Asymmetric horseshoe-like assembly of peroxisomal yeast oxalyl-CoA synthetase

Yeast OCS forms an asymmetric three-layered hexameric assembly

To understand the structural arrangement of yeast OCS, we first determined the crystal structure of the enzyme at 2.9 Å resolution (Figure 1, Supplementary Table S1). Within each asymmetric unit of these crystals, we found two hexameric assemblies with an identical arrangement, composed of trimers of dimers (Figure 1A, Table 1, Supplementary Figure S1).

Figure 1:

X-ray crystal structure of the hexameric OCS (wt) assembly. (A) Three OCS dimers associate into an asymmetric hexameric assembly with a horseshoe-like overall shape. OCS dimer-1 protomers D1(A) and D1(B) are colored in marine blue and cyan, respectively; D2(A) and D2(B) in green and dark green, respectively; D3(A) and D3(B) in yellow and orange, respectively. Upper panel: View along a pseudo-threefold axis in two opposite orientations. As the protomers of the three OCS dimers are related by two-fold symmetry orthogonal to the pseudo-threefold axis, only the front layer of the three OCS dimers is visible. For the purpose of clarity, the visible OCS protomers are indicated by pie charts in the same colors and the angles, defining the relations between the three OCS dimers (cf. Table 2). Two CTDs out of six OCS protomers were included into the overall structure of the OCS hexamer-1, used in this illustration (in light pink). The OCS hexameric assemblies in the lower panel are rotated by 90° around a horizontal axis with respect to the upper panel. Due to the overall horseshoe-like arrangement, a long deep crevice generated by the OCS hexameric assembly becomes visible (lower left panel). The overall dimensions of the OCS hexameric assembly are indicated as well. (B) Molecular details of OCS dimer/dimer assembly, exemplified for the D1/D2 interface indicated by box 1 in panel A. The D2/D3 interface is virtually identical (Table 1). Colors for dimer D1 are as in panel A. Surface areas of dimer D1 that are in contact with dimer D2 are shown in the colors used for D2. Polar residues that contribute to specific dimer D1/D2 interface interactions are colored by their side chain charges (negative, red; positive, blue) and are labeled. The only sidechain specific polar interaction between D1 and D2 is generated by D26 and K352 and is two-fold repeated with opposite orientations, involving D1(A)/D2(A) and D1(B)/D2(B). The respective surface areas on D1(A, B) are boxed. (C) The D36-K352 interaction is supported by additional hydrogen bonds with the OCS main chain backbone (right panel).

While the electron density of all N-terminal domains (NTD) of the two OCS hexamers was well interpretable for the complete NTD sequence, the density accounting for the smaller CTD was less well defined to various degrees. We quantified the expected density for each CTD by map-model correlation analysis, revealing correlation coefficients (cc) between 0.15 and 0.83 (Table 2). As an analysis of contacts with symmetry-related OCS molecules did not reveal major differences, the varying levels of CTD density are a likely a consequence of inherent flexibility of the OCS NTD/CTD arrangement, in-line with previous findings of other members of the ANL family (Gulick 2009). To this end, we used cc = 0.7 as a threshold for structural interpretation, to model two CTDs of the first OCS hexamer (chains D1(A) and D2(B)) and one CTD of the second OCS hexamer (chain D2(A)). Due to limited interpretable density, the CTD sequences that were modeled remained fragmented (Table 2). Modeling of these three CTDs allowed us to assigning OCS residues 1–429 to comprise the NTD, followed by a seven-residues flexible hinge segment (430–436), which connects to the CTD (residues 437–543). As the PTS1 motif required for peroxisome targeting is at the C-terminus it is not visible in any of our structures (Supplementary Figure S1).

