Two families of enzymes collaborate to synthesize this group of molecules (Fig. 1): the IP6Ks (E.C. 2.7.4.21, Kcs1 in yeasts; (Saiardi et al., 1999)) and the PPIP5Ks (E.C. 2.7.4.24, Vip1 in yeasts; (Mulugu et al., 2007)). The Kcs1/IP6K family (Saiardi et al., 2001) are responsible for adding an extra phosphate to the 5-position of three different substrates: Ins(1,3,4,5,6)P₅, InsP₆ and 1/3-PP-InsP₅ (Draskovic et al., 2008; Lin et al., 2009). These particular kinases possess a PxxxDxKxG catalytic domain, and a remote, catalytically-essential SSLL (or similar) tetrapeptide. Vip1 (Mulugu et al., 2007) and the two mammalian PPIP5Ks (Choi et al., 2007; Fridy et al., 2007) add a 1/3-diphosphate to both InsP₆ and 5-InsP₆ (Lin et al., 2009). The uncertainty concerning the precise placement of the 1/3-diphosphate group reflects an analytical impediment created by the axis of symmetry in the inositol ring. That is, 1-PP-InsP₅ and 3-PP-InsP₅ are an enantiomeric pair, and a stereoselective technique is needed in order to distinguish between them. A recent molecular modeling study tentatively proposed that the Vip1/PPIP5Ks are 3-kinases (Thorsell et al., 2009); on the contrary, some unpublished data that were cited in a recent review (Best et al., 2010) suggest a 1-kinase specificity. The placement of this diphosphate at the 1-position in Fig. 1 is completely arbitrary and should not be taken as indicative of any preference that we might have about which one of the two alternative proposed structures is correct. Unlike the IP6Ks, the PPIP5Ks do not phosphorylate Ins(1,3,4,5,6)P₅ (Choi et al., 2007).

The mammalian PPIP5Ks are veritable behemoths of around 120-160 kDa (Choi et al., 2007; Fridy et al., 2007). The PPIP5Ks utilize an ATP-grasp catalytic domain which has been mapped to just the N-terminal one third of these proteins (Mulugu et al., 2007). What, then, is the significance of the other two-thirds of these proteins, including an intriguing, catalytically-inactive phosphatase-like domain? It's a fair bet that the answer(s) to that question will illuminate some important cell-signaling functions of the PPIP5Ks.

A variety of tri- and putative tetra-phosphate products have been shown to be produced in vitro upon the prolonged incubation of either the Kcs1/IP6Ks or the Vip1/PPIP5Ks with their respective substrates (Draskovic et al., 2008; Losito et al., 2009; Saiardi et al., 2000). Some of these products have also been observed when the relevant kinases are over-expressed in yeasts (Draskovic et al., 2008). We have not included these molecules in Fig. 1, out of concern that their detection in intact cells relies upon kinase over-expression, and so we consider that their physiological significance remains to be proven. It has been proposed (Draskovic et al., 2008) that the cellular levels of these particular polyphosphates have been underestimated due to their instability, both in the acidic conditions used to quench cells and also at the low pH typically employed for HPLC analysis. However, we have not found that to be the case with the techniques in routine use in our laboratory (unpublished data).

When the diphosphoinositol polyphosphates were first discovered, attention was drawn to their rapid rates of turnover through kinase/phosphatase substrate cycles (Menniti et al., 1993; Stephens et al., 1993b). Indeed, the family of phosphatases (DIPPs in mammals, E.C. 3.6.1.52, Ddp1 in yeast) that degrade the polyphosphates exhibit values for their specificity constants (kcat/Km) that are sufficiently large as to approach the theoretical limit, as defined by diffusion-controlled encounter between enzyme and substrate (Caffrey et al., 2000; Safrany et al., 1998a). While the rapid turnover of the diphosphoinositol polyphosphates is surely functionally important - because otherwise it would be a wasteful ongoing expenditure of cellular energy - its significance has yet to be explained. Each of the DIPPS can remove a diphosphate group from all of the diphosphoinositol polyphosphates (Caffrey et al., 2000; Hua et al., 2003; Leslie et al., 2002; Safrany et al., 1998a). The Human Genome Organization Gene Nomenclature Committee (HGNC) has allocated the DIPPs to the NUDT gene family (NUDT = “Nudix [Nucleoside Diphosphate attached moiety ‘x’]-Type motif), in honor of their eponymous catalytic site (Gx₅Ex₅[UA]×REx₂EExGU, or similar, in which U is an aliphatic, hydrophobic residue (McLennan 2006)).

