Crystallization and structure determination of the Mtb EF-Tu•EF-Ts complex
To disclose the molecular basis of the EF-Tu reactivating mechanism in Mtb, we investigated the crystal structure of the full-length EF-Tu and EF-Ts complex. We expressed and purified EF-Tu and EF-Ts in Escherichia coli cells. EF-Tu and EF-Ts formed a 1:1 complex in solution, evidenced by the undeniable fact that the complex got here out on the elution peak of 76 mL on a Superdex200 16/600 GL column, which corresponds to a molecular mass of 71 kDa (Fig. 1a, b). Also, the EF-Tu or EF-Ts was eluted out on the peaks of 84 mL or 86 mL on the gel-filtration column, corresponding to a mass of 43 kDa and 30 kDa, respectively (Fig. 1a, b). The Mtb EF-Tu•EF-Ts crystals were obtained after several rounds of optimization, and the complex structure was solved at a resolution of two.8 Å by molecular alternative method using the EF-Tu•GDP (PDB: 7VOK) and the AlphaFold-predicted EF-Ts (https://alphafold.ebi.ac.uk/entry/P9WNM1) because the searching models.
Fig. 1: Crystal structure of Mtb EF-Tu•EF-Ts complex.
a The dimensions-exclusion chromatography (SEC) profiles of Mtb EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex on a Superdex200 16/600 column. b The SDS-PAGE results of EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex, corresponding with (a). M protein marker, L protein loading sample. c Schematic domains of Mtb EF-Tu and EF-Ts. EF-Tu consisted of three domains, including Domain I (gray), Domain II (blue), and Domain III (purple). EF-Ts consisted of three domains, including the N-terminal domain (wheat), the C-terminal domain (green), and the core domain in the center (yellow). d A ribbon representation of Mtb EF-Tu•EF-Ts complex structure, coloured with that in (c). EF-Tu and EF-Ts formed a 1:1 symmetric complex. The binding interface was labeled with a black rectangle for key residues depicted with stick models, and the cartoon is shown with a 50% transparency to spotlight the important thing residues. e A ribbon representation of the detailed binding interface between EF-Tu and EF-Ts. The involved key residues were labeled and shown as sticks. f The electrostatic potential surfaces of Mtb EF-Tu•EF-Ts complex (left), EF-Tu (middle), and EF-Ts (right). Red denoted negative potential, and blue denoted positive potential.
The structure of Mtb EF-Tu•EF-Ts complex
The Mtb EF-Tu accommodates three domains, named Domain I (residues 1–200), II (residues 213–296), and III (residues 306–396) (Fig. 1c and Supplementary Fig. 1a). Domain I, also called the GTP-binding domain or Ras-like domain, hydrolyzes GTP to GDP in an Mg2+-dependent manner. Domains II and III are oligonucleotide-binding domains, which bind to each charged tRNA and EF-Ts in T. thermophilus28. The Mtb EF-Tu structure showed three domains forming a flattened triangular shape with a hole in the center, arranged like other homolog structures (Fig. 1d and Supplementary Fig. 1a). Furthermore, Domains I and III were closer than Domain I and II, making face-to-face contact and resulting the side-chain interactions. Also, the flattened triangular shape permits a high degree of inter-domain flexibility, which advantages EF-Tu to bind with sorts of substrates within the polypeptide synthesis29,30. The EF-Ts comprises 13 α-helices and 6 β-strands, containing three domains, the N-terminal domain, the core domain, and the C-terminal domain (Fig. 1c and Supplementary Fig. 1b). The core domain comprises two subdomains, denoted as subdomain N (residues 61–65, 68–73, and 133–139) and subdomain C (residues 144–149, 157–166, and 259–267), which consists of 6 β-strands surrounding by the α-helices (Supplementary Fig. 1b).
