Single‐particle analysis of urea amidolyase reveals its molecular mechanism

Abstract

Urea amidolyase (UA), a bifunctional enzyme that is widely distributed in bacteria, fungi, algae, and plants, plays a pivotal role in the recycling of nitrogen in the biosphere. Its substrate urea is ultimately converted to ammonium, via successive catalysis at the C‐terminal urea carboxylase (UC) domain and followed by the N‐terminal allophanate hydrolyse (AH) domain. Although our previous studies have shown that Kluyveromyces lactis UA (KlUA) functions efficiently as a homodimer, the architecture of the full‐length enzyme remains unresolved. Thus how the biotin carboxyl carrier protein (BCCP) domain is transferred within the UC domain remains unclear. Here we report the structures of full‐length KlUA in its homodimer form in three different functional states by negatively‐stained single‐particle electron microscopy. We report here that the ADP‐bound structure with or without urea shows two possible locations of BCCP with preferred asymmetry, and that when BCCP is attached to the carboxyl transferase domain of one monomer, it is attached to the biotin carboxylase domain in the second domain. Based on this observation, we propose a BCCP‐swinging model for biotin‐dependent carboxylation mechanism of this enzyme.

1. INTRODUCTION

Urea is a metabolic product in many living organisms after degrading nitrogen‐containing molecules. Transformation of urea to ammonium is an imperative step in nitrogen recycling in the biosphere.1 Urea amidolyase (UA), widely distributed in bacteria, fungi, algae, and plants, is a single polypeptide enzyme responsible for catalyzing this reaction in an ATP‐dependent manner, enabling urea to be utilized as a nitrogen source.2, 3, 4, 5, 6, 7, 8, 9 This is distinct from urease, a nickel‐containing enzyme that also degrades urea but in an ATP‐independent manner.10, 11, 12

Kluyveromyces lactis UA (KlUA) is composed of 1,829 amino acids, forming a compact homodimer in solution with a molecular weight of 400 kD.13, 14, 15 The conversion of urea begins at its C‐terminal urea carboxylase (UC) domain, processing the carboxylation of urea to generate the intermediate product allophanate, which then presumably diffuses into the N‐terminal allophanate hydrolyse (AH) domain to be further hydrolyzed into ammonium.14, 15, 16 The UC domain belongs to the biotin‐dependent carboxylase family which share common features of domain organization.2, 17 They all consist of biotin carboxylase (BC), biotin carboxyl carrier protein (BCCP), and carboxyl transferase (CT) domains, but in different arrangements (Figure 1a). Biotin, covalently tethered to the BCCP domain as a cofactor, undergoes carboxylation in the BC domain and transcarboxylation of carboxyl to the substrate bound in the CT domain which diversifies in functions in the family.2, 18, 19 Since the biotin is conjugated to the lysine of BCCP through a series of methylene groups and rotable single bonds, it can reach a length of ~16 Å if fully extended20; therefore, the arm length between the BC and CT domains can be up to ~30 Å.21 Nevertheless, the measured distances between the active sites of BC and CT are in the range of 55–85 Å among the family members (60 Å in KlUA case),14, 20 which is a much longer distance for an extension of a biotin fixed in one location. It is proposed that BCCP must swing between the two domains to make two successive reactions possible. Thus, it is supposed that only BCCP swinging between the two domains can make this reaction possible. Although crystal structures of several biotin‐dependent carboxylases have been determined in the holoenzyme form, such as pyruvate carboxylase (PC),22, 23 propionyl‐coenzyme A carboxylase (PCC),24 3‐methylcrotonyl‐CoA carboxylase (MCC),25 UC,14 acetyl‐CoA carboxylase (ACC),26 all of them captured only a single conformation with the BCCP domain located either in‐between BC and CT, or close to the CT domain (summarized in Table S1). Obviously, further structural evidence still needs to be provided to corroborate the proposed BCCP‐swinging model.

Figure 1.

