Modified mevalonate pathway of the archaeon Aeropyrum pernix proceeds via trans-anhydromevalonate 5-phosphate

Significance

Herein, the partially identified “modified” mevalonate pathway of the majority of archaea is elucidated using information from comparative genomic analysis. Discovery of two enzymes, mevalonate 5-phosphate dehydratase and trans-anhydromevalonate 5-phosphate decarboxylase, from a hyperthermophilic archaeon, Aeropyrum pernix, shows that the pathway passes through a previously unrecognized metabolite, trans-anhydromevalonate 5-phosphate. The distribution of the known mevalonate pathways among archaea and other organisms suggests that the A. pernix-type pathway, which is probably conserved among the majority of archaea, is the evolutionary prototype for the other mevalonate pathways involving diphosphomevalonate decarboxylase or its homologs.

Abstract

The modified mevalonate pathway is believed to be the upstream biosynthetic route for isoprenoids in general archaea. The partially identified pathway has been proposed to explain a mystery surrounding the lack of phosphomevalonate kinase and diphosphomevalonate decarboxylase by the discovery of a conserved enzyme, isopentenyl phosphate kinase. Phosphomevalonate decarboxylase was considered to be the missing link that would fill the vacancy in the pathway between mevalonate 5-phosphate and isopentenyl phosphate. This enzyme was recently discovered from haloarchaea and certain Chroloflexi bacteria, but their enzymes are close homologs of diphosphomevalonate decarboxylase, which are absent in most archaea. In this study, we used comparative genomic analysis to find two enzymes from a hyperthermophilic archaeon, Aeropyrum pernix, that can replace phosphomevalonate decarboxylase. One enzyme, which has been annotated as putative aconitase, catalyzes the dehydration of mevalonate 5-phosphate to form a previously unknown intermediate, trans-anhydromevalonate 5-phosphate. Then, another enzyme belonging to the UbiD-decarboxylase family, which likely requires a UbiX-like partner, converts the intermediate into isopentenyl phosphate. Their activities were confirmed by in vitro assay with recombinant enzymes and were also detected in cell-free extract from A. pernix. These data distinguish the modified mevalonate pathway of A. pernix and likely, of the majority of archaea from all known mevalonate pathways, such as the eukaryote-type classical pathway, the haloarchaea-type modified pathway, and another modified pathway recently discovered from Thermoplasma acidophilum.

The mevalonate (MVA) pathway provides fundamental precursors for isoprenoid biosyntheses, such as isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP). This pathway was discovered in the late 1950s through the study of cholesterol biosynthesis (Fig. 1A) (1, 2). In this pathway, the C₆ intermediate MVA is formed from acetyl-CoA via acetoacetyl-CoA and hydroxymethylglutaryl-CoA. It then undergoes two steps of phosphorylation catalyzed by mevalonate kinase (MVK) and phosphomevalonate kinase (PMK) to yield mevalonate 5-diphosphate (MVA5PP) via mevalonate 5-phosphate (MVA5P). The C₅ compound IPP is synthesized by the decarboxylation of MVA5PP accompanied by a detachment of its 3-hydroxyl group. To catalyze the reaction, diphosphomevalonate decarboxylase (DMD) consumes ATP to temporarily phosphorylate MVA5PP and form mevalonate 3-phosphate 5-diphosphate inside its catalytic pocket as shown recently by our mutagenic study (3). Detachment of the 3-phosphate group of the intermediate triggers decarboxylation to yield IPP. These ATP-dependent enzymes, MVK, PMK, and DMD, belong to the GHMP (galactokinase, homoserine kinase, mevalonate kinase, phosphomevalonate kinase) kinase family and show a certain level of homology. Conversion of IPP into DMAPP is catalyzed by IPP isomerase, which includes two evolutionary independent types of enzymes. This most widely accepted, sometimes called “classical” or “canonical,” MVA pathway exists in almost all eukaryotes and in certain forms of bacteria, such as lactic acid bacteria, whereas the vast majority of bacteria utilize the methylerythritol phosphate (MEP) pathway that proceeds through completely different intermediates from those in the MVA pathway.

