Research ArticleBiochemistry

Crystal structure of the hinge domain of Smchd1 reveals its dimerization mode and nucleic acid–binding residues

See allHide authors and affiliations

Science Signaling  16 Jun 2020:
Vol. 13, Issue 636, eaaz5599
DOI: 10.1126/scisignal.aaz5599

Illuminating dimerization

Proteins of the SMC family are chromosomal organizers involved in sister chromatid cohesion, chromosome condensation, and DNA repair. Unlike other eukaryotic family members, SMCHD1 forms homodimers, rather than heterodimers, and has a distinct domain architecture. Dysregulation of SMCHD1 function results in a form of muscular dystrophy and a developmental disorder. Chen et al. solved the x-ray crystal structure of the Smchd1 hinge domain, which is important for homodimerization and nucleic acid binding. Site-directed mutagenesis studies identified critical residues involved in SMCHD1 function in cells. Together, these data suggest how mutations in the SMCHD1 hinge domain contribute to human disease.

Abstract

Structural maintenance of chromosomes flexible hinge domain containing 1 (SMCHD1) is an epigenetic regulator in which polymorphisms cause the human developmental disorder, Bosma arhinia micropthalmia syndrome, and the degenerative disease, facioscapulohumeral muscular dystrophy. SMCHD1 is considered a noncanonical SMC family member because its hinge domain is C-terminal, because it homodimerizes rather than heterodimerizes, and because SMCHD1 contains a GHKL-type, rather than an ABC-type ATPase domain at its N terminus. The hinge domain has been previously implicated in chromatin association; however, the underlying mechanism involved and the basis for SMCHD1 homodimerization are unclear. Here, we used x-ray crystallography to solve the three-dimensional structure of the Smchd1 hinge domain. Together with structure-guided mutagenesis, we defined structural features of the hinge domain that participated in homodimerization and nucleic acid binding, and we identified a functional hotspot required for chromatin localization in cells. This structure provides a template for interpreting the mechanism by which patient polymorphisms within the SMCHD1 hinge domain could compromise function and lead to facioscapulohumeral muscular dystrophy.

INTRODUCTION

Structural maintenance of chromosomes (SMC) proteins are an evolutionarily conserved family responsible for regulating chromatin and chromosomal organization (1). Conventional SMC proteins contain several structural motifs, including a bipartite adenosine 5′-triphosphate–binding cassette (ABC)–type adenosine triphosphatase (ATPase) domain comprising subdomains contributed by the N and C termini, a centrally-located hinge domain, and a coiled-coil arm formed by the two intervening segments (13). Whereas prokaryotic SMC proteins exist as homodimers, there are three exclusive pairs of SMC heterodimers in eukaryotes for which the hinge domain constitutes the dimerization interface. Together with auxiliary subunits, SMC proteins form large protein complexes that play distinct roles in sister chromatid cohesion, chromosome condensation, and DNA repair.

SMC flexible hinge domain containing 1 (SMCHD1) is considered a noncanonical member of the SMC protein family because it has a domain architecture that is distinct from those of other SMC proteins, including an N-terminal gyrase, Hsp90, histidine kinase, and MutL (GHKL)–type ATPase domain (46) and a hinge domain located at the C terminus (79). Disruption of SMCHD1 function is as an underlying factor in two human disorders: the muscular dystrophy, facioscapulohumeral dystrophy (FSHD) (1012), and the developmental disorder, Bosma arhinia micropthalmia syndrome (10, 13, 14). Therefore, there is increasing interest in uncovering how SMCHD1 functions in development and pathogenesis. At the molecular level, our previous studies demonstrated that SMCHD1 occupies defined genomic regions to exert epigenetic silencing effects (15, 16). We and others reported that Smchd1 mediates long-range chromatin interactions and appears to insulate the chromatin from the action of other DNA binding proteins (1618). The hinge domain plays a crucial role in exerting these functions, because it is this domain that mediates SMCHD1 homodimerization and nucleic acid binding in vitro (9, 15). This functionality is clearly critical, because when compromised by the R1866G substitution present in an FSHD kindred, the silencing capacity of SMCHD1 is reduced (9, 15). In contrast to the antiparallel intramolecular coiled-coil present in canonical SMC proteins, our previous small-angle x-ray scattering (SAXS) studies indicate that SMCHD1 adopts an intermolecular parallel coiled-coil arrangement flanking the hinge domain, consistent with a predicted head-to-head arrangement in full-length SMCHD1 dimers (9). However, the atomic-resolution structure of the SMCHD1 hinge domain and the mechanism underlying its association with nucleic acids remain to be elucidated.

Several crystal structures of SMC hinge dimers have been previously reported, including those of the prokaryotic SMC homodimer (19), the SMC1-SMC3 heterodimer of the cohesin complex (20), SMC2-SMC4 of the condensin complex (21), and the SMC5-SMC6 heterodimer (22). In this study, we present the crystal structure of the homodimer of the hinge domain of mouse Smchd1, which adopts a fold grossly resembling the canonical SMC hinge and forms a homodimer through a similar dimerization interface. Smchd1 has distinctive features compared to those of canonical hinge domains, such as fewer, longer helices, the absence of a pair of β strands at the interface between the two subdomains, and different secondary structural elements at the termini. By performing structure-guided mutagenesis, we defined the dimerization interface, implicated the extended linker region and residues surrounding the “central channel” in nucleic acid binding, and demonstrated the importance of the hinge domain and a functional hotspot within this domain for chromatin localization in cells. Together, our findings provide insights into SMCHD1 function at the molecular level and provide a basis for interpreting how mutations in the SMCHD1 hinge domain might affect function and contribute to human disease.

RESULTS

The Smchd1 hinge domain contains an unconventional subdomain I structure

We solved the structure of the mouse Smchd1 hinge domain (residues 1683 to 1899) by x-ray crystallography to 3.3-Å resolution after anomalous phasing using mercury derivatized crystals (Table 1 and fig. S1). The overall architecture of the Smchd1 hinge domain has similarities to previously published SMC hinge domain dimers, forming a doughnut shape with the N and C termini on the same face (Fig. 1, A to D). A distance-matrix alignment (DALI) structure search identified the Thermotoga maritima SMC (TmSMC) hinge domain structures as the closest structural homologs of the Smchd1 hinge domain Protein Data Bank (PDB) [PDB 1GXK and 1GXJ; root mean square deviation (RMSD) of 3.1 Å (aligning 131 Cα) and 3.7 Å (aligning 131 Cα), respectively] (19, 23). The TmSMC hinge was the first-hinge domain structure reported, and as a homodimer, 1GXJ provides a reference point for discussing the features of the Smchd1 structure. Similar to the TmSMC and other canonical SMC proteins, the SMCHD1 hinge monomer comprises subdomains I and II connected by a flexible linker. The TmSMC subdomains are pseudosymmetric, with subdomain I containing a three-stranded β sheet (S1 to S3) and subdomain II containing a five-stranded β sheet (S4 to S8) with a pair of β strands (SI and SI′) contributing to an auxiliary intersubdomain interface (Fig. 1, E to G). Within the TmSMC homodimer, β sheets contributed by subdomains I and II of dimer-opposed protomers contribute to a composite extended β sheet. In contrast, Smchd1 lacks the N-terminal strand, with only S1 and S2 from subdomain I contributing to the intersubdomain β sheet. In addition, it lacks the two antiparallel β strands (SI and SI′) that are central to the intersubdomain interface in TmSMC, which are replaced by extended subdomain interface helices in Smchd1 (Fig. 1, B and C). Furthermore, the Smchd1 hinge domain is distinct because of a short-loop insertion in S7 that generates an additional β strand, S8, at the C terminus. As a result, S7 loops out of the plane of the intersubdomain β sheet, and instead, S8 forms a β sheet with strands S3 to S6 from subdomain II as part of the intersubdomain extended β sheet. There are also differences between the α helices present in the TmSMC and Smchd1 hinge domains. In the TmSMC hinge, subdomain I is flanked by H1 to H2 and H6, and subdomain II is flanked by H11 (Fig. 1F). However, in Smchd1, the equivalent of the helices H1 to H2 is not observed in the density, although the corresponding region was present in the construct crystallized. As a result, the subdomain I and II β sheets in the hinge domain of Smchd1 are flanked by only one α helix denoted H3 in subdomain I and H6 in subdomain II (Fig. 1, B and C), which relative to their counterparts in TmSMC, H6 and H11 are shortened and lengthened by one helical turn, respectively. Grossly, the intersubdomain orientations of the Smchd1 and TmSMC hinge domains differ markedly, with a rotation of the Smchd1 subdomain II by 18° or 27° relative to the closed (1GXK) or open (1GXJ) TmSMC reference structures. Such a rotation places the two subdomains I further apart and the two subdomains II closer within the Smchd1 hinge dimer (Fig. 1, B, D, F, H, and I).

