The inorganic pyrophosphatases of microorganisms: a structural and functional review

PPase diversity

To date, two main c-PPase families have been identified: family I and family II. In contrast, family III is less studied, has an enzymatic mechanism that involves the binding of magnesium-linked PPi, requires four kinetic constants for the binding and nucleophilic attack of the substrate, and involves haloalkane dehalogenases (see below) (Baykov et al., 1999). Using bioinformatic tools, we show in Fig. 3 examples of the sequence diversity in family I and II representatives. A third family has been proposed, but little information is available on these enzymes (see below). In Fig. 3A, the sequence diversity reflects that the enzymes have incorporated additional sequences, but the main catalytic residues remain the same. Family II, for instance, has a conserved SRKKQ signature at the C-terminal end. However, both families have conserved estimated physicochemical features in most full-length sequences (Fig. 3A). Curiously, Giardia intestinalis encodes a family II protein with regulatory domains that have been identified in bacteria, such as Clostridium perfringens (Tuominen et al., 2010), and has been proposed as a target for inhibition by AMP or ADP or activation by ATP or diadenosine polyphosphates (Anashkin et al., 2015), suggesting that the regulatory role of the CBS domain links PPi hydrolysis with the capacity for biosynthesis in bacteria. Family II proteins tend to cluster together even with extra regulatory domains. At the same time, family I seems to diverge more between organisms distant from bacteria and early divergent protists (Figs. 4B and 4C).

As shown in Fig. 3, in a bioinformatic comparison of a sample of cytoplasmic PPases from different organisms, families I and II displayed low sequence and structural homology. Although the overall catalytic mechanism is similar, regulatory differences have been found (Baykov et al., 2017; Sarmina, Peña Segura & Celis, 2017). Figure 3 shows that the regions conserved between the two families of cytoplasmic PPases are limited. The regions conserved between family I and family II members differed strongly, except that the estimated physicochemical properties were similar in most regions of the protein sequence; this can be seen more clearly when comparing a 2D pairwise alignment map. Family II PPases cluster together, and two clear clusters of family I proteins can be identified (bioinformatic analysis is shown in Fig. 3B; UMPA analysis is shown in Fig. 3C).

In Fig. 4, the phylogenetic distribution analyzed for this review of PPases seems to cluster in terms of their source. Bacterial family I PPases are more distant from their eukaryotic counterparts, except for Chlamydia trachomatis PPase (UniProt B0B8Z8), which is located as a common ancestor for two branches of bacterial enzymes and a branch of protists (Chlamydomonas reinhardtii and Trichomonas vaginalis). The other branch that clusters together is for amoebozoan and fungal PPases. The distribution of invertebrate and vertebrate enzymes is consistent with the clustering shown in Figs. 3B and 3C. Family II enzymes are clustered together, and surprisingly, G. an intestinalis PPase is a family II enzyme (UniProt V6TM56) that contains a regulatory domain (see below).

The family I PPases have strong similarities. Figure 5A shows that the extensively studied yeast PPase (PDB ID: 1E9G) strongly resembles the modeled enzymes from Aspergillus nidulans and Candida albicans. However, Drosophila and Caenorhabditis elegans, in which no determined structure exists, both contain a predicted disordered N-terminal end (Fig. 5B). Nevertheless, the active site region is conserved compared to that of yeast PPase.

The N-terminal extensions found in Drosophila and C. elegans enzymes led us to search the literature for studies on m-PPase activity in other organisms. We found that two reports suggested that Entamoeba histolytica had PPase activity in postnuclear homogenates with localization at the cell membrane and distinctive biochemical features, such as activity at pH 5.0 and no divalent cation requirement (McLaughlin, Lindmark & Müller, 1978a). Additionally, the authors discuss the possibility that the enzyme is enclosed in an organelle. In a follow-up study, McLaughlin, Lindmark & Müller (1978b) demonstrated that the enzyme is heat stable, which is consistent with other family I c-PPases and the temperature sensitivity observed in some family II examples (Klemme & Gest, 1971; Romero, García-Contreras & Celis, 2003; Celis et al., 2003). However, the presence of this enzyme in the membrane was intriguing.

In Fig. 5C, using the Protter tool, analysis of the single pyrophosphatase (244 residues) annotated in UniProt (in different strains S0AZM4, M2Q2R6, A0A5K1U1F2, M7W242, M3SAR5, N9V599, and C4M982) revealed a C-terminal helix that is not found in typical PPases and has no residues that suggest binding to the membrane (Fig. 5D). E. histolytica HM-1:IMSS-B (UniProt M3TGX3) encodes a classic enzyme (175 residues). Recently, an enzyme was found by mass spectrometry in the mitosome of this parasite to colocalize with mitochondrial chaperonin 60 (Cpn60) (Mi-ichi et al., 2009) and it may play a role in sulfate metabolism via physical interactions with sulfate activation enzymes, since in E. histolytica, the c-PPase is not fused to APS kinase, ATP-sulfurylase, or a recognizable domain that suggests its involvement in sulfate metabolism (Bradley et al., 2009; Mi-ichi et al., 2009). Additionally, a similar extension was observed for the Saccharomyces carlsbergensis (Saccharomyces pastorianus) PPase (Lichko & Okorokov, 1991). In Fig. S2, the modeled PPase from S. pastoris shows a similar helix, in this case in the N-terminal end, which could explain the limited interaction with the membrane. However, no hydrophobic residues are found in this protein since the predicted surface charge shows some neutral residues, but several positive residues are shown; perhaps it interacts with another protein that is associated with the membrane, resulting in the determination of enzymatic activity as being associated with a membrane. Further biochemical data are needed to clarify the c-PPase activity in these organisms.

Cytoplasmic PPases, as stated above, are divided into two families, and a recently discovered third family has a complex biochemistry. The classification depends on the sequences and structure that impact their features. The family I PPases are present in all kingdoms of life and use Mg²⁺ as their cofactor. Their structure is either dimeric or hexameric (Heikinheimo et al., 1996; Harutyunyan et al., 1997; Baykov et al., 2017). Family II PPases are relatively rare, are found mainly in bacteria, and are more active when Mn²⁺ is used as their substrate cofactor (Fabrichniy et al., 2004), as they contain either Mn or Co bound to the enzyme. The proposed family III PPases belong to the haloalkane dehalogenase family, are found in some bacteria, and are Ni ²⁺-dependent (Lee et al., 2009). To date, the only structure reported for this family is Bacteroides thetaiotaomicron BT2127 inorganic pyrophosphatase (UniProt Q8A5V9) (Huang et al., 2011).

