2.1. Cloning

After codon optimization for expression in E. coli, genes encoding the mevalonate kinase from Methanosarcina mazei (MKMaz) and the two putative mevalonate kinases from Ramazzottius varieornatus (MKvar; BDGG01000012.1, GAV05667.1) and Methanococcoides burtonii (MKbur; WP_011500381.1, Q12TI0) were ordered from Twist Bioscience (San Francisco, California, USA; Supplementary Fig. S1). The genes were cloned into a pBAD vector, adding a sequence encoding a 6×His tag to the N-terminus (Esquirol et al., 2022). Subcloning was performed in DH5α cells (New England Biolabs, USA).

2.2. Protein expression

Vectors built using an acceptor vector (cassettes 5 and 6) producing His-tagged protein were used to transform E. coli BL21 (λDE3) cells (New England Biolabs, USA). Bacteria were grown on lysogeny broth (LB) medium containing 100 µg ml⁻¹ ampicillin for the pBAD vectors at 37°C with shaking at 180 rev min⁻¹ until OD₆₀₀ reached 0.5. Induction was triggered by the addition of 0.2% L-arabinose and the temperature was decreased to 28°C. The cells were incubated overnight before harvesting by centrifugation at 5000g for 15 min using an Avanti J-E centrifuge (Beckman Coulter, Indianapolis, USA).

The temperature was kept at 4°C during all of the following purification steps. The cells were resuspended in buffer A (50 mM Tris, 100 mM NaCl, 5 mM imidazole pH 8) and lysed by passage through an Emulsiflex-C5 homogenizer (Avestin, Canada) five times at 137 MPa. Lysis was followed by centrifugation at 18 000g for 30 min to pellet the cellular debris. The soluble fraction was then filtered through a 0.22 µm syringe filter (Millipore, USA). The soluble fraction was loaded onto a HisTrap FF column (Cytivia) and eluted with buffer B (50 mM Tris, 100 mM NaCl, 500 mM imidazole pH 8) using an ÄKTApure (GE Healthcare Life Sciences). The purity of the sample was assessed by SDS–PAGE analysis on Mini-PROTEAN Precast Gels (Bio-Rad). Fractions containing the protein of interest were concentrated using an Amicon Ultra centrifugal filter and rinsed with a buffer consisting of 50 mM Tris, 100 mM NaCl pH 8. A final purification step was performed on a Superdex 200 10/300 GL size-exclusion column (GE Healthcare) equilibrated with 50 mM Tris, 100 mM NaCl pH 8. Final concentration of the proteins was performed using an Amicon Ultra centrifugal filter. The concentrated proteins (MKvar at 8.8 mg ml⁻¹, MKbur at 6 mg ml⁻¹ and MKmaz at 4.7 mg ml⁻¹) were aliquoted in 100 µl tubes, snap-frozen in liquid nitrogen and kept in a freezer at −80°C until further use.

2.3. Mass spectrometry

To validate the sequence of His-tagged mevalonate kinase from R. varieornatus (UniProt Accession No. A0A1D1VW28), a gel band of purified MKvar was subjected to manual in-gel reduction, alkylation and tryptic digestion. The sample was reduced with 10 mM DTT (Sigma) for 30 min, alkylated for 30 min with 50 mM iodoacetamide (Sigma) and digested with 375 ng Trypsin Gold (Promega) for 16 h at 37°C. Extracted peptides were then analysed by matrix-assisted laser de­sorption ionization (MALDI) using an UltrafleXtreme MALDI-TOF mass spectrometer (Bruker). Equal volumes of tryptic peptide and matrix solutions were mixed and deposited onto ground-steel BC target plates (MTP 384, Bruker). External calibration was performed using Peptide Calibration Standard II (Bruker, m/z mass range 700–3500 Da) and MALDI-TOF spectra were searched against a decoy database containing sequences of the MKvar construct plus common proteomics contaminants using the MASCOT peptide mass fingerprint search engine (version 2.3; Matrix Science).

