A two-way switch for inositol pyrophosphate signaling: evolutionary history and biological significance of a unique, bifunctional kinase/phosphatase

Abstract

The inositol pyrophosphates (PP-InsPs) are a unique subgroup of intracellular signals with diverse functions, many of which can be viewed as reflecting an overarching role in metabolic homeostasis. Thus, considerable attention is paid to the enzymes that synthesize and metabolize the PP-InsPs. One of these enzyme families - the diphosphoinositol pentakisphosphate kinases (PPIP5Ks) - provides an extremely rare example of separate kinase and phosphatase activities being present within the same protein. Herein, we review the current state of structure/function insight into the PPIP5Ks, the separate specialized activities of the two metazoan PPIP5K genes, and we describe a phylogenetic analysis that places PPIP5K evolutionary origin within the Excavata, the very earliest of eukaryotes. These different aspects of PPIP5K biology are placed in the context of a single, overriding question. Why are they bifunctional: i.e., what is the particular significance of the ability to turn PP-InsP signaling on or off from two separate ‘switches’ in a single protein?

1. Introduction to the roles of 5-InsP₇ and InsP₈ in metabolic homeostasis.

Stimulus-activated phospholipase-C releases inositol 1,4,5-trisphosphate Ins(1,4,5)P₃, the prototypical inositol phosphate signaling molecule that mobilizes cellular Ca²⁺ stores (Berridge and Irvine, 1989). Nevertheless, Ins(1,4,5)P₃ serves another vital role; it is phosphorylated by a series of kinases to yield inositol hexakisphosphate (InsP₆), which offers a relatively abundant precursor pool for a separate and unique set of diffusible intracellular messengers: the “inositol pyrophosphates” (PP-InsPs; Fig. 1A,B).

Figure 1. Two perspectives on the synthesis and metabolism of the PP-InsPs.

Panel A depicts the original ‘diamond-shaped’ viewpoint that envisages two parallel routes of InsP₈ synthesis through the collaborating kinase activities of PPIP5Ks and IP6Ks (Fridy et al., 2007; Mulugu et al., 2007). Since the PPIP5Ks have kinases and phosphatase activities, separate directional arrows are shown. For the purposes of clarity, the schematic does not show the separate phosphatases (DIPPs) that hydrolyze both the 1- and 5-phosphate diesters. Yeasts and plants also possess a different class of PP-InsP 5-phosphatase (Steidle et al., 2016; Wang et al., 2018). Panel B shows a more recent proposal that the metabolic pathway is largely cyclical in nature (Gu et al., 2017a; Shears, 2018). Here, 5-InsP₇ is considered to be the predominant substrate for the kinase activity of the PPIP5Ks (see text for details). The major roles of the DIPPs are viewed as being (a), dephosphorylation of InsP₈ through 1-InsP₇ and (b), helping to prevent 1-InsP₇ from accumulating to significant levels inside cells.

Following the discovery of PP-InsPs in the early 1990s (Europe-Finner et al., 1991; Menniti et al., 1993; Stephens et al., 1993), they were handed the role of being unwanted cousins at a family wedding: they did not attract much attention (see (Shears, 2007)). Arguably, two advances in particular dragged this topic into the cell-signaling mainstream: first, the introduction of genetic methods to study PP-InsP actions, following cloning of an inositol hexakisphosphate kinase (IP6K, E.C. 2.7.4.21; (Saiardi et al., 1999)), which adds a diphosphate group to InsP₆ (Fig. 1A,B); second, the eye-catching discovery that the IP6K product, 5-InsP₇, can exert biological actions upon transfer of its ‘energetic’ β-phosphate to certain proteins in a non-enzymatic reaction (Ganguli et al., 2019; Saiardi et al., 2004).

5-InsP₇ (more technically: 5-diphosphoinositol 1,2,3,4,6-pentakisphosphate) is the most abundant of the PP-InsPs. Abundance, is of course, relative; the concentration of 5-InsP₇ inside mammalian cells is only about 2 to 5 μM (Harmel et al., 2019; Shears, 2018). There is even less InsP₈ (officially: 1,5-bis-diphosphoinositol 2,3,4,6-tetrakisphosphate; Fig. 1A,B); its levels are usually 5- to 10-fold lower than those of 5-InsP₇ (Shears, 2018). Despite existing at these relatively low levels, both 5-InsP₇ and InsP₈ possess a number of vital regulatory roles, many of which are viewed as acting at the interface of cell-signaling and metabolic homeostasis (Azevedo and Saiardi, 2017; Chakraborty, 2017; Shears, 2018).

For example, we have published evidence that InsP₈ acts to restrain cellular ATP synthesis by down-regulating glycolysis and mitochondrial oxidative phosphorylation (Gu et al., 2017a). Thus, elimination of InsP₈ synthesis in CRISPR-generated, PPIP5K-null HCT116 cells leads to elevated levels of ATP (Gu et al., 2017a). Consistent with the latter result, there is ATP accumulation in IP6K1/IP6K2 double knockout cells, which we argue is not due to loss of 5-InsP₇, but instead reflects the collateral damage to InsP₈ synthesis (Fig. 1A,B) (Wilson et al., 2019).Consequently, insulin-mediated increases in InsP₈ levels (Nair et al., 2018), may be viewed as a signaling process that contributes to bioenergetic homeostasis through control of ATP synthesis.

Recent studies with Schizosaccharomyces pombe have revealed that there is a separate interaction of InsP₈ with mitochondria that has bioenergetic significance: the InsP₈ is required for the microtubule-dependent, active positioning of these organelles (Pascual-Ortiz et al., 2018). Dynamic control over mitochondrial placement is part of an overarching homeostatic process that matches energy supply to demand in cellular microcompartments (Moore and Holzbaur, 2018). The participation in this process of the dynein family of cytoskeletal motor proteins (Moore and Holzbaur, 2018) is very pertinent, because dynein dynamics are themselves regulated by 5-InsP₇ (Chanduri et al., 2016). Thus, the two PP-InsPs may cooperate to regulate mitochondrial motility, a process that has also been linked with a quality control pathway – mitophagy - that recycles these organelles when they become aged and/or damaged (Moore and Holzbaur, 2018).

In Arabidopsis, InsP₈ plays a role in jasmonate-dependent transcriptional responses that confer resistance to herbivores and necrotrophic fungi (Laha et al., 2015). It has also been suggested that, in Planta, InsP₈ synthesis is positively correlated with availability of Pi and ATP (Zhu et al., 2019); that is, InsP₈ synthesis is tied to bioenergetic health. It is easy to view this as a protective mechanism: immune responses are energetically costly, which is why they are only induced when needed (Lazzaro, 2015). Draining resources can impair general fitness of an organism by depriving other vital, energy-consuming processes. Perhaps the requirement for InsP₈ in the jasmonate pathway may be a means by which the cell restrains immune responses when energy resources are inadequate.