The Mtb EF-Tu•EF-Ts complex accommodates one EF-Tu and one EF-Ts per asymmetric unit (Fig. 1d), different from that in E. coli and T. thermophilus that consists of two EF-Tu and two EF-Ts molecules27,28,31. In E. coli, the amino-terminal region of EF-Ts interacts with the nucleotide-binding Domain I of EF-Tu, and the opposite half binds to Domain III. As well as, residues K51, G54, D80, F81, I125, and G126 of E. coli EF-Ts are vital for the interaction with EF-Tu; nonetheless, the structure and sequence alignment shows that only D80 (corresponding to D77 in Mtb EF-Ts) is the conserved energetic site27. In T. thermophilus, EF-Ts is a dimer, wherein each subunits contribute to the bipartite interface. Also, the conserved TDFV sequence of EF-Ts contacts with EF-Tu Domain I in E. coli, T. thermophilus, and B.mitochondrial32. Much like homologs in other species, the Mtb EF-Tu certain with the N-terminal domain and the core domain of EF-Ts through Domains I and III (Fig. 1d, e). We selected the pair of amino acids with a relative distance inside 3 Å as interface residues. Twenty residues were involved within the interface with hydrogen bonds or salt bridges (Fig. 1d, e). Also, the electrostatic potential surfaces evaluation showed that the EF-Tu•EF-Ts complex was stuffed with negative potential surfaces (Fig. 1f). Notably, the binding interface of EF-Tu was stuffed with negative potential, while that in EF-Ts was positive potential, indicating the charge-charge interaction promoting the complex formation, which is critical for protein stacking33,34.
Mtb EF-Ts exhibits a high binding affinity to EF-Tu
The interface-involved residues of EF-Tu were situated in Domain I and III, while the related residues of EF-Ts were mainly situated within the N-terminal domain and the core domain (Fig. 1d). Briefly, residue K24 on the α2 helix of EF-Ts formed salt bridges with residues D144 and D145 of EF-Tu, and the K24 side chain formed a hydrogen bond with D145. Residue R13 on the α1 helix of EF-Ts formed hydrogen bonds with P114 and E155 of EF-Tu, respectively. Furthermore, residue D77 of EF-Ts formed a hydrogen bond with residue H87, and a salt bridge with A88 of EF-Tu. Residue N82 of EF-Ts formed a hydrogen bond with the side chain of EF-Tu E120. As well as, the N atom on the side chain of EF-Ts-H149 formed a salt bridge with the O atom of the D357 side chain of EF-Tu. Also, the side chain of EF-Ts R151 formed a hydrogen bond with N358 of EF-Tu, and D154 formed a salt bridge with EF-Tu Q127 (Fig. 2a). Next, we investigated the interaction between EF-Tu and EF-Ts using the isothermal titration calorimetry (ITC) method. The result showed that EF-Tu certain with EF-Ts at a 1:1 ratio with a Kd value around 1.36 µM (Fig. 2b), different from the Kd values for EF-Tu and EF-Ts from E. coli and Bovine mitochondria (2 nM and 5.5 nM, respectively)35,36. Reports showed that the equilibrium dissociation constants would change amongst different species despite the high sequence similarity37. Moreover, Mtb EF-Tu shared 74% and 71% similarities in amino acid sequence with that in E. coli and T. thermophilus, and the interfaced-related residues were highly conserved (Supplementary Fig. 2). Nonetheless, EF-Ts’ amino acid sequence similarity between Mtb and E. coli, or T. thermophilus was 39% or less, except that residues R13, K24, and D77 in Mtb were conserved (Supplementary Fig. 3).
Fig. 2: Mtb EF-Tu exhibits a high binding affinity with EF-Ts.
a Ribbon representations of crucial residues within the interface between EF-Tu (cyan) and EF-Ts (light pink). The residues were labeled and shown as sticks. The hydrogen bonds amongst interacted residues are shown in black dashed lines. b The isothermal titration calorimetry (ITC) result showed that the Kd value between EF-Tu and EF-Ts was 1.36 µM. The binding molar ratio was around 1:1. c The SEC results of the EF-Tu•EF-Ts complex and different mutants. The mutants Tu/Ts-R13, Tu/Ts-N82, and Tu/Ts-H149 dramatically modified their solution status. Tu/Ts is abbreviated for EF-Tu/EF-Ts. d The SDS-PAGE results of the EF-Tu•EF-Ts complex and different mutants, corresponding to (c). M protein marker, L protein loading sample.
