A universal polyphosphate kinase powers in vitro transcription

Abstract

Polyphosphate kinases (PPKs) catalyze phosphoryl transfer between polyphosphates and nucleotides. Polyphosphates are a cost-effective source of phosphorylating power, making PPKs attractive enzymes for nucleotide production. However, at present, applications that require the simultaneous utilization of diverse nucleotides are not possible due to the restricted substrate profiles of PPKs. Here, we present a universal PPK capable of efficiently phosphorylating all eight common ribonucleotides (purines and pyrimidines, monophosphates and diphosphates) to triphosphates. Under optimal conditions, ~70% triphosphate conversion was observed for each substrate. To demonstrate the biotechnological potential of a universal PPK, we developed a one-pot assay for PPK-powered in vitro transcription. Primitive biology likely relied on enzyme promiscuity to support nascent metabolism with a compact proteome. This work highlights how applying the same principle to synthetic biology can facilitate the construction of complex in vitro reaction systems.

Currently polyphosphate kinases (PPKs), which catalyse phosphoryl transfer between polyphosphates and nucleotides, have restricted nucleotide substrate profiles. Here the authors discover and characterise a PPK capable of phosphorylating all eight common ribonucleotides.

Introduction

In 1956, Kornberg reported the discovery of polyphosphate kinases, enzymes that transfer phosphate groups between nucleotides and polyphosphate (PolyP)^1,2. Inspired by the structural simplicity of PolyP, its facile production from orthophosphate by heating, and its role in intracellular phase phenomena³, Kornberg hypothesized that PolyP served as a primitive source of phosphorylating power^4,5. Nearly 50 years later, Kornberg reported a second family of polyphosphate kinases (PPK2)^6,7 belonging to the ancient and ubiquitous P-Loop NTPase enzyme lineage^8,9. Whereas PPK1 favors PolyP elongation, PPK2 favors nucleotide phosphorylation and is well suited for energy regeneration, in which a pool of nucleotide triphosphates is maintained to drive catalysis^4,10. The first PPK2 enzymes to be characterized could phosphorylate either nucleotide monophosphates (Class II) or diphosphates (Class I). However, in 2014, the discovery of Class III PPK2 enzymes¹¹ that can convert AMP to ATP further expanded the scope of energy regeneration by PolyP ($0.016/mmol, Sigma Aldrich)^6,7, which is cheaper than more popular stores of phosphorylating power such as phosphoenolpyruvate ($55/mmol, Sigma Aldrich) or creatine phosphate ($37/mmol, Sigma Aldrich)¹². Increasingly, PPK2 is emerging as the key to developing a robust and cost-effective PolyP-driven energy regeneration system with broad biotechnological applicability, including in vitro chemical production¹³.

While metabolic and genome engineering of microbes for chemical production has advanced rapidly, the unpredictable, difficult-to-control nature of living organisms can limit the utility of these approaches¹⁴. Consequently, various in vitro platforms for chemical production have been developed—including reactions of immobilized enzymes, cascades mediated by multiple purified enzymes, and reactions carried out in crude cell extracts¹⁵. Many of these systems require energy regeneration, and proof-of-principle demonstrations of PPK2-powered in vitro chemical synthesis indicate significant progress towards realizing the goal of PolyP-driven regeneration^6,7,16–18, particularly for simple reactions that require a single nucleotide. Excitingly, a broad-specificity PPK2 enzyme that efficiently phosphorylates both adenine and guanine ribonucleotides (but not pyrimidines) was recently identified¹⁹ and applied to in vitro translation²⁰, which requires the regeneration of two nucleotides, ATP and GTP. However, reaction systems that require all four common nucleotides (ATP, GTP, CTP, and UTP) remain out of reach. While ATP and GTP often serve as energy sources, UTP and CTP also play important roles in metabolism. For example, UTP and CTP can activate sugars²¹ and lipid precursors²², respectively, that precede the production of various chemicals. Thus, a single enzyme capable of regenerating all common ribonucleotides would facilitate the design of on-demand in vitro metabolic pathways that bridge diverse nucleotide requirements. Does such a universal PPK2 (uPPK2) enzyme—defined here as an enzyme that can efficiently produce all four common nucleoside triphosphates by phosphoryl transfer from PolyP—exist?

Here, we report the discovery of a uPPK2 and demonstrate its utility by powering in vitro transcription (IVT). Four Class III PPK2 enzymes were selected and tested for their universal nucleotide phosphorylation activity. Despite close evolutionary relationships, only one enzyme—the PPK2 from Mangrovibacterium marinum, designated MAN—was able to phosphorylate all four common nucleotide monophosphates (AMP, GMP, CMP, UMP) and diphosphates (ADP, GDP, CDP, UDP) to their corresponding triphosphates (ATP, GTP, CTP, UTP), achieving ~70% conversion in each case. At elevated temperatures, adenine polyphosphates up to AP30, the longest nucleotide polyphosphate reported to date^23–25, could be produced. Key residues involved in the universal activity of MAN were investigated, and all analyzed mutants lost universal activity, revealing a rugged functional landscape. Finally, we demonstrate that MAN-mediated, one-pot synthesis of NTPs from nucleoside monophosphates (NMPs) can directly power IVT. We conclude that enzyme promiscuity can be a design principle for constructing simple reaction systems—a notable parallel to the proposed role of promiscuous enzymes in primitive biology^26,27. Given the simplicity of MAN-based nucleotide phosphorylation, we envision significant potential for MAN and similar enzymes in supporting diverse in vitro metabolic pathways¹⁵.

Results

Sequence selection strategy

An ideal uPPK2 should phosphorylate both pyrimidines and purines, and produce triphosphates from both monophosphate and diphosphate nucleotides—eight substrates in total. To find a naturally occurring uPPK2, we focused on Class III PPK2 enzymes^10,11,19. The typical Class III PPK2 enzyme is composed of a single P-loop NTPase domain of approximately 290 residues that oligomerizes in solution. Although an analysis of PPK2 gene complement across prokaryotes challenges a class-based heuristic for functional prediction^28,29, Class III enzymes have two attractive properties: First, they have been demonstrated to convert NMPs to NTPs either via two successive phosphoryl transfer reactions or by pyrophosphoryl transfer^{10,11,16,19,24}. Second, some Class III enzymes can accept both adenine and guanine nucleotides due to a key substrate-broadening mutation, an Asn at position 138 (numbering based on the PPK2 enzyme from Cytophaga hutchinsonii, referred to here as CHU¹⁹). By flipping the side-chain carboxamide of Asn, the enzyme can accommodate both the C6 amino group of adenine and the C6 carbonyl group of guanine (Fig. 1A). We hypothesized that a uPPK2 enzyme may employ the same mechanism to achieve broad purine specificity. And, with its ability to act as either a hydrogen bond donor or acceptor, N138 may also support the binding of both pyrimidines in the right context. For these reasons, we focused our search on Class III PPK2 enzymes with an Asn at position 138.

Fig. 1. Gene selection strategy.

