Filamentation and inhibition of prokaryotic CTP synthase with ligands

Structure of ecCTPS bound with CTP, NADH, and DON

Previous studies have revealed that CTPS possesses the intriguing ability to spontaneously assemble into filaments in vitro under different conditions ¹⁹ . Structures of CTPS filament have been determined in various eukaryotic organisms, including humans, Drosophila, and budding yeast. Previously, researchers observed and characterized CTPS filaments from prokaryotes in vitro and in vivo.

To establish a more robust structural foundation for the study of prokaryotic CTPS filaments, we employed a recombinant expression approach to purify and isolate the CTPS protein from Escherichia coli (Figure S1). Utilizing cryo‐EM and single particle analysis, we resolved the structure of purified ecCTPS bounded with CTP, NADH, and DON, achieving a resolution of 2.9 Å (Figures 1A and S2–4). During the sample preparation process, under conditions with product CTP, we observed a distinct filamentous arrangement of ecCTPS in the cryo samples (Figure S3). The results of the two‐dimensional (2D) classification provided a clear visualization of the ecCTPS filament, wherein it assembles into filaments with CTPS tetramers as the helical unit (Figure S3).

Consistent with the findings of Barry et al., the interface of the ecCTPS filament comprises two major interaction regions (Figure 1B). One of these regions is located in the linker domain of CTPS, where chains A and A′ engage in a symmetric complementary way (Figure 1C). Notably, residue F281 appears to form Van der Waals interactions with F281′. These interactions likely contribute to the stability of the ecCTPS filament. The second interface is situated between helices 330–335 within the GAT domain (Figure 1D).

Evolution of CTPS filament assembly

In comparison to eukaryotic CTPS, the assembly interactions of ecCTPS filament exhibit significant distinctions. The assembly interface of solved eukaryotic CTPS filaments is primarily centered on helix 12, a sequence that is absent in the E. coli counterpart (Figure S5). To systematically analyze this phenomenon from an evolutionary perspective, we selected CTPS sequences from 112 species to construct a phylogenetic tree. This data set included 50 prokaryotes, 23 eukaryotes, and 39 archaea. The presence or absence of helix 12 was used as a classification criterion (Figures 2A and S6).

The evolutionary analysis revealed that among the 112 selected species, helix 12 is present in nearly all eukaryotes (>95%) and absent in all prokaryotes and the majority of archaea (92%) (Figure 2B). Based on the presence or absence of helix 12 for sequence classification, we observed that the majority of organisms possessing helix 12 were eukaryotes, with only a small fraction belonging to archaea. Conversely, among the organisms lacking helix 12, prokaryotes and archaea constituted the predominant groups. Notably, within the analyzed data set, only one eukaryotic organism was found to be devoid of helix 12 (Figure 2C).

The insertion of helix 12 into the GAT domain allows ecCTPS to form filaments with substrates UTP and ATP. Based on both previous research and our experimental findings, it has been established that the wild‐type ecCTPS is capable of forming filaments in the presence of product CTP. However, filament formation was not observed in the presence of substrates UTP and ATP ²⁰ .

To elucidate the necessity and role of helix 12 in CTPS filament formation, we engineered, expressed, and purified ecCTPS with an inserted hCTPS helix 12 (referred to as ecCTPS^helix12+) (Figure S7).

We employed AlphaFold2 to predict the structure of ecCTPS^helix12+. Among the five structures predicted, rank 2 exhibited the best performance across interchain predicted alignment error and predicted local distance difference test, leading us to select this model for subsequent analysis (Figure S8). Structural alignment revealed that ecCTPS^helix12+ possesses a markedly different secondary structure compared to wild‐type ecCTPS. The insertion of the human CTPS sequence resulted in the extension of the originally shorter loop 337–341, forming a complete helix that resembles α‐helix 12, which is responsible for filament assembly in eukaryotic CTPS (Figure 3A). When we aligned the predicted model with hCTPS1 filament based on the filament assembly interface, we observed a high degree of similarity, with a root‐mean‐squared deviation (RMSD) of 0.479 Å. H346 and W349 exhibited patterns consistent with H355 and W358 of hCTPS1, which play vital roles in the filament assembly (Figure 3B). This implies that ecCTPS^helix12+ possesses a structural basis for assembling into filaments that resemble those in hCTPS1.

To further investigate the nature of ecCTPS^helix12+, we prepared samples of it in both substrate and product states. Upon analyzing cryo samples of ecCTPS^helix12+, we observed that this engineered variant not only formed filaments in the presence of product CTP but also exhibited filament formation in the presence of substrates UTP and ATP (Figure 3C). Intriguingly, the arrangement of the formed filaments exhibited a connected X‐shaped pattern, resembling the previously observed helix 12–12–mediated helical interface in eukaryotic CTPS filaments ²⁰ . This result demonstrates that the insertion of hCTPS helix 12 is sufficient to alter the assembly mode of ecCTPS filaments, suggesting the pivotal role of helix 12 in determining the CTPS filament assembly pattern.