Fig. 1.

Variation and distribution of the MVA pathways. (A) The MVA pathways known to date and discovered in this study. The names of enzymes are shown in boxes, which are colored in light blue, green, or pink when the enzymes are DMD homologs. IDI, isopentenyl diphosphate isomerase. (B) Distribution patterns of DMD homologs and the enzymes studied in this work. Each box represents an archaeal species selected on the basis of the one-species-for-each-genus rule (SI Appendix, Table S1). Boxes colored in light blue, green, pink, and gray indicate archaea possessing the (putative) genes of DMD, PMD, M3K/BMD, and a DMD homolog of unknown function, respectively, while white boxes mean their absence. Similarly, boxes colored in red represent the presence of the putative ortholog genes of proteins described on the left.

The “modified” MVA pathway was first proposed in 2006 by Grochowski et al. (4) based on the discovery of a new enzyme, isopentenyl phosphate kinase (IPK), and on data from comparative analyses of archaeal genomes. For archaea, which do not possess the MEP pathway, the MVA pathway is requisite for the biosynthesis of specific membrane lipids and other isoprenoids, such as respiratory quinones and dolichols. These organisms do have the putative genes of most enzymes in the aforementioned eukaryote-type MVA pathway; it is curious, however, that almost all archaea apparently lack the genes of one or two enzymes of the pathway, typically both PMK and DMD (5–7). Thus, Grochowski et al. (4) proposed a bypass pathway, called the modified MVA pathway, in which isopentenyl phosphate (IP) was formed from MVA5P by an undiscovered decarboxylase and was then phosphorylated by IPK, which is conserved in almost all archaea, to yield IPP (Fig. 1A). The decarboxylase [i.e., phosphomevalonate decarboxylase (PMD)] was recently identified from a halophilic archaeon, Haloferax volcanii (8), and a Chloroflexi bacterium, Roseiflexus castenholzii (9). The discovery substantiated the existence of the proposed modified pathway in these organisms. The pathway is, however, considered to be exceptional in the domain Archaea, because the gene of PMD, which is a close homolog to DMD, is conserved in all haloarchaea but not in most archaea. Different MVA pathways have been found from other unusual archaea that also possess DMD homologs, such as those of the orders Sulfolobales and Thermoplasmatales. The archaea of the order Sulfolobales, such as Sulfolobus solfataricus, are known to possess a eukaryote-type MVA pathway, but these are rare exceptions in archaea (10). In contrast, recent studies have proven that the archaea of the order Thermoplasmatales, such as Thermoplasma acidophilum and Picrophilus torridus, possess a distinctly modified MVA pathway, in which MVA is first converted into mevalonate 3-phosphate (MVA3P) by a DMD homolog, mevalonate 3-kinase (M3K) (Fig. 1A) (11–13). MVA3P is then phosphorylated by a non-GHMP family kinase, MVA3P 5-kinase, to form mevalonate 3,5-bisphosphate. The decarboxylation of the intermediate is catalyzed by another DMD homolog, bisphosphomevalonate decarboxylase (BMD), to yield IP (14). Interestingly, BMD does not require ATP to react, which suggests that the two functions of DMD (or PMD), phosphorylation and decarboxylation, were separately inherited by M3K and BMD, respectively. Therefore, all of the MVA pathways elucidated to date involve the DMD homologs, which are absent in the great majority of archaea (Fig. 1B and SI Appendix, Fig. S1).

This situation motivated us to search for undiscovered enzymes involved in the MVA pathway of the majority of archaea. We believed that the organisms would possess an isozyme of PMD, which shows no homology to DMD. Comparative genomic analysis, however, led to an unexpected discovery from the hyperthermophilic archaeon Aeropyrum pernix of two previously unidentified enzymes that convert MVA5P into IP via an intermediate, trans-anhydromevalonate 5-phosphate (tAHMP). This discovery meant that the majority of archaea, in which the putative orthologs of these enzymes are conserved, likely utilize the modified MVA pathway that goes via tAHMP and thus, is distinct from the known MVA pathways.