Schematic drawings of domain organization in representative biotin‐dependent carboxylases. The amino acid numbers marked above are defined based on the individual domain boundaries. RePC, Rhizobium etli pyruvate carboxylase; ScACC, Saccharomyces cerevisiae acetyl‐CoA carboxylase; HsPCC, human propionyl‐coenzyme A carboxylase; HsMCC, human 3‐methylcrotonyl‐CoA carboxylase. The biotin carboxylase (BC), carboxyl transferase (CT), and biotin carboxyl carrier protein (BCCP) domains are colored in light red, gold, and dark red, respectively. In KIUC, CT–BCCP linker (a.a. 1,742–1,759) is in‐between the CT and BCCP domains. In RePC, the central allosteric domain [CAD (a.a. 417–489, 1,002–1,073)] is colored in white. In ScACC, the BC–CT interaction domain (BT) and central region (CR) is colored in light and dark gray, respectively. Instead of a single chain sequence, HsPCC and HsMCC are split into α and β subunits

In the present study, we prepared monodispersive, the full‐length KlUA samples in three functional states by using negative‐staining procedure and reconstructed individual three‐dimensional structures. Unexpectedly, the ADP‐bound structures of the homodimeric enzyme with or without urea exhibit an asymmetric conformation with two BCCP domains preferentially bound to two different locations. This feature is specific to the ADP‐bound form because without ADP, BCCP domain appears completely disordered. Furthermore, this asymmetric conformation could also be taken as a possible synergistic catalytic mechanism for KlUA. When the BCCP of monomer A is carboxylyzed in the BC domain, the BCCP of monomer B is transferring carboxyl to urea in the CT domain. This may also explain the much more efficient activity of KlUA in homodimer form than the monomeric form. Besides, our model capturing BCCP concurrently presented in both BC and CT within a homodimer could be taken as a further evidence for the BCCP‐swinging model proposed previously. Combining all these observations together, we conclude that the translocation of KlUA–BCCP domain is most likely in an alternating mode, a simple and efficient way.

2. MATERIALS AND METHODS

2.1. Protein expression and purification

The expression and purification of the full‐length KlUA was performed as described in previous work.13, 14 Briefly, the KlUA gene was inserted in the vector pET28a (Novagen). The plasmid was then transformed into Escherichia coli BL21 Star (DE3), which were cultured in lysogeny broth (LB) medium with 50 mg/L kanamycin and then induced with 0.3 mM isopropyl β‐d‐1‐thiogalactopyranoside. When the optical density at 600 nm reached 1.0 at 16°C, proteins were purified using successive nickel‐nitrilotriacetic acid (Qiagen), anion‐exchange (HitrapQHP, GE Healthcare) and size exclusion columns (Sephacryl S300 HR, GE Healthcare). After concentration, proteins were aliquoted and flash frozen in liquid nitrogen and stored at −80°C until use. Lys1729 was biotinylated as examined by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS‐PAGE), since after incubating with avidin, the protein migrated more slowly than the control.

2.2. Negatively stained sample preparation

For the ADP‐bound sample with or without 5 mM urea, the full‐length KlUA was diluted to 1 mg/ml using buffer containing 20 mM Tris‐HCI (pH 7.5), 110 mM NaCl, 2 mM dithiothreitol (DTT), and 5 mM ADP. Following incubation on ice for 15–30 min, the protein solution was diluted to an appropriate concentration with the same buffer for negative staining. A 4‐μl droplet was applied onto a glow‐discharged 300‐mesh copper grid covered with a thin carbon film (Xinxingbairui Company, Beijing). After 30 s, the excess solution was blotted by filter paper from the edge of the grid, and the sample was stained with 1% uranyl acetate for 30 s. The staining solution was then removed until a thin layer was left. After air drying, the sample was loaded into the electron microscope for visualization and data collection. For the apo‐state specimen, the sample preparation and negative‐staining procedure were the same as described above except for ADP and urea.

