Independent Mechanisms Target SMCHD1 to Trimethylated Histone H3 Lysine 9-Modified Chromatin and the Inactive X Chromosome

Abstract

The chromosomal protein SMCHD1 plays an important role in epigenetic silencing at diverse loci, including the inactive X chromosome, imprinted genes, and the facioscapulohumeral muscular dystrophy locus. Although homology with canonical SMC family proteins suggests a role in chromosome organization, the mechanisms underlying SMCHD1 function and target site selection remain poorly understood. Here we show that SMCHD1 forms an active GHKL-ATPase homodimer, contrasting with canonical SMC complexes, which exist as tripartite ring structures. Electron microscopy analysis demonstrates that SMCHD1 homodimers structurally resemble prokaryotic condensins. We further show that the principal mechanism for chromatin loading of SMCHD1 involves an LRIF1-mediated interaction with HP1γ at trimethylated histone H3 lysine 9 (H3K9me3)-modified chromatin sites on the chromosome arms. A parallel pathway accounts for chromatin loading at a minority of sites, notably the inactive X chromosome. Together, our results provide key insights into SMCHD1 function and target site selection.

FIG 1

Biochemical characterization of SMCHD1 complexes. (A) Canonical SMC complex, represented here by cohesin, composed of an SMC1-SMC3 heterodimer and an Scc1 (kleisin) subunit. The SMC hinge (HD) and Walker A/B ATPase (A/B) are denoted. (B) Schematics illustrating the domain architecture of canonical SMC proteins and SMCHD1. GHKL ATPase (GHKL), coiled-coil (cc), and bromo-adjacent homology (BAH) domains are denoted. (C) Western blots showing expression of SMCHD1-FLAG or SMCHD1 in parental (PGK12.1) and stable (clone C6) cell lines used in mass spectrometry experiments. Arrowheads indicate full-length endogenous or FLAG-tagged SMCHD1 protein (227 kDa). The four lanes shown for each blot are the input (In), flowthrough (FT), and elutions 1 and 2 (E1 and E2, respectively) from anti-FLAG IP. SMCHD1 is present only in eluents from clone C6. (D) Silver staining of IP material for mass spectrometry from control (PGK) and clone C6 nuclear extracts. (E) Western blot of fractions following size exclusion chromatography of nuclear extracts from C6 clone for FLAG-tagged SMCHD1 (top) and from PGK12.1 ESCs for endogenous SMCHD1 (bottom). (F and G) Western blots of fractions generated from sucrose gradient separation of nuclear extracts from the C6 clone (anti-FLAG) and PGK12.1 (anti-SMCHD1), for all fractions (F) and only selected fractions (G). Molecular mass standards (in kilodaltons) for panels E to G are labeled above each blot.

FIG 2

SMCHD1 forms a functional homodimer. (A) Reciprocal immunoprecipitation and Western blotting of SMCHD1-HA and SMCHD1-FLAG from nuclear extracts of HEK293T cells cotransfected with SMCHD1 plasmids. In, input (10%); IP, immunoprecipitate (15%). (B) Silver-stained IP sample from TAP-tagged mouse SMCHD1 expressed in human HEK293T cells. Both human and mouse SMCHD1 peptides were detected at similar levels. (C) Anti-FLAG Western blotting of fractions following size exclusion chromatography of rSMCHD1. (D) Anti-FLAG Western blotting of fractions from sucrose gradient sedimentation of rSMCHD1. Molecular mass standards (kilodaltons) for panels C and D are labeled above each blot. (E) Electron microscopy images of negatively stained rSMCHD1. Selected images are further magnified and presented at the right. Bars, 100 nm (left image) and 10 nm (right images). (F) Schematic of the proposed homodimeric form.