A Side-chain flipping of N138 to mediate binding of adenine (PDB ID 6AN9) and guanine (PDB ID 6ANH)¹⁹. The primary polar interactions with adenine involve R117, H118, and N138, which form a network of hydrogen bonds with each other, the base, and two water molecules^19,76. The hydrogen bonding network upon binding guanine is unclear, as the lower resolution of the crystal structure complicates the interpretation of absent waters. B Class III PPK2 gene tree (left) and a zoom-in on the major N138 clade (right). The ring around the phylogenetic tree indicates the residue at position 138 (green for Asp, orange for Glu, and magenta for Asn). Genes selected for experimental characterization, as well as the previously characterized CHU enzyme, are indicated. C Multiple sequence alignment (CHU numbering) of selected constructs highlighting key functional regions^19,46–48, including the Walker A (blue), Walker B (yellow), and N138 (magenta). All selected sequences contain a variation of the Walker A motif, canonically GxxxxGK[TS], that lacks a conserved Ser/Thr at the last position, as is common among PPK2 enzymes⁸. Residues of the base binding pocket that differ between CHU and MAN are highlighted in orange (see also Fig. 4). Source data for this figure is available in the Source Data file.

A gene tree of Class III PPK2 reveals that enzymes with N138 belong to a limited number of clades (Fig. 1B). The major clade, to which most N138 enzymes belong, also contains the previously characterized CHU enzyme¹⁹. In addition to the above, the selection of PPK2 genes for functional profiling followed two simple criteria: First, PPK2 sequences should be of canonical length, roughly 290 residues^19,28. Second, genes should come from mesophiles. We focused on mesophiles for two reasons: First, mesophile proteins are expected to be active at moderate temperatures, making them compatible with various enzyme assay conditions, including cell-free protein synthesis^16,30,31. Second, recent work has shown that high temperature can delay the exploration of novel phenotypes³², suggesting that a uPPK2 may emerge first in mesophiles, especially if stability-function trade-offs³³ are significant. We note that any selection criteria may also exclude variants of interest. In total, four genes were selected from the major N138 clade of Class III (Fig. 1B, C and Supplementary Table 1): The PPK2 enzymes from Pedobactor sp. (PED), Mangrovibacterium marinum (MAN), Arenitalea lutea (ARE), and Leeuwenhoekiella palythoae (LEE). The pairwise amino acid identities of the selected enzymes range from 44.9 to 68.0%. MAN and LEE genes are the only PPK2 genes within their parent organism, while ARE and PED genes are one of two or more PPK2 genes.

MAN is a universal PPK2

Phosphoryl transfer activity of each selected PPK2 was quantified by HPLC (Fig. 2A) for eight nucleotides (four nucleoside monophosphates and four nucleoside diphosphates, hereafter referred to as NMPs and NDPs, respectively) and three PolyP chain lengths (10, 60, and 700 phosphate residues on average). Although the resulting functional profiles, reported here as fractional conversion to NTP, are relatively idiosyncratic (Fig. 2B), several general trends emerge: First, all enzymes tested are able to catalyze phosphoryl transfer to both mono- and diphosphate nucleotides, consistent with other Class III PPK2 enzymes. Second, PolyP-60 is preferred by all four enzymes, though the influence of PolyP length is not uniform and even adenine nucleotide utilization, which is an enduring feature of the PPK2 family, is sensitive to PolyP length (e.g., MAN with Poly-60 vs. PolyP-700). Finally, purines are generally preferred over pyrimidines. However, despite all proteins having N138, GTP production varied greatly, due in part to over-phosphorylation (Supplementary Fig. 3). ARE and LEE are the most closely related enzymes (68.0% identity), yet PED and ARE (45.6% identity) exhibited the most similar functional profiles—greater sequence similarity does not necessarily imply greater similarity in functional profiles, at least at these identity ranges.

Fig. 2. Substrate profiling of selected Class III PPK2 enzymes.

A Overview of the enzyme profiling assay. The product distribution was analyzed by HPLC after 60 min incubation at 37 °C using the standard protocol (see Supplementary Fig. 2A for an analysis of authentic samples). B Percent conversion to NTP for selected enzymes (18.9 µM; ARE, LEE, MAN, and PED are orange, blue, magenta, and green, respectively) reacting with the indicated nucleotide (4 mM) and PolyP chains of varying length (65 mM phosphate units) in the presence of 10 mM Mn²⁺. Refer to Supplementary Fig. 3 for product distributions. C Product distribution of MAN and CHU under the same conditions as in (B). NPn, NMP, NDP, and NTP are indicated in teal, blue, yellow, and red, respectively. This coloring scheme is used in all subsequent figures. The product distributions for MAN reactions are also provided in the Source Data file. Plotted values are the average of 4 independent runs. Error bars are standard deviations. Source data for this figure is available in the Source Data file.

While most enzymes failed to phosphorylate one or several nucleotides, only MAN catalyzed significant triphosphate production for all eight nucleotides and all three PolyP lengths (Fig. 2B). In reactions with PolyP-60, where MAN is the most efficient, 69–78% NTP conversion for NMPs, NDPs, purines, and pyrimidines was achieved. Moreover, an analysis of MAN-catalyzed NTP production with PolyP-60 (Fig. 2C) reveals a remarkably consistent product distribution for all common ribonucleotides, with only minor production of over-phosphorylated nucleotides (NPn). MAN could also catalyze phosphoryl transfer to inosine monophosphate, a key purine precursor, resulting in the synthesis of inosine triphosphate, although significant over-phosphorylation also occurred (Supplementary Fig. 4). For each common ribonucleotide, NDP was not generated from mixtures of NMP and NTP in the presence of MAN (Supplementary Fig. 5), suggesting that PolyP is the primary phosphoryl donor for NMP phosphorylation even in the presence of NTP.

Recently, CHU has emerged as a valuable resource for biotechnology applications, highlighting the potential utility of broad-specificity PPK2 enzymes for nucleotide regeneration^18,20,34. Noted for having a relatively high k_cat/K_M of 10³ s⁻¹M⁻¹ for GMP and GDP, CHU is an ideal point of comparison for MAN. We observe that under the same assay conditions, CHU produces significant amounts of GPn (guanine nucleotides with four or more phosphate groups) and has weak activity against CMP. These limitations are similar to those of the other enzymes tested: PED and ARE exhibit weak activity against CMP, GDP, and GMP, while LEE exhibits no activity against CMP and UMP, with at best weak activity against CDP and UDP. Based on these data, we designate MAN a uPPK2 enzyme—unique among the proteins characterized from the major N138 clade of Class III PPK2 enzymes.

