Membrane-bound pyrophosphatases (mPPases) couple pyrophosphate (PPi) hydrolysis to H+ and/or Na+ transport. In the present study, we describe a novel subfamily of H+-transporting mPPases that are only distantly related to known mPPases and show an unusual pattern of regulation by Na+ and K+.
Membrane-bound pyrophosphatases (mPPases; Transporter Classification Database number 3.A.10) couple pyrophosphate (PPi) hydrolysis to transmembrane transport of H+, Na+ or both against their electrochemical potential gradients [1,2]. An mPPase is present in ∼25% of prokaryotic species, many protists and all plants . The enzyme localizes to the cell membranes as well as to the membranes of cell organelles, such as vacuoles in plants and acidocalcisomes in protists and bacteria [3–5]. The PPi-powered ion transport system provides the cell with an energy reserve that appears to be particularly physiologically important during various stresses, such as energy restriction or exposure to salinity or toxins [6–10]. mPPases hold significant biotechnological potential because their engineered overexpression significantly increases the stress tolerance of model and agricultural plants .
mPPases constitute a unique ancient protein superfamily that does not share apparent sequence homology with other proteins. mPPases are structurally relatively simple as they are composed of two identical subunits, which mainly fold as α-helices that span and protrude from the membrane. Each subunit is typically formed by 16 transmembrane helices organized into inner and outer concentric rings of 6 and 10 helices respectively [12,13]. The inner ring constitutes the ion transport funnel, which is ‘gated’ at the approximate midpoint of the membrane. The cytoplasmic extension of the funnel houses the PPi hydrolysis site, which is enriched with evolutionarily old aspartate residues . Approximately half of the identified mPPases require K+ ions for maximal activity . K+ binds in the funnel and interacts with the substrate PPi. K+-independent mPPases have a specifically conserved lysine residue, the terminal NH3+ group of which probably fills the space occupied by K+ in K+-dependent mPPases . mPPase activity is absolutely dependent on Mg2+ ions, consistent with the crystallographic observation of several Mg2+ ions around PPi in the active site [12,13].
mPPases differ in their transport specificities as a result of subtle sequence changes adopted during a long evolutionary history. Whereas the ancestral mPPase apparently operated as a Na+ pump (Na+-PPase), three evolutionarily independent lineages from Na+ to H+ transport (H+-PPases) and one to Na+ and H+ co-transport (Na+,H+-PPase) have been identified [16,17]. The latter transition appears relatively unstable as the phylogenetic clade of Na+,H+-PPases additionally includes enzymes that apparently have reverted to exclusive Na+ pumping . A rationale for the surprising frequency of Na+→H+ pumping transition during the mPPase evolutionary history was recently provided by the finding that Na+-PPases are capable of low-efficiency H+ transport at sub-physiological Na+ concentrations . The evolutionary pressure thus modified a pre-existing function rather than creating it de novo.
The functional divergence of mPPases, the availability of crystal structures and the relative simplicity of the enzyme make the mPPase protein superfamily a promising model system for studying structure–function relationships and evolutionary changes in a membrane environment. In the present study, we identified a novel, highly divergent bacterial mPPase subfamily and expressed and functionally characterized its two representative members from Chlorobium limicola [Cl-PPase(2)] and Cellulomonas fimi (Cf-PPase). Our results demonstrate that the divergent sequences indeed encode PPi-energized H+ pumps with unusual responses to Na+ and K+.
mPPase protein sequences were retrieved from the KEGG (Kyoto Encyclopedia of Genes and Genomes) Database with BLAST (Basic Local Alignment Search Tool)  using H+-PPase of Rhodospirillum rubrum as a query sequence. The full-length sequences thus obtained were aligned with MUSCLE (Multiple Sequence Comparison by Log-Expectation) version 3.6  using default settings. The alignment was manually processed by eliminating incomplete and redundant sequences (leaving 130 taxa) and sequence regions, including indels and ambiguously aligned residues (leaving 398 amino acid residue columns). The resulting sequence block served as input for calculation of a phylogenetic tree representing the full evolutionary diversity of mPPases (Figure 1A). To analyse the divergent mPPase subfamily in more detail, we used only the sequences assigned to this subfamily in the preliminary analysis (Figure 1B). The latter sequence block contained 47 taxa and 450 columns. Both trees were calculated using MrBayes 3.1.2 . In both cases, two parallel tree estimations were performed by calculating four Markov chain Monte Carlo chains for five or two million generations in each run for the trees in Figures 1(A) and 1(B) respectively. The parallel runs converged well with an average S.D. of split frequencies of 0.007 and 0.016 respectively. A sample tree was saved for every 1000 chains and a single consensus tree was summarized from the resulting sample trees using 25% of the samples as burn-in. Phylogenetic analyses performed using different sequence sets or software settings produced qualitatively similar results.