The three OCS dimers found in both hexamers assemble into an unusual three-fold repeated stack, forming a horseshoe-like arrangement (Figure 1A, Supplementary Figure S1). This assembly is composed of two orthogonal layers where each of the three two-fold symmetry related OCS homodimers contribute one protomer to the first layer and the other protomer to the second layer, thus generating an asymmetric trimer of OCS dimers. The three-fold repeated OCS dimer layers are related by rotational angles between pairs of OCS dimers 1/2 and 2/3 of about 113° each (Figure 1A, Table 1), thus considerably deviating from three-fold symmetry. The two OCS layers with three protomers each are in congruent positions when looking along the pseudo-threefold axis. Within this OCS dimer layer arrangement, OCS dimers 1 and 2 as well as OCS dimers 2 and 3 interact via relatively small two-fold repeated interfaces of about 400 Å² each (Figure 1B, Table 1, Supplementary Table S4). In these two virtually identical OCS dimer/dimer interfaces, we found a single side chain-specific salt bridge between D36 and K352 (Figure 1C). As result of these dimer/dimer arrangements involving OCS dimer 1 /dimer 2 and OCS dimer 2 /dimer 3, there is a deep and narrow groove between OCS dimers 1 and 3 (Figure 1A, left panel). The visible CTDs are not involved in any of these interfaces. In addition, each of the three OCS homodimers are formed through an elongated and two-fold symmetric interface of around 1100 Å² between OCS NTDs, supported by an array of distinct specific interactions (Figure 2A–B). The overall dimensions of the NTD OCS hexameric arrangement are about 11 nm within each of the two trimeric OCS layers and 9 nm across the trimeric OCS layers (Figure 1A). When including the modeled CTDs the overall particle size increases to about 14 nm × 11 nm, respectively.

Figure 2:

X-ray structure of the dimeric OCS (KD) mutant assembly: (A) Size exclusion profiles of hexameric OCS (wt, blue) and dimeric K352D (KD, green). Calibration standards are shown as well. (B) Dimer interface of the KD variant, which is identical to the interfaces of the three dimers D1, D2 and D3 of the OCS (wt) hexameric assembly (cf. Figure 1). The colors are identical to those used for dimer D1 in the hexameric OCS (wt) assembly (Figure 1). Polar residues that contribute to specific dimer interface interactions are labeled. (C) Surface presentation of the dimeric OCS (KD) variant. The OCS CTD is visible in only in one of the two protomers of the OCS (KD) dimer (light pink). The NTD/CTD hinge region is colored in hot pink. Boxed inlet: Composite 2Fo-Fc OMIT map of the OCS (KD) D1(A) CTD in ribbon presentation. The boundaries of visible CTD fragments are labeled (cf. Table 2). (D) Molecular details of NTD/CTD interface of protomer D1(A) of the wt OCS hexamer-1 (left panel) and of protomer D2(A) of the wt OCS hexamer-2 (right panel). The CTD of D2(A) of the wt OCS hexamer-2 is rotated by 118° with respect to the orientation of th CTD of D1(A) of the wt OCS hexamer-1 (cf. Table 2). This change is most significantly visualized by CTD residue positions T519 and T521, which are involved in specific interactions with the NTD in the second conformation only (right panel). The NTD/CTD arrangements found in protomer D2(B) of wt OCS hexamer-1 and D(A) of the OCS (KD) variant are similar to that of protomer D1(A) of the wt OCS hexamer-1 (left panel) (cf. Table 2, Supplementary Figure S1).