Controllable production of ATP and APn

In addition to PolyP length, MAN activity may also be modulated by temperature^11,35,36, dication availability^11,19,35,37, and PolyP concentration^35,36,38 (Fig. 3). Optimal ATP synthesis was observed between 37 and 45 °C using AMP and PolyP-60 as substrates (Fig. 3A, left). At lower temperatures, the rate of the reaction is reduced, and the proportion of unreacted AMP increases. At 55 °C, however, the HPLC chromatogram is characterized by a broad, wavy tail at high retention times (Fig. 3A, right). Previously, this wavy pattern was identified as over-phosphorylated nucleotides by LC-MS/MS^23–25, indicating that the product distribution has shifted to over-phosphorylation. Careful inspection of the chromatogram reveals nucleotides with 30 or more attached phosphate groups. Time course data at 55 °C (Supplementary Fig. 6) reveals that ATP production occurs primarily within the first 10 min. At 30 min, APn production becomes detectable and by 45 min APn is the dominant species in solution. These data suggest that over-phosphorylation at higher temperatures is due to partial unfolding of the enzyme³⁹, and this interpretation is consistent with M. marinum (the parent organism of MAN) having an optimal growth temperature of 33 °C⁴⁰. To our knowledge, AP30 is the longest reported adenine nucleotide to date, with AP9 holding the previous record²³. The biological occurrence of adenine-capped PolyP molecules (such as AP10 and AP30) remains to be elucidated^23,25. Compared to AMP, only modest GPn formation occurs at 55 °C, and over-phosphorylation of pyrimidines is not observed (Supplementary Fig. 7).

Fig. 3. Effect of reaction conditions on MAN catalysis.

Effect of temperature (A), metal ion (B), and PolyP concentration (C) on MAN catalysis. Unless otherwise indicated, reactions were performed at 37 °C and contained 4 mM AMP, 65 mM phosphate units of PolyP-60, and 10 mM Mn²⁺. HPLC analysis performed using the standard protocol (see Supplementary Fig. 2A for an analysis of authentic samples). The product distribution for each condition is also provided in the Source Data file. Plotted values are the average of 4 independent runs. Error bars are standard deviations. Source data for this figure is available in the Source Data file.

MAN, like other PPK2 enzymes^19,37, can accept diverse divalent metal ions (Fig. 3B). The highest NTP conversions were observed with Mn²⁺ and Mg²⁺, which have roughly equivalent product distributions. Conversely, Ca²⁺, Co²⁺, and Ni²⁺ promote APn production, though not as much as increased temperature. PolyP concentration effects are also significant and PolyP-length dependent: Whereas high concentrations of PolyP-700 and PolyP-60 suppress reactivity, decreasing the concentration of PolyP-10 suppresses over-phosphorylation. Very short polyphosphates are significantly less reactive (Supplementary Fig. 8), and no product formation in the presence of diphosphate (20 mM phosphate units) was observed. Triphosphate and tetraphosphates yield approximately 12–20% conversion to ATP, 27–33% conversion to ADP, and no APn formation, with tetraphosphate being generally more active. These results, along with the ladder of increasing APn sizes observed at high temperature, each with decreasing concentration, suggest sequential phosphoryl transfers and not direct attachment of nucleotides to PolyP molecules. The cyclic polyphosphate trimetaphosphate⁴¹ is largely nonreactive, with only minor ( < 3%) production of ADP, perhaps from contaminating linear polyphosphates. Finally, hexametaphosphate, a common additive in the food and cosmeceutical industries that often contains related PolyP forms^42,43 yielded significant production of ATP with only modest (6%) over-phosphorylation. In short, the extent of over-phosphorylation by MAN can be controllably suppressed or exploited by various assay conditions– as detailed in Fig. 3, Supplementary Fig. 8, and the Source Data file– and is modest for the primary PolyP forms used in industry.

MAN kinetic characterization

Kinetic parameters were determined for MAN (Table 1, Supplementary Fig. 9) in the presence of Mg²⁺ at 30 °C. The dication Mg²⁺ was selected because it is a common component of IVT and translation systems. 30 °C was chosen to facilitate direct comparison with prior work on other Class III PPK2 enzymes¹⁹ and because this temperature is close to the optimal growth temperature of the parent organism. Binding of monophosphates is slightly preferred over diphosphates, except for cytosine nucleotides, where the trend is reversed. Turnover numbers are also consistently higher for monophosphates than for diphosphates, with a 16% (guanine) to 510% (cytosine) increase. Binding of purine monophosphates (K_M = 2.9–3.6 mM) is preferred over pyrimidine monophosphate (K_M = 7.2–38 mM) and turnover numbers show a clear preference for purines (k_cat = 42–98 s⁻¹) over pyrimidines (k_cat = 0.49–4.1 s⁻¹) as well. Guanine nucleotides are associated with higher errors than the other nucleotides (coefficient of variation ranging from 0.37-0.92 for GMP and GDP versus 0.09-0.24 for all others), perhaps due to the propensity for guanine nucleotides to assemble in solution^44,45. Consequently, apparent second-order rate constants (k_cat/K_M) are higher for purines than pyrimidines, and all nucleotides except CMP and CDP have a second-order rate constant greater than 10² s⁻¹M⁻¹. AMP has the highest catalytic efficiency of 3.3 × 10⁴ s⁻¹M⁻¹, perhaps reflecting the primary biochemical role of this enzyme. Catalytic efficiencies of 10³–10⁴ s⁻¹M⁻¹ for purine nucleotides are consistent with previous studies of Class III PPK2 enzymes, which range from 1.0 × 10² to 2.1 × 10⁵ s⁻¹M^{−1 19}. To our knowledge, a full characterization of pyrimidine catalytic efficiencies has not been reported. The kinetic characterization confirms activity against all substrates while also revealing clear nucleotide preferences.

Table 1.

MAN kinetic parameters

Substrate	KM (mM)a	kcat (s−1)	kcat/KM (s−1M−1)
AMP	2.9 ± 0.5	98.0 ± 9.8	3.3 × 104
ADP	3.0 ± 0.7	76.4 ± 8.1	2.5 × 104
GMP	3.6 ± 2.5	49 ± 18	1.4 × 104
GDP	13 ± 12	42 ± 25	3.2 × 103
UMP	7.2 ± 1.7	4.1 ± 0.4	5.7 × 102
UDP	12.5 ± 2.1	2.2 ± 0.2	1.7 × 102
CMP	38 ± 8	2.5 ± 0.4	6.6 × 10
CDP	6.6 ± 1.5	0.49 ± 0.06	7.4 × 10

^aStandard errors calculated from the fit to the Michaelis–Mention equation (Supplementary Fig. 9; see “Methods” for more details).

N138 contributes to utilization of all nucleotides

N138 is thought to facilitate binding and catalysis of both adenine and guanine nucleotides (Fig. 1A). The role of N138 in AlphaFold 3 (AF3) models of MAN with AMP (Supplementary Fig. 10A) are consistent with experimental data¹⁹. In the predicted MAN-GMP complex, however, the rotamer of N138 is flipped as expected but, surprisingly, so too is the ring of the guanine base (cf. Fig. 1A). Thus, an interaction is formed between the side-chain nitrogen amide of N138 and the amino group at position C2. In models of MAN with CMP and UMP (Supplementary Fig. 10A), N138 is predicted to be at least 4.3–4.9 Å from a suitable hydrogen bond donor or acceptor on the pyrimidine ring, suggesting modest involvement. Instead, the conserved R117 forms an interaction with the base. However, as no pyrimidine-bound Class III PPK2 structures have been reported to date, and the details of the guanidine nucleotide binding mode are not consistent with experimental data on another Class III PPK2¹⁹, it is unclear to what extent nucleotide binding modes can be predicted by AF3. Thus, we turned to mutational analysis to study the role of N138 in MAN (Fig. 4A, B), which revealed a fairly complex role in nucleotide utilization: At high enzyme concentrations and long incubation times (10 µM and 60 min, respectively; Supplementary Fig. 11A, B) mutation to Ala, Asp, or Glu resulted in only modest differences in NTP production, with N138 being the most active overall. To better assess differences in enzymatic rates, the enzyme concentration was reduced to 0.2 µM (AMP, ADP, GMP, and GDP) or 6 µM (UMP, UDP, CMP, and CDP) and incubation times were reduced to 10 min (CMP and CDP) or 5 min (all other nucleotides) (Fig. 4B and Supplementary Fig. 11C). Under these conditions, mutation of N138 resulted in a non-uniform decrease in NTP production for all nucleotides tested, with AMP being the most robust. Mutation to Ala and Glu preferentially retained GMP and CDP activity, respectively, while mutation to Asp was generally the most disruptive. In all cases, decreased NTP production was due to reduced activity and not over-phosphorylation. Given the proximity of N138 to the base (Figs. 1A and 4A) and the non-uniform loss of activity across nucleotides (Fig. 4B), we conclude that this position contributes to the broad substrate profile of MAN, as initially hypothesized, but is not the sole causative factor of broad substrate specificity.

