Structure of the DASH/Dam1 Complex Shows its Role at the Yeast Kinetochore-Microtubule Interface

Abstract

Kinetochores connect mitotic-spindle microtubules with chromosomes, allowing microtubule depolymerization to pull chromosomes apart during anaphase while resisting detachment as the microtubule shortens. The DASH/Dam1 heterodecamer (DASH/Dam1c), an essential component of yeast kinetochores, assembles into a microtubule-encircling ring. The ring associates with rod-like Ndc80 complexes to organize the kinetochore-microtubule interface. We report the cryo-EM structure (at ~4.5 Å resolution) of a DASH/Dam1c ring and a molecular model of its ordered components, validated by evolutionary direct-coupling analysis. Integrating this structure with that of the Ndc80 complex and with published interaction data yields a molecular picture of kinetochore-microtubule attachment, including how flexible, C-terminal extensions of DASH/Dam1c subunits project and contact widely separated sites on the Ndc80c rod and how phosphorylation at previously identified sites might regulate kinetochore assembly.

One Sentence Summary:

A cryo-EM derived structure of the DASH/Dam1 complex suggests how microtubule depolymerization drives chromosome separation.

Kinetochore linkages between chromosomes and spindle microtubules (MTs) persist despite rapid gain and loss of tubulin subunits during cell division (1). The capacity to follow cycles of polymerization and depolymerization without dissociation is an important and puzzling property of the kinetochore-MT interface. The puzzle is particularly evident for “point-centromere” kinetochores of budding yeast, each of which connects a single MT with a small (~125–200 bp) and well-defined centromere (2, 3). Although motor proteins participate during an initial, side-on attachment to MTs (4), the energy for chromosome movement during anaphase comes from the GTPase activity of tubulin and the polymerization-depolymerization events it governs (5–7).

Stable association with the plus end of a MT while the latter lengthens and shortens depends most directly on two essential molecular components of the budding-yeast kinetochore -- the heterotetrameric Ndc80 complex (Ndc80c) and the heterodecameric DASH/Dam1 complex (DASH/Dam1c). Ndc80c is an end-to-end, rod-like assembly of two heterodimers (Ndc80:Nuf2 and Spc24:Spc25), with a ~580 Å long, coiled-coil shaft and small, globular domains at either end (8, 9). It establishes the principal kinetochore axis, by associating with the MT at its Ndc80:Nuf2 end and with chromosome-proximal substructures at its Spc24:Spc25 end. A single kinetochore contains multiple copies of Ndc80c; the most recent estimates from quantitative fluorescence microscopy and biochemistry are in the range 6 to 10 (10–12).

DASH/Dam1c is a more compact structure that assembles into a ring encircling the MT, organizing the Ndc80c rods and potentially other kinetochore components (13, 14). Ring formation under physiological conditions in vitro by recombinant DASH/Dam1c requires MTs, and at high enough concentrations, DASH/Dam1c rings (or helical spirals) decorate the entire MT length. The abundance of the ten subunits in yeast cells indicates that kinetochores in vivo have no more than one or two DASH/Dam1c rings (Table S1) (15). DASH/Dam1c and Ndc80c interact with each other in addition to their separate interactions with MTs (16). Specific phosphorylation regulates most of these contacts, suggesting that a DASH/Dam1c:Ndc80c assembly controls both the timing of end-on attachment and the mechanical coupling of chromosome separation to MT shortening (17, 18). DASH/Dam1c is not present in most metazoans, which instead have the three-protein Ska complex (19). The two complexes appear to have analogous functions (20–22).