To confirm the accurate mass of MKvar, liquid chromatography–mass spectrometry (LC-MS) was carried out using an UltiMate 3000 HPLC system (Thermo Fisher Scientific) coupled to a maXis II Q-TOF mass spectrometer (Bruker), as described previously (Newman et al., 2019). Briefly, a 10 ml aliquot of purified protein (1 mg ml⁻¹) was loaded onto a 50 × 4.6 mm, 5 µm particle-size, 300 Å pore-size PLRP-S column (Agilent) pre-equilibrated with 0.1% formic acid. The protein was eluted from the column at a flow rate of 250 µl min⁻¹ by applying a linear 30 min gradient from 0 to 80% mobile phase B [mobile phase A, 0.1%(v/v) formic acid; mobile phase B, 90%(v/v) acetonitrile/0.1%(v/v) formic acid] and ionized using an Apollo II electrospray ion source (Bruker) with the nebulizer pressure set to 1.8 bar and dry gas maintained at 220°C at a flow rate of 8 l min⁻¹. High-resolution LC-MS data were analysed using the Protein Metrics Intact Mass parsimonious charge-state deconvolution algorithm (Bern et al., 2018).

2.4. Structure determination of MKvar

MKvar protein (3 mg ml⁻¹ in 50 mM Tris–HCl, 50 mM NaCl pH 8 with either 2 mM mevalonate or 2 mM ATP) was set up in initial crystallization screens at both 20 and 8°C. An initial hit which grew overnight [0.2 M calcium chloride, 28%(v/v) polyethylene glycol 400, 0.1 M sodium HEPES 7.5, protein and mevalonate] was used to seed into other PEG 400 conditions. Finally, additive screening experiments were set up around hits from the optimization step. Crystals grown from the additive screening showed diffraction limits from 6 Å to better than 2.5 Å. The best crystals contained protein treated with mevalonate and were harvested from drops containing 0.126 M calcium acetate, 24.9%(v/v) polyethylene glycol 400, 0.09 M sodium HEPES pH 8.4 with the additives Gly-Gly-Gly (0.03 M) or taurine (0.01 M) at 20°C. All crystallization trials were set up in SD-2 plates (SwissSci) with drops consisting of 150 nl protein solution and 150 nl reservoir solution equilibrated against 50 µl reservoir solution. Crystals were harvested by adding 1.3 µl reservoir solution to the crystal drop and then gently pulling the crystals out with a MiTeGen mylar loop. No additional cryoprotection was added as the growth conditions were sufficiently cryoprotecting.

Two data sets from the MX1 beamline at the Australian Synchrotron were processed using XDS (Kabsch, 2010) and AIMLESS (Evans & Murshudov, 2013) to give a total of 720° of data to higher resolution and merged as the unit-cell parameters were essentially the same; this gave a 2 Å resolution data set in space group P2₁2₁2₁. All data sets were collected at 100 K, collecting 360° of data in 36 s.

The initial molecular-replacement (MR) solution obtained using PDB entry 2r42, CHAINSAW (Stein, 2008) and Phenix (Liebschner et al., 2019) was poor. It was clear that there were two copies of the protein in the asymmetric unit and there was decent density for a portion of the structure near the N-terminus and then again near the C-terminus. Several mevalonate kinase structures were overlaid and common features were manually cut out to build significant portions of the structure using PDB entries 2r42, 1kkh, 1kvk, 2hfu, 4hac and 4rkz. After two rounds of this process, a model that was about 60% of the full-length protein gave a significantly better score in Phenix: a refined LLG of 248 and a TFZ of 17.8 (the previous values were an LLG of 111.9 and a TFZ of 12.8).

Buccaneer (Cowtan, 2006) was used to attempt to build a model into the density, but this failed. A model was built into the density manually with Coot (Emsley et al., 2010) for four rounds of manual and REFMAC (Murshudov et al., 2011; Agirre et al., 2023) refinement before setting up Phenix Autobuild (Terwilliger et al., 2008). Phenix Autobuild gave a more complete structure, which required another nine rounds of manual rebuilding (Coot) and refinement (REFMAC) to give a final structure.

