Transient excited states of the metamorphic protein Mad2 and their implications for function

Abstract

The spindle checkpoint complex is a key surveillance mechanism in cell division that prevents premature separation of sister chromatids. Mad2 is an integral component of this spindle checkpoint complex that recognizes cognate substrates such as Mad1 and Cdc20 in its closed (C-Mad2) conformation by fastening a “seatbelt” around short peptide regions that bind to the substrate recognition site. Mad2 is also a metamorphic protein that adopts not only the fold found in C-Mad2, but also a structurally distinct open conformation (O-Mad2) which is incapable of binding substrates. Here, we show using chemical exchange saturation transfer (CEST) and relaxation dispersion (CPMG) NMR experiments that Mad2 transiently populates three other higher free energy states with millisecond lifetimes, two in equilibrium with C-Mad2 (E1 and E2) and one with O-Mad2 (E3). E1 is a mimic of substrate-bound C-Mad2 in which the N-terminus of one C-Mad2 molecule inserts into the seatbelt region of a second molecule of C-Mad2, providing a potential pathway for autoinhibition of C-Mad2. E2 is the “unbuckled” conformation of C-Mad2 that facilitates the triage of molecules along competing fold-switching and substrate binding pathways. The E3 conformation that coexists with O-Mad2 shows fluctuations at a hydrophobic lock that is required for stabilising the O-Mad2 fold and we hypothesize that E3 represents an early intermediate on-pathway towards conversion to C-Mad2. Collectively, the NMR data highlight the rugged free energy landscape of Mad2 with multiple low-lying intermediates that interlink substrate-binding and fold-switching, and also emphasize the role of molecular dynamics in its function.

Introduction

The single-sequence-structure-function paradigm in structural biology postulates that amino acid sequences of proteins code for a unique three-dimensional structure that is necessary for performing function. Metamorphic proteins are exceptions to this paradigm and exist in two or more conformations of comparable stability that interconvert in the absence of ligands or cofactors [1–4]. Since the discovery of the two distinct structural forms of Mad2 in 2004 [5, 6], the class of metamorphic proteins has expanded to include proteins such as KaiB, a core protein in the cyanobacterial circadian clock [7], IscU, a component of the iron-sulphur cluster biogenesis pathway [8], the chemokine lymphotactin [9] and the Escherichia coli virulence factor RfaH [10].

Wild-type (wt) Mad2 is a dimer made up of 23.5 kDa protomers that each adopt a HORMA (Hop1p, Rev7 and Mad2) fold [11–13]. The core of this fold consists of three α-helices A, B and C, a β-hairpin between helices A and B, and a three-stranded antiparallel β-sheet consisting of strands 4-6 (Figure 1A-C, grey). This core remains unchanged between the two metamorphic forms C- and O-Mad2. In contrast, a substantial part of the protein stretching over 61 amino acids including the N-terminus (M1-S16, blue), the seatbelt (Y156-S178, red) and the C-terminal β-hairpin (strands 8’,8”, residues E179-K200, green) undergoes rearrangement during fold-switching of C-Mad2 to O-Mad2. The disordered N-terminus and the first two turns of helix A in C-Mad2 become strand 1 in O-Mad2 that hydrogen bonds with strand 5 of the core. The C-terminal β-hairpin unravels into a long strand (strand 8) that moves to the other side of the core β-sheet, while the seatbelt elongates strand 6 and extends into another short strand (strand 7) that hydrogen bonds to the newly formed strand 8 (Figure 1A, 1B). The interconversion of O-Mad2 to C-Mad2 is very slow and takes ~ 9 hrs, and the equilibrium ratio of O-Mad2 to C-Mad2 is estimated to be 1:8 (303 K, pH 6.8) with C-Mad2 being the more stable conformer [5, 14, 15].

Figure 1. Metamorphosis of Mad2 in the spindle checkpoint complex.

A-C) Structures of C-Mad2 (PDB ID: 1S2H [5]), O-Mad2 (PDB ID: 1DUJ [82]) and C-Mad2 bound to the Mad2-binding-peptide 1 (MBP1, PDB ID: 1KLQ [19]). The core of the protein which remains unaltered between O- and C-Mad2 is shown in grey. Residues from the N-terminal β-strand in O-Mad2 (blue) dissociate from strand 5 and extend the helix A in C-Mad2. This paves the way for the C-terminal β7-β8 hairpin, which is hydrogen bonded to strand 6 in O-Mad2, to move over and hydrogen bond to strand 5 in C-Mad2. During this process, the C-terminal segment of strand 7 unfolds and extends the long loop (seatbelt) connecting strands 8’ and 8”. The substrate-bound conformation of Mad2 closely resembles C-Mad2, with the peptide inserting between the seatbelt and strand 6. D) Mad2 molecules produced by ribosomal biosynthesis exist predominantly as O-Mad2. In the presence of unattached kinetochores during metaphase, Mad1 molecules associate with the kinetochores and recruit C-Mad2. The Mad1-C-Mad2 complex at the kinetochore then catalyses the conversion of O-Mad2 to C-Mad2 via the formation of a hetero-oligomeric Mad1-C-Mad2-O-Mad2 complex. The newly generated C-Mad2 then binds Cdc20 and inhibits the anaphase-promoting-complex (APC) [16].

Mad2 is a key constituent of a cell surveillance mechanism called the spindle checkpoint complex that arrests premature mitotic cell division. Prior to the metaphase-anaphase transition, sister chromatids attach themselves to the mitotic spindle through their kinetochores. All kinetochores must be bound before the separation of chromatids commences, and incomplete attachment can cause aneuploidy in the daughter cells. C-Mad2 is recruited to unattached kinetochores by binding to its substrate Mad1, while the pool of unbound Mad2 in the cell exists predominantly as the O-conformer. Mad1-bound C-Mad2 then forms a heterodimer with O-Mad2 and drives the interconversion of O-Mad2 to C-Mad2. The newly generated C-Mad2 is then free to bind and sequester Cdc20 in order to delay the onset of anaphase (Figure 1D) [5, 16–18]. Both the substrates Mad1 and Cdc20 bind C-Mad2 by inserting their binding motifs into the seatbelt, where the substrate polypeptide adopts a strand conformation and hydrogen bonds with strand 6 of the Mad2 core (Figure 1C) [19].

Mad2 is an example of the complex interplay that occurs between the conformational free energy landscape of a protein and its function, and much of this interplay remains uncharacterized at the molecular level. How metamorphic proteins such as Mad2 interconvert efficiently without falling into kinetic traps and subsequently misfolding or aggregating remains poorly understood. Mad2 is a central hub in the spindle checkpoint mechanism, where it recognizes Mad1 associated with kinetochores, sequesters Cdc20 and also facilitates the fold-switching of O-Mad2 to C-Mad2. Whether all these functions are performed by the native state of C-Mad2, or whether Mad2 is able to deploy low-lying conformations on the free energy landscape in order to perform some of these functions remains an open question. The role of dynamics in Mad2 function has already been anticipated in literature, where the model hypothesized for the insertion of substrate into the binding cleft involves transient opening of the seatbelt region [20–22]. However, there is currently no experimental evidence supporting the role of dynamics in regulating the function of Mad2.

The link between dynamics and function has been strengthened by recent breakthroughs in NMR spectroscopy which have enabled the detection of sparsely and transiently populated (excited) biomolecular conformations [23]. Experiments such as chemical exchange and dark state exchange saturation transfer (CEST and DEST) [24–27], Carr-Purcell-Meiboom-Gill (CPMG) [28–30] and off-resonance rotating frame (R1ρ) relaxation dispersion [31–33], and paramagnetic relaxation enhancement (PRE) [34] are sensitive to states populated to as low as 0.5 % with lifetimes (τ_E) varying between 10^-4 to 10^-1 s. The application of these methods has illustrated the importance of low-lying transiently populated states (excited states) in protein folding and maturation [35–39], DNA binding [40], light sensing [41], signal transduction [42], enzyme activity [43, 44], riboswitch function [45], and DNA base repair activity [46].