Fig. 4. Mutation analysis of MAN and CHU.

A Overlay of the base binding pockets of MAN (AlphaFold3 structure AF-A0A2T5C445-F1-v4)⁷⁷ and CHU (PDB ID 6AN9)¹⁹. Conserved residues other than H118, R117, and N138 that mediate binding are not drawn. Sequence differences in the primary base binding pocket are drawn as cyan and magenta sticks for CHU and MAN, respectively. Sequence differences involving residues that point away from the base binding pocket and were not characterized further are shown as yellow sticks. AF3 models of MAN bound to all four nucleotides are presented in Supplementary Fig. 10. Additional sequence information can be found in Fig. 1C. B Characterization of position 138. Note that activities are normalized to MAN activity, indicated by the fuchsia outer ring. C Characterization of base binding site chimeras. Refer to the main text for more details. D Overlay of the polyphosphate binding track, following the coloring conventions in (A). E Characterization of polyphosphate binding site chimera. C and E report the percentage of NTP production. Plotted values are the average of 4 independent runs. Error bars are standard deviations. Source data for this figure is available in the Source Data file.

Truncation of the C-terminal α-helix promotes purine utilization and suppresses pyrimidine utilization

The active site of PPK2 enzymes is centered on the Walker A motif, which adopts the canonical P-loop conformation^19,46–48. As with thymidylate kinases⁴⁹, an α-helix hairpin hood rests atop the P-loop. Upon oligomerization, the C-terminal α-helix of one protomer interacts with the hood of an adjacent protomer (Supplementary Fig. 12A), potentially influencing the catalytic properties of the enzyme. Previously, truncating the C-terminal α-helix of CHU resulted in improved purine utilization¹⁹, a result reproduced here (Supplementary Fig. 12B). Truncation of the C-terminal α-helix in LEE and PED produced similar results (Supplementary Fig. 12B). The MAN and ARE truncation variants failed to express and could not be characterized. These data suggest that truncation is a general strategy for designing purine-preferring Class III PPK2 enzymes (in cases where the resulting protein is well behaved) but at the expense of pyrimidine utilization. Class III PPK2 enzymes that naturally lack the C-terminal α-helix (e.g., UniProt accession number A0A1M7MEJ7), although uncommon, may represent the natural exploration of this engineering approach. The exact mechanism of truncation is unclear, but it may relate to changes in the oligomerization state affecting conformational dynamics. Indeed, size exclusion chromatography (SEC) detected a change in oligomerization state upon truncation of the C-terminal α-helix, though the estimated masses indicated a hexamer to trimer transition (Supplementary Table 2). As crystal structures suggest a tetramer (e.g., PDB ID 5LC9) with two distinct dimer interfaces and our constructs have an N-terminal MBP tag that may result in an elongated shape, we suspected that the masses calculated by SEC, though not the existence of an oligomerization state change, may be artefactual. Analytical ultracentrifugation (AUC) (Supplementary Fig. 13) of full-length and truncated variants of CHU, LEE, and PED demonstrated a transition from a dimer to a monomer upon truncation in each case (Supplementary Table 2), consistent with our expectation that the populated states should be monomers, dimers, or tetramers.

Base binding preferences are encoded in the base binding pocket and beyond

What residues underlie the broad specificity of PPK2 enzymes, and can we explain the functional differences between MAN and CHU? In the region of the base binding pocket, just four residues (positions 104, 141, 148, and 151; CHU numbering) differ between MAN and CHU (Fig. 4A and Supplementary Fig. 10B, cyan and magenta sticks; Fig. 1C, orange highlight). Although these residues do not make direct interactions with the base in the most common adenine binding mode (represented in Fig. 4), an alternative binding mode found in Class I PPK2 (Supplementary Fig. 14A, B) has the base flipped up⁴⁷, potentially within interaction distance of these residues. Other differences between MAN and CHU (Fig. 4A, yellow sticks) are more distant and have their side chains oriented away from the base binding pocket. To better understand the determinants of substrate specificity, these four positions in the MAN binding site were mutated to the corresponding CHU residue and vice versa. MAN→CHU (base), where the parent enzyme is MAN, shows a contraction of the substrate profile, with activity against CMP becoming CHU-like and activity against guanine nucleotides largely unchanged. On the other hand, CHU→MAN (base) (Fig. 4C and Supplementary Fig. 14C), where the parent enzyme is CHU, shows improved activity against GMP and GDP but reduced activity against pyrimidines (Fig. 4C and Supplementary Fig. 14C). Neither of the chimeric constructs resulted in a complete conversion of base binding preferences.

The partial conversion of MAN and CHU activities by binding site chimeras suggests that base binding preferences are determined in part by non-local interactions, for example, through long-range dynamics (consistent with the truncation analysis) or the positioning of the polyphosphate substrate. To test the latter possibility, residues along the path of the polyphosphate chain (as inferred from PDB ID 5LLF; Fig. 4D) in CHU were mutated to be MAN-like. The resulting chimera, CHU→MAN (PolyP), was more active against GMP and GDP but lost activity against the pyrimidines UMP, CDP, and CMP (Fig. 4E and Supplementary Fig. 14C). The MAN→CHU (PolyP) construct was degraded in Escherichia coli and could not be purified. Taken together, these data suggest that the substrate preferences of Class III PPK2 enzymes—particularly for nucleotides other than ATP—are encoded by residues distributed across the protein, and not simply near the base binding pocket.