PDBeFold (Krissinel & Henrick, 2004) showed that we did not choose the most structurally homologous available PDB entries for the initial MR. PDB entry 4ut4 has 17% sequence identity and an r.m.s.d. of 2.0 Å to our structure over 308/309 aligned amino acids, with a Q-score of 0.49 and a Z-score of ∼13.5. PDB entry 2r42 was 19th on the list, with a Q-score of 0.39, a Z-score of 9.9, 333 aligned residues and a sequence identity of 33% with an r.m.s.d. of 2.9 Å.

Using the SSM algorithm (Krissinel & Henrick, 2004) in Coot the numbers are slightly different: an r.m.s.d. of 3.05 Å over 335 aligned residues with 21 gaps and 34% sequence identity for PDB entry 2r42. When visualized, it is clear that the two structures have the same fold, but much of their secondary structures (particularly the helices and loops) are out of registration with each other, which was why it was a difficult MR solution.

For MKbur (6 mg ml⁻¹), crystals were obtained from an optimization plate set up with 20% PEG 8000, 200 mM MgCl₂, 100 mM Tris pH 8.5 with 3% trehalose as an additive. Well diffracting crystals that diffracted to about 2.2 Å resolution were obtained from several drops in this additive plate, so the additives were not the important factor here (they were all quite different). Crystals were obtained at both 4 and 20°C, so temperature was not particularly important in this case either. In both cases, MKvar and MKbur, the protein was flash-frozen in liquid nitrogen in small (100 µl) aliquots and defrosted just prior to crystallization trials.

An initial MR (Phaser; McCoy et al., 2007) solution for MKbur was found in space group P2₁ using the MKvar structure as a starting point. A solution for a single protomer was found; this was modified manually in Coot (Emsley et al., 2010) and a second round of Phaser was then run to obtain two molecules in the asymmetric unit. The coordinates were rebuilt manually in Coot and were refined with REFMAC (Murshudov et al., 2011; Agirre et al., 2023).

The structures were validated using the tools in Coot and the final structures were verified using the validation tools from the RCSB PDB during the deposition process. Ramachandran statistics from the RCSB PDB are provided in Table 1. Coordinates and structure factors have been deposited in the PDB as entries 8tfo and 8teb.

2.5. Protein alignment

For sequence alignment, the T-Coffee server was used (Notredame et al., 2000). All images were produced with PyMOL (version 1.8, Schrödinger; DeLano & Lam, 2005; DeLano, 2009) and the PDB entries for mevalonate kinases from Homo sapiens (PDB entry 2r3v; Fu et al., 2008), Rattus norvegicus (PDB entry 1kvk; Fu et al., 2008), Staphylococcus aureus (PDB entry 2x7i; Voynova et al., 2004; Oke et al., 2010), Methanocaldococcus jannaschii (PDB entry 1kkh; Huang et al., 1999; Yang et al., 2002), Streptococcus pneumoniae (PDB entry 2oi2; Andreassi et al., 2007, 2009) and Methanosarcina mazei (PDB entry 6mde; Miller & Kung, 2018).

2.6. Enzyme assay

2.7. Differential scanning fluorimetry (DSF)

The three mevalonate kinases were each diluted to a final concentration of ∼0.07 mg ml⁻¹ (1.5 µM) in a final volume of 512 µl crystallization buffer (50 mM Tris–HCl, 50 mM NaCl pH 8.3); 8 µl 1:20 diluted SYPRO dye (Sigma) was added and 24 replicate wells for each protein were set up in a 96-well plate (Thermo AB-800 white PCR plate). Lysozyme at 0.02 mg ml⁻¹ in the same buffer, with the same amount of SYPRO dye, was used as a control (also 24 replicate wells). The temperature was increased from 20 to 100°C in increments of 0.5°C every 5 s using a Bio-Rad CFX96 RtPCR machine. The results were viewed using the Bio-Rad CFXmanager software and the melting temperature was derived using Meltdown (Rosa et al., 2015).

