Dear Editor,

Dosage compensation of the male X-chromosomal genes in Drosophila results from acetylation of histone H4 Lys16 (H4K16) along the male X-chromosome by the MSL (male-specific lethal) complex1. The MSL complex comprises five proteins (MSL1, MSL2, MSL3, MLE and MOF), as well as two non-coding RNAs (roX1 and roX2). The enzymatic activity of histone acetyltransferase (HAT) MOF (males-absent on the first) is tightly regulated by MSL1 and MSL32,3,4. MSL1 tethers MOF with the regulatory factor, the MRG domain of MSL3, through two adjacent regions at its C-terminus4,5. Only in complex with MSL1 and MSL3, MOF is capable of specifically acetylating nucleosomal H4K164,5.

In addition to the MSL complex, MOF also resides in an NSL (nonspecific lethal) complex that plays an important role in genome-wide chromatin modification and transcriptional regulation6,7. Besides MOF, the NSL complex includes six other components (NSL1, NSL2, NSL3, MCRS2, MBD-R2 and WDS). MOF is directly associated with an MSL1-like protein, NSL17. Compared with the MSL complex, although an MSL3-like factor is absent from the NSL complex, MOF and NSL1 alone are sufficient for acetylating nucleosomal H4K166,7,8. It remains unclear how the nucleosomal HAT activity of MOF is regulated in both complexes.

The two MOF-containing complexes have conserved mammalian orthologs, which are essential for genome-wide H4K16 acetylation8,9. We found that human MSL1473-520 and NSL1885-934 are the minimal MOF-binding motifs (MBM) of MSL1 and NSL1, respectively (Figure 1A and Supplementary information, Figure S1A). Sequence alignment of MSL1MBM and NSL1MBM across species reveals two highly homologous regions, indicating that MSL1 and NSL1 might interact with MOF in similar manners (Figure 1B).

Figure 1
figure 1

Structural and functional analysis of the MSL and NSL complexes. (A) Domain organization of the MSL and NSL complexes. Numerals indicate residue numbers at the boundaries of subdivisions. Protein interactions are indicated with gray-shaded areas. CC, coiled coil; NHAM, nucleosomal HAT activation motif; MBM, MOF-binding motif; CD, chromodomain; M3BM, MSL3-binding motif. (B) Sequence alignment of MSL1MBM and NSL1MBM across species. Structure-based secondary structure assignments are shown as cylinders (α helices) and arrows (β strands). Conserved residues are highlighted in green and the MOF-interacting residues are denoted with orange circles. (C) In vitro HAT assays were performed with MOFHAT and the MSL complexes composed of MOFHAT, MSL3MRG and different MSL1 fragments. (D) In vitro HAT assays were performed with the MSL complexes composed of MOFHAT, MSL1473-591 and several deletion mutants of MSL3MRG. (E) In vitro HAT assays were performed with MOFHAT in complex with a hybrid protein, MSL1473-520-(GGS)6-MSL3MRG. (F) In vitro HAT assays were performed with the NSL complexes composed of MOFHAT and different NSL1 fragments. (G) Sequence alignment of the NHAM motif of NSL1 across species. Conserved residues of NSL1NHAM and NSL1MBM are highlighted in magenta and green, respectively. Yellow circles denote the NSL1NHAMresidues that are crucial for activating the nucleosomal HAT activity of MOF. (H) In vitro HAT assays were performed with the NSL complex comprising MOFHAT and different alanine-substituted mutant NSL1850-932 fragments. (I) Overall structure of the MOFHAT-MSL1MBM complex. MOFHAT and MSL1MBM are colored in yellow and green, respectively. CoA is superposed from the MOFHAT-CoA structure (PDB: 2PQ8). (J) Detailed interactions between MOFHAT and the β-strand module of MSL1MBM. Two slightly different conformations within an asymmetric unit are shown separately. Interacting residues of MSL1MBM and MOFHAT are shown in ball-and-stick models. Hydrogen-bonding interactions are denoted with magenta dashed lines. (K) The middle loop region of MSL1MBM makes less contribution to the MOFHAT-MSL1MBM interaction. Interacting residues are shown in ball-and-stick models, and hydrogen-bonding interactions are denoted with magenta dashed lines.

