Structural basis for the structural dynamics of human mitochondrial chaperonin mHsp60

Human mitochondrial chaperonin mHsp60 is essential for mitochondrial function by assisting folding of mitochondrial proteins. Unlike the double-ring bacterial GroEL, mHsp60 exists as a heptameric ring that is unstable and dissociates to subunits. The structural dynamics has been implicated for a unique mechanism of mHsp60. We purified active heptameric mHsp60, and determined a cryo-EM structure of mHsp60 heptamer at 3.4 Å. Of the three domains, the equatorial domains contribute most to the inter-subunit interactions, which include a four-stranded β sheet. Our structural comparison with GroEL shows that mHsp60 contains several unique sequences that directly decrease the sidechain interactions around the β sheet and indirectly shorten β strands by disengaging the backbones of the flanking residues from hydrogen bonding in the β strand conformation. The decreased inter-subunit interactions result in a small inter-subunit interface in mHsp60 compared to GroEL, providing a structural basis for the dynamics of mHsp60 subunit association. Importantly, the unique sequences are conserved among higher eukaryotic mitochondrial chaperonins, suggesting the importance of structural dynamics for eukaryotic chaperonins. Our structural comparison with the single-ring mHsp60-mHsp10 shows that upon mHsp10 binding the shortened inter-subunit β sheet is restored and the overall inter-subunit interface of mHsp60 increases drastically. Our structural basis for the mHsp10 induced stabilization of mHsp60 subunit interaction is consistent with the literature that mHsp10 stabilizes mHsp60 quaternary structure. Together, our studies provide structural bases for structural dynamics of the mHsp60 heptamer and for the stabilizing effect of mHsp10 on mHsp60 subunit association.

mHsp60 (Fig. 2a). Top and bottom views of the 2D class averages shows seven-fold symmetry characteristics (Fig. 2b), confirming the heptameric arrangement of the mHsp60 subunits. The two stripes in the side view correspond to the apical and equatorial domains, indicating a single-ring conformation. Thus, EM studies validated our purified mHsp60 as a single-ring heptamer. The final cryo-EM map was estimated to 3.4 Å ( Table 2, Fig. S2) with various local resolutions (Fig. 2c). The map (Fig. 2d) shows that the equatorial domain has the highest resolution (3.3 Å), while the apical domain has the lower resolution range (3.9-4.5 Å). Since high resolution maps reflect a homogenous ensemble of conformational states, our observations indicate that the mHsp60 equatorial domains are structurally more homogenous and better ordered than the apical domains, in agreement with crystal structural analysis of GroEL 30,31 . The homogenous and well-ordered structure of the equatorial domains is attributed by their extensive inter-subunit interactions; they contribute ~ 75% of the total inter-subunit interface in the heptamer. The regional map in the equatorial domain that is involved in the inter-subunit interactions (below) is well resolved with clear sidechain information (Fig. 2e).
Overall structure of mHsp60. The mHsp60 heptameric ring is larger than a GroEL ring. The diameter of the mHsp60 ring is 148.8 Å, 8 Å wider than that of the GroEL ring (PDB: 5w0s; 3.5 Å cryoEM structure)-the diameter of the chaperonin ring is determined by the equatorial domains. Since the apical domains contribute to the height of the chaperonin ring, their lower resolution regional map in mHsp60 may not estimate the height of the mHsp60 ring (74.4 Å) accurately compared to that of the GroEL ring (69.1 Å); however, mHsp60 appears taller than the GroEL ring (Fig. 3a). A larger mHsp60 ring is also evident by comparing cryo-EM reconstructions (Fig. S3) of mHsp60 and GroEL (EMDB: 8750). The larger size of the mHsp60 ring suggests that the seven mHsp60 subunits may not interact as strongly as the GroEL subunits in the GroEL ring.
Between mHsp60 and GroEL, the three domains, apical, intermediate and equatorial domains, are arranged similarly but differ in their conformations. While conformations of the equatorial domains overlay well, those of the apical domains deviate (Fig. 3b). Overall, mHsp60 consists of shorter α helices and β strands than GroEL (Fig. 3c). 292 mHsp60 residues while 392 GroEL residues are involved in forming either α helices or β strands (Fig. 3c). The lower secondary structure content of mHsp60 suggests that the mHsp60 subunits are less well folded than the GroEL subunits.