Continuous purine nucleotide regeneration by MAN

Previously, PPK2 was used for nucleotide recycling^16,20,30,31; specifically, for the regeneration of ATP and GTP from ADP/AMP and GDP, respectively. To test whether MAN can also catalyze the simultaneous regeneration of ATP and GTP, we employed a reconstituted in vitro translation system⁵⁰. As shown in Supplementary Fig. 15A, B, MAN successfully catalyzes regeneration as indicated by the production of GFP. As MAN can produce APn and GPn under certain conditions, which may act as inhibitors, the reaction mixture was analyzed after 120 min of continuous recycling to see if APn or GPn was present. MAN did not catalyze the production of significant amounts of APn or GPn under these conditions (Supplementary Fig. 15C).

uPPK2-powered in vitro transcription

With the ability to phosphorylate all four ribonucleotides, MAN represents a significant new opportunity for the biotechnological application of PPK2 enzymes. To underscore this point, we established an IVT system powered by MAN-derived NTPs. Given the differences in catalytic efficiency between purine monophosphates and pyrimidine monophosphates (Table 1), two reaction schemes were considered: A one-step reaction in which all NMPs were added simultaneously and incubated for 90 min, and a two-step reaction in which pyrimidines were incubated first for 60 min and then purines were added and the mixture was incubated for an additional 30 min (Fig. 5A). Notably, both reaction schemes can be performed in a single vessel (one pot). The PPK2-synthesized NTPs were then used to drive IVT of a Pepper-containing construct (188 nucleotides) that forms a fluorescence complex upon binding the substrate HBC530⁵¹, or GFP mRNA (971 nucleotides), the fidelity of which was confirmed by in vitro translation into active GFP (described below).

Fig. 5. In vitro transcription utilizing NTPs produced by Class III PPK2 enzymes.

A NTP production and in vitro transcription reaction schemes. This panel was created in BioRender. Matsuura, T. (2025) https://BioRender.com/m88d406. B HPLC chromatograms for one-step and two-step NTP production catalyzed by MAN (left). HPLC analysis performed using the enhanced protocol (see Supplementary Fig. 2B, C for an analysis of authentic samples). Time course of Pepper RNA production during in vitro transcription with various Class III PPK2 enzymes (right). Solid lines are the average of 3 independent runs, dashed lines are the individual runs, and shaded bands are standard deviations. For GFP mRNA transcription, see Supplementary Fig. 17. Source data for this figure is available in the Source Data file.

The product distributions of the one-step and two-step reactions showed significant differences, particularly with respect to cytosine nucleotides (Fig. 5B). In the one-step reaction, significant amounts of the CMP substrates remained and only a weak CTP peak was observed, consistent with this base having the lowest apparent second-order rate constant. Consequently, IVT reactions performed with this sample yielded modest transcription. The two-step reaction mixture, on the other hand, contained significant amounts of all four NTPs and the subsequent IVT produced nearly as much RNA (90.5% yield), nearly as quickly, as the NTP positive control, with both reactions reaching completion at ~4 h. As significant amounts of unreacted NTPs remain in both the MAN synthesized-NTP and authentic-NTP reaction mixtures, even after the reaction plateaus (Supplementary Fig. 16), the endpoint of the reaction appears to relate to a loss of T7 polymerase activity. Two-step reactions using CHU or truncated CHU (ΔC-CHU) failed to produce a detectable amount of RNA, even at the level of the one-step MAN reaction, highlighting the uniquely broad substrate specificity of the MAN enzyme. Finally, to further confirm the fidelity of MAN-driven IVT, GFP mRNA was synthesized, purified, and then translated in vitro. Active GFP production was virtually identical to an authentic nucleotide control (Supplementary Fig. 17).

Discussion

We demonstrate that MAN is a universal PPK capable of efficiently phosphorylating all eight common ribonucleotides—purines and pyrimidines, NMPs and NDPs—with PolyP. However, our analysis suggests that PPK2 enzymes in general (and MAN in particular) have a relatively rugged functional landscape, particularly for non-adenine bases. This conclusion follows from the diverse properties of N138 Class III PPK2 enzymes (Fig. 2) and the idiosyncratic, non-uniform changes in substrate preference upon mutation (Fig. 4B and Supplementary Fig. 11) and chimerization (Fig. 4C and E). The general sensitivity of these enzymes to reaction conditions (Fig. 3) may also be an indication of a rugged functional landscape. Although the source of this complex behavior is unknown, several lines of evidence suggest that conformational dynamics and/or stability may play a role. First, the C-terminal α-helix does not directly interact with the base, yet truncation of this α-helix consistently shifts substrate specificity to purines (Supplementary Fig. 12). Second, mutation of the PolyP binding site also alters substrate preferences (Fig. 4E), again without directly interacting with the base of the nucleotide. Finally, the chimera constructs (Fig. 4C and Supplementary Fig. 14) cannot be compactly described as transitions between MAN and CHU functional profiles, despite substituting all non-identical residues near the base-binding site. The coupling of conformational dynamics with function may also contribute to the abrupt transition from ATP formation to APn formation with increasing temperature.

We have previously shown that the number of PPK2 genes within a single organism can vary widely, with some organisms having as many as seven PPK2 genes^28,29. The fact that MAN is the only PPK2 identified in the proteome of Mangrovibacterium marinum, given its notably broad substrate specificity, may suggest a kind of functional consolidation. That said, the substrate breadth of MAN in vivo is unknown, and few reports have addressed the role specialization of PPK2 in organisms with multiple PPK2 genes.

The potential biotechnological applications of uPPK2 enzymes are exciting: For example, we envision that MAN or enzymes like it, with the ability to phosphorylate diverse nucleotides, will reduce the cost of NTP production, and thus the cost of RNA production as well. The one-pot nucleotide synthesis and IVT reaction scheme presented here marks a significant step towards this goal. Reagent costs and synthesis pipelines for the large-scale production of nucleotides and nucleic acids are considered proprietary information and are not publicly disclosed. However, a price comparison between NTPs ($1095/mmol, sum of 1 mmol of ATP, GTP, UTP, and CTP) and NMPs ($78/mmol, sum of 1 mmol of AMP, GMP, UMP, and CMP) confirms that the phosphorylation step is a key contributor to NTP production costs (Supplementary Table 3). This price difference is likely because NMPs can be readily produced, for example, by simply hydrolyzing biologically derived RNA⁵². Subsequent NMP phosphorylation, if performed enzymatically, will be different if using cell-pathway inspired reaction schemes (according to KEGG database⁵³) or a uPPK2 (Supplementary Fig. 18 and Supplementary Table 3). With cellular reactions, the synthesis of one mol unit of NTPs from NMPs needs 6 enzymes and 8 mol unit of creatine phosphates ($296/8 mmol). On the other hand, uPPK2-mediated synthesis of one mol unit of NTP requires one enzyme and 8 mol unit of polyphosphate ($0.128/8 mmol). Moreover, while the cellular-inspired reaction process requires at least 6 pots, the uPPK2-based process requires only one pot, further reducing production costs. In summary, NMP to NTP synthesis by MAN is far cheaper than that by cell-pathway inspired reaction.

The broad specificity of MAN also makes it a potential starting point for the development of diverse high-specificity enzymes, echoing the natural process of sub-functionalization^54,55 after gene duplication and the role of promiscuity in enabling functional exploration⁵⁶. As such, the application of MAN or related enzymes may emerge as facile approaches to phosphorylate diverse nucleotides, including engineered nucleotide pharmaceuticals. The demonstration that inosine monophosphate, a key precursor in purine biosynthesis, can also be phosphorylated (Supplementary Figure 4) provides tentative support for this potential application. Deoxynucleotide phosphorylation for cost-effective DNA production is yet another possible application of MAN, and such studies are currently underway. Finally, while we are aware of no reconstituted in vitro metabolic pathway that requires the continuous regeneration of all 4 NTPs, optimized MAN-like variants with tuned catalytic efficiencies for each nucleotide may open up possibilities for the design of reaction systems such as this.