Bioenergetic significance for 5-InsP₇ is also evident from evidence that fluctuations in its levels operate as an energy-monitoring rheostat (Gu et al., 2017a; Rajasekaran et al., 2018). This process arises from the Km value of InsP₆ kinases for ATP being approximately 1 mM, which is within the range of values in which cytoplasmic [ATP] fluctuates (Gu et al., 2017a; Rajasekaran et al., 2018; Saiardi et al., 1999; Wundenberg et al., 2014). ATP-dependent changes in 5-InsP₇ concentration have considerable physiological relevance. For example, nutrient-stimulated elevations in [ATP] in pancreatic β-cells stimulate rises in 5-InsP₇ levels (Rajasekaran et al., 2018), which promote insulin secretion (Illies et al., 2007). Furthermore, a requirement of rRNA synthesis for 5-InsP₇ has been proposed to be a mechanism by which the rate of ribosome biogenesis matches cellular energy supply (Thota et al., 2015), which is an important precaution for a process that consumes up to 80% of the cellular energy required for proliferation (Schmidt, 1999). Other effects of 5-InsP₇ include inhibition of the AKT protein kinase, which attenuates metabolic signaling by insulin (Chakraborty et al., 2010). Genetic targeting of 5-InsP₇ synthesis protects mice from gaining weight while consuming a high-fat diet (Chakraborty et al., 2010; Zhu et al., 2016). The participation of 5-InsP₇ in metabolic regulation may be highly conserved; the fungal pathogen, Cryptococcus neoformans, relies on 5-InsP₇ to adjust its metabolism to allow growth in the glucose-poor environment of the host lung (Lev et al., 2015). Thus, the development of drugs that specifically target the InsP₆-kinase activity of C. neoformans could have human health benefits.

It is very relevant to the remit of PP-InsP signaling that the maintenance of cellular [ATP] depends upon the availability of inorganic phosphate (Pi) (Gu et al., 2017b; Lonetti et al., 2011; Wilson et al., 2019); PP-InsPs have been known for many years to participate in regulating cellular Pi homeostasis (Norbis et al., 1997; Saiardi et al., 2004; Saiardi et al., 1999; Schell et al., 1999). This process has considerable metabolic significance beyond serving ATP synthesis: Pi is one the cell’s most influential modulators of cellular metabolism (Alam et al., 2017).

Mechanistic insights into [Pi]-regulation by PP-InsP signals have begun to emerge; in animal cells, PP-InsPs have been found to regulate certain transmembrane Pi-transport proteins (Potapenko et al., 2018; Wild et al., 2016; Wilson et al., 2019). In plants, PP-InsPs assist transcription factors to induce the expression of Pi-responsive genes (Dong et al., 2019; Wild et al., 2016). Recently, Wilson et al (2019) reported that IP6K1/IP6K2 knockout HCT116 cells exhibit a slower rate of XPR1-dependent cellular Pi efflux than wild-type cells (Wilson et al., 2019); the latter study proposes that either 5-InsP₇ or InsP₈ is required for Pi efflux. Our studies have determined that it is InsP₈ that regulates XPR1 activity (X. Li and S. B. Shears, unpublished data).

The different protein targets that mediate PP-InsP activities - i.e., transcription factors and phosphate-transporters - all have in common strongly electropositive regions: SPX domains (Dong et al., 2019; Wild et al., 2016; Wilson et al., 2019; Zhu et al., 2019). The latter studies have assayed the rank order affinities of the SPX domains for InsP₆ and PP-InsPs. However, this body of work has indicated that preferential binding in vitro of a particular PP-InsP to an SPX domain is often, by itself, insufficient to explain functional specificity in vivo. For example, inorganic polyphosphate synthesis by the vacuolar transporter chaperone (VTC) from Saccharomyces cerevisiae is stimulated by PP-InsP binding to an SPX domain (Wild et al., 2016). Subsequent work has shown that InsP₈ is more potent than 5-InsP₇ at enhancing VTC-mediated polyphosphate synthesis, yet genetic experiments with intact yeast exclude that InsP₈ is required for polyphosphate synthesis in vivo (Gerasimaite et al., 2017). Thus, specificity of PP-InsP action upon SPX domains in vivo may also require contributions from additional protein interactions that remain to be characterized (Wild et al., 2016; Wilson et al., 2019).

To fully understand cellular functions for 5-InsP₇ and InsP₈ in a signaling context, it has become a priority to understand the regulation of their enzymatic interconversion and hence prevailing levels. A fascinating class of enzyme has this responsibility, which take the alternative names of Vip1, Asp1, VIH1/2 and PPIP5K1/2 (EC 2.7.4.24), in S. cerevisiae, S. pombe, Arabidopsis and mammals, respectively (Fig. 1A,B). Although this disparate nomenclature is inconvenient, it has developed from logical priorities of gene nomenclature (see below).

InsP₈ synthesis and metabolism: mutually competing kinase and phosphatase activities.

Naming an enzyme “PPIP5K” (Fig. 1A,B) does not exactly install a headline-grabbing sound-bite. Nevertheless, international protein nomenclature guidelines request that the name of a protein be an accurate reflection of its main function: in this case, it was logical that the mammalian kinase (‘K’) that phosphorylates PP-IP5 (i.e., the original nomenclature for InsP₇ (Shears et al., 1995)), was to become known as PPIP5K (Huang et al., 1998) (and see E.C. 2.7.4.24). Thus, in this review, the PPIP5K nomenclature is used in reference to the enzymes in eukaryotes, but with the important exceptions of the fungal and plant kingdoms, due to their priorities of pre-existing nomenclature. For example, the single gene encoding the S. cerevisiae ortholog of PPIP5K was already named Vip1, before its kinase/phosphatase activities were discovered by John York’s laboratory (Fridy et al., 2007; Mulugu et al., 2007). (As it happened, the human ortholog could not adopt VIP, as that was already ascribed to vasoactive intestinal peptide).

Vip1 was originally christened by Kathy Gould’s group. They discovered that Vip1, like its S. pombe ortholog Asp1, has a role in regulating yeast actin cytoskeleton dynamics (Feoktistova et al., 1999). The functional similarities of the two genes is conveyed (Gould, personal communication) by an asp also being a species of snake (Vipera aspis). The more obvious route would have been to pass the Asp nomenclature onto S. cerevisiae, but that was prohibited because that name was itself already allocated to the L-asparaginase genes. Similarly, the Vip nomenclature is also not universally adopted in the plant kingdom because Arabidopsis Vip refers to VirE2 Interacting Protein 1. Instead, the plant orthologs have been christened VIH1 and VIH2 (Vip1 homologs 1 and 2 (Laha et al., 2015)). Hence the large and sometimes confusing catalog of names for the same, highly conserved gene.