3.1. Difference in thermostability of the mevalonate kinases

Two uncharacterized mevalonate kinase candidates were identified by using protein BLAST to search the genomes of the eukaryote R. varieornatus (Hashimoto et al., 2016) and the psychrophilic archaeon M. burtonii (Allen et al., 2009) using the sequences of MKmaz and rat (Rattus norvegicus) mevalonate kinase (MKrat; Fu et al., 2002; Supplementary Fig. S1). The survival of extremophile organisms often relies on the production of isoprenoid derivatives; for example, in the case of M. burtonii an unusual lipid content in the membrane is thought to allow resistance to cold temperatures (Nichols et al., 2004). Therefore, we sought to investigate the mevalonate kinases from these cold-adapted or cold-tolerant organisms.

The MK genes encoding MKvar and MKbur were ordered, cloned and expressed with a C-terminal 6×His tag for MKbur and MKmaz and an N-terminal 6×His tag for MKvar for purification by Ni–NTA and size-exclusion chromatography. Recombinant MK proteins were found to elute at around 180 mM imidazole for MKvar and at 100 mM imidazole for both MKbur and MKmaz (Supplementary Fig. S2).

Differential scanning fluorometry (DSF) shows melting temperatures of 48.4, 65.7 and 66.7°C for MKbur, MKvar and MKmaz, respectively (Fig. 2b, Supplementary Fig. S3). The structures of the archaeal MKs MKmaz and MKbur are highly similar and their sequence alignment shows 61.1% identity and an r.m.s.d. of 0.77 Å (Supplementary Table S1); however, their thermostability assay and DSF results show a striking 20°C difference. Measuring residual activity after incubation at temperatures ranging from 25 to 80°C for 5 min revealed that the T₅₀ values were 40°C for MKbur and 62°C for both MKvar and MKmaz (Fig. 2a). The specific activity of MKmaz and MKbur incubated at 4, 15 and at 37°C was measured over time and while the highest specific activity for MKmaz was obtained at 37°C, MKbur appears to be most active at 15°C. Interestingly, in the case of MKbur activity can be detected with the same order of magnitude as in the 15°C samples after 2 h at 4°C (Fig. 2).

Figure 2 (a) Residual activity of the MKs after 5 min incubation at a given temperature as a percentage of the highest specific activity. (b) Differential scanning fluorometry showing the melting temperature of the protein. (c, d) Specific activity of (c) MKbur and (d) MKmaz after incubation for 0, 1, 2 and 24 h at 4, 15 and 37°C (n = 3).

Chu et al. (2007) linked the thermostability of another archaeal MK, that from Methanocaldococcus jannaschii, to a disulfide bond linking a cysteine located in motif II and a cysteine located in motif III. A mutation of an alanine residue in motif III into a cysteine in MKrat was also shown to increase the thermostability of the enzyme. Although MKmaz, MKvar and MKbur possess the cysteine in motif III, MKmaz and MKvar lack the cysteine in motif II. It would be interesting to perform mutagenesis to assess its potential impact on thermostability.

In their review, Liu and coworkers list several characteristics of cold-activated enzymes: differences in the primary structure, such as proline-poor sequences, the presence of glycine, serine and histidine in the central domain, and differences in hydrogen bonding, where less bonding grants more flexibility; flexibility of the active site at the expense of binding efficiency is also reported (Feller & Gerday, 2003). Finally, enzymes from psychrophiles have fewer disulfide bonds, greater hydrophobicity and hydrophobic surfaces, and also appear to have a larger active-site cavity compared with mesothermophilic and thermophilic homologues (Liu et al., 2023). Futher studies will be needed to understand the structural bases that allow MKbur to function at very low temperature.

