Review article

Inositol pyrophosphates: between signalling and metabolism

Miranda S. C. Wilson, Thomas M. Livermore, Adolfo Saiardi


The present review will explore the insights gained into inositol pyrophosphates in the 20 years since their discovery in 1993. These molecules are defined by the presence of the characteristic ‘high energy’ pyrophosphate moiety and can be found ubiquitously in eukaryotic cells. The enzymes that synthesize them are similarly well distributed and can be found encoded in any eukaryote genome. Rapid progress has been made in characterizing inositol pyrophosphate metabolism and they have been linked to a surprisingly diverse range of cellular functions. Two decades of work is now beginning to present a view of inositol pyrophosphates as fundamental, conserved and highly important agents in the regulation of cellular homoeostasis. In particular it is emerging that energy metabolism, and thus ATP production, is closely regulated by these molecules. Much of the early work on these molecules was performed in the yeast Saccharomyces cerevisiae and the social amoeba Dictyostelium discoideum, but the development of mouse knockouts for IP6K1 and IP6K2 [IP6K is IP6 (inositol hexakisphosphate) kinase] in the last 5 years has provided very welcome tools to better understand the physiological roles of inositol pyrophosphates. Another recent innovation has been the use of gel electrophoresis to detect and purify inositol pyrophosphates. Despite the advances that have been made, many aspects of inositol pyrophosphate biology remain far from clear. By evaluating the literature, the present review hopes to promote further research in this absorbing area of biology.

  • bisdiphosphoinositol tetrakisphosphate (IP8)
  • 5-diphosphoinositol pentakisphosphate (IP7)
  • inositol hexakisphosphate (IP6) kinase (IP6K)
  • inositol phosphate
  • metabolism
  • polymer inorganic polyphosphate (polyP)
  • PP-IP5 kinase (PPIP5K)


The inositol phosphate family represents one of the most important signalling families, using myo-inositol as the basic building block. Each of the six hydroxy groups of the inositol ring can be phosphorylated in a combinatorial manner, generating many signalling molecules, the best understood of which is the Ca2+ release factor IP3 (myo-inositol 1,4,5-trisphosphate) [1,2]. Two subfamilies of inositol phosphate exist: the lipid-bound, commonly called PIs (phosphatidylinositols), of which seven species have been identified; and the cytosolic inositol phosphates, extracted from cells using water-based acid solutions and thus referred to as ‘soluble’. The traditional method to resolve the soluble inositol phosphates relies on their differential negative charge. Owing to the different number and position of the phosphate groups, different members of the family elute at different times during SAX (strong anion exchange) chromatography [3]. Peaks eluting after the fully phosphorylated ring of IP6 (inositol hexakisphosphate; also known as phytic acid) were first noted in the early 1990s [47] and were suggestive of a molecule more polar still than IP6. The discovery of inositol pyrophosphates, however, is generally accredited to two papers, from two research groups, both published in the Journal of Biological Chemistry in early 1993 [8,9].

Phytic acid is the precursor of the best characterized inositol pyrophosphates, IP7 (or PP-IP5; 5-diphosphoinositol pentakisphosphate) and IP8 [or (PP)2IP4; bisdiphosphoinositol tetrakisphosphate], that possess, as the name suggests, seven or eight phosphate groups attached to the six-carbon inositol ring, thus possessing one and two pyrophosphate moieties respectively (Figure 1). Although the pyro- nomenclature is now commonly used, it is important to mention that the correct IUPAC definition for this moiety is diphospho-, thus the correct chemical name of IP7 is diphosphoinositol pentakisphosphate (or PP-IP5). The past 20 years have seen this class of molecules as well as related enzymes and gene names referred to in several ways. Table 1 displays these different nomenclatures, indicating the scientific name and the common name used in the present review.

Figure 1 Linear pathway of inositol pyrophosphates synthesis

This pathway depicts the possible routes of inositol pyrophosphate synthesis in cells, starting from myo-inositol. S. cerevisiae exclusively uses the lipid route to synthesize inositol phosphates [119], whereas D. discoideum utilizes the cytosolic route to synthesize IP6 [15], the enzymology of which is not fully elucidated. The interested reader is invited to read the reviews [11,120,121] which describe the synthesis of IP6 in more detail. This Figure highlights the metabolic steps most relevant in vivo, supported by yeast genetic studies [25,26,40,53,119]. There are multiple PP-IP4 isoforms synthesized by IP6K in vitro [(1/3)PP-IP4 and 5PP-IP4 are shown here] [19]. The broken line represents the axis of symmetry through myo-inositol between carbons 2 and 5, making positions 1,3 and 5,6 enantiomeric. P indicates a single phosphate (PO32−) group; R and R′ represent fatty acid chains; broken arrows represent multiple steps of modification that are not shown in detail; and kinases catalysing each step are indicated in red (mammalian) and blue (S. cerevisiae).

View this table:
Table 1 Inositol pyrophosphate nomenclature

(a) Summary of the nomenclature of the inositol pyrophosphates, with the names used in the present review in the left-hand column, the correct IUPAC nomenclature, the alternative names sporadically present in the literature and the identified biological isomers. (b) Details of the mammalian inositol pyrophosphate synthesis genes, including the chromosomal localization of the human isoforms, as well as previous names and the corresponding S. cerevisiae gene name.

Inositol pyrophosphates have the distinctive property of containing high-energy phosphate bonds. The free energy of hydrolysis of the pyrophosphate moiety is similar to that of the high-energy bond found in ATP [9,10]. Like ATP, IP7 and IP8 are found in all eukaryotic cells analysed so far, ranging from fungi to humans. Inositol pyrophosphates control disparate cell biological processes, from regulating telomere length to controlling vesicular trafficking, from insulin secretion to chemotaxy. However, an emerging hypothesis places the inositol pyrophosphates as conserved and fundamental regulators of cellular energy metabolism, enabling their many and varied downstream effects. This new vision will be the driving theme of the present review.