Fig. 1 The overall structure of the hinge domain of SMCHD1 reveals differences compared to canonical SMC proteins.

(A and B) Schematic diagrams of SMCHD1 (A) and the canonical SMC (B) protein domain architecture. SMCHD1 and canonical SMC proteins dimerize through their hinge domain, which is composed of subdomain I (slate and cyan) and subdomain II (green). The canonical SMC dimer has long intramolecular antiparallel coiled coil arms that are composed of N- (blue) and C-terminal (pink) residues flanking the hinge domain, and the ABC-type ATPase domain constituted of N- and C-terminal subunits. The Smchd1 dimer contains shorter intermolecular coiled-coils that are connected to its N-terminal region (gray) encapsulating a GHKL-type ATPase domain (red). (C and D) The crystal structure of the Smchd1 hinge domain (C) and a canonical SMC hinge [PDB: 1GXK;(D)] as shown from above the torus in cartoon representation. The component protomers are demarked by a dashed border. The color scheme is the same as in Fig. 1A but with protomer 1 in purple (subdomain I) and orange (subdomain II) for contrast. (E and F) Diagrams depicting the topology of the Smchd1 monomer [residues 1710 to 1884; (E)] and the canonical SMC hinge domain from T. maritima (F). (G and H) Cartoons comparing the hinge domain monomers of Smchd1 (G) and T. maritima (H). (I) Superimposition of the Smchd1 and TmSMC hinge domains [from (G) and (H), respectively]. Secondary structure is annotated sequentially for Smchd1 and according to the original TmSMC description (19); however, short helical segments (consisting of less than two hydrogen bonds) have been omitted from the renderings.

Table 1 X-ray crystallography data collection and refinement statistics.

Statistics for the highest-resolution shell are shown in parentheses.

View this table:

Our previous SAXS analyses of the extended hinge domain (9), and an electron microscopy study of full-length SMCHD1 (4), support a role for the coiled-coil regions immediately flanking the hinge domain in homodimerization in a head-to-head orientation. However, these coiled-coil regions were not observed in our crystal structure. The construct subjected to crystallization encompassed residues 1683 to 1899; however, density was scarce for the N-terminal (1683 to 1709) and C-terminal (1887 to 1899) regions despite mass spectrometry (MS) analysis confirming that the crystallized protein was intact (fig. S2A). As a result, densities for between 27 and 32 amino acids at the N terminus and 13 to 15 at the C terminus were unresolved, varying among the six hinge domain copies in the asymmetric unit. It was possible to assign two additional helices in the electron density of 12 and 21 amino acids within the asymmetric unit (fig. S2, B and C). However, because these helical segments contained no selenomethionines (SeMets) and had poor side-chain density, sequence assignment was not possible, and they were modeled as polyalanine helices. The absence of sequence information prevented us from deducing their connectivity to the hinge domain, and it remains unclear why only two additional helices were present in an asymmetric unit containing six copies of the hinge domain. One possibility is that these helices could contribute to the coiled-coil regions (comprising residues 1683 to 1709 and residues 1885 to 1899) predicted to flank the core hinge domain. Whether these flanking coiled-coils adopt a closed conformation, as observed in structures of the yeast SMC2-SMC4 complex and Pyrococcus furiosis SMC hinge domain (21), or an open conformation, as seen in the Geobacillus stearothermophilus (24) and Escherichia coli MukB SMC hinge domain (25) structures (fig. S3), in the structure presented here remains unclear in the absence of ordered coiled-coils.

The Smchd1 hinge domain differs from other SMC hinge domain structures

Several structures of eukaryotic heterodimeric SMC complexes have been reported, including mouse cohesin SMC1-SMC3 (20), human condensin SMC2-SMC4 (26), and yeast SMC5-SMC6 (22) (fig. S4, A to H). The sequence of the Smchd1 hinge domain shows little homology to these proteins, with amino acid identity ranging from 10 to 20%. Structural alignments identified mouse SMC3 as the closest heterodimeric SMC hinge domain structural homolog, with an RMSD of 3.4 Å (aligning 124 Cα), despite a modest 12% sequence identity for the aligned regions. Human SMC4 was the next most similar structurally, with an RMSD of 3.8 Å (aligning 135 Cα) and a 22% sequence identity.

Although the overall fold of Smchd1 is similar to that of the prokaryotic TmSMC and its eukaryotic orthologs SMC1 to SMC6, the hinge domain of Smchd1 has several divergent structural features. The loss of H1/H2 and the N-terminal β strand from the Smchd1 hinge is a distinguishing feature, because whereas H1 is absent from human SMC2 (26), concurrent loss of the N-terminal β strand has not been observed in any structures reported to date (fig. S4, A to I). Furthermore, there are differences between the intersubdomain interfaces between Smchd1 and canonical SMC hinge domains. In Smchd1, the subdomain I helices at the interface, H1 and H2, are extended by 1.5 and 1 turns, respectively, relative to their counterparts in TmSMC (H3 and H4; Figs. 1, E and F, and 2), with the equivalent of H5 absent from Smchd1. The subdomain II helices at the interface are conserved in Smchd1, but H5 is one turn shorter than H10, its counterpart in TmSMC (Fig. 1, C and G). Another distinguishing feature of the Smchd1 hinge domain is the extended linker region that bridges subdomains I and II (residues 1778 to 1805) between S2 and S3 (Figs. 1, C and G, and 2). Among copies of Smchd1 in the asymmetric unit, this was the structural element that exhibited the greatest heterogeneity, consistent with loop flexibility, whereas the length of this linker varies between SMC hinge domains from one residue shorter in human SMC2 to 10 residues shorter in TmSMC (Fig. 2), it is notable that the linker of Smchd1 has the greatest abundance of positively charged amino acids, suggesting a possible role in interactions, such as with nucleic acids.

Fig. 2 Multiple sequence alignment of the hinge domains of the Smchd1 hinge domain and canonical SMC proteins.