Here, we detect the existence of millisecond timescale conformational fluctuations in both C- and O-Mad2 using CEST and CPMG NMR experiments. These fluctuations occur in regions of the protein that adopt different conformations in C- and O-Mad2, while the core of the protein that remains intact in both metamorphic structures is rigid on the millisecond timescale. Two distinct exchange processes are observed in C-Mad2; the first process results in the transient insertion of the N-terminus of one molecule of C-Mad2 into the seatbelt region at the substrate binding site of another C-Mad2 molecule. The second process occurs at the interface of helix C and the β-hairpin composed of strands 8’ and 8”, as well as in the seatbelt region. This process likely leads to the disengagement of the β-hairpin and the seatbelt from the protein core prior to substrate binding. The N-terminus is also dynamic in the O-Mad2 topology and populates an excited state where its interaction with helix C is disrupted. Our data highlights the presence of pervasive conformational frustration in regions of Mad2 that are involved in both substrate recognition and fold-switching. This frustration generates as many as three thermally accessible conformations with populations between 2-10 % and millisecond lifetimes. We propose that these excited states establish vital links between metamorphosis and function in Mad2 by possessing structural features that are in-between the various endpoints of the Mad2 functional cycle.

Results

R133A Mad2 is monomeric and undergoes fold-switching

Full-length wild-type (wt) Mad2 is a 205-amino acid protein with a molecular weight of 23.5 kDa. In solution, Mad2 is predominantly a dimer (Figure S1). In order to characterize the dynamics of Mad2 without interference from canonical dimerization, we used the monomeric R133A variant of Mad2 (Mad2*) (Figure S2) [47]. Mad2* retains the ability to interconvert between the two metamorphic states, and homogeneous samples of O- and C-Mad2* can be prepared using anion exchange chromatography [5].

Circular dichroism spectra of O- and C-Mad2* show minima around 208 and 222 nm, and a positive excitonic band between 190 and 200 nm, indicating that both forms are folded with comparable secondary structure (Figure 2A, 2D). The fluorescence excitation and emission maxima of the three Trp residues in O- and C-Mad2* are also comparable, reflecting the similarity in the environments of W75, W100 and W167 (located in helix B, strand 5 and the seatbelt respectively) in both conformations (Figure 2B, 2E). Size exclusion chromatography multi-angle light scattering (SEC MALS) data confirm that both O- and C-Mad2* are monomers with molecular weights of 23.9 ± 0.3 kDa and 21.7 ± 0.1 kDa respectively (Figure 2C and 2F). Finally, both O- and C-Mad2* also give high quality ¹H-¹⁵N HSQC NMR spectra with well-resolved peaks in both the 1H and ¹⁵N dimensions characteristic of globally folded proteins (Figure 2G-H). However, the two 1H-¹⁵N HSQC spectra are very different from each other with the largest differences localizing to the regions of fold-switching, highlighting the metamorphic nature of Mad2 (Figure S3).

Figure 2.

Biophysical characterization of O- and C-Mad2*. Far-UV circular dichroism (A,D) Trp fluorescence excitation (red) and emission (green) (B,E) spectra, SEC-MALS profiles (C,F) and 1H-¹⁵N HSQC spectra (G,H) of O-Mad2* (A, B, C and G) and C-Mad2* (D, E, F and H).

Millisecond dynamics in C-Mad2* observed in ¹⁵N-CEST and ¹⁵N-CPMG profiles

In order to search for higher free energy alternate conformations of Mad2*, we ran ¹⁵N chemical exchange saturation transfer (CEST) [24] and Carr-Purcell-Meiboom-Gill relaxation dispersion (CPMG) [48, 49] experiments on C-Mad2*. CPMG data are sensitive to conformational exchange in the microsecond-millisecond timescale (k_ex ~ 200 – 10000 s^-1), while routine applications of CEST detect exchange with rate constants ranging between 5-500 s^-1. CEST has also been used to study characterize exchange (up to k_ex/Δω = 5) in both nucleic acids and proteins for rate constants as large as 11000 s^-1 [50, 51].

In the ¹⁵N-CEST experiment, ¹⁵N z-magnetization is irradiated with a weak radiofrequency (RF) field of amplitude B¹ for an exchange duration T_EX and the amount of z-magnetization is quantified at the end of T_EX. The chemical shift offset at which this B₁ field is applied is systematically varied across the entire ¹⁵N chemical shift range of interest and a ¹H-¹⁵N correlation spectrum is acquired for each offset position. Resonance intensities (I) in these correlation spectra are quantified and normalized against the corresponding intensities in a spectrum acquired using a T_EX value of 0 (I₀). Residue-specific CEST profiles are then constructed as the ratio I/I₀ as a function of the offset at which the weak RF field is applied. CEST profiles in systems lacking detectable exchange show a single ‘dip’ in intensity caused by saturation of the major state resonance. On the other hand, slow exchange manifests as a second ‘dip’ in intensity of the major state peak that occurs at the ¹⁵N chemical shift of the residue in the excited state because the perturbation of excited state magnetization by the weak B1 field is transferred to the major state via conformational exchange [24, 25].

In the ¹⁵N-CPMG experiment, τ_cp-180°-τ_cp trains are applied on transverse in-phase or anti-phase ¹⁵N magnetization for an exchange duration T_EX and the residual amount of transverse magnetization at the end of T_EX is quantified through a ¹H-¹⁵N correlation spectrum. For each residue, the effective transverse relaxation rate constant (R_{2, eff}), which is a measure of the broadening resulting from μs-ms timescale exchange, is calculated from the peak intensity (I) in the correlation spectrum as $R_{2,eff}=\frac{1}{T_{EX}}\ln (\frac{I}{I_0}).$ Here, I₀ is the peak intensity in a spectrum acquired with T_EX = 0 (I₀). Residue-specific CPMG profiles are then generated by plotting R_{2, eff} against the CPMG pulsing frequency $(v_{CPMG}=\frac{1}{4{\tau }_{CP}}).$ Conformational exchange on the μs-ms timescale appears in CPMG profiles as a decrease in R_{2, eff} with increasing CPMG pulsing frequency, because higher frequency of τ_CP-180°-τ_CP trains are able to quench exchange broadening more efficiently [28, 49, 52].

The ¹⁵N-CEST profiles of residues Gln4, Leu5 and Ser6 show clear signatures of millisecond timescale conformational exchange with isolated dips in intensity distinct from the saturation dip present in each CEST profile (Figure 3A). Exchange is also visible in CPMG profiles for these residues and the small magnitude of R_ex (= 5 s^-1) suggests that the exchange is slow (Figure 3B). In order to quantify the minor state population and lifetime, we fit the two B₁-field CEST and one B₀-field CPMG data for these three residues globally to a two-state Bloch-McConnell model of conformational exchange [53, 54]. The data fit well to the two-state model with a ${\chi }_{red}^2$ value of 1.29. The population of the excited state (p_E1) is 3.5 ± 0.3 %, while the lifetime $({\tau }_{E1}=\frac{100}{(100-p_{E1})k_{ex,CE1}})$ is 6.6 ± 0.1 ms. Errors were determined using Monte Carlo analysis and the corresponding distributions for p_E1 and k_ex,CE1 are shown in Figures S4A and S4B. We next constructed 1D (Figure S4C-D) and 2D (Figure 3C) ${\chi }_{red}^2$ surfaces to evaluate the reliability of the thermodynamic and kinetic parameters determined from ¹⁵N-CEST and ¹⁵N-CPMG data. Although the ${\chi }_{red}^2$ surface for k_ex,CE1 has a minimum, the minimum is shallow for p_E1 suggesting that only k_ex,CE1 can be extracted reliably and not p_E1. In addition, the p_E1/k_ex,CE1 2D ${\chi }_{red}^2$ surface shows that p_E1 and k_ex,CE1 are inversely correlated, indicating that the least-square minimization algorithm can compensate a higher value of p_E1 by reducing k_ex,CE1 while maintaining the same fit quality (Figure 3C).

