Histone methylation regulates diverse chromatin-templated processes, including transcription. Many transcriptional corepressor complexes contain lysine-specific demethylase 1 (LSD1) and CoREST that collaborate to demethylate mono- and dimethylated H3-K4 of nucleosomes. Here, we report the crystal structure of the LSD1-CoREST complex. LSD1-CoREST forms an elongated structure with a long stalk connecting the catalytic domain of LSD1 and the CoREST SANT2 domain. LSD1 recognizes a large segment of the H3 tail through a deep, negatively charged pocket at the active site and possibly a shallow groove on its surface. CoREST SANT2 interacts with DNA. Disruption of the SANT2-DNA interaction diminishes CoREST-dependent demethylation of nucleosomes by LSD1. The shape and dimension of LSD1-CoREST suggest its bivalent binding to nucleosomes, allowing efficient H3-K4 demethylation. This spatially separated, multivalent nucleosome binding mode may apply to other chromatin-modifying enzymes that generally contain multiple nucleosome binding modules.

Histone methylation regulates diverse chromatin-templated processes, including transcription. The recent discovery of the first histone lysine-specific demethylase (LSD1) has changed the long-held view that histone methylation is a permanent epigenetic mark. LSD1 is a flavin adenine dinucleotide (FAD)-dependent amine oxidase that demethylates histone H3 Lys4 (H3-K4). However, the mechanism by which LSD1 achieves its substrate specificity is unclear. We report the crystal structure of human LSD1 with a propargylamine-derivatized H3 peptide covalently tethered to FAD. H3 adopts three consecutive gamma-turns, enabling an ideal side chain spacing that places its N terminus into an anionic pocket and positions methyl-Lys4 near FAD for catalysis. The LSD1 active site cannot productively accommodate more than three residues on the N-terminal side of the methyllysine, explaining its H3-K4 specificity. The unusual backbone conformation of LSD1-bound H3 suggests a strategy for designing potent LSD1 inhibitors with therapeutic potential.

Histone modifications in chromatin regulate gene expression. A transcriptional co-repressor complex containing LSD1–CoREST–HDAC1 (termed LCH hereafter for simplicity) represses transcription by coordinately removing histone modifications associated with transcriptional activation. RE1-silencing transcription factor (REST) recruits LCH to the promoters of neuron-specific genes, thereby silencing their transcription in non-neuronal tissues. ZNF198 is a member of a family of MYM-type zinc finger proteins that associate with LCH. Here, we show that ZNF198-like proteins are required for the repression of E-cadherin (a gene known to be repressed by LSD1), but not REST-responsive genes. ZNF198 binds preferentially to the intact LCH ternary complex, but not its individual subunits. ZNF198- and REST-binding to the LCH complex are mutually exclusive. ZNF198 associates with chromatin independently of LCH. Furthermore, modification of HDAC1 by small ubiquitin-like modifier (SUMO) in vitro weakens its interaction with CoREST whereas sumoylation of HDAC1 stimulates its binding to ZNF198. Finally, we mapped the LCH- and HDAC1–SUMO-binding domains of ZNF198 to tandem repeats of MYM-type zinc fingers. Therefore, our results suggest that ZNF198, through its multiple protein-protein interaction interfaces, helps to maintain the intact LCH complex on specific, non-REST-responsive promoters and may also prevent SUMO-dependent dissociation of HDAC1.

Small ubiquitin-like modifier (SUMO) regulates diverse cellular processes through its reversible, covalent attachment to target proteins. Many SUMO substrates are involved in transcription and chromatin structure. Sumoylation appears to regulate the functions of target proteins by changing their subcellular localization, increasing their stability, and/or mediating their binding to other proteins. Using an in vitro expression cloning approach, we have identified 40 human SUMO1 substrates. The spectrum of human SUMO1 substrates identified in our screen suggests general roles of sumoylation in transcription, chromosome structure, and RNA processing. We have validated the sumoylation of 24 substrates in living cells. Analysis of this panel of SUMO substrates leads to the following observations. 1) Sumoylation is more efficient in vitro than in living cells. Polysumoylation occurs on several substrates in vitro. 2) SUMO isopeptidases have little substrate specificity. 3) The SUMO ligases, PIAS1 and PIASx␤, have broader substrate specificities than does PIASy. 4) Although SUMO1 and SUMO2 are equally efficiently conjugated to a given substrate in vitro, SUMO1 conjugation is more efficient in vivo. 5) Most SUMO substrates localize to the nucleus, and sumoylation does not generally affect their subcellular localization. Therefore, sumoylation appears to regulate the functions of its substrates through multiple, context-dependent mechanisms.Covalent conjugation of SUMO 1 (sumoylation) is an important post-translational modification that regulates protein functions in eukaryotes (1-7). Three isoforms of SUMO, SUMO1, SUMO2, and SUMO3, exist in mammals (2). SUMO1 consists of 101 amino acids and shares about 50% sequence identity with SUMO2/3 and 18% sequence identity with ubiquitin (2).Similar to the ubiquitin system (8, 9), conjugation of SUMO to substrate proteins is mediated by a cascade of enzymes, including SUMO isopeptidases (SENPs), SUMO-activating enzyme (a heterodimer of Aos1-Uba2), SUMO-conjugating enzyme (Ubc9), and SUMO ligases (1-7). SUMO precursor proteins are processed by a SUMO protease, exposing diglycine motifs at their C termini. In an ATP-dependent reaction, the active site cysteine of Aos1-Uba2 forms a thioester with the C terminus of SUMO. Aos1-Uba2 transfers SUMO to the Ubc9 SUMO-conjugating enzyme again as a thioester. Ubc9 then transfers SUMO to the ⑀-amino group of a lysine residue in the substrate, forming an isopeptide bond. Unlike ubiquitination, Ubc9 can catalyze efficient sumoylation of many substrates in the absence of SUMO ligases largely because of the ability of Ubc9 to directly recognize ⌿KXE (⌿, a hydrophobic residue; X, any residue) sumoylation consensus motifs on substrates (10, 11). However, SUMO ligases can increase the rates of sumoylation, especially in vivo (1-7).Several types of SUMO ligases have been identified, including the PIAS family of proteins (12), RanBP2 (13), and Pc2 (14). Interestingly, these SUMO ligases exhibit distinct patterns of subcellular localization (6). Furth...