The gene encoding the cDPGS was successfully cloned in the pLATE51 expression vector in frame with the N-terminal His6x-tag sequence and under the control of the lactose inducible promoter. The His-tagged cDPGS protein was successfully over-expressed in a soluble form in Escherichia coli BL21 (DE3). cDPGS was purified from the cell extracts by Ni²⁺-NTA affinity chromatography with a recovery yield of 20 mg L⁻¹. This was followed by size exclusion chromatography (SEC), that showed that the purified cDPGS elutes in a dimeric form of ~100 kDa (Supplementary Figure S1). As previously reported (Lehmacher et al., 1990; Van Alebeek et al., 1991, 1994) the presence of at least 300 mM KCl is essential for the purification of cDPGS to maintain the enzyme correctly folded and in an active state. Protein purification using 500 mM NaCl instead of KCl lead to poor recovery.

The purified cDPGS was assayed for its thermal stability using differential scanning fluorimetry (DSF) Vivoli et al., 2014). However due to the high thermostability of the enzyme its melting only starts to appear at 95°C. Due to the instrument limitations it was not possible to obtain a precise apparent T_m value for this enzyme. This demonstrates that cDPGS is a highly thermostable enzyme with a melting temperature above 95°C (Supplementary Figure S2).

To confirm the expression of an active correctly folded protein suitable for crystallization studies, the activity of the cDPGS was measured by monitoring the ADP formation using a linked assay with pyruvate kinase (PK) and l-lactate dehydrogenase (LDH) (Scheme 2). The cDPGS was found to be active with a specific activity of 0.039 U/mg (Supplementary Figure S3). The direct production of cDPG was monitored by HPLC–MS to confirm the correct identification of the product (Figure 1 and Supplementary Figure S4).

The cDPGS crystallized readily in many screening conditions and the structure of the apo-cDPGS was solved by the single wavelength anomalous dispersion (SAD) method using data collected from a selenomethionine protein crystal. The space group was determined to be I222 with a single protomer of cDPGS in the asymmetric unit. The structure was refined to 1.7 Å resolution and was found to contain the entire polypeptide chain. This has been refined to R and R_free values of 17.7 and 21.3%, respectively (Table 1).

The overall structure of the cDPGS showed only limited similarity to other known protein structures according to the DALI structural comparison server (Holm and Laakso, 2016; Holm, 2020). The results showed a few structural homologs with a DALI Z score of >7.0. The best scoring homologs were an Aquifex aeolicus tetraacyldisaccharide 4′-kinase (PDB entry 4EHX, Z score = 15.9), a Klebsiella pneumoniae urease accessory protein UreG (PDB entry 5XKT, Z score = 14.0) and the Helicobacter pylori hydrogenase/urease nickel incorporation protein H (PDB entry 4LPS, Z score = 13.2). Structural alignment with these proteins showed a core region that is conserved amongst other NTP-dependent enzymes (Figure 2). However, a portion of the structure, including a 127 amino acid N-terminal domain, shows no structural similarity to other known structures in the PDB. A protein BLAST (Altschul et al., 1997) search of the N-terminal domain sequence against the PDB did not produce any hits. The DALI server did not find any known protein domains with a fold close to that of the N-terminal domain of cDPGS.

The structure of cDPGS can be split into two domains. The smaller N-terminal domain contains a six-stranded β-sheet of mixed type with direction + + + + + − and connectivity -1x,-1x,3x,1x,1 (Richardson, 1981) flanked by 4 α-helices (α1–α4). The β-sheets of the two N-terminal domains in the cDPGS dimer form a large intersubunit 12 strand β-sheet on the molecular dyad. The C-terminal domain of cDPGS roughly aligns with the unusual P-loop kinase tetraacyldisaccharide 4′-kinase (LpxK) from A. aeolicus (Emptage et al., 2012). The cDPGS C-terminal domain and LpxK show the same Rossmann like α/β/α sandwich fold connected by two twisted, antiparallel β-strands. The larger C-terminal domain (residues 136–460) contains a twelve-stranded β-sheet (β7 – β18) surrounded by twelve α-helices. The strand direction is − − − − + + + + + + + − and connectivity is 2x,2,-1,-2x,-2x,-1x,-1x,-3,-1x,2x,1x (Supplementary Figure S5).