Phylogeny of membrane-bound PPase protein sequences
Expression of divergent mPPases
The genomic DNA of C. limicola DSM245 and C. fimi DSM20113 were purchased from DSMZ (Deutsche Sammlung von Microorganismen und Zellkulturen GmbH). The putative divergent membrane PPase genes (KEGG accession numbers: Clim_1174 and Celf_1319 respectively) were amplified from genome DNA using Phusion DNA polymerase (Thermo Scientific) and the primer pairs 5′-tatacatatgaaaaaacctttcaatgtcgtaag-3′ (forward) and 5′-tatagcggccgttagtgttcgtccaccttgattc-3′ (reverse) for C. limicola and 5′-tatacatatgtccacatcgcagtcg-3′ (forward) and 5′-tataaagcttttagacgcgggtgggc-3′ (reverse) for C. fimi. The reverse primers for His8-tag attachment to the C-terminus of Cl-PPase(2) and Cf-PPase were 5′-tatactcgaggtgttcgtccaccttgattctc-3′ and 5′-tataaagcttgacgcgggtgggc-3′ respectively. The Cl-PPase(2) XhoI restriction site was removed from the nucleotide sequence using the primers 5′-tttgccaactctaggactgc-3′ (forward) and 5′-gcagtcctagagttggcaaa-3′ (reverse). The primers contained an NdeI restriction site (shown in italics) and either NotI, XhoI or HindIII restriction sites, incorporated to facilitate cloning of the genes under the control of the IPTG-inducible T7 promoter in the pET36b plasmid (Novagen).
mPPase genes were expressed in Escherichia coli  and inverted membrane vesicles (IMVs) were isolated as described previously . IMVs were suspended in storage buffer [10 mM Mops–tetramethylammonium hydroxide (TMAOH), pH 7.2, 1 mM MgCl2, 0.9 M sucrose, 5 mM dithiothreitol and 50 μM EGTA], frozen in liquid N2 and stored at −85°C. IMVs were quantified based on their protein content, which was measured using a Bradford assay .
PPi hydrolysis measurements
E. coli does not have an endogenous mPPase and its soluble PPase is efficiently washed away during the IMV isolation protocol, allowing the direct use of E. coli IMVs in activity measurements. The PPi hydrolysis activity of IMVs was continuously recorded at 25°C using an automatic phosphate analyser, as described previously . Analyser sensitivity was adjusted to 2 μM Pi per recorder scale. The reaction medium (25–40 ml) typically consisted of 0.1 M Mops–TMAOH (Mops–KOH or Mops–NaOH, pH 7.2), 0.05–40 mM free Mg2+, 0.5–1000 μM Mg2–PPi complex, 0–150 mM KCl, 0–100 mM NaCl and 5–40 μM EGTA. The highest concentration of Mg2–PPi used was determined by the upper limit of its solubility. The reaction was initiated by adding IMVs. Reaction rates were calculated from the initial slopes of the Pi liberation traces. The results of duplicate measurements were usually consistent within 10%.
Membrane transport measurements
PPi-energized H+ transport into IMVs was assayed by measuring fluorescence quenching of the pH-sensitive probe ACMA (9-amino-6-chloro-2-methoxyacridine) and Na+ transport into IMVs was assayed by determining the accumulation of 22Na+ in IMVs at 23°C using a membrane filtration procedure, as described previously . Changes in membrane potential due to mPPase H+ transport activity were recorded using the fluorescence probe DiBAC4(3) [(bis-(1,3-dibutylbarbituric acid)trimethine oxonol] (Life Technologies). IMVs (0.3–0.6 mg/ml) were pre-incubated for 2 h at 25°C in a buffer containing 0.1 M Mops–TMAOH (pH 7.2), 25 mM K2SO4, 5 mM MgSO4 and 25 nM DiBAC4(3). mPPase- or ATPase-driven ion transport was initiated by the addition of 1 mM TMA4PPi or 0.5 mM ATP respectively. The excitation and emission wavelengths were 488 and 520 nm respectively.