Results

Search for Enzymes Involved in the MVA Pathway.

To find candidates for the undiscovered enzymes involved in the modified MVA pathway, genes conserved in the archaea that lack the genes of DMD homologs were searched from the genomes of 88 archaeal species using the MBGD website (mbgd.genome.ad.jp) that can create sets of putative ortholog genes. The candidate genes that we searched for were expected to be absent in the archaea possessing the DMD homolog genes, such as those of the class Halobacteria and the orders Sulfolobales and Thermoplasmatales. By allowing for differences in several genomes, two gene sets, which are the putative orthologs of A. pernix genes APE_2087.1 and APE_2089, were selected as the candidates that best fit the requirements (Fig. 1B and SI Appendix, Table S1). These genes of A. pernix likely compose an operon that is annotated in the database as the genes encoding the large and small subunits, respectively, of putative aconitase. A group of aconitase homologs that includes the A. pernix proteins was previously named “aconitase X (AcnX)” by Makarova and Koonin (15), and several bacterial members of this group were recently shown to catalyze the dehydration reactions in hydroxyproline metabolism (16, 17). These facts suggest the possibility that the proteins APE_2087.1 and APE_2089 are the subunits of an enzyme hereafter designated as ApeAcnX, which might be a dehydratase or a decarboxylase that catalyzes the decarboxylation evoked by dehydration in the MVA pathway.

Identification of MVA5P Dehydratase from A. pernix.

Each of the archaeal proteins APE_2087.1 and APE_2089 was recombinantly expressed in Escherichia coli cells as a fusion with an N-terminal polyhistidine tag. Purification of APE_2087.1 by affinity chromatography yielded a brown-colored solution, which suggested that the protein has an Fe-S cluster as with other aconitase homologs. The protein aggregated immediately after purification, but copurification with APE_2089 yielded stable proteins (Fig. 2A). When copurified with APE_2089, however, the brown color of APE_2087.1 disappeared in a day after exposure to air. Fig. 2B shows the UV-visible spectrum of the APE_2087.1/APE_2089 solution copurified under anaerobic conditions, the color of which persisted for more than a week. A peak at around 400 nm suggests the existence of an Fe-S cluster. The solution of the proteins, regarded as ApeAcnX, was reacted with radiolabeled intermediates of the MVA pathways, such as MVA, MVA5P, and MVA5PP, and the mixtures were analyzed by normal-phase TLC. Only MVA5P was converted into an unknown compound with an R_f of 0.50, which is lower than that of IP at ∼0.6. This showed that ApeAcnX shows enzyme activity other than decarboxylation toward MVA5P (Fig. 2C). Conversion of MVA and MVA5PP was not observed, which indicates that the enzyme reaction is highly specific (SI Appendix, Fig. S2). The maximum ratio of the product of ApeAcnX to MVA5P was ∼20%, although an excess amount of the enzyme was used for the reaction, suggesting equilibrium with the substrate. The product could be recovered from a TLC plate and was reacted again with the enzyme (Fig. 2C). After the reaction, a major part of the product was converted back into MVA5P.

Fig. 2.

Elucidation of the function of ApeAcnX. (A) SDS/PAGE of copurified ApeAcnX. (B) UV-visible spectrum of 4 mg/mL ApeAcnX solution. (C) Normal-phase TLC analysis of the ApeAcnX reaction product. Lane 1, [2-¹⁴C]MVA5P reacted without ApeAcnX; lane 2, [2-¹⁴C]MVA5P reacted with ApeAcnX; lane 3, the ApeAcnX product recovered from TLC and reacted without ApeAcnX; lane 4, the ApeAcnX product recovered from TLC and reacted with ApeAcnX. ori, Origin; s.f., solvent front. (D) ¹³C-NMR spectra of the samples before (Left) and after (Right) reaction with ApeAcnX. Signals derived from the substrate [U-¹³C]MVA5P and the ApeAcnX product from [U-¹³C]MVA5P are indicated by overlaying blue and red bars, respectively (SI Appendix, Fig. S3).