FIG 3

SMCHD1 homodimerizes through the SMC hinge domain. (A, left) Thermotoga maritima Smc hinge dimer structure (PDB accession number 1GXL), shown as light and dark blue monomers. (Right) The SWISS-MODEL-predicted SMCHD1 hinge structure based on the T. maritima hinge is also composed of two monomers, shown in gold and orange. (B) Graphical representation of the FLAG- and HA-tagged SMCHD1 derivatives used in cellular assays. (C) Anti-FLAG immunoprecipitation and Western blotting of SMCHD1 derivatives tagged with either FLAG or HA and cotransfected into HEK293T cells. In, input (10%); IP, immunoprecipitate (15%). (D) ClustalW2 alignment of the hinge domains from several canonical SMC proteins and SMCHD1. Five species are represented in this alignment: Thermotoga maritima (Tm), Bacillus subtilis (Bs), Saccharomyces cerevisiae (Sc), and Mus musculus (Mm). The conserved glycine residues mutated in the SMCHD1 hinge AAA mutant are labeled and highlighted in gray. (E) Hinge dimer ribbon diagram with conserved glycine residues mutated in the SMCHD1 hinge AAA mutant highlighted. These three glycine residues are predicted to be positioned at the interface of the two hinge domain monomers. (F) HD dimerization analysis by size exclusion chromatography. The A₂₈₀ peak for the WT (black) and AAA mutant (gray) hinges are shown (top), and the corresponding 0.5-ml fractions were run on an SDS-PAGE gel and Coomassie stained (bottom).

FIG 4

The GHKL ATPase domain hydrolyzes ATP. (A) Structures of MutL homolog 1 (MLH1) from Homo sapiens (PDB accession number 3NA3) (bronze) (left) and the predicted SMCHD1 ATPase domain (green) (right), shown with ATP (orange) and the catalytic glutamic acid (blue/red) highlighted. (B) ClustalW2 alignment of the conserved GHKL ATPase domains from the DNA mismatch repair protein MutL and DNA gyrase subunit B (GyrB) alongside SMCHD1. The conserved N, G1, and G2 motifs are outlined, and the catalytic glutamic acid that is mutated in the SMCHD1 E147A mutant constructs is highlighted in gray. (C and D) ATPase activity assays using purified WT or E147A mutant SMCHD1 ATPase domains and rSMCHD1 (C) or the WT SMCHD1 ATPase domain in the presence of various concentrations of the GHKL ATPase inhibitor radicicol (D). Statistical significance was determined by one-tailed Student's t test; the asterisk indicates a P value of 0.0017. Results show the mean calculated activities (n = 3), and the error bars show standard deviations.

FIG 5

SMCHD1 conserved domains are required for chromatin loading. (A, top) Western blot showing subcellular fractionation of endogenous SMCHD1 and Smchd1^−/− ES23 cells stably expressing WT SMCHD1-FLAG and mutant derivatives (Fig. 3B) in cytoplasmic (Cy), soluble nuclear (N), and chromatin-bound (Ch) fractions. Endogenous SMCHD1 from WT E14 ESCs is present in all fractions but predominantly in and approximately equally divided between the nuclear and chromatin-bound lanes. (Bottom) Histone H3 as a control for the chromatin-bound fraction and tubulin as a control for the cytoplasmic fraction. (B) Rabbit polyclonal anti-SMCHD1 antibody was tested by Western blotting against nuclear extracts from human HEK293T cells transfected with an SMCHD1-HA-monomeric GFP (mGFP) plasmid, wild-type mouse C127 fibroblasts, Smchd1^−/− MEFs, Smchd1^−/− ESCs stably expressing SMCHD1-FLAG (ES23⁺), and Smchd1^−/− ESCs (ES23). Bands at the predicted molecular masses are visible in extracts containing SMCHD1 but not in extracts from Smchd1^−/− cells. Asterisks indicate HA-mGFP-tagged protein. (C) Localization of SMCHD1 mutants in Smchd1^−/− mouse embryonic fibroblasts. Cells were transfected with plasmids containing WT and derivative SMCHD1 constructs and stained with DAPI to mark DNA (blue), anti-K27me3 antibody for Xi (red), and anti-HA for SMCHD1 (green). The percentage of cells showing a focus for Xi is shown at the top right of each antibody panel, representing the means from 3 replicates (n > 50 cells). Bars, 5 μm.

