The Significance of the Bifunctional Kinase/Phosphatase Activities of Diphosphoinositol Pentakisphosphate Kinases (PPIP5Ks) for Coupling Inositol Pyrophosphate Cell Signaling to Cellular Phosphate Homeostasis*

Proteins responsible for Pi homeostasis are critical for all life. In Saccharomyces cerevisiae, extracellular [Pi] is “sensed” by the inositol-hexakisphosphate kinase (IP6K) that synthesizes the intracellular inositol pyrophosphate 5-diphosphoinositol 1,2,3,4,6-pentakisphosphate (5-InsP7) as follows: during a period of Pi starvation, there is a decline in cellular [ATP]; the unusually low affinity of IP6Ks for ATP compels 5-InsP7 levels to fall in parallel (Azevedo, C., and Saiardi, A. (2017) Trends. Biochem. Sci. 42, 219–231. Hitherto, such Pi sensing has not been documented in metazoans. Here, using a human intestinal epithelial cell line (HCT116), we show that levels of both 5-InsP7 and ATP decrease upon [Pi] starvation and subsequently recover during Pi replenishment. However, a separate inositol pyrophosphate, 1,5-bisdiphosphoinositol 2,3,4,6-tetrakisphosphate (InsP8), reacts more dramatically (i.e. with a wider dynamic range and greater sensitivity). To understand this novel InsP8 response, we characterized kinetic properties of the bifunctional 5-InsP7 kinase/InsP8 phosphatase activities of full-length diphosphoinositol pentakisphosphate kinases (PPIP5Ks). These data fulfil previously published criteria for any bifunctional kinase/phosphatase to exhibit concentration robustness, permitting levels of the kinase product (InsP8 in this case) to fluctuate independently of varying precursor (i.e. 5-InsP7) pool size. Moreover, we report that InsP8 phosphatase activities of PPIP5Ks are strongly inhibited by Pi (40–90% within the 0–1 mm range). For PPIP5K2, Pi sensing by InsP8 is amplified by a 2-fold activation of 5-InsP7 kinase activity by Pi within the 0–5 mm range. Overall, our data reveal mechanisms that can contribute to specificity in inositol pyrophosphate signaling, regulating InsP8 turnover independently of 5-InsP7, in response to fluctuations in extracellular supply of a key nutrient.

Phosphate has multiple functions that direct the survival of all living organisms: in its organic form, P i is a component of genomic material, it serves as an energy currency, and it is ubiquitous in cell signaling. Thus, P i homeostasis is essential to life, but the mechanisms by which this occurs in humans and other metazoans are largely unknown (1,2). Most of the previous work in this field of research has focused on yeast models (3)(4)(5).
In particular, recent studies with Saccharomyces cerevisiae have revealed a new function in P i homeostasis for inositol pyrophosphates (5). The latter are soluble, intracellular signals that contain multiple phosphates and diphosphates; up to seven (InsP 7 ) 4 or eight (InsP 8 ) phosphates in total are crammed around a six-carbon inositol ring (see Refs. 6 -8 and Fig. 1). In S. cerevisiae, levels of one inositol pyrophosphate, 5-InsP 7 , track perturbations to P i homeostasis (5).
This P i -sensing activity of 5-InsP 7 appears to reflect it being synthesized by a kinase class (kcs1 in yeast; IP6Ks in metazoans) that exhibits an unusually low affinity for ATP (9,10). Consequently, cellular levels of 5-InsP 7 in yeast decrease in response to the drop in [ATP] that accompanies extracellular [P i ] depletion (5,11). Furthermore, these ATP-driven changes in 5-InsP 7 levels appear to comprise a dynamic signaling response because 5-InsP 7 regulates proteins that maintain P i homeostasis through interactions with their SPX domains (5). However, it is not known to what extent this signaling response is applicable to metazoan cells, which lack orthologs of many of the yeast genes that function in P i sensing and P i homeostasis (2). * This work was supported in part by the Intramural Research Program of the NIEHS, National Institutes of Health. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 1 Supported by Swiss National Science Foundation Grant PP00P2_157607. 2 Partially supported by National Center for Advancing Translational Sciences, National Institutes of Health Grant UL1TR001111. 3 To whom correspondence should be addressed: Laboratory of Signal In the current study, we have searched for links between P i homeostasis and inositol pyrophosphates in a human model system: the HCT116 intestinal epithelial cell line. This choice reflects the physiological relevance of both small and large fluctuations in [P i ] within the gastrointestinal tract (12). One of our goals has been to investigate whether there are changes in extracellular [P i ] that might cause intracellular [ATP] and  ] to co-vary, which as mentioned above is considered to be an IP6K-dependent phenomenon.