To determine the structure of the ApeAcnX product, we performed NMR analysis using a ¹³C-enriched substrate. The enzyme reaction with [U-¹³C]MVA5P resulted in the emergence of small NMR signals supposedly derived from the product along with the signals of unreacted MVA5P (Fig. 2D, Table 1, and SI Appendix, Fig. S3). Their chemical shifts and coupling constants suggest that the product is derived from the 2,3-dehydration of MVA5P. Moreover, the chemical shifts of the emerged signals correspond well with those of the trans-anhydromevalonate moiety of pestalotiopin A [(E)-5-acetoxy-3-methylpent-2-enoic acid] (SI Appendix, Fig. S4) reported by Xu et al. (18). These facts indicated that ApeAcnX has the activity of MVA5P dehydratase, which produces tAHMP. Electrospray ionization–MS (ESI-MS) analysis of the reaction products from either nonlabeled MVA5P or [U-¹³C]MVA5P also detected ions corresponding to tAHMP (SI Appendix, Figs. S5 and S6).

Table 1.

¹³C NMR data for the ApeAcnX product from [U-¹³C]MVA5P

Compound and carbon no.	Chemical shift, ppm	Coupling pattern*	1JC-C values, Hz
Product (tAHMP)
1	177.0	d†	260
2	122.8	dd	284/260
3	145.1	ddd (app. td)	284/162/162
4	40.1	dd (app. t)	162/150
5	62.6	d	150
6 (3-CH3)	17.8	d	162
Pestalotiopin A (partial) (18)
1	172.7	—	—
2	120.6	—	—
3	151.8	—	—
4	40.4	—	—
5	63.1	—	—
6 (3-CH3)	18.3	—	—

app., Apparent; d, doublet; t, triplet.

^*Patterns resulted from ¹J_C-C coupling are indicated.

^†An additional 25-Hz coupling, which might have resulted from ³J_C-C coupling with C4, was observed, whereas a corresponding coupling was not clearly observed with the relatively broad signal of C4. The coupling might contribute to the broadening of the C4 signal along with the ³J_C-P coupling.

Identification of tAHMP Decarboxylase.

If tAHMP is an intermediate of the MVA pathway of A. pernix, there will be an enzyme that connects between tAHMP and IP, because A. pernix has a putative ortholog gene of IPK. We noticed that the gene of the UbiD-type decarboxylase homolog likely forms an operon with the genes of the MVA5P dehydratase subunits in the genomes of some archaea including methanogens, such as Methanosarcina acetivorans (SI Appendix, Table S1). Because UbiD catalyzes the decarboxylation of 3-polyprenyl-4-hydroxybenzoate in the bacterial biosynthetic pathway of ubiquinone (19), this type of decarboxylase is thought to be involved in the biosynthesis of respiratory quinones that also are found in some archaea; however, methanogens do not have respiratory quinones. This situation implies the involvement of UbiD-type decarboxylase in the modified MVA pathway of general archaea. In addition, the above-described candidate genes selected by comparative genomic analysis included, but in lower ranks, putative ortholog genes encoding UbiD-like proteins and those encoding UbiX-like proteins, which are regarded as the partners of UbiD-type decarboxylases (20–22) (Fig. 1B and SI Appendix, Table S1). Although these putative ortholog genes are also found in some archaea utilizing the known MVA pathways, such as several haloarchaea and all archaea of the orders Thermoplasmatales and Sulfolobales, this might be because their apparent distribution patterns are affected by incorporation of the genes of UbiD/UbiX homologs responsible for respiratory quinone biosynthesis or other forms of metabolism. For example, A. pernix has two genes of the putative orthologs of UbiD-type decarboxylase, APE_1571.1 and APE_2078; the latter is highly homologous to the UbiD homolog singly possessed by methanogens and thus, is likely involved in the MVA pathway.