To discover the critical residues for the EF-Tu/EF-Ts complex’s interaction, we mutated seven residues of EF-Ts into alanine, including R13, L21, K24, D77, N82, H149, D154, and EF-Tu N358. When EF-Tu was co-purified with the wild-type or mutant of EF-Ts on the scale exclusion chromatography, it clearly showed that mutants R13A, N82A, and H149A dramatically disrupted the interaction profiles with EF-Tu, indicating the three residues are indispensable for the EF-Tu/EF-Ts formation (Fig. 2c, d).
EF-Ts exhibits a various binding conformation amongst different species
EF-Ts functions as a guanine nucleotide exchange factor and catalyzes the response of EF-Tu from the inactive form (GDP-bound) to the energetic form (GTP-bound)38. As EF-Ts shared low similarities amongst different species (Supplementary Fig. 3), and preferred to bind with and reactivate EF-Tu at different ratios between Mtb (1: 1, EF-Ts: EF-Tu) and T. thermophilus, (2:1, EF-Ts: EF-Tu)28,31, we compared the EF-Tu•EF-Ts complex structures using the ChimeraX and PymoL software. Superimposing the Mtb EF-Tu•EF-Ts complex with that of E. coli (PDB ID:1EFU) or T. thermophilus (PDB ID: 1AIP) revealed similar features of overall folds, which showed high similarities among the many Cα atoms with RMSD values of 1.214 Å or 1.309 Å, respectively (Fig. 3a). Briefly, the structural differences existed mainly in EF-Ts, while EF-Tu had similar architecture (Fig. 3a, b). Different from the Mtb EF-Ts, the EF-Ts of E. coli was composed of 13 α-helices and 6 β-strands, which assembled in 4 critical domains, the N-terminal domain (residues 1–54), the core domain (residues 55–179), the dimerization domain (residues 180–228), and the C-terminal domain (residues 264–282)27. The C-terminal domain was stretched out to interact with Domain I in EF-Tu of E. coli, whereas it was absent in Mtb EF-Ts (Fig. 3c, d and Supplementary Fig. 3)27,28, implying a definite binding conformation in numerous bacteria.
Fig. 3: EF-Ts exhibits a various binding conformation amongst different species.
a Superimposition of EF-Tu•EF-Ts complex structures amongst Mtb (marine, PDB: 7VMX), E. coli (light pink, PDB: 1EFU), and T. thermophilus (green, PDB: 1AIP). The RMSD values between Mtb and E. coli or T. thermophilus were 1.214 Å or 1.309 Å, respectively. b Superimposition of EF-Tu structures amongst Mtb, E. coli, and T. thermophilus. c Superimposition of EF-Ts structures amongst Mtb, E. coli, and T. thermophilus. The numbers denoted significant differences among the many three structures. d The detailed difference of EF-Ts structures amongst three species, in keeping with (c). ‘1’, the additional helix of E. coli EF-Ts for binding with EF-Tu Domain I. ‘2’ and ‘4’, the helices within the T. thermophilus EF-Ts, absent in Mtb and E. coli, which weren’t involved in binding with EF-Tu. ‘3’, the α10 and α11 helices that shifted about 4.7 Å or 6.0 Å inside three species, respectively.
The second significant difference was that the T. thermophilus EF-Ts had one extra helix between the β1 and β2 strands. Because the EF-Tu•EF-Ts complex is an asymmetrical heterotetramer in T. thermophilus, of which one EF-Tu interacts with two subunits of EF-Ts, forming a bipartite interface; the additional helix could explain why one EF-Tu molecule needed to be reactivated by two EF-Ts molecules in T. thermophilus (Fig. 3d). Mtb and E. coli EF-Ts is a monomer with a structural repeat that mimics the dimeric interface in T. thermophilus EF-Ts28. Also, the helices, α10 and α11, of Mtb EF-Ts dramatically shifted with distances around 4.7 Å and 6.0 Å to that of E. coli or T. thermophilus (Fig. 3d).