The simple myo-inositol ring forms the basis of the diverse family of inositol phosphates. Existing in the chair formation, this six-carbon sugar molecule has one axis of symmetry, through its 2 and 5 positions (Figure 1). This stereochemistry arises due to the axial nature of the hydroxy group on position 2, whereas the other five hydroxy groups are in the equatorial plane. An important consequence of this is the expansion of the range of phosphorylated states that can result from this building block. With each carbon representing a distinct stereochemical environment, it has been calculated that there are 63 possible varieties of inositol phosphate, even before the pyrophosphate isomers are considered. More than 30 different soluble inositol phosphates have been identified in eukaryotic cells [11].

In the yeast experimental model Saccharomyces cerevisiae, the synthesis of the soluble pool of inositol phosphates begins with the well-studied second messenger IP3 (Figure 1). This molecule is the cleavage product of the lipid PIP2 (phosphatidylinositol 4,5-bisphosphate) by phospholipase C [12,13]. Once liberated from its DAG (diacylglycerol) tail the soluble molecule can be dephosphorylated to myo-inositol, or it can be further phosphorylated to the fully phosphorylated IP6 ring and beyond, to form the inositol pyrophosphates IP7 and IP8 (Figure 1) [8,9]. In the model amoeba Dictyostelium discoideum, the synthesis of the higher phosphorylated forms of inositol does not require phospholipase C activity, and instead occurs by a direct sequential phosphorylation route from myo-inositol to IP6 (Figure 1) [9,15].


Inositol pyrophosphates are largely synthesized from IP6, generating IP7 and IP8. IP6, the most abundant of the inositol phosphates, can reach concentrations of 10–60 μM in mammalian cells and 0.7 mM in D. discoideum [16], whereas IP7 can reach concentrations of 0.5–1.3 μM in mammalian cells [17,18] and 0.3–0.5 mM in the amoeba [17]. Two classes of enzymes are responsible for the synthesis of pyrophosphates: the IP6Ks (IP6 kinases) and the PPIP5Ks (PP-IP5 kinases), Kcs1 and Vip1 respectively in S. cerevisiae. These two distinct classes of enzyme exhibit catalytic activity against different positions on the ring, with IP6Ks placing a phosphate group at the β position of C-5 on the fully phosphorylated IP6 ring [19], whereas the PPIP5Ks phosphorylate position 1 [20,21] (J.D. York, personal communication) (Figure 1). The isomeric form of IP8 present in mammalian cells has, therefore, been identified as (1,5)PP-IP4 [18] (Figure 1). Interestingly, despite similar enzymology, D. discoideum instead contains the (5,6)PP-IP4 isomer of IP8 [10,22], possibly reflecting different metabolism. Although in vitro PPIP5Ks do not act on the principal IP5 (inositol pentakisphosphate) isomer I(1,3,4,5,6)P5 [23,24], the IP6K enzymes are able to phosphorylate it, mainly, but not exclusively, at the enantiomeric positions 1,3, generating (1,3)PP-IP4. This can then be further phosphorylated by IP6K or PPIP5K to (PP)2-IP3 (bis-disphosphoinositol trisphosphate) [25]. Yeast Kcs1 has been shown in vivo to generate inositol pyrophosphate even using IP3 and IP4 (inositol tetrakisphosphate) [26]. It is worthy of mention that in vitro the IP6Ks are also able to form a triphosphate PPP-IP5 form of ‘IP8’ [19]. Therefore taking into account the diverse substrates and the formation of the pyrophosphate moiety at different positions of the inositol ring, inositol pyrophosphates have the potential to become a very large family of molecules.

The IP6Ks belong to the same family of inositol phosphate kinases as the IP3K (IP3 3-kinase) and the IPMK (inositol phosphate multikinase). This family is characterized by the presence of the conserved PxxxDxKxG motif in the inositol-binding region (Pfam family PF03770) [27,28]. Phylogenetic analysis indicates that in fact IP6Ks are likely to be the most ancient members of this family [29]. The IP6Ks were first characterized and cloned in the Snyder laboratory shortly after the identification of IP7 [27,30]. Three mammalian genes, IP6K1, IP6K2 and IP6K3, were identified [31], one of which, IP6K2, had previously been identified as PiUS (Pi uptake stimulator) [27,32,33]. In vitro all three mammalian IP6Ks, as well as the yeast Kcs1, are capable of phosphorylating IP6 to IP7, as well as IP5 to PP-IP4 [31,34,35]. The physiological relevance of these derivatives is much debated and is likely to be highly contingent both on the relative concentrations of IP5 and IP6 in a given cell as well as the substrate specificity of the kinase involved. In yeast the presence of PP-IP4, for instance, is much increased in the ipk1Δ [IP5-2K (IP5-2 kinase); IPPK in mammals] strain that is unable to synthesize IP6 (Figure 1) and thus has elevated levels of IP5. It is also clear that the relative affinity for IP6 over IP5 of a given IP6K varies between organism and isoform; human IP6K1 for instance exhibits a 5-fold difference in Km between IP6 and IP5, whereas IP6K2 has a 20-fold higher affinity for IP6 over IP5 [27].

In 2007 the York laboratory identified the second class of pyrophosphate-producing kinases in the form of the yeast enzyme Vip1 [36]. Two mammalian homologues, PPIP5K1 and PPIP5K2, were also identified [23,37]. These reasonably large proteins, approximately 150 kDa in size, do not belong to the inositol phosphate kinase family. In addition to the kinase domain, they also possess a histidine acid phosphatase-like domain in the C-terminal portion of the protein. It has been postulated that the phosphatase domain is catalytically inactive, and is responsible for the allosteric regulation of PPIP5K by IP6 [38]. However, truncated PPIP5K constructs lacking the phosphatase domain show increased activity when overexpressed [36], indicating that this domain is active. It may antagonize the kinase domain, specifically dephosphorylating 1PP-IP5, the IP7 isomer produced by the kinase domain (J.D. York, personal communication). Although in vitro PPIP5K can promptly phosphorylate IP6 to 1PP-IP5 [39], in vivo this activity is masked by phosphatase activity and cannot be easily detected, suggesting that the major physiological target of the PPIP5Ks is IP7, as also suggested by the kinetic parameters [23]. This is further supported by the rise in levels of IP7 in vip1Δ mutants, whereas levels of IP6 remain unchanged [40].