Multiple sequence alignments and secondary-structure predictions (23) allowed us to identify likely α-helical regions in the ten DASH/Dam1c subunits and to make corresponding truncations in all of them (Table S2). We prepared multi-gene expression vectors containing these truncated subunits, individually and in various combinations (see Materials and Methods, Fig. S1) and analyzed integrity of the expressed complexes by affinity purification and gel electrophoresis (Fig. S2A, B). We did so for nine fungal species in addition to S. cerevisiae (Table S3, Fig. S2C). A 115 kDa Chaetomium thermophilum DASH/Dam1c construct, comprising the ten subunits, crystallized as hexagonal rods, but Bragg peaks extended only to very low resolution (60–40 Å) (Fig. S2D–J). The apparent unit cell dimensions (Table S4) suggested that the crystallized DASH/Dam1c had formed rings or helical spirals, like those observed when it has assembled around MTs (13, 24, 25). Dynamic light scattering (DLS) and negative-stain EM showed that C. thermophilum DASH/Dam1c assembled into tubes upon incubation with increasing concentrations of monovalent salts (e.g. sodium formate or ammonium formate), consistent with the high-salt conditions that gave the crystals.

For cryo-EM, we locked assembled DASH/Dam1c with bifunctional crosslinkers in high-salt buffer and purified the tubes by size exclusion chromatography in low salt (Fig. S3). The tubes are one-dimensional crystals formed by alternate stacking of DASH/Dam1c rings with 17 protomers (C₁₇ symmetry, Fig. S4). Poor long-range order limited the resolution of an initial helical reconstruction (Fig. S5). We therefore carried out a symmetric reconstruction at 6.5 Å resolution of individual DASH/Dam1c double rings, by masking images and reference volumes (Fig. S6). At this stage, we validated our reconstruction by tilt-pair analysis, which also established the correct enantiomer (Fig. S7). We then independently refined sets of four adjacent protomers in a ring (“asymmetric refinement” in FREALIGN), improving the nominal resolution to 4.5 Å (Fig. S6). Three-dimensional classification did not yield better maps, presumably because selecting particles from stacks had eliminated bad ones and because the rings in a stack have a continuum of small structural distortions or variations.

We built a de novo DASH/Dam1c molecular model into the asymmetric reconstruction. We could obtain a reliable model because the structure consists mostly of α-helices, which are well resolved at 4.5 Å resolution and for which more than one large side chain can establish correct sequence register for the rest of the helix (Fig. S8). After an initial manual build, we used RosettaCM (26) to find low energy conformations that fit the density map and then refined the structure with PHENIX (27). We validated the model by predicting molecular contacts within DASH/Dam1c by direct coupling analysis (DCA) of residue co-evolution (28). For this purpose, we identified and annotated all DASH/Dam1c genes from about 1300 fungal genomes as input for the DCA (see Materials and Methods). Fig. S9 shows residue pairs with the highest coupling scores mapped onto our molecular model. The analysis confirmed the subunit assignment and sequence register. DCA has successfully predicted folded protein structure (29); its application to validating residue assignments is likely to be a powerful tool for interpreting cryo-EM maps, when a suitably large and diverse set of sequences is available. The final model contains amino-acid residues for Ask1 (13–78), Dad1 (18–76), Dad2 (25–95, 109–116), Dad3 (18–82), Dad4 (3–70), Dam1 (53–107), Duo1 (49–121), Hsk3 (22–77), Spc19 (7–112) and Spc34 (3–48, 112–199). Data collection and model statistics are in Table S5.

DASH/Dam1c is a 160 Å long, rod-shaped complex, with two coiled-coil arms that merge in the middle, where a protruding domain extends roughly perpendicular to the arms (Fig. 1). The overall shape agrees well with a published, low-resolution envelope of the S. cerevisiae complex (24). Both coiled-coil domains are parallel five-helix bundles, with the N termini of the subunits at the distal ends of the arms, generally consistent with previous assignments (30). Arm I contains Ask1, Dad2, Dad4, Hsk3 and the N-proximal segment of Spc19; arm II, Dad3, Duo1, Dad1, Dam1 and the N-proximal segment of Spc34. The subunits of the two five-helix bundles interdigitate in the middle of the complex, forming what was called at low resolution a “central domain”. The C-terminal halves of Spc19 and Spc34 form the base and extension of the protrusion domain. Part of the base, formed by Spc34, is not α-helical, less conserved, and likely flexible. We did not attempt to model it (Fig. S10A).