Using Fold Seek, we found that family III enzymes are not annotated as pyrophosphatases, but rather, using a model from the Thermococcus onnurineus enzyme (UniProt B6YSF3, Fig. 6A) (Lee et al., 2009), most hits are pyrimidine 5′ nucleotidases, haloacid dehalogenases or probable hydrolases. Figure 6 shows that the structures of these enzymes are similar to those of the pyrimidine 5′ nucleotidase PynA from Streptococcus pneumoniae (Fig. 6B), YfnB from B. subtilis (PDB 3I76, Fig. 6C), and a hypothetical protein from Pyrococcus horikoshii (PDB 2OM6, Fig. 6D). The structural alignment shown in Fig. 7E indicates that all of these proteins are closely related. However, bearing a different multimeric structure, the hypothetical protein from Pyrococcus horikoshii is a dimer, while YnfB from B. subtilis is a trimeric protein (Fig. 6). More work is needed to characterize and differentiate these enzymes from other functional metabolic enzymes since the results obtained with Fold Seek suggest that other enzymes bear a similar fold and may behave as moonlighting enzymes in addition to having PPase activity.

Family II enzymes

Cytoplasmic PPases are enzymes needed to maintain the biosynthetic nature of metabolic pathways. Although the overall functions of family I and II enzymes are the same, there are some differences in their specific mechanisms and structures (Sarmina, Peña Segura & Celis, 2017). The family I PPases are usually homohexameric (except in eukaryotic organisms for which they are dimeric) with a monomer with a molecular mass of ∼20 kDa. Family II PPases are usually homodimeric enzymes with larger subunits (∼34 kDa per subunit); the active site is located at the interface of the two subunits, and the active site contains a conserved signature of DHH residues placed into two domains (DHH and DHHA2) (bioinformatic comparison shown in Fig. 7B, B. subtilis enzyme) (Merckel et al., 2001; Sarmina, Peña Segura & Celis, 2017). Additionally, the structure depends on an open and closed state for catalysis, as shown for the B. subtilis enzyme (Ahn et al., 2001). In the case of B. subtilis and Streptococcus gordonii, the two domains are linked by a flexible hinge, and the active site is formed between the two subunits (Fig. 1B); the distance between catalytic residues depends on the binding of the substrate, and the metal bound to the enzyme forms an activated water molecule that results in nucleophilic attack on the PPi (Ahn et al., 2001). The two regions comprising the active site were demonstrated to be directly involved in the two charged residues in the conserved SRKKQ C-terminal signature in the B. subtilis enzyme, and Arg²⁹⁵ and Lys²⁹⁶ are directly involved in substrate binding (Shizawa et al., 2001). One exception in the conformation of the active site is the case of Staphylococcus aureus. The enzyme is in a closed conformation in the absence of substrate, bound PNP (imidodiphosphate), sulfate or chloride, or reaction products; this is due to the position in this enzyme of the SRKKQ motif, which is outside of the active site and is proposed to be mobile upon substrate binding (Gajadeera et al., 2015). After searching the literature, we were not able to find a follow-up study identifying the role of the SRKKQ signature in the S. aureus enzyme, suggesting that further research is needed to elucidate the mechanistic effect of a closed conformation in PPi hydrolysis and perhaps the role of this enzyme as a therapeutic target in multidrug-resistant strains.

Important discoveries regarding the differences between family I and II PPases in photosynthetic bacteria exist. First, the family II PPases of Cereibacter sphaeroides (Rhodobacter sphaeroides) are highly temperature sensitive; the cytosolic enzyme was purified without the use of a previously reported method for Rhodospirillum rubrum family I PPase that involved heating the crude extract at 60 °C for 5 min. The C. sphaeroides enzyme lost all activity under these conditions (Romero, García-Contreras & Celis, 2003; Celis et al., 2003), and the enzymes lost activity if no Co²⁺ was included during purification. Additionally, the C. sphaeroides enzyme shows activity at a wide range of pH values (8.5−9.5), and the catalytic parameters for PPi hydrolysis with Mn²⁺ or Mg²⁺ are the same (Celis et al., 2003). In the R. rubrum enzyme, free Mg²⁺ had an activating effect, while C. sphaeroides had a strong inhibitory effect (Celis et al., 2003). The activation energies of the family I R. rubrum enzyme and family II C. spaheroides enzyme are different; that of the R. rubrum enzyme is 67.3 kJ/mol, while that of the C. sphaeroides enzyme is 36.94 kJ/mol (Ordaz et al., 1992; Celis et al., 2003). These results are in agreement with the temperature sensitivity of the C. sphaeroides enzyme. Interestingly, the c-PPase activity in C. sphaeroides and Rhodobacter capsulatus is not inhibited significantly by NaF, imidodiphosphate, or methylene diphosphate, which are strong inhibitors of family I enzymes (Celis et al., 2003). Nevertheless, Methanococcus janaschii has been demonstrated to protect against NaF inhibition (Kuhn et al., 2000). There is limited evidence regarding the inhibition of family II enzymes by methylene diphosphate or imidodiphosphate in addition to the B. subtilis enzyme.

In a follow-up study, Sarmina, Peña Segura & Celis (2017) demonstrated that c-PPases from photosynthetic bacteria can hydrolyze PPi in the absence of metal cations. In experiments with purified enzymes that were desalted by dialysis and Sephadex G-25, the family I enzyme lost all activity if no Mg²⁺ was present, but family II enzymes retained the same approximate activity (Sarmina, Peña Segura & Celis, 2017). Additionally, Sarmina, Peña Segura & Celis (2017) demonstrated that the metal bound to the enzyme is essential for activity since EDTA completely inhibited the activity at 100 µM EDTA. The presence of the chelator does not affect the dimeric nature of the C. sphaeroides enzyme, as demonstrated by native PAGE analysis (Sarmina, Peña Segura & Celis, 2017). The depletion of the metal ion is not restored by incubating the enzyme with the Mg²⁺ ion; thus, the strong dependence of Co²⁺ on enzyme activity is unremarkable, and Co²⁺ seems to be the key to EDTA resistance in this enzyme. The crystal structure of the C. sphaeroides enzyme sheds light on the involvement of Co²⁺ in the catalytic activity of this enzyme and its role in catalysis in the absence of Mg²⁺. The metal ratio per subunit is relevant in the catalytic steps of enzymes (Zyryanov et al., 2004).