The Mtb EF-Tu•GDP complex structure
To elucidate the molecular basis of the inactive type of Mtb EF-Tu through the elongation cycle, we expressed and purified the full-length EF-Tu protein, and successfully obtained the EF-Tu•GDP complex structure at a resolution of three.4 Å using the E. coli EF-Tu (PDB ID: 1EFC) as a searching model. The ultimate structure accommodates 4 EF-Tu molecules per asymmetric unit (Table 1). Each EF-Tu molecule was certain with a GDP and an Mg2+ ion (Fig. 4a). Domain I of Mtb EF-Tu underwent a structural rearrangement of roughly 90° rotation to the opposite domains (Fig. 4a), just like that in E. coli which occurred around switches 1 and a couple of when GTP replaced GDP. This movement formed a binding site of aminoacyl-tRNA to interact with all three domains of EF-Tu. Domain II was composed of 84 residues from 213 to 296, with 6 antiparallel β-strands forming a β-barrel (Supplementary Fig. 1a), different from that in E. coli EF-Tu. Residues of Domain II took responsibility for discriminating between the charged and uncharged tRNA within the binding pocket. Furthermore, Domain III of Mtb was composed of 91 residues from 306 to 396, with 6 antiparallel β-strands forming a β-barrel. In Mtb, Domains II and III were connected by a shorter peptide, but in E. coli, there was yet one more β-strand between the 2 domains20. Also, Domain II and III formed antiparallel β-barrels to manage the activity of Domain I, increasing the affinity of GDP over GTP (Fig. 4)38.
Table 1 Data collection and refinement statistics.Fig. 4: Critical residues within the GDP-binding pocket.
a A ribbon representation of the EF-Tu•GDP structure, which is coloured as Fig. 1a. GDP and Mg2+ were situated in Domain I. GDP is shown as sticks, and Mg2+ is shown as a green sphere. b Representation of GDP-binding sites within the EF-Tu•GDP complex. The broken black lines represented the hydrogen bonds formed between EF-Tu and GDP. The related residues were labeled and shown as sticks. c The electrostatic potential surface of the EF-Tu•GDP complex. Red indicated the negative potential, and blue indicated the positive potential. GDP is shown as sticks, and Mg2+ is shown as a green sphere. d Superimposition of Mtb EF-Tu between the EF-Tu•GDP (yellow) and EF-Tu•EF-Ts (marine) complexes. The RMSD value for the Cα alignment between the 2 structures was 0.711 Å. e The detailed shift of GDP-bound residues between the EF-Tu•GDP and EF-Tu•EF-Ts complexes. The broken black lines represented the shifting of labeled residues.
The GDP-binding sites are conserved amongst different species
The GDP molecule was certain with Domain I of EF-Tu, which consisted of 5 loops, named G-1 to G-5, just like other nucleotide-binding proteins (Fig. 4b and Supplementary Fig. 2)28,31,39. G-1 was known as the P-loop attributable to the conserved amino acid sequence like GX4GK(ST) from residues G19 to K25, which connected the β1 strand and α1 helix chargeable for phosphate binding. Also, residues K25, T26, and T27, situated on the α1 helix, forming hydrogen bonds with the α- and β-phosphate of GDP. G-2 contained the second β-strand and its preceding loop, which was certain with the γ-phosphate of GTP described in other homologous structures. G-3 contained the conserved sequence of DX4G, residues from 83 to 86 in Mtb EF-Tu, which connected the β5 strand and α3 helix. Especially, residue D83 was certain with the Mg2+ ion through several water molecules as described before. G-4 contained β5 strand and its preceding loop, and had the conserved amino acids sequence of (N/T) (K/Q) XD, which was NKAD in Mtb EF-Tu spanned from residues 138 to 145 that formed a loop to attach β7 strand and α5 helix. Also, residues D141 and A176 formed hydrogen bonds with the guanine ring of GDP. The G-5 region was composed of β8 strand and α6 helix, which spanned from residues P170 to A176, and was removed from the GDP and Mg2+ molecules (Fig. 4b).
