Molecular basis for inner kinetochore configuration through RWD domain–peptide interactions

Abstract Kinetochores are dynamic cellular structures that connect chromosomes to microtubules. They form from multi‐protein assemblies that are evolutionarily conserved between yeasts and humans. One of these assemblies—COMA—consists of subunits Ame1CENP‐U, Ctf19CENP‐P, Mcm21CENP‐O and Okp1CENP‐Q. A description of COMA molecular organization has so far been missing. We defined the subunit topology of COMA, bound with inner kinetochore proteins Nkp1 and Nkp2, from the yeast Kluyveromyces lactis, with nanoflow electrospray ionization mass spectrometry, and mapped intermolecular contacts with hydrogen‐deuterium exchange coupled to mass spectrometry. Our data suggest that the essential Okp1 subunit is a multi‐segmented nexus with distinct binding sites for Ame1, Nkp1‐Nkp2 and Ctf19‐Mcm21. Our crystal structure of the Ctf19‐Mcm21 RWD domains bound with Okp1 shows the molecular contacts of this important inner kinetochore joint. The Ctf19‐Mcm21 binding motif in Okp1 configures a branch of mitotic inner kinetochores, by tethering Ctf19‐Mcm21 and Chl4CENP‐N‐Iml3CENP‐L. Absence of this motif results in dependence on the mitotic checkpoint for viability.

We could not obtain substantial amounts of soluble full-length K. lactis Ame1-Okp1 from expression of an Okp1-Ame1 dicistron.
Appendix Figure S4 SEC of reconstituted K. Vanderwaltozyma polyspora (XP_001645198.1), Zygosaccharomyces rouxii (XP_002498459.1), Lachancea lanzarotensis (GenBank data base accession code: CEP61777.1), Candida glabrata (GenBank data base accession code: KTA95770.1). Through analyses of our mass spectra of samples from our limited proteolysis of K. lactis COMA-Nkp1-Nkp2, we found Nkp1 and Nkp2 to be the least protease-sensitive proteins in COMA-Nkp1-Nkp2 (Tables EV1;2; Fig  EV4). With our limited proteolysis experiments, we found that Nkp1-Nkp2 is especially protease-sensitive in the C-terminal part of Nkp1 and Nkp2, in the vicinity of residue 150 in Nkp1, and residue 116 in Nkp2 (Table EV3). Residues around these sites of Nkp1 and Nkp2-possibly in projecting loops-are probable contacts for Okp1-Ame1 C-termini, because the regions surrounding these sites are stabilized in COMA-Nkp1-Nkp2 (see A or B). Nkp1-Nkp2 spontaneously proteolysed in Nkp1-Nkp2; but was less prone to proteolysis when associated with COMA/Okp1-Ame1 (see Fig 1A; Appendix Fig S1A; Table EV1). With mass spectrometry, we found the following fragments of Nkp1 and Nkp2-the result of spontaneous proteolysisin our purified Nkp1-Nkp2 sample: Nkp1 residues 1-60; and lower amounts of Nkp1 fragments with residues 1-57, 25-174 or 25-175; for Nkp2: 1-68, and a lower amount of an Nkp2 fragment with residues 1-67. E) Representative plots showing deuterium-exchanged peptides, after 10 sec, 60 sec, or 1200 sec of deuterium exchange of Nkp1 alone or Nkp1 in Nkp1-Nkp2. Nkp1 residues 1-60 are flexible in Nkp1, but stabilized in Nkp1-Nkp2. Our observations suggest that the Nkp1 N terminus contacts Nkp2. The Nkp1 C-terminal part, which becomes structured upon binding of Nkp1-Nkp2 to COMA, is partially ordered in Nkp1 alone, but is not ordered in Nkp1-Nkp2, suggesting conformational changes in Nkp1 upon binding to Nkp2. F) Representative SEC chromatogram, and image of SDS-PAGE gel with fractions from principal SEC peak, of an Nkp1-Nkp2 variant with N-terminally truncated Nkp1 (residues 13-210), and C-terminally truncated Nkp2 (residues 1-143); L: sample loaded on column. The N-terminal 12 residues of Nkp1 and C-terminal 8 residues in Nkp2 are dispensable for Nkp1-Nkp2 formation. But we found that Nkp2 (without C terminal 19 residues) did not co-purify with polyhistidine-tagged Nkp1 without its N terminal 30 residues (data not shown), when we co-expressed coding regions for these variants; consistent with our suggestion (see E) that the Nkp1 N terminus contacts Nkp2.     Proteolysis sites from our limited proteolysis Based on our deuterium-exchange data of Ctf19-Mcm21, we re-assigned electron density for a protein fragment that is not directly connected with the globular Ctf19-Mcm21 D-RWD domains in our previously reported crystal structure of fulllength K. lactis Ctf19-Mcm21 (Schmitzberger & Harrison, 2012), which we had determined with X-ray diffraction data to 3.9 Å resolution (PDB code 3ZXU). We previously interpreted this density as residues 69-87 of Ctf19. Our deuteriumexchange data, however, show that Ctf19 residues 36-55 are ordered in Ctf19-Mcm21 (see D). We thus think it is more plausible this density corresponds to residues 36-54.

