Three-dimensional topology of the SMC2/SMC4 subcomplex from chicken condensin I revealed by cross-linking and molecular modelling

Abstract

SMC proteins are essential components of three protein complexes that are important for chromosome structure and function. The cohesin complex holds replicated sister chromatids together, whereas the condensin complex has an essential role in mitotic chromosome architecture. Both are involved in interphase genome organization. SMC-containing complexes are large (more than 650 kDa for condensin) and contain long anti-parallel coiled-coils. They are thus difficult subjects for conventional crystallographic and electron cryomicroscopic studies. Here, we have used amino acid-selective cross-linking and mass spectrometry combined with structure prediction to develop a full-length molecular draft three-dimensional structure of the SMC2/SMC4 dimeric backbone of chicken condensin. We assembled homology-based molecular models of the globular heads and hinges with the lengthy coiled-coils modelled in fragments, using numerous high-confidence cross-links and accounting for potential irregularities. Our experiments reveal that isolated condensin complexes can exist with their coiled-coil segments closely apposed to one another along their lengths and define the relative spatial alignment of the two anti-parallel coils. The centres of the coiled-coils can also approach one another closely in situ in mitotic chromosomes. In addition to revealing structural information, our cross-linking data suggest that both H2A and H4 may have roles in condensin interactions with chromatin.

Keywords: SMC; coiled-coil; condensin; cross-linking; mass spectrometry; structure.

Figure 1.

Cross-linking of isolated condensin I complex. (a) Composition of condensin complex purified from mitotic DT40 cells using tagged SMC2 or CAP-H. (b) Cross-linker titration of condensin holocomplex. A fixed amount of isolated complex (at 0.05 μg μl⁻¹) was incubated with increasing amounts of BS3 cross-linker, subjected to SDS–PAGE and analysed by mass spectrometry. Based on gel mobilities, we postulate that band i represents an assortment of cross-linked dimers, band ii is likely to be cross-linked trimers and band iii is likely to be the cross-linked condensin pentamer.

Figure 2.

Cross-linking reveals close contacts between the SMC2 and SMC4 coiled-coil domains. Cross-link maps for (a) band i (b) band ii (c) band iii and (d) SMC2/SMC4 subcomplex visualized using xiNET (www.crosslinkviewer.org) [57]. Dashed green lines show links within subunits. Dashed blue lines show links between subunits. The coiled-coils of SMC4 are shown in red, whereas the coiled-coils of SMC2 are purple. CAP-H, CAP-G and CAP-D2 cross-link to the head and coiled-coil domains, but not to the hinges.

Figure 3.

Cross-linking of condensin in situ in isolated mitotic chromosomes. (a) Immunoblot of the isolated chromosomes cross-linked with increasing amounts of BS3, probed using CAP-H antibodies. Purified non cross-linked condensin (lane 1) serves as control. (b) Protocol of sample preparation for cross-linking/targeted mass spectrometric analysis of condensin and cohesin on chromosome. (c) Chromosome scaffolds visualized by SDS–PAGE and silver staining: XS, isolated chromosomes; XS^xl, cross-linked chromosomes; P, non-cross-linked pellet after scaffold extraction; P^xl, cross-linked pellet; S, non-cross-linked supernatant; S^xl, cross-linked supernatant. The chromosome scaffold preparation step reduced the sample complexity from over 4000 to 610 proteins.

Figure 4.

Condensin cross-links detected in situ in mitotic chromosomes. Linkage map of condensin complex cross-linked in situ in mitotic chromosomes visualized using xiNET (www.crosslinkviewer.org) [57]. Three linkages connect SMC2 with SMC4, two of them in the middle of the coiled-coils. One linkage connects the head of SMC2 with CAP-H. Nine intramolecular linkages provide information about the topology of SMC4 and SMC2 proteins. Four linkages indicate direct interactions between H2A or H4 and condensin.

Figure 5.

Homology models of SMC2 and SMC4 head domains. Ribbon diagrams of the bipartite head domains of chicken (a) SMC2 (residues M1–E167 and L1030–K1177) and (b) SMC4 (residues L79–E249 and L1129–A1280). Intradomain cross-links between lysines (orange spheres) are annotated with their Xwalk SAS distances [70]. Unlinked lysines are marked by grey spheres. The inferred location of the ATPase active site is pointed out on SMC4 (hidden in the view of SMC2). Images produced with UCSF Chimera v. 1.9.

