Inositol lipids: from an archaeal origin to phosphatidylinositol 3,5-bisphosphate faults in human disease



The last couple of decades have seen an extraordinary transformation in our knowledge and understanding of the multifarious biological roles of inositol phospholipids. Herein, I briefly consider two topics. The first is the role that recently acquired biochemical and genomic information – especially from archaeons – has played in illuminating the possible evolutionary origins of the biological employment of inositol in lipids, and some questions that these studies raise about the ‘classical’ biosynthetic route to phosphatidylinositol. The second is the growing recognition of the importance in eukaryotic cells of phosphatidylinositol 3,5-bisphosphate. Phosphatidylinositol 3,5-bisphosphate only entered our phosphoinositide consciousness quite recently, but it is speedily gathering a plethora of roles in diverse cellular processes and diseases thereof. These include: control of endolysosomal vesicular trafficking and of the activity of ion channels and pumps in the endolysosomal compartment; control of constitutive and stimulated protein traffic to and from plasma membrane subdomains; control of the nutrient and stress-sensing target of rapamycin complex 1 pathway (TORC1); and regulation of key genes in some central metabolic pathways.


2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid






archaetidylinositol 3-phosphate




myo-inositol 1,4,5-trisphosphate


myo-inositol 1-phosphate


d-inositol 3-phosphate


last common ancestor of all eukarya


myo-inositol 3-phosphate synthase


mammalian target of rapamycin complex 1


type III phosphoinositide 3-kinase


phosphatidylinositol 3-phosphate 5-kinase


phosphoinositide (s)


β-propellers that bind phosphoinositides




phosphatidylinositol 3,4,5-trisphosphate


phosphatidylinositol 3,5-bisphosphate


phosphatidylinositol 4,5-bisphosphate


phosphatidylinositol 3-phosphate


phosphatidylinositol 4-phosphate


phosphatidylinositol 5-phosphate


WD-40 protein interacting with phosphoinositide


The transformation of biological knowledge that has occurred during the last half-century has included remarkable revelations about the biology of the simple polyol inositol, especially as a constituent of essential membrane phospholipids. Knowledge was very limited at the start of that period – we knew that some organisms needed myo-inositol (Ins) for growth; that there were three simple glycerophosphoinositides, i.e. phosphatidylinositol (PtdIns) and its much less abundant phosphorylated derivatives PtdIns 4-phosphate (PtdIns4P) and PtdIns 4,5-bisphosphate (PtdIns(4,5)P2) (collectively known as phosphoinositides (PPIn)); that plants contained an Ins sphingolipid; and that mycobacteria contained mannosylated PtdIns derivatives. Much of the limited amount of work performed had focused on possible functions for PPIn, especially PtdIns(4,5)P2, in mammalian brain: see [1] for the early history of the field.

But what was the biological value of making PPIn? We had no idea. Now we have eight ‘standard’ PPIn – PtdIns, three PtdIns phosphate isomers, three PtdIns bisphosphate isomers, and PtdIns 3,4,5-trisphosphate (PtdIns(3,4,5)P3) [1-3]. There are also Ins lipid-based anchors for membrane proteins, which may have as their cores either PtdIns or inositolphosphoceramide [4], and there are diverse Ins-based sphingolipids – notably in plants [5], fungi [6], and protists. And, the biosynthesis of inositolphosphoceramide, the simplest Ins sphingolipid, has become a candidate drug target in the search for therapies against trypanosome-mediated diseases [7].

At least five of these PPIn – PtdIns, PtdIns4P, PtdIns(4,5)P2, PtdIns 3-phosphate (PtdIns3P), and PtdIns 3,5-bisphosphate (PtdIns(3,5)P2) – are almost ubiquitous in eukaryotes, in which they have many widely conserved functions. Each PPIn appears to be to some degree confined – both physically and functionally – to one (or sometime more) cell compartment(s) or to a functionally linked series of cell compartments (see, for example, Fig. 1 and [2, 3, 8]). Note, however, the emphasis in the legend to Fig. 1 that any such 2-dimensional cartoon represents an oversimplification, and it is either known or at least likely that several PPIn execute specific functions at multiple locations in the cell.

Figure 1.

Subcellular distributions of the seven PPIn. Each PPin is made and functions at particular subcellular location(s). This illustration schematically identifies a predominant functional location for each PPIn in eukaryotic cells. Note that the ‘predominant’ in this statement is very important, as PPIn often have functions at more than one site within cells. For example, much recent work on the synthesis and functions of PtdIns4P has focused on its roles in the trans-Golgi region. This is despite the fact that it has long been known that it has other functions elsewhere, e.g. as substrate for PtdIns(4,5)P2 synthesis at the plasma membrane (reviewed in [99]). Only recently, however, has the original localization of PtdIns 4-kinase activity at the plasma membrane of mammalian cells [105, 106] been given a clear molecular basis, by the demonstration that the plasma membrane only properly preserves its functional identity so long as PtdIns4P is being synthesized there by PtdIns 4-kinase-IIIα [107]. Reproduced, with permission, from [5].