The cDPGS structure shows a tight dimer that buries a surface area of 2844.1 Ų, accounting for 13.7% of the total solvent accessible area of the protomer (Figure 3A). The interface is stabilized by 36 hydrogen bonds and 24 salt bridges between the interacting subunits as estimated by PISA (Krissinel and Henrick, 2007). These include an intersubunit 12-strand β-sheet formed by the N-terminal domains. The dimer interface of cDPGS clearly shows a large positively charged patch at the interface of the C-terminal domain that interacts with the negatively charged N-terminal domain of the opposing molecule (Figure 3B). The high thermal stability of cDPGS appears to be due to both hydrophobic interactions and a high number of hydrogen bonds and ion pairs. Similar interactions have previously been observed in other thermostable proteins with temperature optima up to approximately 75°C (Sayer et al., 2012; Ferrandi et al., 2018).

Multiple sequence alignments using the results from the structural comparison server revealed a number of highly conserved residues (Figure 4) that tend to cluster around the nucleotide and ligand binding sites, particularly a P-loop/ Walker A motif GxxGxGK[T/S] (Walker et al., 1982). This loop typically binds the phosphate groups of phosphorylated ribonucleotides and catalyzes phosphoryl transfer (Romero Romero et al., 2018).

To elucidate the substrate binding residues of cDPGS, the structure of the 2,3 DPG, ADP/Mg²⁺ bound enzyme complex was solved by molecular replacement to a resolution of 2.2 Å resolution, resulting in R_work and R_free values of 21.5 and 25.6%, respectively (Table 1).

The 2,3 DPG binding site is located between the C-terminal domain of the one protomer and the N-terminal domain of the other protomer of the dimer (Figure 5A). The presence of the bound ligands is confirmed by the Fo-Fc difference map calculated with ADP Mg²⁺ and 2,3 DPG at zero occupancy. The ADP and Mg²⁺ are clearly resolved and reside in the pocket formed within the C-terminal domain (Figure 5B). The 2,3 DPG is clearly resolved in only three of the four protomers making up the asymmetric unit of the crystal. A phosphate molecule has been modeled in the active site of the 4th protomer. The Walker A/P-loop surrounds the pyrophosphate moiety of the ADP while coordinating with the Mg²⁺ atom. The adenosine moiety is bound in a pocket formed by α5, α14, β13, β14 and β15. The Walker B loop borders the 2,3 DPG binding site without directly interacting with it.

A superposition of the apo- and ADP/Mg²⁺ bound forms of cDPGS reveals a large movement of several α-helices that close around the nucleotide, which highlight the significant conformational changes that accompany ligand binding (Figure 6A). The helices α5, α8, α9 and α14 show a motion of up to 8.5, 7.0, 6.7, and 2.6 Å, respectively. The loops formed by residues 186–192, 396–400 and 420–427 make up the “hinge” regions. A closer look at the active site revealed that the closure is triggered by the binding of the substrate and ADP which brings the Walker A/P-loop close enough to contribute to the stabilization of 2,3 DPG. Residues K145 and R146 move closer to the substrate to form multiple hydrogen bonds with it (Figure 6B). The 2,3DPG is also stabilized by E45 from the opposing subunit (Figures 7A,B). The pyrophosphate moiety of ADP is held in place by an intricate network of hydrogen bonds with backbone amide nitrogen atoms, various side chains, and water molecules (Figure 7C). The magnesium atom is octahedrally coordinated by the side chain oxygens of D208, T150, the O3B oxygen of ADP, O9 oxygen of 2,3DPG and two water molecules. The domain movement causes the interaction of several residues with the ADP. Outside the P-loop, residues E305, P335, T353 and E404 appear to have a significant role in ATP binding since they form hydrogen bonds to the adenosine, ribose hydroxyl groups, or the α-phosphate of ADP.