The early studies with Vip1 (Fridy et al., 2007; Mulugu et al., 2007) did not focus on the phosphorylation of InsP₇ to InsP₈, but instead highlighted its ability to also convert InsP₆ to 1-InsP₇; this underlies the proposition of a ‘diamond-shaped’ route to InsP₈ synthesis (see Fig. 1A). However, we have several reasons for concluding that InsP₆ phosphorylation to 1-InsP₇ is only a relatively minor pathway in mammalian cells, and others have drawn the same conclusion from experiments with yeast and plant cells (Onnebo and Saiardi, 2009; Zhu et al., 2019).

1. There is evidence that compartmentalization of InsP₆ reduces its availability to compete with 5-InsP₇ phosphorylation by the PPIP5Ks. Although the consensus in the literature is that total cellular levels of InsP₆ are 15 μM or more, there is an emerging appreciation within the field that much of this polyphosphate is not free in the cytoplasm. For example, there is currently renewed interest in an earlier study that provided evidence of InsP₆ immobilization to membranes through cation-mediated ionic 'bridges' with anionic phospholipids (Poyner et al., 1993). In addition, intracellular InsP₆ is a non-exchangeable structural cofactor for a number of proteins in yeasts and animal cells. Two particularly well-characterized examples are Adenosine Deaminase Acting on RNA type 2 (ADAR2) (Macbeth et al., 2005), and yeast N-terminal acetyltransferase, NatE (PDB entry 4XNH); the growing number of these types of proteins that have been identified implies they are sufficient to sequester a substantial proportion of the cell’s InsP₆ away from the cytoplasm (John York, personal communication).

2. Kinetic assays with the kinase domain of PPIP5K2 – under both first-order and second-order conditions (Wang et al., 2012; Weaver et al., 2013) – demonstrate that 5-InsP₇ is strongly preferred as a substrate over InsP₆ and hence would likely out-compete InsP₆ in vivo. Kinetic experiments with full-length PPIP5Ks also show a several-fold higher catalytic efficiency for 5-InsP₇ over InsP₆ (Fridy et al., 2007), although such data may be somewhat quantitatively compromised by the PPIP5K phosphatase domains, which consume the PP-InsP product of the kinase reaction (see below). Thus, a more accurate determination of the kinase parameters in the context of the full-length enzyme can be obtained using a phosphatase-dead construct that can be generated by a single amino-acid mutation (Gu et al., 2017b). In such circumstances, 5-InsP₇ is still the preferred substrate (Gu et al., 2017b).

3. There has been atomic-level validation of the enzyme favoring the substrate with a 5-diphosphate group; the binding of 5-InsP₇ within the catalytic pocket is specifically enhanced through ionic interactions of the β-phosphate with Lys214 and a magnesium atom (Wang et al., 2012).

4. The expectation that the substrate of an enzyme will accumulate in intact cells when that enzyme is knocked out, is consistent with the observation that levels of 5-InsP₇ are elevated in vip1Δ S. cerevisiae (Onnebo and Saiardi, 2009).

5. The steady state levels of 1-InsP₇ are normally barely detectable (Gu et al., 2016; York et al., 2005), except in the non-physiological context of ksc1Δ yeast, i.e., InsP₆ phosphorylation to 1-InsP₇ by Vip1 is only observed in cells that do not express InsP₆ 5-kinase activity, and hence cannot synthesize any of the preferred Vip1 substrate, 5-InsP₇ (Wilson et al., 2019; York et al., 2005). It seems noteworthy to mention here that another circumstance in which 1-InsP₇ levels have been proposed to be elevated in yeast – during Pi starvation (Lee et al., 2007) – has not been possible for other groups to replicate (Wild et al., 2016). In fact, Pi is an inhibitor of the 1-phosphatase activity (Gu et al., 2017b); withdrawal of this macronutrient would be expected to activate the 1-phosphatase. Levels if 1-InsP₇ (and InsP₈) would then be reduced, not increased.

Not only do we argue that cellular InsP₈ is mainly synthesized from 5-InsP₇, we also propose that DIPP-dependent InsP₈ dephosphorylation to InsP₆ mainly proceeds through 1-InsP₇ (Kilari et al., 2013). This pathway of 5-phosphatase dephosphorylation of InsP₈ is reinforced in plant and yeast cells by the catalytic activity of Siw14/Atg5000 (Steidle et al., 2016; Wang et al., 2018). So, we propose that at least some of the metabolic flux through the PP-InsP pathway may be cyclical in nature (Fig. 1B). In any case, we now view the functions of the PPIP5Ks largely in terms of their ability to regulate levels of InsP₈, rather than 1-InsP₇.

As mentioned above, the PPIP5K/Vip1 family of enzymes contain a phosphatase domain that competes with the activity of the kinase domain. This observation was first described by John York’s group both in published manuscripts (Fridy et al., 2007; Mulugu et al., 2007) and in contemporaneous presentations at a number of scientific symposia. That is, the enzyme contains separate 1-kinase and 1-phosphtase activities. Others have confirmed that the Asp1, and the VIH orthologs in Arabidopsis, are also active PP-InsP phosphatases (Pascual-Ortiz et al., 2018; Pohlmann et al., 2014; Zhu et al., 2019). Unfortunately, we initially and erroneously concluded that the PPIP5K phosphatase domains are inactive (Gokhale et al., 2011). We were misled by our use of recombinant constructs that were C-terminally truncated which, we later discovered, lacked residues that are necessary for activity (Wang et al., 2015). In that latter study, we reinvestigated this issue, and confirmed that the human PPIP5K and Asp1 enzymes both contain 1-InsP₇/InsP₈ 1-phosphatase activity.

Structural Considerations

There are two PPIP5K genes in mammals: types 1 and 2 (Fig. 2). These encode relatively large proteins: 160 kDa (for PPIP5K1) and 140 kDa (for PPIP5K2). The kinase domain is self-contained within the initial 25% of the N-terminal, amino-acid sequence (Fig. 2). The phosphatase domain is approximately twice as large, and occupies a central region; the remaining residues comprise an intrinsically disordered domain (IDR) that have protein scaffolding functions (Machkalyan et al., 2016). In contrast to the high degree of conservation of the catalytic domains of PPIP5K1 and PPIP5K2, the two IDRs diverge substantially, indicative of functional specialization. This C-terminal protein-protein binding domain is absent from lower eukaryotes.

Figure 2. Domain and motif schematic for human PPIP5K1 and PPIP5K2.