Protein sequences of Mus musculus (Mm) Smchd1 and canonical SMC proteins—including MmSMC1 and SMC3, Homo sapiens (Hs) SMC2 and SMC4, and T. maritima (Tm) SMC—were aligned using the program MultAlin (50). α Helices (H) and β strands (S) are colored in red and yellow, respectively, the 310 helices are shaded in purple, and the loop connecting subdomains I and II is colored in blue. The secondary structures of mouse Smchd1 and TmSMC as defined by DALI are annotated above and below the aligned sequences, respectively. The image was generated using ESPript3.0 (51). Gray text indicates regions of the SMCHD1 structure that could not be built because of unresolved electron density.

The Smchd1 hinge dimer interfaces comprise canonical and noncanonical interactions

The Smchd1 hinge domain exists as an obligate homodimer with twofold rotational symmetry (Fig. 1C). As for canonical SMC proteins, dimerization is mediated by two key interfaces between SMC pairs (fig. S4). In Smchd1, two helices contributed by subdomain I (H3) and subdomain II (H6) of opposing Smchd1 protomers form the first interface, which includes a salt bridge between Arg1762 (R1762) of H3 and Asp1842 (D1842) of H6 (Fig. 3A and fig. S4A). These helices cap a second interface comprising an extended β sheet formed by β strands S1 and S2 from the subdomain I of one protomer (which also contributes H3 to the dimer interface) and β strands S3 to S8 of subdomain II from the symmetrical protomer. This interaction involves backbone-backbone hydrogen bonds between Val1774 (V1774) of S2 in subdomain I and Phe1874 (F1874) of S7 in subdomain II of the dimeric partner (Fig. 3A). Broadly, a similar interface is formed in TmSMC in which H6 of subdomain I of one protomer and H11 from the other form a cap over a β sheet containing an additional β strand in addition to those in Smchd1 (Fig. 1D).

Fig. 3 Dimerization of the Smchd1 hinge domain occurs via canonical and noncanonical interfaces.

(A) Magnified view of the dimerization interface of the Smchd1 hinge domain as highlighted with subdomain I in slate and subdomain II in green and the side view at 90° as indicated. Key residues coordinating the dimer interface are shown as sticks with numbers indicated and highlighted according to their subdomain. Electrostatic interactions between residues are indicated by dashed lines. (B to G) Small-angle x-ray analysis of WT [B; replotted from our earlier report; (9)] and the representative dimer interface mutants Y1765A (C), V1774G (D), R1848A (E), K1873A (F), and G1872A/G1875A/G1876A (G). Top: Scattering intensity profiles are shown where the background-subtracted SAXS data are indicated as black circles representing mean intensity I(q) as a function of momentum transfer q in Å−1. Guinier plots for qRg ≤ 1.3 are shown as insets, where linearity indicates that high–molecular weight and polydisperse particles do not measurably contribute to the scattering. The radius of gyration and initial scattering intensity I(0) were approximated using the Guinier equation with PRIMUS (43), giving values as indicated in table S1. Bottom: Pair distribution functions, P(r) plots were calculated using GNOM (44). The Rg and maximum particle dimension Dmax calculated from the P(r) analysis are as reported in table S1. Data in (C) to (E) were collected with a PILATUS 1M detector; a PILATUS 2M detector was used to collect data for the G1872A/G1875A/G1876A mutant.

Furthermore, in addition to the canonical dimer interfaces noted earlier, several distinct interactions may also contribute to the Smchd1 hinge dimer interface. A noncanonical β strand, S8, present at the C terminus of Smchd1 forms part of the extended β sheet with strands S1 to S6, which displaces the S7 strand out of the plane and, in so doing, imparts a defining feature of the SMCHD1 hinge domain (Fig. 1, C and E, and fig. S4, A and B). This feature is likely stabilized by distinct contacts between H3 of subdomain I and S6 and S7 of subdomain II from the other protomer. These contacts include a hydrogen bond between the hydroxyl of Tyr1765 (Y1765) in H3 and the S6 Asp1866 (D1866) carboxylate, and aliphatic interactions between Tyr1765 of H3 and Phe1874 of S7 (Fig. 3A). Moreover, S7 not only mediates intersubunit interactions with S2 and H3 but also interacts with H2 from the other protomer through a salt bridge between the side chains of Lys1873 (L1873) of S7 and Asp1749 (D1749) of H2 (Fig. 3A). As is typical of other SMC hinge domains, Smchd1 contains a triglycine motif at the C terminus of the domain that is involved in intersubunit interactions and houses the aforementioned Lys1873 and Phe1874. Note, however, that the Smchd1 motif is GX6GX2GG rather than the more typical SMC motif GX6GX3GG. This difference contributes to the kink between S7 and S8, which, in turn, is an important determinant of the unconventional interaction between protomers. Furthermore, Arg1848 (R1848) within H6 of subdomain II interacts with the side chain of Ser1779 at the start of the extended linker of the opposing monomer, potentially strengthening its canonical interaction with H3 of subdomain I from the other protomer.

Mutagenesis of dimer interface residues does not alter the overall structure of the Smchd1 hinge dimer

To examine whether residues located at the crystallographic dimer interfaces were important to dimer formation, we generated 10 mutants in which interface residues were substituted, including a triple mutant replacing all three glycine residues at the consensus site. All of the mutants were successfully expressed and purified analogously to the wild-type (WT) protein, with all mutants eluting from size exclusion chromatography (SEC) as a single peak that resembles the profile of the WT hinge dimer (fig. S5). This indicates that all interface mutants remained dimeric in solution under these purification conditions.

To confirm the dimerization state and examine gross structural changes caused by these mutations, we used SAXS to study five representative mutants: Y1765A, V1774G, R1848A, K1873A, and G1872A/G1875A/G1876A from within H3, S2, H6, S7, and loop regions flanking S7, respectively. Background-subtracted scattering profiles of WT and mutant Smchd1 hinge domains (Fig. 3, B to G, top) were subjected to Guinier (Fig. 3, B to G, top inset) and interatomic distance distribution (Fig. 3, B to G, bottom) analyses. The linearity of the Guinier plots indicates that the samples were monodisperse, enabling us to perform more detailed analyses of the size and shape of the mutants. Our previous studies established that the radius of gyration values (Rg) of WT Smchd1 is ~27.0 Å, with a maximum particle dimension (Dmax) of 100 Å (9). These parameters correspond to a homodimeric configuration, as validated with orthogonal analytical ultracentrifugation studies (15). Here, we observed similar parameters for the Smchd1 hinge domain mutants (Fig. 3, C to G, and table S1). Together with our SEC data, our SAXS analyses support the idea that Smchd1 interface mutants form stable homodimers in solution, with overall sizes and shapes grossly resembling those of the WT protein.

Dimer interface residues contribute to protein thermal stability

We reported previously that a mutation related to FSHD2, R1867G (R1866G in humans), did not change the overall conformation of the Smchd1 hinge domain homodimer but reduced the thermal stability of the protein as assessed by differential scanning fluorimetry (DSF) (9). Because the residue-neighboring Arg1867, Asp1866, mediates dimerization contacts, we postulated that mutation of the dimer interface residues would affect dimer stability. The melting temperature (Tm) for WT and R1867G mutant dimers were measured as 53° and 42°C, respectively (Table 2 and fig. S6), comparable to our previous results (9), whereas most of the dimer mutants showed compromised stability in parallel assays. The exceptions were D1842A, which exhibited a Tm comparable to that of the WT dimer, and K1873A, which showed an increased thermal stability (Tm of 68°C). Six of the eight mutants that showed decreased thermal stability exhibited Tm values between 41° and 47°C, consistent with a role for the mutated residues in stabilizing the dimer contacts proposed based on our hinge domain crystal structure. Introduction of V1774G or the triple glycine mutant even further decreased thermal stability, with these mutants having Tm values of 35° and 38°C, respectively.