Perhaps as a consequence of 5-InsP 7 being the most abundant of the inositol pyrophosphates, it has been the focus of much of the literature in this field (6,8,13,14). In the current study, we also study a different inositol pyrophosphate, InsP 8 (Fig. 1). We describe some new features to InsP 8 turnover that solidify its own, independent cell signaling credentials. This information arises out of our focus on the PPIP5Ks (Fig. 2). The latter enzymes are of general interest; in addition to hosting a kinase domain that phosphorylates 5-InsP 7 to InsP 8 , PPIP5Ks posses a separate phosphatase domain that dephosphorylates InsP 8 back to 5-InsP 7 (15)(16)(17).
That is, PPIP5Ks interconvert substrates and products in apparent "futile cycles" (Figs. 1 and 2). Kinase/phosphatase and other covalent modification cycles are a nexus for regulatory inputs into metabolic and signaling pathways (18); in fact this phenomenon is considered a core motif in the field of systems biology (19). However, in general, such competing catalytic activities are hosted by separate proteins for the purposes of compartmentalization and for promoting signaling fidelity (20). Only in rare cases have these apparent benefits been selected against in order that the mutually antagonistic catalytic activities co-exist within a single protein (19,21,22). The PPIP5K family is one of these exceptions; representatives from humans, yeasts, and plants contain kinase and phosphatase domains (16), indicating that this bifunctionality has survived at least 1.5 billion years of evolutionary pressure (23).
Among a number of proposed advantages of having competing catalytic activities in a single, bifunctional protein are the following: (a) preventing signaling incoherence that can otherwise arise due to stochastic fluctuations in the degrees of expression of two separate proteins; (b) robustness, i.e. invariance to quantitative changes of the system's components, including substrate concentration; and (c) increased "parametric sensitivity," that is, a situation in which signaling output is amplified following relatively small changes in the concentration of a particular parameter, such as an enzyme regulator (19,21). However, the significance of these phenomena in vivo is dictated by the catalytic parameters of the mutually antagonistic domains (19,21). Hitherto, we lacked this information. The full kinetic profile for PPIP5Ks has not previously been determined in the full-length versions of these enzymes. Moreover, there is no information in the literature describing the existence of a modulator of either the kinase or the phosphatase activity of any mammalian PPIP5K. The current study addresses these important gaps in our understanding of inositol pyrophosphate turnover. We demonstrate that P i regulates the catalytic activities of the PPIP5Ks. Furthermore, our data indicate that InsP 8 and 5-InsP 7 each act through separate mechanisms to individually sense extracellular P i status. The schematic describes all known mammalian enzyme classes that interconvert InsP 6 with InsP 8 . Note that current thinking (28,29) has the major route from InsP 6 to InsP 8 in mammalian cells progressing through 5-InsP 7 rather than 1-InsP 7 . FIGURE 2. Domain graphic for human PPIP5Ks. Domain graphics are shown for the human PPIP5Ks used in this study (type 1, BC057395.1; type 2, XM_005271938). For PPIP5K1, amino acid residues defining each domain are numbered as in a previous study, which also defined the intrinsically disordered domain (IDR) (49). These boundaries were matched to those of the corresponding domains in PPIP5K2 by sequence alignments using Clustal Omega. The aligned intrinsically disordered domain boundaries in PPIP5K2 are consistent with those independently predicted from the PSIPRED Protein Sequence Analysis Workbench. The percent sequence identities across each of the three domains are also indicated. Also indicated are the nature and the locations of our engineered mutations in the kinase and phosphatase domains.

Results and Discussion
The Effects of Extracellular [P i ] Starvation and Replenishment upon Levels of Inositol Pyrophosphates in HCT116 Cells-Previous work with S. cerevisiae (24) has shown that extracellular [P i ] is sensed by the IP6K that synthesizes the intracellular inositol pyrophosphate 5-InsP 7. During a period of P i starvation, there is a decline in cellular [ATP]. The unusually low affinity of IP6Ks for ATP compels 5-InsP 7 levels to fall in parallel (5). As noted in a recent review of this field (24), yeasts are the only organisms that have previously been used to study 5-InsP 7 turnover in response to fluctuations in extracellular P i availability. We have addressed this gap in the field by using the HCT116 human intestinal epithelial cell line as a model system. We prelabeled cells with [ 3 H]inositol for several days, and then we investigated the effects of removal and replenishment of extracellular P i . We used HPLC to assay the intracellular levels of "higher" InsPs, including InsP 5 , InsP 6 , and the inositol pyrophosphates.