Yčas²⁶ and Jenson²⁷ envisioned promiscuity as an organizing principle for primitive enzyme-mediated metabolism because diverse substrates could be acted on by a limited repertoire of enzymes. These insights have been borne out in part by studies of ancestral proteins, some of which have been shown to be generalists with broad substrate profiles^57,58. While ancestral enzyme studies demonstrate that generalists are an achievable functional state, supporting their potential role in primitive metabolism, these enzymes are rarely integrated into larger reaction systems that depend on their generalist properties. Similarly, Kornberg hypothesized that polyphosphate was a primitive store of phosphorylating power due to its simple structure and synthesis. The demonstration that PolyP can be used as a regeneration system for in vitro translation provides additional support for key primitive roles. Yet, as noted above, direct phosphorylation of all common ribonucleotides would allow further simplification of NTP-utilizing reaction systems, and suggest how such systems could have emerged with fewer components. Finally, Tawfik and others have argued for the primordial origin of the P-Loop NTPase family^9,59. A recent model of the joint evolution of enzymes and metabolism supports this view⁶⁰, as do experiments on simplified P-Loop NTPase-derived peptides, which can bind to polyphosphates and nucleotides⁶¹, exhibit helicase-like activity⁶², and weakly catalyze phosphoryl transfer⁶³. Here, we combine the themes of Yčas, Jenson, Kornberg, and Tawfik by showing that a single promiscuous P-Loop NTPase enzyme can mediate the production of diverse NTPs from polyphosphate to drive a core cellular process.

In summary, a uPPK2 enzyme capable of efficiently phosphorylating the common ribonucleotides was identified and characterized. To demonstrate the biotechnological utility of uPPK2 enzymes, a one-pot IVT assay powered entirely by uPPK2-synthesized nucleotides was developed. Our results highlight the value of primitive biology perspectives—such as the roles of enzyme promiscuity and PolyP—in the development of increasingly demanding reconstituted cell-free reaction systems¹⁵.

Methods

Sequence analysis and gene tree construction

Full-length PPK2 gene sequences were retrieved from the InterPro database⁶⁴ entry IPR022488 (Polyphosphate kinase-2-related) on 28 August 2022. Clustering and representative sequence selection were performed using CD-HIT version 4.8.1⁶⁵ with a 60% identity cutoff and a word size of 4. The resulting 2118 representative sequences were aligned using the L-INS-i algorithm implemented in MAFFT version 7.525⁶⁶. To improve phylogenetic tree building, gaps and regions of poor alignment were removed using trimAL version 1.4⁶⁷ using the gappyout option. Alignment quality was assessed visually in Jalview version 2.11.4⁶⁸. A phylogenetic tree was constructed using IQ-TREE version 2.3.6⁶⁹ with automatic model selection and a single thread, as multiple threads degrade performance⁷⁰. The phylogenetic tree was analyzed using the Python package Environment for Tree Exploration version 4⁷¹ and visualized using the Interactive Tree of Life version 7.0⁷². PPK2 clades were assigned using the clade-specific residue at position 137^11,19, and named according to established conventions.

Cloning and mutagenesis

Class III PPK2 genes from Arenitalea lutea (ARE, A0A1M6FUJ9), Leeuwenhoekiella palythoae (LEE, A0A1M5Z1E8), Mangrovibacterium marinum (MAN, A0A2T5C445), and Pedobacter sp. (PED, A0A519X5M4), and their truncated C-terminal α-helix variants were synthesized (Eurofins Scientific) and cloned into the pQI-MBP plasmid. pQI-MBP, constructed in our group, is a derivative of the pQE30 plasmid (Qiagen) in which MBP is cloned downstream of the T5 promoter and also encodes the LacIq sequence. The full plasmid sequence is given in Supplementary Data 1. The previously characterized PPK2 gene from Cytophaga hutchinsonii (CHU, A0A6N4SMB5)¹⁹ and its truncated C-terminal α-helix variant were also synthesized. Each construct contains an N-terminal 6xHis tag for purification followed by a maltose-binding protein (MBP) tag to improve expression and solubility. For the preparation of MAN mutants, site-directed mutagenesis was performed using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs). Chimeras were prepared using the HiFi DNA Assembly Kit (New England Biolabs) according to the manufacturer’s instructions. All protein sequences are provided in Supplementary Table 1.

Expression and purification

PPK2 proteins were produced by heterologous expression in E. coli XL10-Gold (Agilent). Briefly, 1 L cultures of terrific broth (TB) were shake-incubated at 37 °C until exponential growth and an optical density at 600 nm between 0.4 and 0.8 AU. At this point, the temperature was decreased to 25 °C and isopropyl-β-D-thiogalactopyranoside was added to a final concentration of 0.5 mM. Expression proceeded overnight for approximately 16 h. Cells were harvested by centrifugation at 6000 × g for 10 min at 4 °C. Cell pellets were stored at −80 °C until lysis. Cells were resuspended in pre-chilled 50 mM HEPES-KOH pH 7.6, 300 mM NaCl, 0.01 mM DTT, and 10% glycerol, re-pelleted as above, and resuspended once again. Cells were lysed by sonication on ice, and the lysate was clarified by centrifugation at 20,000 × g for 30 min at 4 °C. Clarified lysate was incubated with TALON metal affinity resin (Takara Bio Inc.) for 30 min at 4 °C with gentle mixing. The resins were rinsed with 10 column volumes of 50 mM HEPES-KOH pH 7.6, 300 mM NaCl, 0.01 mM DTT, 15% glycerol, and 10 mM imidazole and PPK2 was eluted from the beads in 50 mM HEPES-KOH pH 7.6, 300 mM NaCl, 0.01 mM DTT, 15% glycerol, and 300 mM imidazole. Purified protein was passed through a PD-10 desalting column (Cytiva) pre-equilibrated with 50 mM HEPES-KOH pH 7.6, 100 mM KCl, 10 mM MgCl₂, and 30% glycerol. Protein purity was assessed by SDS-PAGE using Coomassie Brilliant Blue stain (Supplementary Fig. 19. Protein concentrations were measured using absorbance at 280 nm and calculated with extinction coefficients (ARE: 113220 M⁻¹ cm⁻¹, LEE: 111730 M⁻¹ cm⁻¹, MAN: 117230 M⁻¹ cm⁻¹, and PED: 121700 M⁻¹ cm⁻¹) estimated by Benchling (https://www.benchling.com/) based on the protein sequences. Purified protein samples were stored at −80 °C.