2.3. Data collection and image processing

Micrographs of the full‐length KlUA with ADP bound were collected on a Thermo Fisher Talos F200C equipped with Ceta camera at a nominal magnification of ×57,000 yielding a pixel size of 2.56 Å. The other specimens were imaged on a Thermo Fisher F20 equipped with a Gatan US4000 camera at a nominal magnification of ×50,000 with a pixel size of 1.7 Å. Electron dose of all specimens did not exceed 25 e⁻/Ųs. CTFFIND427 was used to estimate the contrast transfer function (CTF) parameters and the quality of each data set. Underfocus values for all data sets ranged from 0.6 to 2.2 μm. Only those micrographs without obvious drift and astigmation were used for the manual or automatic particle‐picking using the e2boxer.py program of EMAN228 or XMIPP29 software. Following 2–3 iterations of two‐dimensional (2D) classification performed by RELION2.1,30, 31 the particles showing high quality in class average were selected and proceeded to next iteration. All the routines used were wrapped in the SCIPION software package.32 The crystal structure of the KlUA–ΔBCCP homodimer resolved in our previous work was subjected to a Gaussian low‐pass filter of 60 Å and used as the reference model for the three‐dimensional (3D) classification. Only the class showing good electric potential density distribution of the overall architecture was used to perform 3D auto‐refinement with or without applying an overall soft mask. No symmetry was applied for the overall structure. For the purpose of improving the density of the central AH dimer, we performed a masked refinement33, 34 focusing on this region with local C2 symmetry applied. To further confirm the characteristic density that may assigned to BCCP, we subtracted the mask based on AH dimer from the one based on the overall structure using University of California San Francisco (UCSF) Chimera volumetric operation (VOP).35 The generated mask was applied to the UC portion for further 3D auto‐refinement without C2 symmetry constraint. The data processing procedure was summarized in Figure S1.

2.4. Rigid‐body docking

We used UCSF Chimera software to identify a dimer from the crystal structure determined previously in our laboratory of KlUA (PDB ID: 5I8I), and docked it into the electron microscopy (EM) density map at ADP–urea‐bound state by using the coordinate‐to‐map fitting procedures (i.e., Fit‐in‐map routine) to figure out the presumable density for the BCCP domain. For the ADP‐bound and an inactive and unbound state (APO) states, we docked the atomic structures of UC‐ΔBCCP (segment from PDB ID: 5I8I), complete UC (PDB ID: 3VA7) and AH dimer (PDB ID: 4ISS) separately into the corresponding subunit EM densities instead of the homodimer, since a slight rotation between the UC and AH domains of individual monomers has been found compared to the homodimer crystal structure.13

3. RESULTS

3.1. BCCP is asymmetrically located in two positions in homodimer for ADP–urea‐bound state

Although we obtained the crystal structure of the complete UC region including BCCP domain in our previous study,14 when we attempted to apply the same crystal growth conditions (containing both ADP and urea) into the full‐length KlUA crystallization experiment, it failed unless the BCCP domain was truncated.13 This suggests that the BCCP is not attached to the CT domain as was in the previous experiment. In other words, the previously obtained crystal structure probably captured one conformation of UC. To confirm our speculation, we initially performed in this study negatively stained single‐particle electron microscopy experiment on soluble full‐length KlUA which was fully biotinylated (Figure 2a) in the ADP–urea‐bound state. From 247 micrographs screened by CTF estimation, 7,869 particles were manually picked (Figure 2b). After two iterations of 2D classification using RELION, 7,557 particles were retained for the 3D classification using the low‐pass filtered crystal structure of the symmetric KlUA homodimer13 as the initial model to avoid model bias. More than one‐third of the particles contributed to a further 3D auto‐refinement (for details see Table S2). Intriguingly, when we focused on the UC portion of the entire reconstruction (Figure 2c), an extra, protruding electric potential density could be distinguished at either the BC or CT domain of individual monomers (both extra protruding densities were pointed by black arrows in Figure 2c). No atomic structure can be assigned to it when we docked the crystal structure of KlUA–ΔBCCP homodimer (PDB ID: 5I8I) into the refined structure (crystal structure colored in dark purple in Figure 2c). Then we tried to fit the UC structure (a.a. 617–1829, PDB ID: 3VA7), whose BCCP domain could occupy an extra density close to the CT domain (Figure 2d). We therefore inferred that the extra protruding density close to the BC domain in the opposite monomer is contributed by BCCP as well. To validate our speculation, we carried out a masked refinement on the UC portion of the homodimer. As expected, the density of BCCP close to CT domain became more noticeable, with the density of BC and CT becoming more compact (Figure 2d), while no further improvement was observed for the other monomer (UC2 in Figure 2c). The masked refinement processing confirmed that the protruding density is not from noise and the BCCP domain is not fixed in a unique position in the ADP–urea‐bound state. Considering the C2 symmetric characteristic of the central AH dimer in our previously determined crystal structure,13, 15 we also performed a masked refinement on the central density map with C2 symmetry applied, which improved the corresponding density to be more compact as well (colored in light purple mesh in Figure 2e) and showed a better fitting of the crystal structure of AH dimer (cyan compared to purple). The masked refinement based on not only the overall structure, but also the individual portions improved the resolution as shown in the Gold‐standard fourier shell correlation (FSC) curves at 0.143 criterion (Figure 2f).