Calculations and data analysis
PPi and Mg2+ form two types of complexes, Mg–PPi and Mg2–PPi, with the latter being the likely true substrate of mPPase . In addition, free PPi forms weak complexes with Na+ and K+ . The concentrations of the Mg2–PPi complex and free Mg2+ (essential cofactor) at pH 7.2 were calculated using 0.859, 0.161 and 0.140 mM as apparent dissociation constant values for Mg–PPi in the absence of alkali metal ions and in the presence of 100 mM Na+ or 100 mM K+, respectively; for Mg2–PPi, a value of 2.84 mM was used in all cases [28,29]. Non-linear least-squares fits to determine parameter values and their S.E.M. were performed using the program SCIENTIST (Micromath).
Bioinformatic analysis of mPPase sequences
A previous phylogenetic analysis  identified a sequence from Chlorobium tepidum that showed only modest identity with mPPases and, accordingly, was separated from mPPases by a very long branch in the phylogenetic tree. This and similar ‘odd’ sequences were long disregarded by us because their low occurrence and large difference from ‘canonical’ mPPases suggested that they were either pseudogene products or proteins that did not belong to the mPPase protein superfamily. Additionally, introduction of these sequences into the analysis led to a shorter block of reliably aligned residues, inaccuracy in the placement of K+-independent and K+-dependent family borders and generally lower clade credibilities within the tree. However, as the number of sequenced prokaryotic genomes has increased, it has become clear that such mPPase-like sequences are not a rarity and are found in bacteria belonging to different classes, strongly suggesting that the sequences correspond to functional proteins.
Our search of the KEGG database  containing 2828 fully sequenced prokaryotic genomes revealed 686 full-length mPPase genes, of which 47 encode the aforementioned mPPase-like sequences. These sequences are only 23%–34% identical to those of other mPPases, which is reflected in the long distance between the two groups in the phylogenetic tree (Figure 1A). On the other hand, the divergent sequences constitute a relatively homogenous group, as indicated by the high intragroup sequence identity (≥56%). Because of this high sequence similarity, phylogenetic analyses could only partly resolve the evolutionary relationships among divergent mPPases (Figure 1B). All divergent sequences were found in bacteria, with the exception of the euryarchaeon Methanosarcina barkeri (Supplementary Table S1). The taxonomic distribution of the host bacteria covers nine of 30 phyla. Although divergent mPPases are typically present in only 1%–2% of sequenced bacteria in a specific phylum, they are found in 25% of Acidobacteria and 38% of Chlorobi. The hosts inhabit diverse aquatic and terrestrial environments or may live as animal colonialists/symbiotes.
With a sequence length of 750–907 residues, divergent mPPases resemble plant H+-PPases (750–850 residues), making them the largest known prokaryotic mPPases (typically 650–750 residues). Most of the extra residues are incorporated as extensions to the proteins' N- and C-termini, two cytoplasmic loops between transmembrane helices (TMH) 7 and 8 and 11 and 12 and the periplasmic loop connecting TMH 10 and 11. Signal peptides in N-terminal extensions, which are never found in canonical bacterial mPPases but are occasionally observed in eukaryotic mPPases , were predicted by SignalP  and Phobius  in 23 and 40 divergent mPPases respectively (Supplementary Table S1). Thus, divergent mPPases may be sorted into different cellular locations, such as acidocalcisomes  in some organisms. Two fundamentally different algorithms, TopPredII  and TMHMM2 , predicted an extra TMH in the C-terminal extensions of all divergent mPPases, which possess 17 TMHs as opposed to the 16 TMHs found in almost all canonical mPPases. This extended topology appears to be an additional lineage-specific feature of divergent mPPases.
Despite the low overall sequence identity to typical mPPases, the divergent mPPase-like sequences contained all residues that constitute the core catalytic machinery of mPPases, as revealed by X-ray structures of Vigna radiata H+-PPase  and Thermotoga maritima Na+-PPase . Thus, all 12 residues that directly interact with PPi, the nucleophilic water and the five Mg2+ ions that contact PPi are fully conserved in the divergent sequences. These residues include Lys240/250, Lys715/694 and Lys744/730 (PPi ligands); Asp276/287 and Asp745/731 (bind nucleophilic water); and Asp243/253, Asp247/257, Asp272/283, Asp504/507, Asn550/534, Asp712/691 and Asp741/727 (Mg2+ ligands; residue numbering used throughout this section refers to Cl-PPase(2)/V. radiata H+-PPase). Furthermore, six other aspartate residues that are found in the substrate-binding pocket, but do not interact directly with PPi, the water or Mg2+, are also conserved in the divergent sequences.