Polyphosphate kinase activity assay

Polyphosphate kinase activity profiling against NMPs and NDPs was performed in 50 mM MOPS-NaOH pH 7.0, 10 mM MnCl₂ at 37 °C. Reaction mixtures contained 65 mM phosphate units, 4 mM nucleotide, and 18.9 µM PPK2, and had a final volume of 20 µL. Six polyphosphate forms were tested: diphosphate (2 phosphate units; Sigma-Aldrich), triphosphate (3 units; Sigma-Aldrich), tetraphosphate (4 units; Sigma-Aldrich), PolyP-10 (12.5 units; Bioenex, Japan), PolyP-60 (65 units; Bioenex, Japan), PolyP-700 (750 units; Bioenex, Japan), trimetaphosphate (a cyclic triphosphate; Sigma-Aldrich), and hexametaphosphate (a mixture of polyphosphate forms used in the food and cosmetic industries, Sigma-Aldrich). After 60 min, the samples were put on ice and the reactions were terminated by the addition of trichloroacetic acid (TCA) to a final concentration of 2.5%. The samples were then centrifuged at 20,000 × g for 30 min at 4 °C, which removed the protein component. The supernatant was analyzed by high-performance liquid chromatography (HPLC) as described below.

MAN temperature profiling was performed as above except that only AMP and PolyP-60 were tested, and the reaction temperature ranged from 20 to 55 °C. MAN metal profiling was performed as above except that only AMP and PolyP-60 were tested and 10 mM MgCl₂, CaCl₂, CoCl₂, or NiCl₂ were used in place of MnCl₂. For chimeras, reaction mixtures contained 50 mM MOPS-NaOH pH 7.0, 10 mM MnCl₂, 50 mM phosphate units of PolyP-60, 4 mM nucleotide, and 18.9 µM enzyme. For point mutants at position 138 (CHU numbering), reaction mixtures contained 50 mM Tris-HCl pH 8.0, 10 mM MgCl₂, 20 mM phosphate units from PolyP-10, 4 mM nucleotide, and either 0.2 µM (AMP, ADP, GMP, and GDP) or 6 µM (UMP, UDP, CMP, and CDP) enzyme. Reactions were conducted at 30 °C for either 5 min (AMP, ADP, GMP, GDP, UMP, and UDP) or 10 min (CMP and CDP).

High-performance liquid chromatography (HPLC)

The supernatants from TCA precipitation were diluted in two volumes of 50 mM triethylammonium acetate (an ion pairing agent), 2 mM ethylenediaminetetraacetic acid, pH 7.0 (Buffer A). Samples containing only one nucleotide base type (e.g., only adenine nucleotides) were analyzed using the standard protocol. Samples were injected onto an XBridge BEH C18 column (2.5 µm particles, 4.6 × 30 mm, Waters Corporation) for analysis. Gradient elution was performed at 20 °C with a flow rate of 1 mL/min. Buffer A was the starting buffer and 100% acetonitrile served as the elution buffer (Buffer B). Buffer B was increased linearly from 0% to 2.5% over 2 min, and then further increased to 5% over the next 4 min. Absorbance at 260 nm was recorded. Samples containing multiple nucleotide base types (Fig. 5B), where improved resolution is required to distinguish between similar species, were analyzed using the enhanced protocol, in which two XBridge BEH C18 columns were connected in series. Gradient elution was performed at 23 °C with a flow rate of 1 mL/min. Buffer A, as described above, supplemented with 0.6% acetonitrile, was used as the starting buffer and 100% acetonitrile served as the elution buffer (Buffer B). Buffer B was held at 1% for the first 3 min, increased to 5.5% over the next 5 min, and maintained at 5.5% for 2 min. Nucleotide concentrations were determined from the peak areas of the chromatogram using LabSolutions version 5.124 (Shimadzu). Representative HPLC runs of authentic nucleotides using both protocols are presented in Supplementary Fig. 2. Peak areas were converted into nucleotide concentrations according to a standard curve obtained with authentic nucleotides. Nucleotide extinction coefficients were obtained from Promega and NEB chemical data sheets.

Size exclusion chromatography (SEC)

SEC was performed on a Superdex 200 Increase 5/150 GL column (Cytiva) in 50 mM sodium phosphate pH 7.0, 300 mM NaCl, 0.005% (v/v) Tween-20, and 10% (v/v) glycerol (Buffer A). MBP-tagged wild-type enzymes (CHU, LEE, and PED) and their corresponding MBP-tagged C-terminal α-helix truncated variants were diluted in Buffer A to a final concentration of 0.25 mg/mL, and 10 µL of each sample was injected per run. Chromatographic runs were conducted at a flow rate of 0.15 mL/min, and elution profiles were monitored at 280 nm. Molecular weight estimation was performed using a gel filtration standard (Bio-Rad Gel Filtration Standard, #1511901) containing thyroglobulin (670 kDa), γ-globulin (158 kDa), ovalbumin (44 kDa), myoglobin (17 kDa), and vitamin B12 (1.35 kDa) to generate a calibration curve. The molecular weights of enzyme samples were estimated by interpolating their retention times against the calibration curve derived from the gel filtration standards.

Analytical Ultracentrifugation (AUC)

Ultracentrifugation experiments were performed using a ProteoLab XLI analytical ultracentrifuge equipped with an AN-50 Ti rotor (Beckman Coulter). Protein samples with A₂₈₀ ≅ 1 were prepared in 50 mM HEPES-KOH pH 7.6, 500 mM KCl, 10 mM MgCl₂, and 0.005% Tween 20. 400 µL of each sample and reference buffer were injected into a 12 mm sample cell with a sapphire window. Velocity measurements were conducted at 50,000 rpm and 20 °C. For each sample, 100 scans were collected over the course of 6–12 h, depending on the number of samples being measured simultaneously. The density and viscosity of the buffer were calculated by SEDNTERP version 3.0.4⁷³. Time course data was analyzed by SEDFIT version 16.50⁷⁴ using a resolution of 100, a partial specific volume of 0.73, a frictional ratio of 1.2, and a confidence level of the F-ratio (f/fo) of 0.95.

Steady-state kinetic analysis of MAN

Kinetic analyses of MAN were performed under initial velocity conditions in 50 mM Tris-HCl pH 8.0, containing 10 mM MgCl₂ and 20 mM phosphate units from PolyP-60 at 30 °C. For each nucleotide substrate, a series of substrate concentrations were tested with a fixed MAN concentration as follows: AMP (0.89, 2.67, 6.23, 9.78, and 19.8 mM; MAN 0.42 µM), ADP (0.89, 2.67, 6.23, 9.78, and 19.8 mM; MAN 2.67 µM), GMP (0.89, 2.67, 6.23, 9.78, and 19.8 mM; MAN 0.85 µM), GDP (0.89, 2.67, 6.23, 9.78, and 19.8 mM; MAN 1.28 µM), UMP (2.67, 7.55, 15.1, 30.2, and 40 mM; MAN 0.94 µM), UDP (2.67, 7.55, 15.1, 30.2, and 40 mM; MAN 3.84 µM), CMP (2.67, 7.55, 15.1, 30.2, and 40 mM; MAN 7.69 µM), and CDP (2.67, 7.55, 15.1, and 30.2 mM; MAN 7.69 µM). Reaction mixtures (total volume 5.5 µL) were prepared using an Echo 525 acoustic liquid handler (Beckman Coulter) by combining concentrated stock solutions (1 M Tris-HCl, pH 8.0; 440 mM MgCl₂; 275 mM PolyP-60; 97.8 mM nucleotide; and 188 µM MAN) dispensed from 384-well polypropylene source plates into 384-well assay plates, followed by dilution with Milli-Q water to the final volume. Reactions were incubated at 30 °C and, for each time point, three independent reaction wells were prepared. Separate plates were prepared for each time point: 0, 1, 2, and 5 min for AMP, ADP, GMP, and GDP, and 0, 5, 10, and 15 min for UMP, UDP, CMP, and CDP. At the designated time, each plate was quenched by adding 16.5 µL of 3.3% (w/v) TCA, followed by centrifugation at 6000 × g for 30 min. The resulting supernatants were analyzed by HPLC.

