Structure determination of Y. enterocolitica urease by cryo-EM

We have used single particle cryo-EM to determine the structure of the fully assembled Y. enterocolitica urease. We acquired 4494 movies of urease particles using a Titan Krios transmission electron microscope (TEM) equipped with a K2 direct electron detector and an energy filter (see “Methods” for details). Approximately half of the movies (2243) were acquired by illuminating three locations (shots) per grid hole using beam-image shift in order to speed up the data collection¹⁴, whereas the remaining movies were recorded without this feature i.e., just a single shot at the center of the hole. This allowed us to measure and assess the extent of beam tilt and other optical aberrations, as well as the behavior of sample drift between each condition and beam-image shift position. Typical micrographs from the imaged grids are shown in Supplementary Fig. S1a and a summary of data collection information is given in Table 1.

Each dataset was processed separately for 3D reconstruction following the strategy depicted in Supplementary Fig. S2. The first obtained 3D map, at an overall resolution of 2.6 Å, revealed that this urease assembly is a dodecamer of tetrahedral (T) symmetry with a diameter of ~170 Å. The separate processing of each dataset yielded refined 3D maps at nominal resolutions of 2.10 Å and 2.20 Å for the multi-shot and single−shot cases, respectively (see “Methods”).

For comparison, we also processed the merged set of particles from both datasets altogether. We observed on the 2D class averages a preferential orientation for the threefold symmetric view of urease, and also the presence of isolated monomers and broken assemblies (Supplementary Fig. S1b). The presence of such incomplete assemblies was further confirmed by performing 3D classification without imposing symmetry, as shown in Supplementary Fig. S1c. The 3D class corresponding to the complete dodecameric assembly of urease contained 119,020 particles, of which 69,512 (58.4%) came from the multi-shot and 49,518 (41.6%) from the single-shot dataset. With respect to the number of particles picked from each dataset, 64.7% of the particles from the multi-shot and 56.8% from the single-shot datasets were retained at this stage and throughout the final reconstruction. While coma-free alignment was performed and active beam-tilt compensation in SerialEM¹⁵ was used on our data collections, after performing beam tilt refinement in RELION-3¹⁶ we observed that the single-shot case has a residual beam-tilt higher than the smallest residual observed in the multi-shot case (Supplementary Table S2). These two values are however very close to zero and are possibly within the error margin of the post hoc beam tilt refinement procedure.

The reduced need to move the specimen stage in beam-image shift mode not only speeds up data collection but also minimizes stage drift. The second and third shots from the multi-shot dataset have comparatively less drift than both the first multi-shot and the single shot, as suggested by the parameter values obtained from the Bayesian polishing training¹⁷ on each beam-tilt class separately (Supplementary Table S3). As all the three multi-shots are taken in nearby areas within the same foil hole, this observation is consistent with movement by the annealing behavior of the vitreous ice layer and its carbon support after pre-irradiating the specimen as reported previously¹⁸.

At this point, the nominal resolution of the map after 3D refinement was 2.05 Å. Finally, correcting for residual higher-order aberrations in CTF refinement¹⁹ (Supplementary Fig. S3) yielded a map at a global resolution of 1.98 Å (Supplementary Fig. S4a). Local resolution estimation reveals that the core of the map is indeed at this resolution level or better (Fig. 2a and Supplementary Fig. S4b), and the local resolution-filtered map was then used for model building as explained in the next section. Despite the 12-fold symmetry of the urease assembly, a limiting factor in the resolution of the map is the strong presence of preferential orientation, as confirmed by the plot of the final orientation assignments (Supplementary Fig. S4c). The estimated angular distribution efficiency is 0.78²⁰. An overview of the cryo-EM map and its main features are depicted in Supplementary Movie 1.