The ubiquity of PPIn in eukaryotes suggests that the ability to make some form of PPIn probably evolved in some unidentified ancestor of all extant eukaryotes, and this, in turn, suggests that the development of some of their functions in vesicular trafficking and in cytoskeletal regulation must have been underway before early eukaryote diversification [9-11].

Nature has made use of at least six of the nine stereoisomers of inositol (hexahydroxycyclohexane) (see Fig. 1 of [9]). Although Ins has attracted most attention, other inositol isomers (including neo-inositol, d-chiro-inositol, scyllo-inositol, muco-inositol, and epi-inositol) occur in nature in various guises [9, 10].

For example, some of the most abundant phosphate compounds in many soils and sediments are polyphosphates of several inositols [12-14]. Entamoeba histolytica and Mastigamoeba (previously Phreatamoeba) balamuthi, two evolutionarily ‘primitive’ amoebae, make neo-inositol hexakisphosphate by an undefined route [15]. At present, it is not known what proportion of the megatonnes of these compounds that exist in the environment are residues from the decay of plants rich in phytate (myo-inositol hexakisphosphate) and how much is newly made by the biota resident in soils and sediments.

Since the discovery 30 years ago that Ins 1,4,5-trisphosphate (Ins(1,4,5)P3) is a Ca2+-mobilizing second messenger, at least in metazoans [16], there has been an explosion of interest in the functions fulfilled within cells by the many soluble, nonlipid, phosphorylated derivatives of Ins. The unexpected track onto which this discovery directed us has led to the discovery of many functions for other Ins derivatives [17, 18]. However, these are large and separate topics that are beyond the scope of this article.

Inositol lipids – diversity and evolutionary origin in the Archaea

There is fairly general agreement that the last common ancestor of all eukaryotes (LECA) incorporated functional elements from both bacterial and archaeal ancestors [19, 20]. This raises the question of whether LECA obtained its ability to make PPIn and Ins polyphosphates from one of these progenitors or evolved these abilities subsequently, during some very early period in eukaryotic diversification. If the former is true, might the archaeal or bacterial donor organism have already involved PPIn in ‘primitive’ versions of any of the functions that they fulfil in extant eukaryotes?

There have been relatively few biochemical analyses of the phospholipid compositions and metabolism of Ins and its derivatives in bacteria or archaeons, but the many bacterial and archaeal genomes that have recently been sequenced have proved very illuminating. Most of the few archaeons that have been biochemically analysed contain diradylglycerol-based lipids that include an Ins 1-phosphate (Ins1P) headgroup identical to that in eukaryotic PPIn. By contrast, few bacteria contain Ins-containing membrane lipids (for details and references, see [9, 10, 20]). Archaeons, many of which live in hostile and largely abiotic environments such as hot springs, need a source of Ins with which to make their PPIn, so it can be assumed that most of them make their own Ins. All biological Ins is made by a single type of NADH-dependent d-myo-inositol-3-phosphate synthase (MIPS). This converts the ubiquitous central metabolite glucose 6-phosphate, via a 5-ketoglucose-6-phosphate intermediate, to Ins 3-phosphate (Ins3P) (which is synonymous with l-myo-inositol-1-phosphate) [21, 22]. Ins3P is dephosphorylated by inositol monophosphatase to yield free Ins for later reactions. Figure 2 summarizes this pathway and a number of the other metabolic reactions that are commonly involved in the metabolism of Ins phospholipids in the various classes of organism.

Figure 2.

Outline of metabolic pathways for PPIn in major groups of organisms. The boxes indicate in which groups of organisms there is evidence for each pathway (see the text for details). Glc6P, glucose 6-phosphate; InsPase, inositol monophosphatase; PIS, PtdIns synthase; PI4K, PtdIns 4-kinase; PIPKI, PtdIns4P 5-kinase; PIPKII, PtdIns5P 4-kinase; PI3KIII, PtdIns 3-kinase; Fig 4/Sac3, PtdIns(3,5)P2 5-phosphatase; MTM, myotubularin (PtdIns3P/PtdIns(3,5)P2 3-phosphatases); PI3KI, PtdIns(4,5)P2 3-kinase; PTEN, PtdIns(3,4,5)P3 3-phosphatase.