To investigate the OCS dimer module as a separate entity, we converted the positive charge of K352, which forms the salt bridge across the OCS dimer protomer interface, into a negative charge by mutating it to aspartate (Figure 1C). As predicted, the OCS K352D (KD) variant lost its ability to form a hexameric assembly as seen by a delayed elution volume by size exclusion chromatography (Figure 2A). We then solved the X-ray crystal structure of the KD mutant to 2.5 Å resolution. In the crystal structure, the OCS (KD) variant forms a homodimer, which is identical to the one observed in the OCS (wild-type, wt) hexameric structure (Figure 1A, Figure 2B–C). Due to higher resolution of the structure of the OCS (KD) variant, atomic B factor refinement allowed insight into the mobility distribution of the overall OCS fold. As expected, the so-called P-loop, which has a crucial role in nucleotide binding of members of the ANL family (Gulick 2009) belongs to the most flexible segments of the catalytic NTD (Supplementary Figure S2). Similar to our observations in the hexameric OCS assembly there was only sufficient density for model building of a partially resolved CTD of one of the OCS protomers, supporting a semi-flexible arrangement of the NTD relative to the CTD in the crystal lattice (Figure 2C). At this point we analyzed the nature of the NTD/CTD positions of all OCS protomers where it was possible to assign density to the CTD. While the arrangements of two OCS promoters of the wt hexameric assembly and of the OCS (KD) variant are identical, the CTD of the third OCS protomer with modeled CTD is rotated by 118° when using the first NTD/CTD arrangement as reference (Figure 2D, Table 2). The rotation is due to major changes of the dihedral angles of residues 433–435 within the NTD/CTD-connecting flexible hinge. Both arrangements are characterized by a scarce number of interactions between residues from the NTD and CTD (Figure 2D). While the first arrangement has a single hydrogen bond between main chain carbonyl group of T430 (NTD) and S445 (CTD) in common (Figure 2D, left panel), the second NTD/CTD arrangement is held by a specific interaction between the tyrosine hydroxyl group of Y420 from the NTD and T519 and T521 from the CTD, close to the OCS C-terminus (Figure 2D, right panel). The latter two residues are not visible in some of the OCS protomers with the first NTD/CTD arrangement (Table 2). When comparing the two arrangements, for instance the positions of T519/T521 are different by about 3 nm on those superimposed protomers. This illustrates the scale of movement induced through this rotation (Figure 2D).

OCS dimeric layer self-assembly is concentration-dependent

We independently determined the structure of OCS by single particle cryo-EM (Figure 3, Supplementary Figure S3, Supplementary Table S2). After motion correction, the micrographs showed a homogeneous spread of globular particles (Supplementary Figure S3A–B), hinting towards OCS oligomers comprised of multiple copies of OCS homodimers. In addition to hexameric OCS accounting to 63% of all particles, we also detected populations of smaller particles with 33% associated with OCS tetramers and 4% with OCS dimers (Figure 3A). The 2D class averages accounting for the hexameric assembly revealed a large complex with the same characteristic horseshoe-like assembly as observed in the crystal structure (Figure 3A, right panel, Figure 1).

Figure 3:

Cryo-EM structures of hexameric, tetrameric and dimeric OCS assemblies. (A) Representative cryo-EM 2D class averages of OCS (wt) dimeric, tetrameric and hexameric assemblies, with the percentage of particles for each oligomeric assembly indicated. (B) Density maps of dimeric (left panel), tetrameric (central panel) and hexameric (right panel) OCS (wt) cryo-EM samples in two orientations, rotated by 90° around a horizontal axis. The density map colors and labels are as in Figure 1. For comparison, structural cartoons of these assemblies are shown below. (C) Scheme indicating a plausible OCS dimer self-assembly path based on the distribution of cryo-EM particles and biophysical data (cf. Figure 4). Dimeric OCS self-assembly requires the formation of a two-fold repeated salt bridge between neighboring OCS dimers (cf. Figure 1). According to this scheme, OCS dimer self-assembly beyond the observed hexameric arrangement is not permissive, as this would lead to steric clashes.

For the hexameric OCS assembly, we obtained a 3D reconstruction at an average resolution of 3.1 Å (Figure 3B, right panel, Supplementary Figure S4B), which agrees well with the crystal structure. Hence, the complete crystallographic hexameric complex was first fitted into the cryo-EM density with a simple rigid body fitting and finally refined against the cryo-EM map. Unlike our observations in the OCS crystal structures, the cryo-EM map showed no density for any CTD, thus structural interpretation was limited to the analysis of the NTD arrangement (Supplementary Figure S3B). The root-mean-squares deviation (rmsd) between the refined cryo-EM and crystallographic model NTDs (based on C_α atoms of 2553 residues in total) was 1.26 Å, which indicates a high similarity between both models.