Our study established a connection between protein sequences and the assembly of CTPS filaments while also providing an evolutionary perspective on the formation of CTPS filaments.

DON binding and ammonia tunnel formation of ecCTPS

CTPS is composed of GAT and AL domains, and during the reaction, the intermediate product ammonia is transferred between the two domains through an internal ammonia tunnel. DON is an irreversible inhibitor of CTPS that can covalently bind to CTPS and inhibit its activity.

In 1971, Levitzki et al. demonstrated that even after DON binding, prokaryotic CTPS can still use NH₄ ⁺ as a substrate for CTP synthesis ³² . In our obtained structure, DON forms a covalent bond with C379, a residue that is critical for the catalytic function of ecCTPS. Surrounding amino acids, such as F353, contribute to the binding and stabilization (Figure 4A). Upon the conjunction of DON, the GAT domain adopts a closed conformation, preventing external NH₄ ⁺ from accessing the ammonia tunnel through the catalytic pocket of GAT for further utilization by the AL domain (Figure 4B).

Through computational analysis, we observed the presence of a solvent‐accessible channel on the side of F353 close to the AL domain, connecting an internal cavity within the GAT domain (Figure 4B,C). The minimum diameter of this channel exceeded 1.7 Å, allowing the passage of NH₃ or NH₄ ⁺ (Figure 4D). This cavity is also positioned above the ammonia tunnel that connects to the AL domain, where we identified a minimum diameter of the ammonia tunnel above 1.3 Å, allowing the transfer of ammonia. The existence of this entrance to the ammonia tunnel explains the feasibility of ecCTPS utilizing NH₄ ⁺ as a substrate when DON is bound to the GAT pocket.

Diverse binding modes of CTP in CTPS

The product CTP serves as a natural inhibitor of various eukaryotic CTPSs and can bind to the substrate binding pockets of UTP or ATP in two distinct inhibitory modes: competitive and noncompetitive with substrate UTP ¹⁷ , ¹⁸ .

In our structure of ecCTPS, CTP binding was observed in the UTP binding pocket (Figure 5A), consistent with previously resolved ecCTPS–CTP complexes ¹⁸ , ²¹ (Figure 5B). CTP shares the triphosphate binding site with UTP, and interactions between coenzyme magnesium ions and the triphosphate portion further stabilize CTP binding. The cytidine portion and ribose of CTP are enveloped by three adjacent protomers, which are stabilized via electronic interactions (Figure 5A,B).

Previous research has revealed that product CTP can bind to CTPS near the substrate ATP binding pocket. This binding mode, initially discovered in dmCTPS in 2021, emerged years after the classical CTP binding site was identified ¹⁸ . As a result, this unique CTP binding mode was coined the “nonclassical binding of CTP”. Subsequently, the existence of the nonclassical CTP binding mode was also confirmed in both hCTPS1 and hCTPS2 ¹⁷ .

In this nonclassical binding mode, the triphosphate portion of CTP shares the same pocket as ATP, while the cytidine portion of CTP adopts a distinct binding mode alongside the adenine portion of ATP (Figure 5C). However, we did not observe the density of CTP in proximity to the ATP binding pocket in our map. Aligning the structure with those of dmCTPS and hCTPS containing nonclassical CTP binding, we identified a clash between the CTP portion of the eukaryotic CTPS model and the solvent‐accessible surface of the ecCTPS model due to hindrance by amino acids H75 and I20 (Figure 5D). This clash impedes CTP from binding to ecCTPS in a noncompetitive inhibitory mode similar to that in eukaryotic CTPS within the ATP pocket.

NADH binds to ecCTPS in the ATP binding pocket

In earlier research, Habrian et al. discovered that NADH can synergistically inhibit CTPS along with CTP ¹³ . However, a molecular basis for this regulation was not provided. We observed NADH binding in the ATP binding pocket in our model (Figure 5D). At the 1σ counter level, the adenine portion of NADH displayed clear and distinct density, while the remaining portion lacked stable density (Figure 6A).