Figure 2.

Biotinlyation and reconstruction of the full‐length Kluyveromyces lactis UA (KlUA) in ADP–urea‐bound state. (a) 8% sodium dodecyl sulfate–polyacrylamide gel electrophoresis in the absence or presence of avidin, with the higher molecular weight shift in the right lane of the protein confirming the biotinylation. (b) The left panel shows a typical negatively stained micrograph (scale bar 100 nm), with white circles indicating the dimer in various orientations. The related two‐dimensional (2D) power spectrum calculated by CTFFIND4 with an estimated defocus of −0.7 μm is on the upper right panel, while the lower panel shows the representative 2D class averages. (c) Two views of the reconstruction (showed in mesh mode, two urea carboxylase (UC) portions and allophanate hydrolyse (AH) dimer portion colored in light gray and light purple respectively) with the crystal structure of KlUA–ΔBCCP in dimer form (PDB ID: 5I8I, colored in dark purple) docked in it. The black arrows point to the vacant densities which are thought to be occupied by biotin carboxyl carrier protein (BCCP) domains. (d) The density map of UC1 portion after masked refinement and with the crystal structure of UC (colored in yellow, PDB ID: 3VA7) docked in it, compared with that of UC–ΔBCCP from the KlUA–ΔBCCP crystal structure (colored in dark purple). The right panel defined different domains of UC by different colors. From N‐ to C‐terminus, the biotin carboxylase, carboxyl transferase (CT), and BCCP domains are colored in yellow, light blue, and green, respectively. The CT–BCCP linker is colored in red. (e) Two views of the density map of the central AH dimer after masked refinement (showed in light purple mesh) and with C2 symmetry applied. A better fitting can be observed by comparing the occupation of the crystal structure of AH dimer (PDB ID: 4ISS) before (colored in dark purple) and after (colored in cyan) masked refinement. (f) Gold‐standard fourier shell correlation (FSC) curves indicate a final resolution of 29 Å (unmasked), 24 Å (masked), 18 Å (AH dimer masked, C2 symmetry applied) and 23 Å (UC portion masked), respectively

3.2. BCCP is asymmetrically located in either BC or CT domain in homodimer for ADP‐bound state

Stabilizing the location of BCCP should be the obstacle for resolving the structure of the full‐length KlUA at higher resolution. Considering the two‐step catalytic activity of the KIUC, we thought that the absence of urea may help BCCP stabilize at the BC domain as no substrate would attach it to the CT domain. Thus we performed the same experimental procedure as described above without urea added (Figure S2, for details see Table S2). Unexpectedly, almost a half of the particles used for 3D classification processing refined to a better‐defined density map exhibiting characteristics consistent with the ADP–urea‐bound state. The BCCP was located close to either the BC or CT domain in each monomer (Figure 3a). Considering the extended densities reconstructed at the central AH dimer (as marked by asterisks in Figure 3a), we performed the masked refinement focusing on not only UC, but also the central AH dimer. The outline of the AH dimer was improved significantly, as the density that inferred to be a merging with the adjacent BC or CT region of UC became more distinguishable and well defined (Figure 3b). For the UC region, no obvious improvement was gained (Figure 3b). Taken together, it demonstrates that whether urea is added or not, BCCP in the ADP‐bound structure is asymmetrically located in either the BC or CT domain. This serves as further evidence to understand the nonsynchronous translocation of BCCP in a KlUA homodimer for achieving more effective activity.

Figure 3.

