Stress conditions (e.g. anoxia) frequently result in a decrease of [ATP] and in an increase of [ADP] and [AMP], with a concomitant increase of [Mg2+] and other cations, e.g. Ca2+. The elevation of [Mg2+] is linked to the shift in the apparent equilibrium of adenylate kinase. As a result, enzymes that use Mg2+ as a cofactor are activated, Ca2+ activates calcium-dependent signalling pathways, and PPi can serve as an alternative energy source in its active form of MgPPi or Mg2PPi. Under anoxic conditions in plants, an important source of PPi may come as a result of combined reactions of PK (pyruvate kinase) and PPDK (pyruvate, phosphate dikinase). The PPi formed in the PPDK/PK cycle ignites glycolysis in conditions of low [ATP] by involving PPi-dependent reactions. This saves ATP and makes metabolism under stress conditions more energy efficient.
Stress factors severely affect the energy state and ATP levels in living cells. Changes in ATP content and hence the cellular energy charge status result either from limitations/inhibition of ATP synthesis or from an increased ATP consumption. ATP deprivation is especially evident under low oxygen stress, but also during chilling, drought, phosphate starvation or in meristematic tissues . The first consequence of anoxia is a decrease and then stopping of aerobic respiration , hence ATP synthesis via oxidative phosphorylation is inhibited. Other means of ATP synthesis, i.e. substrate phosphorylation (resulting mainly from lactic and alcoholic fermentation)  and alternative respiration using nitrite, have a limited capacity for ATP synthesis .
An important consequence of the decrease in [ATP] is a release of divalent cations such as Mg2+, Ca2+ and Mn2+ . Although the total concentration of Mg2+ in plant tissues is in the order of 10 mM , a significant amount of Mg2+ is bound to different metabolites, in particular nucleoside tri- and di-phosphates and, under normal physiological conditions, the steady cytosolic concentration of free Mg2+ can be as low as 0.2–0.4 mM . Generally, nucleoside triphosphates bind Mg2+ very tightly and those complexes function as cofactors and energy-rich compounds. Nucleoside diphosphates bind Mg2+ less firmly, whereas nucleoside monophosphates bind Mg2+ very weakly and usually participate in metabolism as Mg2+-free forms. A decrease in pH levels observed under stress and, in particular, in hypoxic conditions, shifts chemical equilibria towards release of Mg2+, further increasing its concentration. In these conditions, the species that normally only loosely bind Mg2+, such as PPi, turn into Mg2+-complexed physiologically active forms.
Under the conditions of ATP deficiency, plants can use PPi as another energy currency. This molecule is a byproduct of several biosynthetic processes, and its utilization is usually linked to a PPase (inorganic pyrophosphatase) activity. This is because PPi accumulation affects the equilibrium of the biosynthetic reactions, and it has to be quickly removed . PPase is inhibited by lower pH values (characteristic for many stresses) , making PPi a valuable player in plant bioenergetics under stress, especially in the cytosolic compartment. The plant cytosol has been proposed previously as an environment where PPi can serve as an alternative to ATP as the energy donor .
In the present review, we analyse the operation of plant metabolism that, upon stress-induced elevation of [Mg2+], uses PPi as energy currency, and we discuss physiological conditions that are favourable for this operation. The use of PPi during early stages of glycolysis makes this process independent of ATP. Similarly, PPi can replace ATP for efficient pumping of protons into the vacuole (preventing acidosis). Our and other [11–14] analyses of sources of PPi in stressed cells point to possibly critical roles of two enzymes, PPDK (pyruvate, phosphate dikinase) and PK (pyruvate kinase), involved in net PPi production under stress.
Mg2+ AND ADENYLATES UNDER ANOXIA
A fall in total [ATP] leads to an increase in [Mg2+] . This may compensate for low ATP production in such a way that many enzymes limited by low [Mg2+] become activated and thus cause an increase in corresponding metabolic fluxes. The release of Mg2+ also results in displacement of apparent equilibria (Kapp) of certain enzymes. Most importantly, increased [Mg2+] leads to a displacement in the Kapp of AK (adenylate kinase) [7,15]. Also, an increase of [Mg2+] from 0 to 1 mM results in the displacement of the Kapp of aconitase, and the citrate/isocitrate ratio rises from 9 to 21 . This is unfavourable for the TCA (tricarboxylic acid) cycle turnover and stimulates efflux of citrate from mitochondria . Yet another important consequence of increased [Mg2+] relevant to the present review is the activation of PPi-dependent glycolysis and the vacuolar proton-pumping PPase (see below).