Using the prediction rules for cation specificity devised for canonical mPPases , one could predict H+ transport function for the divergent sequences. Indeed, the occurrence of lysine, not alanine, in the position that aligns with that determining K+ dependence in all other mPPases  (position 553 in Cl-PPase(2) and 585 in Cf-PPase) suggests that divergent mPPases belong to a K+-independent family, all of whose known members are exclusively H+ transporters . Furthermore, the divergent sequences do not contain the tetrade D−T/S−F−M corresponding to the Bacteroides vulgatus Na+,H+-PPase signature, but do contain the Glu residue [position 286 in Cl-PPase(2) and 319 in Cf-PPase] that is absent from K+-dependent plant-type and Flavobacterium johnsoniae-type H+-PPases .
Cloning and expression of C. limicola and C. fimi mPPases
No protein bands clearly attributable to heterologously expressed Cl-PPase(2) or Cf-PPase were detected upon SDS/PAGE of IMVs isolated from transformed E. coli strains and protein staining, probably because of low expression levels and the presence of similarly sized E. coli membrane proteins (Supplementary Figure S1). To unambiguously demonstrate the expression, we engineered His8-tagged Cl-PPase(2) and Cf-PPase, which were detected by Western analysis using an anti-His6-tag antibody (Figure 2A). For an unknown reason, Cf-PPase produced three close antibody-reactive bands, whereas a single band was observed with Cl-PPase(2). A 31-residue N-terminal signal sequence was predicted to be present in Cl-PPase(2) but not in Cf-PPase protein sequence. Its presence in mature recombinant proteins remains unclear because attempts to purify the proteins for mass spectrometric analysis have been so far unproductive. Importantly, no antibody-reactive band was detected in IMVs prepared from E. coli expressing Cl-PPase(2) or Cf-PPase without a His8-tag or those transformed with the cloning vector only. Enzymatic activity measurements provided further evidence of the successful expression of Cl-PPase(2) and Cf-PPase in E. coli (Figure 2B). Cl-PPase(2) and Cf-PPase IMVs exhibited significant PPi hydrolysis activities that were almost completely insensitive to the soluble PPase inhibitor fluoride but were significantly attenuated by the mPPase-specific inhibitor AMDP (aminomethylenediphosphonate) [38,39]. No PPi hydrolysis was detected when Mg2+ was omitted from the reaction medium or was replaced by Mn2+, Zn2+, Co2+, Ni2+ or Ca2+. On the basis of these data, we conclude that both Cl-PPase(2) and Cf-PPase are indeed functional mPPases and constitute the founding members of a new mPPase subfamily, which we have termed ‘divergent mPPases’.
Expression of divergent Cl-PPase(2) and Cf-PPase in E. coli
Cation transport activities
Proton transport activity was assayed using ACMA as fluorescent reporter of ΔpH generation across IMV membranes. The fluorescence of the indicator dye becomes quenched upon acidification of the IMV lumen. As clearly shown in Figure 3(A), both enzymes indeed mediated PPi-dependent fluorescence quenching that was reversible upon the addition of the proton gradient disruptor ammonium chloride. The H+ transport signals of Cl-PPase(2) IMVs were decreased in the presence of the mPPase-specific inhibitor AMDP, were completely abolished by the protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) and were absent from IMVs prepared from E. coli cells transformed with ‘empty’ pET36b plasmid (Figure 3A). Similar effects of AMDP and CCCP were observed with Cf-PPase IMVs. Cl-PPase(2) retained H+ transport activity in the presence of 100 mM Na+, but the initial rate and signal amplitude were diminished (Figure 2A). In contrast, K+ (150 mM) enhanced H+ transport activity. Divergent mPPases are therefore fully capable of H+ transport and differ from Na+-PPases, which pump H+ only at low (<5 mM) Na+ concentrations .
H+ transport activity of divergent mPPases
Cl-PPase(2) was able to transport H+ over a wide range of Mg2–PPi concentrations (Figure 3B). Interestingly, the H+ transport activity progressively decreased at substrate concentrations above 100 μM. Using DiBAC4(3)  as fluorescent membrane potential indicator, we found that the H+ transport reaction in IMVs was accompanied by formation of membrane potential, positive inside (Figure 3C). PPi-dependent DiBAC4(3) fluorescence quenching was not observed with IMVs prepared from E. coli cells transformed with ‘empty’ pET36b plasmid; these IMVs were however able to build up an ATP-dependent membrane potential. In all cases, the fluorescence quenching was reversed upon addition of the potassium ionophore valinomycin.