Schematics are shown for isoform 5 of PPIP5K1 (NCBI: XP_016878236.1), isoform 8 of PPIP5K2 (NP_001332804.1) and isoform 9 of PPIP5K2 (NP_001332805.1). Amino acid numbering for each domain is based on a previous study (Machkalyan et al., 2016). “IDR” = intrinsically disordered domain (Machkalyan et al., 2016). “CPA” = contains penta-arginine (Yong et al., 2015). See the text and Fig. 5 for explanation of the ‘ARK’ and ‘AWK’ motifs.

We have sought to determine the structures of the PPIP5Ks: such information deciphers catalytic mechanisms, defines regulatory processes at a molecular level, clarifies evolutionary relationships, and allows rational design of chemical probes that can be used to study biological functions of the enzymes in intact cells. Obviously, the key goal is to solve the structure of the full-length protein, but this is large and difficult to express, so we have taken a reductionist approach: we have solved the crystal structure of the kinase domain (Wang et al, 2012). The latter study showed ATP to be bound between two sets of anti-parallel beta sheets - a so-called 'ATP-grasp' kinase. As for the determinants of substrate specificity, there are architectural and electrostatic constraints that precisely accommodate 5-InsP₇. For example, there are two near parallel grooves on the surface of the catalytic pocket that form the shape of a ‘staggered-H’ that matches the dimensions of the substrate.

For such a highly-charged substrate, it could have been anticipated that an important role in ligand-protein binding is also played by electrostatic steering; electropositive residues that act as a 'tractorbeam' that pulls negatively-charged substrate towards the entrance to the catalytic site (Wade et al, 1998). However, PPIP5Ks utilize a more advanced and highly unusual mechanism to acquire substrate from the bulk phase: an actual ligand-binding site on the protein's surface. Once within this 'capture-site', the bound substrate is optimally oriented for subsequently being flipped into the catalytic pocket (Wang et al, 2014b; Wang et al, 2012). This “catch-and-pass” reaction mechanism was recently modeled by application of molecular dynamics simulations (An et al., 2019): we rationalized how this enzyme can harness random molecular motions to significantly reduce the entropy toll that might otherwise have been required for a coordinated deterministic transport of the substrate. As these molecular motions carry 5-InsP₇ towards the catalytic site, Glu192 takes on the role of a molecular ratchet, by moving towards the space vacated by the substrate. That is, the Glu192 electrostatically enforces one-way substrate delivery (An et al., 2019). This specialized role has been verified experimentally – a Glu192Gln mutation dramatically reduces kinase activity (Wang et al., 2014).

To date, the structure of the phosphatase domain has not been determined. We know the phosphatase domain belongs to the histidine acid phosphatase family, due to its RH[G/A]xRxP consensus sequence (Fridy et al., 2007; Mulugu et al., 2007); the catalytic function of this motif has been confirmed for Asp1, Vip1, VIH2 and PPIP5K1/2 by showing reductions in catalytic activity upon mutagenesis to Ala of the Arg and His residues (Gu et al., 2017b; Pascual-Ortiz et al., 2018; Zhu et al., 2019).

Remote from the histidine acid consensus is the H[I/V/A] ‘M2’ catalytic motif (Fridy et al., 2007; Mulugu et al., 2007; Pascual-Ortiz et al., 2018). Interestingly, the phosphatase activity in Asp1 is inactivated by mutation to Asp of the aliphatic residue that follows the M2-His (Pascual-Ortiz et al., 2018). This observation distinguishes the PPIP5K family from many other histidine acid phosphatases, which utilize Asp (or Glu) in the His+1 position as a proton donor for the hydrolytic reaction (Rigden, 2008).

At least in some other histidine acid phosphatases, this M2-His is itself important by virtue of its polar interactions with the substrate (Lindqvist et al., 1994; Schneider et al., 1993). Indeed, we have previously demonstrated that the equivalent residue in Asp1, His807, makes significant contributes to InsP₈ hydrolysis; phosphatase-domain constructs of Asp1 that host a His807Ala mutation have only about 10% of the activity of wild-type enzyme (Wang et al., 2015). We considered that further validation of the catalytic contribution of this His residue is important to the conclusions we make from sequence alignments that are discussed below. Thus, we performed new assays that use full-length Asp1: the His807Ala mutant exhibited only 10-15% of the activity of wild-type protein, against both 1-InsP₇ and InsP₈ (Fig. 3). These data contrast somewhat with an earlier study (Pascual-Ortiz et al., 2018) that concluded this same His807Ala mutant enzyme has residual catalytic activity that could not be distinguished from that of wild-type enzyme. However, in the latter study the phosphatase assays were performed in vitro for 16 hours against only 1-InsP₇, the weaker of the two PP-InsP substrates (Fig. 3), in incubation media that contained considerable quantities of ATP, which inhibits this family of phosphatases (Dong et al., 2019). Our 30 minute assays are more sensitive (Fig. 3 and (Wang et al., 2015)). While Pascual-Ortiz (2018) have also argued that the His807Ala mutation is functional in vivo (by supporting hypersensitivity to thiabendazole, a microtubule poison), this toxin-dependent phenotype was observed upon plasmid-based overexpression of the mutant protein, apparently at 20-fold higher levels than that of endogenous Asp1 (Pascual-Ortiz et al., 2018; Pohlmann and Fleig, 2010). In such experiments, the elevated levels of the mutant enzyme may functionally compensate for its inherently lower activity.

Figure 3. Asp1 phosphatase activities towards 1-InsP₇ and InsP₈; the effect of the His807Ala mutation.

Full-length recombinant Asp1 and the Asp1^H807A mutant were prepared and assayed as previously described (Wang et al., 2015).

The similarities in the domain organization of PPIP5K1 and PPIP5K2 proteins (Fig. 2) suggests that they have arisen from a gene duplication event. Duplicated genes are initially redundant, and even potentially hazardous, because they offer a substrate for homologous recombination and the accompanying loss or gain of genomic information; a frequent outcome is that one of the duplicates is inactivated with a half-life of a few million years (Force et al., 1999). Thus, duplicated genes that are maintained have generally acquired separate functions and/or new activities. What might be the nature of these specialized, individual functions?

One clue as to a biological difference between PPIP5K1 and PPIP5K2 comes from the considerable sequence divergence of the region that is C-terminal to the phosphatase domain: the intrinsically disordered region (IDR) (Fig. 2 and (Machkalyan et al., 2016)). These are protein-protein interaction domains (molecular ‘scaffolds’) that presumably have isoform-specific binding functions (Machkalyan et al., 2016). The IDR in PPIP5K1 interacts with proteins that participate in vesicular trafficking, cytoskeletal function, and lipid metabolism; no specific biological activity for any of these particular interactions has been identified to date.