Reduced inter-subunit interactions in mHsp60. The equatorial domains contribute most of the intersubunit interactions, accounting for ~ 75% of the total inter-subunit interface and 44 residues involved in intersubunit interactions. The apical domains contribute nine residues for the inter-subunit interactions, while the intermediate domains contribute six residues. Overall, mHsp60 has a much smaller inter-subunit interface than GroEL, ~ 1258 Å 2 versus ~ 1674 Å 2 . In the following, we compare the molecular interactions on the inter-subunit www.nature.com/scientificreports/ interface between mHsp60 and GroEL. We focus on the equatorial domains because they contribute most of the inter-subunit interactions and their regional maps are well resolved with clear sidechain information.
Center on the inter-subunit interface is a four-stranded β-sheet: β1 and β18 from one subunit connected with β2 and β3 from the other subunit (Fig. 3a). β strands are shorter in mHsp60 than in GroEL: 12 mHsp60 residues compared with 19 GroEL residues are involved in the β-sheet (Table 3). Specifically, β1 and β18 are shortened to  Table 1 www.nature.com/scientificreports/ three residues, the minimal length for β-strands, from five and seven residues in GroEL, respectively. Together, four hydrogen bonds from the backbones are lost in the reduced-size β-sheet of mHsp60 (more below). The shortened β strands in mHsp60 are associated with the unique mHsp60 sequences flanking the N-or C-termini of the β strands. At the outer side of the β-sheet, the GroEL N37 interacts closely with T516 (2.7 Å) and is proximity with T517 (4.5 Å); both T516 and T517 are from the other GroEL subunit (Table 3). These interactions impose restrains on the backbone of N37, leading to a twisted N-terminus of β2 at N37 (Fig. 4A). Also, they together with the R36-E518 interaction (next) result in a tight twist N-terminus of β18. The GroEL R36 forms a strong charge-charge interaction with E518 (2.8 Å) and a weak charge-polar interaction with M114 (4.2 Å) from the neighboring subunit. These restrains on T517 (via N37) and E518 (via R36) explain the T517-E518 segment tilts ~ 45° from the rest of β18 (Fig. 4A). Together, these inter-subunit sidechain interactions are important in forming the twisted conformation of the two interacting inter-subunit strands β2 and β18. However, most of inter-subunit sidechain interactions are lost in mHsp60 (Fig. 4B). The absence of T35 interactions with A516 and T515 (Table 3) releases the restrains on T35 and A516 from forming twisted N-termini of β2 and β18. Similarly, the much-reduced R34-E517 interactions (4.0/4.7 Å) and the absence of R34-V112 interaction release the restrain on E517 from a tight bent of β18. Thus, the three mHsp60 unique residues, T35, A516 and V112, diminish the restrains required to form the twisted conformations found in GroEL β2 and β18, leading to shortened β2 and β18 in mHsp60. Importantly, the three residues are conserved in higher eukaryotic chaperonins (Fig. 3c).
At the inner side of the inter-subunit β-sheet, the GroEL K4 sits snugly in the center of a negative cluster. K4 interactions with both E59 and E61 from the neighboring subunit, albeit modest, position K4 to form an effective interaction with D523 (3.6 Å). The K4-D523 sidechain interaction brings the backbones of K4 and D523 close to form hydrogen bonding, adding twisted ends to β1 and β18, respectively (Fig. 4C). In comparison, with the absence of mHsp60 K2-E522 sidechain interaction, likely affected by the different K2-S57 and K2-D59 interactions when compared with those of K4-E59 and K4-E61 (Table 3), the backbones of K2 and E522 are too far to engage in hydrogen bonding, excluding both residues from β1 and β18, respectively (Fig. 4D). Thus, the three mHsp60 residues, S57, D59 and E522, contribute to the shortened β1 and β18. Importantly, the three residues are conserved in higher eukaryotic chaperonins (Fig. 3c), as found with T35, A516 and V112 (above). Unfolding free energy. The above structural analysis indicates a less stable mHsp60 heptamer than a GroEL heptamer. Consistently, we found that the unfolding free energy (ΔG unfolding ) of mHsp60 is lower than GroEL SR (Table 4, Fig. 3d). Our structural analysis also shows the importance of the regions including the N-terminus in the inter-subunit interactions. D3G mutation, associated with MitCHAP60 disease 32 , has been found to destabilize the mHsp60 heptamer 33 . D3 in mHsp60 is the first residue in the already minimal three-residue β1. Since Gly has low propensity to form a β strand, D3G mutation most likely demolishes β1, downgrading the four-stranded to a three-stranded, β18-β2-β3, β-sheet. Consistently, we found that ΔG unfolding of mHsp60 D3G was much smaller than the wildtype mHsp60 (Table 4, Fig. 3d).