First (Fig. 3A), we analyzed cells that had been incubated under P i -replete conditions (i.e. 2 mM [P i ] (25,26)). As in all metazoan cells, levels of InsP 5 and InsP 6 are much higher than those of the inositol pyrophosphates (InsP 7 and InsP 8 ); thus for clarity, those regions of the chromatographs that include the inositol pyrophosphates are replotted on an expanded y axis (Fig. 3, A, B, C, and D). The depiction of InsP 7 as being the 5-isomer is based on previous work with HCT116 cells in which our HPLC procedures were shown to resolve 5-InsP 7 from 1-InsP 7 ; the latter comprises Ͻ2% of total InsP 7 (27). Indeed, it is generally believed that most of the cell's InsP 8 is synthesized from 5-InsP 7 rather than 1-InsP 7 (28,29).
We next investigated the effects of perturbations to P i homeostasis. We challenged cells with 6 h of P i starvation (Fig. 3B). This procedure did not impact levels of InsP 5 or InsP 6 ( Fig. 3, A, B, E, and F), but InsP 7 levels decreased by ϳ70% (Fig. 3, A, B, and G). This is the first demonstration, for any metazoan cell type, that levels of InsP 7 are sensitive to a change in extracellular [P i ]. Nevertheless, InsP 8 levels fell by 98%, which is a much greater response to the P i depletion protocol (Fig. 3, A, B, and H). It has not previously been reported that InsP 8 levels are influenced by extracellular [P i ] in any organism (e.g. see Ref. 24). In other words, our data describe a new connection between nutrient status and the poise of a cell signaling cascade.
Next, we substituted the last 15 min of the P i starvation protocol with the readdition of 2 mM P i . The levels of 5-InsP 7 were not influenced by this 15-min period of P i replenishment ( Fig. 3, B, C, and G). In contrast, there was a substantial rescue of the InsP 8 peak (Fig. 3, B, C, and H). These data testify to an acute effect of extracellular P i status upon InsP 8 synthesis that is independent of the supply of 5-InsP 7 (i.e. the major InsP 7 precursor for InsP 8 ; see above). We revisit this point below.
In further experiments, a 3-h period of P i starvation was followed by 3 h of 2 mM P i readdition. After this protocol, levels of 5-InsP 7 were similar to those found in cells incubated with 2 mM P i for 6 h (Fig. 3, A, D, and G). In contrast, levels of InsP 8 in P i -starved/P i -replenished cells were 80% higher than those of control cells (Fig. 3, A, D, and H), suggestive of a longer term adaptive response to a period of P i starvation. Indeed, a major conclusion to draw from all of these experiments is that InsP 8 reacts more dramatically to fluctuations in extracellular [P i ] than does 5-InsP 7 .

The Effects of Extracellular [P i ] Starvation and Replenishment upon Levels of Intracellular [P i ] and [ATP] in HCT116
Cells-How do intracellular inositol pyrophosphates sense short term fluctuations in extracellular P i levels? As a first step toward answering that question, we investigated whether intracellular [P i ] tracked the imposed changes in extracellular [P i ]. We found that our 6-h P i starvation protocol reduced intracellular [P i ] by 55% (Fig. 3I); only 15 min of P i replenishment was sufficient to completely restore intracellular [P i ] (Fig. 3I). The recovery in InsP 8 lags behind that for P i (Fig. 3, H and I); the rate of change in InsP 8 levels depends upon dynamic fluctuations in InsP 8 synthesis and metabolism (see below).
How might inositol pyrophosphate levels be modified by changes in intracellular [P i ]? Previous work has indicated that P i levels influence ATP production (30). Indeed, we found that our P i starvation protocol reduced [ATP] by 66% (Fig. 3J), i.e. to a value of about 1.5 mM, based on the concentration of ATP normally being around 5 mM (31)(32)(33). Thus, the associated drop in 5-InsP 7 levels ( Fig. 3, A, B, and G) may reflect the unusually low affinity (K m approximately 1 mM) of IP6Ks for ATP (9,10). Both ATP and 5-InsP 7 were at normal levels when a 3-h period of P i starvation was followed by 3 h of replenishment with 2 mM [P i ] (Fig. 3

, G and J).