Y. enterocolitica urease assembles as a tetramer of trimers

Model building was initiated from available crystallographic models with subsequent fitting and refinement against the cryo-EM map. The model was built and refined for one asymmetric unit containing one copy of the ureA, ureB, and ureC protein each. The model was then expanded using NCS (see “Methods”). The complete model covering the whole oligomeric assembly contains 9552 residues, 3672 waters and 24 nickel ions (two per active site, twelve active sites) (Table 1). The quality of the model was assessed with the cryo-EM validation tools in the PHENIX package²¹. The map allowed for the building of all residues of ureA (1–100), and residues 31–162 of ureB and 2-327/335–572 of ureC (Supplementary Fig. S5). The hetero-trimer formed by the three protein chains (ureA, ureB, ureC) (Fig. 1b) oligomerizes into a homo-trimer. The homo-trimer is arranged in a tetramer-of-trimers making the full complex a dodecamer of the hetero-trimer (Fig. 2b). There are four large oval shaped holes between the trimers (64 Å long, 12 Å wide, high electrostatic potential) and four smaller holes at the center of the trimer with a diameter of 6 Å (low electrostatic potential), as shown in Fig. 2b, and the center of the enzyme assembly is hollow (Fig. 2). The holes provide ample opportunity for diffusion of the uncharged substrate and product, and the hollow inside potentially leads to a local increase of reaction product. The assembly has the same symmetry as the urease homolog in H. pylori, which was postulated to increase stability and/or resistance to acidic environments²².

For analysis of the protein sequences, the ConSurf Server^23,24 was used with the sample list of homologs option to get a diverse set of 150 sequences. The protein chains of Y. enterocolitica urease are highly conserved across different organisms. The ureA chain is split after a LVTXXXP motif and is 99–100 amino acids long in most cases, with a sequence identity of 55.7%. The ureB chain of Y. enterocolitica has between 20 and 30 N-terminal amino acids more compared to the other sequences (except Kaistia sp. SCN 65-12), which share an identity of 51.5%. This N-terminal extension is located on the outside of the holoenzyme where ureA and ureB chain split occurs and are too disordered to be modeled in the structure (Supplementary Fig. S5). The charges and properties of this stretch of amino acids vary and if they still serve a function remains unclear. The last 20 amino acids of the C-terminus of ureB are only represented in half of the compared sequences and accurate sequence conservation could not be determined in this part. This stretch contains a loop and a C-terminal helix (Supplementary Fig. 5). The ureC protein of the compared sequences has a shared sequence identity of 60.3%. All amino acids involved in catalysis are highly conserved (Supplementary Fig. S5). The ureA and ureB chains show lower conservation compared to ureC. They are not involved in catalysis but in scaffolding, so the differences could stem from their role in different types of oligomeric assembly (Fig. 3).

To investigate this aspect further, we compared the presented structure to the ureases of H. pylori, S. pasteurii, and K. aerogenes. Sequence identity scores among these ureases are provided in Supplementary Table S4. The H. pylori urease is made up of two protein chains ureA (that contains the equivalent of ureA and ureB in Y. enterocolitica) and ureB (that is the equivalent of ureC in Y. enterocolitica) (Fig. 1c and Fig. 3c). It assembles into a T-symmetric oligomer like in Y. enterocolitica and the crystal structure was solved to 3 Å. S. pasteurii and K. aerogenes ureases both assemble into a trimer from the hetero-trimeric unit (Figs. 1c,  3b). S. pasteurii urease has been solved by X-ray crystallography in different conditions (for example, PDB IDs: 2UBP, 3UBP, 4CEU)^25,26. Here we use the highest resolution urease structure, that was solved in the presence of the inhibitor N-(n-Butyl)thiophosphoric Triamid (NBPT) to 1.28 Å, for comparison (PDB ID: 5OL4)²⁷. The crystal structure of K. aerogenes used for comparison in this paper has a similar resolution range and was solved to 1.9 Å in absence of substrate or inhibitors (PDB ID: 1EJW)¹¹.

There are two main regions with high root mean square deviations (RMSDs) when comparing these three ureases to the Y. enterocolitica model (Supplementary Fig. S6 and Supplementary Table S5). The first region with high deviation is the mobile flap, which opens and closes over the active site (residues 312–355 of ureC). Both the open and the closed conformations of the mobile flap have been observed in crystal structures, stabilized at pH values lower and higher than the pKa of the conserved His323, respectively⁸. The residues of its connecting loop could not be built with confidence in the cryo-EM model (residues 326–333 of ureC). The other region with large differences is on the edges of ureA and ureB where the interactions with the next protomer occur. The H. pylori assembly contains an additional C-terminal loop (residues 224–238 of ureA) after the top alpha helix (residues 206–223 of ureA). This helix (central helix) forms the threefold axis of three neighboring trimers and the loop binds in a head-to-tail fashion to the next trimer forming the tetramer (Fig. 3a, c)²². The core of the assembly is identical in its structure. For whole-chain superposition scores and RMSD values between the compared models please see Supplementary Table S5.