Enhanced subunit interactions in mHsp60 heptamer in response to mHsp10 binding. The mHsp60 subunit interactions are enhanced upon mHsp10 binding. We compared our structure of single-ring mHsp60 (3.4 Å) with the structure of single-ring mHsp60-mHsp10-ADP 26 (3.8 Å, PDB: 6MRD), both solved via cryo-EM (Fig. 5a,b). We first focused on the inter-subunit interface in the equatorial domains, since the domains provide the major inter-subunit interactions (above). Strikingly, mHsp10 binding induces significant structural changes in the equatorial domain, promoting the subunit association. The shortened the inter-subunit β1-β18-β2-β3 sheet is largely restored to that found in GroEL: β1 restores from three to five residues, while β18 increases from three to five resides (Fig. S4). Moreover, the inter-subunit β sheet serves as a pivoting point to tilt the neighboring equatorial domains towards each other (Fig. 5c,d). As a result, the interface formed via the equatorial domains increases significantly by ~ 29% from 938 Å 2 in the mHsp60 alone structure to 1211 Å 2 in the mHsp60-mHsp10 structure. In addition, both the apical and the intermediate domains undergo large domain movements (below), increasing their contribution to the inter-subunit interaction from 25 to 35%. Overall, the inter-subunit interface of the mHsp60 heptamer increases drastically by ~ 49% from 1258 Å 2 in the mHsp60 alone structure to 1871 Å 2 in the mHsp60-mHsp10 structure. The mHsp10-induced drastic increase in intersubunit interface in mHsp60 heptamer is consistent with the observation that the presence of mHsp10 (and ATP) prevents mHsp60 from dissociation 20 . Table 1. The molecular properties of mHsp60, GroEL SR and GroEL. a SEC was carried out using a Superdex 200 10/300 GL column (GE Healthcare Life Sciences). b Values in parentheses refer to Polydispersity Index (PDI). c ATPase activity is normalized to the monomer concentration. d MDH yield was at 60 min after the unfolded MDH was added to the chaperonin systems, and was normalized to native MDH. Without the chaperonin system, the unfolded MDH did not refold (< 5%).   (Fig. 5a,b). Specifically, α14 of the intermediate domain rotates downwards to the equatorial domain in mHsp60-mHsp10, closing the nucleotide binding site and locking the bound nucleotide. The three stranded β4-β13-β5 sheet in apo mHsp60 is tilted upward and reduced to a two stranded β4-β13 sheet in mHsp60-mHsp10. The connected three-stranded β6-β12-β11 sheet in mHsp60 is swung ~ 150° and increased to a four-stranded β5-β6-β12-β11 sheet in mHsp60-mHsp10. α9 and α10, the main binding site for the substrate proteins 35,36 , originally facing the chaperonin cavity in mHsp60, are repositioned upward to interact with mHsp10 in mHsp60-mHsp10. Importantly, the re-positioned apical and intermediate domains are in close contact with their counterparts in the neighboring subunits. As a result, their combined inter-subunit interface increases by > 100% from 320 Å 2 in the mHsp60 alone structure to 660 Å 2 in the mHsp60-mHsp10 structure, contributing to stabilizing the mHsp60 heptamer.

Discussion
Due to its essential role for mitochondrial function and its important roles in human health, mHsp60 has been proposed as potential biomarkers and drug targets [5][6][7][8] . However, detailed molecular and structural investigations on mHsp60 mechanism have been hampered by the difficulties in purifying mHsp60 due to its intrinsic dynamics. The intrinsic dynamics, although a hurdle to detailed molecular investigations on mHsp60, has been implicated in the unique mechanism of mHsp60. Structures of mHsp60 in complex with mHsp10 have been reported 25,26 , however, these structural studies do not address the structural dynamics of mHsp60, because the mHsp60 subunit association is stabilized in the mHsp60-mHsp10 complex. Here we report the structure of mHsp60 alone, the initial functional state in the mHsp60-mHsp10 chaperone reaction. Our structural analysis provides explanations for the dynamics of the mHsp60 subunit associations and how the mHsp60 subunit association is enhanced upon mHsp10 binding.
The dynamics of mHsp60 subunit associations is due to the unique sequences at the inter-subunit interface. Many of the six unique mHsp60 sequences result in diminishing the inter-subunit sidechain interactions as Table 2. Cryo-EM data collection, refinement and validation statistics.  A four-stranded β-sheet, consisting of two strands β1 and β18 from one subunit (grey) and two strands β2 and β3 from the other subunit (orange), is located at the center of the cross-ring inter-subunit interface. Residues involved in inter-subunit interactions are highlighted in Fig. 4, and the atomic distances are summarized in Table 3. The mHsp60 heptamer and one GroEL heptamer were superimposed. (b) Overlay of a mHsp60 subunit with a GroEL subunit. The two equatorial domains were superimposed. (c) Sequence alignment among chaperonins from various sources. Secondary structures of mHsp60 and GroEL are based on Chimera: stripped rectangles for α helices and solid rectangles for β strands.