A Okp1
Appendix Figure S6  , that we show as blue mesh, and that we calculated after molecular replacement, automated rebuilding, and refinement of the Ctf19-Mcm21 D-RWD domains only-without modeling any Okp1 residues (thus unbiased for Okp1 fragment) or water molecules. For map calculation, we included X-ray diffraction data to 2.1 Å. We show the Ctf19-Mcm21 D-RWD domains as secondary structure cartoon-diagrams; and some sidechains in stick representation. Viewing perspective is similar as for Fig 6D. Two Ctf19-Mcm21-Okp1 assemblies are the asymmetric unit in our crystals. From our diffraction data, we modeled most of the Ctf19-Mcm21 D-RWD domains, except for a few residues at their C-termini, for which no substantial electron density is present. D) Image of silver-stained SDS-PAGE gel with fractions from our Ctf19 107-270 -Mcm21 108-293 -Okp1 295-360 crystals. We washed crystals in three separate drops with crystallization solution. Fractions from these drops are designated W1, W2, W3. After our washes, we dissolved crystals in SDS sample-buffer (lane designate Cr.); a fraction of purified Ctf19 107-270 -Mcm21 108-293 -Okp1 295-360 we show as control. E) Secondary structure cartoon-diagrams of the superposed structures of the D-RWD domains of Ctf19-Mcm21 (PDB code: 3ZXU; orange) and Ctf19-Mcm21 bound with Okp1 (our study; PDB code: 5MU3; colours are as for Fig 6D). We superposed the structures using the main-chain atoms of Ctf19 D-RWD -Mcm21 D-RWD . A few residues we show in stick representation. The relative arrangement of the two Ctf19-Mcm21 dimers that form the asymmetric unit in our Ctf19 D-RWD -Mcm21 D-RWD -Okp1 295-360 crystals differs from that of our previously reported crystal structure of full-length Ctf19-Mcm21 (Schmitzberger & Harrison, 2012). This observation suggests that the crystal contacts for both structures are not biologically relevant; i.e. for COMA multimerization. We observe structural differences between the two Ctf19-Mcm21 structures mostly in the C-terminal RWD domains (RWD-C). Because the Mcm21 D-RWD domains are structurally more similar between the two structures, and the differences for Ctf19 RWD-C are present for both molecules of the asymmetric unit, we assume the change in conformation in Ctf19 RWD-C with Okp1 bound relative to Ctf19 that is not bound with Okp1, is due to Okp1 binding. This assumption is consistent with the change in the hydrogen-bonding network of Ctf19 RWD-C that our hydrogen-deuterium exchange data show (Fig 6B).

A B
Increasing elution volume  showing that no substantial amount of Okp1_cmΔ-Ame1 co-purifies with Ctf19-Mcm21. Gel lane was deliberately overloaded with sample (to potentially see residual Okp1_cmΔ-Ame1). Migrating position of full-length Okp1 in full COMA is indicated (compare with gel image in Fig 1A). B,C) Representative images of Western blots of immunoprecipitated fractions from S. cerevisiae extracts with Okp1 versions C-terminally tagged with six flag epitopes (6×flag) and Ctf19 C-terminally tagged with four myc epitopes (4×myc). We show blots of extracts from S. cerevisiae that have either full-length Okp1 (fl) or Okp1 that lacks the Ctf19-Mcm21 binding motif (cmΔ). We used magnetic Protein G coupled beads that were not coated with anti-flag antibodies as a control. Images of Ponceau S-stained nitrocellulose membranes that we used for Western blots we show below the respective blot images. Full-length Ctf19 was present in lower amount from S. cerevisiae extracts with our Okp1_cmΔ mutant than from clones with Okp1_fl. We could not produce substantial amounts of recombinant K. lactis Ctf19-Mcm21 D-RWD domains in the absence of Okp1 (variants). We conclude from these observations and the Ctf19-Okp1 contacts in our structure that Ctf19 and Okp1 stabilize each other. We prepared extracts from our native laboratory S. cerevisiae clone (S288C type; identifier DDY904 in Table EV6) without additional genetic modifications, or clones with an Okp1 version C-terminally tagged with six flag epitopes (Okp1-6×flag) and Mcm16 C-terminally tagged with thirteen myc epitopes (Mcm16-13×myc), after metaphase growth-arrest with nocodazole. We show blots of extracts from S. cerevisiae that have either Okp1_fl or Okp1_cmΔ. Filled circles indicate presence of Okp1_fl-6×flag, Okp1_cmΔ-6×flag, or Mcm16-13×myc. Mcm16-13×myc from the cell-extract fraction migrated at a lower position on our SDS-PAGE gel than Mcm16-13×myc that immunoprecipitated (and was thus purified) with Okp1_fl; an effect that we observed frequently.   