On the other hand, there is no reason to suspect that synthesis of InsP 8 by the PPIP5Ks would be influenced by these changes in [ATP]. In contrast to IP6Ks, the PPIP5Ks exhibit a low K m value for ATP (20 M (29)), so their kinase activities would be saturated even by the reduced levels of [ATP] caused by 6 h of P i starvation (Fig. 3J). We therefore searched for other mechanisms by which P i might regulate InsP 8 turnover. There are precedents for P i itself being a physiologically relevant regulator of both kinases and phosphatases (34,35). Thus, we decided to study the synthesis and metabolism of InsP 8 by the human bifunctional PPIP5Ks in vitro and then investigate whether these catalytic activities are regulated by P i . To our knowledge, no previous study has characterized the competing kinase and phosphatase activities of full-length PPIP5Ks purified from a human expression system.
Kinetic Analysis of the Bifunctional Human PPIP5Ks: the Interconversion of 5-InsP 7 and InsP 8 -Mammals express two PPIP5Ks, type 1 and type 2 (Fig. 2). HEK cells were used as hosts for the expression of full-length, recombinant wild-type PPIP5K1 and PPIP5K2, each of which were engineered to possess an N-terminal FLAG tag to facilitate purification with anti- The identification of InsP 7 as the 5-isomer is based on our previous work with HCT116 cells in which our HPLC procedures were shown to resolve 5-InsP 7 from 1-InsP 7 ; the latter is a minor isomer that comprises Ͻ2% of total InsP 7 (27). The graphics below the x axes correspond to those described above that depict the various extracellular [P i ] conditions. Scatter plots (error bars represent S.D.) in I and J show total intracellular P i (malachite green/molybdate method; n ϭ 8 -10) and ATP (n ϭ 4 -7), respectively, determined in parallel experiments with non-radiolabeled cells. The mean values for cell P i content (nmol/mg of protein) from left to right are as follows (error bars represent S.D.): 62 Ϯ 2.3, 28 Ϯ 1.9, 62 Ϯ 2.7, and 53 Ϯ 2.2; these values closely match the P i levels that were obtained by an independent enzymatic method (60 Ϯ 5.6, 32 Ϯ 2.5, 59 Ϯ 2.9, and 50 Ϯ 2.7; n ϭ 6). MARCH 17, 2017 • VOLUME 292 • NUMBER 11 JOURNAL OF BIOLOGICAL CHEMISTRY 4547 FLAG resin. Upon SDS-PAGE, both recombinant proteins eluted at their expected molecular sizes (Fig. 4).
Analysis of the wild-type kinase activities by HPLC confirmed that the InsP 8 product accumulated (Fig. 5, A, B, and C). Nevertheless, when these assays were performed with phosphatase mutants (PPIP5K1 R399A and PPIP5K2 R388A ), a 2.4 -4-fold larger net formation of InsP 8 was observed (Fig. 5, A, B, and C). Thus, the phosphatase activities constrain but do not overwhelm kinase activities.
We next verified that wild-type PPIP5K1 and PPIP5K2 both dephosphorylate InsP 8 to 5-InsP 7 (Fig. 5, D, E, and F). We also confirmed that the single site mutations in the phosphatase domain nearly completely impaired InsP 8 hydrolysis (Fig. 5, D, E, and F). This observation also usefully confirms that there is negligible contamination of our enzyme preparations with any other InsP 8 phosphatase activities. In fact, the only other class of mammalian enzyme known to hydrolyze InsP 8 is DIPP (40), and, even if that had been present, it would have dephosphorylated InsP 8 to InsP 6 without appreciable accumulation of 5-InsP 7 (see below).
Clearly, the phosphatase mutants of the PPIP5Ks yield the more accurate rates of the 5-InsP 7 kinase reactions (Figs. 5 and 6A). In this respect, we found that PPIP5K2 is 6-fold more active than is PPIP5K1. Our phosphatase mutants were also useful in confirming that zero-order conditions prevail when physiologically relevant concentrations of 5-InsP 7 are phosphorylated by PPIP5Ks: reaction rates were not increased when the 5-InsP 7 concentration was increased from 1 to 5 M (representative examples: PPIP5K1 R399A , 1.2 and 0.9 nmol/mg of protein/min, respectively; PPIP5K2 R388A , 6.4 and 5.8 nmol/mg of protein/min, respectively).