Fig. 1. Cryo-EM structure of the DASH/Dam1c heterodecamer.

(A) Electron cryo-micrograph of DASH/Dam1c helical tubes (1D-crystals) in vitreous ice recorded at 1.9 μm defocus. The scale bar corresponds to 50 nm. 1D-crystals formed by sequential stacking of DASH/Dam1c rings, composed of 17 heterodecamers. (B) Ribbon diagram of the DASH/Dam1c structure with subunits colored: Ask1 (red), Dad2 (green), Dad4 (yellow), Hsk3 (cyan), Spc19 (salmon), Dad3 (firebrick), Duo1 (palegreen), Dad1 (paleyellow), Dam1 (palecyan) and Spc34 (wheat). On top, the view of the complex is along the approximate two-fold symmetry axis. (C) Secondary structure diagram. The same color scheme is used as in (B). α helices are shown as cylinders. First and last residue numbers included in the model, are shown (numbers in parenthesis are the corresponding residues in Saccharomyces cerevisiae). C-terminal residue numbers of all subunits are shown. α helices not modeled in the structure, but inferred from secondary structure prediction, are shown in gray. Zig-zag lines show flexible loops, N- and C-terminal extensions that were not included in the expression construct or observed in the cryo-EM map (functionally important termini of Dam1, Duo1, Ask1 and Spc19:Spc34 are in bold). The description of a previous low-resolution structure called Arm I the “top rod” and Arm II, the “bottom rod” (24).

A layer of buried hydrophobic amino-acid side chains lines the core of the central domain (Fig. S10B), and hydrophobic residues also line the interiors of the two arms. The structure shows that removal of any of the ten subunits would expose hydrophobic patches and probably induce degradation of DASH/Dam1c in the cell. Indeed, omitting Hsk3 from co-expression disrupts arm I, leading to dissociation of Ask1, Dad2 and Dad4, and leaving a hexameric complex of the remaining subunits. Similarly, omitting Dam1 disrupts arm II so that Duo1, Dad1 and Dad3 no longer associate with the complex (31).

Apart from the protrusion domain, the DASH/Dam1c heterodecamer has approximate internal two-fold symmetry relating the left and right arms (Fig. 1B). Each subunit in one arm has a corresponding subunit in the other arm, with similar organization and three-dimensional conformation. Sequence analysis has shown homology between some DASH/Dam1c subunits, (20). Our structure shows that each subunit has a structural paralog, and we can assign the five pairs as Ask1:Dad3, Dad2:Duo1, Dad4:Dad1, Hsk3:Dam1 and Spc19:Spc34. Contemporary DASH/Dam1c may thus have evolved from one or more gene duplication events.

The stacking interactions between rings in our in vitro assembled tubes are probably adventitious contacts resulting from the high protein and high salt concentrations that allow tube formation and crystallization. There is good evidence, however, that the lateral interactions between protomers within a ring, is the physiological mechanism of DASH/Dam1c oligomerization. (i) The same relative heterodecamer orientation (Fig. 2A) is present in a low-resolution reconstruction of a S. cerevisiae DASH/Dam1c dimer (24). (ii) The diameter of the ring is very similar to that of the ring formed by S. cerevisiae DASH/Dam1c (25) and consistent with the requirement that it surround a MT. (iii) Surface amino-acid residues involved in the lateral inter-complex interactions are conserved (Fig. 2B). (iv) We detect coevolution of residues in the inter-complex interface (Fig. 2C, bottom right).

Fig. 2. Structural basis of DASH/Dam1c oligomerization.