In the MSL complex, the nucleosomal HAT activity requires the MRG domain of MSL3 (MSL3MRG)10. MSL3MRG shares substantial similarities with other MRG family proteins (Supplementary information, Figure S1C); MSL3MRG can be superimposed onto many other MRG domains with root-mean-square deviation values of less than 2.0 Ã…. Notwithstanding the high degree of overall structural conservation, the connecting loop regions have great variations in MSL3MRG compared with other MRG domains. For example, MSL3MRG has a long loop (∼150 residues) between helices α4 and α5, whereas helices α4 and α5 of other MRG domains are connected by a short four-to-five-residue turn (Supplementary information, Figure S1C). Deletion of this loop greatly impaired the nucleosomal HAT activity of the MSL complex (Figure 1D), suggesting that it plays a crucial role in HAT activity regulation.

MSL1 utilizes two separate regions to bind to the HAT domain of MOF (MOFHAT) and MSL3MRG; MSL1 residues 550-591 (MSL3 binding motif; MSL1M3BM) C-terminal to MSL1MBMmediate the interaction with MSL3MRG (Figure 1A)4,10. MSL1473-591 that includes both MSL1MBM and MSL1M3BM associates with MOFHAT and MSL3MRG simultaneously to form a stable ternary complex that is capable of acetylating nucleosomal substrates (Figure 1C and Supplementary information, Figure S1B). Notably, a hybrid molecule, in which MSL3MRG was directly connected to the C-terminus of MSL1MBM by an 18-amino-acid linker – (GlyGlySer)6, was sufficient to stimulate the activity of MOFHAT (Figure 1E), suggesting that in the MSL complex MSL1 likely functions as a scaffold to tether MSL3MRG and MOFHAT together for optimal enzymatic activity regulation.

Our previous studies indicated that NSL1 alone is sufficient to support the HAT activity of MOF on nucleosomal substrates8. Further mapping narrowed the necessary region of NSL1 down to a short fragment (NSL1850-932) that only contains NSL1MBM and 35 amino acids immediately N-terminal to NSL1MBM (Figure 1F). As NSL1MBM itself did not activate the nucleosomal HAT activity of MOFHAT (Figure 1F), the observed stimulatory effect of NSL1850-932 likely resulted from NSL1850-884. Hereafter, we will refer to NSL1850-884 as the nucleosomal HAT activation motif of NSL1 (NSL1NHAM) (Figure 1A and 1G). Mutagenesis analysis of NSL1NHAM revealed that substitution of the four N-terminal arginine residues with alanines completely abolished the HAT activity of the NSL complex on HeLa nucleosomes. This result is consistent with the observation that the MOFHAT–NSL1856-932 complex does not possess nucleosomal HAT activity (Figure 1F and 1H). In addition, alanine substitution of resides 864NIVI867 in NSL1NHAM also led to decreased activities (Figure 1H). Notably, both the arginine residues and 864NIVI867 are highly conserved across species (Figure 1G), consistent with their crucial roles in MOF activation.

To examine the structural basis of how the nucleosomal HAT activity of MOF is regulated in the MSL and NSL complexes, we determined the crystal structure of MOFHAT in complex with MSL1MBM at a resolution of 2.05 Ã… (Supplementary information, Table S1). The formation of the MOFHAT–MSL1MBM complex involves an extensive set of interactions and causes the burial of 1,580 Ã…2 of surface area at the interface. MSL1MBM adopts an extended conformation and surrounds the N-terminal half of MOFHAT, with both termini close to the central catalytic site of MOFHAT (Figure 1I). Comparison with the MOFHAT structure complexed with cofactor CoA (PDB: 2PQ8) suggests that the MOFHAT–MSL1MBM interaction does not interfere with MOFHAT binding to CoA (Figure 1I). MSL1MBM can be divided into three binding modules. From the N- to the C-terminus, MSL1MBM contains a β strand (residues 477-480), an extended region (residues 481-499), and an α helix (residues 500-520) (Figure 1I). Both the β-strand and α-helix modules form extensive contacts with MOFHAT. In contrast, the middle loop region makes limited contribution to the interaction (Figure 1I). This is consistent with the observation that both the length and the sequence of the loop region are less conserved than the β strand and α helix modules (Figure 1B). Of note, Drosophila NSL1MBM contains a large 168-residue loop region (Figure 1B), explaining the failure to detect MBM in dmNSL1 by bioinformatics10.