In contrast, we were unable to detect any 22Na+ transport inside IMVs containing Cl-PPase(2) or Cf-PPase (Figure 4) using our standard assay of Na+ transport activity . However, such activity was evident in IMVs containing a Na+-transporting mPPase from C. limicola [16,18], confirming the validity of our Na+ transport assay. These results define divergent mPPases as monospecific H+ transporters.
Na+ transport in IMVs containing divergent mPPases or Na+-transporting Cl-PPase
Authentic and His8-tagged recombinant Cl-PPase(2) exhibited similar hydrolytic and transport activities in IMVs in the above tests. However, to avoid any possible, subtle effects of the His8-tag, we performed all subsequent experiments using the non-tagged version of the enzyme.
K+ and Na+ as modulators of hydrolytic activity
Similar to its effect on H+ transport activity (Figure 3A), Na+ inhibited the hydrolytic activity of Cl-PPase(2), both in the absence and presence of 50 mM KCl (Figure 5A). This inhibitory effect did not result from a change in ionic strength because the inert electrolyte TMACl did not affect Cl-PPase(2) at similar concentrations. The Na+ dependencies were quite steep and could not be described by a hyperbolic function. A Hill plot analysis of the Na+ dependence yielded a Hill coefficient of 1.9 ± 0.1 (Figure 5B). Accordingly, eq (1), where v0 and vN are rate values at zero and infinite effector (N) concentrations, provided a perfect fit. Values of these parameters, as well as that for the apparent dissociation constant, KN, for the inhibiting Na+ ion, are found in Table 1. K+ ions were also inhibitory, but at higher concentrations (Figure 5A), as reflected in an order of magnitude higher KN value. In the presence of both K+ and Na+, K+ potentiated the effect of Na+ (Figure 5A), decreasing KN for the latter cation by ∼1.5-fold (Table 1). Na+ ions were also inhibitory at 5, 100 and 1000 μM substrate concentrations in the presence of 5 mM Mg2+.
Modulation of Cl-PPase(2) activity by NaCl and KCl
|Enzyme||Modulating cation||v0 (μmol min−1 mg−1)||vN (μmol min−1 mg−1)||KN*(mM)|
|Enzyme||Modulating cation||v0 (μmol min−1 mg−1)||vN (μmol min−1 mg−1)||KN*(mM)|
Mean binding constant calculated assuming highly cooperative binding of two alkali metal ions (see text for details).
Effect of Na+ measured in the presence of 50 mM K+.
The effects of Na+ and K+ on activity were modulated by Mg2+, as shown in Figure 5(C). In relative terms, the inhibition measured at 20 μM substrate increased with decreasing [Mg2+] because Na+ and K+ shifted the activity compared with the Mg2+ concentration profile to higher metal ion concentrations. All three curves in Figure 5(C) could be described by eqn (2), yielding quite similar Mg2+-independent values of activity (vmax) and substantially different apparent Mg2+-binding constants (KM,app; Table 2). The last term in eqn (2) accounts for a small decline in activity at high Mg2+ concentration.
Figure 5(D) shows how Na+ and K+ effects on Cl-PPase(2) depended on substrate concentration. Rate profiles measured over a wide range of substrate concentrations were bell-shaped, both in the absence and presence of alkali cations, indicating inhibition by excess substrate. The profiles were modelled quantitatively in terms of 8, where E denotes enzyme, A1 and A2 are specific activities of the mono- and di-substrate complexes, respectively, and Km1 and Km2 are the corresponding Michaelis constants. The parameter values in 8 were obtained by fitting eqn (3), where [E]t is total enzyme concentration, to the substrate profiles of hydrolysis rate (v) and are shown in Table 3.
|Enzyme||A1 (μmol min−1 mg−1)||A2 (μmol min−1 mg−1)||Km1 (μM)||Km2 (μM)|
|Enzyme||A1 (μmol min−1 mg−1)||A2 (μmol min−1 mg−1)||Km1 (μM)||Km2 (μM)|
The original dependencies for Cl-PPase(2) are shown in Figure 5(D).
Measured in the presence of 100 mM NaCl.
Measured in the presence of 100 mM KCl.