In family II c-PPases from photosynthetic bacteria, once the Co²⁺ ion is chelated, the activity cannot be restored even though the oligomeric state is maintained (Sarmina, Peña Segura & Celis, 2017), which is consistent with the finding that in the B. subtilis enzyme, the active site has three metal ions that coordinate the PPi oxygens and results in high mobility in the active sites of family II enzymes (Fabrichniy et al., 2007). The metal ratio has been determined by proton beams to induce X-ray emission (PIXE), which allows one to know the exact number of metal ions per subunit. Solís et al. (2011) demonstrated that in family II c-PPase from Rhodobacter capsulatus, the enzyme possesses three Co²⁺ ions per enzyme, which contrasts with the finding in PDB 1K23, where the B. subilis enzyme contains only two Mn²⁺ ions and has been proposed to form an activated water molecule that bridges the two ions and is involved in nucleophilic attack on the substrate (Ahn et al., 2001).

Further research is needed to elucidate the different mechanisms involving the metal ions bound to family II c-PPases. A tool for determining the available Mn²⁺ and Co²⁺ in family II enzymes is arsenazo(III) (2,2′-[1,8-dihydroxy-3,6 disulfo-2,7-naphthalene-bis(azo)]dibenzenearsonic acid) reactions (Zyryanov & Baykov, 2002). This technique allows the quantification of the amount of free metal ions tested as well as those derived from hydrolyzed enzymes, which facilitates quantification and ratio estimation.

When the first family II enzyme was discovered (Young et al., 1998; Shintani et al., 1998), it took almost ten years to discover novel family II enzymes (Jämsen et al., 2007). The novel enzymes have extended domains that are regulated by ADP and AMP between the N- and C-terminal domains that form the mobile active site described in previous studies (Kajander, Kellosalo & Goldman, 2013). These enzymes contain a cystathione β-synthase (CBS) domain and are regulated by AMP and ATP; removing the CBS domain favors its activity and eliminates its regulation by phosphorylated nucleotides (Salminen et al., 2014).

Using bioinformatic analysis, as shown in Fig. 7A shows the alignment of examples from family II enzymes, including enzymes from photosynthetic bacteria, Asgard archaea and other archaeal organisms, and, surprisingly, from the intestinal parasite G. intestinalis. The family II CBS domain-containing enzymes were thought to belong exclusively to prokaryotes but are now reported in other protists, such as Emiliania huxleyi, Fragilariopsis cylindrus, Giardia lamblia (UniProt ID V6TM56 and V6U0W2), Guillardia theta, Salpingoeca rosetta, Symbiodinium minutum, and Thalassiosira pseudonana, and four green algae, Bathycoccus prasinos, Micromonas commoda, Ostreococcus lucimarinus, and Ostreococcus tauri (Baykov et al., 2017). As Baykov et al. (2017) reported, the ever-growing databases may contain enzymes with novel features that deserve further study. We analyzed the structure of these enzymes with a two-domain active site containing extensions with CBS domains. In Fig. 7B, we show the models retrieved from the AlphaFold2 database using a rainbow color scheme. The figure indicates that the N- and C-terminal ends form the active site and are between the regulatory domains. We could not find a report describing the presence of a family II c-PPase and an m-PPase. However, in the case of Thermotoga maritima, the presence of a family II enzyme (UniProt ID Q9WZ56) also correlates with the presence of a K⁺-stimulated m-PPase (UniProt ID Q9S5X0).

The findings for T. maritima led us to search for pathogenic organisms that may also encode a family II protein and a single m-PPase. Our survey revealed that Clostridium tetani encoded a family II c-PPase (UniProt accession number A0A4Q0VED8_CLOTA) and a putative K⁺-stimulated m-PPase (UniProt accession number A0A4Q0V5P6_CLOTA). In Fig. S1, we show that the structure of the cytoplasmic family II PPase contains the regulatory CBS domain and present a structural alignment showing that the catalytic domain is highly similar to that of the B. subtilis c-PPase (Fig. S1B). Additionally, m-PPase displayed a fold consistent with that of H⁺ m-PPases. The role of m-PPase in C. tetani physiology remains to be determined. Nevertheless, the enzyme contains only three intrinsic tryptophan residues. The highly conserved Trp⁶⁰² plays a role in stabilizing the protein structure (Chen et al., 2014). These enzymes and their role in metabolic dynamics in this organism remain to be elucidated; specifically, c-PPase is highly active (Chen et al., 2014). These enzymes may be targeted for inhibition.

The family II CBS-containing enzymes shown in Fig. 7 suggest important differences among the enzymes, specifically in the G. intestinalis enzyme, where two loops are modeled. Complementary to the overall structure comparison, in Table 1, we show the data from BLAST analysis using the G. intestinalis sequence, and the resulting hits indicate that the CBS domain is variable among proteins. Conducting the same analysis with HMMR (Fig. 8A) revealed that the domain is present mostly in eukaryotic cells but has low conservation significance. As shown in Fig. 8B, via bioinformatic analysis, structural comparison with the B. subtilis enzyme revealed that the catalytic residues are in the same pocket as those in the B. subtilis enzyme. Additionally, the overall comparison between B. subtilis c-PPase and the selected examples shown here (Fig. 8C) indicates that the catalytic domain is highly conserved, and except for two loops in the G. intestinalis enzyme, the CBS regulatory domains are structurally conserved. These two loops are either relevant for the enzyme’s structure or are intrinsically disordered regions.

Family II enzymes have been shown to form dimers, which are important for catalytic reactions, as shown for B. subtilis enzymes (Ahn et al., 2001; Shizawa et al., 2001). A search in the PDB provided no results for the crystal structures of these enzymes. Thus, the exact assembly mode and symmetry at the crystal structure level remain to be elucidated. However, Dadinova et al. (2020) reported the formation of tetrameric structures by small-angle X-ray scattering of purified recombinant proteins. Using the G. intestinalis sequence, the dimeric model is inconsistent with the findings in bacterial c-PPases determined by Dadinova et al. (2020), who reported that the interaction interface is the catalytic domain, not the regulatory domain. In the model, the interface is the regulatory domain for G. intestinalis (Fig. 8D). Long loops are not present in other bacterial CBS family II c-PPases and interfere with the prediction, suggesting that experimental determination of the crystal structure is needed. Anashkin et al.’s (2020) model of the Clostridium perfringens c-PPase showed that the tetrameric structure is a dimer similar to that of G. intestinalis (Fig. 8D). The catalytic domains, in turn, form the tetrameric form of the enzyme in its active form (Anashkin et al., 2020).