In kcs1Δ yeast, the inability to detect 1PP-IP5 in vivo could be the result of multiple activities, i.e. by the intrinsic acidic phosphatase domain of PPIP5K, and by the action of the DIPP (diphosphoinositol-phosphate phosphohydrolase) proteins [41]. DIPPs prefer to hydrolyse at position 1; in fact (1,5)PP2-IP4 is hydrolysed in a specific order, first position 1 and then position 5 [42]. Four mammalian enzymes, DIPP1, 2, 3 and 4, have been identified, whereas only one DIPP protein, Ddp1, exists in S. cerevisiae [43,44]. These proteins belong to the large family of Nudix (nucleoside diphosphate linked moiety X) hydrolases, which contain a 23-amino-acid catalytic domain named the MutT motif or nudix box [45]. Nudix hydrolases catalyse the cleavage of the phosphoanhydride bond in substrates that are usually nucleoside diphosphates linked to other molecules generally through phosphodiester bonds. Therefore DIPPs are very promiscuous and are able to degrade inositol pyrophosphates as well as nucleotide analogues such as Ap6A (diadenosine hexaphosphate) [42,46]. More recently, it has been shown that DIPPs also degrade the ubiquitous polyP (inorganic polyphosphate) [25]. Thus the enzymes that degrade inositol pyrophosphate [41] are also capable of acting on polyP [4749], a cellular phosphate reservoir important for the regulation of energetic metabolism, as discussed below. This ability to act on polyP is also shared by the S. cerevisiae homologue Ddp1 [25].


Previous reviews have emphasized the lack of evidence suggesting that inositol pyrophosphates rapidly and dynamically respond to a particular stimulus [50,51]. It is unlikely that inositol pyrophosphates participate in a classical receptor-mediated signalling event, exemplified by phospholipase C activation that leads to the rapid hydrolysis (in seconds) of PIP2, and a dramatic increase in IP3 [12]. However, many experiments have been identified describing modulation of the cellular levels of inositol pyrophosphates [40,5052]. Together the evidence points to a link between inositol pyrophosphates and the cellular metabolic status.

Shortly after the discovery of inositol pyrophosphates it became apparent that, although the cellular concentration remains largely constant in most cells, in mammalian cells at least the turnover of IP7 and IP8 is in fact very rapid [8,52]. The Shears laboratory, using the phosphatase inhibitor fluoride as a metabolic trap, showed that in a variety of mammalian cells 50% of the IP6 pool, as well as 20% of the IP5 pool, were converted into pyrophosphates each hour [8]. Furthermore the IP7 pool in primary hepatocytes was demonstrated to turn over ten times every 40 min, compared with just 10% of the IP6 pool in the same time [52]. Although these observations have led to suggestions that inositol pyrophosphates have a dynamic role with a potential ‘molecular switch’ activity [53], the physiological significance remains unknown, mainly because fluoride is a commonly used protein phosphatase inhibitor and thus has multiple effects on cell signalling. It has been assumed that the stabilizing effect of fluoride reflects its ability to inhibit the activity of DIPP phosphatases [41]. However, fluoride does not elevate the inositol pyrophosphate concentration in yeast, where Ddp1 is also sensitive to fluoride inhibition [24]. Similarly, fluoride does not alter the inositol pyrophosphate level in D. discoideum or plant cells (A. Saiardi, unpublished work), thus the effect of fluoride seems to be specific to mammalian cells, where the mechanism is unknown. Aside from fluoride, a few other treatments have been shown to regulate inositol pyrophosphates in mammalian cells. Induction of apoptosis in ovarian cancer cells by 8 h of cisplatin or staurosporine treatment led to a 4-fold increase in IP7 levels, and a slight increase in IP8 [54]. Inositol pyrophosphate levels also change during the cell cycle. In rat mammary tumour cells, the level of IP7 was twice as high in G1-phase as in the other phases of the cell cycle [55]. Variation in the IP7 level through the cell cycle was also recently seen in yeast, although here IP7 peaked in S-phase rather than G1-phase [56].

In MEFs (mouse embryonic fibroblasts), overnight serum-starvation depleted IP7 levels, which could be restored with 1 h of IGF-1 (insulin-like growth factor 1) treatment [57]. Interestingly, starvation also induces the most dramatic ‘physiological’ change in inositol pyrophosphate levels described so far. The social amoeba Dictyostelium is single-celled while food is available; starvation induces a cAMP-mediated chemotaxic aggregation into a multicellular ‘slug’, followed by further differentiation into a fruiting body. Starvation has been reported to increase IP7 and IP8 levels up to 25-fold within hours of nutrient removal [58,59]. A newer methodology for investigating endogenous unlabelled inositol phosphate levels in the amoeba also found an increase in inositol pyrophosphates during development, but to a much lower level (A. Saiardi and T.M. Livermore, unpublished work). Treatment with cAMP directly has been shown to rapidly elevate IP7 and IP8 levels in Dictyostelium within 60 s, although the cells had been pre-starved for 5 h and treated with caffeine, an adenylate cyclase and PIKK [PI3K (phosphoinositide 3-kinase)-related] family inhibitor that was used to reduce endogenous cAMP synthesis [59].

Nutrient starvation also regulates inositol pyrophosphate levels in yeast. It has been reported that a decrease in extracellular phosphate results in an increase in IP7 in S. cerevisiae [60]. However, the observation of the opposite result [24], i.e. a decrease in IP7 levels after phosphate starvation, is easier to explain as this treatment dramatically decreases ATP levels [61] and therefore IP7 synthesis. Restoring the phosphate concentration in the medium has an immediate effect in restoring ATP synthesis and leads to a rapid increase in inositol pyrophosphate levels [25].