We further obtained cryo-EM structures of the tetrameric assembly and the OCS dimer (Figure 3B, left and central panel), which are however affected by anisotropic resolution due to preferred orientation of the respective particles. Therefore, we performed rigid body fitting of the available molecular model of the OCS dimers and did not attempt further model refinement. In summary, the distribution of these oligomeric states in solution indicates the tight OCS dimer as the basic building block. The tight dimers further oligomerize in a cyclic sequential manner via a single salt bridge towards the formation of the hexameric assembly (Figure 3C).

Interestingly, we observed tetramers and hexamers only in cryo-EM samples and exclusively dimers during negative stain EM, that is typically performed at much lower concentrations (Supplementary Figure S4A–B). Chemical crosslinking before negative stain EM stabilized the hexameric horseshoe-like conformation (Supplementary Figure S4C–D). These observations suggest that the protein might be prone to stepwise concentration-dependent self-association of OCS homodimers rather than concerted hexameric assembly.

To further characterize OCS dimer self-association, we analyzed OCS (wt) and the OCS (KD) variant at different concentrations by SAXS (Supplementary Table S3). OCS (wt) SAXS profiles show indeed a concentration-dependent change in the scattering intensities, as the sample was diluted from 7 to 0.5 mg/mL (Figure 4A, Supplementary Figure S5A). These changes correlate to a decrease in the apparent molecular weight (MW) and radius of gyration (R _g ) of the particle population, consistent with mixtures of tetramers and hexamers, as observed in the cryo-EM samples (Figure 4B, Supplementary Figure S5C–D). These observations for OCS (wt) contrast with those for OCS (KD) mutant, where the SAXS profiles and the corresponding MW and R _g as well as other structural parameters such as the Porod volume and real-space scattering-pair distance distribution are consistent with OCS dimers throughout the concentration series (Figure 4C–D, Supplementary Figure S5B–D). Indeed, by taking the X-ray crystal structures and subsequently including the mass and spatial disposition of both the NTD and CTD of each protomer in the OCS assemblies, the OCS (KD) mutant SAXS data is well described by the conformation of crystallographic dimer showing a χ² fit = 1.0 (Figure 4C) whereas the fit to the OCS (wt) SAXS data requires the application of a volume-fraction weighted ratio of crystallographic hexamers and tetramers at each sample concentration resulting in χ² fit range of 1.5–3.9 (Figure 4A–B).

Figure 4:

Biophysical and biochemical impact of concentration-dependent OCS assembly. (A) SAXS curves of OCS (wt) at different protein concentrations, with refined OCS tetramer/hexamer fit in red. Calculated χ² values are 3.9 for 7 mg/ml, 1.6 for 2.5 mg/ml and 1.2 for 0.5 mg/ml. The SAXS curves are vertically offset for clarity. (B) Estimation of OCS (wt) tetrameric/hexameric assembly ratio according to protein concentration. OCS (wt) shows a shift from hexamer to tetramer with decreasing sample concentration. (C) SAXS curve of the OCS (KD) variant at 2.5 mg/mL with a complete OCS dimer model fitted. The calculated χ² is 1.0. (D) Estimates of OCS (wt) and OCS (KD) MWs at different concentrations using SEC-MALLS. OCS (wt) shows a concentration-dependent change in MW, while OCS (KD) even at the highest concentration measured remains dimeric. (E) Melting temperatures of OCS (wt) and OCS (KD) measured by NanoDSF. OCS (KD) has a significantly lower melting temperature than OCS (wt). Measurements were carried out in triplicate. Student’s t-test ***probability < 0.001. (F) Elution profile of 1 µM OCS (wt), OCS (KD) and OCS (K523A) catalytic reaction after 1 min with 100 µM oxalate. ATP and AMP elute at distinct times and OCS (K523A) mutant show no formation of AMP. (G) Quantification of AMP products. Measurements were carried out in triplicate.

To confirm the concentration-dependent change in OCS oligomeric state, we also performed SEC-MALS experiments using three different OCS (wt) injection corresponding to concentrations of 7, 2.5 and 0.5 mg/mL sample. We observed a decrease in MW averages obtained for decreasing OCS (wt) sample injection concentrations, while even at the highest sample injection concentration tested for the OCS (KD) variant the MW average showed that the mutant MW corresponds to a dimer (Figure 4D).