Table 2 Summary of melting temperature (Tm) and dissociation constant (Kd) values for the WT Smchd1 hinge and for those of the indicated mutants.

N.D., not determined.

View this table:

The introduction of glycines into ordered regions destabilizes secondary structures (27). In addition, V1774G mutation within S2 will result in loss of van der Waals interactions to the adjacent Phe1874 in S7 from the other protomer at the dimerization interface. The combination of these two factors likely contributes to the observed decrease in the thermostability of the V1774G mutant. Note that the D1842A mutation did not compromise thermal stability, but mutation of a residue with which it forms a salt bridge, Arg1762 (fig. S1D), did. Whereas Asp1842 is involved exclusively in an interaction with Arg1762, the R1762 mutation would be expected to abrogate multiple interactions, which would have a deleterious effect on thermal stability. For example, Arg1762 is sandwiched between Asp1766 and Asp1842; thus, loss of Arg1762 would be expected to destabilize the protein due to the repulsion between the two negatively charged aspartates. In addition, mutation of Arg1762 would preclude the formation of a charged pair with Asp1766, one helical turn away, and likely contributes to destabilization of this helix. Because Gly1872, Gly1875, and Gly1876 are in loop regions flanking S7, their mutation to alanine likely destabilizes the S7 to S8 loop by interfering with backbone hydrogen bonding to Gln1878 and Asn1879 (fig. S4B), thereby reducing the thermostability. Therefore, these results support an integral role for multiple intersubunit contacts, including between residues in S2 and S7, in stabilizing hinge dimers. Although it appears that, overall, these mutations do not block dimer formation, likely due to a role for the coiled-coil regions flanking the core hinge domain, destabilization of the homodimer was evident from the reductions in thermal stability.

Note that the substitution of another S7 residue, Lys1873, with alanine unexpectedly increased the Tm to ~68°C. The side chain of Lys1873 is orientated in parallel to the positively charged side chain of the neighboring Arg1869 in the adjacent S6 strand, with both side chains directed toward the central cavity of the Smchd1 dimer. We hypothesized that the K1873A substitution might eliminate an electrostatic repulsive interaction between Lys1873 and Arg1869, thus increasing thermal stability. We tested this possibility by generating an R1869A mutant, which we found also exhibited an increase in Tm to that of K1873A (Table 2). Therefore, the observed stabilizing effect of these alanine substitutions is likely explained by attenuation of repulsive steric interactions between the positively charged side chains.

The Smchd1 hinge domain binds to nucleic acids through two positively charged patches

Analysis of the electrostatic surface potential of the Smchd1 hinge domain structure revealed the domain to be highly basic in charge, as befitting a domain previously shown to bind to nucleic acids in vitro (15). We identified three clusters of positive charge: cluster 1, cluster 2 on the opposite face, which is contributed by the extended linker, and cluster 3 surrounding the central channel (Fig. 4, A and B). We substituted the positive residues in these clusters with alanine to deduce nucleic acid–binding residues using a fluorescence polarization assay. None of the tested cluster 1 mutants had a marked bearing on nucleic acid binding (Fig. 4C). In contrast, the mutations R1790A, R1796A, and K1799A in cluster 2 led to, respectively, 5-, 2.8-, and 2.2-fold reductions in nucleic acid–binding affinity, suggesting that residues within cluster 2 contribute to nucleic acid binding (Fig. 4D). Our previous work using electrophoretic mobility shift assays and analytical ultracentrifugation assays established the importance of the FSHD2-associated substitution, R1867G, in DNA binding (15). Here, using a fluorescence polarization assay, we validated the importance of Arg1867 (residing in cluster 3) for nucleic acid binding (Fig. 4, A and E, and Table 2). Our assay examined the binding of the recombinant Smchd1 hinge domain to a single-stranded DNA sequence containing the HS1-b binding site sequence harboring a methylated cytosine, although we would expect comparable binding to double-stranded DNA and unmethylated DNA based on our previous studies (15). In addition to R1867G, the cluster 3 mutants, R1869A and K1873A, exhibited respectively 15- and 5-fold reductions in nucleic acid–binding affinity relative to that of the WT Smchd1 hinge domain. Note that among the interface residues, besides K1873A, which also contributes to cluster 3, the only tested mutations that showed deficits in DNA binding were R1762A and R1848A, with 2.1- and 7.6-fold reductions in dissociation constant (Kd), respectively (Fig. 4F). This suggests that these residues perform dual roles in mediating interprotomer contacts and ligand binding. We then examined whether binding might occur through the torus of the hinge domain dimer by introducing mutations to occlude DNA from entering the central hole of the dimer (Fig. 4G). We selected Ser1870 for substitution because it is not engaged in crucial structural interactions in our dimer structure, and we chose His1856 because it lies on the rim of the torus. Introduction of methionine, asparagine, or glutamine in place of Ser1870, or substitution of His1856 with the bulkier residue tryptophan, did not measurably affect nucleic acid binding (Fig. 4G), consistent with nucleic acid binding not occurring by threading through the central channel. Together, these analyses implicate two sites surrounding the Smchd1 hinge domain torus in nucleic acid binding (Fig. 4, B and H to J): binding site 1 centered around the linker (cluster 2) and binding site 2 on the opposite face (cluster 3).

Fig. 4 The nucleic acid–binding activity of the Smchd1 hinge homodimer is potentiated by two positively charged interaction sites.

(A) Electrostatic surface potential representations of the Smchd1 hinge homodimer viewed down the side of the torus. Positively charged residues are shown as sticks in dark blue. Lys1718, Arg1719, and Arg1771 at the top side of the dimer are grouped as cluster 1 (black). Lys1789, Arg1790, Arg1796, and Lys1799 located at the bottom side of the dimer are grouped as cluster 2 (cyan). Residues in close proximity to the pathogenic mutation Arg1867, including Arg1869, Lys1873, and Lys1880, are grouped as cluster 3 (blue). (B) Cartoon representation of the Smchd1 homodimer in the same orientation as that shown in (A) highlighting residues that interact with nucleic acid on one protomer. The thin red line indicates the possible position of nucleic acid binding as deduced from these experiments. (C to G) DNA binding fluorescence polarization (mP) of 12.5 nM 6-FAM–labeled, 20-base pair single-stranded DNA by wild-type (WT) and clusters 1 (C), cluster 2 (D), cluster 3 (E), dimer interface (F), and the torus (G) mutants of the Smchd1 hinge dimer at indicated protein concentrations. Data points are plotted for technical duplicates for three (WT and torus mutants) or two (all other mutants) independent experiments fitted to a hyperbolic binding curve. Kd values that were calculated from the fitted curves are reported in Table 2. (H to J) The Smchd1 hinge dimer has two clusters of positively charged residues implicated in nucleic acid binding. The first interaction site, comprising Arg1790, Arg1796, and Lys1799 from cluster 2 (cyan), is positioned on the surface of the dimer where the intersubdomain linker resides. The second interaction site is located at the central channel of the hinge dimer on the opposite side of the dimer comprising Arg1867 and Lys1873 from cluster 3 (blue). Residue Arg1848 (colored in magenta in the structural figures), which is also positioned at the central channel, is located between interaction sites 1 and 2 and is also implicated in nucleic acid binding. Panels (H) and (J) show the same view of the hinge domain dimer to show the torus residues selected for mutation (J) and cluster 2 residues (H). Distances between the residues located at the two nucleic acid interaction sites are marked beside corresponding dashed lines in (H) and (I).