Table EV6), Okp1_cmΔ with genomically encoded OsTIR1 (Okp1_cmΔ OsTIR1), a native clone with OsTIR1 (OsTIR1), two unique clones of Okp1_fl with OsTIR1 and Mad1-degron-9×myc (mad1-degron-9×myc OsTIR1 Okp1_fl), or two unique clones of Okp1_cmΔ with OsTIR1 and Mad1-degron-9×myc (mad1-degron-9×myc OsTIR Okp1_cmΔ), on solid raffinose-galactose agar with 1 mM of the auxin analogue 1napthylic acetic acid (NAA), which induces degradation of Mad1-degron-9×myc. F) Representative Western-blot image of S. cerevisiae extracts from cultures grown in liquid raffinose-galactose medium at 30 °C of a native laboratory S. cerevisiae (native; S288C type; identifier DDY904 in Table EV6), or clones with mad1-degron-9×myc OsTIR1 Okp1_fl (identical clone as the one further towards the top of the two that we show growth of in E), or with mad1-degron-9×myc OsTIR1 Okp1_cmΔ (identical clone as the one further towards the top of the two that we show growth of in E). We prepared samples from cultures either from before addition of NAA or from specific time points after addition of NAA. We used extracts from a S. cerevisiae clone with Ndc80-13×myc as a control (lane 1). We show image of Ponceau Sstained nitrocellulose membrane used for Western blot below the blot image. The higher level of Mad1-degron-9×myc, after addition of NAA, in extracts from Okp1_cmΔ with OsTIR1 and Mad1-degron-9×myc, compared with that in extracts from Okp1_fl with OsTIR1 and Mad1-degron-9×myc, indicates up-regulation of transcription of mad1-degron-9×myc or translation of Mad1-degron-9×myc in Okp1_cmΔ cells. This observation is consistent with our growth assays on solid medium that we show in E. The growth phenotype of mad1-degron-9×myc OsTIR1 Okp1_cmΔ is not as pronounced as that of our mad1Δ Okp1_cmΔ clone (see Fig  8C), indicating that mitotic checkpoint activity is still present. G) Evaluation of chromosome mis-segregation in meiosis. Typical images of haploid spores from tetrad dissection of homozygous diploid S. cerevisiae with Okp1_fl/Okp1_fl, Okp1_cmΔ/Okp1_cmΔ, or ctf19_1-954Δ/ctf19_1-954Δ, after two days of growth on solid YPD agar, showing spore viability after meiosis. We placed the four spores from tetrads column-wise. Perhaps reflecting a differential centromere-localization

. K Q E S R R I L A E R H F Q N I N R K L E Y A L E V Q R G K L A K E H
Ame1 segment 1 K. lactis Ame1 S. cerevisiae Ame1 V. polyspora Ame1 Z. rouxii Ame1 E. gossypii Ame1 L. thermotolerans Ame1 C. glabrata Ame1

Chemicals, reagents, and consumables
For our experiments, we used high-grade chemicals from Sigma-Aldrich (Merck KGaA), Merck Millipore (Merck KGaA), Thermo Fisher Scientific, Bio-Rad Laboratories, Carl Roth, Bernd Kraft, or PanReac AppliChem, unless specified otherwise. Plasmid and PCR-product purification reagents, and DNA-purification columns were from Qiagen, Epoch Life Science, or Carl Roth.

Molecular cloning and plasmid preparation for recombinant protein production
We amplified DNA coding-regions for Ame1, Ctf19, Mcm21, Nkp1, Nkp2, or Okp1 by polymerase chain-reaction (PCR) from isolated Kluyveromyces lactis genomic DNA or isolated Saccharomyces cerevisiae genomic DNA, and prepared polycistronic constructs for Ame1-Okp1 variants, Ctf19-Mcm21-Okp1 variants, COMA variants, or Nkp1-Nkp2 as previously described (Schmitzberger & Harrison, 2012). We inserted PCR products into a ligation-independent cloning compatible pET3aTR expression plasmid, or into a similar plasmid encoding an N-terminal polyhistidine tag (His; with sequence MKSSHHHHHHENLYFQSNA), with a tobacco-etch virus (TEV) protease cleavage-site. We prepared polycistronic plasmids encoding K. lactis COMA variants with Ame1 variants or Okp1 variants, with the order ( ; and Ame1-Okp1 variants with the order His-Okp1, Ame1, by overlapping PCR as described (Schmitzberger & Harrison, 2012). We prepared a dicistron that encodes K. lactis Nkp1-Nkp2 or S. cerevisiae Nkp1-Nkp2, with the coding-region order His-Nkp1, Nkp2, and a polycistron encoding K. lactis Nkp1-Nkp2 variants with truncations of Nkp1 or Nkp2 analogously. We carried out site-directed mutagenesis of Okp1 in our K. lactis COMA-encoding polycistron to generate Okp1_cmΔ with PfuTurbo DNA Polymerase AD (Agilent Technologies) and deoxynucleotide mix from QuikChange Multi Site-directed Mutagenesis Kit (Thermo Fisher Scientific). We used restriction enzymes from New England BioLabs or Fermentas/Thermo Fisher Scientific; and T4 DNA Polymerase (LIC-qualified), for ligation-independent cloning, from Novagen/Merck Millipore.
Coding regions for COMA proteins from S. cerevisiae were inserted in plasmid pST39 (Tan, 2001) with the order of coding regions Mcm21, Ctf19, Ame1, and Okp1; and encoding a polyhistidine tag with a precision-protease cleavage site that is N-terminal of Ame1 (with sequence MGSSHHHHHHSLEVLFQGPH). The polycistron for production of S. cerevisiae Ame1-Okp1 was as described (Hornung et al, 2014); this construct encodes Okp1 followed by Ame1 with a C-terminal polyhistidine tag (HHHHHH).
For construct details and plasmids, see Table EV5. Our plasmids and PCR primer-sequences are available upon request.
We stained our SDS-PAGE gels with a solution containing 0.06 % (w/v) Coomassie Brilliant Blue R. If we stained gels with silver stain, we used Silver Stain Plus Kit (Bio-Rad Laboratories).