Substrate binding and hydrolysis at a fixed Mg2+ concentration (S=Mg2–PPi)
The polarity of the K+ effect reversed at high substrate concentrations (>60 μM), where K+ became an activator (Figure 5D). Kinetic modelling of the data in terms of 8 and eqn (3) revealed two K+-dependent effects that explain this dual behaviour. First, K+ increased the Km2 value for the binding of the second substrate molecule 3.4-fold (Table 3) while leaving binding of the first substrate molecule unchanged, thereby shifting the balance of the ES and ES2 enzyme complexes towards the more active ES state. Second, the catalytic activity of the ES complex decreased by ∼40% in the K+-bound state, thereby contributing to the inhibitory action of K+ at <60 μM Mg2–PPi, where the monosubstrate enzyme species is dominant. In contrast, Na+ was inhibitory over the entire range of substrate concentrations (Figure 5D). The effects of K+, Na+ and Mg2–PPi on the hydrolytic activity of Cl-PPase(2) explained the effects of these compounds on the H+-transport activity shown in Figure 3.
Detailed kinetic analysis of Na+ and K+ effects on Cl-PPase(2)
In this section, effects of the two cations on Cl-PPase(2)-catalysed hydrolysis were measured over wide ranges of substrate and cofactor Mg2+ concentrations. Profiles analogous to those shown in Figure 5(D) were obtained at 10 fixed Mg2+ concentrations in the 0.05−40 mM range in the absence of alkali cations and in the presence of 100 mM Na+ or K+. The substrate concentration range was constrained by the upper limit of Mg2–PPi solubility (1 mM). At low Mg2+ concentrations, the maximal substrate concentration was further decreased (to 500, 200, 50 and 25 μM at 0.5, 0.2, 0.1 and 0.05 mM Mg2+, respectively) to maintain the concentration of metal-free PPi, a strongly anionic species that can non-specifically affect the reaction, below 6 mM.
The rate versus substrate concentration dependencies were clearly bell-shaped at high, but not at low, Mg2+ concentrations (Figure 6). This difference was most clearly evident for the Na+-inhibited reaction, where this transition was observed at higher Mg2+ concentrations. In terms of 8, the appearance of the bell shape was indicative of A1 > A2, whereas the opposite relation explained the form of the profiles at low Mg2+ levels. Thus, the degree of substrate inhibition decreased with decreasing [Mg2+] and was eventually replaced by substrate activation, which became increasingly evident as Mg2+ concentration was further decreased.
Hydrolysis rate versus substrate concentration profiles for Cl-PPase(2) at different Mg2+ concentrations
Computational analyses of the data in Figure 6 were done in two steps. First, each rate profile shown in Figure 6 was quantitatively modelled in terms of 8 to obtain the apparent values of the four parameters involved (A1, A2, Km1 and Km2). In the next step, their dependencies on Mg2+ concentration (Figure 7) were subjected to further analysis to derive the kinetic scheme (2) and the true (Mg2+-independent) values of the respective parameters. The resulting 2 differs from 8 by including two metal-binding equilibria described by the dissociation constants KM1 and KM2. Accordingly, K′m1 and A′1 refer to the Mg2+-bound enzyme. This scheme was formulated based on the analysis that follows.
Parameters of eqn (3) for Cl-PPase(2) as a function of free Mg2+ concentration
Overall scheme of substrate and Mg2+ binding during catalysis
The A1 profiles (Figure 7) indicated that Mg2+ removal from E2S renders it inactive and showed that the alkali cations, especially Na+, facilitate this dissociation reaction. The value of the Mg2+-binding constant, KM2, obtained by fitting eqn (4) to these profiles, increased 200-fold in the presence of 100 mM Na+ and less markedly in the presence of K+ (Table 4). In contrast, the limiting value of A1, A′1, was unaffected by the alkali cations. The Km1 profile was also markedly affected by inclusion of Na+ and revealed a transition between two limiting values of Km1 upon Mg2+ binding or, equivalently, different affinities of substrate-free and substrate-bound enzyme for Mg2+ (eqn 5). The KM2/KM1 ratio was found to be 110 with Na+ compared with approximately 3 without alkali cations (Table 4). The lack of Mg2+-concentration dependence of Km1 in the presence of K+ suggested that KM1 does not differ significantly from KM2. Importantly, the limiting values of Km1, K′m1, were again similar in all cases.
|Parameter||Unit||No alkali metal||Na+ added||K+ added|
|A′1||μmol min−1 mg−1||0.41±0.03||0.38±0.02||0.34±0.03|
|A′2||μmol min−1 mg−1||∼0.05||∼0.05||∼0.05|
|Parameter||Unit||No alkali metal||Na+ added||K+ added|
|A′1||μmol min−1 mg−1||0.41±0.03||0.38±0.02||0.34±0.03|
|A′2||μmol min−1 mg−1||∼0.05||∼0.05||∼0.05|
A high value of Km2 and a low value of A2 together with the limitation on permissible substrate concentration precluded estimation of these parameters with comparable accuracy, insofar as they showed a smaller variation with [Mg2+]. Within the error of measurements, A2 values did not show significant variation with [Mg2+], were generally similar in the three cases and were approximately 10-times lower than the corresponding maximal A1 values. The difference between A2 and A1 values decreased with decreasing Mg2+ concentration; in the case of the Na+-inhibited reaction, A1 eventually became much lower than A2. Km2 values measured without alkali cations were significantly lower than those measured with Na+ or K+ at low [Mg2+], but tended to reach the same level at high Mg2+ concentrations (Figure 7). Estimated values of all parameters are summarized in Table 4.