The interface between MOFHAT and the α-helix module of MSL1MBM is almost identical to that of a recently published crystal structure of the MOF-MSL1 complex10 (Supplementary information, Figure S2). However, although a similar MSL1 construct was used in both studies, the N-terminal conformation of MSL1MBM is completely absent in their structure, possibly resulting from crystal packing effect (Supplementary information, Figure S3). Our structure clearly demonstrates that the N-terminus of MSL1MBM forms an antiparallel intermolecular β-sheet with strand β2 of MOFHAT, resulting in a twisted five-stranded β sheet extending over both molecules (Figure 1J). In addition to these main-chain contacts, the intermolecular β sheet is further stabilized by electrostatic contacts, as well as van der Waals interactions (Figure 1J). On one side of the β sheet, the side chains of MSL1MBMArg479 and His481 adopt two different conformations within an asymmetric unit, but both participate in hydrogen-bonding interactions with the side chain of MOFHATGlu188. Another hydrogen bond is formed between the side chain of MOFHATAsp190 and the backbone of Ser477 from MSL1MBM (Figure 1J). In contrast, the interaction between MOFHAT and MSL1MBM is highly hydrophobic on the other side of the intermolecular β sheet. The aromatic side chain of MSL1MBMTrp478 sits snugly on a hydrophobic surface formed by Tyr187, Pro321, and Pro322 of MOFHAT, while MSL1MBMPro476 packs against the side chain of MOFHAT Trp192 (Figure 1J). Deletion mutant of the β-strand of MSL1MBM failed to form a stable MSL complex (Supplementary information, Figure S4A), underscoring the importance of the N-terminal interaction of MSL1MBM with MOF. As residues 888-892 of NSL1MBM are almost identical to residues 476-480 of MSL1MBM (Figure 1B), the same interface should also mediate the MOF-NSL1 interaction. Indeed, NSL1 with a polyalanine substitution for the N-terminal β-strand cannot form a stable complex with MOFHAT (Supplementary information, Figure S4B).

The middle extended region of MSL1MBM surrounds the zinc finger motif of MOFHAT with few direct contacts. Val483 and Pro485 of MSL1 packs on a hydrophobic patch on one side of the MOFHAT αA helix (Figure 1K). On the other side, the side chain of Glu496 forms hydrogen-bonding interactions with Ser222, Lys218 and Tyr219 of MOFHAT (Figure 1K). This region of MSL1MBM is highly variable across species in both sequence and length (Figure 1B), mostly likely serving as a flexible linker between the β-strand and α-helix modules of MSL1MBM.

From our structural and functional analyses, it is clear that NSL1NHAM and MSL3MRG are functional counterparts in the two MOF-containing complexes. Notably, the tie-like conformation of the MOF-binding motif suggests that NSL1NHAM and MSL3MRG likely are placed close to the active site of MOFHAT from opposite directions by the N-terminal β strand or the C-terminal α helix, respectively. Although the detailed mechanisms of the activation of MOFHAT on nucleosomal substrates by NSL1NHAM and MSL3MRG remain to be explored, our results presented here provide a basic structural insight into the regulation of MOF in the MSL and NSL complexes.

Accession numbers

Atomic coordinates and structure factors have been deposited with the Protein Data Bank accession code 4DNC.