In the assays of phosphatase activities toward InsP 8 described above, we incubated the PPIP5Ks with 0.05 M InsP 8 , which is the estimated concentration of this inositol pyrophosphate in HCT116 cells cultured in DMEM/F-12 (27). In this reaction condition, the InsP 8 phosphatase activities (0.8 -1.5 nmol/mg of protein/min; Fig. 6B) lie in a range that is similar to that for the kinase activities (determined by assaying the phosphatase mutants: 0.7-4.5 nmol/mg of protein/min; Fig. 6A). These same data further indicate that, for PPIP5K2, the 5-InsP 7 kinase/InsP 8 phosphatase ratio slightly favors the kinase activity. For PPIP5K1, the phosphatase activity is favored. Such subtle differences between the two PPIP5K isoforms may serve some as yet unsuspected physiological purpose.
Our next goal was to determine how the InsP 8 phosphatase activities might respond to elevated concentrations of substrate. Therefore, we assayed phosphatase activities against 1 M InsP 8 ; the reaction rates for both PPIP5Ks are 15-20-fold higher than those observed with 0.05 M InsP 8 (Fig. 6B). Thus, we conclude that the phosphatase activities in vivo may be modulated by fluctuations to supply of substrate (i.e. InsP 8 ). In contrast, the 5-InsP 7 kinase activities proceed under zero-order conditions in vivo (see above and Ref. 29). Mathematical modeling of just such a situation for a kinase/phosphatase bifunctional protein (zero-order for the kinase; first-order for the phosphatase) has established that a biological outcome is a degree of concentration robustness for the kinase product (21), in this case InsP 8 . In other words, it can be concluded that InsP 8 is inherently insensitive to changes in 5-InsP 7 levels. This property of PPIP5Ks could promote specificity of inositol pyrophosphate signaling by stabilizing InsP 8 levels during periods of stimulus-dependent regulation of 5-InsP 7 levels. Conversely, such concentration robustness indicates that the response of InsP 8 to changes in P i homeostasis (Fig. 3) reflects active regulation of InsP 8 turnover and not mass action effects due to fluctuations in 5-InsP 7 supply.
There are three additional conclusions that can be drawn from our kinetic data (Figs. 5 and 6). First, these results provide the first direct demonstration that the phosphatase activity of mammalian PPIP5Ks has the potential to significantly restrict net kinase activity in the context of the full-length proteins. That is, we conclude that so-called futile cycling by bifunctional PPIP5Ks is a realistic physiological scenario. Second, the activity of the kinase domain in full-length PPIP5K2 is 30-fold below its potential maximal capacity (190 nmol/mg of protein/min; as recorded for the isolated kinase domain expressed in Escherichia coli (29)). This catalytic constraint upon the kinase domain, when expressed as a full-length protein in a human cell type, could reflect covalent modification and/or a conformational constraint enforced by the other protein domains. Irrespective of the mechanism, the submaximal kinase activity in the full-length protein can be viewed as a selective advantage: at steady-state levels of 5-InsP 7 (1 M) and InsP 8 (0.05 M), the rates of the kinase and phosphatase reactions are similar, so in such a situation, the slower these individual rates, the lower the amount of ATP that must be expended to maintain a given cellular content of InsP 8 . A third conclusion that emerges from the phosphatase and kinase activities of full-length PPIP5Ks being similar is that it is intuitive that it maximizes the sensitivity with which a modulator of either activity could affect net flux through the cycle. This is a topic we return to later (see below).