Appendix: RWD-domain interactions at the inner kinetochore We purified K. lactis Nkp1-Nkp2 (Appendix Fig S1A) and our Nkp1-Nkp2 truncation variant (Appendix Fig S5F) as we describe above. We produced Nkp1 alone from our Nkp1-Nkp2 polycistron-Nkp1 was present in excess in our purifications. For Nkp1 and Nkp1-Nkp2 samples that we used for deuteriumexchange that we show in Appendix Fig S5E, lysis buffer was with cOmplete, EDTA-free Protease Inhibitors Cocktail tablets (one tablet for 50 ml buffer; Roche Diagnostics) and ~ 1 mM PMSF instead of the general recipe that we describe above. We separated Nkp1 from Nkp1-Nkp2 on a HiTrap Q HP column (Nkp1 eluted in the flow-through), and purified it as in our general purification procedure that we describe above.
We purified Ame1 1-260 -Okp1 123-336 that we used for our deuterium-exchange experiments (Fig 6A; Appendix Fig S6A,C) or our Ame1 1-260 -Okp1 108-308 variant (Appendix Fig S4B) as in our general procedure that we describe above, with the following modification. For SEC, we used a buffer with 25 mM Hepes pH 7.5, 500 mM NaCl, 5 % (v/v) glycerol, 0.5 mM TCEP. Using a higher NaCl concentration, than in our general purification procedure, reduced aggregation.
We purified samples of Ctf19 107-270 -Mcm21 108-293 -Okp1 108-336 and Ctf19 107-270 -Mcm21 108-293 -Okp1 123-336 that we show in Fig 4D, as in our general procedure that we describe above, with the following modifications. We pooled the flow-through fraction from chromatography on a HiTrap Q HP column, and subsequently purified it with cation-exchange chromatography on a 5 ml HiTrap SP HP column, with a buffer of 30 mM Hepes pH 8.1, 100 mM NaCl, 10 % (v/v) glycerol, 0.5 mM TCEP and a 100-1000 mM NaCl gradient.
K. lactis MIND core (MIND-C1; with Dsn1 residues 230-479, Mtw1 residues 1-233, full-length Nnf1, full-length Nsl1) was reconstituted and purified as previously described (Dimitrova et al, 2016), in a buffer of 30 mM HEPES, pH 7.5, 200 mM NaCl, 1 mM TCEP. In the same buffer, we reconstituted K. lactis COMA-MIND, for our mass-spectrometry experiments that we show in Fig EV1A, by combining purified COMA and purified MIND-C1 samples, which we had stored on ice after purification, and eluting from a SEC column (Superdex 200 10/300; GE Healthcare Life Sciences).
We affinity-purified K. lactis COMA that lacks the Ctf19-Mcm21 binding motif (Okp1_cmΔ; Appendix Fig S8A) with lysis buffer with cOmplete, EDTA-free Protease Inhibitor Cocktail tablets (one tablet for 50 ml buffer; Roche Diagnostics) and 1 mM PMSF (Sigma-Aldrich), instead of the protease inhibitors in our general procedure that we describe above.
For Ni 2+ -affinity-based binding assays (see below), we did not cleave the polyhistidine tag of our purified proteins, and proceeded directly from Ni 2+ -affinity chromatography to anion-exchange chromatography. We purified recombinant S. cerevisiae COMA according to our general procedure that we describe above, but without TEV-protease cleavage and without 2 nd Ni 2+ -affinity chromatography. S. cerevisiae Chl4-Iml3 was prepared as previously described (Hinshaw & Harrison, 2013).
We produced S. cerevisiae Okp1-Ame1-His in lysogeny broth (LB) medium. We purified recombinant S. cerevisiae Okp1-Ame1-His or S. cerevisiae His-Nkp1-Nkp2, for our binding experiments (see Fig EV3B), similar to our general proteinpurification protocol that we describe above with some modifications. For purification of S. cerevisiae Okp1-Ame1-His or S. cerevisiae His-Nkp1-Nkp2, we resuspended the cell pellet in a buffer of 20 mM Na-phosphate pH 7.5, 500 mM NaCl, 20 mM imidazole, or in a similar lysis buffer as previously described (Hornung et al, 2014). After sonication and centrifugation, we incubated our cleared lysates with Ni 2+ -NTA, washed Ni 2+ -NTA with a buffer of 20 mM Na-phosphate pH 7.5, 500 mM NaCl, 20 mM imidazole; and eluted with a buffer of 20 mM Na-phosphate pH 7.5, 500 mM NaCl, 200 mM imidazole. We purified pooled Ni 2+ -NTA eluates of S. cerevisiae Okp1-Ame1-His or S. cerevisiae His-Nkp1-Nkp2 with a Superdex 200 10/300 column on an Äkta FPLC system, and with a buffer of 20 mM Na-phosphate pH 6.8, 200 mM NaCl, 2.5 % (v/v) glycerol, 0.5 mM TCEP. We concentrated eluted protein fractions, flash froze the concentrate in liquid N 2 , and stored it at −80 °C, until usage.
We stained our SDS-PAGE gels with a solution containing 0.06 % (w/v) Coomassie Brilliant Blue R.