The hinge domain targets Smchd1 to chromatin

Mutational studies identified clusters 2 and 3 and the interface residue, Arg1848, as important contributors to nucleic acid binding in vitro. It remained of interest whether these residues contributed to the cellular function of Smchd1. Accordingly, we introduced the cluster 2 mutations (R1790A and R1796A), the interface mutations (D1842A and R1848A), and the cluster 3 mutations (R1867G, R1869A, and K1873A) into full-length mouse Smchd1 and transiently expressed these constructs in human embryonic kidney–293 cells in which endogenous SMCHD1 was silenced by a short hairpin DNA (shRNA; Fig. 5 and figs. S7 and S8A). shRNA-mediated silencing was restricted to the human ortholog, thereby enabling us to examine the effects of hinge domain mutations on the function of mouse Smchd1. Note that all of the transfections were performed in parallel under identical conditions, including transfecting the cells with identical amounts of each expression construct and using the same imaging parameters for each sample. Consequently, differences in the localization of mutant Smchd1 proteins relative to that of WT Smchd1 should reflect intrinsic differences in protein behavior, which were observed across multiple independent experiments (fig. S8). D1842 was included as an additional control because this mutation did not alter either the Tm or the nucleic acid–binding activity (Fig. 4F, Table 2, and fig. S6). In nonsilenced control 293 cells, endogenous SMCHD1 formed nuclear foci corresponding to localization to the inactive X chromosome (Fig. 5A and fig. S7A), as reported previously (28). Foci were depleted upon knockdown of endogenous SMCHD1 (Fig. 5B and fig. S7B), and successful knockdown was evidenced by the loss of SMCHD1 staining in individual cells across the microscope field. Expression of WT Smchd1 restored focal Smchd1 enrichment in the nuclei, although these foci were distinct from those formed by endogenous SMCHD1 (Fig. 5C). A similar phenotype to overexpression of WT Smchd1 was observed for cells expressing R1790A, R1796A, D1842A (control), R1869A, and K1873A (fig. S7), but cells expressing R1848A and the FSHD-associated mutant R1867G showed diffuse localization and abrogation of focus formation (Fig. 5, D and E). These data suggest that, on the whole, mutations that affected nucleic acid binding in vitro did not compromise chromatin binding, likely owing to some compensation from residues in the adjacent nucleic acid–binding cluster. It was clear, however, that Arg1848 and Arg1867 contributed to a functional hotspot in the hinge domain that played an essential role in localizing Smchd1 to chromatin and that their loss could not be compensated for by residues in the adjacent nucleic acid–binding clusters.

Fig. 5 The mutations R1848A or R1867G or compromising the integrity of the hinge domain alter the nuclear localization pattern of Smchd1.

(A to F) Immunofluorescence microscopy analysis of (A) control, nonsilencing shRNA–transduced, or shRNA-mediated SMCHD1 knockdown (KD) 293 cells in the absence (B) or presence of WT (C) or mutant (D and E) full-length Smchd1 or a variant missing the hinge domain (F). Maximum intensity projection images are shown as representative of n > 150 nuclei positive for Smchd1 overexpression per sample. Data are representative of three independent experiments. All images were obtained with identical settings between controls and all transfected cells to enable comparison between the images provided in the figure. Scale bars, 20 μm. DAPI and SMCHD1 staining and merged channels are displayed in fig. S7, and repeat experiments are shown in fig. S8.

Because the R1848A and R1867A mutant hinge domains exhibited reduced thermal stability in vitro, which could be attributed to the loss of several interactions (fig. S8, B and C), we further examined whether hinge domain integrity contributed to Smchd1 chromatin localization. Introduction of mutations that compromised thermal stability within hinge domain-coiled coil dimers, V1774G and G1872A/G1875A/G1876A (fig. S7, K and L), or deletion of the hinge domain (ΔL1710-M1884; Fig. 5F and fig. S7M), in full-length Smchd1 led to abrogation of focus formation. These data support an essential role for the Smchd1 hinge domain in chromatin localization and demonstrate that mislocalization can arise when hinge domain integrity is compromised.

DISCUSSION

Here, we present the crystal structure of the Smchd1 hinge domain. The overall topology of this domain grossly resembles the previously described TmSMC hinge domain homodimers and the eukaryotic SMC hinge domains of the cohesin SMC1-SMC3 (20), human condensin SMC2-SMC4 (26), and yeast SMC5-SMC6 (22) heterodimers, despite having sequence identities of only 10 to 20%. Common to these structures, the Smchd1 hinge domain contains a conserved, extended β sheet formed by subdomains I and II contributed by the two protomers in the dimer, which is capped by an α helix from each subdomain. However, there are several distinguishing features of the Smchd1 hinge domain compared to the other solved SMC hinge domain structures. Further to the absence of the H1-H2 and S1 and the addition of the C-terminal S8, the relative orientations of the two component subdomains in hinge of Smchd1 differ from conventional SMC hinge domains. The first indications of differences in subdomain orientations arose during unsuccessful attempts to solve the structure by molecular replacement, which necessitated experimental phasing. The likely basis for this is the rotation of subdomain II by 18° or 27° relative to the respective closed or open TmSMC reference structures. These rotations arise from the presence of three key nonconserved residues in the intersubunit interface: Val1774 (S2) and Tyr1765 (H3) in subdomain I and Phe1874 (S7) in the dimer-opposed subdomain II. In addition to their roles in mediating dimer formation, as evidenced by the decreased thermal stability of the V1774G, Y1765A, and F1874A mutants, these residues likely determine the preference of Smchd1 for self-association rather than forming heterodimers with other SMC proteins. The atypical subdomain orientations in the Smchd1 hinge domain structure is enabled by the unconventional GX6GX2GG motif, which leads to S7 dropping out of the plane of the extended β sheet and the β strand peculiar to Smchd1, S8, forming part of the β sheet.

Our structure enabled us to visualize the core hinge domain of Smchd1, but we did not observe electron density for the flanking coiled-coil regions except for two unassigned helices in the asymmetric unit. This was unexpected, considering that the coiled-coil forming sequences were present within the crystallized protein, and because previous SAXS analyses of the hinge domain and flanking coiled-coils (9) and electron microscopy of full-length SMCHD1 (4) indicated the propensity for the coiled-coil regions flanking the core hinge domain to interact in a head-to-head manner. In contrast, structures of previously reported SMC domains form coiled-coils through head-to-tail (N- to C-terminal coil) coiled-coils that may arrange into either an open or closed conformation (fig. S3) (21, 24, 25). The lack of electron density for the flanking coiled-coil regions in the Smchd1 structure indicates that these regions are likely to be flexible, although whether they adopt an open or closed conformation cannot be deduced from the structure reported here. Furthermore, the N terminus of the Smchd1 hinge domain lacks the H1-H2 typical of canonical SMC domains. One explanation for this absence could be that this is a site of crystal contacts in both the tetragonal and rhombohedral crystal forms, but more broadly is suggestive of lability at this site that would enable exposure of the N-terminal region of subdomain I to facilitate crystal contact formation. Although the atomic detail of the coiled-coils was not revealed in our crystal structure, the importance of these regions in augmenting homodimer stability is supported by our studies of dimer interface mutants, which led to reduced stability but no evidence of abrogated dimer formation. Mutations at the intersubunit interface identified from our hinge domain structure, including at the variant hinge domain GX6GX2GG motif, did not compromise dimer assembly but did affect stability. This was unexpected because mutation of this motif, identified as putative dimerization interface residues based on homology to other SMC hinge domains, was concluded to play a key role in Smchd1 dimerization in a previous study (4). In that study, triple mutation to alanine caused a modest shift in elution time in analytical SEC (similar to our own observations; fig. S5A), although the proposed disruption remained to be established biophysically. In our study, mutation of the three interface glycines or other interface residues led to reductions in thermal stability of the dimer, while not breaking the Smchd1 hinge homodimer, with SAXS analyses consistent with retention of the homodimer configuration. These observations are consistent with the flanking coiled-coils augmenting dimer formation, even when the core hinge domain intersubunit interface is compromised by mutation.