Additionally, Zamakhov et al. (2023) confirmed the tetrameric nature of these enzymes by purifying and characterizing the enzyme of Desulfobacterium hafniense by transmission electron microscopy and biochemical approaches. They found that a highly flexible region is responsible for limiting the crystallization of these enzymes, which is consistent with the findings of G. intestinalis. More studies are needed to assess the role of the flexible region and the regulatory mechanism of CBS-containing enzymes.

Thus, the future of research into enzymes bearing CBS domains is to obtain good-quality crystal structures and identify the dimerization surface in these proteins. Additionally, the DHH and RKKQ signatures are present in this enzyme. This local folding is consistent with the catalytic activity of a family II enzyme (Fig. 8B) (Shizawa et al., 2001). Future mutagenesis studies can uncover the role of the SRRKQ signature in CBS domain-containing c-PPases.

The finding of homologous CBS-containing c-PPases in the TM6 candidate phylum and Verrucomicrobia bacterium, two potential groups linking bacteria and higher eukaryotic organisms (Devos, 2021), suggests that the evolution of highly regulated CBS domain-containing proteins may ultimately involve bacterial CBS c-PPases as ancestors and thus a novel linage of proteins that are regulated by internal inhibition, since the CBS domain is present in enzymes and transporters in all kingdoms of life (Baykov, Tuominen & Lahti, 2011) and is linked to important conformational changes in the enzyme, resulting in strong regulatory flexibility (Ereño Orbea, Oyenarte & Martínez-Cruz, 2013). Thus, the presence of these genes in this group may shed light on the regulatory and metabolic mechanisms needed for these organisms to thrive.

Identifying CBS-containing c-PPases by searching for novel enzymes in sequenced genomes (from cultured organisms to environmental samples) is a much-needed task. The discovery of novel enzymatic systems has revealed the origin of metabolic pathways. However, detailed biochemical analysis is the only approach for assessing certain active pathways. PPi- and CBS-containing enzymes play a role in controlling glycolytic enzymes (for example, phosphoenolpyruvate carboxyltransferase); it has been shown that the binding of regulatory nucleotides and polyphosphates results in the control of metabolic enzymes under low-energy conditions and thus reduces or inhibits biosynthetic reactions, counteracting the effect of PPi on the progression of synthetic reactions (Müller et al., 2012; Baykov et al., 2017).

The core metabolic toolbox of eukaryotic cells comprises 50 enzymes (Müller et al., 2012), mostly dependent on phosphorylation reactions. Müller et al. (2012) proposed that this basic core is reduced in parasitic organisms. This reduction is also found in microorganisms that thrive in sugar-rich environments, such as parasitic protists. Organelles associated with energy generation have also been minimalized into mitosomes or hydrogenosomes, and the close link between polyphosphates and other phosphorylated molecules acts as an alarm signal (Reeves, 1984; Müller et al., 2012; Michels et al., 2006; Wimmer, Kleinermanns & Martin, 2021). Additionally, Danchin, Dondon & Daniel (1984) reported that 2-ketobutyrate acts as an alarmon in E. coli, which clearly points to specialization in a specific environment and may be linked to the origin of mitochondria and metabolic pathways in higher eukaryotes. The ability of G. intestinalis to encode a family II c-PPase may be related to cytosolic ATP synthesis and is the first step toward the complex regulatory mechanism in which the accumulation of PPi and polyphosphates acts as a ratchet pawl to prevent the halting of synthetic pathways (Lloyd, Ralphs & Harris, 2002).

Using the tool Fold Seek (van Kempen et al., 2023), we searched for structures such as PDB 1WPM, corresponding to the dimeric form of the B. subtilis family II c-PPase, and similar structures were also found in Enterococcus faecium (UniProt A0A133CPF0) and S. aureus (UniProt Q2FWY1), which possess a typical Mn-dependent family II c-PPase. However, using the same tool, we found that Trichuris trichiura (whipworm) (UniProt A0A077ZGU3), Plasmodium falciparum (UniProt Q8IM69), and Trypanosoma cruzi (UniProt Q4DJ30), among other hits, may possess an Mn-dependent family II PPase. Similarly, other exopolyphosphatases may exhibit the same folding and perhaps the same biochemical activity in higher organisms, suggesting that c-PPases may be the intermediate evolutionary ancestors of other phosphate metabolism enzymes in higher eukaryotes. The same analysis revealed that 37 PDB entries had similar folds and, in all instances, were structures of family II c-PPases. One example is PDB 4PY9, the crystal structure of an exopolyphosphatase-related protein from Bacteroides fragilis, which has a conserved fold and sequence with the B. subtilis family II c-PPase. Overall, proteins that may be related to family II c-PPases still need to be identified. Cumulative evidence indicates that these enzymes are not restricted to bacteria and are found in microbial eukaryotes, as previously suggested (Baykov et al., 2017).

However, the identification of c-PPases is still lacking. In the inorganic c-PPase of Medicago trunculata (MtPPA1), an N-terminal sequence of 29 to 32 residues (MSEETKDNQRLQRPAPRLNERIL) was missing in the crystal structure, and it was hypothesized that this sequence targeted the enzyme to the mitochondria due to its highly similar sequence and properties of a signal peptide (Grzechowiak et al., 2019). The N-terminal fragment is reported to be sequentially removed since the sequence missing in the crystal is variable and seems to be dependent on the actual activity of the enzyme (Grzechowiak et al., 2019). BLASTp analysis of the N-terminal fragment from MtPPA1 indicated that some plant processing and moonlighting functions of cytoplasmic PPases may play a deeper role in plant biology, as PPi plays a central role in sugar accumulation.

Are m-PPases present in fungi?