It is unclear what are the relative rates of flux in vivo through the two complementary pathways of (PP)₂-InsP₄ synthesis (see Fig. 1). It is known that steady-state levels of 5 PP-InsP₅ greatly exceed those of 1/3-PP-InsP₅ (Albert et al., 1997), but that observation by itself is not indicative of relative metabolic fluxes. Some kinetic data obtained from assays with the recombinant kinases (Choi et al., 2007), and some pharmacological data obtained from intact cells (Padmanabhan et al., 2009), together are consistent with 5-PP-InsP₅ being the most important intermediate. However, the kinetic data do not take into account the possibility that subcellular compartmentalization of substrates and/or enzymes might occur in vivo. Also, the DIPPs may have hindered earlier attempts to determine which of the synthetic pathways is preferred in vivo (Padmanabhan et al., 2009), especially if the phosphatases show asymmetric activities against the two diphosphate groups in (PP)₂-InsP₄, as has been proposed (Shears et al., 1995).

For example, the levels of diphosphoinositol polyphosphates increase dramatically in Dictyostelium in response to starvation (Luo et al., 2003), although the significance of this adaptation is unknown. Slime molds are generally considered not to be a representative model organism for this field of research: this organism accumulates >100-fold higher levels of the diphosphoinositol polyphosphates than do yeast and mammalian cells, and the isomers are different; the 4/6,5-isomer of (PP)₂-InsP₄ is unique to Dictysotelium (Laussmann et al., 1996). Hence the current consensus (Burton et al., 2009) that whatever actions these polyphosphates are accomplishing in slime molds, they probably are not applicable to other eukaryotes.

Nevertheless, cellular stress is a recurring theme in studies into changes in the cellular levels of diphosphoinositol polyphosphates. In mammalian cells, (PP)₂-InsP₄ synthesis is up-regulated in cells subjected to hyperosmotic pressure (Pesesse et al., 2004). The general biological relevance in mammals of cells being exposed to osmotic stress has not always been appreciated. Instead it has been a popular opinion for much of the previous 70 years (Bourque et al., 1997; Darrow et al., 1935) that metazoan cells, with the obvious exception of those of the renal medulla, are largely protected from the potentially deleterious effects of anisosmotic gradients, by virtue of their being bathed in an osmotically stable extracellular fluid. However, it takes as little as 25-50 mOsM hyperosmotic stress to activate (PP)₂-InsP₄ synthesis in mammalian cells (Pesesse et al., 2004). Recent research has uncovered increases in extracellular osmolarity that can exceed 100 mOsM in the extracellular environment bathing airway epithelial cells, lymphocytes and the various cell-types in bones and cartilage (Alfieri et al., 2007; Go et al., 2004; Knothe Tate 2003). For example, lymphocyte development depends upon their ability to adapt to the hyperosmotic environment of the thymus (Go et al., 2004). Bones also provide an osmotically-stressful environment for cells: there are fluctuating osmotic pressure gradients between osteocytes and their surrounding proteoglycan filled lacunocanalicular system which are important for mechanochemical coupling and for driving bulk fluid flow through the bone tissue (Knothe Tate 2003). Hyperosmotic stress can also occur in airway epithelial cells when the composition of the airway surface liquid layer is compromised by inadequate airway humidification, such as during rapid breathing (e.g. during exercise), breathing of dry/cold air, and in some airway diseases (Song et al., 2001). More generally, even in cell types that do not routinely experience any of these significant fluctuations in extracellular osmolarity, it is now accepted that mechanisms must be in place to adapt to the alterations in intracellular osmolarity (50-100 mOsM) that inevitably accompany normal cellular activities: ion-transport across the plasma membrane, uptake and release of sugars and amino-acids, and polymerization/depolymerization of macromolecules such as glycogen and proteins (Schliess et al., 2002). It is our hypothesis that (PP)₂-InsP₄ plays some as yet undefined role in the mechanisms by which cells adapt to the inherent dangers of osmotic stress, which include strain upon the cytoskeleton, perturbation of chromatin structure, DNA damage, and inhibition of DNA repair (Chiasson et al., 2003; Kültz et al., 2001).

Thermal stress - both heat and cold - can elevate cellular levels of (PP)₂-InsP₄ (Choi et al., 2005). External organs, such as skin and testis, must be especially adaptable to substantial variations in environmental temperature. Potentially deleterious or even lethal elevations in core temperature can occur even in euthermic organisms, due to fever, environmental stress and exertional heat illness (Sonna et al., 2002; Sonna et al., 2004). Such conditions can increase the rate of protein degradation, there can be cytoskeletal disruption, and damaging alterations in membrane permeability and ion homeostasis (Dorion et al., 2002; Seno et al., 2004; Sonna et al., 2002). We (Choi et al., 2005) have suggested that diphosphoinositol polyphosphates might mediate adaptive responses to these consequences of thermal stresses.