Like other SMC dimer structures and modeled counterparts (20, 21), the Smchd1 hinge domain has highly basic surfaces, of which only clusters 2 and 3, but not cluster 1 (residues adjacent to H1 and H3 in subdomain I), markedly contributed to nucleic acid interaction in vitro. The Smchd1 DNA binding clusters were not inside the donut of the dimer structure, because mutation of torus residues to residues with bulkier side chains that would occlude DNA binding had no effect on nucleic acid binding. Instead, nucleic acid binding was mediated by two surface patches (clusters 2 and 3). This contrasts with the cohesin SMC1-SMC3 heterodimer central pore, where mutation of three positively charged SMC1 residues in the lumen of the hinge donut abolished DNA loading, but not recruitment to chromatin or cohesin catalytic activity (29). Unexpectedly, most of the residues that we have implicated in DNA binding here are not well conserved between SMC domains, whereas all are highly conserved in vertebrate Smchd1 sequences, with only charge-conservative substitutions observed. Consistent with its high conservation among Smchd1 orthologs, the DNA binding residue Arg1867 in S6 (cluster 3) is mutated in patients with FSHD and is associated with loss of SMCHD1 function (15). In addition, the cluster 2 residues, Arg1790, Arg1796, and Lys1799, all of which reside in the extended linker and contribute to DNA binding, are only individually present in human SMC4, mouse SMC1A, and mouse SMC1A/TmSMC, respectively. Our studies implicate the interface residue Arg1848 and residues in clusters 2 and 3 in nucleic acid binding in vitro, although most of these residues were dispensable for chromatin localization in cells when introduced into full-length Smchd1. Only R1848A and the cluster 3 mutant R1867G [originally identified in patients with FSHD; (15)] compromised the formation of nuclear foci. Both Arg1848 and Arg1867 form multiple hydrogen bonds to surrounding residues proximal to, and including, the Smchd1 dimer interface (fig. S8, B and C), which may contribute to their key role in chromatin localization. Together, these data indicate that other residues in nucleic acid–binding clusters can compensate for mutations that diminish nucleic acid affinity, except for a functional hotspot encompassing Arg1848 and Arg1867. We cannot exclude the possibility that compromise of hinge domain stability upon introduction of these mutations contributes to the loss of chromatin localization, as was observed when we deleted the hinge domain or introduced destabilizing mutations within the hinge domain. The precise contributions of the highly basic nucleic acid–binding clusters and intradomain dynamics await further detailed examination.

The mode by which Smchd1 interacts with DNA remains unclear, although several lines of evidence enable us to propose a model. First, our previous work showed that Smchd1 is specifically associated with regions of the chromatin where it is necessary for transcriptional silencing and that Smchd1 exhibits a slight preference for binding to methylated DNA in vitro (15, 16). By contrast, we found that any association with RNA occurs without sequence specificity (30), a finding reminiscent of the nucleic acid–binding capacity of another epigenetic regulator complex, polycomb repressive complex 2 (PRC2), which similarly lacks sequence specificity in nucleic acid interactions (31). Analogous to our findings with Smchd1, PRC2 exhibits reduced function when site-specific mutations decrease nucleic acid binding. Furthermore, in this study, the nucleic acid–binding capacity results in prolonged residence time on the DNA, and it was proposed that PRC2 is not targeted by the DNA binding activity, but rather that the nonsequence specific affinity acts simply to retain the complex on chromatin long enough for it to enzymatically modify the histones (31). We previously found that the targeting of Smchd1 to chromatin is dependent on PRC1-mediated ubiquitylation of histone 2A Lys119 (H2AK119ub) (30), which was subsequently validated by others (32). Here, we propose that this pathway is responsible for targeting, but as for PRC2, the nucleic acid–binding capacity of the hinge domain is required to stabilize the binding of Smchd1 to chromatin to enable productive silencing. The Smchd1 hinge domain structure reported here provides a foundation for understanding the molecular basis of how chromatin binding can be compromised in disease, such as by the FSHD2-associated SMCHD1 mutation R1866G, although the precise mechanism and dynamics underlying SMCHD1 chromatin binding and transcription silencing remain to be further explored in the future.

MATERIALS AND METHODS

Recombinant protein production

The DNA sequence encoding mouse Smchd1 residues 1683 to 1899 was subcloned into pPROEX HTb (Invitrogen) for expression. Mutations were introduced by oligonucleotide-directed mutagenesis polymerase chain reaction (PCR), and the inserted sequences were verified by Sanger sequencing (Micromon). Proteins were expressed and purified from BL21 CodonPlus E. coli as described previously (9, 15). Briefly, cells were cultured shaking in Super Broth to an optical density at 600 nm of ~0.6 to 0.8 at 37°C before the temperature was reduced to 18°C, and protein expression was induced by treatment with 0.5 mM isopropyl-β-d-thiogalactopyranoside overnight. Cells were lysed by sonication in lysis buffer [0.5 M NaCl, 20 mM tris-Cl (pH 8), 20% (v/v) glycerol, 5 mM imidazole (pH 8), and 0.5 mM tris(2-carboxyethylphosphine) (TCEP)] supplemented with 1 mM phenylmethylsulfonyl fluoride. His-tagged hinge domains were purified by Ni chromatography (Fast Flow resin, QIAGEN, or HisTag resin, Roche) and, after extensive washing, were eluted in lysis buffer containing 250 mM imidazole (pH 8). After cleavage of the His tag with TEV (Tobacco Etch Virus) protease (leaving an additional Gly-Ala-Met-Gly-Ser overhang sequence encoded by the vector), the protein was buffer-exchanged into lysis buffer and subjected to further Ni chromatography, and then the unbound fraction containing the hinge domain was concentrated and further purified by Superdex 200 SEC (GE Healthcare) in 100 mM NaCl and 20 mM Hepes (pH 7.5). Protein purity was evaluated by reducing SDS–polyacrylamide gel electrophoresis with SafeStain visualization, and fractions were pooled, concentrated to 5 to 10 mg/ml as estimated from the A280 measurement, aliquoted, snap-frozen in liquid N2, and stored at −80°C until required. SeMet-labeled recombinant proteins were produced similarly to the native recombinant protein, except that cells were cultured in a chemically defined culture medium as previously described with modification (33). One liter of the culture medium contains 100 ml of 10× stock solution (0.4 M Mops, 40 mM tricine, 100 μM FeSO4, 2.76 mM K2SO4, 5 μM CaCl2, 5.28 mM MgCl2, 0.4 M NaCl, 3 μM Na2MoO4, 400 μM H3BO3, 30 μM CoCl2, 10 μM CuSO4, 80 μM MnCl2, and 10 μM ZnSO4), 0.508 g of NH4Cl, 0.23 g K2HPO4, 0.84 g of NaHCO3, 0.05 g of thiamine, 20% (w/v) glucose, 100 mg of phenylalanine, 100 mg of threonine, 100 mg of lysine, 50 mg of valine, 50 mg of leucine, 50 mg of isoleucine, and 60 mg of SeMet.