Analytical size-exclusion chromatography
We carried out analytical SEC for samples that we show in Fig 5A,B and Fig EV3A on a Superdex 200 10/300 column on an Äkta FPLC (GE Healthcare Life Sciences) system, at 4 °C. For SEC that we show chromatograms of in Fig 5A,B, we used an Nkp1-Nkp2 sample that after purification we had stored at −80 °C; and samples of Ctf19-Mcm21-Okp1 variants that after purification we had stored on ice. We loaded ~ 4 nanomole of each protein assembly in a volume of ~ 500 µl. We incubated combinations of protein samples on ice for ~ 3000 sec, before injecting on the column. For this experiment and most of our other analytical SEC experiments, we eluted with a buffer of 24 mM or 25 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol, 0.5 mM TCEP. We incubated combinations of samples of different protein assemblies for the chromatograms that we show in Fig EV3A on ice for ~ 3600 sec, before injecting the combined sample on the column. For the SEC that we show chromatograms of in Fig EV3A, we used a sample of full-length Ctf19-Mcm21 that after purification we had stored at −80 °C, and an Nkp1-Nkp2 sample that after purification we had stored on ice. We injected 12-25 nanomole of our protein Appendix: RWD-domain interactions at the inner kinetochore samples in a volume of ~ 100 µl. For the SEC that we show in Appendix Fig S7B, we used a Superdex 200 HiLoad 10/600 prep grade column, and we used protein samples that we had stored on ice for a few days after purification. We injected samples in a volume of 2 ml. We combined 159 nanomol of Ctf19 D-RWD -Mcm21 D-RWD -Okp1 295-383 sample with 123 nanomole Nkp1-Nkp2 sample (in a volume of ~ 1 ml), and incubated this mixture for ~ 600 sec on ice before loading it on the column. Ctf19 D-RWD -Mcm21 D-RWD -Okp1 295-383 co-eluted with Nkp1-Nkp2 (Appendix Fig S7B); in a buffer of 25 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol, 0.5 mM TCEP, but dissociated from Nkp1-Nkp2 after cryo-freezing in liquid N 2 and storage at −80 °C. For our sample analysis that we show in Fig EV3B, we carried out SEC with a Superdex 200 PC 3.2/300 column (GE Healthcare Life Sciences) on an Ettan LC system (GE Healthcare Life Sciences), with a buffer of 20 mM Na-phosphate pH 6.8, 200 mM NaCl, 2.5 % (v/v) glycerol, 0.5 mM TCEP; we incubated combinations of protein samples on ice for ~ 3000 sec, before we injected them on the column; we injected protein samples at a concentration of ~ 5 µM in a volume of 50 µl (we estimated concentrations with Bradford spectrophotometric assay, with Protein Assay Dye Reagent Concentrate (Bio-Rad Laboratories), with bovine serum albumin as standard).

Limited proteolysis experiments, and mass spectrometry of resultant protein fragments or proteins
We used purified protein samples, which we had stored on ice after purification, for our limited proteolysis experiments with trypsin (Sigma-Aldrich) or elastase (Worthington Biochemical Corporation).
For limited proteolysis without subsequent SEC, we incubated ~ 910 picomole of COMA-Nkp1-Nkp2 (Table EV2) in a molar ratio of 10:1 with elastase or trypsin, in a buffer of 25 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol, 1.5 mM Lmethionine for 180 sec or 600 sec, at 22-25 °C. In a similar buffer with 2.5 mM Lmethionine, we incubated ~ 5.8 nanomole of K. lactis Nkp1-Nkp2 in a molar ratio of 10:1 with elastase or trypsin, for 300 sec or 600 sec, at 22-25 °C (Table EV3). We stopped our limited proteolysis reactions by addition of solid high-grade guanidine hydrochloride until saturation of the solution.
David S. King analysed protein fragments with mass spectrometry by electrospray ionization with Fourier-transform ion resonance-cyclotron or ion-trap mass spectrometers at the Howard Hughes Medical Institute mass spectrometry facility (University of California, Berkeley); and analysed mass spectra. He analogously analysed our purified Ctf19 107-270 -Mcm21 108-293 -Okp1 variant samples Appendix: RWD-domain interactions at the inner kinetochore (Figs 4D;5A,B), or other protein samples, such as Nkp1-Nkp2, which we had not actively proteolysed.

Binding assays with in vitro translated proteins
Coding regions, as PCR products or in plasmids, for proteins for in vitro translation with a Kozak translation-initiation sequence were prepared as previously described (Hinshaw & Harrison, 2013). We produced S 35 -labeled proteins, either from PCR products (for experiments with K. lactis COMA) or, for experiments with S. cerevisiae COMA, from pET3aTR plasmids, by in vitro translation in rabbit-reticulocyte lysate (TnT lysate systems; Promega Corporation) with S 35 -labeled L-methionine, following the manufacturer's instructions. For each binding experiment, we incubated S 35labeled proteins with 10 µg or 15 µg of purified K. lactis COMA (Fig EV6A), which after purification we had stored on ice or at −80 °C, in a buffer of 15 mM Hepes pH 7.5, 110 mM NaCl, 10 mM imidazole, 2.5 % (v/v) glycerol, 0.05 % (v/v) octyl phenoxypolyethoxylethanol (Nonidet P-40), 0.5 mM TCEP (IVT buffer) for 3600 sec on ice. After addition of 30 µl of Ni 2+ -NTA slurry (equilibrated in IVT buffer) to this mixture, we incubated under rotation for 1800 sec at 4 °C. We used Ni 2+ -NTA purified polyhistidine-tagged maltose binding protein, which we had stored at −80 °C after purification, as a negative control; and 10 % (v/v) of the reaction mixture (from sample for maltose-binding protein) as positive control. After washing Ni 2+ -NTA resin three times with 500 µl of IVT buffer, we eluted proteins by addition of 25 µl of SDS sample-buffer with 500 mM imidazole, and denatured eluted proteins by heating for 10 min at 100 °C. We separated denatured proteins on 10-20 % (w/v) acrylamide Criterion Tris-HCl or TGX gels (Bio-Rad Laboratories). After drying gels, we transferred signals to a phosphor-imaging plate, and recorded phosphorescence signals on a personal molecular imager (Bio-Rad Laboratories). For binding experiments with S. cerevisiae COMA (Fig EV6B), our protocol was similar. For each binding experiment, we used 15 µg of recombinant purified S. cerevisiae COMA, which we had stored at −80 °C after purification; our IVT buffer was with 20 mM Hepes pH 7.5, 150 mM NaCl, 20 mM imidazole, 5 % (v/v) glycerol, 0.05 % (v/v) Nonidet P-40, 0.5 mM TCEP.