Subfunctionalization of the IDRs may further arise from the existence of multiple isoforms; alternate RNA splicing yields a variety of PPIP5K1 and PPIP5K2 isoforms that vary in sequence within the IDR region (see, for example, (Yousaf et al., 2018)). One well-characterized example is an RRRRRS sequence that is encoded by some PPIP5K2 mRNAs (Yong et al., 2015). This is a functional nuclear localization sequence (NLS), which our data indicate to be silenced by phosphorylation of its C-terminal Ser (Yong et al., 2015). Sequence alignments indicate some PPIP5K1 isoforms contain a corresponding region, but the NLS is functionally disrupted by an Arg-to-Cys substitution (Fig. 4). Indeed, there is no evidence that any PPIP5K1 is present in the nucleus (Fridy et al., 2007; Gokhale et al., 2013; Gokhale et al., 2011; Yong et al., 2015).

Figure 4. Conservation of the CPA domain.

Shown is an alignment (Clustal Omega) of the CPA domains of PPIP5K2, isoform 9, and PPIP5K2 from the Elephant shark (C. milii), which is selected for this comparison as it is the evolutionarily oldest living vertebrate (Venkatesh et al., 2014). The identical sequences (95%) are highlighted in yellow. Also illustrated is the poor conservation in PPIP5K1 of the PPIP5K2-specific CPA domain and its eponymous RRRRRS sequence (27% identity highlighted in cyan). The identical amino acid residues in the two PPIP5Ks from C. milii are also noted with asterisks. Human PPIP5K Accession numbers are given in the Figure 2 legend. Accession numbers for C. milii PPIP5K1 and PPIP5K2 (from ensemble.org) are ENSCMIT00000019741 and ENSCMIT00000043862, respectively.

As noted previously (Yong et al., 2015), the penta-arginine motif lies within a larger domain (named CPA for ‘contains penta-arginine’) that is extremely well conserved. This is well-illustrated in Fig. 4, which shows that the CPA region of human PPIP5K2 is 95% identical to the ortholog in the elephant shark, the evolutionarily oldest living vertebrate (Venkatesh et al., 2014). In contrast, the corresponding regions of elephant shark and human PPIP5K1s share only 27% identity (Fig. 4). Tight and specific conservation of the CPA domain only in PPIP5K2 orthologs speaks strongly to its biological value.

Our further analysis of multiple sequence alignments has uncovered another potential example of subfunctionalization of the two PPIP5K genes: residues 294-296 of human PPIP5K2 comprise Ala-Trp-Lys (‘AWK’; Fig. 5A,B,C). Structural data show Trp295 is surface exposed (Wang et al., 2012). We hypothesize that there must be considerable significance to having such a hydrophobic amino acid on the surface of a soluble protein, since otherwise it would be expected to encounter severe negative evolutionary pressure. In fact, others have noted that conservation of Trp on a protein surface very likely reflects it being part of a protein-protein binding site (Ma et al., 2003). Interestingly, we have not found an example of either Trp, or any other hydrophobic residue, in the corresponding position of human and other vertebrate PPIP5K1s (residues 305-307 in human, Ala-Arg-Lys; an “ARK” motif).

Figure 5. A surface exposed Trp in the kinase domain of PPIP5K2 is highly-conserved.

Panels A and B show multiple sequence alignments (Clustal Omega) of representative vertebrates within a region which contains the surface-exposed Trp295 (yellow highlight) in the kinase domain of human PPIP5K2. The alignments also show that Arg is the corresponding amino-acid residue in PPIP5K1. Panel C shows the equivalent regions of three groups of PPIP5Ks: the top section is representative of metazoans that contain only one PPIP5K1 gene; the middle section are representative examples of organisms with duplicated but relatively undifferentiated PPIP5K1 genes, denoted ‘a’ and ‘b’; the lower section shows differentiated PPIP5K1 and PPIP5K2 from coelacanth and elephant shark. In all panels, common names are used for clarity, and the Trp residue in the PPIP5K2-specific ‘AWK’ motif is highlighted in yellow. Human PPIP5K Accession numbers are given in the Figure 2 legend. Others are as follows (from Genbank or ensembl.org, except as noted). PPIP5K1: C. milli, ENSCMIT00000019633; Cat, XP_011281336.1; Chicken, XP_015147636.1; Coelocanth, ENSLACP00000012723; Cow, NP_001098824.1; Dog, XP_005638405.1; Dolphin, XP_026939920.1; Elephant, XP_023414532.1; Horse, ENSECAP00000014895; Microbat, XP_014315001.1; Mouse, NP_848910.3; Opossum, XP_001365498.2; Panda, XP_002913227.2; Xenopus, ENSXETP00000061250. PPIP5K2: C. milli, ENSCMIT00000043862; Cat, ENSFCAP00000018399, Chicken, XP_004949357.1; Coelacanth, XP_014342401.1; Cow, NP_001178081.2, Dog, XP_005618108.1, Dolphin, ENSTTRP00000000996, Elephant, XP_023409576.1; Horse, XP_005599579.1; Microbat, XP_006081247.1; Mouse, NP_776121.4; Opossum, XP_007486299.1, Panda, XP_019659196.1, Xenopus, XP_017952576.1. Proposed PPIP5K1 (see text): C. elegans, NP_740855.2; Centipede, SMAR005830-PA; Choanaflagellate, XP_001746616.1, Fruit fly, NP_001097041.2; Scabies mite, KPM06372.1; Spider mite, XP_015794659.1; Sponge, XP_019852345.1; Trichoplax, XP_002108151.1. Proposed PPIP5K1a (see text): Acornworm, XP_006815874.1; Sea squirts Ciona intestinalis and Ciona savignyi, ENSCINP00000031546 and ENSCSAVP00000013135, respectively; Sea urchin, SP-HISPPD2A. PPIP5K1b (see text): Acornworm, Sakowv30022281m (from https://groups.oist.jp/molgenu/hemichordate-genomes); Sea squirts C. intestinalis and C. savignyi, ENSCINP00000018023 and ENSCSAVP00000013136, respectively; Sea urchin, XP_011670011.1.

There are also some catalytic differences between the two PPIP5K2 isoforms. For example, PPIP5K2 is inherently more kinase-competent (as explained above, in full-length proteins this feature is most clearly observed when the phosphatase-domain is rendered inactive with a single-site mutation (Gu et al., 2017b)). Additionally, Pi rheostatically modulates the kinase activity of PPIP5K2, but not that of PPIP5K1 (Gu et al., 2017b). Thus, the relative levels of expression of PPIP5K1 and PPIP5K2 might be important in determining InsP₈ levels, and their responses to certain stimuli, in a cell- and tissue-dependent manner.