Protein crystallization

Smchd1 protein (residues 1683 to 1899) at a concentration of 10 mg/ml was subjected to robotic crystal trials (C3 Facility, CSIRO) using the sitting-drop vapor diffusion method. Initial crystals were obtained in a condition comprising 0.1 M sodium bicine (pH 9.35), 0.18 M MgCl2, 20% (v/v) 2-methyl-2,4-pentanediol, or 20% (v/v) 2-propanol at 8°C. Conditions were further optimized using the hanging-drop vapor diffusion method. Diffraction quality crystals with rhombohedral symmetry were grown by mixing 1 μl of protein (10 mg/ml) with 1 μl of reservoir solution [0.1 M tris-Cl (pH 9.0), 0.21 M MgCl2, and 20% (v/v) 2-methyl-2,4-pentanediol] at 8°C. Crystals typically appeared after 20 days. Diffraction quality crystals with tetragonal symmetry were grown by mixing 1 μl of protein (5 mg/ml) with 1 μl of reservoir solution [0.1 M tris-Cl (pH 10.0), 0.24 M MgCl2, and 24% (v/v) 2-propanol] at 8°C. Crystals typically appeared after 7 days. Crystals were transferred to a cryoprotectant solution containing the reservoir solution and 30% (v/v) glycerol and were flash-cooled for x-ray diffraction data collection. SeMet-derivatized crystals with tetragonal symmetry were obtained from SeMet-labeled recombinant protein crystallized under the same conditions as for the native protein.

Structure determination

The protein showed modest sequence homology to SMC hinge domains for which previous structures have been solved, with the TmSMC hinge domain (PDB 1GXJ) providing the greatest identity at 21% (19). Attempts to solve the structure by molecular replacement with these models failed, and so anomalous phasing techniques were used. The SeMet derivative were crystallized in the tetragonal (P41212) spacegroup and diffracted to 3.3 Å; however, the anomalous signal was too weak to identify the seven potential selenium sites. To solve the phase problem, we derivatized the rhombohedral (R32:H) crystal form by soaking native crystals with a cryoprotectant solution supplemented with 100 μM methylmercury chloride and obtained a dataset that diffracted to 4.2 Å. This dataset had sufficient anomalous signal at 7 Å to identify three mercury atoms covalently attached to the Cys1752 thiol in the asymmetric unit, revealing three copies of the Smchd1 hinge domain in the asymmetric unit (fig. S1, A to C). Through a combination of density modification and noncrystallographic symmetry averaging in PHENIX (34), it was possible to identify secondary structure features in the maps that were aligned to the TmSMC hinge domain [PDB 1GXJ; (19)], generating a partial dimer model. The partial dimer was used as a search model for molecular replacement searching for a single copy in the higher-resolution SeMet dataset (TFZ 21.1) (35). This solution was used as a partial model for MR-SAD searching for selenium sites (34). The phases from MR-SAD were used in combination with the partial dimer model for a phased MR placing the remaining two dimers in the asymmetric unit (36). The resulting model with three Smchd1 hinge domain dimers was used in another MR-SAD protocol to identify the selenium locations and assign sequence to the model (37). The final P41212 model contained six copies of the hinge domain arranged into three dimers. The final model was refined in buster (version 2.10.3) (38) with Rwork and Rfree values of 21.4 and 24.1%, respectively, indicating a fit to the data comparable to other structures of similar resolution in the PDB. The diffraction data and structure refinement statistics are summarized in Table 1.

Small-angle x-ray scattering

SAXS data were collected at the Australian Synchrotron using the in-line SEC setup (39, 40), as previously described (5, 9, 10, 41). Data collection and analysis statistics are shown in table S1. Recombinant protein (50 μl of recombinant protein at a concentration of ~5 mg/ml) was resolved by injection onto an in-line Superdex 200 5/150 column (GE Healthcare) in 200 mM NaCl, 20 mM tris-Cl (pH 8), 10% (v/v) glycerol, and 0.5 mM TCEP (Pierce) and eluted in the path in the beam through a quartz capillary, as described previously (9, 10, 41). Two-second exposures of diffraction data were collected with a 1M PILATUS detector (PILATUS 2M detector for G1872A/G1875A/G1876A) and radially averaged. Buffer-only scattering collected earlier in the data collection were averaged and used to subtract background scatter from data from the apex of the SEC peak using Scatterbrain software (S. Mudie, Australian Synchrotron). Data analyses were performed with the ATSAS suite (42) as described previously (5, 9, 10, 41). Guinier analyses were performed with PRIMUS (43) to examine scatter at very low Q (qRg ≤ 1.3) to estimate the radius of gyration, Rg, and zero angle intensity [I(0)], with linearity indicating the absence of both high–molecular weight aggregates and interparticle interference. The real-space interatomic distance distribution function P(r) and the maximum dimension of the particle Dmax were computed by indirect Fourier transform with GNOM (44), which also enabled estimation of Rg and I(0).

Differential scanning fluorimetry

Analysis of the thermal stabilities of WT and mutant Smchd1 proteins (residues 1683 to 1899) was performed by DSF, essentially as described previously (45). Briefly, protein at a concentration of 0.05 or 0.1 mg/ml in 100 mM NaCl and 20 mM tris-Cl (pH 8) was subjected to thermal denaturation more than 20° to 100°C with a Bio-Rad CFX96 reverse transcription–PCR machine in a 20-μl volume containing 0.3 μl of SYPRO Orange (Sigma-Aldrich) with fluorescence measured at 530 nm. The protein Tm was calculated as the midpoint for the protein-unfolding transition by fitting the sigmoidal melt curves of quadruplicate samples to the Boltzmann equation using GraphPad Prism, as described previously (15, 46), and is reported as the mean of the Tm values calculated at each of the protein concentrations (0.05 and 0.1 mg/ml). Data points after the fluorescence intensity maximum were excluded from fitting.

Nucleic acid–binding assays

6-Fam fluorescently labeled single-stranded DNA (12.5 nM; high-performance liquid chromatography–purified, Integrated DNA Technologies) corresponding to the HS5-1b sense-methylated oligonucleotides (5′-6-Fam-ATCTGCCACCTGGTGTTC*GA, where the asterisk denotes methylation), as previously reported (15), was incubated with recombinant Smchd1 hinge at final concentrations of 0 to 100 μM diluted in protein sample buffer [100 mM NaCl and 20 mM Hepes (pH 7.5)] in 10 μl of reactions. Reactions were set up in 384-well low-flange black flat bottom plates (Corning) in duplicate and incubated at room temperature for 15 min in the dark. The emission polarization values were measured with the Envision Multilabel plate reader (PerkinElmer) with a 480-nm excitation filter, a 535-nm static and polarized filter, and fluorescein isothiocyanate fluorescence polarization (FP) dual mirror.