Figure 3. Detection and structural characterization of the excited state E1 in equilibrium with C-Mad2*.

A) ¹⁵N-CEST (47.5 Hz) and B) ¹⁵N-CPMG profiles of residues Q4, L5 and S6 showing signatures of conformational exchange. Solid lines are global fits of ¹⁵N-CEST and ¹⁵N-CPMG data to the two-state Bloch-McConnell equations. Dotted lines in panel A indicate positions of the ground (C-Mad2*) and excited (E1) states. The chemical shift difference between the two is indicated in each CEST profile in panel A. C) 2D ${\chi }_{red}^2$ surfaces depicting the change in fit quality from the best fit as a function of the population of the excited state (p_E1) and the exchange rate constant between the ground and the first excited state (k_ex,CE1). The allowed values of p_E1 and k_ex,CE1 fall approximately in the region of the surface bounded by the red contour. D) Structure of C-Mad2 (PDB ID: 1S2H [5]) showing the nitrogen atom of residues changing in conformation in E1 as yellow spheres. The colour scheme of C-Mad2 is the same as in Figure 1A. E) Comparison of the ¹⁵N chemical shift differences extracted from ¹⁵N-CEST and ¹⁵N-CPMG data (purple) with those expected for the O-Mad2↔C-Mad2 transition (green). F) Structure of dimeric wild-type C-Mad2 (PDB ID 2VFX [20]) showing three monomers from the crystal structure involved in intermolecular interactions. While the monomer on the right-hand side binds to the central monomer via the canonical dimer interface, the N-terminus of the monomer on the left-hand side forms a short β-strand and inserts into the seat-belt of the monomer in the middle. (Inset) Zoomed-in view of the seatbelt region of the central monomer. The inserted peptide is depicted both in cartoon and stick representations. Backbone hydrogen bonds tethering the peptide to the seatbelt and strand 6 are shown as yellow lines. G) ¹⁵N-CEST profiles of L5 in apo (red) or MBP1 peptide-bound (green) C-Mad2*. Solid lines are fits to the two-state Bloch-McConnell equations while the dashed lines indicate the chemical shift positions of the ground (C-Mad2*) and excited (E1) states. E1 is destabilized in peptide-bound C-Mad2* and cannot be detected in ¹⁵N-CEST data.

We noticed that the best-fit parameter estimation procedure allows solutions with large p_E1 and small k_ex,CE1 with minimal penalty in ${\chi }_{red}^2$(Figure 3C). Resonances for minor states with p_E > ~10 % and k_ex < ~200 s^-1 are generally visible as distinct peaks in 2D ¹H-¹⁵N correlation spectra and exchange between them can be detected using the magnetization transfer NMR experiment. In this experiment, cross-peaks are observed between ¹H-¹⁵N correlations (diagonal peaks) belonging to each state of a particular residue and the relative intensity of the cross-peak is dependent on p_E and k_ex, as well as the exchange duration T_EX and the longitudinal relaxation rate constants (R₁) of the two states [55]. In order to get tighter bounds on the populations and rate constants, we ran a ¹⁵N magnetization exchange difference experiment [56] on C-Mad2*, in which only diagonal and cross-peaks from resolved resonances exchanging magnetization during T_EX remain in the 2D ¹H-¹⁵N correlation spectrum and all other resonances are subtracted out. The 2D exchange spectrum of C-Mad2* does not show any peaks above noise level for a T_EX of 400 ms. We next simulated magnetization exchange profiles using pairs of (p_E1, k_ex,CE1) that are allowed solutions $(\text{Δ}{\chi }_{red}^2\le 1)$ from fits of ¹⁵N-CEST and ¹⁵N-CPMG data and used the signal-to-noise of the exchange spectrum, as well as the intensity of the major state resonance, to determine which pairs of (p_E1, k_ex,CE1) will give rise to observable exchange cross-peaks. These p_E1, k_ex,CE1 pairs (for eg. (20 %, 150 s^-1) and (1 %, 600 s^-1)) were then eliminated from the list of possible solutions for p_E1, k_ex,CE1 to give a final estimate of the bounds for p_E1 and τ_E1 for the minor state (E1) of C-Mad2* as 2-10 % and 2-32 ms (k_ex,CE1 = 35 – 500 s^-1). The allowed values of p_E1 and k_ex,CE1 correspond approximately to the region defined within the red contour in Figure 3C.

The N-terminus of one molecule of C-Mad2* inserts into the substrate binding site of a second molecule in the excited state E1

We next enquired into the molecular identity of the excited state E1. CEST and CPMG data are valuable aids in determining excited state structures because they provide residue- and nucleus-specific chemical shift differences (Δϖ_GE) between the ground and excited conformations. The N-terminus of C-Mad2*, which is the site of the exchange process leading to E1 (Figure 3D), is also a region that participates in O-Mad2↔C-Mad2 fold-switching. Accordingly, we first compared the Δϖ_CE1 values for Q4-S6 to the differences expected if C-Mad2 was adopting a O-Mad2-like conformation at the N-terminus. Δϖ_CE1 values are much larger than the chemical shift differences expected from the O-Mad2↔C-Mad2 interconversion (since both O- and C-Mad2 are disordered at the N-terminus), confirming that E1 is not a O-Mad2-like conformation in the Q4-L6 region (Figure 3E). The N-terminus of C-Mad2 is also far from the native dimer interface, confirming that E1 is not a transient canonical dimer formed by the dimerization-deficient R133A mutant.

The Δϖ_CE1 values for E1 are large, -4.7 ppm for Q4, 6.5 ppm for L5 and 8.6 ppm for S6. This suggests that the N-terminus of C-Mad2 may be undergoing a change in secondary structure in E1 from its random coil native conformation. In particular, the signs of the two largest Δϖ_CE1 tend towards the values expected for a random coil↔β-strand conversion [57]. Intriguingly, the crystal structure of wt C-Mad2 (PDB ID: 2VFX [20]) shows the existence of intermolecular contacts between two C-Mad2 dimers, where the N-terminus of one C-Mad2 molecule occupies the substrate binding site of another C-Mad2. The N-terminal insert adopts a β-strand conformation in this structure and Q4, S5 and L6 form backbone hydrogen bonds with W167 and E168 in the seatbelt, as well as with I155 and T157 in β6 of the partner protomer (Figure 3F) [20].

In order to test whether E1 is formed because of two C-Mad2* molecules associating via the substrate binding site of C-Mad2*, we carried out experiments on a C-Mad2*-substrate complex using Mad2-binding peptide 1 (MBP1, SWYSYPPPQRAV) which mimics C-Mad2 substrates [19]. We first compared the ¹⁵N chemical shifts of E1 with those of MBP1 bound to C-Mad2* available from literature (BMRB ID: 5299). Only two sequence positions of C-Mad2*, Q4 and L5, are directly comparable with their counterparts in MBP1 (S4 and Y5), as position 6 in MBP1 is a Pro residue. Binding to C-Mad2* results in similar patterns of chemical shift perturbation in the two substrates; position 4 experiences a small negative Δϖ and position 5, a large positive Δϖ compared to the free substrate, with the magnitude of Δϖ being approximately two-fold smaller for MBP1 binding to C-Mad2* (Figure S5).