With the report of PPase activity bound to the membrane in S. carlsbergensis, we questioned whether other fungi may encode an m-PPase. It has been assumed that no m-PPase homologs exist in protists, yeast, fungi, or mammals. However, m-PPase is now recognized as a relevant enzyme in protists (Serrano et al., 2007), especially in choanoflagellates (Serrano, 2004). The enzyme was lost in the early evolution of animals and fungi, as the enzyme might have been lost in a choanoflagellate ancestor of higher eukaryotes (King & Carroll, 2001; Serrano et al., 2007).

An intriguing question regarding the evolutionary distribution of these enzymes is that they are not found in algae but in Arabidopsis, mung bean, and barley. Additionally, m-PPases are found in archaea, which may provide a link for the presence of these enzymes in higher eukaryotic organisms. Nevertheless, these enzymes have been found in all organisms except fungi and mammals, and further insight is needed to clarify the evolutionary link of these enzymes in eukaryotes.

However, as shown in Fig. 12A, we conducted a BLASTp search in Fungal DB (Basenko et al., 2018) using Vigna radiata (mung bean) membrane-bound PPase and identified 61 proteins with putative m-PPase sequences and structures homologous to those of V. radiata m-PPase; the last hits were either fragments of proteins or unrelated proteins. Additionally, the predicted structure of some examples from this analysis indeed presented a predicted structure with m-PPases (Fig. 12B and, for comparison, please see Fig. 2). The organisms encoding m-PPases include oomycetes and one example of an ascomycete, from which one example is a nonrelated enzyme. This suggests that with the increasing number of available genomes, the presence of m-PPases may open a new avenue for assessing the function of these enzymes in fungal organisms. For example, Pythium ultimum is a plant pathogen, so studying these enzymes may also shed light on the possible control of plant pathogens bearing this enzyme. Additionally, BLASTp was used to detect fragments with homology to membrane PPases. We collected the best and worst hits and searched for AlphaFold2 models (Fig. 12), and the results suggested that not only do the sequences show homology but also that the structural models are consistent with those of membrane PPases. Therefore, this enzyme is present in fungal genomes, but little is known about the expression pattern of this enzyme and its possible kinetic parameters in the fungal membrane. One remaining question is what role the m-PPase may play in fungi and why the enzyme was not lost in all these organisms (Holmes, Kalli & Goldman, 2019).

One hypothesis is that the oomycete cell wall is different from that of true fungi because the former is composed of cellulose and β-glucans. In contrast, the latter contains both β-glucans and chitin as structural polysaccharides (Mélida et al., 2013; Gow, Latge & Munro, 2017). The donor sugar for the synthesis of both oomycete polysaccharides is UDP-glucose, which is synthesized by UDP glucose pyrophosphorylase, which acts on glucose-1-phosphate and UTP, generating UDP-glucose and PPi (Gow, Latge & Munro, 2017). Therefore, the active synthesis of the cell wall involves the constant hydrolysis of PPi generated during UDP-glucose synthesis. Since glucan and cellulose synthesis are performed by membrane-bound enzymes (Gow, Latge & Munro, 2017), it is feasible to hypothesize that membrane-bound PPases can prevent the accumulation of PPi, which remains to be proven experimentally. One putative m-PPase, the sole fungal species listed in Fig. 12, was identified in Sordaria macrospora. The search for putative orthologs in other fungi was unsuccessful, but they are highly similar to other enzymes from bacteria. The function of this apparent unique fungal m-PPase remains to be evaluated (Maeshima, 2000; Holmes, Kalli & Goldman, 2019). Additionally, this enzyme’s subcellular localization is key for determining its role in either cell wall biosynthesis or proton pumping activity in another cell compartment for an unknown process. These hypotheses remain to be explored experimentally.

Molecular biology applications

In vitro transcription for generating large amounts of RNA is taking the central stage with the development of RNA-based vaccines and therapeutics (Perenkov et al., 2023; Malla et al., 2023) and the study of RNA structure −function relationships (Tersteeg et al., 2022). To produce large amounts of RNA, c-PPase plays a key role in preventing the accumulation of PPi as a byproduct and thus inhibiting the enzymes for transcribing the RNA, in particular by preventing the accumulation of Mg-PPi, since Mg²⁺ ions are essential cofactors for T7 RNA polymerase, reducing yield and thus preventing the precipitation of Mg-PPi (Tersteeg et al., 2022). Additionally, identifying and purifying novel enzymes for in vitro transcription has been a goal for increasing yield. An example of a versatile application is the enhancement of PCRs in the presence of thermostable enzymes such as the c-PPase from Thermococcus onnurineus NA1 (Li, Yang & Gao, 2022) and the increased activity of metabolic enzymes such as UDP-glucose and UDP-galactose (Li, Yang & Gao, 2022), which have important industrial applications such as DNA synthesis, DNA and RNA sequencing, and sugar-nucleotide synthesis. Thus, producing c-PPases in a cost-effective method that renders highly active enzymes that are easy to express and purify is desirable because of increasing interest in mRNA-based therapeutics.

Another highly relevant example for this review is the use of c-PPase to develop novel reporter systems. For example, Biswas et al. (2013) developed a screening method for the bacterial DNA primase DnaG, which is needed for chromosomal replication and is structurally distinct from its eukaryotic counterpart. DnaG is an RNA polymerase that releases PPi. By coupling PPi hydrolysis, Pang, Garneau-Tsodikova & Tsodikov (2017) were able to use Mycobacterium tuberculosis primase and c-PPase to screen for inhibitors of both enzymes, selecting nine positive compounds, three of which were confirmed to be inhibitors of DnaG (Biswas et al., 2013). The method uses two purified enzymes: the primase transcribes an RNA primer, which in turn releases PPi; the c-PPase activity is detected by transforming malachite green from a yellow solution into a green solution. This is the first nonradioactive primase activity assay that has served to screen for inhibitory compounds for either enzyme. Thus, this is a tool for the search for novel antimicrobial agents for relevant pathogenic bacteria, a latent threat due to the high rate of antibiotic resistance.

Quantification of the effect of PPi on human health

With the above examples, the correct quantification of PPi has become important in biotechnology and health. For example, we recommend the review by Orriss (2020) on the role of extracellular PPi as a water softener. One example of the importance of PPi is its role in mineralization. Overall, there are several conditions in which the amount of PPi is relevant, especially in tissue calcification disorders (Ralph et al., 2022). A reduced concentration of circulating PPi, which is the main inhibitor of calcium hydroxyapatite deposition in soft connective tissue, leads to ankylosis (Ralph et al., 2022). Thus, improving PPi quantification in serum is a cornerstone for future diagnostic methods.