3.2. Novel MKs display the conserved motifs of the GHMP kinase superfamily

Mevalonate kinases belong to the GHMP kinase superfamily, which includes enzymes such as galactokinase, homoserine kinase, MK and phosphomevalonate kinase. The sequence alignment of MKmaz, MKrat, MKvar and MKbur is shown in Supplementary Fig. S4. Archaeal and eubacterial MK sequences tend to be shorter than those of eukaryotic MKs and differ in the N-terminal region of the enzyme, lacking two α-helices and the loop between them (Fig. 3, Supplementary Fig. S4). The C-terminus of the MKrat sequence is thought to be involved in metabolite-mediated allosteric inhibition of the enzyme, and it is notable that the eukaryotic sequence does not align with the C-termini of the feedback-resistant archaeal proteins (MKmaz and MKbur; Fu et al., 2008; Supplementary Fig. S5).

Figure 3 Cartoon representation of the monomers of (a, b) MKvar and (c, d) MKbur showing the N-terminal domains in yellow and the C-terminal domains in green (a, c) and with labelled helices in red, loops in green and β-sheets in yellow (b, d). An MVA molecule is shown as blue sticks (a, b) in the active site of MKvar; the locations of GHMP motifs I, II and III are indicated in pink, blue and grey, respectively, and the location of an Mg2+ ion involved in ATP binding is indicated by a cyan sphere in MKbur (c, d).

We confirmed the identity of MKvar before crystallization and structural characterization using MALDI-TOF analysis of a tryptic digest. The sample was confidently identified as a tardigrade mevalonate kinase (UniProt Accession No. A0A1D1VW28) with 62% sequence coverage (Supplementary Fig. S7). Intact mass analysis by LC-MS suggested that the protein was very pure, with a major mass of 45 277.0 Da, corresponding to the theoretical mass of His-tagged MKvar. A mass addition of 178 Da suggested gluconoylation of the His tag. E. coli has been previously reported to (phospho)gluconoylate sequences (Oke et al., 2010; Supplementary Fig. S7).

As expected, the sequences contain the conserved motifs of the GHMP enzyme superfamily (Yang et al., 2002; Andreassi & Leyh, 2004; Fu et al., 2002). Motif I is shown in pink in Fig. 3 and Supplementary Fig. S5. The second conserved GHMP motif is motif II, which is shown in blue (Fig. 3, Supplementary Fig. S5); this motif forms a loop that is involved in ATP binding (–PXGXGLGSSAA–). Finally, motif III, which is shown in grey (Fig. 3, Supplementary Fig. S5), is a conserved region which forms a glycine-rich loop and is known to be involved in substrate binding. The aspartate, which is the main catalytic residue (as determined by mutagenesis studies; Potter & Miziorko, 1997), is identified with a star, and other conserved amino acids involved in stabilizing the substrates are marked by arrows (Fu et al., 2002; Andreassi & Leyh, 2004; Potter & Miziorko, 1997; Supplementary Fig. S5).

The data-collection and refinement statistics for the MKvar and MKbur crystal structures can be found in Table 1. The His tag was not resolved in the structure. Like all GHMP superfamily enzymes, each monomer is composed of two domains: an N-terminal domain and a C-terminal domain separated by a cleft (Fig. 3a). Both the mevalonate-binding (motif I and III) and the ATP-binding (motif II) sites are located in the cleft between the two domains, where the mevalonate molecule can be seen (Fig. 3b). Motif I is found in β2 in both MKs. Motif II is located in a loop between β7 and α5 in MKvar and between β6 and α2 in MKbur; it is identical in all of the enzymes from the superfamily and is not thought to rearrange upon binding to ATP (Andreassi et al., 2007; Fig. 3). Finally, motif III is in a loop between β13 and β14 in MKvar and β10 and β11 in MKbur. Andreassi and coworkers describe the loop (between β12 and α8 in MKvar and between β9 and α5 inr MKbur) as a mobile section that forms a lid on the active site that protects MVA and ATP from solvent (Andreassi et al., 2007; Fig. 3). The catalytic aspartate residue is found at the N-terminus of α7 in MKvar and the N-terminus of α4 in MKbur; this residue is highly conserved across the GHMP family, and although these enzymes act on different substrates and have different regulatory mechanisms (bi-bi, allostery), the involvement of the aspartate residue in the phosphorylation reaction is conserved (Roy et al., 2019).