Next, we probed the existence of E1 in the C-Mad2*-MBP1 complex. Size exclusion chromatography data confirms that C-Mad2* remains a monomer when bound to MBP1 (Figure S6). The ¹H-¹⁵N HSQC spectrum of a sample containing 180 μM C-Mad2* and 360 μM MBP1 matches with the spectrum of MBP1-bound C-Mad2* from literature (Figure S7A) [19]. There are distinct resonances for the MBP1-bound form and no residual intensity for free C-Mad2*, confirming that C-Mad2* is fully bound to the peptide in this sample (Figure S7B-D). The ¹⁵N-CEST profile of L5 in the C-Mad2*-MBP1 complex has only a single dip in intensity and does not show the presence of the minor dip seen in apo C-Mad2*, demonstrating that the excited state E1 has been destabilized by substrate binding (Figures 3G and S7E). MBP1 can therefore compete with the N-terminus of C-Mad2* for binding to the substrate recognition site of C-Mad2* and this provides direct evidence that the excited state E1 is formed due to the N-terminus of C-Mad2* mimicking a substrate and binding in the seatbelt region of another C-Mad2* molecule.

It should be noted that E1 is an asymmetric homodimer of C-Mad2* and each nucleus shuttles between three chemical environments during dimerization, one in the monomer and one each in the two protomers of the asymmetric dimer. However, the chemical environments of Q4, L5 and S6 in the monomer are very similar to those in the protomer of the asymmetric dimer whose substrate binding site is occupied. Therefore, we expect the chemical shifts and relaxation rate constants for Q4, L5 and S6 at these two sites to be comparable and different from the values at the third site where the N-terminus has inserted into the seatbelt region of a second protomer. In this limit, three-state asymmetric dimerization can be modelled as a two-state process [58] in which the parameters for dimerization are given in terms of the population (p_E1) and the rate constant (k_ex,CE1):

$$K_d=\frac{A_T{(1-p_{E1})}^2}{p_{E1}(1+p_{E1})}$$

$$k_{on}=\frac{k_{ex}{p}_{E1}(1+p_{E1})}{A_T(1-p_{E1})}$$

and

$$k_{off}=k_{on}{K}_d$$

where A_T is the total Mad2 monomer concentration (A_T = 133 μM). Using the best-fit parameters for p_E1 (3.5 ± 0.3 %) and k_ex,CE1 (155 ± 18 s^-1), the K_d for asymmetric dimerization is calculated to be 3.4 mM, while the association and dissociation rate constants are 4.4 ± 0.3 x 10⁴ M^-1 s^-1 and 152 ± 17 s^-1 respectively.

Conformational fluctuations at the β-hairpin 8’-8” are distinct from N-terminal conformational exchange

In addition to the dynamics at the C-Mad2* N-terminus, the ¹⁵N-CPMG profiles of many residues in the β8’-β8” hairpin, helix C and the seatbelt show evidence of μs-ms timescale fluctuations (Figures 4A and S8) and the R_ex values for these residues ranges from 3.5 to 16.8 s^-1. The CPMG profiles of all these residues can be modelled globally using the two-state Bloch-McConnell equations [53, 54] with an overall ${\chi }_{red}^2$ of 0.95, suggesting that they all belong to the same conformational exchange event (II) and report on a single excited state E2. The best-fit parameters for k_ex,CE2 and p_E2 are 544 ± 146 s^-1 and 6 ± 2 % respectively.

Figure 4. Detection and structural characterization of excited state E2 in equilibrium with C-Mad2*.

A) ¹⁵N-CPMG profiles of residues sensing the fluctuations of C-Mad2* (ground state) to the excited state E2. Solid lines are fits of the ¹⁵N-CPMG data to the two-state Bloch-McConnell equations. B) 2D ${\chi }_{red}^2$ surfaces depicting the change in fit quality from the best fit as a function of the population of the excited state (p_E2) and the exchange rate constant between the ground and the second excited state (k_ex,CE2). C) Structure of C-Mad2 (PDB ID:1S2H [5]) showing the nitrogen atom of residues with significant R_ex in ¹⁵N-CPMG profiles (E2) as yellow spheres. The colour scheme of C-Mad2 is the same as in Figure 1A. D) ¹⁵N-CPMG profiles of residues showing exchange with E2 in apo (red) or MBP1 peptide-bound (green) C-Mad2*. Solid lines are fits to the two-state Bloch-McConnell equations. E2 is destabilized in peptide-bound C-Mad2* but can still be detected in ¹⁵N-CPMG data. E) Comparison of ¹⁵N-CPMG profiles of T136 in C-Mad2* (180 μM) in the presence of 360 μM (green) and 600 μM (orange) MBP1 peptide. F) Bar plot depicting the residue-specific decrease in R_ex detected in ¹⁵N-CPMG profiles of C-Mad2* (red) in the presence of 360 μM (green) and 600 μM (orange) MBP1.

Processes I and II occur in distinct regions of Mad2* and are unlikely to correspond to the same exchange event (Figure S9). In order to establish whether E1 and E2 result from the same exchange process, we first fit the ¹⁵N-CEST (two B₁ fields) and ¹⁵N-CPMG profiles (700 MHz) of Q4, L5 and S6 together with the ¹⁵N-CPMG profiles for process II (Figures S9A and S9B). The ${\chi }_{red}^2({\chi }_{red}^2=1.52)$ increases significantly compared to separate fits of process I $({\chi }_{red}^2=1.29)$ and process II $({\chi }_{red}^2=0.95),$ highlighting the incompatibility of the thermodynamics and kinetics between the two processes. Next, we fixed p_E2 and k_ex,CE2 to the best-fit values derived from modelling process I (p_E1 = 3.5 % and k_ex,CE1 = 155.4 s^-1) and fit the CPMG profiles for process II. The resulting fits are poor and the ${\chi }_{red}^2$ increases to 1.85, compared to 0.95 when p_E2 and k_ex,CE2 are allowed to float without constraints (Figure S9C). Fitting the same CPMG profiles using other pairs of allowed solutions of p_E1 and k_ex,CE1 (Figure 3C, red contour) also does not improve the fit quality (${\chi }_{red}^2$ varies from 2.2-3.2) (Figure S9D). Similarly, the ${\chi }_{red}^2$ for modelling CEST and CPMG data of process I increase considerably (from 1.3 to 6.2) when p_E1 and k_ex,CE1 are fixed to the populations extracted from fitting CEST profiles of process II, and the quality of the fits is visually poor (Figures S9E and S9F). In order to probe the origins of the incompatibility between processes I and II, we calculated the exchange regime of process I using the α factor, which is defined for systems where p_G >> p_E as:

$$\alpha =\frac{2{(k_{ex}/\text{Δ}\omega )}^2}{1+{(k_{ex}/\text{Δ}\omega )}^2}$$

and varies from 0 for slow exchange through 1 for intermediate exchange to 2 for fast exchange. Based on the best-fit values of k_ex,CE1, p_E1 and Δϖ_CE1, α for process I varies between 0.005-0.015 (0.05 – 0.15 if the upper bound of k_ex,CE1 is used), clearly showing all three residues are in slow exchange. In this regime, the maximum R_ex for a process defined by k_ex,CE1 and p_E1 is R_ex = P_E1k_ex,CE1, which is 5.5 s^-1 for process I. The CPMG profiles of residues in such slow exchange have plateaus at both low and high CPMG pulsing frequencies [59]. However, several residues belonging to process II such as T136 (R_ex = 17 s^-1) and V181 (R_ex = 17 s^-1) have R_ex significantly larger than 5.5 s^-1, while not having a plateau at the low frequency end. The distinction between processes I and II can therefore be inferred directly from a visual inspection of their respective CPMG profiles and is not merely a result of the fitting routine.