Reconstruction of the full‐length Kluyveromyces lactis UA (KlUA) in ADP‐bound, an inactive and unbound state (APO) states and their comparison with ADP–urea‐bound state on urea carboxylase (UC) portions. (a) The reconstructed ADP‐bound model is shown in mesh mode (colored in purple), with the crystal structure of allophanate hydrolyse (AH) dimer (colored in cyan, PDB ID: 4ISS) and UC–ΔBCCP (colored in gold, from PDB ID: 5I8I) docked in it. The black arrows point to vacant densities with no docked structures. The extended densities reconstructed at the central AH dimer are marked by asterisks, which were improved significantly after the masked refinement. (b) The entire crystal structure of UC (in orange), AH dimer (in cyan) and UC–ΔBCCP (in gold) docked into the UC1 portion, center AH dimer and UC2 portion related to (a) after performing masked refinement. The black dotted circle indicates the presumable density for accommodating the biotin carboxyl carrier protein (BCCP) domain. (c) Structures of APO state (in yellow mesh) at two views for a better exhibition of the two UC portions (with the crystal structure of AH dimer and UC–ΔBCCP docked). The UC structures (docked with crystal structure of UC for UC1 and UC–ΔBCCP for UC2) in the lower center panel were generated with masked refinement. The superimposed structures of two opposite UC portions of KlUA in three states are represented in the orange box on the sides. (d) Difference maps (in mesh mode) generated by University of California San Francisco (UCSF) Chimera volumetric operation (VOP) of two pairs (ADP–urea‐bound state and APO state, ADP‐bound state and APO state) were presented for UC1 and UC2 respectively. The crystal structure of UC (left panel) and UC–ΔBCCP (right panel) were used as references for locating the difference maps. The color scheme is the same to panel (c)

3.3. BCCP is disordered in the APO structure

Given the behavior of BCCP in the previous conditions, we also examined the behavior of BCCP in the KlUA APO structure. We repeated the sample preparation, data acquisition and processing procedure (Figure S3, for details see Table S2) as described before. Different from our prior specimens, the refined structure showed no additional density obviously close to either the BC or CT domain for accommodating BCCP (Figure 3c). It indicates that in the APO state, without any attraction from BC or CT domain, BCCP is much more flexible, hence the related electric potential density may be averaged out during reconstruction. We performed 3D classification and masked refinement focusing on the UC portion for the best class. However, this could not separate any cluster representing a conformation like the ligand‐bound states (Figure 3c). Additionally, in our previously determined crystal structure,13 the UC portions of individual monomers have a slight 10° rotation relative to each other, which becomes more noticeable in the EM structure with a rotation up to 45° (Figure 3c). This flexibility could be attributed to not only a 15‐residue linker between the AH and UC domains, but also the relatively weak interaction between N‐terminus of AH and CT.13 Further, when superimposing the individual UC domains of a dimer among the three states, the protruding density in ADP‐ and ADP–urea‐bound states for accommodating BCCP domain is more obvious than that of the APO state (Figure 3c). Difference map between each pair of states was generated for the two UC portions (Figure 3d), which clearly shows the extra density. Combining the lack of additional density holding BCCP domain in the APO state, and the structural characteristic of the protruding density in the ADP‐ and ADP–urea‐bound states, we conclude that the BCCP of KlUA may swing between the BC and CT domain within a monomer in a spontaneous mode (Figure 4).

Figure 4.

Biotin carboxyl carrier protein (BCCP)‐swinging model based on the resolved full‐length Kluyveromyces lactis UA (KlUA) dimer in two different states. The center allophanate hydrolyse dimer is colored in blue, the urea carboxylase (UC) portions on each side are colored in orange and gold, and the BCCP domain is colored in dark red. For UC in monomer 1 (UC 1), the BCCP domain anchors close to the carboxyl transferase domain, while in UC 2, BCCP locates close to biotin carboxylase domain in two slightly different positions (in light and dark red) as represented in the KlUA ADP–urea‐ and ADPbound states