In the dodecameric assembly seven different interfaces are formed between the hetero-trimers (Fig. 4a, b and Supplementary Fig. S7). Intra-trimer interactions occur between the three basic hetero-trimers in one assembled trimer, forming a threefold symmetry axis (Fig. 4a). The interactions between these trimers to form the tetramer then make up a different threefold symmetry axis (Fig. 4b). The three largest interfaces (interfaces 1–3) are formed intra-trimeric between ureC of one hetero-trimer and ureC, ureA, and ureB of the next trimer (Fig. 4a). The three ureA proteins make up the intra-trimer-core (first threefold axis) with interface 4 (Fig. 4c and Supplementary Fig. S7). Comparison of the interface areas formed in the trimer assembly shows no substantial differences between the four organisms (Supplementary Fig. S7). Inter-trimer interfaces (interfaces 5, 6, 7) formed in the dodecameric Y. enterocolitica and H. pylori ureases have similar areas (Fig. 4b and Supplementary Fig. S7). Part of the interactions occur between ureB and ureC forming interfaces with each other (interface 4, 6). The other interaction is between the three ureB proteins and forms interface 7 and the inter-trimer-core (second threefold axis) with their central helices (Fig. 4b). Y. enterocolitica does not have the same oligomerization loop after the central helix proposed for H. pylori. However, there is a short loop before the central helix, which is extended in Y. enterocolitica. It binds into a pocket of ureC of the neighboring trimer in interface 6 (Fig. 4e). These types of loops or extensions are missing from S. pasteurii and K. aerogenes ureB proteins. S. pasteurii has the central helix, but there is no extended loop before or after it (Supplementary Fig. S6b). K. aerogenes urease does not have a helix nor a loop in this region (Fig. 3b). This suggests that the presence of oligomerization loops in ureB is crucial for determining the oligomeric state of the enzyme.

The dodecameric holoenzyme structure of ureases might aid in stabilizing the protein at acidic pH, and in combination with 12 active sites producing ammonia enables the formation of a pH-neutralizing microenvironment around the assembly²². This ensures the continued function of the enzyme and makes this type of oligomeric assembly essential to survival of Y. enterocolitica in the host. It is remarkable that Y. enterocolitica is the first organism outside the Helicobacteraceae family to have a known dodecameric urease. Considering the different subunit organization between these ureases, it raises the question of what particular events in the evolutionary history of Y. enterocolitica could have led to this type of assembly²⁸.

The empty active site is filled with water

At the global resolution of 1.98 Å, detailed structural features can be observed. All throughout the highly resolved areas of the protein, salt bridges, backbone and side chain hydration, and alternative side chain conformations can be visualized (Supplementary Fig. S8a–c). Furthermore, the high resolution allows for a detailed description of the nickel-metallo-center and the active site. The active site is located on the ureC protein at the edge of the hetero-trimer and is wedged in between the ureA and ureB proteins of the next hetero-trimer in the homo-trimeric assembly (Fig. 5a).

The catalysis of ammonia and carbamate from urea occurs in two steps (Fig. 1a). Urea first interacts with the nickel ions through its carbonyl oxygen and amino nitrogens. The active site contains two Ni²⁺ ions which are coordinated by six different amino acids (Fig. 5b). Both Ni²⁺ ions are coordinated by the carbamylated Lys222*. Ni(1) is additionally coordinated by His224, His251 and His277 and Ni(2) by His139, His141, and Asp365. Close to the active site is a methionine (Met369), which can be modeled in different alternative conformations. One conformation could potentially reach the active site. There is no described function for this amino acid (Fig. 5c and Supplementary Fig. S9).

The active site is protected by a helix-turn-helix motif, called the mobile-flap. Its function is to coordinate the access of substrate to the catalytic site and the release of the product from it^3,4. The protonation state of a conserved histidine on the mobile flap (His325) is essential for catalysis by determining opening and closing of the mobile flap and strongly depends on the solution pH^4,8 (Fig. 5a). By closing of the mobile flap His325 moves closer to the active site, stabilizing the distal amine of urea in the active site pocket^3,4,8,9. After closing of the mobile flap, the urea molecule chelates the two Ni ions in the active site, and following the nucleophilic attack by the bridging hydroxide onto the urea C atom, a proton is transferred to the distal amine group from the metal-bridging C–OH group, yielding an ammonia molecule after breakage of the resulting C-NH3+ bond. Flap opening then releases ammonia and carbamate, where the latter spontaneously hydrolyzes into another molecule of ammonia and bicarbonate. The mobile flap of the cryo-EM structure presented here is modeled in an open position, however the local resolution is lower than in the surrounding areas, indicating flexibility. The sample was frozen in a buffer of pH 7.0, where the mobile flap of urease can adopt both open and closed conformations⁸. The twelve active sites of each particle adopting different conformations are averaged by single particle reconstruction with symmetry imposition into an mainly open conformation. In the absence of substrate or inhibitors in the sample, the mobile flap cannot be stabilized in one conformation (Fig. 5a). Coordinated water molecules can be seen in the empty pocket of the active site, which do not only form hydrogen bonds with side chains or the protein backbone, but also with each other constituting a hydration network (Supplementary Fig. S8d).