While the CPMG profiles of all the residues reporting on process II fit well to a two-state exchange model, 1D (Figure S10) and 2D (Figure 4B) ${\chi }_{red}^2$ surfaces for p_E2 and k_ex,CE2 are flat indicating that thermodynamic and kinetic parameters cannot be reliably extracted from these profiles alone. In addition, the absence of a low frequency plateau in CPMG profiles places the residues in the intermediate-fast exchange regime where p_E and Δϖ are correlated. Accordingly, Δϖ_CE2 is also not a parameter that can be interpreted from the fits of the CPMG profiles. In order to get insights into the structure of E2, we first predicted the chemical shifts of the excited state based on some conformations that C-Mad2* could potentially adopt in the excited state, namely conformations that resembled O-Mad2*, substrate-bound C-Mad2* and the unfolded state. We then globally fit the CPMG profiles of all residues varying only p_E2 and k_ex,CE2 while keeping these excited state chemical shifts fixed to the predicted value. The fits are very poor (Figure S11) and the ${\chi }_{red}^2$ increases from 0.95 for the unconstrained fit to 2.5, 3.7 and 2.1 for the case of exchange with conformations similar to O-Mad2*, substrate-bound C-Mad2* and the unfolded state respectively. This analysis establishes that fluctuations of C-Mad2* to E2 involves neither local or global unfolding, nor transitions paralleling fold-switching or substrate recognition.

Yu and coworkers have proposed a model for substrate binding to C-Mad2 in which the seatbelt and the β8’-β8” hairpin disengage from the protein core to form an ‘unbuckled’ state during complex formation. This detachment can occur either before substrate binding (conformational selection), or after substrate binding and subsequent reorganization (induced fit) [20]. Intriguingly, CPMG dispersions are seen for several residues in the seatbelt and the β8’-β8” hairpin, as well as in helix C that forms the point of attachment of the β-hairpin to the core of C-Mad2 (Figure 4C). In addition, all these dispersions can be fit together and are likely to be reporting on the same exchange event. We therefore propose that process II represents a transient opening of the substrate binding cleft that occurs via the disengagement of the seatbelt and the β8’-β8” hairpin from strand 5 and helix C.

In order to determine how substrate binding modulates process II, we acquired ¹⁵N-CPMG data on a sample containing 180 μM C-Mad2* and 360 μM MBP1. The R_ex for all the residues belonging to process II systematically reduce in the C-Mad2*-substrate complex (Figure 4D) further confirming that all these CPMG profiles mark the transition to the same excited state E2 which is less accessible in the substrate-bound state. The presence of measurable R_ex in CPMG profiles of MBP1-bound C-Mad2* raises the question whether these dispersions result from the association and dissociation of MBP1 with C-Mad2* due to the existence of a small population of apo C-Mad2* at equilibrium. In order to assess this possibility, we acquired ¹⁵N-CPMG data on a second sample of 180 μM C-Mad2* containing a higher concentration of MBP1 (600 μM). The ¹H-¹⁵N HSQC spectra at 360 μM and 600 μM MBP1 overlay very well, indicating that binding is already saturated at 360 μM MBP1 (Figure S12). If the dispersions originate from peptide binding and release, the magnitude of R_ex will decrease because there will be a smaller amount of unbound C-Mad2* in the 600 μM MBP1 sample than in the 360 μM one. Conversely, if the dispersions are a consequence of fluctuations intrinsic to the C-Mad2*-MBP1 complex, they should remain unchanged in both samples. Figure 4E shows an overlay of the CPMG profiles of T136 from the 360 μM MBP1 sample in blue and the 600 μM sample in orange respectively. The two profiles overlay perfectly, confirming that the excited state E2 is also accessed by MBP1-bound C-Mad2* also, although E2 is destabilized in the MBP1-C-Mad2* complex. Similar trends are seen in all the residues belonging to process II, where there is a reduction in R_ex between apo and substrate-bound C-Mad2*, while R_ex is comparable for the samples containing 360 μM and 600 μM MBP1 (Figure 4F). The binding of C-Mad2* to substrates such as Cdc20 is essential for their activation. It is likely that substrate binding rigidifies C-Mad2* by forming interactions with both the seatbelt and strand 6 that tether the seatbelt to the core of the protein. This suppresses the disengagement of the seatbelt and the β8’-β8” hairpin and lowers the population of the unbuckled conformation. However, a basal level of opening and closing is still present that enables substrate release once activation is complete.

Dynamics in O-Mad2*

Finally, we asked whether O-Mad2 is a more rigid reservoir of the protein and whether fold-switching to O-Mad2 is beneficial because millisecond timescale processes are suppressed in in this alternate conformation. Acquiring long (> 48 h) CEST experiments on O-Mad2 is challenging because O-Mad2 interconverts to C-Mad2 on this timescale [5]. We therefore acquired ¹⁵N-CPMG experiments to probe millisecond timescale fluctuations in O-Mad2*. Figures 5A and S13 show ¹⁵N-CPMG profiles of residues in the regions involved in fold-switching, namely Q9 and G10 at the N-terminus, I135 in helix C and I31 in helix A. All these residues show measurable R_ex, indicating that the conformational free energy landscape in the vicinity of O-Mad2* is also rugged with the presence of a low-lying thermally accessible conformation (E3). It is challenging to determine p_E3, k_ex,OE3 and Δϖ_OE3 from these CPMG profiles since all of them fall in the fast exchange regime where there is a tight correlation between p_E3 and Δ_ϖOE3. However, the location of the residues undergoing exchange on the structure of O-Mad2 (Figure 5B) provides insights into the excited state E3. The exchanging residues form a hydrophobic cluster which tethers β-strand 1 to the core of O-Mad2* (Figure 5B). It is therefore likely that this hydrophobic pocket is rearranged or disrupted in the excited state E3. Since this hydrophobic cluster is important in maintaining the structural integrity of O-Mad2, it is possible that the transition to E3 represents an early event in the metamorphosis of O-Mad2 to C-Mad2.

Figure 5. Conformationally excited state E3 in equilibrium with O-Mad2*.

A) ¹⁵N-CPMG dispersion profiles of residues with detectable Rex. Solid lines are fits of the profiles to the two-state Bloch-McConnell equations. B) Residues showing fingerprints of exchange in ¹⁵N-CPMG are plotted as yellow spheres on the structure of O-Mad2 (PDB ID: 1DUJ [82]). The colour scheme of O-Mad2 is the same as in Figure 1B.

Discussion

Mad2 is a metamorphic protein which switches between distinct open O-Mad2 and closed C-Mad2 conformations as part of its functional cycle in the spindle checkpoint complex [5, 18]. In order to explore the conformational free energy surface of Mad2, we have used ¹⁵N-CEST [24] and ¹⁵N-CPMG [48, 49] NMR experiments which can detect transiently (>100 μs) and sparsely (> 0.5 %) populated protein conformations. These experiments highlight the unusually rugged free energy surface of Mad2 and reveal the existence of three sparsely populated conformations, two in equilibrium with C-Mad2 and one with O-Mad2.

Our results suggest a model in which the intricate equilibrium between the open, closed and transiently populated conformations of Mad2 is modulated by cellular factors to achieve context-dependent signal activation (Figure 6). A distinguishing feature of the protein-protein interactions established by Mad2 that is shared by more than 20 other HORMA domains is the encircling of peptide segments in substrate proteins by the seatbelt [12]. The opening of the seatbelt is, therefore, critical to the function of all these HORMA domains. CPMG dispersions in C-Mad2 show significant R_ex for residues in the seatbelt and the C terminal β-hairpin composed of strands 8’ and 8” immediately downstream, as well as in helix C which tethers the hairpin to the core of C-Mad2 (Figure 4). This suggests that E2 which is detected through these dispersions represents the unbuckled state of C-Mad2 in which the β-hairpin transiently disengages from the core and simultaneously exposes the substrate-binding site of C-Mad2. Interestingly, traces of the unbuckled state have been observed via hydrogen/deuterium exchange-mass spectrometry (HDX-MS) in the meiotic HORMA domain Hop1 [22]. SEC-MALS data on Hop1 reveal the coexistence of compact and extended monomeric conformations. HDX-MS protection factors in the seatbelt and the β-hairpin increase significantly upon peptide binding, suggesting that the extended and compact conformations differ primarily in the seatbelt region [22]. However, the unbuckled state has never been experimentally detected for Mad2, though its existence has been hypothesized as necessary for substrate binding [13, 20–22].