Within the ATP binding pocket, the adenine portion of NADH exhibits a mode of interaction similar to that of ADP with the ecCTPS. The N1 atom forms a set of hydrogen bonds with the nitrogen atom on the V241 backbone, with a distance of 3.0 Å. Additionally, the N6 atom forms another set of stronger hydrogen bonds with the oxygen atom on the K239 backbone, with a distance of 2.5 Å. Furthermore, adenine is within 3.8 Å of the plane of the R211 side chain, forming a strong pi–pi interaction, which, along with surrounding hydrophobic amino acids, stabilizes the binding of adenine (Figure 6B). According to our model, the adenine portion of NADH exhibits a stronger interaction with the ATP binding pocket of ecCTPS compared to the adenine portion of ADP (Figure 6B). This could be attributed to the failure of NADH's remaining functional group to achieve stable binding, thereby providing the adenine portion of NADH with ample space to optimize its binding. This also suggests that NADH's interaction with ecCTPS may depend on the adenine portion, accounting for the relatively weak inhibition of ecCTPS reported by Habrian et al. ¹³

Synergistic inhibition of ecCTPS by CTP with NADH or adenine

To validate the binding effect of NADH and its regulatory role on ecCTPS, we examined the inhibitory effects of both NADH and adenine in combination with CTP on CTPS activity (Figure 6C). In the control setup, CTP alone significantly inhibited CTPS activity. The presence of NADH further suppressed the reaction, while the combination of adenine and CTP exhibited a similar degree of inhibition to that of the combination of CTP and NADH. Enzyme activity assays confirmed the synergistic inhibition of CTPS activity by NADH and CTP and indicated that adenine also possesses this capability.

Our structure provides a structural foundation for the synergistic inhibition of ecCTPS by NADH and CTP. Furthermore, through enzyme activity assays, we propose that NADH's inhibition of CTPS is likely dependent on the affinity between the adenine portion and the ATP pocket of ecCTPS.

ecCTPS plasmid construction and protein purification

The full‐length ecCTPS gene was cloned into a pET28a vector with a C‐terminal 6× His tag and transformed into E. coli Transetta (DE3) cells for expression. Transformed cells were at first cultured in 6–8 ml LB medium at 37°C with 220 rpm for about 12 h and then transferred to 1 l LB medium. When its OD₆₀₀ reached a range of 0.6–0.8, 0.1 mM isopropyl β‐d‐1‐thiogalactopyranoside was added for induction. Cells were pelleted by centrifugation at 4500 rpm for 10 min followed by resuspension in cold lysis buffer (50 mM Tris‐HCl pH 7.5, 500 mM NaCl, 10% glycerol, 20 mM imidazole, 1 mM PMSF, 5 mM β‐mercaptoethanol, 5 mM benzamidine, 2 μg/ml leupeptin, and 2 μg/ml pepstatin). The mixed suspension of bacteria and lysis buffer was disrupted by high pressure at 800 bar and centrifuged at 16,000g for 30 min at 4°C to collect supernatant, which was then incubated with equilibrated Ni‐Agarose (Qiagen) for 1.5 h. Next, Ni‐Agarose was washed by washing buffer (50 mM Tris‐HCl pH 7.5, 500 mM NaCl, 10% glycerol, 40 mM imidazole, 5 mM β‐mercaptoethanol), and proteins were eluted with elution buffer (30 mM Tris‐HCl pH 7.5, 100 mM NaCl, 240 mM imidazole). SuperoseTM 6 Increase 10/30 GL column and AKTA Pure (Cytiva) were used for further purification of peak fractions. Finally, CTPS was eluted with buffer containing 100 mM NaCl and 30 mM Tris‐HCl pH 7.5 and stored at −80°C for further use.

CTPS activity assay

Purified CTPS protein was incubated in the reaction buffer containing 100 mM NaCl, 30 mM Tris‐HCl pH 7.5, 1 mM ATP, 1 mM UTP, 0.2 mM GTP, and 10 mM MgCl₂ for 15 min at 37°C. To initiate the reaction, 10 mM prewarmed glutamine was added to the mixture. For inhibitor assay, 0.4 mM CTP and 2 mM NADH or adenine were added to the mixture before incubation. Absorption of a wavelength of 291 nm of each reaction mixture was measured using SpectraMax i3 as the indication for CTP production at individual time points ³³ , and measurements were made at 10 or 15 s intervals until the absorption value tended to be constant.

Phylogenetic tree construction

A total of 112 sequences of CTPS of different organisms were acquired from the Swiss‐Prot database ³⁴ . A phylogenetic tree was built by Unweighted Pair Group Method (UPGMA) ³⁵ using MEGA11 ³⁶ . Poisson regression method was chosen for the modeling and pairwise deletion was applied for gaps treatment. MUSCLE ³⁷ program was used for alignment of the sequence of Homo sapiens CTPS1 (UniProtKB:P17812) and CTPS2 (UniProtKB:Q9NRF8), Schizosaccharomyces pombe CTPS (UniProtKB:O42644), Drosophila melanogaster CTPS (UniProtKB:Q9VUL1), E. coli CTPS (UniProtKB:P0A7E5), Mycobacterium tuberculosis CTPS (UniProtKB:P9WHK7), Thermus thermophilus CTPS (UniProtKB:Q5SIA8), and Sulfolobus solfataricus CTPS (UniProtKB:Q980S6). The result of sequence alignment was visualized by ESPript 3 ³⁸ which rendered sequence similarities and structure information taking the crystal structure of the tetrameric form of human CTPS1 (PDB EntryID:7MGZ) as a reference.