Phylogenetics

The cloning in 2007 of PPIP5K orthologs revealed their existence in exist in animals, plants and yeasts (Choi et al., 2007; Fridy et al., 2007; Mulugu et al., 2007), so it has been clear for some time that these proteins are evolutionarily ancient. Nevertheless, we can ask several questions: when in eukaryotic evolution did the PPIP5K1/PPIP5K2 gene duplication lead to the functional diversification of these two proteins? Can it be determined which of these two PPIP5K genes is ancestral? Are these proteins and their constituent domains restricted to eukaryotes or do they have origins in non-eukaryotic domains of life?

To pursue these queries, we searched for PPIP5K orthologs from other a number of eukaryotic genomes that we considered might be instructive, especially those that are the more completely annotated. We screened publicly available repositories including NCBI, Ensembl (http://ensembl.org; part of the European Bioinformatics Institute), the Joint Genome Institute (JGI, http://genome.jgi.doe.gov), as well as more focused collections: Echinobase (echinobase.org), and the Genome Portal at JGI (Joint Genome Institute; https://genomeportal.jgi.doe.gov/pages/tree-of-life.jsf). In searching for PPIP5K orthologs, we used BLAST and hidden Markov models (HMMs) with the assistance of the Hmmer v3.1 hmmsearch function (Eddy, 2011) to identify the histidine acid phosphatase superfamily domain. Almost without exception, all of the PPIP5K genes that we identified encoded the kinase and phosphatase domains.

We constructed a phylogenetic tree with representatives from the major Phyla in the Kingdom Animalia (Fig. 6). These data show that pre-Deuterostome Phyla all have only one PPIP5K-like gene. We propose this ancestral gene is more PPIP5K1-like, based first on it not encoding the PPIP5K2-specific CPA domain (not shown, but see Figs. 2, 4). Second, this ancient gene family encodes a PPIP5K1-specfic ARK motif (or near equivalent: [A/S][R/K][Q/K/R]) rather than the corresponding AWK motif in PPIP5K2, in which the hydrophobic Trp has the unusual characteristic of being surface-mounted (see above and Fig. 5A,B,C). The presence of a single PPIP5K1-like gene in both lower metazoans and the choanoflagellates (Fig. 6), the unicellular sister group of animals (Sebe-Pedros et al., 2017), further shows that multicellularity per se does not depend upon PPIP5K gene duplication.

Figure 6. Diversification of PPIP5Ks within the Metazoans.

A Bayesian inference tree of PPIP5K genes shown with the posterior probability support at each of the nodes. The tree was rooted with the choanoflagellate Monosiga brevicollis PPIP5K as a unicellular outgroup to Metazoans. The Metazoans that contain only one PPIP5K1 gene are highlighted in green. The early Chordates, Hemichordates, and Echinoderm that contain two copies of an undifferentiated PPIP5K1 gene are highlighted in blue. The higher Chordates which contain differentiated PPIP5K1 and PPIP5K2 genes are highlighted in purple and red, respectively. Only representatives of major Orders are shown in the tree throughout. Common names are used in the figure for clarity. For the non-vertebrates, fruitfly (D. melanogaster), Trichoplax (Trichoplax adherens), centipede (Strigamia maritima), spider mite (Tetranychus urticae), scabies mite (Sarcoptes scabiei), sea urchin (Strongylocentrotus purpuratus), velvet spider (Stegodyphus mimosarum), acorn worm (Saccoglossus kowalevskii) Lancelet (Branchiostoma floridae) are shown. Branch lengths are measured in the number of substitutions per site.

Our phylogenetic data identify the origin of a PPIP5K duplication early in the Deuterostome Superphylum: we identified two PPIP5K genes (‘a’ and ‘b’ in Figs. 5C, 6) in an Echinoderm, the sea urchin (Strongylocentrotus purpuratus), and a Hemichordate, the acorn worm (Saccoglossus kowalevskii). We also found evidence of two PPIP5K1 genes in representatives of early chordates, i.e, the Tunicates (the sea squirts, Ciona savignyi and Ciona intestinalis; Fig. 6). These separate sets of duplicated genes all possess the PPIP5K1-like ARK signature (Fig. 5C); we could not see any other evidence of the putative subfunctionalization we describe above. Such duplication predates the chordate whole genome duplication events (Holland and Ocampo Daza, 2018). It is the latter which may explain the origin of functionally-distinct PPIP5K1 and PPIP5K2. The Class Elasmobranchii (cartilaginous fish) and the elephant shark (Callorhinchus milii) are the oldest representatives of the Vertebrae in which the PPIP5K2-specific CPA domain and the AWK motif are evident (Fig. 4,5 and see above).

Our phylogenetic analysis includes representatives of the Teleosts, which retain the signature of another whole genome duplication event that was fish-specific (Taylor et al., 2003). We examined ten such genomes (tetradon, fugu, stickleback, tilapia, platyfish, molly fish, medaka, cod, cave fish, and zebrafish; data for cod and zebrafish are included in Fig. 6). Within these genomes, we found two PPIP5K1-type genes (named ‘K1_1’ and ‘K1_2’) to be widely distributed. In contrast, only a single PPIP5K2-like gene was identified, and only in the genomes of cod, cave fish and zebrafish. As mentioned above, a common fate of one of a pair of duplicated genes is its inactivation, unless subfunctionalization and/or neofunctionalization occurs (Force et al., 1999). It is worth noting that we observed (not shown here) that the zebrafish PPIP5K2 gene encodes both the putative protein-protein interaction motif (AWK), and in one of two transcripts (Ensembl: ENSDARP00000115535 and ENSDARP00000097491), the regulated nuclear localization sequence (RRRRS) within a CPA domain (see above discussion). It is surprising that these specific properties have not led to the wider retention of PPIP5K2s throughout the Teleosts.

It could have been useful to classify the nature of the PPIP5Ks in another invertebrate Chordate, the lancelet (Branchiostoma floridiae) in which we found only one PPIP5K gene, and also the lamprey (Petromyzon marinus) which appears to have two PPIP5K genes, but the genome sequencing of these genes is too incomplete to justify making any conclusions.

We next searched for candidate proto-PPIP5K1 genes; we screened the genomes of single cell protists, given that they represent the basal branch of Eukaryota. Our search methodology described above did not identify any PPIP5K genes in any available protist genomes. We therefore used HMMer 3.1 v1 to construct a custom HMM that is more specific for PPIP5K-like kinase and phosphatase domains; as input, we used eight distantly related metazoan PPIP5K proteins (< 80 % sequence identity). Despite the evolutionary distance between protists and metazoans, and the sparseness of the available genome assemblies, our custom HMM identified a bifunctional PPIP5K gene in at least some representatives of all Protista Phyla. The genes were prevalent in the Chromalveolata. As for the Rhizaria, only five genomes are available, but three do have a bifunctional PPIP5K. In the Amoebozoa, we found only one available genome that contains a PPIP5K; this sole exception to an apparent Phylum-wide PPIP5K-gene loss is the Intraphylum Mycetozoa (the slime molds); interestingly, these organisms do not possess a PPIP5K-like phosphatase, either within the C-terminus of the PPIP5K-kinase, or as an independent gene. This likely helps explain the slime mold’s ability to synthesize unusually high (60 to 180 μM) levels of InsP₇ and InsP₈ (Europe-Finner et al., 1991; Pisani et al., 2014).