A highly conserved threonine residue at the N-terminus of the α8 helix of MKvar/the α5 helix of MKbur (Figs. 3 and 4b) was also found to be involved in ATP and mevalonate binding, and its mutation led to an increase in K_m and a decrease in k_cat (Cho et al., 2001). This mutation is also linked to the development of pathology in humans (Andreassi et al., 2007).

Figure 4 Zoom in around the active sites of MKbur in orange and MKvar in green, where (a) and (b) are centred around the mevalonate shown in cyan lines and (c) and (d) are centred around the ATP shown in lines and the magnesium ion (Mg2+) represented as a cyan sphere (the localization of ATP is estimated by alignment with MKrat crystallized with ATP in the active site; Fu et al., 2002). Conserved amino acids are shown as sticks, with motifs I, II and III in pink, blue and grey, respectively. The loops involved in binding the nucleotide moiety are indicated in dark red and the lid is shown in yellow in (c) and (d). Distances in Å between elements are represented by yellow dotted lines.

A loop region between α3 and α4 in MKvar is not seen in the X-ray structure, suggesting high mobility of this loop; mobility in this region is also observed in some other MK structures (Fu et al., 2008). The main structural differences observed between the MKs are found in the N-terminal domain, mirroring what is seen in the sequence alignment. The N-terminal sections missing from eubacterial and archaeal MKs mostly form the two helices α2 and α3 in MKvar and can also be found in rat and human MK (Fu et al., 2008).

The MK structures were run through PDBeFold and the closest matches can be found in Supplementary Table S1, arranged by Q-score. Interestingly, the most structurally homologous match to MKvar was not a mevalonate kinase but was PDB entry 4ut4, the heptokinase WcbL from Burkholderia pseudomallei (Vivoli et al., 2015), which was followed by a range of archaeal and eubacterial MKs and a phospho­mevalonate kinase from Streptococcus pneumoniae. MKbur returned very high similarity to MKs from archaeal origin, with the highest Q-scores of 0.9 for MK from Methanosarcina mazei and 0.6 for MK from Methanocaldococcus jannaschii.

3.3. Functional characterization of MKvar and MKbur

Steady-state kinetic values for (RS)-mevalonate and ATP were determined using a coupled enzyme assay as used previously (Primak et al., 2011; Vinokur et al., 2014; Fu et al., 2008) and apparent kinetic parameters are given in Table 2; kinetic values for previously published mevalonate kinases are given in Table 3. Phosphatidic compounds are common inhibitors of mevalonate kinases (Fu et al., 2008; Nyati et al., 2015; Andreassi et al., 2007). The MK from Staphylococcus aureus (MKStaph) was reported to be inhibited in the presence of its product mevalonate 5-phosphate at 1.56 mM (Voynova et al., 2004; Table 3). Voynova and coworkers determined that the product inhibition of MKStaph was a sign of an ordered sequential mechanism involving first the binding of mevalonate before ATP and then the release of the product before the release of ADP (Voynova et al., 2004). MKvar was found to be inhibited in the presence of >2 mM mevalonate substrate. Although IC₅₀ and K_i are not directly comparable, globally the range of sensitivity is as follows: the most sensitive are the mammalian and mosquito MKs, inhibition of which by phosphatidic compounds is within the 10⁻⁸–10⁻⁷ M range, then yeast MKs, which are inhibited at around 10⁻⁶ M, followed by eubacterial MKs and Methanocaldococcus jannaschii MK, which are inhibited in the 10⁻⁵–10⁻³ M range (Table 3); finally, inhibition is undetected in the archaeal MKs.