(A) Close-up view of two adjacent protomers (labeled decamer 1 and 2) within the DASH/Dam1c ring. Interfaces 1 and 2 are boxed. The subunits are colored as in Fig. 1. (B) Surface residue conservation of DASH/Dam1c in the fungal kingdom. Conservation scores were calculated with ConSurf (46) and mapped onto our structure. The protomers are flipped open, exposing interface 2. Black lines surround the inter-complex binding footprints. The inset shows a close-up view of interface 1 for decamer 1 only. (C) Interface 2 has patches of polar and hydrophobic contacts. Top left, surface representation of the two protomers colored according to the electrostatic potential, calculated with DelPhi (47). Bottom left, close-up view of the polar contact patch. Relevant amino-acids are labeled with residue numbers (S. cerevisiae numbers in parenthesis) and conservation score (small circle, color as in B). Top right, close-up view of the hydrophobic patch. Bottom right, interface molecular contact inferred from DCA of residue coevolution shown as a black line.

There are two interfaces between DASH/Dam1c heterodecamers within the lateral, interprotomer contacts. At interface 2, arm I of one complex (decamer 2) binds the central domain and proximal parts of arm II of an adjacent complex (decamer 1). At interface 1, the base of the protrusion domain of decamer 2 engages the arm II tip of an adjacent decamer 1 (Fig. 2A). From our multiple amino-acid sequence alignments, we calculated DASH/Dam1c residue conservation scores within the fungal kingdom. Fig. 2B shows that the highest conservation scores of surface residues overlap with the footprints of the oligomerization interfaces. At interface 2 (Fig. 2C, top left), we find a tight cluster of charged residues where Ask1 (Asp33) and Dad4 (Arg15) of decamer 2 face Dad3 (Arg59, Glu62) of decamer 1 (Fig. 2C, bottom left); these four residues have the highest conservation scores. Moreover, conserved Pro4 at the tip of the Dad4 helix (decamer 2) contacts a patch of hydrophobic residues (Thr70, Lys73, Ala74, Tyr77) at the end of the Dad3 helix (decamer 1) (Fig. 2C, top right). Because we could not model part of the base of the protrusion domain, folded from Spc34 sequence, we could not analyze specific interactions at interface 1.

Under our experimental conditions, 17 C. thermophilum DASH/Dam1c protomers polymerized into a ring with outer diameter of 560 Å, about twice that of a MT (Fig. 3A). The ring has one protomer more than the 16-membered rings reported for the S. cerevisiae complex (25). Insertion of an additional DASH/Dam1 complex does not substantially change the diameter of the ring (Fig. S11A), which could vary among species. Even within a species, there may be some variation from ring to ring in the number of protomers. When encircling a MT, the gap between the MT lattice surface and DASH/Dam1c is about 60 Å. Oligomerization interfaces 1 and 2 determine the relative orientation of adjacent protomers; ring closure defines their orientation with respect to the MT.

Fig. 3. The DASH/Dam1c ring with flexibly projecting extensions.

(A) Structure of the 17-membered ring. Heterodecamers are show as ribbons and colored as in Fig. 1. Protomers are alternately shown with transparent surface. The MT axis (side view, left) and lattice surface (top view, right) are shown as dashed lines. (B) Close-up views of the anchor points on the ring, where flexibly projecting extension are attached. C_α atoms of terminal amino acids modeled in the cryo-EM map are shown as spheres with residue numbers (S. cerevisiae numbers in parenthesis). Flexible extensions are drawn schematically as dashed lines, except for Spc19/Spc34 (center left panel), for which we indicate a C-terminal heterodimeric coiled coil based on secondary structure prediction and molecular contacts inferred from DCA of residue coevolution.

Near the N terminus of Dam1, phosphorylation of Ser20 (in S. cerevisiae) by Ipl1/AuroraB inhibits cooperative binding of DASH/Dam1c to MTs (17, 32). The location of the Dam1 N- terminus, on the outside of the ring (Fig. 3B, center right), suggests that Ser20 phosphorylation could modulate DASH/Dam1c oligomerization either by destabilizing an interaction with a neighboring protomer or by stabilizing an autoinhibitory interaction within the same protomer. The phosphorylation site is conserved among budding yeast (Data S1), and the length of the Dam1 N-terminal extension could accommodate either option.