The molecular mechanisms that maintain energy homeostasis are a fundamental necessity for cell survival (Hardie 2004). Some recent developments ((Bennett et al., 2006; Choi et al., 2008; Lee et al., 2007), for example) have led to the suggestion (Shears 2009) that diphosphoinositol polyphosphates might couple signaling pathways to the energetic status of the cell. This hypothesis envisages that the levels of PP-InsP₅and (PP)₂-InsP₄ in a cell are directly tied to its metabolic health. For example, we have found that (PP)₂-InsP₄ synthesis is inhibited by an elevation in cellular [AMP] (Choi et al., 2008), which is a sentinel for bioenergetic stress (Hardie 2004). (PP)₂-InsP₄ suffers the same fate when cells are incubated with 5-aminoimidazole-4-carboxamide ribonucleoside (Choi et al., 2008), which is a pharmacological mimic of bioenergetic stress (Hardie 2004). The molecular mechanisms that are involved have not been established, but it has been determined that AMPK is not involved (Choi et al., 2008).

If the cellular levels of the diphosphoinositol polyphosphates truly reflect - or even regulate - the metabolic health of the cell, this may eventually help us understand how and why an elevated [cAMP] can reduce (PP)₂-InsP₄ levels (Safrany et al., 1998b). By using several commercially available cAMP analogues as pharmacological tools, we have excluded both PKA (Safrany et al., 1998b) and EPAC (unpublished data) from mediating this effect of the cyclic nucleotide upon (PP)₂-InsP₄ turnover. So perhaps decreased (PP)₂-InsP₄ synthesis is a downstream consequence of high levels of cAMP sometimes being registered as a metabolic crisis. That proposal is made because others (Daval et al., 2005; Yin et al., 2003) have shown that, in adipocytes, elevated [cAMP] can activate AMPK, which is normally indicative of a cell undergoing metabolic stress (Hardie 2004). We have found a similar effect of cAMP upon AMPK in DDT₁-MF₂ cells (unpublished data). We are currently using proteomic techniques to identify sites on the IP6Ks and PPIP5Ks that might undergo covalent modification in response to cellular stress.

Yeasts may use diphosphoinositol polyphosphates to respond to a shortage of certain nutrients. This is a conclusion to emerge from the work of O'Shea and colleagues (Lee et al., 2007), who have shown that the synthesis of total PP-InsP₅ levels are up-regulated in yeast cells grown for 1 to 2 hr in low-phosphate media. Genetic evidence indicates that it is the 1/3-isomer of PP-InsP₅ that must have been elevated (Lee et al., 2007), although this was not directly confirmed. Again, the mechanism that regulates PP-InsP₅ synthesis is unknown.

The changes in levels of diphosphoinositol polyphosphates that are described above are clearly of a global nature, in the sense that they can be observed within a cell population. Additionally, we should consider the possibility that there may be more subtle alterations in metabolic turnover in specific regions of single cells; such effects are difficult to record in the absence of the appropriate molecular probes.

We also have to consider that we may have been somewhat misled by our initial ligand-binding studies (Fleischer et al., 1994; Shears et al., 1995; Ye et al., 1995) being performed in the absence of Mg²⁺. Since that cation chelates some of the phosphate group's negative charge, the lack of Mg²⁺ may have led to an over-estimation of the affinity with which PP-InsP₅ binds to a protein (Shears 2009). Moreover, a highly-phosphorylated molecule (such as PP-InsP₅) can interact with a protein through delocalized and non-specific electrostatic interactions (Lemmon et al., 2002). These can substitute for the more physiologically-relevant and specific ligand interactions that are normally driven by a geometrically-precise arrangement of fewer phosphate groups (Lemmon et al., 2002). Perhaps in this particular series of experiments (Fleischer et al., 1994; Shears et al., 1995; Ye et al., 1995) the detection of PP-InsP₅ as a ligand in vitro merely reflected the occurrence of inositol lipid binding in vivo. Indeed, most of the current literature on AP-180 is consistent with the inositol lipids being the biologically-significant ligands, and the binding of the diphosphoinositol polyphosphates is now largely ignored (Legendre-Guillemin et al., 2004).