Inositol pyrophosphates as metabolic regulators

Rather than demonstrating a role for inositol pyrophosphates as classical second messengers, these data showing alterations in inositol pyrophosphate levels are more coherent with a view of pyrophosphates as metabolic messengers or energetic sensors. In light of this new definition, the apparent induction of IP7 in response to insulin signalling can be explained as the levels of IP7 returning to baseline after serum-deprivation and therefore low metabolic rate. Similarly, the changes seen throughout the starvation-induced development of Dictyostelium may reflect adaptations to the metabolic state of the starved cells. Hyperosmotic and thermal stress were shown to specifically induce IP8 in mammalian cells, but not in S. cerevisiae [62,63]. This was initially thought to be regulated by the MAPKs (mitogen-activated protein kinases), but further investigation found that the MAPK inhibitors used were actually affecting the energy charge of the cell, and it was this off-target effect that altered the IP8 level [64]. Pharmacological treatment with AICAR (5-amino-4-imidazolecarboxamide riboside), which is processed into an AMP mimic, caused a reduction in IP8 but not IP7 in mouse cells. The increase was independent of AMPK (AMP-activated protein kinase), again suggesting an independent role for inositol pyrophosphates in energy sensing. In mammalian cells, treatment with the Ca2+ regulator thapsigargin caused a slight decrease in IP7 levels [52]. This is now thought to be a consequence of the off-target reduction in ATP concentration subsequently reported for thapsigargin treatment [51].

Support for the concept of inositol pyrophosphates, and particularly the products of IP6K, as energy sensors can be found in the biochemical properties of the inositol phosphate kinases. The usual Km for ATP of the inositol phosphate kinases IP3K, IMPK, IPPK and PPIP5K are all in the micromolar range, approximately 20–100 μM [65]. Conversely, the IP6Ks have a Km for ATP between 1.0 and 1.4 mM [27,30,34], similar to the intracellular ATP concentration. This makes IP6K activity and therefore IP7 synthesis extremely sensitive to fluctuations in the intracellular ATP level. The recent review by Wundenberg and Mayr suggests an elegant ‘energostatic’ model that describes how metabolic regulation involving inositol pyrophosphates may work [65].


To transduce a signal, the mechanism of action used by many members of the inositol phosphate family of molecules is via binding to a particular receptor, such as the IP3 receptor for IP3 [2], or binding to proteins containing specific domains such as PH (pleckstrin homology), PX (phagocyte oxidase homology) or FYVE (for Fab1, YOTB, Vac1 and EEA1) domains [66]. Consequently inositol pyrophosphates may also signal through allosteric interactions with proteins. However, the unique presence of a highly energetic pyrophosphate bond has also opened the possibility of phosphotransfer reactions, covalently modifying the protein targets. The literature contains evidence for both mechanisms of action, which are not mutually exclusive and may coexist in cells (Figure 2).

Figure 2 Proposed mechanisms of action for inositol pyrophosphates

The allosteric mechanism of action (top), i.e. the binding of an inositol pyrophosphate to a protein effector, has been suggested for the S. cerevisiae protein Pho81 [67] and the PH domains of the D. discoideum protein Crac [59] and the mammalian Akt [57]. However, this simple binding mechanism does not make use of the pyrophosphate bond. One can hypothesize that a protein-binding partner may cycle through conformational states by binding and hydrolysis of an inositol pyrophosphate, similar to the regulation of GTPases. Pyrophosphorylation of proteins (bottom) instead takes advantage of the ability of inositol pyrophosphates to participate in phosphotransfer reactions. A pre-phosphorylated serine residue in a negatively charged local environment (shown in red) can become the recipient of the β-phosphate of the pyrophosphate moiety, in the presence of divalent cations [72,73].

The ability of inositol pyrophosphates to allosterically regulate a binding partner has been proposed for PH domain-containing proteins, such as the mammalian growth factor/insulin signalling kinase Akt/PKB (protein kinase B), and Pho81, a yeast CKI [CDK (cyclin-dependent kinase) inhibitor] that regulates the phosphate response. In S. cerevisiae, the cyclin–CDK–CKI complex Pho80–Pho85–Pho81 is involved in sensing phosphate availability. The Vip1-generated IP7, 1PP-IP5, was found to specifically bind to and inhibit this complex, whereas IP6 or 5PP-IP5, the dominant IP7 species in most cells, had no effect [67]. No mammalian homologues for these Pho proteins have been discovered, so this may be a strictly S. cerevisiae-specific regulatory mechanism, which is not conserved even in the fission yeast Schizosaccharomyces pombe [68].

PH domain binding

The majority of work in this area has been to determine whether inositol pyrophosphates can bind to PH domain-containing proteins whose main ligand is the phosphoinositide lipid PIP3 (phosphatidylinositol 3,4,5-trisphosphate). In vitro IP7, and to a lesser extent its precursor IP6, were able to compete with PIP3 or labelled IP4 (the PIP3 head group) for binding to the recombinant PH domains of several proteins, including the D. discoideum chemotaxis protein Crac (cytosolic regulator of adenylate cyclase) and mammalian Akt [59]. Because the concentration of inositol pyrophosphates is unusually high in Dictyostelium, it was suggested that the physiological regulation of PH domain proteins by inositol pyrophosphates would be restricted to this amoeba. However, it has previously been reported that mammalian PDK1 (3-phosphoinositide-dependent protein kinase 1)-mediated phosphorylation of Akt is strongly inhibited by IP7 with an IC50 of 1 μM in the presence of PIP3, whereas the IC50 of IP7 inhibition decreases to an astonishing 20 nM in the absence of PIP3 [57]. The authors propose that the direct binding of IP7 to the PH domain of Akt is responsible for this inhibition [57]; however, appropriate binding assays were not reported. Because the cellular IP7 concentration is in the range 0.5–1.3 μM, cytosolic Akt must then be constitutively bound and inhibited by IP7. If this is true, Akt needs to release IP7 before its activation by PIP3–PDK1 can take place at the plasma membrane.