Nanoflow electrospray-ionization mass spectrometry of protein assemblies
For our nanoflow electrospray-ionization mass-spectrometry (for review see (Sharon & Robinson, 2007)), we used purified protein samples that we had flash-frozen in liquid N 2 , and had stored at −80 °C and on solid CO 2 . We transferred proteins to a buffer of 200 mM ammonium acetate pH 6.7-7.3, with Amicon concentrators. We sprayed samples usually at a concentration of 2-10 µM, unless specified otherwise. We determined concentrations with a PicoDrop UV/VIS spectrophotometer. We acquired mass spectra or tandem mass-spectra (Benesch et al, 2006), in positive ion mode, on a high mass quadruple time-of-flight (Q-TOF)-type instrument (Sobott et al, 2002) adapted for a QSTAR XL platform (MDS Sciex) (Chernushevich & Thomson, 2004). To spray samples, we used in-house prepared gold-coated glass capillaries (Nettleton et al, 1998). Optimized instrument parameters were as follows: ion-spray voltage 1300 Volt, declustering potential 100 Volt, focusing potential 200 Volt and collision energy up to 200 Volt, MCP 2350 Volt. In tandem mass-spectrometry, we selected the relevant m z -1 range in the second quadrupole. The proteins in that range were subjected to acceleration in the collision cell. We used argon as a collision gas at maximum pressure. We calibrated all spectra externally using a cesium-iodide solution (100 mg ml -1 ). We derived mean values and standard deviations for masses (see inset tables in our figures) with the MassLynx software (Waters Corporation), selecting masses from a series of identified m z −1 peaks in our spectrum. We carried out partial, in-solution disruption of purified K. lactis COMA (Fig 2A,B; Appendix Fig S2A) by adding acetic acid to a concentration of 5 % (v/v) in 100 mM ammonium acetate, to a pH 4.0. After addition of acetic acid, we incubated this mixture on ice for 900 sec, before spraying the solution in the mass spectrometer. For our analysis of Ctf19-Mcm21-Okp1 that we show in Appendix Fig  S2B, we analogously incubated COMA with 100 mM ammonium acetate pH 3.7, 5 % (v/v) acetic acid. For our analyses of purified samples of COMA (Figs 1C;EV1B) or COMA-MIND (Fig EV1A) we sprayed in 200 mM ammonium acetate pH 6.7-7.3. We sprayed COMA-Nkp1-Nkp2 (Figs 1B;EV2A,B) in 200 mM ammonium acetate pH 7.4.
Hydrogen-deuterium exchange of proteins followed by mass spectrometry of peptides For our deuterium-exchange experiments, we used purified protein samples that we had flash-frozen in liquid N 2 , and had stored −80 °C and on solid CO 2 . In the following, we describe experiments and analyses of proteins in COMA and COMA-Nkp1-Nkp2. The procedure for our other samples was similar, unless stated otherwise.
We constructed a reference list of pepsin proteolysed K. lactis COMA-Nkp1-Nkp2 peptide masses using a non-deuterated sample as follows. We diluted 5 µl aliquot of protein stock (1.5-3 mg ml −1 (estimated from absorbance at 280 nm) in 24 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol, 0.5 mM TCEP) in a 1:10 ratio, by adding 45 µl of a buffer with 20 mM Tris, pH 8.0, 150 mM NaCl (H 2 O Reaction buffer). We acidified this sample by mixing with 10 µl of 2 M glycine pH 2.5 (H 2 O Stop Buffer). The sample was digested online using an immobilized pepsin column (Poroszyme, Applied Biosystems, Thermo Fisher Scientific) with 0.07 % (v/v) formic acid in water as mobile phase (flow rate 3.33 µl sec -1 ). Digested peptides were passed over a C18 trapping column (ACQUITY BEH C18 VanGuard Pre-column, Waters Corporation), and subsequently over a reversed-phase chromatography column (ACQUITY UPLC BEH C18 column, Waters Corporation) with a 6-40 % (v/v) gradient of acetonitrile in 0.1 % (v/v) formic acid at 0.66 µl sec −1 using the nanoACQUITY Binary Solvent Manager. Total time of a single run was 810 sec. All fluidics, valves, and columns were maintained at 0.5 °C, using the HDX Manager (Waters Corporation). The pepsin column was kept at 13 °C inside the temperaturecontrolled digestion compartment of the HDX manager. The C18 VanGuard Precolumn outlet was coupled directly to the ion source of a SYNAPT G2 HDMS mass spectrometer (Waters Corporation). For our experiments comparing assemblies COMA and COMA-Nkp1-Nkp2 (Fig 3A,B; Appendix Fig S3A,B,C,E), the mass spectrometer was working in ion-mobility mode (Pringle et al, 2007). For our other deuterium-exchange experiments, we did not use ion-mobility mode. Leucineenkephalin solution (Sigma-Aldrich) was used as a lock mass. For protein identification, mass spectra were acquired in MSE mode over the m z -1 range of 50-2000. We used the following spectrometer parameters; electrospray-ionization: positive mode, capillary voltage: 3 kV, sampling cone voltage: 35 V, extraction cone voltage: 3 V, source temperature: 80 °C, desolvation temperature: 175 °C, desolvation-gas flow: 222 ml sec -1 . Peptides were identified using ProteinLynx Global Server software (Waters Corporation). We input the experimental data of Appendix: RWD-domain interactions at the inner kinetochore identified peptides that included mass to charge (m z -1 ), charge, retention time, and ion-mobility drift time in the DynamX software program (version 2.0; Waters Corporation).