The resolution in the active site is sufficient for complete atomic description of the coordinated Ni²⁺ ions, including the carbamylated lysine. The protonation states of the active site residues are also represented in the map (Fig. 5b). One of the hydroxide molecules in the active site is essential as it performs the nucleophilic attack on urea while other molecules are displaced by urea and the closing of the mobile flap^3,4,8,9.

Comparison to the crystal structure of K. aerogenes of similar nominal resolution (1.9 Å) shows differences in the visualization of these features. This crystal structure was solved in absence of inhibitors or substrate such that the active site is also empty and the mobile flap in an open conformation (Fig. 5a). The details of the map provides finer details around the Ni²⁺ ions in the cryo-EM map than the crystallographic data. The protonation of the histidines is clearly visible in the cryo-EM density (Fig. 5b). The positions of the side chains and the Ni²⁺ ions in the active site are very similar to the Y. enterocolitica urease structure with a RMSD of 0.270 Å (Supplementary Table S5). The highest resolution S. pasteurii crystal structure was solved in presence of the inhibitor NBPT, which displaces the essential water molecules needed for the reaction from the active site. The closing of the mobile flap displaces the rest of the waters and brings the catalytic His323 closer to the active site. The tight packing of side chains prevents urea from entering the active site, efficiently blocking it (Fig. 5b, e). The active site residues and Ni²⁺ ions have a RMSD of 0.293 Å between S. pasteurii and Y. enterocolitica.

Nickel atoms come closer together

The distance between the Ni²⁺ ions is 3.7 Å in X-ray structures of K. aerogenes and S. pasteurii, but only 3.2 Å in the Y. enterocolitica cryo-EM model (Fig. 5b, d). Short distances of 3.1–3.3 Å were described for S. pasteurii and K. aerogenes at high resolutions for structures in presence of β-Mercaptoethanol (β-ME)²⁹. Knowing that metallic cores are particularly sensitive to radiation³⁰, we tried to determine the extent to which radiation damage can explain the shorter distance between the Ni²⁺ ions. For this purpose, we generated per-frame reconstructions for the first 25 frames of our data collection, refined the model on each of them (see “Methods”), and measured the distances between the residues involved in ion coordination, shown in Fig. 6. Bayesian particle polishing¹⁷ was run again before calculating each per-frame reconstruction. At the beginning of the exposure, in which the frames contribute more to the full reconstruction due to dose weighting^17,31, there is a trend of the ions coming closer together (Fig. 6a−i). While we cannot determine exactly how this arises from radiation damage, it is likely a result of several interactions in the active site changing simultaneously along the exposure. For example, both Ni(1) and Ni(2) tend to come closer to the carbamylated Lys222 (Fig. 6a-iv, v) as Asp365 vanishes (Fig. 6a-vi), which can be seen in the Supplementary Movie 2. Aspartic acid is known to have its side chain damaged very early on³². The dynamic interplay between residues along the exposure (Fig. 6b) is likely due to the different rates at which specific types of bonds and residues are damaged³³: first negatively charged residues, then positively charged ones followed by aromatic side chains, as also observed in Supplementary Movie 2. A possible explanation of how these events may account for the shorter distance between the Ni²⁺ ions is then that their bridging hydroxide molecule becomes deprotonated into the oxide form, either by radiation damage directly or by local pH changes arising from it. The oxide form is known to have a more favorable ferromagnetic interaction with the two nickel ions²⁵, although not found in ureases under native conditions³⁴. Furthermore, the B-factors suggests that some residues in the active site, and in particular the Ni²⁺ ions, are indeed damaged more strongly than the rest of the protein, right from the beginning of the irradiation as shown in Fig. 6c. We note however that, in the present analysis, radiation damage cannot be completely disentangled from other effects such as residual sample movement, which is especially difficult to correct in the initial frames of the exposure. The later part of the exposure must also be interpreted with caution, as atomic coordinates become less reliable, which is verified by the overall increase in B-factors in Fig. 6c and the error bars in Fig. 6a.