Table 3 Kinetic constants and inhibition profiles for various mevalonate kinases MVA, mevalonate; PP-MVA, pyrophosphate mevalonate; GPP, geranyl pyrophosphate; FPP, farnesyl pyrophosphate; 5-P-MVA, mevalonate 5-phosphate. Organism Kingdom Km, MVA/ATP (µM) kcat, MVA (s−1) kcat/Km (s−1 mM−1) Ki (µM) References Human Eukaryotes 40.8/178 ND ND 0.035 (FPP) Fu et al. (2008) Rattus norvegicus Eukaryotes 35/950 22 628 0.35 (FPP) Chu & Li (2003), Fu et al. (2008), Voynova et al. (2004), Chu et al. (2007) Aedes aegypti Eukaryotes 90/140 10 111 0.55 (GPP), 0.44 (FPP) Nyati et al. (2015) Saccharomyces cerevisiae Eukaryotes 131/650 38 290 2.3 (GPP), 1.9 (FPP) Primak et al. (2011) Staphylococcus aureus Eubacteria 41/339 ND ND 1560 (5-P-MVA), 46 (FPP) Voynova et al. (2004), Oke et al. (2010) Streptococcus pneumoniae Eubacteria 236/372 11 47 >100 (FPP), PP-MVA† Primak et al. (2011), Andreassi et al. (2007) Methanocaldococcus jannaschii Archaea 68.5/92 28.5 419 45% (10 µM GPP)‡, 35% (10 µM FPP)‡ Huang et al. (1999), Chu et al. (2007) Methanosaeta concilii Archaea 17/74 14 823 Not detected ‡ Kazieva et al. (2017) Methanocella paludicola Archaea 15/119 7 466 Not detected ‡ Kazieva et al. (2017) Nitrosopumilus maritimus Archaea 461/1006 12 26 Not detected ‡ Kazieva et al. (2017) †Activity was detected. ‡Percentage residual activity.

The active sites of MKvar and MKbur are highly conserved compared with the active sites of the other characterized MKs (Fig. 4); all of the conserved amino acids are shown in sticks for MKvar and align with those in MKbur, as well as the motifs I, II and III characteristic of the GHMP family.

The C5 hydroxyl group of the mevalonate substrate is thought to be deprotonated by the catalytic aspartate residue, which then proceeds to perform a nucleophilic attack on the γ-phosphate of ATP (Fig. 5).

Figure 5 Scheme of the mevalonate kinase mechanism that converts mevalonate and ATP into mevalonate 5-phosphate and ADP, where the amino-acid numbering represents the sequences of MKbur/MKvar.

The mevalonate molecule is stabilized by interactions with the conserved amino acids located near the catalytic amino acid Asp224/Asp138 (in MKvar/MKbur). The catalytic role of this aspartate residue was demonstrated by Potter and Mizioro, who showed that mutation of the aspartate to an alanine or an arginine led to no detectable activity (Potter & Miziorko, 1997). The roles of Ser221/Ser135, His20/His16, Lys14/Lys9 and Ala354/Ala259 in MKvar/MKbur have been deduced from their conservation amongst the GHMP family as well as by their close location to the substrate (Sgraja et al., 2007; Huang et al., 2016; Miller & Kung, 2018; McClory et al., 2019; Figs. 4a and 4b).

The highly conserved Glu213/Glu127 and Ser163/Ser95 (in MKvar/MKbur) have been found to be involved in ATP binding and more specifically to coordinate the Mg²⁺ ion that activates the γ-phosphate of ATP (Yang et al., 2002; Fu et al., 2002; Figs. 4c and 4d); when this Glu was mutated to an alanine by Potter and Mizioro it turned the human MK into a labile protein, and when it was replaced by a glutamine it lowered the K_m for ATP and the V_max (Potter & Miziorko, 1997). The conserved Thr residue involved in binding the tail of the ATP can be seen in the yellow lid portion (Thr263/Thr174 in MKvar/MKbur; Figs. 4c and 4d; Cho et al., 2001). A conserved Arg residue in MKrat (Arg241), corresponding to Arg261 in MKvar, was found to be essential in binding the tail of ATP and moves inside the active site upon binding the substrates (McClory et al., 2019); no equivalent arginine residue can be found in MKbur or MKmaz, where it is replaced by a serine.