The results of this analysis suggested that Na+ and K+ compete with Mg2+ for binding to substrate-free enzyme and enzyme–substrate complex, thus suppressing the hydrolysis reaction that requires enzyme-bound Mg2+ in addition to substrate-bound Mg2+. Na+ appears to be unable to functionally replace Mg2+ at their common site, as indicated by the steady decline in A1 (by up to three orders of magnitude) at low [Mg2+] in the presence of Na+ (Figure 7). The effect of K+ on A1 is much smaller, making it possible that the enzyme retains some activity (≤10%) in the K+-bound state. Greater effects of Na+ and K+ on KM2 than on KM1 indicated that bound substrate strengthens alkali cation binding. On the other hand, the much weaker sensitivity of catalysis in the disubstrate enzyme complex to changes in [Mg2+] implies tighter Mg2+ binding or weaker alkali cation binding.
In the present study, we report the discovery of a group of sequences that align with previously known mPPases but are separated from them by a very long phylogenetic/evolutionary distance. Based on the functional properties of the two characterized divergent proteins, these sequences constitute a group of fully functional H+ pumps, thus named ‘divergent H+-PPases’. They retain the core structural features (overall membrane topology, except for an additional C-terminal TMH, as well as all active site and gate residues) and catalytic properties (specificity for pyrophosphate and dependence on Mg2+ as cofactor) common to all mPPases. The divergent H+-PPases are structurally similar within the group but differ considerably from other mPPases. Because of the limited intra-group variation, the divergent H+-PPases are probably relatively recent descendants of the mainstream mPPases. However, phylogenetic analyses alone were unable to pinpoint the precursor of the divergent H+-PPase subfamily because of its large separation from the other subfamilies. Instead, this inference is drawn from a comparison of Na+ and K+ effects on different mPPase subfamilies.
Na+-PPase, the likely precursor of the entire mPPase superfamily [16,18,41], has the largest repertoire of transport activities and regulation capabilities. This protein is capable of transferring both Na+ and H+ across membranes , absolutely requires Na+ for both PPi hydrolysis and ion transport reactions and is further activated by K+ [16,18,41]. The H+ transport activity of Na+-PPases progressively decreases as the Na+ concentration is increased above 0.1 mM and is completely absent at physiological levels of Na+ (∼10 mM) . This finding, taken together with the reported effects of mutations, suggest the presence of two Na+-binding sites in Na+-PPase, one of which is inhibitory for H+ transport (but not for Na+ transport) and is located outside the ion-conductance channel . Na+,H+-PPases differ from Na+-PPases by the diminished affinity of the latter site, allowing them to retain H+ transport capability at high Na+ levels and thus the ability to transport both Na+ and H+ under physiological conditions . Neither of these sites overlaps with the K+-binding site that is found in K+-dependent H+-PPases , Na+-PPase  and Na+,H+-PPase , but is absent from K+-independent H+-PPases . Divergent H+-PPases belong to this latter category because they contain a lysine at the position that determines the K+ dependence of all known mPPases . Previously known evolutionary routes to exclusive H+ transport involve loss of Na+ dependence, both as activator and inhibitor, but allow loss of K+ dependence, retention of K+ dependence and even enhanced K+ dependence in the plant-type H+-PPase lineage [1,2,5]. In contrast, evolution to divergent H+-PPases apparently involved the loss of the activator Na+ site and hence Na+ transport function in Na+-PPases, but retention of the inhibitory Na+ site. The small effects of K+ on divergent H+-PPases may result from its weak binding to the Na+ site, a feature that may remain undetected in K+-dependent mPPases because of the much larger effect of K+ at its specific site in this family.