Design and structure prediction of engineered protein

ecCTPS^helix12+ was designed by inserting helix 12 from hCTPS1 into ecCTPS after E399. The structure of ecCTPS^helix12+ was predicted using ColabFold ³⁹ , with the protein sequence “MTTNYIFVTGGVVSSLGKGIAAASLAAILEARGLNVTIMKLDPYINVDPGTMSPIQHGEVFVTEDGAETDLDLGHYERFIRTKMSRRNNFTTGRIYSDVLRKERRGDYLGATVQVIPHITNAIKERVLEGGEGHDVVLVEIGGTVGDIESLPFLEAIRQMAVEIGREHTLFMHLTLVPYMAASGEVKTKPTQHSVKELLSIGIQPDILICRSDRAVPANERAKIALFCNVPEKAVISLKDVDSIYKIPGLLKSQGLDDYICKRFSLNCPEANLSEWEQVIFEEANPVSEVTIGMVGKYIELPDAYKSVIEALKHGGLKNRVSVNIKLIDSQDVETRGVEEEPVRYHEAWQILKGLDAILVPGGFGYRGVEGMITTARFARENNIPYLGICLGMQVALIDYARHVANMENANSTEFVPDCKYPVVALITEWRDENGNVEVRSEKSDLGGTMRLGAQQCQLVDDSLVRQLYNAPTIVERHRHRYEVNNMLLKQIEDAGLRVAGRSGDDQLVEIIEVPNHPWFVACQFHPEFTSTPRDGHPLFAGFVKAASEFQKRQAKHHHHHH” serving as the input. Among the five prediction results, the model of rank2 exhibited the best performance and was selected for the subsequent analysis.

Cryo‐EM grid preparation and data collection

For preparing the ecCTPS filament sample, CTPS was incubated with 0.58 mM DON, 1 mM CTP, 0.5 mM NADH, and 10 mM MgCl₂ at a final concentration of 7 μM for 30 min at 0°C. Samples were prepared with 300 holey golden film (M01Au300‐R1.2/1.3) and FEI Vitrobot (4°C temperature, multiple rounds of sample application and blotting before vitrification, 3.5 s blotting time, −1 blot force). Images were taken with a Gatan K3 summit camera on a FEI Titan Krios electron microscope operated at 300 kV. The magnification was 22,500× in superresolution mode with the defocus range of −1.2 to −1.8 μm and a pixel size of 1.06 Å. The total dose was 50e−/Ų subdivided into 40 frames at 2.8s exposure using SerialEM.

Image processing

We used MotionCor2 for alignment of the raw movie frames, and CTFFIND4 for contrast transfer function (CTF) estimation. A total of 2760 images were selected for subsequent analysis. By employing template‐free particle picking in Relion, we obtained coordinates for 3,401,539 particles. The single particle analysis strategy was carried out during the remaining processes. Particles were downsampled to a pixel size of 2.12 Å for the 2D classification. After this step, we retained 1,934,043 particles that exhibited the most consistent and well‐defined features for three‐dimensional (3D) classification. Continuing with the 3D classification, we applied C1 symmetry. This step helped to further refine the particle selection, and we selected 1,300,114 particles for D2 symmetry 3D classification. Subsequently, we re‐extracted particles to bin1 and performed 3D refinement. A soft mask focusing on the center CTPS tetramer is applied in all remaining refinement steps. CTF refinement and Bayesian polishing were applied to improve the map resolution. After these procedures, we obtained a map with a resolution of 2.9 Å. Subsequent quality enhancement of the map was performed using ResolveEM within Phenix, resulting in a final resolution of 2.8 Å.

Model building and refinement

We employed Model PDB ID 5U3C as the initial model. Initially, a rigid body fitting was conducted using Phenix ⁴⁰ , followed by manual coordinate adjustments using Coot ⁴¹ . Subsequently, real‐space refinement was carried out once again using Phenix.

Statistical analysis

Results of the CTPS activity assay were analyzed using GraphPad Prism 8 ⁴² and were shown as means ± SD of three or more independent experiments. The MUSCLE ³⁷ program was used for sequence alignment.