Kinetic Analysis of the Bifunctional Human PPIP5Ks: the Interconversion of InsP 6 and 1-InsP 7 -PPIP5Ks are also capable of catalyzing a separate cycle of competing kinase and phosphatase activities that interconvert InsP 6 with 1-InsP 7 (Fig. 1). We next used the wild-type and phosphatase mutant PPIP5K constructs to determine the relative rates of these particular reactions. For the kinase assays (Fig. 6C), we set InsP 6 concentrations to 1 M. Most estimates of cellular levels of InsP 6 are 10 M or more (36 -38), but the use of 1 M increases assay sensitivity while still being more than sufficient to saturate PPIP5Ks with this substrate and obtain V max values (15,29). For the phosphatase assays (Fig. 6D), we initially used a concentration of 0.05 M 1-InsP 7 (because cellular levels of the latter are 2-10% of total InsP 7 (27,28)). The maximum rates of the kinase activities toward InsP 6 (0.2-1 nmol/mg of protein/min; Fig. 6C) are 4 -6-fold lower than the kinase activities toward 5-InsP 7 (Fig. 6A). Furthermore, the rates of the phosphatase activities toward 0.05 M 1-InsP 7 (0.2-0.4 nmol/mg of protein/min; Fig.  6C) are 4-fold lower than the phosphatase activities toward InsP 8 (Fig. 6A). That is, kinetic parameters dictate that substrate cycling through InsP 6 and 1-InsP 7 is quantitatively less significant than is cycling through 5-InsP 7 and InsP 8 . The latter conclusion is directly relevant to current thinking (6, 28) that the major route from InsP 6 to InsP 8 progresses through 5-InsP 7 rather than 1-InsP 7 .
Although there is no direct evidence that the intracellular concentration of 1-InsP 7 in mammalian cells exceeds 0.05 M (27), we also studied in vitro PPIP5K phosphatase activity toward this substrate at a concentration of 1 M (Fig. 6D). At  MARCH 17, 2017 • VOLUME 292 • NUMBER 11 this higher substrate concentration, the phosphatase activity was double that against 0.05 M 1-InsP 7 (Fig. 6D), indicating that this activity (like the phosphatase activity toward InsP 8 ; see above) does not operate under zero-order conditions in vivo; instead it is capable of being regulated by substrate supply. Armed with an improved kinetic understanding of the operation of the PPIP5Ks, we next sought information concerning how changes in intracellular P i might mediate the effects upon InsP 8 turnover described above.

PPIP5K Bifunctionality and Phosphate Homeostasis
Regulation of PPIP5Ks by Inorganic Phosphate in Vitro-We investigated whether P i might directly modify the catalytic activities of the PPIP5Ks. We found that P i reduced the rate of InsP 8 hydrolysis by our preparations of recombinant PPIP5K1 and PPIP5K2 in a dose-dependent manner (90 and 40% inhibition, respectively, by 1 mM P i ; Fig. 7, A, B, C, and D). At the very least, these data indicate that cytoplasmic P i will alter the kinase/phosphatase poise of the PPIP5Ks in favor of InsP 8 synthesis. Moreover, these data suggest a mechanism by which P i starvation of HCT116 cells promotes a loss of InsP 8 levels (Fig.   3, A, B, and H): the associated depletion of intracellular P i (Fig.  3I) could relieve its inhibition of InsP 8 phosphatase activity. Equally, the restoration of InsP 8 levels upon readdition of extracellular P i (Fig. 3, C, D, and H) may be explained by the replenishment of intracellular P i (Fig. 3I), which inhibits the PPIP5K phosphatase activities (Fig. 7, B and D). Our data therefore represent the first demonstration of a mechanism for regulating the catalytic activities of PPIP5K in response to an extracellular stimulus (in this case nutrient availability). It is of further interest that DIPP-mediated InsP 8 phosphatase activity proceeds without InsP 7 accumulation (Fig. 7, E and F), and is not inhibited by P i (Fig.  7, E and F). That is, P i is not a general inhibitor of all inositol pyrophosphate phosphatases. Moreover, this observation excludes any residual possibility that contaminating DIPPs contribute to the phosphatase activities of our PPIP5K preparations.
Finally, we found that not only does P i inhibit PPIP5K2 phosphatase activity (Fig. 7D) but also its kinase activity is activated 2-fold as the concentration of P i was raised from 0 to 5 mM (Fig.  8). This is the first description of a mechanism by which the kinase activity of any PPIP5K may be regulated. Reciprocal regulation of competing catalytic activities by a single regulator is a particularly sensitive control mechanism. Thus, it is possible that, in vivo, quite small fluctuations in cytoplasmic [P i ] could be amplified into proportionately large changes in net [InsP 8 ] synthesis by cytoplasmic (15) PPIP5Ks. Remarkably, the activation of kinase activity is also specific to PPIP5K2: P i has no effect upon the kinase activity of PPIP5K1 (Fig. 8). These data underscore a significant difference in the regulation of the activities of the tightly conserved catalytic domains of the two PPIP5Ks.