The major effect of Na+ on Cl-PPase(2) is to decrease the affinity of the enzyme–substrate complex for the essential activator Mg2+. The corresponding binding constant, KM2, increased 200-fold in the presence of 100 mM Na+ (Table 4). The Na+-binding constant can thus be estimated to be 7 mM, assuming that Na+ binding is proportional to the second power of Na+ concentration, as suggested by Figure 5(A). The structure of the V. radiata H+-PPase in complex with a very close PPi analogue, imidodiphosphate, contains five Mg2+ ions, all in direct contact with substrate. Thus, one likely possibility is that Na+ inhibition results from its displacement of one of these Mg2+ ions. A further intriguing possibility is that this Na+/Mg2+ exchange is somehow involved in Na+ transport by Na+-PPases and that the inhibition of divergent H+-PPases is a consequence of the inability to transfer Na+ because of altered ion channel and/or gate cation specificity. These interpretations are consistent with the inability of Na+ to suppress Mg2+ binding in the Na+-transporting PPase of Methanosarcina mazei . However, further studies are clearly needed to test this hypothetical mechanism.
Na+ inhibition of Cl-PPase(2) also suggests a mechanism for the co-operation of Cl-PPase(2) with the Na+-transporting PPase that is also present in C. limicola cells [16,18]. Cl-PPase(2) apparently works in parallel with an H+-transporting ATPase to maintain an H+ gradient across the cell membrane, whereas the Na+-PPase uses PPi energy to pump out Na+, which is detrimental to the cell above its physiological level of about 10 mM. Therefore, the susceptibility of Cl-PPase(2) to inhibition by Na+ may represent a regulatory mechanism that allows the host cell to invest its available PPi pool to preferentially fuel H+ or Na+ transport at normal or elevated Na+ levels respectively. Importantly, although Na+-PPase is also inhibited by excess Na+ , this inhibition is observed at much higher Na+ concentrations than are required to inhibit Cl-PPase(2). Furthermore, this inhibition depends only on [Na+] for Na+-PPase , but on [Na+]2 for Cl-PPase(2), a relationship that allows efficient inhibition over a narrow [Na+] range.
A search through complete bacterial genomes revealed that divergent H+-PPases are accompanied by another mPPase in 20% of cases (Supplementary Table S1), but some of the alternative mPPases appear to be K+-independent H+-PPases, which are not regulated by Na+. Such organisms, as well as those containing only a divergent H+-PPase, use other mechanisms of Na+ efflux that depend on ATP or membrane potential as the energy source. In these organisms, increases in PPi level resulting from divergent H+-PPase inhibition by excessive Na+ are expected to inhibit numerous biosynthetic reactions that consume ATP and produce PPi as a byproduct , thereby also making more energy (ATP) available for Na+ transport.
However, complete description of PPi utilization pathways in a bacterial cell must include soluble PPases, also present in most mPPase-containing bacteria (Supplementary Table S1) . Cellular PPi concentration is around 1 mM in different bacteria  despite the presence of excessive soluble PPase activity. Clearly, soluble PPases must be under strict control to maintain such high levels of PPi substrate for cation transport by divergent and other mPPases. The mechanism of such control is still poorly understood and may vary in different bacteria. Acquisition of nucleotide-binding CBS (cystathionine-β-synthase) domains by some soluble PPases, which allow differential regulation of PPase activity by adenine nucleotides , is perhaps the most comprehensible case. Other probable mechanisms include down-regulation of PPase activity by bound cellular proteins  and mobilization of the PPi pool within acidocalcisome-like structures . Differential regulation of soluble and membrane PPases at the level of gene expression was also documented .
In summary, we have characterized a unique subfamily of membrane-bound, Na+-regulated PPases that transport H+ ions and are sporadically present in several bacterial phyla. Our findings substantially expand the known sequence space of functional mPPases and support the hypothesis that selection pressure favours membrane H+ coupling versus Na+ coupling in modern organisms [49–52].
Cellulomonas fimi PPase
carbonyl cyanide m-chlorophenylhydrazone
Chlorobium limicola H+-translocating PPase
bis-(1,3-dibutylbarbituric acid)trimethine oxonol
inverted membrane vesicle
Kyoto Encyclopedia of Genes and Genomes
Heidi Luoto and Erika Nordbo performed the experiments. Heidi Luoto and Anssi Malinen performed the phylogenetic analysis. All authors were involved in designing the study, interpreting the results and writing the manuscript.
We are grateful to Edita Jetullahi (University of Turku) for the preparation of TMA4PPi and the CSC–IT Center for Science for the computational resources used in the phylogenetic analysis.
This work was supported by the Academy of Finland [grant number 139031]; the Russian Foundation for Basic Research [grant number 12-04-01002]; the Turku University Foundation [grant number 10506]; the Emil Aaltonen Foundation; and the Ministry of Education and the Academy of Finland.