Based on the apo structure and the ADP/Mg²⁺ complex structures, we propose a classical phosphoryl transfer mechanism for cDPGS as observed in other NTP dependent enzymes (Matte et al., 1998; Guixé and Merino, 2009; Gerlits et al., 2015). The conformational changes that accompany binding of ATP and 2,3 DPG bring the ATP into the proximity of 2,3 DPG in a sterically and energetically strained conformation, forming a ‘near attack complex’ (Bruice, 2002). Positively charged residues (K145, R146, K149, K406) and the magnesium ion will shield the negative charges on the phosphates, allowing the reactants to approach. The domain closure creates a hydrophobic environment allowing the phosphoryl transfer from the ATP γ-phosphate to the 2-phosphate of the 2,3 DPG through an S_N2 reaction, with the transition state stabilized by the magnesium ion, to form the intermediate. A second S_N2 reaction is initiated by attack of the 3-phosphate of the intermediate on the 2-pyrophosphate, releasing an inorganic phosphate. This step is facilitated by the magnesium ion, K145, and R146 providing positive charges to stabilize the transition state. Release of phosphate makes the reaction highly exergonic under standard conditions, providing the energy for the formation of the cDPG phosphoanhydride (Scheme 3).

The genes coding for Mf-cDPGS in its E. coli codon optimized version was kindly donated by the Siebers lab (Essen Germany) and sub-cloned into the pLATE51 vector (ThermoFisher) in frame with a N-terminal 6x His-Tag. pLATE51/cDPGS (Supplementary Figure S7) was introduced into E. coli DH5α (New England Biolabs, C2987H) using standard techniques. The plasmid was purified and sequenced to confirm the presence of the gene. The plasmid pLATE51/cDPGS (Addgene: 201561) was introduced into E. coli BL21 competent cells (Agilent Technologies, 230,132) according to the manufacturer’s instructions.

The transformants obtained were grown in LB supplemented with 100 μg/mL ampicillin (LB_amp100) medium (50 mL) overnight and then inoculated in 0.5 L LB _amp100 at 37°C and 200 rpm. When the OD₆₀₀ reached 0.5–0.6, gene expression was induced by the addition of 0.5 mM IPTG, and the culture was maintained at 25°C for 24 h. Cells were harvested by centrifugation (4,750 g for 30 min), and frozen at −20°C before proceeding with the protein purification.

A selenomethionine derivative (SeMet) of the enzyme was also produced in E. coli using the protocol described by Studier. Briefly the protein was expressed in PASM-5052 auto induction media, supplemented with 100 μg/mL ampicillin, 200 μg/mL each of 17 aa (no C, Y, M), 10 μg/mL methionine, 125 μg/mL selenomethionine and 100 nM vitamin B₁₂ (Studier, 2005).

The frozen cell paste expressing cDPGS was thawed and re-suspended in 50 mM Tris–HCl pH 7.5, 300 mM KCl, and 20 mM imidazole. The cells were disrupted by sonication at 10 μm (Soniprep150; MSE) on ice for 4 min and the cell debris was removed by centrifugation at 24000 g at 4°C for 30 min. The clarified cell lysate was then heat-treated at 60°C for 30 min before being centrifuged at 24000 g at 4°C for 30 min to remove any denatured proteins. The protein was purified using a 1 mL HisTrap FF crude column (Cytiva) using a two-step gradient. Step one increased the imidazole concentration to 40 mM for 10 column volumes (CV), whilst step two increased the concentration to 300 mM imidazole for 20 CV. The eluate from the second step was applied to a calibrated Superdex 200 HiLoad 16/60 gel filtration column (Cytiva) and eluted with one column volume of 20 mM Tris–HCl pH 7.5, 300 mM KCl, at 1.0 mL min⁻¹.