Snyder and colleagues (Luo et al., 2003) have reported that PP-InsP₅ competes with PtdIns(3,4,5)P₃ for binding to certain PH domains (Luo et al., 2003), but Downes and colleagues (Downes et al., 2005; Komander et al., 2004) have been unable to reproduce that observation. In any case, in mammalian cells the physiological levels of PP-InsP₅ (<5 μM, see above) make it an unlikely competitor for PtdIns(3,4,5)P₃, the concentration of which may increase to approx. 200 μM after activation of the PtdIns 3-kinase pathway (Stephens et al., 1993a). Furthermore, the plasma membrane residence of proteins endowed with a PH-domain does not entirely depend upon electrostatic interactions, but also involves hydrophobic interactions between protein and membrane (Manna et al., 2007). The latter phenomenon will reduce the significance of any electrostatic competition between PP-InsP₅ and PtdIns(3,4,5)P₃ that might occur.

The transfer of the phosphate group from the diphosphoinositol polyphosphate to the protein substrate is an especially remarkable phenomenon because it occurs independently of protein kinase activity (Saiardi et al., 2004). It seems that the co ordination of the phosphate's negative charge by Mg²⁺ is key to enabling the transphosphorylation to occur (Bhandari et al., 2007; Saiardi et al., 2004). There is also a requirement that the target proteins must first be “primed” by an initial casein kinase 2 (CK2)-dependent phosphorylation event (Bennett et al., 2006; Bhandari et al., 2007). In fact, experimental evidence now points to the diphosphoinositol polyphosphates actually further phosphorylating the same serine residue that is initially phosphorylated by CK2 (Bhandari et al., 2007). That is, the protein target becomes diphosphorylated, which is a novel means of covalent modification.

These in vitro data are very impressive and, potentially, they fill a significant gap in our understanding of how diphosphoinositol polyphosphates can regulate cellular function. Furthermore, because all of the diphosphoinositol polyphosphates can transphosphorylate proteins (Bhandari et al., 2007), there is less concern when an apparent biological function of these molecules shows no specificity for one specific diphosphate isomer. Such as was the case when all isomers of PP-InsP₅ were found to be equally effective at enhancing insulin secretion from pancreatic beta-cells (Illies et al., 2007). In such an event, specificity may not matter because 5-PP-InsP₅ is the only isomer that is present at a high enough concentration to elicit the biological effect. Incidentally, 5-PP-InsP₅ appears to act by increasing the size of the readily-releasable pool of insulin granules (Illies et al., 2007), which offers some possible directions for further homing in on the molecular mechanisms involved. Another notable feature of this study is that InsP₆ did not imitate the effects of the 5-PP-InsP₅ (Illies et al., 2007).

However, it has not yet proved possible to obtain direct and convincing evidence that this transphosphorylation process actually occurs in vivo. Despite recent improvements in the ability of mass spectrometry to measure changes in protein phosphorylation (Steen et al., 2006), it remains challenging to unequivocally identify a serine diphosphate in a peptide fragment obtained from a cell extract. Perhaps in the future it might be possible to develop antibodies against diphospho-serine. In the meantime, Snyder, Saiardi and colleagues (Azevedo et al., 2009; Bhandari et al., 2007; Saiardi et al., 2004) have tried indirect approaches to investigate if protein phosphorylation by the diphosphoinositol polyphosphates is physiologically relevant. For example, in some experiments they used yeast cells in which the InsP₆ kinase (Kcs1) that synthesizes 5-PP-InsP₅ was genetically eliminated (Saiardi et al., 2004). It was the absence of the phosphate donor activity of the PP-InsP₅ that was proposed to account for the lowered degree of phosphorylation of endogenous Nsr1. This is certainly an intriguing observation, but a change in the phosphorylation status of Nsr1 could instead arise independently of PP-InsP₅ synthesis per se, and might instead reflect one of the cell's many and complex adjustments that compensate for the kcs1Δ genotype.

It has also been observed that the deletion of the Nsr1 gene in S. cerevisiae caused a doubling of intracellular levels of PP-InsP₅ and (PP)₂-InsP₄ (Saiardi et al., 2004). This increase was proposed to reflect a reduced demand for diphosphoinositol polyphosphate turnover, since one of the proposed targets of phosphorylation was now eliminated (Saiardi et al., 2004). However, this proposal might now be questioned by the expansion of the number of proteins now put forward as substrates for transphosphorylation (Bhandari et al., 2007). If there really are such a large number of protein substrates, removing just one of them would not be expected to significantly impact the cellular levels of diphosphoinositol polyphosphates. Especially if the putative serine-diphosphate is long-lived, as has been proposed (Burton et al., 2009), since this also limits the impact of the phosphorylation process upon the turnover of the phosphate donors.