Thus, we constructed the coexpression system of UbiD and UbiX homologs from A. pernix (APE_2078 and APE_1647, respectively) in E. coli cells. Because UbiX is known to be a flavin prenyltransferase that produces prenylated flavin mononucleotide (prFMN), which is a coenzyme required by UbiD (20–22), only APE_2078 was expressed as the fusion protein with a C-terminal polyhistidine tag, while APE_1647 was expressed without an affinity tag. Using the APE_2078 protein partially purified with a nikkel affinity column (Fig. 3A), we tested its enzyme activity by attempting a conversion of radiolabeled tAHMP, which had been purified by TLC, into another compound. TLC analysis of the reaction mixture showed that a radioactive spot with an R_f of 0.63 emerged with the disappearance of the spot of tAHMP (Fig. 3B). Because the R_f value of the product approximated that of IP, we verified the formation of IP by adding T. acidophilum IPK, Sulfolobus acidocaldarius geranylgeranyl diphosphate (GGPP) synthase, ATP, DMAPP, and Mg²⁺ in the same reaction. Through the reaction with the enzymes with known functions (23, 24), the product was converted into GGPP as shown by the reversed-phase TLC analysis of the alcohol from GGPP in Fig. 3C. This clearly proved that the product was IP, indicating that the UbiD homolog from A. pernix, APE_2078, definitely had tAHMP decarboxylase activity.

Fig. 3.

Elucidation of the function of APE_2078. (A) SDS/PAGE of a partially purified APE_2078. (B) Normal-phase TLC analysis of the reaction products from [2-¹⁴C]tAHMP. (C) Reversed-phase TLC analysis of the hydrolyzed products from the reaction with [2-¹⁴C]tAHMP or [4-¹⁴C]IP in the presence of T. acidophilum IPK and S. acidocaldarius GGPP synthase. ori, Origin; s.f., solvent front.

Verification of the MVA Pathway of A. pernix.

Because A. pernix possesses the putative genes of the enzymes responsible for the production of MVA5P from acetyl-CoA and for the conversion of IP into downstream metabolites, such as IPP and DMAPP, the discovery of MVA5P dehydratase and tAHMP decarboxylase strongly suggests the existence of a modified MVA pathway, which passes through the intermediate tAHMP (Fig. 1A). Therefore, we checked to see if the cell-free extract from A. pernix possessed the enzyme activities that would convert tAHMP into a downstream compound, IPP. Radiolabeled putative intermediates of the modified MVA pathway of A. pernix (MVA, MVA5P, tAHMP, IP, and IPP) along with an intermediate of the eukaryote-type MVA pathway (MVA5PP) were reacted with the cell-free extract in the presence of ATP, Mg²⁺, S. acidocaldarius GGPP synthase, and DMAPP. Radiolabeled GGPP was synthesized as the index of IPP formation by the action of enzymes contained in the cell-free extract and was extracted from the assay mixture with 1-butanol to be analyzed by reversed-phase TLC after phosphatase treatment (Fig. 4). The TLC autoradiogram indicated that tAHMP could be converted into IPP as well as MVA, MVA5P, and IP, whereas the conversion of MVA5PP was not observed. The conversion efficiency of tAHMP was, however, obviously lower than that of the downstream intermediate IP, suggesting that tAHMP decarboxylase activity in the cell-free extract was weak. Moreover, the conversion from tAHMP seemed inefficient even compared with those from the upstream intermediates MVA and MVA5P. This situation might be explained by the results from normal-phase TLC analysis of the assay mixture without GGPP synthase and DMAPP (SI Appendix, Fig. S7). IP was completely converted into IPP by the reaction, showing strong activity of IPK in the cell-free extract. Nevertheless, IP seemed to accumulate in the reaction with tAHMP, suggesting the inhibition of IPK by tAHMP.

Fig. 4.

Conversion assay with A. pernix cell-free extract. Radiolabeled GGPP was extracted from the reaction mixture containing ¹⁴C-labeled intermediates (A. pernix cell-free extract, ATP, Mg²⁺, S. acidocaldarius GGPP synthase, and DMAPP) to be analyzed by reversed-phase TLC after phosphatase treatment. ori, Origin; s.f., solvent front.