Measuring inorganic phosphate has been the standard method for quantifying PPase activity (Sumner, 1944). Thus, most methods rely on purified c-PPase to convert PPi into Pi.

A recent paper described the use of coupled enzymatic reactions using ATP sulfurylase to generate ATP and then quantified it by firefly luciferase activity (quantifying the bioluminescence) and included an internal ATP standard to correct for sample-specific interference since the authors aimed to determine the plasma levels of PPi (Lundkvist et al., 2023). This assay shows excellent precision and accuracy, and interestingly, the anticoagulant EDTA blocked the conversion of ATP into PPi in plasma after blood collection.

However, few studies have focused on improving the direct measurement of PPi. In this review, we propose that the use of nanotechnology in the future may further enhance the direct measurements of PPi. For instance, Zhang et al. (2023a; 2023b) developed a carbon quantum dot system in which highly versatile carbon quantum dots are coupled to nanotubes synthesized using metal–organic frameworks. The detection method worked via a “turn-off” mechanism, where Cu²⁺ ions quench the fluorescence from the quantum dots, and the detection is based on the decomposition of the metal–organic framework containing the Cu²⁺ ions, which are released by the coordination of the Cu²⁺ ions with PPi, leading to fluorescence recovery. The fold change in fluorescence determines the presence of c-PPase (human serum PPase) due to the hydrolysis of PPi (Zhang et al., 2023a). This method was proven to work not only for measuring PPase activity but also for measuring the amount of PPi in the sample. This is an alternative, more sensitive method for detecting PPi and PPase activity because of the low toxicity of the sensor components. A similar reporter system was described by Zhou et al. (2016) using bimetallic nanoclusters, where Cu²⁺ ions quench the high fluorescence emitted by an Au–Ag nanocluster, and the recovery of the signal is achieved by the coordination of PPi with the Cu²⁺ ion, which in turn allows the nanocluster to emit fluorescence. In the presence of PPase activity, the fluorescence is quenched by the hydrolysis of PPi, releasing the Cu²⁺ ion.

Overall, nanotechnology-derived sensing methods are powerful tools for the sensitive and accurate measurement of PPi and PPase activity and constitute a platform for the development of high-throughput screening methods for novel inhibitors of PPase enzymes for the treatment of a myriad of diseases (Tian et al., 2019).

PPases as biological targets

The search for novel PPases, specifically in pathogenic bacteria, protists, and fungi, is relevant for identifying novel treatment targets. One example is the identification, molecular cloning, and functional characterization of a gene encoding an H⁺-pyrophosphatase in the protozoan scuticociliate parasite Philasterides dicentrarchi, which infects turbot (Mallo et al., 2015). The enzyme resembles the plant vacuolar m-PPase and is highly sensitive to PPi analogs, inhibiting the growth of the ciliate parasite and resulting in a promising target for controlling scuticociliatosis.

The second area where PPi is important is human health. Numerous reports have assessed the role of the PPi or c-PPase in homeostasis. One example is the role of the PPi in tumor biology, which has been addressed extensively. Evidence suggests that it is an essential housekeeping enzyme that is overexpressed in tumors and is a key player in cell signaling, both in the regulation of its activation and in the role of tumor cell metabolism (Wang et al., 2022). The overexpression of c-PPase in tumors can potentially be targeted by anticancer drugs (Niu et al., 2021; Menteş & Yandım, 2023). One example of the role of c-PPase is the heightened expression of PPA1, the housekeeping isoform of c-PPase in humans, in colorectal tumors (Niu et al., 2023). Niu et al. (2023) reported that PPA1 is highly expressed in different cell lines and colorectal tumor samples and is concurrent with activation of the PI3K/Akt pathway. This report highlights the importance of activating the PI3K/Akt signaling pathway since this pathway controls cell proliferation and stemness properties in colorectal tumor cells.

Previously, Niu et al. (2021) obtained the crystal structure of human c-PPase (PDB 6C45) and determined that the active site has the possibility of binding JNK1-derived phosphor peptides, suggesting its possible role as a protein phosphatase, in agreement with previous observations of the interrelationships between JNK1 activity and PPA1 abundance (Wang et al., 2017). Structural models showing the binding of JNK1-derived peptides are a starting point for revealing novel functions of c-PPases. Additionally, the authors propose that PPA1 activation is a putative novel biomarker and a potential target to block tumor progression due to its important metabolic role (Niu et al., 2023).

PPase inhibitors

The discovery of novel inhibitors for PPase enzymes is not an easy task. The difficulty lies in designing inhibitors that selectively and effectively target the inorganic pyrophosphatase of a desired organism, for example, a bacterial or protist pathogen, without causing undesirable side effects or interfering with the activity of the host PPase or essential cellular processes and pathways. The complexity of the enzyme’s active site (and differences in each family of enzymes), its involvement in multiple cellular pathways, and the need for high specificity and low toxicity in off-targets pose significant challenges in inhibitor discovery and development. Overcoming these hurdles requires a deep understanding of the enzyme’s structure −function relationship with respect to the target organism and innovative approaches in drug design to achieve the desired therapeutic outcomes and, as stated previously, high-throughput methods.

Sodium fluoride is the main inhibitor of family I enzymes. However, the quest for novel inhibitors of family II enzymes is limited to the use of pyrophosphate analogs, such as imidodiphosphate, the most potent inhibitor, and the modest effect of other derived molecules (Zyryanov, Lahti & Baykov, 2005). Additionally, the analogs studied by Zyryanov, Lahti & Baykov (2005) inhibit family I enzymes. The structures of such analogs can be further engineered to find novel inhibitors. However, as stated in the previous sections, novel molecules may help uncover other inhibitors to target family II enzymes. In a paper published in 2017, the inhibitory effect of fructose-1-phosphate on the E. coli enzyme was described (Vorobyeva et al., 2017). The binding and inhibition maxima of fructose-1-phosphate were observed at 1.1 mM, and the authors suggested that fructose-1-phosphate exerts a regulatory effect on metabolism since a previous study showed the interaction of c-PPase with fructose-1,6-bisphosphate aldolase and 5-keto-4-deoxyuronate isomerase (Rodina et al., 2011). Thus, the analysis of protein −protein interactions in other organisms by pull-down experiments and the exploration of phosphate-sugar analogs may also provide fertile ground for finding novel inhibitors and identifying the specific role of c-PPases in cell metabolism, as shown here for E. coli. Another example of a novel PPase inhibitor that targets medically relevant microorganisms was reported by Pang et al. (2016), who described derivatives of 2,4-bis(aziridine-1-yl)-6-(1-phenylpyrrol-2-yl)-s-triazine, an allosteric inhibitor of bacterial PPases. The reported compounds inhibited the Mycobacterium tuberculosis enzyme by binding to an unprecedented target, the interface between monomers, in a species-selective manner. Further work will uncover novel compounds that may point to specific treatments for bacteria and perhaps other organisms in the future.