We carried out hydrogen-deuterium exchange experiments essentially as described (Kupniewska-Kozak et al, 2010) for our non-deuterated samples, with our Reaction buffer containing D 2 O (99.8 % (v/v); Armar Chemicals); we adjusted pH read (uncorrected meter reading) using DCl or NaOD (Sigma-Aldrich). After mixing 5 µl protein stock with 45 µl D 2 O Reaction buffer, the exchange reactions proceeded for 10 s, 60 sec, or 1200 sec, at 22-25 °C. We quenched deuterium-exchange reactions by reducing the pH read to 2.5, through adding the reaction mixture into a microcentrifuge tube that contained Stop buffer (2 M glycine pH read 2.5), which we had cooled on ice. Immediately after quenching, we manually injected the sample into the nanoACQUITY (Waters Corporation) Ultra Performance Liquid Chromatography (UPLC) system. We carried out pepsin digestion, liquid chromatography, and mass-spectrometry analyses as described for our nondeuterated samples. We carried out two kind of control experiments to determine experimental in-exchange and back-exchange values, as previously described (Kupniewska-Kozak et al, 2010). Briefly summarized, to determine the minimum exchange of peptides-our in-exchange control, we added D 2 O reaction buffer to Stop buffer cooled on ice. We added this mixture to protein stock and immediately pepsin digested it, before we carried out liquid chromatography coupled to massspectrometry analysis, as we describe above. We calculated the deuteration level of peptides from our in-exchange control experiment using DynamX. We used values from this analysis as the minimum exchange ( !" ! ; see next paragraph). For our determination of back-exchange peptide values, we combined 5 µl protein stock with 45 µl of D 2 O Reaction buffer, incubated this mixture ~ 15-19 hrs at 4 °C, before we combined it with Stop buffer, and analyzed it analogously as our in-exchange control. We used determined peptide deuteration-values from our back-exchange experiment as the maximum deuterium exchange ( !" !"" ). For each one of our different (protein assembly) sample comparisons, we used respective uniform in-exchange and backexchange controls. We repeated all deuterium-exchange and in-exchange and backexchange control experiments at least three times.
We calculated peptide deuteration-levels with DynamX (version 2.0), using our pepsin-proteolysed peptide list obtained from the ProteinLynx Global Server. We selected peptides in DynamX with the following acceptance criteria: minimum intensity threshold 2000, minimum products per amino acids 0.2. We analysed isotopic envelopes from deuterium-exchange with DynamX with the following parameters: RT deviation ± 15 sec, m z −1 deviation ± 12.5 parts per million (ppm), drift time deviation ± 2 time bins (if applicable). We manually verified each isotopic envelope, which had been identified and assigned to peptides by the automated analysis with DynamX, to ensure valid calculation of the mean peptide masses from our deuterium-exchange experiment ( !" ) and from our in-exchange and backexchange experiments ( !" ! and !" !"" ). We discarded ambiguous or overlapping isotopic envelopes from further analyses. We exported our selected deuteriumexchange data to Excel (Microsoft Corporation), and calculated the hydrogendeuterium exchange mass shifts and fraction of exchange. We calculated the fraction of peptide deuterium-exchange (f) with the formula: !"" -!" ! Appendix: RWD-domain interactions at the inner kinetochore We calculated mean values and standard deviations for the exchange fraction (f) from at least three independent experiments. Most of our standard deviations were below 2 % of our mean determined deuteration fractions. We visualized our data with a previously described Excel macro template file (Black et al, 2007) that plots deuterium-exchanged peptide representations that are coloured by fraction of deuterium exchange, underneath their corresponding position in a linear amino-acid sequence representation of the protein the respective peptides originate from.

Dynamic light scattering measurements and multi angle laser-light scattering measurements
For our light scattering measurements, we used protein samples that we had stored on ice after purification. We measured dynamic light-scattering data of K. lactis COMA at a concentration of 21 µM (assuming a COMA dimer), and of K. lactis Nkp1-Nkp2 or K. lactis COMA-Nkp1-Nkp2 at similar concentrations, after filtration through a 0.22 µm Ultrafree-MC GV filter, in a quartz cuvette on a Dynapro instrument (Wyatt Technology) with a laser of 826.2 nm, at 15 °C. Multiple monomodal auto-correlation function were fit in cumulant-expansion analysis mode for our data with the Dynamics software version 5.26.56 (Wyatt Technology) (see Appendix Fig S2C).