Elevated Extracellular [P i ] Specifically Promotes InsP 8 Accumulation-Our main experimental paradigm has been to impose a nutritional challenge upon cultured cells: deprive and then restore the levels of P i found in serum. Other groups have studied the impact upon signal transduction cascades when cells are incubated with "high" [P i ] (5-10 mM) (2,30). We noted that such dramatic fluctuations in extracellular P i concentration (up to 10 mM or more) are physiologically relevant for cells that line the gastrointestinal tract (12). We therefore used the HCT116 model to study the effects upon inositol pyrophosphate turnover of elevating extracellular P i from 1 to 6 mM for periods of up to 1 h. This procedure did not significantly affect levels of either InsP 5 , InsP 6 , or 5-InsP 7 (Fig.  9, A, B, and C). The absence of an effect upon 5-InsP 7 illustrates how the reaction of inositol pyrophosphates to changes in P i homeostasis may depend upon the nature of the experimental protocol. We propose that levels of 5-InsP 7 do not respond to the high [P i ] protocol because [ATP] and hence IP6K activity (see above) are also not altered (Fig. 9D).
In response to elevated extracellular [P i ], intracellular [P i ] trended higher (by 9 -14%; Fig. 9E legend) but not with statistical significance. Nevertheless, InsP 8 levels responded dramatically: a 2.3-fold increase within 15 min and a 5-fold elevation after 1 h (Fig. 9, A, B, and F). This response of InsP 8 consolidates our discovery (see above) that it is more sensitive to changes in extracellular [P i ] than is 5-InsP 7 . In view of how acutely and specifically InsP 8 responds to high [P i ], we sought further evidence for the participation of PPIP5Ks in this event. We gener-FIGURE 6. Kinase and phosphatase activities of recombinant PPIP5Ks. All assays were performed as described under "Experimental Procedures" with either wild-type or phosphatase mutant versions of PPIP5Ks (K1, PPIP5K1; K2, PPIP5K2) plus either 1 or 0.05 M substrate as indicated beneath each scatter plot. A and B show kinase and phosphatase activities toward 5-InsP 7 or InsP 8 , respectively (as indicated by the graphic above the graphs). C and D show kinase and phosphatase activities toward InsP 6 or 1-InsP 7 , respectively (as indicated by the graphic above the graphs). Error bars represent S.D. and are derived from three to four biological replicates.
ated an HCT116 cell line in which both PPIP5Ks have been knocked out using CRISPR (27). These cells do not synthesize InsP 8 irrespective of the concentration of extracellular [P i ] (Fig.  9G).
Concluding Comments-Until quite recently, signal transduction cascades and metabolic circuits were not recognized to be intimately connected; now it is appreciated that a key aspect of cellular and organismal homeostasis is the acute regulation of cell signaling pathways by the levels of a particular metabolite or nutrient (41). In particular, inositol pyrophosphates have gained considerable attention for their actions that dovetail signaling with metabolism (6,7,24,42). In our study, we add several new aspects to this important research topic by showing that InsP 8 levels sense and respond to fluctuations in the extracellular levels of a vital nutrient, P i .
Our in vitro data describe physiologically relevant mechanisms by which P i may directly regulate the catalytic activities of the PPIP5Ks: inhibition of InsP 8 phosphatase (Fig. 7) and stimulation of 5-InsP 7 kinase (Fig. 8). Prior to this study, no regulators of either the kinase or phosphatase activities of mammalian PPIP5Ks had been described. Reciprocal regulation of any covalent modification cycle by a single modulator is a particularly efficient regulatory paradigm. Moreover, it is intuitive that the sensitivity of such a process is enhanced when the two opposing activities proceed at approximately equal rates as is the case for PPIP5K2 under physiologically relevant conditions (Fig. 6, A  and B). Thus, there may be scenarios in which proportionately large changes in [InsP 8 ] may be promoted by quite small fluctuations in cytoplasmic [P i ] that are beyond the sensitivity of our assays, which record total intracellular [P i ] (Fig. 9E). A more precise spatiotemporal understanding of the inter-relationships between P i and InsP 8 in vivo requires additional information that is not readily obtained: a quantitative resolution of their separation into different cellular pools (2,43). Furthermore, differential levels of expression of PPIP5K1 versus PPIP5K2 could contribute to quantitative and qualitative cell type-specific differences in InsP 8 responses to extracellular [P i ] because this anion only stimulates the kinase domain of PPIP5K2 and not that in PPIP5K1 (Fig. 8). It will be important to determine how widespread in other cell types is the impact of fluctuations in extracellular [P i ] upon intracellular levels of InsP 8 .