Flexible polypeptide extensions that link partner protein complexes are a general feature of kinetochore architecture (33, 34). Apart from ring formation, in which DASH/Dam1c cores interact directly, contacts with both MT and Ndc80c are through subunit extensions predicted to be conformationally variable (Fig. 1C, Data S1). Although these extensions were either omitted from our construct or undefined (owing to flexibility) in the cryo-EM map, the structure of the DASH/Dam1c core and its orientation with respect to the MT together determine the directions in which the flexible extensions project and hence their likely contacts.

The structure of the DASH/Dam1c ring also suggests how to interpret three sets of contact points, detected by chemical crosslinking of Ndc80c with Dam1, Ask1 and Spc34, respectively (35), leading to the interactions illustrated by arrows in Fig. 3B and to the model for the MT-proximal components of the yeast kinetochore shown in Fig 4. The lengths of the C-terminal extensions of Dam1, Ask1, and Spc34 and the directions in which they project might permit all three Ndc80c contacts to be established with a single DASH/Dam1c ring (Fig. 4B). The molecular organization of this assembly is also consistent with the assumption that a DASH/Dam1c ring encircles a MT plus-end with curled protofilaments, as observed in tomographic reconstructions of kinetochore MTs (36).

Fig. 4. Model of the yeast kinetochore-MT interface.

(A) Overview from the side (left) and top (right). (B) Close-up view and interactions of DASH/Dam1c with the MT and Ndc80c. Approximate distances between the attachments of the flexible extensions on the DASH/Dam1c ring and their target sites, defined by crosslinking, are shown. Interactions A, B and C are between Ndc80c and DASH/Dam1 subunits Dam1, Ask1 and Spc34:Spc19, respectively. S. cerevisiae residues known to be phosphorylated by Ipl1/Aurora B are shown in the insets, and arrows indicate release of Ndc80c binding upon phosphorylation.

The C-terminal halves of Dam1 and Duo1 bridge the gap between the DASH/Dam1c ring and the MT and contact the MT surface (31, 37). The points from which the Dam1 and Duo1 extensions emanate are next to each other on the inside of the ring (Fig. 3B, right). Their proximity explains extensive crosslinking of Dam1 and Duo C-terminal extensions with each other (37, 38). Both also crosslink to tubulin at positions on the outer surface of a MT (37). Low-resolution density, likely from Dam1 and Duo1, is present in the reconstruction of a MT-encircling, DASH/Dam1c ring from S. cerevisiae (Fig. S11B) (24, 25). The long Dam1 C-terminal projection in S. cerevisiae crosslinks to Ndc80 (35) and hence has a dual role (Fig. 3B, right). The site of crosslinking, a hairpin at the junction between the Ndc80 globular head and the coiled-coiled shaft, is also the site of a deletion that affects chromosome segregation (18). When Ndc80c attaches to a MT, this loop is close to the MT surface (39). Because of the high symmetry of the DASH/Dam1c ring, however, the same Dam1 subunit need not participate in both MT and Ndc80 contacts at the same time. In some fungal species, including C. thermophilum, the C-terminal extension of Dam1 may not be long enough to extend from a DASH/Dam1c ring to the Ndc80 globular head (Data S1).

The Ask1 C-terminal extension projects from the outside of the ring (Fig. 3B, left). Its contact on the Ndc80:Nuf2 shaft (35), possibly also involving residues of the loop (16), is well within reach of the Ask1 arm, which would emerge from just beneath the Ndc80c shaft as positioned in Fig. 4. The Spc19 and Spc34 C-terminal extensions project from the tip of the protrusion domain. Secondary structure prediction (Data S1) and the second highest peak in our Spc19:Spc34 molecular contact map from DCA suggest that a heterodimeric Spc19:Spc34 coiled coil is flexibly tethered to the structured protrusion domain by two short linkers (Fig. 3B, center left). The observed crosslink toward the C terminus of Nuf2 (35), near the junction with Spc24:Spc25, and binding of S. cerevisiae Spc19 residues 128–165 and Nuf2 residues 399–429 (40), then sets the polarity of the DASH/Dam1c ring with respect to the MT, with the protrusion pointing toward the plus end. As the C-terminal Spc19:Spc34 binding module is anchored to the structured core of the DASH/Dam1c ring through a relatively short linker, the Ndc80:Nuf2 coiled-coil shaft probably bends around the DASH/Dam1c ring, perhaps near the position of the “Ndc80 loop”, bringing the Spc24:Spc25 tip of Ndc80c closer to the kinetochore central axis where it binds the MIND adaptor complex (Fig. 4A).