Color scheme is the same as Fig. 2d. Residues involved in the inter-subunit interactions that are unique to human or higher eukaryote mouse and bovine (although yeast also has Ser at the corresponding S57 in human) are in red. Four residues forming the conserved inter-ring salt bridges are in dark (K105-E435) and light (R453-E462) shades, in GroEL numbering. For convenience, numbering of GroEL secondary structures is used here. (d) Thermal denaturation of single-ring chaperonins. The molar ellipticity (ϑ) was measured at 222 nm. The solid lines are non-linear regression analysis (Eq. 1). The derived ΔH vh , ΔG unfolding and T m values are listed in Table 4. www.nature.com/scientificreports/ found in GroEL. For example, the mHsp60 T35 does not interact with either T515 or A516 of the neighboring subunit, while the corresponding GroEL N37 interacts with T516 and T517 of the neighboring subunit. Moreover, since these unique sequences are located at the ends of β strands, losses of their sidechain interactions fail to engage their backbones to conforming to hydrogen bonding, excluding them from β strand conformation and shortening the β strands. The losses of T35 interactions with T516 and T517 exclude T35 from β2 and T516 and T517 from β18. The importance of sidechain interactions in extending β strands is underscored in GroEL. The N37 interactions with T516 and T517, together with the R36 interaction with E518, restrict the backbones of N37, T517 and E518. As a result, N37 adds a twisted N-terminus to β2, and both T517 and E518 extend a bended N-terminal segment to β18. The reduced inter-subunit sidechain interactions due to the unique mHsp60  www.nature.com/scientificreports/ sequences, together with the resulting shortened inter-subunit β-sheets, decrease the inter-subunit interface of mHsp60, accounting for the structural dynamics of mHsp60 subunit association. Importantly, the unique sequences, T35, S57, D59, V112, A516, and E522, are specific to and conserved within the mitochondrial chaperonins of higher eukaryotes including human. For the lower eukaryotic organisms such as yeast, the sequences at these positions resemble those of GroEL. Moreover, yeast contain sequences for one of the two conserved inter-ring salt bridges, R452-E462 in GroEL, while the higher eukaryotes lack sequences for both salt bridges (Fig. 3c). Thus, the destabilizing sequences are unique to the single ring eukaryotic chaperonins, underscoring their importance in structural dynamics and functional mechanism of the single-ring chaperonins in higher eukaryotic mitochondria. Binding of cochaperonins induces large structural changes in chaperonins in a nucleotide dependent manner. The large structural changes include an upward rotation of α14 in the equatorial domain, an upward tilt of the β5/β13/β4 sheet in the intermediate domain, and a large swing of the β6/β12/β11 sheet and a reorientation of α9 and α10 in the apical domain (Fig. 5a). For GroEL, the large structural changes do not affect the inter-subunit interface: the interface remains largely the same before (1583 Å 2 ) and after (1594 Å 2 ) GroES binding (PDB: 1PCQ; the trans and cis GroEL ring, respectively). In addition, the contributions from individual domains to the inter-subunit interface remain largely the same. For example, the equatorial domains contribute 62% and 64% of  Table 3. Table 4. Thermodynamic parameters from thermal unfolding of mHsp60 and GroEL SR proteins via circular dichoism. a van't Hoff enthalpy, derived from fitting data to Eq. 1. b The free energy of unfolding at 20 °C, derived from Eq. 2. c The temperature of the transition midpoint, derived from fitting data to Eq. 1. www.nature.com/scientificreports/ the total interface before and after GroES binding, respectively. Moreover, structure of the inter-subunit β1-β18-β2-β3 sheet remains the same after GroES binding. This structural analysis suggests that the subunit association in the GroEL ring, i.e., the stability of the GroEL ring, is not affected by GroES binding. In a striking contrast, mHsp10 binding affects inter-subunit interaction in mHsp60 drastically. Upon mHsp10 binding, the overall inter-subunit interface in mHsp60 increases by 49%, and the contributions from the equatorial domains and the apical/intermediate domains increase by 29% and 105%, respectively. In addition, the shortened inter-subunit β1-β18-β2-β3 sheet of mHsp60 is restored to the same as in GroEL or the GroEL-GroES complex (Fig. S4). Thus, unlike GroEL, mHsp60 subunit association, i.e., mHsp60 heptameric stability, is enhanced in the presence of bound mHsp10. The mHsp10 effect on mHsp60 stability implicates the importance of structural dynamics in heptameric mHsp60, underscoring the mechanistic differences between mHsp60-mHsp10 and GroEL-GroES.