In the absence of a reliable method for directly assaying diphosphorylation of proteins, we, too, have looked for indirect evidence of its occurrence in vivo. Using the human homologue of Nsr1 - nucleolin - as a model, we (Yang et al., 2008) searched for evidence that its phosphorylation by diphosphoinositol polyphosphates might be physiologically relevant. We made the assumption that, if Snyder and colleagues (Bhandari et al., 2007; Saiardi et al., 2004) are correct, the degree of nucleolin phosphorylation should increase as the cellular levels of (PP)₂-InsP₄ and/or PP-InsP₅ are elevated. We also noted previous experiments demonstrating that the phosphorylation of nucleolin is associated with its transfer from the nucleolus into the nucleoplasm (Kim et al., 2005). Thus, the extent to which nucleolin accumulates in the nucleoplasm can be anticipated to provide a readout of its degree of phosphorylation. We therefore manipulated cellular levels of diphosphoinositol polyphosphates in an osteosarcoma cell line using a combination of hyperosmotic stress, and some pharmacological tricks (Yang et al., 2008). We found that a hyperosmotic challenge caused nucleolin to accumulate in the nucleoplasm -- suggesting its degree of phosphorylation was increased -- but this response occurred independently of changes in levels of diphosphoinositol polyphosphates (Yang et al., 2008).

Azevedo et al. (Azevedo et al., 2009) recently used a “back-phosphorylation” assay to study if protein diphosphorylation might occur in vivo. The protein target that was studied was AP3B1, the β-subunit of the AP3 adaptor complex. The basis of this assay is that following the isolation of AP3B1 from intact cells, it would only be diphosphorylated by [³²P]PP-InsP₅ in vitro if it had not already been diphosphorylated by PP-InsP₅ in vivo (Fig. 2a). So, AP3B1 was exogenously expressed in a kcs1Δ strain of S. cerevisiae, in which diphosphoinositol polyphosphates and hence the capacity for transphosphorylation were both virtually eliminated. When AP3B1 was extracted from this strain of yeast and incubated in vitro with [³²P]PP-InsP₅, there was considerable transphosphorylation of the adaptor (Fig. 2a). AP3B1 was also expressed in either wild-type S. cerevisiae, or in a strain (vip1Δ) that has elevated PP-InsP₅ levels. The AP3B1 obtained from these strains exhibited little or no transphosphorylation in vitro. Thus, the authors argued that the adaptor protein must already have been diphosphorylated by cellular PP-InsP₅ in vivo (Fig. 2b). However, there is another interpretation of these results that becomes clear once it is recalled that in vitro, diphosphoinositol polyphosphates can only phosphorylate an appropriate Ser residue that is first primed by phosphorylation by casein-kinase II (CK2) (Bhandari et al., 2007). It is certainly the case that the AP3B1 that was isolated from kcs1Δ yeast must have already been mono-phosphorylated by CK2 in vivo, or the transphosphorylation by PP-InsP₅ would not have occurred in vitro (Fig. 2a). So, let us suppose that for some reason the expression of kcs1 in intact cells causes AP3B1 not to be monophosphorylated by CK2 in vivo (Fig. 2c). In such a situation, AP3B1 cannot then be transphosphorylated by PP-InsP₅ in vitro (Fig. 2c). In fact, genetic interaction studies (Fiedler et al., 2009), which measure the extent to which the function of one gene depends on the presence of a second gene, have found an association between Kcs1 and casein kinase (Cka2) in S cerevisiae. Therefore, it is possible that the back-phosphorylation assay actually could be recording the degree of CK2-mediated monophosphorylation of the appropriate Ser in AP3B1 in vivo (Fig. 2).

InsP₆ is quite an effective inhibitor of protein phosphorylation by diphosphoinositol polyphosphates (Saiardi et al., 2004). In eukaryotic cells the cellular levels of InsP₆ are typically at least 20-fold higher than the diphosphoinositol polyphosphates, so the latter will likely only be capable of phosphorylating proteins in an a microenvironment from which InsP₆ is relatively excluded. This scenario is plausible. There is certainly evidence that some InsP₆ is divided into metabolically-separated “pools” (Otto et al., 2007). Other data showing a punctate distribution of the InsP₅ 2-kinase within certain cellular structures such as nucleoli and stress-granules also indicates that intracellular InsP₆ synthesis is compartmentalized (Brehm et al., 2007). Thus, future studies into the possibility that there is compartmentalization of diphosphoinositol polyphosphate synthesis could have a significant impact on the future of the transphosphorylation hypothesis.