FIG 6

LRIF1 mediates indirect interaction of SMCHD1 with H3K9me3-modified chromatin. (A to D) Western blots of bound proteins following peptide pulldown experiments on nuclear extracts from ES23⁺ ESCs with unmodified H3, H3K4me3, H3K9me1/2/3, and H3K27me3 peptides (A and B) and on nuclear extracts from ES23 ESCs complemented with WT or mutant SMCHD1-FLAG derivatives and H3K9me3 peptide (C) and full-length recombinant SMCHD1 (rSMCHD1) and H3K9me3 peptide (D). The flowthrough (FT) lane shows that rSMCHD1 remains in solution and is not bound to the H3K9me3 peptide. (E) Quantitative RT-PCR to verify the loss of transcript in Lrif1^−/− cell lines created by CRISPR mutagenesis of exon 3/4. A schematic of the Lrif1 coding sequence is drawn with numbered exons, with three primer sets spanning intron/exon boundaries for qRT-PCR below (i to iii). Lrif1 expression in the mutant clones is shown as a percentage relative to the expression level in the wild-type parental cell line. Both the ES23⁺ and C127 Lrif1^−/− cell lines show reduced transcript levels with primer set i and a loss of transcript with primer sets ii and iii. (F) Western blots for SMCHD1-FLAG and HP1γ (control) following peptide pulldown experiments with nuclear extracts from Lrif1^−/− ES23⁺ ESCs. (G) Anti-FLAG Western blotting with mCherry-LRIF1 immunoprecipitation of nuclear extracts from HEK293T cells cotransfected with mCherry-LRIF1 and SMCHD1-FLAG derivatives. Input, 10%; IP, 30%.

FIG 7

SMCHD1 does not predominantly colocalize with HP1 proteins at pericentric regions. (A) Western blot showing the SMCHD1-FLAG distribution following subcellular fractionation of WT and Lrif1-null (clone G5) ES23⁺ cells into cytoplasmic (Cy), soluble nuclear (N), and chromatin-bound (Ch) fractions. (B) SMCHD1 staining in WT J1 XY ESCs. Pericentric heterochromatin domains are visualized with HP1α and are also visible in DNA (DAPI) panels. The percentage of cells that contain either broad nuclear staining (top) or pericentric staining (bottom) are listed at the top right corner of the SMCHD1 panels. Scoring data are the means from 3 replicates (n > 150 cells). (C and D) SMCHD1 staining in WT MEFs. (C) Staining for HP1α is enriched at pericentric heterochromatin domains but not Xi, which is enriched for SMCHD1. (D) HP1γ enrichment at pericentric foci and also throughout the chromatin arms is seen in a broad staining pattern more similar to that of SMCHD1. Note the two Xi foci in panel D. Bars, 5 μm. (E) Western blotting for HP1γ on nuclear extracts from WT and HP1γ-null (clone H5) ES23⁺ cells. Histone H3 is shown as a control. (F) Western blot showing the SMCHD1-FLAG distribution following subcellular fractionation of WT and HP1γ-null (clone H5) ES23⁺ cells.

FIG 8

SMCHD1 chromatin loading onto Xi is independent of LRIF1. (A) H3K9me2 and SMCHD1 accumulation on the inactive X chromosome in wild-type mouse MEFs. Note the two foci. (B) Localization of GFP-LRIF1 in WT and Smchd1-null MEFs. Cells are stained for DNA (DAPI) (blue), Xi (H3K27me3) (red), and GFP-LRIF1 (anti-GFP) (green). (C) Localization of SMCHD1 in WT and Lrif1-null (clone G6) C127 cells. Cells are stained for DNA (DAPI) (blue), Xi (H3K27me3) (red), and SMCHD1 (anti-SMCHD1) (green). The percentage of cells showing a focus for Xi is shown at the top right of each antibody panel, representing the means from 3 replicates (n > 50 cells [B] and n > 200 cells [C]). Bars, 5 μm. (D and E) Cell fractionation and Western blotting for endogenous SMCHD1 in WT and Lrif1-null (clone G6) C127 cells. HP1γ is shown as a control.

FIG 9

Model for SMCHD1 recruitment to chromatin. (A) Constitutive heterochromatin such as that found at pericentromeric regions is marked by H3K9me3 and all three HP1 paralogs. Recruitment of SMCHD1 (blue) to these sites by LRIF1 may be blocked by the compact and inaccessible organization of HP1 protein oligomers. (B) Other heterochromatic sites, such as telomeres and subtelomeric D4Z4 repeats, are marked by H3K9me2 and H3K9me3 and HP1γ. SMCHD1 recruitment to these sites is mediated by LRIF1 and is independent of the ATPase and BAH domains. (C) SMCHD1 recruitment to Xi in mouse, which is marked with H3K27me3 and H3K9me2, is independent of LRIF1 but requires ATPase activity and the BAH domain. The molecular mechanism behind SMCHD1 loading onto Xi is unknown but may be an active process similar to the loading of conventional SMC proteins via hinge opening.