Figure 6.

Model depicting the equilibrium between the native (O and C), transiently populated (E1, E2 and E3) and peptide-bound conformations of Mad2. Excited states E1, E2 and E3 detected here using ¹⁵N-CEST and ¹⁵N-CPMG experiments are labelled in red. All speculative processes are indicated in green text.

Once the substrate binding site becomes accessible in E2, it can be occupied by substrate peptides (termed as closure motifs in literature) to provoke downstream signal transduction [60]. C-Mad2 engages closure motifs in a variety of proteins apart from Mad1 and Cdc20 and plays regulatory roles in processes such as insulin receptor endocytosis that are distinct from its function in spindle checkpoint assembly [13, 61, 62]. We observe that the N-terminus of C-Mad2 itself encodes for a closure motif and can bind to another molecule of C-Mad2 at the substrate-binding site (E1) (Figure 3F). The sequence at the N-terminus of Mad2, ¹MALQLS⁶, matches quite well with the consensus motif for Mad2 binding, which has hydrophobic residues at positions 3 and 5, as well as polar residues in positions 4 and 6 [12]. Self-closure motifs have so far not been identified in mitotic HORMA domains. We propose that this self-closure of C-Mad2 can serve as an autoinhibitory mechanism that tempers the affinity of C-Mad2 for other closure motifs and provides a free energy difference that must be overcome by competing substrates prior to signal activation. For example, E1 would retard the binding of C-Mad2 to cytosolic Mad1 or Cdc20 when inhibition of APC/C was not required. The rate constant for the formation of E1 (k_C,E1 = p_E1k_ex,CE1 = 5.5 s⁻¹) is smaller than the rate constant for formation of E2 (k_C,E2 = p_E2k_ex,CE2 = 33.9 s⁻¹), supporting the conjecture that unbuckling of C-Mad2 must occur prior to substrate binding (or alternatively, that E2 is on-pathway for the formation of E1) (Figure 6).

In contrast to Mad2, meiotic HORMA domains in several organisms from fungi to mammals carry closure motifs at their C-terminus. Intra and intermolecular self-closure has been proposed as a regulatory mechanism in meiotic HORMA domains that would control the association of HORMA domains with chromosomes and prevent the formation of aberrant double strand breaks in DNA. Meiotic HORMA domains in Caenorhabditis elegans also assemble into oligomers through the interaction of the C-terminal closure motif with the substrate binding site [12, 60, 63]. Our CEST and CPMG data show the existence of an analogous interaction of the mitotic C-Mad2 with its own N-terminal closure motif that results in the formation of E1; this provides a mechanism by which similar head-to-tail oligomers can be formed by C-Mad2 as and when required during chromosome segregation. Indeed, tetramers of wild-type C-Mad2 have been observed previously using size-exclusion chromatography [47, 64] that support our interpretation of E1 as an interaction between the seatbelt and the closure motif of C-Mad2.

Mad2 is a molecular switch that fluctuates between open (O-Mad2) and closed (C-Mad2) structures. Ribosomal biosynthesis generates O-Mad2, which interconverts to C-Mad2 over the timescale of several hours [5]. However, C-Mad2 associated with Mad1 at unattached kinetochores accelerates this interconversion by forming a heterooligomer with O-Mad2 [18]. In vitro, the unbuckling of C-Mad2 is likely to be the first step in the conversion of C-Mad2 to O-Mad2 as the β7-β8 hairpin needs to dissociate from strand 5 and helix C before associating with strand 6 on the other side of the Mad2 core. The excited state E2 that we observe in CPMG experiments is an unbuckled state and is therefore likely to be an on-pathway intermediate towards the formation of O-Mad2. The E2 conformation in equilibrium with C-Mad2 could accordingly act as a hub for triage of molecules along either substrate binding or interconversion pathways (Figure 6).

We also see the initial steps of O↔C interconversion from the side of O-Mad2 which transiently forms the excited state E3. In E3, the N-terminal β-strand 1 detaches from helix C in preparation for extending helix A that is longer in C-Mad2. In support of this structural interpretation, Mad2^Δ15, Mad2^L13A and Mad2^L13Q predominantly adopt the C-Mad2 conformation [17, 65, 66] suggesting that the hydrophobic interactions between strand 1 and helix C are crucial for stabilizing O-Mad2 and should first be broken when fold-switching commences. E3 is therefore likely to be an on-pathway intermediate in the conversion of O-Mad2 to C-Mad2. Since the formation of excited states E2 and E3 both occur on the millisecond timescale, while O-Mad2↔C-Mad2 conformational switching takes much longer, we postulate that the unbuckling of the seatbelt and β7-β8 hairpin from strand 6 of O-Mad2 should be the rate determining step for O↔C interconversion (Figure 6). In agreement with this hypothesis, O-Mad2 recruited to C-Mad2 molecules at kinetochores has been previously hypothesized to first form an unbuckled intermediate state [13, 21, 22], emphasizing that the Mad1-C-Mad2 complex is able to accelerate this rate-limiting step and thereby catalyze O↔C fold-switching. Moreover, the MBP1 peptide is also able to speed up O↔C interconversion in vitro [5]. Since Mad2 locked in the O-Mad2 conformation cannot bind substrates, MBP1 must be able to couple one of the excited states of O-Mad2 (likely E3) with the unbuckled E2 state in order to effect rapid fold-switching. Our data thus highlights how fold-switching and substrate binding modalities of Mad2 are tightly intertwined.

Intriguingly, Mad2 undergoes phosphorylation at Ser170, Ser178 and Ser195 located in the seatbelt and the β-hairpin 8’,8”. The levels of phosphorylation fluctuate through the cell-cycle and are maximal during mitosis [67]. Phosphorylation at Ser195 locks Mad2 in the open form and eliminates its binding to Cdc20, while dephosphorylation of Mad2 promotes its in vivo association with Mad1 and Cdc20 [67, 68]. While the effect of phosphorylation on the stability of the excited states observed here remains unknown, it is apparent that phosphorylation furnishes an additional way in which the Mad2 binding and fold-switching equilibria can be regulated in response to cellular conditions.

Mad2 is a noteworthy example of the versatility of protein sequences in being able to encode not only multiple structures but also dynamics organized across a hierarchy of energy and timescales. From a structural standpoint on the one hand, Mad2 populates two distinct native conformations O-Mad2 and C-Mad2 at comparable stabilities. In addition, the free energy landscape of Mad2 features at least three other thermally accessible excited states, E1, E2 and E3, all of which are present within a free energy difference of ~3 kcal/mol from O- and C-Mad2. From the perspective of molecular dynamics on the other hand, the conformations in the free energy surface of Mad2 convert across a range of timescales from milliseconds (C-Mad2↔E1, C-Mad2↔E2 and O-Mad2↔E3) to several hours (O-Mad2↔C-Mad2) (Figure S14), indicating the presence of free energy barriers with heights differing by ΔΔG^0,≠ > 10 kcal/mol.

The heterogeneity in the conformational ensemble of Mad2 and the coexistence of multiple conformations at equilibrium indicates that Mad2 is highly frustrated. Frustration in a protein refers to the presence of multiple competing interactions that cannot be simultaneously adopted by a protein. Frustration results in the availability of low-lying conformations in the free energy landscape that are degenerate or near-degenerate in free energy with the native state [69, 70]. Since these alternate conformations can present interaction interfaces that are structurally complementary to binding partners, biomolecular interaction interfaces are enriched in frustration. Our results on Mad2 demonstrate that alternate conformations arise not only at interaction interfaces (E1 and E2), but also in regions that are relevant for fold-switching (E2 and E3). In sharp contrast, the core of the protein, which is not involved either in metamorphosis or in binding partner proteins, is rigid and shows no detectable exchange in the microsecond-second timescale (Figure S15).