Discussion

In this study, we discovered a modified MVA pathway, which passes through a previously unknown metabolic intermediate, tAHMP, from A. pernix. Unlike the three MVA pathways known to date, the fourth MVA pathway lacks the homolog of DMD and instead, utilizes two previously unidentified enzymes, MVA5P dehydratase and tAHMP decarboxylase. MVA5P dehydratase from A. pernix is composed of large and small subunits: APE_2087.1 and APE_2089, respectively. Its putative orthologs from archaea comprise the type IIb subclass of the AcnX family, while some bacterial AcnX proteins of the types I and IIa subclasses were recently revealed to be cis- or trans-3-hydroxy-l-proline dehydratase involved in hydroxyproline metabolism (16, 17). This situation sparked an interest in the evolution of this group of enzymes along with the unknown function of the remaining type IIc subclass proteins. We also showed that the APE_2078 protein exhibits decarboxylase activity toward tAHMP, which is a unique property for a UbiD homolog, because all known UbiD-type decarboxylases react with aromatic substrates, such as 3-polyprenyl-4-hydroxybenzoate and (hydroxy)cinnamic acids (20); the rare exceptions are TtnD from Streptomyces griseochromogenes (25) and SmdK from Streptomyces himastatinicus (26) involved in the biosynthesis of secondary metabolites. The involvement of the enzyme in the MVA pathway is intriguing, because known UbiD-type decarboxylases require prFMN, which is synthesized from a probable downstream metabolite of the pathway, dimethylallyl phosphate. The enzymatic properties of tAHMP decarboxylase, however, must be thoroughly investigated later.

The modified MVA pathway found from A. pernix seems widely distributed among the domain Archaea, with the exceptions of haloarchaea and the orders Sulfolobales and Thermoplasmatales (Fig. 1B and SI Appendix, Fig. S1). The distribution pattern of the four MVA pathways in the domain Archaea suggests that the modified pathway is more primordial than the other pathways, including the eukaryote-type MVA pathway. In contrast, the eukaryote-type and haloarchaea-type MVA pathways are possessed only by a very limited number of species in the domain Bacteria (SI Appendix, Fig. S8), implying that the pathways in bacteria might have horizontal transfer origins. Given the hypothesis that eukaryotes have evolved from the fusion of archaea and bacteria (27), the modified MVA pathway should be considered the prototype for all known MVA pathways. The most conceivable evolutionary scenario of the MVA pathways is that PMD emerged first among the DMD homologs, probably via the evolution from some kinase of the GHMP family, and replaced MVA5P dehydratase and tAHMP decarboxylase to create the MVA pathway currently found in haloarchaea and some Chloroflexi bacteria. The replacement caused the additional consumption of an ATP molecule for the production of each molecule of IPP or DMAPP, but it might have allowed the organisms to save a portion of the cost for producing multiple proteins and a specific coenzyme or to avoid the use of an oxygen-sensitive enzyme. PMD seems suitable for aerobes, such as haloarchaea, while the Aeropyrum-type modified MVA pathway with lower ATP requirement can benefit anaerobes, in which ATP is in short supply. PMD evolved later into other homologs, such as DMD, M3K, and BMD, which caused an emergence of the eukaryote-type and the Thermoplasma-type MVA pathways. Based on these arguments, we propose that the Aeropyrum-type MVA pathway possessed by the majority of archaea should be called the “archaeal MVA pathway,” while the others could be called the “(eukaryotic) MVA pathway,” the “haloarchaea-type MVA pathway,” and the “Thermoplasma-type MVA pathway.”

Materials and Methods

Materials.

Precoated reversed-phase TLC plates, RP18 F_254S, and normal-phase TLC plates, Silica gel 60, were purchased from Merck Millipore. [2-¹⁴C]MVA5P (55 Ci/mol) and [1-¹⁴C]IPP (55 Ci/mol) were purchased from American Radiolabeled Chemicals, Inc. [U-¹³C]MVA was prepared as described elsewhere (28). All other chemicals were of analytical grade.