The GDP molecule interacted with residues D22, G24, K25, T26, T27, N138, D141, and A176 of EF-Tu, which were conserved in Mtb, E. coli, and T. thermophilus (Fig. 4b and Supplementary Fig. 2). Residues D22 and G24 were situated in the primary loop, which connected the β1 strand and α1 helix. Also, D141 was situated within the loop connecting the β7 strand and α5 helix. Furthermore, analyzing the electrostatic potential surface showed that GDP was certain within the pocket stuffed with positive charges (Fig. 4c).
When superimposing the EF-Tu molecules within the EF-Tu•GDP and EF-Tu•EF-Ts structures, it showed similar features of overall folds with an RMSD value of 0.711 Å (Fig. 4d). A big difference existed in Domains I and III, chargeable for the interaction with EF-Ts. The EF-Tu within the recycling complex (EF-Tu•EF-Ts) had a far more compact shape than that within the inactive form (EF-Tu•GDP) (Fig. 4d). Significantly, the residues involved within the GDP-binding pocket modified positions. D22, G24, K25, T26, T27, A176, and D141 of EF-Tu shifted from 2.8 Å to five.5 Å upon conformational change (Fig. 4e), which could explain how EF-Ts help EF-Tu to dissociate with GDP and reactivate EF-Tu through the elongation cycle of protein translation. The pocket was now not suitable for binding with GDP due to translocate of GDP-binding residues after EF-Ts binding, just like the report in E. coli, wherein the movement of residues K136 and N138 of EF-Tu made them away from the nucleotide-binding site and relaxed the interactions inside the base (Fig. 4e)38.
Moreover, we investigated the answer status of EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex by the small-angle X-ray scattering (SAXS) method (Fig. 5a–i). All of the proteins behaved well in solution. The utmost dimension (Dmax) from distance distribution function p(r) for the EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex were 86 Å, 80 Å, and 106 Å, respectively (Fig. 5b, e, h). When superimposed the crystal structures with the envelopes generated from SAXS data, EF-Ts and EF-Tu•EF-Ts complex showed high similarities, whereas EF-Tu had discrepancies resulting from a highly flexible loop between Domain I and Domain II (Fig. 5c, f, i).
Fig. 5: The SAXS models of the Mtb EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex.
a The experimental scattering curve of the Mtb EF-Tu•EF-Ts complex. b The space distribution function curve of the EF-Tu•EF-Ts complex. c The crystal structure of the EF-Tu•EF-Ts complex was fitted into the ab initio envelope obtained from SAXS. d The experimental scattering curve of Mtb EF-Tu. e The space distribution function curve of Mtb EF-Tu. f The crystal structure of EF-Tu was fitted into the ab initio envelope obtained from SAXS. g The experimental scattering curve of Mtb EF-Ts. h The space distribution function curve of Mtb EF-Ts. i The crystal structure of EF-Ts was fitted into the ab initio envelope obtained from SAXS. j Dynamic light scattering (DLS) evaluation of the Mtb EF-Tu, EF-Ts, and EF-Tu•EF-Ts complex. Particle polydispersity was defined in the next terms: polydispersity (Pd), % polydispersity (% Pd).