OCS oligomerization and thermostability

To investigate whether self-assembly of OCS dimers has an effect on thermostability, we determined the OCS (wt) and OCS (KD) variant unfolding temperatures using nanoscale Differential Scanning Fluorimetry (NanoDSF). While OCS (wt) unfolded at 43 °C, the dimeric OCS (KD) variant unfolded at 37 °C, corresponding to a reduction of 8 °C in melting temperature (Figure 4E, Supplementary Figure S6). This implies that self-assembly of OCS dimers has a protective effect on OCS unfolding. A second OCS unfolding event is independent on the type of OCS variant used and hence independent of the OCS oligomerization state (Supplementary Figure S6).

We finally tested whether lack of OCS dimer self-assembly impairs OCS catalytic activity, by monitoring ATP hydrolysis and AMP formation using reverse phase chromatography. As a negative control, we used an OCS K523A active site variant (Niebling et al. 2021), which was expected to be catalytically inactive, in analogy to an equivalent mutant of A. thaliana OCS (Supplementary Figure S7) (Fan et al. 2016). As we did not detect any significant difference in the AMP levels generated (Figure 4F–G), we concluded that lack of OCS self-assembly has no impact on OCS catalytic activity.

Protein purification

Full-length S. cerevisiae OCS was cloned in a petM14 vector with an N-terminal hexa-histidine tag and a HRV3C protease cleavage site (www.embl.org/groups/protein-expression-purification). OCS was expressed in autoinduction medium (Studier 2005), for 5 h at 37 °C and 26 h at 20 °C. Cells were harvested, resuspended in lysis buffer (50 mM Hepes (pH 7.5), 300 mM NaCl, 20 mM imidazole, protease inhibitor (Roche), DNAse (Sigma) and lysozyme (Sigma), homogenized for 1 h at 4 °C and lysed by sonication. Lysate was then cleared by centrifugation and the supernatant filtered at 0.45 µm and loaded onto Ni-NTA resin. Bound proteins were washed with 50 mM Hepes (pH 7.5), 500 mM NaCl, 20 mM imidazole. Protein was eluted with 50 mM Hepes (pH 7.5), 150 mM NaCl, 250 mM imidazole. The eluate was then dialyzed against Hepes (pH 7.5), 150 mM NaCl, 0.5 mM TCEP and simultaneously digested with 1 mg 3C-protease. Undigested protein and 3C-protease were removed by a second Ni-NTA step and flowthrough containing OCS was concentrated to 5 mL for subsequent gel filtration (Hiload 16/60 Superdex 200 pg, GE healthcare). Relevant fractions were pooled together, and the protein was concentrated, flash frozen in liquid nitrogen and stored at −80 °C. After 3C-protease cleavage, five additional residues (GPLGS) from the plasmid backbone remain at the OCS N-terminus.

Crystallization and X-ray structure determination

OCS (wt) was crystallized by vapor diffusion using conditions from the PEG I crystallization suite (Molecular Dimensions). OCS (wt) at a concentration of 6 mg/mL was crystallized in 0.2 M LiCl, 0.1 M Hepes (pH 8.0), 15% w/v PEG-3350. The OCS KD variant was crystallized in 0.15 M ammonium sulfate, 0.1 M MES (pH 6.0), 15% (w/v) PEG-4000. Crystals were soaked in cryo-solution, containing the crystallization mother liquor supplemented with 25% [v/v] ethylene glycol. Crystals were then mounted on cryo-loops (Hampton Research), and flash-cooled in liquid nitrogen. X-ray diffraction data of OCS (wt) and the OCS (KD) variant were collected on the synchrotron radiation beamlines MASSIF-1/ID30A-1 (ESRF, Grenoble) and EMBL P13 at PETRA III (EMBL/DESY, Hamburg), respectively. Data were processed using XDS (Kabsch 2010).