We included Microsporidia in our search for ‘protist’-PPIP5Ks, because a recent study (Choi and Kim, 2017) has breathed new life into an old hypothesis (Vossbrinck et al., 1987) that this group of pathogens have an early Protozoan origin. (We note that many others still argue an alternative origin for Microsporidia, namely as one of the earliest-emerging clades of Fungi (Capella-Gutierrez et al., 2012; Dacks and Doolittle, 2001)). In any case, we found PPIP5K-like genes in two species, Nematocida parisii and Nematocida displodere (Fig. 7), which are two pathogens of C. elegans (Luallen et al., 2016). Our search criteria failed to detect a PPIP5K kinase or phosphatase domain in any other Microsporidia; this situation may reflect the considerable gene loss that has accompanied their evolution into obligate parasites (Luallen et al., 2016).

Figure 7. Candidate PPIP5Ks in protists.

Shown is a multiple sequence alignment (Clustal Omega) of the kinase domain, and the separate RH[G/A]xRxP consensus and the remote M2 motif (here expanded to H[I/L/V/A]), from human PPIP5K2 (isoform 8), PPIP5K1 from C. milii, Asp1 from S. pombe (O74429.1) and putative PPIP5Ks from N. displodere, G. lambia and B. saltans. The key below the alignment describes the significance of those residues that are highlighted in color, as ascertained from structural and mutagenic studies (see (An et al., 2019; Wang et al., 2012; Wang et al., 2014)). The PPIP5K1-like ‘ARK’ motif is highlighted in gray. Note the presence of two RH[G/A]xRxP motifs in the PPIP5K from G. lamblia. In the last alignment block (lower right) the highlighted M2 His residue corresponds to the catalytically-important H807 in Asp1 (see Fig. 3 and (Wang et al., 2015)). Accession numbers (GenBank) are: G. lamblia, EFO64340.1; N. displodere, OAG30117.1. To identify a candidate PPIP5K in Bodo saltans, the transcriptome (ERR152949) was downloaded from the Sequence Read Archive and assembled with Trinity (Grabherr et al., 2011), using default settings from which proteins greater than 100 amino acids were predicted using Transdecoder (Haas et al., 2013). The predicted protein set was then searched with our custom HMMs as described in the text

We believe it is also significant to have identified candidate bifunctional PPIP5Ks in the two Excavata Phyla, which by the consensus of opinion represent the earliest of the Eukaryota (e.g., see (Sebe-Pedros et al., 2017)): the Euglenozoa (e.g., Bodo saltans; Gardia lamblia; Fig. 7) and the Metamonada (Spironucleus salmonicida).

Of course, for each of these candidate proto-PPIP5K genes, the nature of any kinase and/or phosphatase activity remains to be tested experimentally. Nevertheless, some further insights can be gained from a multiple sequence alignment of these kinase domains with the corresponding region of human PPIP5K2 and, for comparative purposes, the PPIP5K1 from C. milii. (Fig. 7). In this alignment, we have highlighted residues known to interact with ATP, magnesium, and the InsP₇ substrate, as determined from the crystal structure of the human PPIP5K2 kinase domain, together with mutagenic experiments (Wang et al., 2012). Significantly, almost all of these catalytically-important residues are either identical or conservatively-substituted in the putative kinase domains of N. displodere, and the representatives of the Excavata, G. lamblia and B. saltans (Fig. 7). The apparent conservation of nucleotide-binding residues is, perhaps, not so unexpected given the high degree of structural similarity of the ATP-biding lobe within kinase-families in general. This can even apply when the kinase catalytic pocket diverges to accommodate alternate substrates with varying physico-chemical properties (Shears and Wang, 2018). So what we think is especially supportive of the possibility of PP-InsP phosphorylation by these Excavata proteins is the tight conservation of an array of the electropositive residues (Fig. 7) that are required not only to bind InsP₇ but also for charge neutralization during the catalytic process (Wang et al., 2012). There is even conservation of the Glu residue (Glu192 in human PPIP5K2; Fig. 7) that operates as a molecular ratchet that ‘pushes’ substrate from the capture site to the catalytic pocket (An et al., 2019). Additionally, the kinase from B. saltans has an appropriately located, PPIP5K1-specific ARK motif (Fig. 7, gray highlight, and see above); a near equivalent is also present in N. displodere (VRK).

Conservation of bifunctionality in the three candidate proto-PPIP5Ks is evident in their RH[G/A]xRxP phosphatase motifs (Fig. 7); interestingly, the protein in G. lamblia has two such signature sequences. The latter observation raises the possibilities of either a second active phosphatase activity, or a separate (regulatory?) ligand-binding site. We also identified an appropriately aligned equivalent to the H[I/V/A] M2 motif (HL, Fig. 7). (This is where our consolidation of the catalytic importance of the M2-His (Fig. 3) assumes particular significance). We therefore conclude that, overall, our data construct a strong case that the most ancient bifunctional PPIP5Ks - or proto-PPIP5Ks - emerged prior to the emergence of the Excavata, some 2.1 billion years ago, even before the plant-animal-fungal radiation (Sebe-Pedros et al., 2017). Finally, we examined available Bacterial, Archaeal, and Viral proteins, but we did not find any genes that we could propose to be PPIP5K-like.

Why bifunctionality?

The evolutionary conservation of a proto-PPIP5K kinase/phosphatase that emerged proximal to the very foot of the eukaryotic tree indicates that the occurrence of two separate and competing catalytic sites within a single protein imparts highly-important biological properties. Yet it is an exceptionally rare phenomenon, with only one other example in eukaryotes: the 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase family (the PFKFBs; E.C. 2.7.1.105).

What is the particular significance of the ability to control whether PP-InsP signaling is on or off from two separate ‘switches’ (the kinase and phosphatase) in a single protein? A well-known beneficial characteristic of any metabolic interconversion is that separate but reciprocal regulation of the ‘forward’ and ‘reverse’ reactions brings considerable amplification to the control of net metabolic flux through the pathway (Newsholme, 1971). Nevertheless, in most cases such cycles are directed by separate proteins. In such a case, a coordinated output from the signaling system may be compromised by stochastic cell-to-cell variation in the expression of each of the individual proteins. This could be a particular problem when a multicellular or tissue-wide coordinated response is demanded. Uncoordinated biological output is avoided when co-expression of the competing catalytic activities is strictly enforced - such as when they both occur in a single protein (Dasgupta et al, 2014). That is probably one key property of the bifunctionality of PPIP5Ks.