4. DISCUSSION

Different from KlUA, certain sequences of other biotin‐dependent carboxylases also participate in the catalytic activities, such as the BC–CT interaction (BT) domain in ScACC,26 HsPCC,24 and PaMCC,25 the central region (CR) in ScACC as well as the central allosteric domain (CAD) in RePC, which make the reaction more complicated. However, the BCCP domain in the highly conserved domain organization plays a central role in the catalysis activity (Figure 1). The distance of active sites between BC and CT domains exceeds the maximum swing of the lysine‐conjugated biotin, thus it has been proposed that the entire BCCP must translocate during catalysis. Apart from KIUC whose catalytic activity is independent of oligomerization, the other enzymes function as dimers or higher‐order oligomers. Based on the oligomers, the BCCP can translocate between the BC domain of one monomer and the CT domain of its neighbor,23, 26 or between the interface of two β‐subunits,24, 25 making the determination of the full‐length crystal structure more difficult (comparison shown in Table S1). The BCCP domain in previously determined structures is located either in the active site of the CT domain or somewhere between the BC and CT domains. Our EM data reported here provide evidence for the BCCP‐swinging model. However, a structure of soluble HsPCC at intermediate resolution provided by another Cryo‐EM study, which shows that BCCP can be positioned differently from corresponding crystal structure.24 KlUA is a single‐chain multidomain enzyme whose catalytic activity does not depend on its oligomerization. Based on the crystal structure in its homodimer form, the swinging of BCCP could only occur within a monomer. All these characteristic makes it an ideal example for the single‐particle analysis.36, 37

In this work, we reconstructed the structure of the full‐length KlUA in three functional states by using negatively stained single‐particle electron microscopy, which provided some unexpected results after 3D classification and masked refinement on the UC portion. A subset of dimeric particles exhibit an asymmetric conformation with the BCCP domain anchoring close to either the BC or CT domain in the ADP‐bound state with or without urea. Meanwhile the BCCP domain behaves more flexibly without a fixed location in the APO state causing the density to be averaged out (Figure 3c). The binding with ligand alone or together with substrate urea contributes to stabilize the location of the BCCP domain, but is not enough to hold it in a unique position. Consequently, this situation can only be captured by single‐particle analysis.38 Besides, the interface between CT and BCCP only buries 680 Ų of surface area.14 The relative weak interaction should also be a factor that helps BCCP domain to leave the active site of CT domain and move toward to the BC active site easily. Based on the structural analysis, we confirm the BCCP‐swinging model between the two successive steps of KlUA catalysis, which may also be extended to the other biotin‐dependent carboxylases.

AUTHOR CONTRIBUTIONS

Y.L. and B.Y. performed the negatively stained sample preparation, data collection, image processing and analysis with the help of L.Z. L.Z. prepared the figures and wrote the manuscript. L.P. and B.C. carried out a further purification of the protein before specimen preparation for electron microscopy. J.Z. constructed the plasmid and performed the expression and purification of KlUA. Y.H.H. participated in the sample preparation and data collection. X.X.Z. and Y.R.Z. manually picked the particles. S.X. provided the purified protein. Y.W. conducted the EM experiment and reviewed the manuscript.

CONFLICT OF INTERESTS

The authors declare no competing financial interests.

Supporting information

Appendix S1: Supporting Information

Figure S1 The data processing procedure for all data sets. The workflow chart takes ADP–urea‐bound sample as an example. The urea carboxylase mask was generated by subtract allophanate hydrolyse mask from the overall mask. The mask files were obtained by RELION.

Figure S2 Reconstruction of the full‐length Kluyveromyces lactis UA in ADP‐bound state. (a) A typical micrograph with 100 nm scale bar and the representative 2D class averages. (b) Gold‐standard fourier shell correlation (FSC) curves show a final resolution of 27 Å (unmasked), 25.6 Å (masked), 22.8 Å (allophanate hydrolyse dimer masked with C2 symmetry applied) and 24 Å (urea carboxylase portion masked).

Figure S3 Reconstruction of the full‐length Kluyveromyces lactis UA in an inactive and unbound state (APO) state. (a) A typical micrograph with a scale bar of 100 nm showing monodisperse particles and representative class averages after 2D classification. (b) Gold‐standard fourier shell correlation (FSC) curves show a final resolution of 33.5 Å (unmasked), 31 Å (masked), and 29 Å (urea carboxylase portion masked).

Liu Y, Yuan B, Peng L, et al. Single‐particle analysis of urea amidolyase reveals its molecular mechanism. Protein Science. 2020;29:1242–1249. 10.1002/pro.3847