shRNA-mediated knockdown of SMCHD1 in 293 cells

For knockdown of human SMCHD1 in 293 cells, we designed shRNA nucleotides targeting the 3′ untranslated region (3′UTR) of SMCHD1 (3′UTR_forward, TCGAGAAGGTATATTGCTGTTGACAGTGAGCGCTTGGTTGTCACTACCTTGCAATAGTGAAGCCACAGATGTATTGCAAGGTAGTGACAACCAAATGCCTACTGCCTCGG; 3′UTR_reverse, AATTCCGAGGCAGTAGGCATTTGGTTGTCACTACCTTGCAATACATCTGTGGCTTCACTATTGCAAGGTAGTGACAACCAAGCGCTCACTGTCAACAGCAATATACCTTC). These were subcloned into the LMPEBFP2 vector (long terminal repeat miR-30 puromycin IRES EBFP2) vector (47). Retroviruses were prepared as previously described (48). 293T cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum at 37°C in a humidified atmosphere with 5% (v/v) CO2. For retrovirus production, cells at 80% confluence were transfected with calcium phosphate with the MD1-gag-pol structural vector, CAG-Eco, and the shRNA retroviral construct in the ratio of 8:24:1. Plasmid DNA was made up in 250 mM CaCl2 and precipitated in 2× Hepes-buffered saline (HBS) solution and then was added to the 293T cells in medium containing 25 μM chloroquine (Sigma-Aldrich). The medium was changed 8 and 24 hours after transfection and collected 48 and 72 hours after transfection by centrifugation to remove residual 293T cells. For transduction of 293 cells with shRNA-expressing retroviral constructs, either a nonsilencing control shRNA or shRNAs designed to knockdown SMCHD1 expression, retroviral supernatant was prepared 1:10 in medium containing polybrene (4 μg/ml; Sigma-Aldrich) and added to 293 cells at 50% confluence. After 24 hours, the medium was changed, and puromycin (5 μg/ml; Sigma-Aldrich) was added for selection of transduced cells.

Transfection of SMCHD1 knockdown 293 cells with plasmids expressing Smchd1 hinge domain mutants

The introduction of point mutations in mouse full-length Smchd1 was accomplished by PCR-mediated site-directed mutagenesis, and deletion of the hinge domain (residues L1710 to M1884) or generation of the mutations G1872A/G1875A/G1876A or L1774G were introduced by Gibson assembly before cloning into pcDNA3 vector at the Bam HI and Xba I restriction enzyme sites. All inserted sequences were verified by Sanger sequencing (Australian Genome Research Facility, Melbourne). For transfection of SMCHD1 knockdown 293 cells with constructs expressing full-length Smchd1 point mutant variants, ~2 × 104 293 cells were seeded in a 12-well plate on a 13-mm coverslip (Marienfield Superior). Twenty-four hours later, the cells were at ~80% confluency and were transfected with 1.2 μg of the appropriate construct by calcium phosphate–mediated transfection. Immunofluorescence microscopy was performed 24 hours after transfection.

Immunofluorescence microscopy

Immunofluorescence microscopy studies were performed as detailed by Chaumeil et al. (49). Briefly, the cells were washed in phosphate-buffered saline (PBS) and fixed in PBS, 3% (w/v) paraformaldehyde for 10 min at room temperature. Cells were washed three times with PBS for 5 min each and then permeabilized on ice with 0.5% (v/v) Triton X-100 in PBS, followed by three washes in PBS for 5 min each. Cells were blocked in 1% (w/v) bovine serum albumin (BSA; Life Technologies) for 15 min, followed by a 45-min incubation in a dark and humid chamber at room temperature with a primary anti-Smchd1 antibody [in-house clone 1D6, previously reported (30)] diluted 1:100 in 1% (w/v) BSA. Cells were washed three times in PBS for 5 min each and incubated for 40 min at room temperature in a dark and humid chamber with a secondary anti-rat 568 antibody (Life Technologies, A-11077) diluted 1:500 in 1% (w/v) BSA. Cells were washed three times in PBS for 5 min each and stained with 4′,6-diamidino-2-phenylindole (DAPI) for 10 min at room temperature, followed by another two washes with PBS. Coverslips were mounted in VECTASHIELD H1000 mounting medium (Vector Laboratories). Cells were visualized with a Zeiss LSM 880 NLO microscope at ×63 magnification, and z stacks were acquired. Images were analyzed with the open source ImageJ distribution package (FIJI).s

Intact protein analysis of crystals using MS

Crystals were dissolved in 3% acetonitrile (ACN) and 2% formic acid (FA). Samples were analyzed by nanoflow liquid chromatography–MS (LC-MS) on a nanoACQUITY UPLC system (Waters) coupled to a maXis II electron-transfer dissociation (ETD) mass spectrometer (Bruker) through a CaptiveSpray ion source with nitrogen gas doped with ACN using a NanoBooster (Bruker). Proteins were directly injected onto a 75 μm × 500–mm column (3.6-μm C4, packed emitter tip, IonOpticks) at 3% buffer B and separated by reverse-phase chromatography on a 45-min linear gradient from 5 to 60% buffer B (A: Milli-Q water and 0.1% FA; B: 99.9% ACN and 0.1% FA) at a constant flow rate of 500 nl/min. The maXis II ETD recorded at 1 Hz (300 to 3000 mass/charge ratio, profile mode) with the instrument controlled using otofControl version 4.0 (Bruker). Data analysis was performed using Compass DataAnalysis version 4.4 (Bruker). Spectra corresponding to Smchd1 were averaged across each LC-MS run before mass list generation (soluble NSF attachment protein).

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/13/636/eaaz5599/DC1

Fig. S1. The Smchd1 rhombohedral crystal structure used to solve the phase problem by single anomalous dispersion and electron density around Arg1762 in the tetragonal crystal structure.

Fig. S2. Coiled-coils flanking the core Smchd1 hinge domain are observed as two unassigned helices in the asymmetric unit, despite being present in intact MS analysis of SMCHD1 hinge dimer crystals.

Fig. S3. Crystal structures of SMC proteins with coiled-coils.

Fig. S4. Dimer interfaces of Smchd1 and canonical SMC hinge domains.

Fig. S5. Size exclusion chromatograms of Smchd1 hinge mutants.

Fig. S6. Tm values of WT and mutant Smchd1 hinge domains measured by DSF.

Fig. S7. The mutations R1848A and R1867G and deletion or compromise of hinge domain integrity alter the nuclear localization pattern of Smchd1.

Fig. S8. The mutations R1848A and R1867G perturb local interactions and abrogate formation of Smchd1 nuclear foci.

Table S1. Data collection and scattering parameters for SAXS analysis.

REFERENCES AND NOTES

Acknowledgments: We thank the MX and SAXS beamline staff at the Australian Synchrotron for assistance with data collection, the C3 Collaborative Crystallization Centre for assistance with robotic crystallization screening and DSF experiments, and P. Colman (Walter and Eliza Hall Institute) for helpful discussions. We thank A. Keniry (Walter and Eliza Hall Institute) for assistance with designing shRNAs. Funding: This study was supported by an Australian Research Training Program scholarship (to A.D.G), a Melbourne Research Scholarship International (to I.W.), a Cancer Council Victoria fellowship (to K.C.), the Bellberry-Viertel Senior Medical Research Fellowship (to M.E.B), and Australian National Health and Medical Research Council fellowships (grant 1079700 to P.E.C.; grants 1105754 and 1172929 to J.M.M.; and grant 1098290 to P.E.C., M.E.B., and J.M.M.). Additional support was provided by the Victorian State Government Operational Infrastructure Support and the Australian National Health and Medical Research Council IRIISS grant (9000433). We thank the InSPIRE program for supporting R.W. Author contributions: K.C., R.W.B., and A.D.G. designed and performed the experiments and analyzed the data. I.W., R.W., M.I., J.J.S., S.N.Y., P.J.H., T.A.W., D.A.H., and A.I.W. performed the experiments, analyzed the data, or both. M.E.B., P.E.C., and J.M.M. led the study and wrote the paper with K.C., R.W.B., and A.D.G., as well as input from the other authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All sequences are available on request. Plasmids require a material transfer agreement (MTA) from the Walter and Eliza Hall Institute. The refined coordinates for the P41212 structure were deposited in the PDB (PDB ID: 6N64). All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article