The FDA-approved drug Osimertinib binds with Mtb EF-Tu and inhibits Mycobacterium growth
To discover the potential inhibitor of Mtb EF-Tu, we screened the FDA-approved drug library (total of 1971 drugs) with the nano-DSF method40. In comparison with EF-Tu itself, sixteen drugs could dramatically affect the steadiness of the EF-Tu protein (Table 2). Significantly, Osimertinib, a drugs for treating non-small-cell lung carcinomas (NSCLC) with specific mutations41, significantly modified the unfolding transition midpoint from 52.5 to 60.4 °C (Fig. 6a, b). Next, 16 drugs were added to different bacterial strains to ascertain the in vivo function individually. The outcomes showed that Osimertinib didn’t change the expansion of E. coli; nonetheless, it dramatically inhibited the expansion of M. smegmatis and Mtb H37Ra strains at around 10 µM concentration (Fig. 6c–e). In contrast, other drugs didn’t show significant antibacterial activities, including hydroquinone and saractinib (Supplementary Fig. 4a, b). Also, Osimertinib significantly inhibited the expansion of the M. Bovis BCG strain (Fig. 6f). The primary-line anti-MTB drug isoniazid (also called INH) inhibits Mtb growth by disturbing the biosynthesis of mycolic acid. Consistently, isoniazid showed a robust ability to inhibit the expansion of Mtb H37Ra and M. smegmatis, but didn’t change the expansion of E. coli (Supplementary Fig. 4d–f). Thus, Osimertinib had an analogous bacterial inhibition capability with isoniazid in bacterial specificity. To further discover whether Osimertinib could directly bind with Mtb EF-Tu, we performed the microscale thermophoresis (MST) method with the purified protein. The result showed that Osimertinib certain to EF-Tu in vitro with medium intensity (Kd = 207 μM) (Fig. 6g), with the binding of GDP and EF-Tu as a positive control (Supplementary Fig. 4c).
Table 2 Drugs modified the thermal stability of Mtb EF-Tu.Fig. 6: The FDA-approved drug Osimertinib binds with Mtb EF-Tu and inhibits Mycobacterium growth.
a, b Thermal unfolding curves and unfolding transition midpoints of EF-Tu and EF-Tu/Osimertinib were detected by the Nano-DSF method. c–f Osimertinib didn’t affect E. coli growth, but significantly inhibited the expansion of M. smegmatis, Mtb H37Ra, and M. bovis BCG strains. Different concentrations of Osimertinib were added to the bacterial strains, and cell densities were detected at different times (c–e). After treatment with various concentrations of Osimertinib, the BCG strains were diluted 100 or 1000 times, and the numbers of colonies were calculated (f). g The MST result showed that Osimertinib is certain with EF-Tu in vitro. h The in silico modeling structure of EF-Tu and Osimertinib complex. EF-Tu was shown in a cartoon model. Osimertinib was shown in a stick model. The error bars represented the usual deviations (SD). *P < 0.05, **P < 0.01, ns no significance.
We next screened 1000’s of crystallization conditions for solving the Osimertinib-bound EF-Tu complex structure; nonetheless, no qualified diffraction datasets were obtained. Thus, we built the in silico modeling complex structure of EF-Tu and Osimertinib using the AlphaFold2 software. The structure showed that Osimertinib certain with the Domain I and II of EF-Tu, composing of flexible loops and different from the positioning of the GDP-binding pocket (Fig. 6h). Also, the electrostatic potential surface of the osimertinib-binding pocket was stuffed with negative charges (Supplementary Fig. 5a). Moreover, the binding pocket of osimertinib within the human EGFR/osimertinib complex structure (PDB ID: 6LUD) consists of flexible loops (Supplementary Fig. 5b). Next, we analyzed the binding sites of 4 known inhibitors in EF-Tu. The crystal structure of antibiotic enacyloxin/EF-Tu complex (PDB ID: 2BVN) shows that enacyloxin locates on the cleft between Domain I and III of E. coli EF-Tu (Supplementary Fig. 5c), whereas antibiotic GE2270A locates on Domain II and doesn’t interact with Domain I and III of E. coli EF-Tu (PDB ID: 3U6K) (Supplementary Fig. 5d). As well as, kirromycin shares an analogous location with enacyloxin (Supplementary Fig. 5e), while pulvomycin locates in the center hole, which interacts with three domains of T. ther EF-Tu (Supplementary Fig. 5f), indicating that Osimertinib docking site may overlap partially with GE2270A and pulvomycin. Thus, Osimertinib might inhibit Mtb growth by blocking the rearrangement of various domains of EF-Tu.