The OCS (wt) structure was solved by molecular replacement (MR) with PHASER (McCoy et al. 2007) using 4-Coumaroyl-CoA Ligase from A. thaliana (3TSY) as a search model. The initial MR solution was rebuilt interactively in COOT (Emsley et al. 2010). Each asymmetric unit contains 12 OCS protomers, grouped into two OCS hexamers, which are virtually identical in the OCS NTD arrangement. GESAMT (Krissinel 2012) superposes 2532 C-alpha atoms of the two hexamers NTD domains with rmsd of 0.525 Å. To simplify building CTD fragments that were relatively poorly resolved in the map, we used AlphaFold2 version 2.1.1 predicted structure (plddt > 90) (Supplementary Figure S12) (Jumper et al. 2021). The predicted CTD structure was rigid body fitted to the map and real space refined in COOT with self-restraints generated at 5 Å cut-off for all the chains. After subsequent refinement in REFMAC (Murshudov et al. 2011) version 5.8.0267 with NCS restrains and automatically defined TLS groups, CTDs with map cc’s < 0.7 (CC_mask calculated using phenix.map_model_cc (Liebschner et al. 2019)) were completely removed from the model. From the remaining CTDs, residues without interpretable 2FoFc density at 1σ level were removed. The resulting model contained two OCS CTDs for the first hexamer and one CTD for the second hexamer (cf. Table 2). The final model was refined to R/Rfree of 0.23/0.25 at 2.87 Å resolution.

The OCS (KD) mutant structure was solved by MR with PHASER using the OCS (wt) NTD (residues 1–431) as a search model. The model was completed with AlphaFold2 predicted CTD structure and refined as described above. For the mutant structure we built a fragmented CTD for one out of two OCS chains in the asymmetric unit (cf. Table 2). The final model refined to R/Rfree of 0.21/0.25 at 2.45 Å resolution. Further details are summarized in Supplementary Table S1. The X-ray structures of OCS (wt) and OCS (KD) have been deposited in the Protein Data Bank with the following codes, respectively: 8AFF, 8AFG.

OCS interfaces were analyzed using the program PISA (Krissinel and Henrick 2007). Angles between OCS (wt) dimers were calculated using the PSICO (https://github.com/speleo3/pymol-psico) plugin in Pymol (PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC). NTD/CTD arrangements were analyzed with the online version of DynDom (Hayward and Lee 2002; Veevers and Hayward 2019).

Negative stain EM

Four µL of OCS sample at a concentration of 0.003 mg/mL in TBS Buffer (50 mM Tris pH 8.0, 150 mM NaCl) was applied for 2 min on a freshly glow discharged carbon coated copper-grid (Agar scientific; G2400C) at room temperature. The sample was blotted with Whatman paper (No 5) and subsequently the grid was washed two times using 10 µL TBS buffer. A 10 µL drop of freshly prepared 0.75% uranyl formate solution was then applied on the grid for 30 s. Afterwards the excess of liquid was blotted away and the specimen was air-dried. Images were recorded with a JEM-1400 (JEOL) or a FEI Tecnai G2 Spirit (FEI) both equipped with LaB₆ cathode and a 4 K CMOS detector F416 (TVIPS). Single particles were selected using crYOLO (Wagner et al. 2019). Reference-free stable class-averages were computed using the ISAC approach of the SPHIRE software package (Moriya et al. 2017).

Crosslinking

Crosslinking with glutaraldehyde was performed in 50 mM Hepes (pH 8.0), 150 mM NaCl. Reaction mixtures with an amount of 0.003 mg/mL OCS were treated with 0.05% crosslinking reagent for 2 min and afterwards quenched by adding Tris (pH 8.0) to a final concentration of 100 mM. Prior to negative stain EM imaging, samples were purified via SEC using a Superose 6 Increase 5/150 GL.