AK equilibrates pools of free and Mg2+-bound adenylates  and establishes a link between [Mg2+] and ratios of free and Mg2+-bound adenylates, as well as the inner membrane potentials of mitochondria and chloroplasts [5,7]. The AK-governed link between contents of adenylates and [Mg2+] allows the calculation of [Mg2+] from the Kapp of AK, which varies from approximately 0.3 to 1.5, depending on [Mg2+] . Under anoxia in germinating lettuce seeds, concentrations of all three adenylates (ATP, ADP and AMP) fall within the AK equilibrium and there is a clear decrease of AK mass action ratio (in particular at 2 kPa oxygen) . This corresponds to an increase of [Mg2+], from low submillimolar levels (0.2–0.4 mM) to millimolar (~1 mM), accompanied by a huge increase of [AMP], a decrease of [ATP] and small or no changes in [ADP] . Even more striking increases of [AMP] under oxygen deficiency were observed in maize [3,20] and Carex , and the overall adenylate contents were well beyond the Kapp of AK. Although the magnitude of [ATP] decrease and [ADP] increase was relatively modest (up to 2–3-fold decrease and increase for ATP and ADP respectively), the hypoxia-induced AMP content increased by 10-, if not 100-fold, when compared with oxygen-sufficient conditions . With nitrate in the medium (compared with chloride), this increase is even higher. Thus in those cases, low oxygen conditions were likely to bring about AK-independent mechanism(s) of AMP production. Simple enzymatic hydrolysis of ADP, e.g. by apyrase , would be, in this case, wasteful from the point of view of cell energetics. A mechanism which consumes ADP and generates alternative energy currency will be analysed later in the present review.
CALCIUM SIGNALLING AS A FEEDBACK OF CHANGES IN [Mg2+] DURING ANOXIA
An important consequence of the AK-governed adenylate equilibrium in the IMS (intermembrane space) in mitochondria is its effect on signalling networks. It has been shown that Mg2+ is an allosteric activator of Ca2+ binding to calmodulin , with the latter activating target enzymes in response to submicromolar increase in [Ca2+]. Changes in internal [Ca2+] can result from changes in [Mg2+] at a millimolar range . When ATP levels decrease under hypoxic conditions and [Mg2+] increases, Ca2+ is released from mitochondria to the cytosol, where its concentration raises above 0.1 μM . On the other hand, as it was established for animal tissues, Mg2+ is a strong inhibitor of Ca2+ uptake into mitochondria . Thus the increase of [Mg2+] in the IMS of mitochondria under intensive respiration can lead to accumulation of Ca2+ in the IMS. This subsequently leads to acti-vation of Ca2+-dependent enzymes, including NADPH and NADH dehydrogenases in the inner mitochondrial membrane facing the IMS  and NAD kinase in the IMS . Thus the AK equilibrium controls Ca2+ homoeostasis, a primary event in the signalling network, which subsequently has profound effects on metabolic events.
The elevation of [Ca2+] upon decrease of [ATP] also occurs because Ca2+ tightly binds to ATP, and the lowering of [ATP] releases Ca2+ in a similar way as Mg2+. The concentration of Ca2+, similar to Mg2+, can also be controlled by the AK-governed adenylate equilibrium , with calcium chelated by the adenine nucleotides to nearly the same extent as magnesium . As a consequence, the ratio of Catotal/Ca2+ is roughly equal to that of Mgtotal/Mg2+ . This also means that an increase of [Mg2+] from 0.1 to 1 mM leads to an increase of [Ca2+] from 0.1 to 1 μM, assuming that the total concentration of calcium is 1000-fold lower than magnesium. Under hypoxic conditions, Ca2+, besides its known participation in the activation of the GABA (γ-aminobutyric acid) shunt via activation of glutamate decarboxylase , can participate in the regulation of vacuolar H+-PPase , in activation of external mitochondrial NAD(P)H dehydrogenases which are also more active at lower pH values (using NADH and NADPH)  and in lower pH-stimulated activation of NAD-kinase . Thus, in calcium signalling, an energy deficit becomes an important feedback of Mg2+ release and pH decrease during hypoxic stress.