Detailed inositol pyrophosphate binding studies were performed with PDK1, which possesses an atypical PH domain containing a large binding pocket. PDK1 was found to interact with IP7 with a considerably lower affinity than that of IP6. Owing to the much higher abundance of IP6 compared with IP7 in mammalian cells, it is unlikely that IP7 is a physiological ligand for PDK1 [69]. The affinity of PDK1 for IP6 may function to retain a pool of this protein in the cytosol. A subsequent study by the same laboratory failed to identify any significant binding between the PH domain of Akt and IP7. Using a FRET (fluorescence resonance energy transfer)-based assay the authors concluded that IP6 and IP7 are not able to displace PIP3 from Akt pre-incubated with the lipid [70]. Despite much work in this area, this aspect of inositol pyrophosphate activity remains controversial, with the main conceptual problems based on structural considerations. Structural analysis of the PH domain from β-spectrin bound to IP3 revealed that binding specificity is determined by the phosphate groups in positions 4 and 5 [71]. It is difficult to envision how 5PP-IP5, with a pyrophosphate moiety in position 5, would fit into a binding pocket designed for only one phosphate at that position. These studies do not paint a clear picture of inositol pyrophosphates acting primarily as ligands of PH domain-containing proteins. Further work will be required to better define, structurally and functionally, the binding pocket recognizing inositol pyrophosphates. It is not clear if the many diverse inositol pyrophosphate species would require specific receptors, and if not, without a specific binding site for each pyrophosphate moiety, how specificity between the pyrophosphate and the precursor would be obtained.

Protein pyrophosphorylation

The published allosteric mechanisms so far also show a lack of target conservation between humans, yeast and amoeba, at odds with the near ubiquitous evolutionary appearance of IP7. It is possible that the primary mechanism of action of inositol pyrophosphates is not binding, but instead relies on the ability of the energetic pyrophosphate group to donate its β-phosphate to acceptor molecules. It has long been suggested that inositol pyrophosphates may act as phosphorylating agents, due to the high free energy of hydrolysis of their pyrophosphate bond(s), which in 1PP-IP5 is comparable with that of ADP [9,10]. This was first tested in vitro using 32P-labelled IP7. When added to eukaryotic cell extracts from yeast, mouse and flies, labelled proteins were obtained [72]. This was enhanced with fluoride but reduced by treatment with λ-phosphatase. The authors [72] focussed on one candidate protein, Nsr1, a nucleolar yeast protein, although the mammalian nucleolar proteins Nopp140 and TCOF1 were also able to be pyrophosphorylated by labelled IP7. Surprisingly, the transfer of the 32P-labelled β-phosphate to Nsr1 was found to be non-enzymatic and did not require any known yeast kinases. However, this modification is not untargeted, and is thought to require serine residues within an acidic region of primary structure. Importantly, endogenous labelling of tagged Nsr1 in yeast given labelled Pi was much reduced in the kcs1Δ mutant, suggesting that IP7 does phosphorylate Nsr1 in vivo.

A follow-up study demonstrated that IP7 was in fact acting on pre-phosphorylated serine residues, resulting in pyrophosphorylation [73]. Other inositol pyrophosphates, 1PP-IP5 and IP8, were found to have similar pyrophosphorylating ability as IP7. The free energy of hydrolysis of the pyrophosphate bond in IP8 is greater than that of IP7 and so this molecule may well be more efficient at modifying proteins, although in vivo IP7 is more abundant. The serine pyrophosphorylation was found to be more acid-labile than traditional serine phosphorylation by ATP, but more resistant to phosphatases. In fact, no protein pyrophosphatase has yet been identified. The physiological effects of pyrophosphorylation by IP7 and other inositol pyrophosphates are not fully understood, although the occurrence of this modification is conserved and has, so far, been demonstrated in both mammalian [74,75] and zebrafish (A. Saiardi, unpublished work) cells. Two studies have been published investigating how pyrophosphorylation may modulate protein activity in yeast and human cells [74,75]. Both demonstrated that pyrophosphorylation mediates and inhibits protein–protein interactions. In these investigations, pyrophosphorylation resulted in regulation of yeast glycolysis, and the release of HIV-Gag virus-like particles from HeLa cells.

Like the allosteric mechanism of action, the protein pyrophosphorylation mechanism also lacks the support of strong experimental evidence, proving that this novel post-translational protein modification occurs in vivo. The strongest evidence indicating that inositol pyrophosphates covalently modify proteins in vivo relies on the fact that many substrates of in vitro phosphorylation by IP7 show an upward gel-mobility shift in extracts from wild-type compared with kcs1Δ yeast [74]. However, no direct biophysical evidence of this modification exists in the literature, thus further work is required to elucidate the exact nature of the covalent modification driven by pyrophosphates in vivo.


Inositol pyrophosphates have been implicated in a variety of diverse activities, from apoptosis and telomere maintenance to trafficking and embryonic development [53,54,76,77]. This diversity and wide range of activities underlines the biological importance of this class of molecules; until recently, it was unclear whether the inositol pyrophosphates regulated each reported cellular activity independently, or if they were controlling some central basic function that sits upstream of the regulated phenotypes. First by theoretical considerations [29], then subsequently on the basis of published results, it has been proposed that inositol pyrophosphates have a role as ‘metabolic messengers’ [51] or ‘regulators of cell homoeostasis’ [65]. The study by Szijgyarto et al. [75] indeed suggests that inositol pyrophosphates are master regulators of cellular energetic metabolism. In this section, we will summarize the literature so far regarding the function of inositol pyrophosphates, with a focus on this concept of metabolic regulation (Figure 3).

Figure 3 Inositol pyrophosphates: the role in metabolism

Primary metabolic influences exerted by inositol pyrophosphates (shown as IP7). The Figure shows three, probably connected, features that link inositol pyrophosphates to metabolism at the molecular (phosphate homoeostasis), cellular (energetic) and organismal (insulin signalling) levels.