3.4. Sensitivity to inhibition

MKs are inhibited as part of the terpene pathway regulatory mechanism, usually by longer prenylphosphates built by the condensation of IPP/DMAPP in the lower part of the pathway (Potter & Miziorko, 1997); however, there is a range of sensitivity depending on the genus of origin (Table 3). In general, longer prenylphosphate compounds, such as FPP, appear to cause greater inhibition than the shorter GPP, which agrees with our observations (Fig. 6).

Figure 6 Residual activity of each mevalonate kinase measured in the presence of 100 µM GPP or FPP (n = 3), where * indicates P ≤ 0.05, ** indicates P ≤ 0.01, *** indicates p-value ≤ 0.001 and *** indicates p-value ≤ 0.0001.

MKmaz and MKbur were found to possess activity in the presence of up to 100 µM of both FPP and GPP, compared with the almost total loss of activity recorded for MKvar, where <20% residual activity remains in the presence of 100 µM GPP and only 0.4% remains with 100 µM FPP (Fig. 6). Kazieva et al. (2017) did not observe any inhibition of MKmaz activity in the presence of 100 µM GPP and FPP, whereas we recorded a loss of activity, with 92% and 70% residual activity remaining in the presence of 100 µM GPP and FPP, respectively (Fig. 6). This decrease might not be a reflection of inhibition; it might simply be a reflection of the thermodynamics of the reaction, or it may be explained by the difference in assay methods. MKbur, like MKmaz, appears to be less sensitive to inhibition.

The N-terminal end of helix α4/α1 (in MKvar/MKbur), marked in dark green in Fig. 7 and known to be involved in inhibition, also varies between feedback-inhibited MKvar and MKbur. The structural basis of the inhibition of MKs has been studied using the MKs from rat and human (Fu et al., 2008). Fu and coworkers determined by mutagenesis that truncating the C-terminal end of human and rat MK (at Arg388, which does not exist in MKvar), and also mutating Thr104 and Ile196 in human MK, decreased the feedback inhibition in human MK (Fu et al., 2008). In MKvar, Thr119 and Ile215 align with Thr104 and Ile196, respectively, in human MK. MKbur has a conserved Ile residue; however, Thr119, which can be found at the N-terminus of the α4 helix (Figs. 4c and 4d) in MKvar, is not present in the N-terminal region of α1 in MKbur. The distance between Thr119 (the N-terminus of the α4 helix) and the ribose of the ATP molecule is ∼3 Å in MKvar, compared with a distance of ∼7 Å to the N-terminus of the α1 helix in MKbur.

Figure 7 Surface representation and surface of (a) MKvar, (b) MKrat, (c) MKmaz and (d) MKbur with mevalonate and ATP shown as sticks. The yellow portion is the lid, blue represents motif II and dark red represents a variable nucleotide-binding loop; the dark green section represents the end of helix α4/α1 (in MKvar/MKbur) or the equivalent in MKrat and MKmaz. The measurement represents the distance in Å between the N-terminus of helix α4/α1 (in MKvar/MKbur) and the loop lid. The localization of ATP is estimated by alignment with MKrat crystallized with ATP in the active site.

Comparison of the active sites of MKvar, MKrat, MKbur and MKmaz shows differences in the cleft region (Fig. 7, Supplementary Fig. S8 and Table S2). The sides of the cleft are composed of two variable regions: a variable loop (in dark red in Figs. 7, 4c and 4d) and the N-terminus of helix α4/α1 on one side (in MKvar/MKbur; dark green in Fig. 7) and a lid portion at the bottom on the other side (in yellow in Figs. 7, 3c and 3d). Further mutational studies would be needed in order to further specify the structural basis of inhibition of MKvar for prenylphosphate compounds or non-inhibition in the case of MKbur.