Mg2+ AND PPi UNDER ANOXIA
PPi is generated in several biosynthetic reactions such as activation of sugars (e.g. synthesis of UDP-glucose and ADP-glucose), activation of amino acids and fatty acids, and during elongation reactions for synthesis of proteins and nucleic acids. In these reactions, the removal of PPi by active PPase drives the equilibrium in the direction of synthesis . On the other hand, Baltscheffsky  recognized that PPi generated during biosynthesis is not always wasted but in certain conditions can be used as energy source. For instance, in developing cereal seeds, PPi produced in the cytosol during starch biosynthesis can drive sucrose conversion into glucose 1-phosphate . The usefulness of PPi as an energy source depends on the energy charge of the cell, [Mg2+] and the actual location of PPi synthesis. PPi is ineffective at high energy charge values, when ATP is the major energy source, and the role of PPi can only be apparent under low [ATP] conditions, which leads to an increase of [Mg2+].
Most PPi-dependent enzymes use a MgPPi complex as a substrate rather than free PPi . The enzymes become more active upon elevation of [Mg2+], and in hypoxia-tolerant plants, the expression of genes for those enzymes is usually strongly increased . For instance, PFP [PPi-dependent PFK (phosphofructokinase)] needs higher [Mg2+] than the MgATP-dependent PFK, owing to the much lower affinity of Mg2+ for the complexation with PPi than with ATP (KMgPPi=1.2 mM−1 compared with KMgATP=73 mM−1) [39,40]. Free PPi probably acts as an inhibitor when the MgPPi complex serves as a substrate . Figure 1(A) shows relative amounts of free PPi, MgPPi and Mg2PPi at different [Mg2+]. The specific form of the MgPPi complex serving as substrate depends on a given enzyme. For instance, some aminoacyl-tRNA synthetases use MgPPi and others prefer Mg2PPi in pyrophosphorolysis . On the other hand, proton-pumping H+-PPase, localized in the tonoplast of vacuole [42,43], uses Mg2PPi rather than MgPPi. This means that for exhibiting efficient activity, this enzyme needs a further increase of [Mg2+]. The activation process is complex, and it involves competitive inhibition by Ca2+, which is released from vacuoles and mitochondria upon low [ATP] conditions  and non-competitive inhibition by the actual substrate Mg2PPi. Free PPi and Mg2+ act as competitive inhibitor and activator respectively . Thus, for the H+-PPase to effectively pump protons into the vacuole, the [Mg2+] should be in the millimolar range. Proton pumping by the H+-PPase may be important to prevent cytoplasmic acidosis; the pump is mainly inactive in non-stressed cells (when H+-ATPase is operating) [45,46] and is activated under anoxia .
Basic characteristics of PPi hydrolysis
The ratio between free PPi, MgPPi and Mg2PPi is under pH control . At lower pH values, which are particularly characteristic of hypoxic stress, the build-up of MgPPi and Mg2PPi can be achieved at lower [Mg2+]. The maximal fraction of the most abundant form, MgHP2O7−, is observed at pH 6–6.5 . According to the analysis conducted by Davies et al. , the pH-dependence of the ratio of standard energies of hydrolysis of PPi and ATP shows that the energy of PPi hydrolysis becomes comparable with the energy of ATP hydrolysis when the pH falls below 7 (Figure 1B). At the hypoxic cytosolic pH range (pH 6.5–6.8) , the energy of PPi hydrolysis does not decrease with the increase of [Mg2+], and at pH 6.4 even increases when the [Mg2+] increases from submillimolar to millimolar values (Figure 1C).
In plants PPases are presumably exclusively located in the vacuole and plastids, and the cytosolic PPi pool is not hydrolysed , although a possibility of PPases in the cytosol has been reported . These PPases are less active at pH values below 7 and are inhibited by Ca2+ accumulating under low [O2] . PPi concentrations are in the range 0.04–0.8 mM, and PPi can be used as an energy donor, instead of ATP, in a variety of situations . PPi generates equilibrium fluxes to bypass and buffer ATP-dependent metabolism . Although the ratios of nucleotides such as ATP/ADP and UTP/UDP change in a similar manner, [PPi] changes independently, thus highlighting its importance as an autonomous energy donor . When the bacterial PPase was expressed in potato plants, this led to decreased vitality under O2 deficiency .
PK/PPDK SUBSTRATE CYCLE
In animal tissues, AK equilibrium is linked to creatine kinase equilibrium and equilibration of adenylates is shifted from AK equilibrium values to higher [ATP] and lower [ADP] supported by creatine kinase activity [29,56]. On the other hand, in anoxic plant tissues, AK equilibrium is probably linked to PK/PPDK equilibration of adenylates. This contributes to higher [AMP] and [PPi] under oxygen stress (Figure 2). Both PK and PPDK, as well as AK, are induced under low [O2] [13,14,57] and can displace the AK equilibrium towards AMP and PPi production. It is worth mentioning that, during photosynthesis in C4 plants, AK has an additional task of recycling the AMP produced by the PPDK reaction . In C4 plants, the PPDK/AK cycle sustains the formation of PEP (phosphoenolpyruvate), the primary carboxylation substrate during C4 photosynthesis.