Figure 6.

Homology models of the SMC2 and SMC4 hinge dimer. The modelled hinge fragments (SMC2 residues R507–A667; SMC4 residues S592–S762) viewed from the side are validated by nine cross-links (a), and the strongly basic surface electrostatics when viewed from the top corroborate the ability of this region to bind DNA (b). Colouring and annotation follows the scheme used in figure 5. In addition, lysines engaged in at least one intermolecular cross-link are shown as red spheres. Images and rendering with UCSF Chimera v. 1.9 interfacing with APBS [74].

Figure 7.

Some of the building blocks used to assemble the central portion of the condensin anti-parallel coiled-coils. Five of the 10 coiled-coil fragments modelled in this study are shown in two views each, providing full annotation detail of intra- and interdomain cross-links (red brackets with Xwalk SAS distances if both lysines are on the same fragment). Intermolecular cross-links are specified in the inner panel images from residues numbered in red font. These fragments span the central portion of the coiled-coil and include two sites with multiple intermolecular links (see also figure 8c). Their location in the three-dimensional model is shown schematically in the overview schematic (SMC2 residue ranges 395–469 + 746–786 (top), 293–386 + 792–895 (bottom); SMC4 residue ranges 479–544 + 793–845 (top), 431–477 + 855–945 (middle), 342–421 + 949–1034 (bottom). Images produced with PyMOL v. 1.7 (Schrödinger, LLC).

Figure 8.

Low-resolution approximation of the three-dimensional structure of the SMC2/SMC4 core of chicken condensin generated through template-assisted rigid assembly of 13 fragments. (a) Ribbon depiction of the 1096 SMC2 residues (92%) and 1111 SMC4 residues (85%) included in the model. Orange and red spheres depict Lys–Cα found in at least one high-confidence cross-link (grey spheres are unlinked lysines). Arrows mark where four sites on SMC2 and SMC4 predicted as possibly irregular in 2002 (loops I and III according to Beasley et al. [43]) line up on the modelled dimer although helical fragments were assembled solely based on the cross-linking data. (b) All-atom depiction of the model. Black lines denote the intramolecular links found between ‘domains' (table 1), which includes those between the anti-parallel helices in the coiled-coils that we used to derive/confirm their approximate relative alignments in each modelled fragment. The Cα–Cα distance average across these interdomain intramolecular cross-links (nine in SMC2; 12 in SMC4) was 16 ± 5.9 Å. The X-walk SAS Cβ-distance average over the 16 in-fragment cross-links among them was 18 ± 5.7 Å. For comparison, the Cα–Cα distance average of the 57 intradomain cross-links (not shown) was 12 ± 4.6 Å and the X-walk SAS Cβ-distance average over the 53 in-fragment cross-links among them was 16 ± 7.3 Å. (c) Red lines depict the 27 intermolecular lysine cross-links easily accommodated in this individual SMC2/SMC4 dimer (three links were rejected as not compatible). These cross-links suggest a close proximity of the two coiled-coils in the rod-like conformation of the heterodimer. The Cα–Cα distance average for these intermolecular cross-links was 21 ± 4.3 Å. Boxes enclose two clusters of intermolecular cross-links that are best modelled as a quadruple-stranded coil. (d) Fit of the assembled model to the spatial junction constraint between modelled fragments (see Results). Average distances per residue are shown for 19 junctions where between two and 10 residues were omitted in the modelling in between fragments, and constraints were imposed. For reference, typical distances for residues in α-helical and β-strand conformations are 1.5 and 3.4 Å, respectively. (e) Histogram of all measurable Cα distances in the model between cross-linked lysines, including the linkages shown in panels b and c and the 57 intradomain linkages. Molecular graphics produced with UCSF Chimera v. 1.9.

Figure 9.

Possible models of condensin complex structure based on cross-linking data. (a) Diagram of condensin as a rod-shaped complex suggested by the cross-linking data. (b) Alternative model suggesting that on chromosomes, cross-links between the SMC2 and SMC4 coiled-coils could arise owing to side-by-side association of condensin holocomplexes. (c) Cross-linking suggests that condensin can form multrimers (possibly trimers in vitro) where CAP-H proteins interact.