As described above, the synthesis of inositol pyrophosphates is tied to ATP levels due to the high Km for ATP of the IP6K enzymes. Surprisingly, cells with no or a decreased amount of inositol pyrophosphates, such as kcs1Δ yeast and IP6K1−/− MEFs, were found to have a greatly increased cellular ATP concentration, whereas ADP and AMP levels were slightly reduced, leading to an overall increase in the adenine nucleotide pools (adenylate pool) and more importantly to an overall increase in AEC {adenylate energy charge; calculated by ([ATP]+0.5[ADP])/([ATP]+[ADP]+[AMP])} [75]. The AEC corresponds to the saturation of phosphodiesteric bonds of the adenylate pool, and thus represents the amount of metabolically available energy stored in the adenine nucleotide pool [78]. Conversely, increasing the amount of IP7 in wild-type yeast by expressing mammalian IP6K1 caused a decrease in ATP concentration. The yeast system was studied in more detail and it was discovered that protein pyrophosphorylation was influencing the interactions of the essential yeast glycolytic transcription factors Gcr1, Gcr2 and Rap1 [79,80]. Pyrophosphorylation of Gcr1 reduced its ability to bind Gcr2, thereby inhibiting the transcription of genes encoding glycolytic enzymes. Accordingly, the kcs1Δ yeast consumed glucose more quickly than wild-type yeast [75].

Surprisingly, given the increase in cellular ATP concentration, the mitochondria were found to be defective in the kcs1Δ yeast. Yeast usually generate energy and intermediates through fermentation, but can be induced to perform mitochondrial metabolism/oxidative phosphorylation by providing only a non-fermentable carbon source such as glycerol; the kcs1Δ yeast were unable to grow on this medium. That yeast preferentially perform fermentation marks them out compared with higher eukaryotes, and the Gcr1 and Gcr2 proteins are not found in mammals. However, the increased ATP level and reduced oxygen consumption of IP6K1−/− MEFs compared with wild-type MEFs demonstrates that the inositol pyrophosphate regulation of fundamental cellular metabolism is evolutionarily conserved. ATP availability, along with NAD/NADH levels, which were also altered, underpins virtually all other cellular activity.

Inositol pyrophosphates also regulate D. discoideum chemotaxis [59]. In this amoeba starvation and subsequent cAMP signalling and chemotaxis causes development into a multicellular fruiting body. The D. discoideum ip6k knockout was more sensitive to cAMP and aggregated more quickly than wild-type cells in response to starvation [59]. It was suggested that this phenotype reflected competition in the wild-type cells between IP7 and PIP3 for binding to the PH domain of Crac, a protein linked to cAMP-dependent chemotaxis. As discussed above, the evidence describing the ability of inositol pyrophosphates to compete with lipid inositols for PH domain binding is conflicting. In the case of Dictyostelium aggregation, PIP3 is unlikely to be the chief regulatory molecule: deletion of all five PI3Ks capable of generating PIP3 did not prevent chemotaxis to cAMP [81]. However, if IP7 affects energetic metabolism and the adenine nucleotide level in this system as in yeast, the high level of ATP may lead to an increased cAMP concentration and subsequently the faster chemotaxic response observed in the D. discoideum ip6k-null strain.

IP6K−/− mouse characterization

Further work has indicated a role for inositol pyrophosphates in metabolic regulation not only at the cellular, but also at the organismal level. Overexpression of IP6K1 in pancreatic β-cells or direct addition of IP7 resulted in an increase in insulin vesicle membrane fusion, i.e. exocytosis [82], whereas knockdown of IP6K1 by RNAi (RNA interference) reduced exocytosis in these cells. This finding was supported by the IP6K1−/− mouse phenotype: it has reduced plasma insulin compared with the wild-type mouse [83]. However, the mutant mouse also displays insulin hypersensitivity, responding more strongly to lower concentrations of insulin [57]. The IP6K1−/− mouse is slightly smaller than the wild-type mouse on a normal diet, owing to a reduction in fatty tissue levels. Even when the mice were fed a high-fat diet, the knockout mouse gained much less weight, resulting in a phenotype more similar to the wild-type mice on a control diet [57]. The defective mitochondria in yeast and IP6K1−/− MEFs, and the requirement of functional mitochondria to synthesize the intermediates for fatty acid synthesis [84,85] could explain the inability of IP6K1−/− mice to gain weight, and provides strong evidence supporting a role for inositol pyrophosphates as regulators of metabolism at the whole organism level. This was explored in Wundenberg and Mayr's recent review [65], providing a theoretical framework to understand the possible regulatory structure behind these observations. Their suggested cellular ‘energostat’ model, with interaction between pyrophosphate levels and AEC, was coupled to the organismal ‘glucostat’ regulation, i.e. the established regulation between insulin and plasma glucose concentration.

The IP6K2−/− mouse did not have defects in insulin signalling, implying that there is incomplete redundancy between the isoforms [86]. This mouse is resistant to ionizing radiation, and its fibroblasts were found to show an up-regulation in DNA repair as well as being resistant to IFN-β (interferon-β) [86]. Conversely, overexpression of IP6K2 sensitized various human carcinoma cells to apoptosis induced by conditions including γ-irradiation, IFN-β, etoposide and cisplatin treatment [54,87,88]. Endogenous IP6K2 expression was induced by IFN-β, and IP6K activity was increased by cisplatin treatment [54] in ovarian cancer cells. Although individual overexpression of IP6K1, IP6K2 and IP6K3 promoted apoptosis, only knockdown of the IP6K2 isoform in HEK (human embryonic kidney)-293 cells increased resistance to staurospaurine [54]. This suggests that IP6K2 may be the most important isoform in cell death regulation by inositol pyrophosphates. The carcinogen 4-NQO (4-nitroquinoline 1-oxide) was more effective at causing squamous cell carcinoma in IP6K2−/− mice than wild-type [86]. Protein–protein interactions between IP6K2 and cell death/survival regulatory proteins have also been discovered. Specifically, IP6K2 is an interacting partner of HSP90 (heat-shock protein 90) and p53; these interactions affect the functionality of these two famous proteins directly, with an unknown mechanism independent of inositol pyrophosphate synthesis [89,90]. In zebrafish, which have two known IP6K isoforms, knockdown of IP6K2 resulted in developmental defects through alteration of Hedgehog signalling [91].