Equilibrium of adenylates by AK in conjunction with PK and PPDK reactions
The expression of the gene coding for PPDK is exceptionally high under low [O2] conditions (at least a 100-fold induction) [13,14]. This seems difficult to explain from the point of view of switching from PK to PPDK, because PK is neither limited nor down-regulated. The only way to explain the importance of PPDK in anaerobic glycolysis is its participation in the substrate cycle where pyruvate is used in a reverse reaction to form PEP and AMP . The total balance of the coupled PK/PPDK reactions is: ADP+Pi=AMP+PPi, the latter serving as an energy source under anoxia. Thus, under low [O2], the PK/PPDK substrate cycle (Figure 3A) becomes the major non-ATP source of energy . The ATP is used there only as a catalyst (it is produced by PK and consumed by PPDK). The anoxia-induced increase in [AMP] [3,20] results, in particular, in stimulation of AMP-activated protein kinase, which is involved in regulation of many cellular processes , and the [PPi] increase [60,61] ignites glycolysis at low [ATP] (see below). On the other hand, [ADP] is not strongly increased in anoxic conditions [3,19,20]. This is consistent with the role of PK/PPDK reactions in PPi formation from ADP and Pi when ATP formation from ADP and Pi is ceased, at least at the level of oxidative phosphorylation.
PK/PPDK substrate cycle (A) and mechanism of regulation of PPDK (B)
The adenylate/Pi/PPi levels regulate PPDK activity through a unique mechanism in which a kinase/phosphatase uses ADP/forms AMP during phosphorylation (formation of the inactive form of PPDK) and uses Pi/forms PPi during dephosphorylation (formation of the active form of PPDK) (Figure 3B). Both PPDK activation and inactivation require Mg2+ , supporting a view that the true substrates are MgADP for phosphorylation/inactivation (suppressed by pyruvate) and MgPi for dephosphorylation/activation (suppressed by MgADP) [63,64]. This unusual mechanism is not directly controlled by energy charge (or ATP/ADP ratio) , but rather by the MgADP/MgPi ratio, which is displaced to a higher [MgPi] only upon a significant release of Mg2+. The apparent stability constant for MgPi is 0.5 mM−1, which is less than for MgPPi (1.2 mM−1) and for MgADP (4 mM−1) , which means that the availability of Mg2+ determines the direction of the PPDK phosphorylation/dephosphorylation reaction. The use of MgADP for phosphorylation and MgPi for dephosphorylation would indicate that phosphorylation occurs at submillimolar [Mg2+], whereas dephosphorylation (reactivation) occurs at millimolar [Mg2+] and is observed when [O2] falls to very low values.
The PK/PPDK cycle is composed of one essentially irreversible reaction (PK) and one reversible reaction (PPDK). This feature is essential for providing a homoeostatic flux control [65,66] when the equilibrium reaction builds up and then buffers and tunes the concentration of a substrate for a non-limiting non-equilibrium reaction. The homoeostatic flux control enzymes (such as PPDK) become important when the flux is increased, such as glycolysis at low [O2]. The reactions exerting homoeostatic flux control are always present within metabolic cycles; in the Calvin cycle these are catalysed by NADP-GAP (where GAP is glyceraldehyde 3-phosphate) dehydrogenase, transketolase and aldolase [64,65], in the TCA cycle these are catalysed by malate dehydrogenase , aconitase, fumarase  and NADP-isocitrate dehydrogenase .
The importance of PPDK is also evident for developing and germinating seeds which commonly exhibit anaerobic metabolism . A relative increase of abundance of the glycolytic enzymes compared with TCA enzymes in maize endosperm is consistent with the demonstration of anoxic conditions during starch accumulation in the endosperm. Under anoxia, provision of substrates for the PK/PPDK reactions can be achieved by anaplerotic reactions catalysed by PEP carboxykinase (PEP feeding) and NAD-malic enzyme (pyruvate feeding). Martin et al.  have shown that, under anoxia, PEP carboxykinase may be subject to a mechanism of post-translational control that selectively inhibits the carboxylating, but not the decarboxylating, activity. Expression of the PEP carboxykinase gene in anaerobic rice was increased by 100-fold, whereas that of PEP carboxylase decreased . Thus the direction of metabolism in anaerobic conditions is from oxaloacetate to PEP, opposite to the C4 cycle in C4 plants. The NAD-malic enzyme activity is promoted by low pH values and relative insensitivity to higher [NADH], the conditions common to anaerobic tissues.