For membrane PPases, triphenylitin (Celis, Escobedo & Romero, 1998) was demonstrated to inhibit the activity of the m-PPase of R. rubrum. Other inhibitors have been found to increase the asymmetry of T. maritima activity (Vidilaseris et al., 2019). Vidilaseris et al. (2019) employed N-[(2-amino-6-benzothiazolyl) methyl]-1H-indole-2-carboxamide (ATC), which is a nonphosphate and nonsubstrate inhibitor of the m-PPase of T. maritima, and showed in structural studies that it inhibits an allosteric mechanism; thus, the asymmetry found in the dimers of m-PPase, i.e., the subunit alternating between substrate binding and catalysis, was revealed. Compounds similar to ATC exhibit reduced inhibition. With the cumulative information presented in this manuscript, we propose that the organisms encoding an m-PPase are relevant to the study of the effects of its inhibition on the physiology of pathogenic organisms, as shown by Vidilaseris et al. (2019); in the same study, no viability effect was found on Plasmodium spp. cells. Mechanistically, the structures of the other m-PPases differed in terms of the location of the ATC target site, which explains the difference in the effect on the T. maritima enzyme. Further research is needed to expand the applicability of m-PPase inhibition in pathogen control.

The role of PPases as biotechnological tools

The heterologous expression of m-PPases is a novel avenue for biotechnological applications. There are some limited examples of heterologously expressed m-PPases in microorganisms, resulting in a detectable phenotype. One example is the complementation of a distinct type of PPi-dependent proton pump that can efficiently substitute for the V-ATPase of S. cerevisiae in diverse physiological scenarios (Pérez-Castiñeira et al., 2011). The authors demonstrated that the overexpression of the K⁺-dependent proton pump m-PPase from A. thaliana alleviates the phenotype of the lack of the yeast m-PPase enzyme, such as bafilomycin resistance, sensitivity to alkaline pH, calcium and zinc sensitivity, and it can acidify internal components (Pérez-Castiñeira et al., 2011). This system is suitable for the heterologous expression of pathogenic organisms that encode m-PPases and for screening for inhibitors that may prove useful as therapeutic agents. Perez-Castiñeira et al. (2002b) also suggested that this system may be useful for generating cells that are tolerant to macrolides and may be further engineered to produce macrolides since they are crucial drugs in cancer and other applications.

Following the above example, Malykh et al. (2023) recently expressed the m-PPase of R. rubrum in E. coli. The result is correct since the authors generated a codon-optimized version of the enzyme (which is an exceptional tool for transferring genes from other species to bacteria and yeast cells) and found that the enzyme complements the absence of the c-PPase (the expression of the m-PPase allowed for the interruption of the c-PPase) and causes a redistribution of the carbon fluxes. The latter result is surprising and confirms that PPi hydrolysis is the ratchet that continuously flows through the synthesis of biomolecules. Nevertheless, the most surprising result is the redirection of the carbon flux from the tricarboxylic acid cycle and pentose phosphate pathways (which are ATP dependent) to an increase in acetate synthesis under aerobic growth conditions; this is correlated with a 1.5-fold increase in concentration versus the biomass generated. Thus, using m-PPase in E. coli and other biological chassis for metabolite expression may increase the production of metabolites independent of biomass growth. Additionally, this paper contributes to the study by Perez-Castiñeira et al. (2002b); even though the paper has been retracted, the authors confirm in their retraction note (doi: 10.1073/pnas.2213841119) that the conclusions are valid. They showed the complementation of S. cerevisiae expressing low levels of IPP1, the essential c-PPase. The mutant strain and plasmids are available for further experimentation,

With the findings that m-PPases are relevant for heterologous expression and have been shown to alter the phenotype of recipient cells, as a closing section, we think that an important area to explore is the use of m-PPases in plant engineering to resist drought and other stresses, with the ever-increasing threat of global warming.

First, a study by Lee et al. (2005) revealed that the metabolism of transgenic Arabidopsis plants expressing E. coli c-PPase decreased as the accumulation of glucose and fructose increased during the day (two- to threefold, 1.6- to 5.7-fold higher total soluble sugars, respectively). In contrast, starch and sucrose did not accumulate except under continuous white light exposure. Leaves in the transgenic plants presented two- to threefold accumulation of Pi and the same-fold accumulation of uridine diphosphate glucose. The metabolic effect reduced the photosynthetic activity by 20–40% but had a negligible impact on leaf size. This work revealed that a low PPi and high Pi content in the cytosol caused drastic metabolic changes, leading to increased transport in source organs (Lee et al., 2005).

The idea that PPi has a profound effect on sugar transport in plants may be relevant in plants that accumulate sugars that have important value for the ever-growing human population, according to a report by Scholz-Starke et al. (2019). An interesting feature of the H⁺ pump m-PPase is that it can perform the reverse reaction: PPi synthesis. The authors found a reverse mode in the flow of H⁺ ions by patch clamp with isolated Arabidopsis vacuoles. They confirmed the use of mutant lines lacking classical PPi-induced outward currents related to H⁺ pumping. Additionally, the authors used a yeast cell line expressing Arabidopsis m-PPase, and m-PPase synthesis of PPi was also observed, which is likely needed for low-oxygen respiration. The flow of H⁺ ions was observed only in the presence of Pi, and no reverse effect was observed. The authors also demonstrated that PPi synthesis in yeast depended on ATP in the vacuolar preparations (Scholz-Starke et al., 2019). Overall, the authors suggest that the synthetic activity of m-PPase beyond R. rubrum may impact the physiology of plants and protists. We hypothesize that this mechanism may also be true for the limited examples of m-PPases in fungi.