Aging, oxidative stress and DNA repair

Metabolism and aging are intimately linked [92,93]. Levels of both IP6 and IP7 were greatly increased in hepatocytes from 10-month-old wild-type mice compared with 2-month-old mice [57]. Further research into possible connections between inositol pyrophosphates and aging would be interesting. One potential link has already been discovered via telomere length. Telomeres are the repetitive DNA sequences that cap chromosomes, preventing degradation. Telomeres shorten through repeated cell division, and there is some evidence that shorter telomeres are associated with organismal aging in mammals [94]. The yeast kcs1Δ strain has been found to have slightly longer telomeres than wild-type yeast; conversely, ipk1Δ yeast, which have an altered inositol pyrophosphate profile with high levels of PP-IP4, have shorter telomeres [76,77]. These effects were dependent on the known telomere regulator Tel1, a PIKK family member and the yeast homologue of ATM (ataxia telangiectasia mutated) [95], suggesting that inositol pyrophosphates are physiological inhibitors of Tel1. In support of this, yeast cell lines lacking inositol pyrophosphates, such as kcs1Δ, were found to be resistant to caffeine and wortmannin, both used as inhibitors of PIKKs, including Tel1 [76].

As well as telomere maintenance, Tel1/ATM and two other closely related PIKKs, Mec1/ATR (ATM- and Rad3-related) and DNA-PKcs (DNA-dependent protein kinase, catalytic subunit) are importantly involved in DNA damage signalling/DNA repair [96]. With Ku (XRCC5/6), DNA-PKcs forms the DNA-PK complex required for the DNA double-strand break repair mechanism non-homologous end-joining, the predominant double-strand break repair mechanism used in mammalian cells. As discussed previously, IP6 has been found to promote Ku-dependent DNA end-joining [97,98]. It is possible that inositol pyrophosphates can also influence this activity.

Yeast, on the other hand, are known to have a high rate of the double-strand break repair mechanism HR (homologous recombination) compared with mammalian cells. Several mutations have been found that artificially increase the rate of HR, leading to excessive deregulated recombination known as hyper-recombination, associated with genomic instability. One such mutation is that of Pkc1 (protein kinase C1). Knockout of KCS1, and therefore a reduction in IP7 and IP8, in the mutant pkc1–4 background restores the normal rate of HR [99,100], hence the Kcs1 name, kinase c suppressor 1. The mechanism behind this is unclear; however, recent evidence has demonstrated that inositol pyrophosphates can also regulate DNA repair in higher organisms [101].

The kcs1Δ yeast strain was found to be more sensitive to the DNA-damage-inducing compound phleomycin than wild-type yeast [40]. Conversely, the kcs1Δ and vip1Δ yeast strains were resistant to cell death and DNA mutation caused by hydrogen peroxide treatment. This was linked to up-regulated activity of the checkpoint kinase Rad53 (CHK2 in humans), and is consistent with the finding that HeLa cells overexpressing IP6K2 were more sensitive to hydrogen peroxide [54]. The difference in response between phleomycin and hydrogen peroxide in yeast may reflect that endogenous hydrogen peroxide, normally generated by mitochondrial activity, is likely to be decreased in the kcs1Δ strain, which could alter the threshold of sensitivity for exogeneously added hydrogen peroxide.

Interestingly, IP6K1 was found to be redox-sensitive, and hydrogen peroxide treatment of wild-type yeast resulted in a reduction in IP7 levels [40]. It would be very interesting to investigate whether IP6Ks and inositol pyrophosphate synthesis is regulated by redox mechanisms under physiological conditions, and if ROS (reactive oxygen species) can regulate this class of enzyme by covalently modifying their cysteine and methionine residues. Although inositol pyrophosphate levels are necessarily regulated by ATP availability, this would suggest further metabolic feedback regulation.

Phosphate metabolism and polyP

Another apparently conserved link between inositol pyrophosphates and metabolism is in phosphate homoeostasis [49]. IP6K2 was first cloned as a protein that stimulated Pi uptake in Xenopus oocytes [27,32,33]. The ability of inositol pyrophosphates to regulate the cellular entry of Pi is conserved: the kcs1Δ yeast mutant displays reduced phosphate uptake [72]. Although in S. cerevisiae the Pho regulon appears to be inhibited by 1PP-IP5 [60], as discussed above, alteration of phosphate metabolism affects KCS1 transcription, thus 5PP-IP5 synthesis but not VIP1 transcription [102], indicating a complex regulatory network.

The analysis of yeast mutants unable to synthesize inositol pyrophosphates has revealed a correlation between a lack of inositol pyrophosphates and a dramatic decrease in the cellular level of polyP [24,103]. The polyP polymer contains tens to hundreds of phosphate residues linked by the same ‘high-energy’ phosphoanhydride bonds found in IP7 and ATP [47]. Many specific biological functions have been attributed to polyP [48], but due to its intrinsic chemical structure, this polymer mainly represents a phosphate buffer that is synthesized and degraded as a function of the phosphate needs of the cell. Key to ATP synthesis is cellular phosphate homoeostasis, which may well be controlled by inositol pyrophosphates through regulation of polyP metabolism. Inositol pyrophosphates not only regulate phosphate metabolism at the intracellular level, but also regulate phosphate homoeostasis of the entire organism. In human serum, the phosphate level is tightly regulated and its level has been genetically linked to a single nucleotide polymorphism in the IP6K3 promoter [104].

Yeast accumulate a large amount of polyP in the vacuole, and it is intriguing that one of the phenotypes of the kcs1Δ yeast is small fragmented vacuoles [34,53], suggesting vesicular trafficking defects. Previous studies identified several trafficking proteins as potential IP7-binding partners, where IP6 and IP7 were found to inhibit clathrin assembly. Unfortunately, these early binding studies were performed with non-physiological salt concentrations [16]. Previously, a different connection between the clathrin adaptor AP3 (adaptor protein 3) complex and IP7 has been identified in HeLa cells. AP3B1 (β-1 subunit of the AP3 complex) is a target of pyrophosphorylation that inhibits a protein–protein interaction between AP3B1 and the motor protein KIF3a [74]. Release of Gag-containing virus-like particles was used as a model to study AP3B1–KIF3a-dependent trafficking: cells with decreased IP7 levels released more virus-like particles, indicating that pyrophosphorylation inhibits the AP3B1–KIF3a interaction in cells [74].