The PK/PPDK substrate cycle can be a primary engine for PPi biosynthesis under anoxia. Other PPi-producing enzymes are related mostly to biosynthetic reactions and are unlikely to be connected to PPi formation under anoxia. This includes the futile cycle of starch formation/breakdown [13,71] and the unconfirmed putative direct synthesis of PPi by the mitochondrial ATP synthase [72,73]. The latter is unlikely to be physiologically important because it cannot be more efficient compared with ATP synthesis in conditions of low [O2] or other limitations of ATP production. It is also possible that mitochondria contain a unique H+-PPase which is capable of coupling H+ gradient and PPi hydrolysis or synthesis ; however, its capacity cannot be high when mitochondrial electron transport is suppressed.
GLYCOLYSIS OPERATING WITH PPi
In non-photosynthetic tissues, e.g. roots, sucrose is the major imported source of carbon, and its oxidation is the major sou-rce of energy there. Sucrose hydrolysis can be catalysed by invertase and/or SuSy (sucrose synthase). Low [O2] conditions are generally limiting to invertase, but not to SuSy . In fact, for Arabidopsis, the only firmly established requirement for SuSy is under anoxic conditions (flooded roots) . The ‘coupling’ of SuSy and UGPase (UDP-glucose pyrophosphorylase) reactions would result in UTP and glucose 1-phosphate formation from sucrose, UDP and PPi . In general terms, the PPi formed from ADP and Pi through the PK/PPDK reactions (Figure 3) feeds UGPase and PFP reactions, thus saving ATP.
For glycolysis that operates under O2-limiting conditions (Figure 4), the overall reaction is the following:
where Pyr is pyruvate. In this PPi-driven glycolysis, instead of four ATP built upon the oxidation of one sucrose, as many as eight ATP are generated at the expense of three PPi. To make these three PPi, three turns of the PK/PPDK cycle are necessary, in addition to four conversions of PEP into pyruvate corresponding to the ‘normal’ glycolysis. This also means that, for operation of such glycolysis, the PK reaction should operate 1.75-fold faster to provide ATP for the PPDK reaction. Indeed, in anaerobic rice, one PK gene (Os09g22410) is up-regulated 8-fold and another (Os01g16960) 4-fold .
Operation of glycolysis from sucrose to ethanol with ATP and PPi
When PPi is created and its level is high, the PPDK equilibrium can be displaced towards ATP formation and PPDK will work in the direction of pyruvate synthesis from PEP, but using AMP and PPi, instead of ADP used in the reaction catalysed by PK. In this case, the summary equation will be:
In this sequence of reactions, not only ADP but also AMP and PPi will be used for ATP formation. In conjunction with AK (reaction: 2ADP→AMP+ATP), PPDK (reaction: AMP+PPi+PEP→ATP+Pi+Pyr) can be also involved in extra ATP formation  in the following combined reaction equation: 2ADP+PPi+PEP→2ATP+Pi+Pyr, and the gain of ATP in glycolysis will be even higher (three ATP from GAP or 12 ATP from sucrose).
The induction of PPi-dependent cytosolic bypasses may help plants to survive certain stresses by circumventing ATP-limited reactions and actually conserve ATP . A possible excess of PPi could be used by the PPi-dependent proton pumps of the tonoplast , contributing to the avoidance of the cytoplasmic acidosis induced by O2 deprivation. This represents a strategy for biochemical adaptation to anoxia aiming at cell survival by switching to low energy consumption and providing salvage maintenance of major metabolic pathways.
For the equilibration of adenylates under low [O2] conditions, one needs to consider both the AK reaction (mainly in IMSs where it promotes release of magnesium, calcium and other metal ions), and the PK/PPDK substrate cycle which forms AMP and PPi from ADP and Pi. The combined action of AK and the PK/PPDK substrate cycle results in the maintenance of high [Mg2+], [MgPPi] and [AMP] under conditions of low [ATP] and moderate [ADP], with PPi serving as an alternative energy currency. Thus the release of Mg2+ under low [O2] has major consequences for cell energetics, making it less ATP-dependent and more efficient under conditions of low energy supply.
- PPDK pyruvate
The work of our laboratories is supported by the Natural Sciences and Engineering Research Council of Canada (to A.U.I.) and by the Swedish Research Council (to L.A.K.).