Cryo-EM structure determination

For cryo-EM, 4 µL of sample at 0.3 mg/mL concentration was applied to freshly glow discharged holey carbon Quantifoil 2/1 grids (#EMS300-Cu). The sample was immediately blotted and plunged into liquid ethane cooled by liquid nitrogen using a Gatan Cryo-plunge 3 device at 95% humidity. The blotting time was 2.5 s at 25 °C (gentle blot). A cryo-EM dataset was collected with a C_s-corrected Titan Krios EM (ThermoFisher) equipped with a K2 direct electron detector (Gatan) and a quantum energy filter (Gatan). Movies were recorded in counting mode using the automated acquisition program EPU (ThermoFisher) at a magnification of 105,000x corresponding to a pixel size of 1.09 Å. 5594 movies were acquired in a defocus range of −0.7 to −2.2 µm. Each movie comprised 60 frames acquired over 12 s with a total dose of ∼63 e/Ų.

Monitoring of image collection as well as image pre-processing was performed using the TranSPHIRE software package (Stabrin et al. 2020). Motion correction was thereby performed on the fly using MotionCor2 (Zheng et al. 2017). The contrast transfer function was estimated using gCTF (Zhang 2016). Ten representative images were then manually picked using the crYOLO-box manager tool, to train crYOLO (Wagner et al. 2019) for further automated particle picking. A total of four million particles were finally selected and extracted using a box size of 182 pixels. SPHIRE (1.0, 1.2, 1.3) (Wagner et al. 2019) was used for further single particle analysis steps. Multiple reference-free 2D classification rounds were then carried out, using the stable alignment and cluster approach (ISAC) (Yang et al. 2012) in SPHIRE (Supplementary Figure S7). After each round of 2D classification, we selected classes containing intact OCS populations, comprising OCS dimers, tetramers and hexamers. The resulting class averages were selected and subsequently used for VIPER (Penczek and Asturias 2014), to generate a low-resolution initial model. The respective class members for each oligomerization state were then subjected to 3D refinement in SPHIRE (Meridien) with C1 (dimeric and tetrameric particles) or C2 (hexameric particles) symmetry, imposed using the VIPER reconstruction as initial reference, respectively. The resulting projection parameters were used to re-center and re-extract the particles. The re-centered particles were again subjected to 3D refinements in SPHIRE (Meridien). Particle “polishing” was then performed using RELION (Zivanov et al. 2019). Final refinements were then performed in SPHIRE using the polished particles of each individual population, respectively. Fourier shell correlation analysis, 3D masking, sharpening and low-pass filtering were automatically performed using the PostRefiner tool of SPHIRE. Further details are given in Supplementary Table S2.

Cryo-EM maps and models were visualized using Chimera (Pettersen et al. 2004) and ChimeraX (Pettersen et al. 2021). Map segmentation was performed using the module Segment Map of the Chimera software package. Structure modeling of OCS was performed using COOT (Emsley et al. 2010). Real space refinement of the crystal structure in the EM density map was performed in PHENIX (Adams et al. 2011).

The resulting model coordinates are available in the Protein Data Bank with the PDB ID 8ATD. The cryo-EM maps have been deposited into the Electron Microscopy Data Bank with the ID EMD-15,646.

Small angle X-ray scattering analysis

Synchrotron SAXS data (I(s) versus s, where s = 4πsinθ/λ, 2θ is the scattering angle and λ the X-ray wavelength, 0.124 nm) were measured from dialyzed solutions of OCS (wt) and OCS (KD) in 50 mM Hepes (pH 7.5), 150 mM NaCl, 0.5 mM TCEP at the EMBL P12 beam line at DESY (Blanchet et al. 2015) (Hamburg, Germany) using a Pilatus 6 M detector at a sample-detector distance of 3 m. OCS (wt) and OCS (KD) protein solutes at 7 or 5, 2.5 and 0.5 mg/mL concentrations were measured at 10 °C. A continuous flow 1 mm cell capillary was used to limit radiation damage. A total of 30 successive 0.1 s frames were collected for both the samples and the corresponding matched buffer. The process of 2D data reduction to generate 1D buffer-subtracted profiles was performed using the SASFLOW pipeline (Franke et al. 2012).

Data were analyzed using modules of the ATSAS 3.0.1 package (Franke et al. 2017). CRYSOL (Franke et al. 2017) was used to calculate the scattering profiles from the atomic coordinates of the OCS (wt) and OCS (KD) X-ray structures. CTDs were added to OCS protomers without CTDs modeled in the respective X-ray structures to create complete OCS model and fit them with the respective experimental SAXS data.