Methods
Protein purification. Gene encoding the mature mHsp60 sequence (without the mitochondrial targeting signal) was synthesized with codons optimized for protein expression in E. coli, and was inserted into pBbE5a with NdeI/BamHI sites. pBbE5a-mHsp60 plasmid was transformed to BL21DE3 for protein expression. Cells were grown in LB at 37 °C until OD 600 = 0.6, protein expression was induced with 0.4 mM IPTG, and cells continued to grow for 5 h. Protein purification was modified from a published protocol 22   www.nature.com/scientificreports/ ammoniate sulfate in Buffer A. After extensive wash including with 20% methanol, 20% ammoniate sulfate in Buffer A, the bound protein was eluted with a linear gradient of 20-0% (NH 4 ) 2 SO 4 . The mHsp60 containing fractions were combined and (NH 4 ) 2 SO 4 was added to a final 60% to precipitate mHsp60. The precipitated mHsp60 was dissolved in 100 mM NaCl in Buffer A, and the solution was subjected to gel filtration purification using a Superdex 200 column (GE Healthcare). Fractions containing single ring mHsp60 were concentrated to a desired concentration, and the freshly concentrated mHsp60 was frozen on the EM grids or used within the same day for analysis. mHsp60 D3G was generated using QuickChange (Qiagen), and purified as above. GroEL and GroEL SR were purified as before 37 .
Dynamic light scattering (DLS). Freshly purified mHsp60, GroEL and GroEL SR were subjected to DLS analysis using a Zetasizer Nano-S dynamic light scattering instrument (Malvern Instruments). Samples were in 100 mM NaCl, 10 mM KHepes, 1 mM EDTA, 1 mM DTT and 5% glycerol, and the temperature was 25 °C. DLS data were collected within few hours of purification and for the next six days. Measurements were repeated more than once for each sample.
ATPase assay. The steady-state ATP hydrolysis rate was measured using the malachite green assay, as described in our previous work 37  The CD signal at 222 nm was recorded by a J-715 CD spectrometer every 2 °C with a 4-s averaging time. Measurements were repeated more than once for each sample. Thermal denaturation was analyzed by fitting the data to a two-state model according to the following equation 38 : where y is the observed CD signal at 222 nm, y N and y D are the derived signals for the native and denatured samples, and m N and m D are the derived baseline slopes for the native and denatured samples, T is the temperature, ΔH vh is the van' Hoff enthalpy, R is the gas constant and T m is the temperature of the transition midpoint. The free energy of unfolding at specific temperature is estimated from a modified Gibbs-Helmholtz equation, assuming no heat capacity change between the denature and native conformations: ΔCp = C p (denatured) − C p (native) = 0 Cryo-EM data collection, image analysis, and model building. Protein sample prepared in the same batch were submitted to two facilities PSU COM and NCCAT for cryo-EM data collection. Detailed methods for cryo-EM are described in the supplementary method section. In general, approximate 4 µl of mHsp60 in ~ 9 mg/ml were applied to Quantifoil™ holey carbon grids. Vitrification was performed using Thermo Fisher Scientific (TFS) Vitrobot Mark IV under 100% humidity. Frozen-hydrated specimens were loaded into Thermo Fisher Scientific Titan Krios operated at 300 kV equipped with Gatan K3 camera. Data collection was done by using TFS EPU or Leginon 39 with (at 30 eV) or without energy filter. Initial data quality assessment was done by RELION and Appion 40 . Dose weighted motion corrected was done by MotionCor2 41 and CTF estimation was done by ctffind4 42 and Gctf 43 . Particles were semi-manually picked by e2boxer.py in EMAN2 44 . Reference-free 2D classification, initial model building, 3D classification, 3D refinement, and CTF refinement were performed in RELION 45 . A total number of 182,600 particles were used to build the final 3D reconstruction at a resolution of 3.5 Å for NCCAT data and a total number of 196,060 particles were used in the final 3D reconstructions at a resolution of 3.5 Å for PSU COM data ( Table 2). To combined final 3D models, the unfiltered odd and even electron density maps from both datasets were scaled based on their size and density and summed in the real