DASH/Dam1c and Ndc80c are both essential for kinetochore-MT attachments that are stable enough to withstand the forces generated by MT depolymerization (41, 42). A rigid DASH/Dam1c ring that interacts through flexible extensions with the MT and Ndc80c has several potential functional advantages. (i) Junctions created by a flexible peptide extension that fits into a specific docking site can adapt to different relative orientations of the partner complexes, as required by variable stoichiometry and unmatched symmetry in the kinetochore (e.g. a 13-protofilament MT surrounded by a 16- or 17-protomer DASH/Dam1c ring and 6–8 Ndc80c complexes). (ii) These junctions allow distinct interactions at different stages of mitosis: budding-yeast kinetochores first attach to the lateral surface of a MT, along which they migrate toward the spindle pole. DASH/Dam1c is dispensable at this stage, but it is necessary for the transition to stable end-on attachments (4). Adaptability of the DASH/Dam1c contacts with Ndc80c is probably a functional requirement for this process. (iii) Peptide segments facilitate regulation of the assembly state and interaction affinities by post-translational modifications in DASH/Dam1c (17, 43). Unfolded and extended polypeptides are good substrates for kinases and other modifying enzymes, as they fit readily into the enzymatic active sites. Ipl1/Aurora B phosphorylation of DASH/Dam1c subunits switches off all three sets interactions with Ndc80c (Fig. 4B) (35) in order to reset MT attachments during activation of the spindle checkpoint. (iv) Docked-peptide interactions are generally of modest strength, so that individual contacts dissociate and rebind rapidly, allowing a disassembling agent, such as the Ipl1 kinase, to access the DASH/Dam1c peptide segments for modification and to release the kinetochore from the MT. (v) Multiple weak contacts, flexible distance constraints, and incommensurate symmetries enable the organized assembly of Ndc80 and DASH/Dam1c to track the plus end of a depolymerizing MT by stochastic, “hand-over-hand” biased diffusion. The role of DASH/Dam1c ring as a structural organizer of Ndc80c is central to this process. (vi) Coordinated structural changes in the cage-like structure of the kinetochore, made possible by flexibility of the DASH/Dam1c:Ndc80c interactions (and of linkages in other parts of the assembly), may be part of the tension-sensing mechanism implicated in setting and releasing the mitotic spindle assembly checkpoint. Compliance of the overall structure can reposition components in response to tension and convert force into changes in molecular proximity.

The molecularly detailed model of the kinetochore-MT connection in budding yeast (Fig. 4), derived from the DASH/Dam1c and Ndc80c structures and from the constraints of published biochemical and genetic data, unifies two proposals for how the energy released by MT depolymerization drives chromosome separation at anaphase (44, 45). One model invokes biased diffusion, in which rapid, asynchronous association and dissociation of multiple, flexible contacts allows any individual contact to break and form again further along the shortening MT. The other invokes a conformational wave (e.g., protofilament curling) that actively pushes the kinetochore in the direction of shortening. Protofilment curling against a closed ring probably exerts force directly, but MT contacts, from the Dam1:Duo1 extensions and from the organized set of Ndc80 complexes, must break and reform to allow MT shortening to progress. Thus, docking and undocking of flexible connections between defined subunits can allow a kinetochore to track the end of the MT through cycles of growth and shrinkage.