Comparative Genomic Analysis.

A search for putative ortholog genes distributed in a certain pattern in representative archaeal species, which were selected by the one-species-for-each-genus rule, was performed using a web service provided by MBGD (mbgd.genome.ad.jp), allowing some discrepancy (similar pattern search) (29). Multiple alignments of the amino acid sequences of homologous proteins were performed using the online version of the MAFFT program (https://mafft.cbrc.jp/alignment/server/) with default settings. Phylogenetic trees were constructed via the neighbor-joining method using a CLC Sequence Viewer, version 7.5 (CLC bio).

Enzyme Preparation.

Recombinant expression and partial purification of ApeAcnX (copurified APE_2087.1/APE_2089), APE_2078 (coexpressed with APE_1647), R. castenholzii PMD, S. solfataricus MVK, T. acidophilum IPK, and S. acidocaldarius GGPP synthase were performed as described in SI Appendix, SI Materials and Methods.

Substrate Preparation.

[2-¹⁴C]MVA and [2-¹⁴C]MVA5PP were prepared from [2-¹⁴C]MVA5P as described elsewhere (10). For the preparation of [4-¹⁴C]IP, 3.64 nmol [2-¹⁴C]MVA5P was reacted with 0.4 mmol purified R. castenholzii PMD, 0.8 µmol ATP, 1 µmol MgCl₂, and 8 µmol sodium phosphate, pH 7.5, in a 200-µL reaction mixture. The enzyme was removed by filtration using a Vivaspin 500 centrifugation filter (10 kDa molecular weight cut off; GE Healthcare), and the filtrate was used as the solution of [4-¹⁴C]IP.

Radio-TLC Assay of ApeAcnX.

To detect the enzyme activity of ApeAcnX, 55 pmol of [2-¹⁴C]MVA5P was reacted with 17 µg of ApeAcnX in a 30-µL reaction mixture containing 3 µmol sodium phosphate buffer, pH 8.0. After 1 h of incubation at 90 °C, a 5-µL aliquot of the mixture was spotted on a Silica gel 60 normal-phase TLC plate and developed with chloroform/pyridine/formic acid/water (12:28:6:4). The distribution of radioactivity on the plate was visualized using a Typhoon FLA 9000 imaging analyzer (GE Healthcare) and quantified using Image Quant TL software (GE Healthcare).

Isolation of the Product of ApeAcnX and Reverse Reaction Assay.

A 50-µL reaction mixture containing 46 nmol [2-¹⁴C]MVA5P, 29 µg of ApeAcnX, and 5 µmol sodium phosphate buffer, pH 8.0, was incubated at 90 °C for 1 h. All of the mixture was linearly spotted on a normal TLC plate. After development with the same solvent system used above, the reaction product was recovered from the plate by scraping the area around its R_f and washing the scraped silica gel with 1 M ammonium acetate, pH 7.5. The ammonium acetate solution containing the radiolabeled product was concentrated by heating and used to assay the reverse reaction as [2-¹⁴C]tAHMP.

For the reverse reaction, a 30-µL reaction mixture containing 55 pmol of [2-¹⁴C]tAHMP, 17 µg of ApeAcnX, and 3 µmol sodium phosphate, pH 8.0, was incubated at 90 °C for 1 h. TLC analysis was performed as described above for the forward reaction.

NMR Analysis.

A 300-µL reaction mixture containing 2 µmol [U-¹³C]MVA, 9 nmol S. solfataricus MVK, 7.5 µmol ATP, 0.3 µmol of MgCl₂, 30 µmol sodium phosphate buffer, pH 7.5, and 10% (vol/vol) D₂O was incubated at 60 °C for 3 h. The enzyme was removed by filtration using a Vivaspin 500 spin column filter. To 260 µL of the filtrate, 520 µg of ApeAcnX was added, and the volume of the solution was adjusted to 600 µL with H₂O and D₂O, keeping the percentage of D₂O at 10%. The solution was then incubated at 90 °C for 2 h. After filtration to remove the enzyme, the ¹³C NMR spectrum of the product from the second reaction was analyzed using an AVANCE III HD 600 NMR spectrometer equipped with a cryoprobe (Bruker). As a negative control, the same volume of buffer was added in place of the ApeAcnX solution.