In an increasingly worrying global warming scenario, plant survival under stress is an important area of concern. Global warming and its impact on crops are more complex than an increase in temperature. Heat stress is linked to other environmental covariables, such as humidity, vapor pressure deficit, soil moisture content, and solar radiation (Djalovic et al., 2023). The current knowledge on the effect of m-PPase on transgenic plants suggests that PPi hydrolysis helps maintain the acidic pH of the vacuole but not reverse it. Expression of m-PPase in the apoplast induces the synthesis of glucose-1-P via UDP-glucose; glucose-1-P, in turn, is transformed into glucose-6-P and then fructose-1,6 biphosphate, which enters into respiration, and the ATP produced generates an H⁺ gradient that is used by sucrose synthase and then increases phloem loading (Primo et al., 2019). The accumulation of PPi, in turn, increases plant biomass and results in increased sugar accumulation, which adds to the current knowledge on plant growth. Further understanding of the growth dynamics and its control in plants is needed, including a model of PPi synthesis.

In the model A. thaliana, Nepal et al. (2020) used AVP1 transgenic plants to screen the effect of the overexpression of m-PPase on the coexpression of myoinositol oxygenase (MIOX) based on the knowledge that there are a myriad of stress conditions in which the single expression of m-PPase and MIOX results in a better response to stress. The results showed that the overexpressing A. thaliana plants exhibited better responses to salinity, drought, and temperature stress. Additionally, the crosses between AVP1 and MIOX transgenic plants yielded more seeds under water limitation stress. Importantly, the authors used a high-throughput method for assessing the chemical composition of plant tissues, applying a Scanalyzer with different sensors to capture and measure subtle phenotypes and thus obtain more information on the outcomes of the transgenic plants. Overall, the method proved useful for screening plants that produce more sugar and grow better under stressful conditions. Additionally, regarding regulation, in most instances where transgenic plants are used, m-PPase is overexpressed from the same plant, which is safe for other uses. One example is how m-PP1 from A. thaliana expressed in wheat (Triticum aestivum) induces photosynthate transport from source leaves to roots, resulting in improved yield (more seeds per plant) (Regmi et al., 2020). The same phenomenon has been reported in the moss Physcomitrella patens (Regmi, Li & Gaxiola, 2017), A. thaliana (Paez-Valencia et al., 2011) and rice (Oryza sativa) (Regmi, Zhang & Gaxiola, 2016).

The overexpression of m-PPase in plants also improved their ability to cope with nitrogen limitation. Concomitantly, the authors also investigated whether m-PPase could interact with other proteins and found that it interacts with the A. thaliana receptor-like protein kinase (AtRLK) protein via a yeast two-hybrid system and colocalization experiments. The most striking finding is that the AtRLK mutant and AtAVP1 mutant exhibit the same phenotype under low-nitrogen conditions (Zhang et al., 2023b). Overall, there are still aspects to uncover for m-PPases in plants; perhaps their roles in subcellular localization and protein interactions in plants may lead to future experiments that will help engineer plants tolerant to the increasing stress conditions in the soil, either as a direct effect of temperature or the secondary effects due to the reduction of rainfall and availability of nitrogen. The novel bacterial m-PPases capable of transporting Na⁺ or H⁺ ions may positively impact plants by enhancing their resistance to stress (Luoto et al., 2011).

Recently, vacuolar m-PPase was shown to be overexpressed in wheat plants resistant to infection by Chinese wheat mosaic virus (CWMVV) via a cell death mechanism related to increased PPi hydrolysis. Additionally, the virus produces a small interfering RNA (siRNA) that targets the 3′UTR of m-PPase, preventing the accumulation of m-PPase mRNA and thus bypassing cell death and, concomitantly, regulating H⁺ accumulation, leading to more favorable cellular conditions for virus replication by generating more alkaline conditions (Yang et al., 2020). The authors evaluated the effect by generating a mutant in the region that produces the siRNA, resulting in a reduction in symptoms. Thus, the enhancement of transport by m-PPase and its effect on the acidification of the vacuole are needed for viral resistance. Further studies regarding other plant hosts and viral infections could generate virus-resistant plants.

Finally, m-PPase may play an important role in resistance to bacterial infections. The Xanthomonas effector XopAP prevents stomatal closure via blockage of vacuolar m-PPase binding to phosphatidylinositol 3,5-biphosphate (Liu et al., 2022). By keeping the stomata open, the infection proceeds. Thus, m-PPase has a secondary role in bacterial infections in plants. Further research is needed to determine the safety of genetically modified plants and the use of m-PPase genes to enhance plant resistance to pathogens and stressful environmental conditions.

Literature survey

The literature was searched through the PubMed, Scopus and Google Scholar search engines from February to September 19, 2023. The keywords used were ‘pyrophosphatase,’ ‘inorganic pyrophosphatase,’ ‘bacteria,’ ‘membrane-bound pyrophosphatase,’ ‘proton pump pyrophosphatase,’ ‘plant pyrophosphatase,’ ‘pyrophosphatase activity detection,’ and Boolean ‘and’ for the combination of these keywords. The authors conducted an independent review of the literature to prevent any bias, and the selected articles were chosen for their relevance to the objectives of this review. Additionally, the authors independently conducted the literature review, and the number of articles addressing the subject in recent years is limited.

Bioinformatic analysis

Complementing the content of this review, sequence analysis was conducted to find new information regarding cytoplasmic and membrane-bound pyrophosphatases using the UniProt database (Release version 2023_03, June 27, 2023; The UniProt Consortium, 2023). Sequence analysis was performed using the UniProt database via BLASTp (Altschul et al., 1990) and the HMMER 2.43 online web server (release date February 2022) (Potter et al., 2018). Protein structure prediction was either retrieved from UniProt through the link to the AlphaFold Database or generated using AlphaFold2 (Jumper et al., 2021) using the collaborative environment with the default settings (Mirdita et al., 2022). Dimer prediction of cytoplasmic pyrophosphatases was performed with the AlphaFold Multimer prediction suite using the default parameters. Phylogenetic analysis was conducted with MEGA version 11.0.13 (Tamura, Stecher & Kumar, 2021). The phylogenetic analysis parameters are indicated in each figure legend. The groups identified were manually colored.

Protein structure alignment was conducted using the default settings with the US-align web server (Zhang et al., 2022).