (ANK1). In a meta-analysis, we reveal two novel previously unrecognized methylation changes at the key AD risk genes APP and ADAM17.ConclusionsOur data highlight the complex interplay between disease, age and cell-type-specific methylation changes in AD risk genes thus offering new perspectives for the validation and interpretation of large EWAS results.Electronic supplementary materialThe online version of this article (10.1186/s13072-018-0211-3) contains supplementary material, which is available to authorized users.

Stable inheritance of DNA methylation is critical for maintaining differentiated phenotypes in multicellular organisms. We have recently identified dual mono-ubiquitylation of histone H3 (H3Ub2) by UHRF1 as an essential mechanism to recruit DNMT1 to chromatin. Here, we show that PCNA-associated factor 15 (PAF15) undergoes UHRF1-dependent dual monoubiquitylation (PAF15Ub2) on chromatin in a DNA replication-coupled manner. This event will, in turn, recruit DNMT1. During early S-phase, UHRF1 preferentially ubiquitylates PAF15, whereas H3Ub2 predominates during late S-phase. H3Ub2 is enhanced under PAF15 compromised conditions, suggesting that H3Ub2 serves as a backup for PAF15Ub2. In mouse ES cells, loss of PAF15Ub2 results in DNA hypomethylation at early replicating domains. Together, our results suggest that there are two distinct mechanisms underlying replication timing-dependent recruitment of DNMT1 through PAF15Ub2 and H3Ub2, both of which are prerequisite for high fidelity DNA methylation inheritance.

SummaryThe RING E3 ubiquitin ligase UHRF1 controls DNA methylation through its ability to target the maintenance DNA methyltransferase DNMT1 to newly replicated chromatin. DNMT1 recruitment relies on ubiquitylation of histone H3 by UHRF1; however, how UHRF1 deposits ubiquitin onto the histone is unknown. Here, we demonstrate that the ubiquitin-like domain (UBL) of UHRF1 is essential for RING-mediated H3 ubiquitylation. Using chemical crosslinking and mass spectrometry, biochemical assays, and recombinant chromatin substrates, we show that the UBL participates in structural rearrangements of UHRF1 upon binding to chromatin and the E2 ubiquitin conjugating enzyme UbcH5a/UBE2D1. Similar to ubiquitin, the UBL exerts its effects through a hydrophobic patch that contacts a regulatory surface on the “backside” of the E2 to stabilize the E2-E3-chromatin complex. Our analysis of the enzymatic mechanism of UHRF1 uncovers an unexpected function of the UBL domain and defines a new role for this domain in DNMT1-dependent inheritance of DNA methylation.

Transcription is the first step on the complex way of converting genetic information into proteins. In recent years, various studies have contributed to a vast knowledge about eukaryotic transcription -on both the global nuclear and the molecular level. However, knowledge pertaining to an intermediate level of transcriptional organization, that is how individual expressed genes are spatially organized, is surprisingly