Initial velocities were obtained from the linear range of substrate depletion; for example, when AMP was the substrate, initial rates were determined from the decrease in the AMP peak even in the presence of other products (ADP, ATP, or polyphosphate adducts). For each substrate concentration, the initial velocity was taken as the slope of a linear regression directly on the unaveraged time course series (N = 3), and the standard error of the slope was used as the error bar. Kinetic parameters (k_cat and K_M) were then determined by fitting the velocity as a function of substrate concentration to the Michaelis–Menten equation, v = V_max[S]/(K_M + [S]), where Vmax = k_cat[E]₀. Nonlinear regression was performed in Python (SciPy v1.13.1), and the standard errors of Vmax and K_M were obtained from the covariance matrix of the fit, which incorporates the individual velocity errors as weights.

AlphaFold3 (AF3) prediction of nucleotide binding modes

The binding of MAN to AMP, GMP, UMP, and CMP was predicted using Protenix (https://protenix-server.com/login), an open-source PyTorch implementation of AF3. Both the amino acid sequence of MAN and the substrates, represented in SMILES notation, were provided as inputs to the model. For the calculations, multiple sequence alignment (MSA) was enabled (use_msa = true). All predictions were carried out using the same random seed (84505).

Continuous ATP and GTP regeneration by MAN

Continuous ATP and GTP regeneration by MAN reactions was performed using an E. coli-based reconstituted in vitro translation system⁵⁰, PUREfrex 2.0 (GeneFrontier), with the modifications indicated below. Here, we used PURfrex2.0 to follow the protocol used in the previous study²⁰, where the ATP/GTP generation was performed using one of the Class III PPK2, CHU. Each reaction contained 0.1 mM ATP, 1 mM GTP, 1 µg GFP mRNA, and Alexa Fluor 647 for fluorescence monitoring, and was incubated at 37 °C for 0, 30, 60, or 120 min. Three conditions were tested: (i) a positive control containing 60 mM creatine phosphate and ATP/GTP recycling enzymes (adenylate kinase, creatine kinase, and nucleoside diphosphate kinase) provided at the default concentrations in PUREfrex2.0, (ii) a negative control lacking both creatine phosphate and the three recycling enzymes, and (iii) a test sample lacking creatine phosphate and the three recycling enzymes but supplemented with 50 mM PolyP-60 and 1 µM MBP-MAN. After incubation, reactions were quenched on ice by adding 30% (w/v) TCA and then diluted threefold with 50 mM triethylammonium acetate (pH 7.0) containing 2 mM EDTA. Samples were centrifuged at 15,000 × g for 10 min at 4 °C, and the resulting supernatants were collected. ATP and GTP quantification was performed following the procedure described in the HPLC section.

Class III PPK2-powered in vitro transcription

IVT was monitored by fluorescence of the Pepper RNA aptamer (Supplementary Table 4) upon binding to benzylidene-cyanophenyl (HBC) derivatives. The in vitro reaction mixture was composed of 40 mM Tris-HCl buffer pH 8.0, 50 mM NaCl, 8 mM MgCl₂, 5 mM dithiothreitol, 0.25 mM spermidine, 0.001% bovine serum albumin, 62 ng of Pepper DNA, 0.04 µM pyrophosphatase (derived from yeast⁷⁵ and purified in-house), 10 µM HBC530 (which binds to the Pepper RNA aptamer), 0.8 units of T7 RNA polymerase from the ScriptMAX Thermo T7 Transcription Kit (Toyobo), and nucleotides prepared as described below. The positive control contained 0.125 mM of each NTP. The test samples contained 0.125 mM of each nucleotide type (AXP, GXP, CXP, and UXP) derived from a PPK2-catalyzed one-pot reaction using NMPs as the initial substrates. Only if the PPK2 phosphorylation reaction is efficient will the nucleotides be predominantly triphosphates. The PPK2 reaction mixture contained 18.9 µM PPK2 enzyme in 50 mM Tris-HCl, 10 mM MgCl₂, 30 mM PolyP-10, and 1 mM NMPs (AMP, GMP, CMP, and UMP). PolyP-10, which is one of the least expensive PolyP forms, was selected for the IVT assay because it yielded the best IVT performance with MAN-driven NTP production (Supplementary Fig. 20). In the one-step reaction, all NMPs were added and incubated for 90 min at 37 °C. In the two-step reaction, CMP and UMP were added and the sample was incubated for 60 min. Then AMP and GMP were added, and the sample was incubated for an additional 30 min. The PPK2 reaction mixtures were used as a source of nucleotides for the IVT. IVT was performed at 37 °C for 20 h in an Mx3005P real-time PCR thermocycler (Agilent). Pepper RNA production was tracked by measuring fluorescence with excitation and emission wavelengths of 492 nm and 516 nm, respectively.

In vitro translation of GFP mRNA produced by MAN-powered IVT

GFP mRNA was synthesized using 335 ng of GFP DNA as a template under the conditions described above for Class III PPK2-powered IVT. The transcription reaction was carried out in four 20 µL samples and incubated at 37 °C overnight. After the reaction, template DNA was removed by treatment with 4 U of DNase I (TaKaRa, 5 U/µL). The reaction mixture was then adjusted to 100 µL with RNase-free water, and total RNA was purified using the RNeasy Mini Kit (QIAGEN). Using this procedure, RNA synthesized from NTPs generated by MAN yielded 111.5 ng/µL, while RNA from the positive control NTPs yielded 179.2 ng/µL. The purified GFP mRNA (final concentration 100 nM) was then used as the template for in vitro translation by PUREfrex 1.0 (GeneFrontier) supplemented with 0.04 µM Alexa Fluor 647 and an RNase inhibitor (0.4 U/µL). Reactions were performed in a real-time PCR instrument (Mx3005P, Stratagene/Agilent) at 37 °C for 20 h. Here, we used PUREfrex 1.0 and not PUREfrex 2.0 as the former is more affordable.

Statistics and reproducibility

Means, standard deviations, and model fits were calculated using SciPy version 1.13.1. No data were excluded from the analyses. No statistical method was used to predetermine sample size. At least three replicates of biochemical assays were performed to assess reproducibility, in keeping with standard practice.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.M., T.W., E.Y., and A.K. performed the experiments, T.W. developed the analysis methods, R.M., L.M.L., and T.M. analyzed the data, R.M., L.M.L., and T.M. conceived the project, and L.M.L., T.M. led the project. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Zonglin Li, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data generated in this study are provided in the Source Data file. The GenBank ID for MAN is PTN09595.1. Source data are provided with this paper.

Competing interests

R.M., T.W., L.M.L., and T.M. are inventors on a Japanese patent application related to this study (Japanese Patent Application No. 2023-182220, filed on October 24, 2023). The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-68012-9.