For multi angle laser-light scattering, we injected our protein samples onto a Superdex 200 10/300 SEC column (GE Healthcare Life Sciences) that was mounted on a high performance liquid-chromatography system (1260 Infinity LC; Agilent Technologies). For samples eluting from this column at a flowrate of 6.7 µl sec −1 , at 22-25 °C, we measured multi angle laser-light scattering data with a laser of 663.9 nm on a Dawn Heleos-II detector (Wyatt-846-H2; Wyatt Technology) and refractive indices with a laser of 658 nm with an Optilab T-rEX instrument (Wyatt-512-Trex; Wyatt Technology). Before injection on the column, we filtered protein samples with 0.22 µm Ultrafree-MC GV filters. For measurements of purified K. lactis COMA (Fig  1D), we injected 100 µl at ~ 43 µM, and eluted with a buffer of 30 mM Hepes pH 7.5, 200 mM NaCl, 1 mM TCEP, 0.4 % (v/v) NaN 3 . Injecting lower amounts of COMA did not yield interpretable multi angle laser-light scattering signals. Our multi angle laserlight scattering data of COMA-Nkp1-Nkp2 under similar solution conditions were inconclusive, with respective to the sample's multimeric state (the calculated molar mass was between that for monomeric COMA-Nkp1-Nkp2 and dimeric COMA-Nkp1-Nkp2). For measurements of purified K. lactis Nkp1-Nkp2 (Appendix Fig S1B), our measurement conditions were similar to those for COMA. We injected 70 µl of sample at ~ 120-150 µM, and eluted with a buffer of 25 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol 0.5 mM TCEP, 0.4 % (v/v) NaN 3 . We used bovine serum albumin for calibration (detector-signal normalization, peak broadening). We analysed our multi angle laser-light scattering data with the Astra software (version 6.1.5.22; Wyatt Technology), with protein concentrations determined from in-line refractive index measurements, and using a dn/dc value of 0.185 ml g −1 (a refractive index of 1.33 was chosen for the aqueous solution). We fit our light scattering data with a first order Zimm function with linear regression, as implemented in the Astra software.

Sedimentation-equilibrium analytical ultracentrifugation analyses
For our sedimentation-equilibrium ultracentrifugation analyses, we used protein samples that we had stored on ice after purification. We recorded sedimentation equilibrium analytical ultracentrifugation data with a ProteomeLab Optima XL-I Appendix: RWD-domain interactions at the inner kinetochore analytical ultracentrifuge (Beckman Coulter) and an An-60 Ti rotor (Beckman Coulter) equipped with a 12-mm wide Epon six-chamber double-sector sample cell. Our protein samples were in a buffer of 25 mM Hepes pH 7.5, 300 mM NaCl, 5 % (v/v) glycerol, 0.5 mM TCEP. We used three different protein concentrations corresponding to an absorbance at 280 nm of 0.25, 0.5, and 0.75, and estimated protein concentrations with the protein-sequence based theoretical extinction coefficients at 280 nm (Gasteiger et al, 2005). We centrifuged our K. lactis COMA sample sequentially at 8000, 10000, 15000, 23000 rotations per minute (rpm) at 4 °C (Appendix Fig S1C). We centrifuged our K. lactis Ctf19-Mcm21 sample sequentially at 9000, 12500, 15000, 18000, 22000 rpm at 4 °C (Appendix Fig S1D). To determine sedimentation, we measured the absorbance at 280 nm every four hours during centrifugation. We fit our Ctf19-Mcm21 data with a global fit (with data from three different concentrations), with a 'Single Species of Interacting System' model, without molecular mass conservation constraints in SEDPHAT. We fit our data of COMA with a global fit (with data from three different concentrations) with a 'Monomer-Dimer Self-Association' model, fixed the molecular mass to that corresponding to monomeric COMA, left molecular mass conservation-constraints out, and allowed the K a to vary during the fit calculation in SEDPHAT. Our sedimentation-equilibrium analytical ultracentrifugation data are available upon request.
We generated S. cerevisiae clones (native, Okp1_fl, or Okp1_cmΔ) that are compatible for auxin-inducible degradation of a target protein, by first integrating in their Ura3 genomic locus the Oryza sativa TIR1 coding region (encoding the F-box transporter inhibitor response 1 protein)-OsTIR1 under a galactose-inducible promoter, as described (Nishimura et al, 2009). To do so, we transformed clones with plasmid BYP7434 (Kubota et al, 2013) (see Table EV5) that we had digested with restriction enzyme StuI. After verifying OsTIR1 genomic integration, we genetically modified mad1 so that it encodes Mad1 that is fused at its C terminus to an auxin-inducible degron and nine myc epitopes.
For a list of S. cerevisiae clones that we used in our study, see Table EV6. Clones (from storage at −80 °C) and PCR primer-sequences for our S. cerevisiae genetic modifications are available upon request.
Our growth experiments with S. cerevisiae clones that encode Mad1 with an auxin-inducible degron (Appendix Fig S9E) were done on solid agar with 2 % (w/v) raffinose and 2 % (w/v) galactose, and 1 mM of the synthetic auxin analogue 1napthylic acetic acid (NAA; from Carl Roth; dissolved in 5 M NaOH); similarly as described (Nishimura et al, 2009).