However, coordinated multicellular responses are less critical to unicellular eukaryotes, and yet they also express bifunctional PPIP5Ks (see above). Thus, the physical linkage of the separate kinase and phosphatase domains must confer other advantages. For example, the kinetic properties of the human PPIP5Ks are such that their kinase domains operate under zero-order conditions, whereas the activities of the phosphatase domains are substrate-limited (i.e, first-order) (Gu et al, 2017). Mathematical modeling of just such a situation for other bifunctional proteins with competing active sites (PFKFBs and some prokaryotic proteins), if extended to the PPIP5Ks, leads to a prediction that InsP₈ exhibits concentration robustness (Gu et al., 2017b; Straube, 2013). That is, PPIP5K kinase activity may be considered to be inherently insensitive to changes in 5-InsP₇ levels. The practical outcome is that bifunctionality of PPIP5Ks can ensure specificity of PP-InsP signaling, by stabilizing InsP₈ levels during periods of stimulus-dependent regulation of 5-InsP₇ levels.

We have proposed that the presence of both kinase and phosphatase domains in the same protein facilitates allosteric communication between them. The benefit of this aspect of bifunctionality is arguably most dramatically demonstrated by one of our recent studies: a PPIP5K2 kinase domain construct that physically lacks most of the phosphatase domain exhibited many-fold higher kinase activity than a full-length version of PPIP5K2 that hosts a catalytically-dead phosphatase domain (Yousaf et al., 2018). In other words, the mere presence of the phosphatase domain (and/or possibly the IDR) somehow constrains kinase activity. Furthermore, we have demonstrated that Pi inhibits the phosphatase activities of PPIP5K1 and PPIP5K2 (Gu et al., 2017b), perhaps (by analogy with other acid phosphatases (Araujo and Vihko, 2013)) through product competition at the active site. In the case of PPIP5K2, Pi also activates the kinase domain (Gu et al., 2017b). In view of our proposal that there is allosteric communication between the two catalytic domains, we propose it is the interaction of Pi with the phosphatase domain that indirectly regulates the kinase; in any case, reciprocal regulation of PPIP5K2 kinase and phosphatase activities by Pi can promote InsP₈ synthesis (Gu et al., 2017b). We further believe this is a key component of a higher-order homeostatic response by which InsP₈ acts as both a sensor and a regulator of cellular Pi transport between the cell and its extracellular environment (X. Li and S.B. Shears, unpublished data).

Remarkably, allosteric communication may be conserved in S. cerevisiae Vip1; its phosphatase activity is not only inhibited by Pi but also by Mg-ATP (Zhu et al., 2019). Moreover, assays with full-length Vip1 show a dramatic flip in activity, in favor of the kinase, when [Mg-ATP] is raised from 5 to 7.5 mM, yet within that concentration range, there is little change to the phosphatase activity per se, which is near-maximally inhibited by 4 mM Mg-ATP (Zhu et al., 2019). Thus, tight evolutionary conservation of bifunctionality appears to ensure that the kinase and phosphatase activities do not operate autonomously, but instead collaborate to regulate InsP₈ synthesis. Moreover, this seems yet another example of the close relationship between bioenergetic homeostasis and PPIP5K catalytic activities.

We recently demonstrated that the inositol lipid, PtdInsP₂, is also an inhibitor of the PPIP5K1 phosphatase domain (Nair et al., 2018); this inhibition was observed when the full-length protein was incubated with a physiologically-relevant concentrations of PtdInsP₂ that was incorporated into large unilamellar vesicles. Perhaps PtdInsP₂, like Pi, also targets the active site. Since some PPIP5K1 is found associated with the plasma membrane (Gokhale et al., 2011), which also hosts much of the cell’s PtdInsP₂, this may permit localized accumulation of InsP₈. ‘Global’ cell-population assays of hormone-dependent elevations in InsP₈ levels have uncovered only relatively modest changes (<2-fold) (Nair et al., 2018). But this measurement may mask much higher, compartmentalized responses.

Our work with recombinant Asp1 led us to propose another possible regulatory mechanism for InsP₇ and InsP₈ interconversion; we have characterized a [2Fe-2S] iron-sulfur cluster in this enzyme’s phosphatase domain that inhibits its catalytic activity (Wang et al., 2015). Perhaps this speaks to iron-dependent regulation of PP-InsP turnover in S. pombe. However, we are aware of some isolated examples of the non-physiological association of iron and sulfide with recombinant proteins expressed in E. coli (Archer et al., 1994; Leartsakulpanich et al., 2000; Stepanyuk et al., 2008). We raise this concern because a recent study could not detect an Fe-S cluster in recombinant Asp1 (Pascual-Ortiz et al., 2018). We would like to see a resolution of this disparity. In any case, we could not detect an Fe-S cluster in Vip. We have not yet been able to interrogate human PPIP5Ks for this property; the spectroscopic and/or structural studies that are essential for defining Fe-S clusters require mg quantities of pure protein, which due to the size of the full-length human proteins, are difficult to prepare.

Bifunctionality is likely to amplify the net effects of small changes to catalytic activity to the separate kinase and phosphatase domains. Supportive evidence comes from our observation of an Arg837His missense variant of the PPIP5K2 gene (rs548137246; 1% minor allele frequency in the South Asian population). This version of the protein has 21% less phosphatase activity than the wild-type enzyme (Yousaf et al., 2018). Since Arg837 lies 8 residues C-terminal of the H[D/I/V/A] M2 motif, the latter should now be considered as possibly part of a larger and more complex structure/function contribution to catalysis. For example, Arg837 may bind the negatively charged PP-InsP substrate; a His substituent would not be so effective in this role, as its side chain is less polar at physiological pH (Mangold et al., 2011). Another possibility is that Arg837 may participate in a structurally-stabilizing salt bridge, which could also be disrupted by the His substitution (Bosshard et al., 2004).

In any case, the Arg837His variant has 60% higher kinase activity, i.e., 3-fold greater than the 21% decrease in phosphatase activity (Yousaf et al., 2018). Since the missense variant lies in the phosphatase domain, we interpret this result as being further evidence that the kinase and phosphatase activities are in allosteric communication. Moreover, the Arg837His variant is associated with a profound phenotype in humans: deafness resulting from hair cell loss within the inner ear (Yousaf et al., 2018). Overall, these data speak to the importance of bifunctionality to human health and well-being through the controlled interconversion of 5-InsP₇ and InsP₈ by the PPIP5Ks. This is an important area of research into the PPIP5K ‘two-way switch’ that we are actively pursuing in our laboratory.