Autophagy and TOR (target of rapamycin)

Overexpression of IP6Ks in HEK-293 cells resulted in inhibition of mTOR (mammalian TOR), and an increase in autophagy thought to lead to cell death [105,106]. It would be interesting to test this in more detail in the context of ATP regulation and AMPK activity. Whether autophagy promotes survival or cell death remains controversial. In yeast, deletion of Kcs1 attenuated autophagy caused by nitrogen starvation [107]. This was associated with defects in autophagosome biogenesis. Given the noted involvement of inositol pyrophosphates in cellular metabolism, it is important to remember that TOR is a crucial regulator of both metabolic/energy stress, via AMPK in mammalian cells, and amino acid availability [108,109].

The inhibition of mTOR by IP6K is also interesting in the context of ribosome biogenesis and rRNA processing. TOR is known to up-regulate ribosome biosynthesis in response to nutrient availability by promoting rRNA synthesis and transcription of ribosomal proteins [110]. The nucleolus plays an important part in this process. Notably, several targets of pyrophosphorylation discovered so far, Nsr1, Srp40/Nopp140 and Tcof1, are nucleolar proteins [72,73]. However, the functional consequences of pyrophosphorylation here are unknown. Mutation of KCS1 was found to rescue cells from a temperature-sensitive mutation of the essential nucleolar protein Rrs1 [111]. These data suggest that inositol pyrophosphates may regulate ribosome biogenesis and thus one of the most energy-intensive cellular processes, essential to growth and cell division [110,112].


After 20 years of research, consensus is beginning to place inositol pyrophosphates as key regulators of cellular metabolism (Figure 3) [57,75]. In the present review we have discussed the major evidence for this, from the level of the cell, with the increase in ATP and AEC in yeast and MEFs lacking inositol pyrophosphates, to the organismal level and the impaired weight gain and insulin hypersensitivity of mice with reduced levels of inositol pyrophosphates. We have also attempted to re-interpret some experimental data in light of this hypothesis. In fact, the wide variety of processes in which inositol pyrophosphates have been implicated, from telomere maintenance to chemotaxis, can be best explained by these molecules acting as basic regulators of metabolism. When considered in this way, it is clear how inositol pyrophosphates are able to exert their influence so widely.

This convergence of opinion [29,51,65] must drive and focus future research. The development of new experimental tools is essential to reveal how, exactly, this control is exerted. The recent application of PAGE [25,113] for analysis of inositol pyrophosphates is already facilitating new lines of inquiry. Further innovation, with the help of our colleagues in organic chemistry, should target production of specific inositol kinase inhibitors as well as non-hydrolysable IP7. Although IP6K inhibitors such as TNP [114] are commercially available, their specificity for IP6K over other members of the inositol phosphate kinase family, particularly IPMK, remains in question. Better inhibitors will allow the reversible depletion of IP7 in cells. Meanwhile, non-hydrolysable forms of IP7 [115,116], where the pyrophosphate mimic cannot be hydrolysed, will facilitate binding studies and help to further demonstrate the requirement of phosphotransfer for the action of pyrophosphates.

In the 20 years since their discovery, we have amassed a good deal of knowledge on the structure, synthesis and role of these molecules, and yet the field remains in relative infancy. It is worthy of note that study of the ‘PI effect’ [117], which led to the discovery that IP3 acts as a second messenger in the mobilization of Ca2+ stores, was already in its 30th year before that landmark discovery was made [118]. Inositol pyrophosphates have been known for 20 years, but we are yet to reveal the full extent of their secrets. We look forward to the next 20 years with interest.


This work was supported by the Medical Research Council funding of the Cell Biology Unit, University College London, London, U.K.


We thank Dr Antonella Riccio and Dr Cristina Azevedo for reading the paper before submission and the members of the Saiardi laboratory for entertaining discussions. We are grateful to Dr John York for sharing unpublished work.

Abbreviations: AEC, adenylate energy charge; AMPK, AMP-activated protein kinase; AP3, adaptor protein 3; AP3B1, β-1 subunit of the AP3 complex; ATM, ataxia telangiectasia mutated; ATR, ATM- and Rad3-related; CDK, cyclin-dependent kinase; CKI, CDK inhibitor; Crac, cytosolic regulator of adenylate cyclase; DIPP, diphosphoinositol-phosphate phosphohydrolase; DNA-PKcs, DNA-dependent protein kinase, catalytic subunit; HEK, human embryonic kidney; HR, homologous recombination; IFN-β, interferon-β; IP3, myo-inositol 1,4,5-trisphosphate; IP4, inositol tetrakisphosphate; IP5, inositol pentakisphosphate; IP6/PP-IP4, inositol hexakisphosphate; IP7/PP-IP5, 5-diphosphoinositol pentakisphosphate; IP8/(PP)2IP4, bisdiphosphoinositol tetrakisphosphate; IP3K, IP3 3-kinase; IP6K, IP6 kinase; IPMK, inositol phosphate multikinase; IPPK/IP5-2K, IP5-2 kinase; MAPK, mitogen-activated protein kinase; MEF, mouse embryonic fibroblast; mTOR, mammalian target of rapamycin; Nudix, nucleoside diphosphate linked moiety X; PDK1, 3-phosphoinositide-dependent protein kinase 1; PH, pleckstrin homology; PI, phosphatidylinositol; PI3K, phosphoinositide 3-kinase; PIKK, PI3K-related; PIP2, phosphatidylinositol 4,5-bisphosphate; PIP3, phosphatidylinositol 3,4,5-trisphosphate; PKB, protein kinase B; polyP, inorganic polyphosphate; PPIP5K, PP-IP5 kinase; TOR, target of rapamycin


View Abstract