MS Analysis.

Procedures for negative ion ESI-MS analysis of the products of MVA5P dehydratase reaction from either nonlabeled MVA or [U-¹³C]MVA are described in SI Appendix, SI Materials and Methods.

Radio-TLC Assay of APE_2078.

A 30-µL reaction mixture containing 55 pmol [2-¹⁴C]tAHMP recovered from a TLC plate as described above, 6.2 µg purified APE_2078, and 3 µmol sodium phosphate buffer, pH 7.5, was incubated at 60 °C for 1 h. Normal-phase TLC analysis of the product was performed using the same procedure described above.

To confirm the production of IP, a 100-µL reaction mixture containing 82 pmol [2-¹⁴C]tAHMP, 3 µg of the purified APE_2078, 0.1 nmol T. acidophilum IPK, an excess amount of S. acidocaldarius GGPP synthase, 0.8 µmol ATP, 3 nmol DMAPP, 1 µmol MgCl₂, and sodium phosphate buffer, pH 7.5, was incubated at 60 °C for 1 h. Then, 200 µL of saturated saline was added to the mixture followed by the extraction of GGPP with 600 µL 1-butanol saturated with saline. Phosphatase treatment of GGPP was performed according to a method described by Fujii et al. (30). To the 1-butanol extract, 2 mL methanol and 1 mL of 0.5 M sodium acetate buffer, pH 4.6, containing 6 U acid phosphatase from potato (Sigma Aldrich) were added. After overnight incubation at 37 °C, geranylgeraniol was extracted from the phosphatase reaction mixture with 3 mL n-pentane. After the addition of 30 nmol farnesol and concentration under an N₂ stream, the pentane extract was spotted on an RP-18 F_254S reversed-phase TLC plate and developed with an acetone/water (9:1) mixture. The autoradiogram of the plate was obtained as described above. The same amount of [4-¹⁴C]IP was used as a control instead of [2-¹⁴C]tAHMP in the absence of APE_2078.

Conversion Assay Using Cell-Free Extract from A. pernix.

A. pernix was provided by the RIKEN BRC through the Natural Bio-Resource Project of the MEXT; cultured at 90 °C in a 250 mL medium, pH 7.0, containing 9.4 g Marine Broth 2216 (Difco), 1.2 g Hepes-NaOH, and 250 mg Na₂S₂O₃·5H₂O; and harvested by centrifugation. Then, 0.5 g of the cells were dissolved in 1 mL of 500 mM 3-morpholinopropanesulfonic acid (Mops)-NaOH buffer, pH 7.0, and disrupted by sonication using a Q125 ultrasonic processor (Qsonica). After centrifugation at 22,000 × g for 30 min at 4 °C, the supernatant was used as A. pernix cell-free extract.

A 100-µL reaction mixture containing 0.1 nmol of a radiolabeled substrate ([2-¹⁴C]MVA, [2-¹⁴C]MVA5P, [2-¹⁴C]MVA5PP, [2-¹⁴C]tAHMP, [4-¹⁴C]IP, or [1-¹⁴C]IPP), A. pernix cell-free extract containing 200 µg protein, 0.8 µmol ATP, 1 µmol MgCl₂, an excess amount of S. acidocaldarius GGPS, 3 nmol DMAPP, and Mops-NaOH buffer, pH 7.0, was incubated at 60 °C for 1 h. The radiolabeled GGPP was extracted with 1-butanol and analyzed by reversed-phase TLC after phosphatase treatment as described above.

Normal-phase TLC analysis of the products from the above reaction without GGPS and DMAPP were performed as described in SI Appendix, SI Materials and Methods.