Finally, Mad2 is an addition to the growing list of examples where transiently and sparsely populated biomolecular conformations are essential for function. NMR spectroscopic methods such as CEST [24, 25], DEST [26, 27, 71], CPMG [28, 29], R1ρ [31–33] and PRE [34, 72] have played a vital role in bringing these ‘invisible’ conformations to light and elucidating the link between excited state structure and function [73–75]. In the case of Mad2, data from these experiments spotlight the existence of an array of transient excited states in equilibrium with the metamorphic forms of Mad2, each of which possess structures uniquely adapted to the role they play in fold-switching and function.

Materials and Methods

Protein overexpression and purification

R133A Mad2 (Mad2*)

The full-length human mad2 gene (Uniprot ID Q13257) with the 133^rd position mutated to alanine (mad2*) was subcloned into a pET-29b (+) expression vector between the NdeI and HindIII restriction sites. The construct has an N terminus hexa-histidine tag followed by a linker (PMSDYDIPTT) and a Tobacco Etch Virus (TEV) protease cleavage site (ENLYFQG) to cleave the tag after purification. Mad2* was overexpressed and purified as reported in the literature [76] with small modifications. The detailed protocol is as follows. Mad2* was overexpressed in BL21(DE3) Escherichia coli cells in either Luria-Bertani (LB) media for unlabeled protein or in 2X M9 minimal media [77] to obtain isotope labeled protein. For U-¹⁵N-labeled Mad2*, ¹⁵N-NH₄Cl (1g/L) was used as the sole nitrogen source; for partially deuterated U-¹⁵N Mad2*, 2X M9 minimal media was prepared in recycled D₂O with ¹⁵N-NH₄Cl and unlabeled glucose as the nitrogen and carbon sources respectively. Cells were grown at 37 °C. When the culture reached an OD₆₀₀ of ~ 0.8, the temperature was lowered to 16 °C. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added to a final concentration of 0.2 mM to induce the overexpression of Mad2* after 30 minutes of shaking at 16 °C. Cells were harvested eight hours after induction by centrifugation at 7500 rpm for 20 minutes at 4 °C. The cell pellet was resuspended in 100 ml lysis buffer (50 mM Tris, 300 mM NaCl, 3 mM β-mercaptoethanol, pH 8.0) to which one cOmplete Mini protease inhibitor cocktail tablet was added. Cells were lysed using a GEA Niro Soavi Panda PLUS homogenizer at 600-700 bar pressure for 10-15 minutes. The lysed cells were incubated with lysozyme (7.5 mg/g of cell pellet), 1 mM phenylmethylsulphonyl fluoride (PMSF), 5 mM MgCl₂, deoxyribonuclease I (DNase I; 50 μg/g of cell pellet) and 0.04 % polyethyleneimine at room temperature on a stirrer for 45 minutes. The lysate was centrifuged at 13000 rpm for 45 minutes to remove cell debris and inclusion bodies. The clear supernatant was filtered using a 0.45 μm filter and loaded onto a His Trap FF 5 ml column. Bacterial proteins were washed out using 25 mM imidazole and His-tagged Mad2* was eluted with 300 mM imidazole. Elution fractions were pooled and dialyzed for 24-36 hours against buffer containing 50 mM Tris, 50 mM NaCl and 3 mM β-mercaptoethanol, pH 8.0, to remove excess imidazole and NaCl. The tag was simultaneously cleaved during dialysis using His-tagged TEV protease (4 mg of TEV per 100 mg of Mad2*). The dialyzed protein was centrifuged at 3500 rpm for 15-20 minutes and the supernatant was loaded on a His Trap FF 5 ml column. The flowthrough and wash (25 mM imidazole) fractions were collected. O-Mad2* and C-Mad2* were separated using anion exchange chromatography with a Mono Q HR 5/5 column (1 ml bed volume). The column was equilibrated with low salt buffer (20 mM Tris, 50 mM NaCl, 3 mM β-mercaptoethanol, pH 8.0). A gradient of 0 % to 40 % of high salt buffer (20 mM Tris, 1 M NaCl, 3 mM β-mercaptoethanol, pH 8.0) over 20 column volumes was used to elute the two conformations of Mad2*. O-Mad2* eluted at a conductivity of 16.6 mS/cm, while C-Mad2* eluted at a conductivity of 24.3 mS/cm. Fractions corresponding to O-Mad2* and C-Mad2* were pooled separately and exchanged into the required buffer using a 10 kDa MWCO Amicon Ultra-15 centrifugal filter unit. The purity and molecular weight of the protein were checked using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Electrospray Ionization Mass Spectrometry (ESI-MS).

L5I Mad2*

The L5I mutation was introduced in the mad2* gene using site-directed mutagenesis. The protein was purified using the same protocol as described for Mad2*. The ¹H-¹⁵N HSQC spectrum of peptide-bound L5I Mad2* was used to confirm the resonance assignment for the L5 residue (Figure S7E, orange spectrum).

MAD2 binding peptide (MBP1)

MBP1 (SWYSYPPPQRAV) [19] was purchased as a lyophilized powder from SynPeptide Company (China). A stock solution of the peptide was made by dissolving MBP1 powder in C-Mad2* buffer (described below) and this stock solution was subsequently used to make all NMR samples containing MBP1.

NMR spectroscopy

U-¹⁵N O-Mad2* (136 μM, 500 μM and 628 μM) and fractionally deuterated U-¹⁵N O-Mad2* (70 μM) samples were prepared in 20 mM Tris buffer pH 7.2 containing 50 mM NaCl, 3 mM dithiothreitol (DTT), 1 mM EDTA, 0.03 % NaN₃ (O-Mad2* buffer) and 10% D₂O. Fractionally deuterated U-¹⁵N C-Mad2* (133 μM for apo C-Mad2* and 180 μM for peptide-bound C-Mad2*) samples were prepared in 50 mM sodium phosphate buffer pH 6.8 containing 50 mM KCl, 3 mM DTT, 1 mM EDTA, 0.03 % NaN₃ (C-Mad2* buffer) and 10 % D₂O.

All NMR experiments were acquired on a 600 MHz (14.1 T) Agilent spectrometer (298 K), a 700 MHz (16.45 T) Bruker spectrometer (303 K), or a 800 MHz (18.8 T) Bruker spectrometer (298 K or 303 K). All three spectrometers were equipped with triple resonance single-axis gradient probes that operated either at room temperature (700 MHz) or were cryogenically cooled (600 and 800 MHz). NMR spectra were processed using NMRPipe [78] and visualized using NMRDraw [78] and NMRFAM Sparky [79] software.

CEST data acquisition

¹⁵N-CEST datasets were acquired on a 133 μM sample of fractionally deuterated U-¹⁵N C-Mad2* as well as 180 μM samples of fractionally deuterated U-¹⁵N C-Mad2* containing 360 μM and 600 μM MBP1 using a previously reported pulse sequence [24] at 25 °C on a 600 MHz spectrometer. ¹⁵N-CEST datasets were collected using two B₁ fields, 23.8 Hz and 47.5 Hz, applied during an exchange time (T_EX) of 300 ms. Acquisition times of (35 ms, 64 ms) (t_1max, t_2max) were used in the indirect and direct dimensions respectively. Eight transients were acquired for each increment and a 1.5 s relaxation delay was used in all experiments. For a B₁ field of 47.5 Hz, 44 2D planes were collected between -1050 Hz to 1050 Hz with a spacing of 50 Hz. 72 2D planes were acquired for the 23.8 Hz B₁ field for the same ¹⁵N-CEST sweep width and a spacing of 30 Hz. B₁ fields were calibrated on an external 1.7 mM ¹⁵N-Trp sample [80] using the method of Guenneugues et al [81].