It is also possible that regulation of PPIP5K activity by P i is not the only cell signaling mechanism that regulates InsP 8 levels, particularly those that might be activated by high extracellular [P i ] (2,30). This would not be surprising considering the many interactions between inositol pyrophosphate signaling and metabolic homeostasis (6,7,24,42) and the domain complexities of the PPIP5Ks (Fig. 2).
Our new data are also intriguing in light of recent data derived from yeast that show that 5-InsP 7 rheostatically regulates mechanisms of P i homeostasis mediated by SPX domains (5). The only human protein known to contain an SPX domain is XPR1, which transports P i out of cells (5). In the current study with a human cell type, we show [ATP] and  ] to co-vary in their response to P i starvation. This is the first time such a response has been observed in any metazoan model. In addition, we show that InsP 8 senses changes in P i status by separate mechanisms that exhibit greater sensitivity as compared with 5-InsP 7 . It could be useful to study whether InsP 8 is a homeostatic regulator of XPR1, perhaps adjusting the rate of cellular P i efflux depending upon organismal P i status. This would be a new direction for this field, which prior to our study had mainly focused on the biological actions of 5-InsP 7 (see Refs. 8 and 14). The possibility of there being other types of InsP 8 receptor should not be ignored. Finally, our results raise the possibility that dysregulation of inositol pyrophosphate signaling by genetic or environmental factors could conceivably contribute to mechanisms of toxicity of unbalanced [P i ] homeostasis (1).

Experimental Procedures
Cell Culture-HEK293T cells and HCT116 cells were obtained from ATCC; HCT116 cells in which both PPIP5K1 and PPIP5K2 were knocked out using CRISPR were derived as described previously (27). The culture medium (Thermo Fisher Scientific) was DMEM/F-12 for HCT116 cells and DMEM for HEK293T cells, each supplemented with 10% fetal bovine serum (Gemini Bio Products) and 100 units/ml penicillinstreptomycin (Thermo Fisher Scientific) at 37°C in 5% CO 2 .
Measurement of Intracellular Levels of P i , ATP, and Inositol Phosphates-For assays of intracellular ATP and P i , 3 ϫ 10 5 cells/well were seeded in a 12-well dish and cultured in DMEM/ F-12 (containing 1 mM P i ) for 2 days at which point cultures were 70% confluent. For some experiments, an additional 5 mM P i was added (as a NaH 2 PO 4 /Na 2 HPO 4 mixture (pH 7.4)), for the times indicated (see "Results and Discussion"). In other experiments, cells were washed three times with P i -free DMEM (ThermoFisher Scientific) and then incubated in either P i -free DMEM or DMEM plus 2 mM NaH 2 PO 4 for the times indicated (see "Results and Discussion"). To terminate these assays, cells were washed three times in ice-cold buffer containing 20 mM HEPES (pH 7.2), 150 mM NaCl and then lysed by agitation in 1 ml of wash buffer containing 1% Triton X-100 for 5 min at 4°C. The ATP was assayed using a commercial kit (Molecular Probes TM , A22066). For the P i assays, the lysate was cleared of debris by a brief centrifugation, and then the resulting supernatant was centrifuged through an Amicon Ultra filter (10-kDa molecular mass cutoff; 5 min at 4°C). The P i in the eluate was generally assayed with a standard malachite green/molybdate method (44). Similar data were obtained (see figure legends) using an enzymatic assay (Cell Biolabs, Inc., catalogue number STA-685) that avoids potential contamination through acid-dependent hydrolysis of organic phosphates. The data that were obtained were normalized to protein concentration, which was measured using the Pierce TM BCA protein assay kit (product number 23225).
To assay inositol phosphates, 1 ϫ 10 6 cells were seeded in a 10-cm dish and cultured for 3 days in 7 ml of medium supplemented with 10 Ci/ml [ 3 H]inositol (American Radiolabeled Chemicals) at which point cultures were 70% confluent. Cells were then incubated in medium with various P i concentrations as described above. Cells were acid-quenched, and the inositol phosphates were extracted and analyzed by a 3 ϫ 250-mm Car-boPac TM PA200 HPLC column (ThermoFisher Scientific) all as described previously (27).