To characterize mixtures of the different oligomeric OCS species, the program OLIGOMER (Konarev et al. 2003) was used. It fits the observed experimental data by a weighted combination of the calculated model scattering curves (form factors) from different quaternary structures. The experimental SAXS data and models are deposited in SASBDB (Kikhney et al. 2020) (https://www.sasbdb.org/) with the accession codes SASDPV3, SASDPW3, SASDPX3, SASDPY3. Further details are presented in Supplementary Table S3.

Size-exclusion chromatography

Purified OCS (wt) and (KD) were analyzed by size-exclusion chromatography using a Superdex 200 Increase 10/300 GL (Cytiva) connected to a Aekta pure chromatography system. 200 µL of protein at 10 mg/mL were injected and analyzed at a flow of 0.5 mL/min. Injection of a standard (Bio-Rad) mixture of protein of known concentration and molecular weight (thyroglobin 670 kDa, γ-globin 158 kDa, ovalbumin 44 kDa, myoglobin 17 kDa and vitamin B₁₂ 1.35 kDa) was used to estimate OCS (wt) and (KD) molecular weight.

Size-exclusion multi angle light scattering

Protein mass measurements were performed on an Agilent HPLC system connected to a MiniDAWN^® TREOS^® multi-angle laser light scattering detector (MALS), with an Optilab T-rEX (RI) refractometer (Wyatt, Germany). 50 µL of sample at 10, 5, 1 mg/mL for OCS (wt) or 5 mg/mL for OCS (KD) were loaded onto a Superdex 200 Increase 5/150 GL (Cytiva) and the MALLS and RI measurements were performed at 25 °C, where the refractive index increment (dn/dc) of the protein samples was set to 0.185 mL/g. The data were processed using Wyatt ASTRA 7.0.1 software (Wyatt).

Nano differential scanning fluorimetry

Melting temperature were measured by NanoDSF using a Nanotemper Prometheus NT.48. Tryptophan intrinsic fluorescence emission was recorded at 330 and 350 nm with 30% excitation power during thermal ramping from 20 °C to 90 °C at a ramping velocity of 1 °C per minute. A volume of 10 μL OCS in 50 mM Hepes (pH 7.5), 150 mM NaCl, 0.5 mM TCEP and at a concentration of 1 mg/mL were loaded in NanoDSF grade standard capillaries. Samples were measured in triplicates and the calculated melting temperatures averaged. Melting temperature were calculated by the software PR. ThermControl (Nanotemper).

Enzymatic activity assay

Enzymatic activity was measured by AMP production at 254 nm after 60 s reaction at 25 °C. The sample volume was 40 μL, containing 50 mM Hepes (pH 7.5), 150 mM NaCl, 2 mM MgCl₂, 0.3 mM CoA, 0.4 mM ATP and 1 µM OCS (wt), OCS (KD) or OCS (K523A, negative control), supplemented with 100 µM oxalate. Reactions were stopped by adding 10 µL of 1 mM HCl and kept on ice. Reaction tubes were centrifuged for 10 min at 4 °C, 14,000 rpm on a tabletop centrifuge to remove aggregated proteins and then sequentially injected into an HPLC (Agilent) equipped with a C18 analytical column (Agilent Poroshell 120).

For high-performance liquid chromatography analysis, solvent A was 50 mM ammonium acetate (pH 5.0) and solvent B was 100% methanol. The flow rate was kept at 1 mL/min and at room temperature. Samples were kept at 10 °C in a 96 well V-bottom plate and sealed to avoid evaporation. The elution method consisted of 0% solvent B for 5 min and then a continuous gradient from 0% to 70% of solvent B for 15 min, and then 100% solvent B for 5 min and re-equilibration of the column in solvent A. Samples with only ATP, AMP and CoA were analyzed to define the elution time of each component. AMP production was assessed by measuring the area under the curve between 0.75 and 1.00 min of elution. Curves were integrated using GraphPad Prism. All reactions were run in triplicate.