CPMG data acquisition

¹⁵N-TROSY-CPMG [49] data were acquired on a 133 μM sample of fractionally deuterated U-¹⁵N C-Mad2* as well as 180 μM samples of fractionally deuterated U-¹⁵N C-Mad2* containing 360 μM and 600 μM MBP1 using a previously reported pulse sequence [48] at 25 °C on a 600 MHz spectrometer. For apo C-Mad2*, one dataset was also acquired on another spectrometer field strength, 800 MHz. An exchange duration (T_EX) of 30 ms was used for ν_CPMG ranging from 33.33 Hz to 1000 Hz. Acquisition times of (35 ms, 96 ms) (t_1max, t_2max) were used in the indirect and direct dimensions respectively. Eight transients were acquired for each increment and a 2.5 s relaxation delay was used in all experiments.

¹⁵N-CW-CPMG [28] data were acquired on a 70 μM fractionally deuterated U-¹⁵N O-Mad2* sample at 25 °C on a 600 MHz spectrometer. An exchange duration (T_EX) of 30 ms was used for ν_CPMG ranging from 33.33 Hz to 1000 Hz. Acquisition times of 27.4 ms, 64 ms (t_1max, t_2max) were used in the indirect and direct dimensions respectively. 32 transients were acquired for each increment and a 2.5 s relaxation delay was used in all experiments. The heating resulting from continuous wave ¹H decoupling during T_EX was calibrated to be 0.4 °C and the experiments were run at 0.4 °C lower temperature to compensate for this heating.

¹⁵N CW-CPMG [28] data were acquired on a 628 μM U-¹⁵N O-Mad2* sample at 30 °C on a 700 MHz spectrometer. An exchange duration (T_EX) of 30 ms was used for ν_CPMG ranging from 33.33 Hz to 1000 Hz. Acquisition times of 32.8 ms, 64 ms (t_1max, t_2max) were used in the indirect and direct dimensions respectively. 16 transients were acquired for each increment and a 2.5 s relaxation delay was used in all experiments. The heating resulting from continuous wave ¹H decoupling during T_EX was calibrated to be 0.6 °C and the experiments were run at 0.6 °C lower temperature to compensate for this heating.

CEST and CPMG data processing and analysis

Global lineshape fitting to extract peak intensities across different ¹⁵N-CEST or ¹⁵N-CPMG planes of a pseudo-3D dataset was implemented using the FuDA (https://www.ucl.ac.uk/hansen-lab/fuda/) software package. For CEST datasets, the ratio of intensities I/I₀ was plotted as a function of the ¹⁵N-offset position. I is the peak intensity in planes acquired with the B₁ field resonant with different chemical shift offsets and I₀ is the peak intensity in the reference plane with T_EX of 0. In case of CPMG datasets, the effective transverse relaxation rate (R_{2, eff}) was plotted as a function of ν_CPMG. R_2,eff was calculated from peak intensities as follows:

$$R_{2,eff}=\frac{1}{T_{EX}}\ln (\frac{I_0}{I})$$

¹⁵N-CEST data for two B₁ fields and different ¹⁵N-CPMG datasets were fit to two-state exchange models using the integrated Bloch-McConnell equations [53] using the software ChemEX (https://github.com/gbouvignies/ChemEx). Global parameters like the exchange rate (k_ex) and the population of minor state (p_E) were extracted from the fitting routine. ${\chi }_{red}^2$ surfaces were used to establish the reliability of the global parameters. For 1D surfaces, values of one parameter were fixed one at a time while the other parameters were floated during the fitting procedure. $2D-{\chi }_{red}^2$ surfaces were run by fixing different pairs of p_E and k_ex during the fitting protocol. Errors in fit parameters were determined from Monte Carlo simulations using 1000 simulated datasets.

O-Mad2* to C-Mad2* conversion

To monitor the conversion of O-Mad2* to C-Mad2*, 136 μM of O-Mad2* sample in O-Mad2* buffer was buffer exchanged to 50 mM sodium phosphate pH 6.8 containing 300 mM KCl, 1 mM DTT, 1 mM EDTA, 0.03 % NaN₃ and 10% D₂O using a 10 kDa MWCO Amicon Ultra-15 centrifugal filter unit. ¹H-¹⁵N HSQC spectra were recorded at 30 °C on an 800 MHz spectrometer till the spectrum resembled that of C-Mad2* and resonance intensities no longer varied with time. The clock recording the time for interconversion was started as soon as the sample was transferred from 4 °C to 30 °C. The noise in the spectrum was used as an estimate of the error in the data points. Peak intensities increasing over time were fit using the following equation:

$$I(t)=[(I_{\infty }-I_0)(1-e^{-kt})]+I_0$$

while peak intensities decreasing over time were fit using the following equation:

$$I(t)=[(I_0-I_{\infty })e^{-kt}]+I_{\infty }$$

Where I₀ is the intensity at time t = 0, I_∞ is the asymptotic intensity at t = ∞ and k is the rate constant. The time-dependent increase and decrease in intensity were globally fit to get the rate of O-Mad2* – C-Mad2* interconversion.

Analytical size exclusion chromatography (SEC)

Analytical SEC was performed on a Superdex 75 Increase 10/300 GL column. 0.2 mg/ml of O-Mad2* and C-Mad2* in their respective buffers (See NMR spectroscopy section in Materials and Methods) with 150 mM salt was used (Figure S2). For Figure S6, 180 μM of C-Mad2* and 360 μM of MBP1 peptide were used. Four standard proteins were used to generate the molecular weight calibration curve: bovine serum albumin (66.0 kDa), ovalbumin (44.3 kDa), cytochrome C (12.4 kDa), and aprotinin (6.5 kDa). Blue dextran was used to measure the void volume of the column.

Size Exclusion Chromatography with Multi-Angle Light Scattering (SEC-MALS)

100 μl of 0.5 mg/ml of O-Mad2* and C-Mad2* in their respective buffers (See NMR spectroscopy section in Materials and Methods) with 150 mM salt were used for SEC-MALS on an analytical Superose 6 Increase 10/300 GL column. Datasets were analyzed using the ASTRA 6.1 software package (Wyatt Technology).

Circular dichroism

8.2 μM of O-Mad2* (20 mM Tris, 20 mM NaCl and 1 mM DTT, pH 7.2) and 8.22 μM of C-Mad2* (20 mM sodium phosphate, 20 mM KCl, 1 mM DTT, pH 6.8) were used to obtain far-UV circular dichroism spectra using a Jasco J-715 spectropolarimeter in a 0.1 cm path length cuvette at 20 °C. Spectra were acquired from 195 nm to 260 nm with a pitch of 0.5 nm and a scanning speed of 50 nm/min. Four scans were acquired for each spectrum. Mean residue ellipticity (MRE, deg cm² dmol^-1) was calculated using the following formula:

$$MRE=\frac{\theta }{10\times l\times c\times n}$$

Where θ is in millidegrees, l is the cuvette path length in cm, c is the molar concentration, and n is the number of residues in the protein. The maximum high-tension (HT) voltage observed during scans was 956 V.

Tryptophan emission and excitation spectra

20 μM of O-Mad2* and C-Mad2* (in O-Mad2* and C-Mad2* buffers) were used to obtain tryptophan emission and excitation spectra using a Jasco FP-6300 spectrofluorometer. The excitation wavelength was set to 295 nm and emission spectrum was recorded from 300 nm to 450 nm. For the excitation spectrum, the dataset was recorded from 220 nm to 330 nm with the emission wavelength set to 335 nm. Spectra were acquired with a pitch of 0.5 nm and a scanning speed of 20 nm/min.

Data availability

Data will be made available on request.