Biochemical and genomic exploration has shown that functional MIPS genes are present in most archaeons (extreme halophiles are a notable exception, for possible reasons discussed in [9]), in a small and select minority of bacteria (including most or all actinobacteria and some hyperthermophiles), and in almost all eukaryotes (although they are often not expressed in all tissues) [9, 10]. There is a suggestion that many of the MIPS genes found in bacteria arrived by lateral gene transfer from archaeons [18]. Among the archaeons, MIPS genes are present in diverse organisms spanning the two most phylogenetically established clades (Euryarchaeota and Crenarchaeota) [10, 23]. Despite the recognition that lateral gene transfers among archaeons are common, this near-ubiquity of archaeal MIPS genes suggests that the first MIPS was probably ‘invented’ by an archaeon that pre-dated the separation of the Euryarchaeota and Crenarchaeota. If this is so, then the synthesis of Ins by MIPS and its incorporation into PPIn – together, presumably, with some ‘early’ functions of PPIn – evolved as a part of the core biochemical apparatus of archaeons long before the emergence of LECA and of all the modern eukaryotic functions of PPIn [9-11].

The diradylglycerol backbones to which the Ins1P headgroup is attached in the PPIn of archaeons are remarkably different from those in eukaryotes and bacteria. The archaeal phospholipid backbones are typically sn-2,3-diphytanylglycerols (which are known as archaeols), but the equivalent structures in most bacterial and eukaryotic phospholipids are sn-1,2-di-acylglycerols. PPIn biosynthesis is initiated by transfer of a phosphatidate (sn-1,2-diacylglycerol 3-phosphate) or an archaetidate (Arc) (archaeol 1-phosphate) moiety from CMP to a hydroxyl group in the headgroup precursor. Evidence from the Koga group suggests that at least some of the CDP-alcohol phosphatidyltransferases that make major phospholipids – those that install serine and glycerol headgroups – in archaeons and in bacteria may not catalytically discriminate to any great degree between these sterically very different diradylglycerols [24-26].

Early studies of PtdIns biosynthesis in eukaryotes led many years ago to universal acceptance of the idea that Ins is incorporated into PtdIns when PtdIns synthase, an endoplasmic reticulum-localized member of a substantial family of CDP-alcohol phosphatidyltransferases, transfers a phosphatidyl residue from CDP-diacylglycerol to the free 1-hydroxyl of Ins [27, 28]. It was therefore a surprise when Morii et al. [29] reported that the biosynthesis of archaetidylinositol (ArcIns) in Methanothermobacter thermoautotrophicus follows a different pathway (see Reactions 1 and 2, and Fig. 2). In this archaeon, a CDP-alcohol archaetidyltransferase links the free 1-hydroxyl of Ins3P  the only known source of which is MIPS-catalysed synthesis de novo – to Arc. This is similar in principle to phosphatidyl transfer during ‘classical’ PtdIns biosynthesis, and is catalysed by a similar enzyme, but there is a key difference: it is to the 1-hydroxyl group of Ins3P rather than the 1-hydroxyl of free Ins that the Arc moiety is transferred:

display math(1)
display math(2)

Prior to this, all reports of Ins lipids that carried one or more monoester phosphates on the Ins ring had come from eukaryotes, in which such PPIns are ubiquitous. Remarkably, however, the headgroup of ArcIns3P, the novel lipid that serves as an intermediate in ArcIns synthesis, is identical to that of eukaryal PtdIns3P, but it is synthesized by a quite different route (Fig. 2). In eukaryotes, the PtdIns3P that is made by type III PPin 3-kinases (PI3KIIIs) is essential both for membrane-sorting events in endosomal trafficking and autophagy and for effective cytokinesis, and its interaction with parts of the membrane-sorting ESCRT-III complex is important in some of these processes. Recent evidence indicates that cell division in some archaeons involves proteins homologous to parts of the eukaryotic ESCRT-III apparatus [30, 31], so it is tempting to speculate that ArcIns3P might prove to have some role in their ‘primitive’ versions of such processes.

A second surprise came when the putative PtdIns synthases (encoded by the pgsA gene) of several Mycobacterium species were introduced into Escherichia coli, a bacterium that neither contains nor can make Ins. PtdIns biosynthesis by the ‘classical’ pathway was expected, and had been reported [32]. However, re-examination of this reaction suggested that mycobacterial ‘PtdIns synthases’ transfer a phosphatidyl residue to the 1-hydroxyl of free Ins very inefficiently [33]. However, they catalyse rapid transfer of a phosphatidyl residue from CDP-diacylglycerol (shown as Ptd-CMP in Reaction 3 to emphasize its role as a phosphatidyl donor) to the 1-hydroxyl of Ins3P, making PtdIns3P [33] – thus following a route similar to that used by archaeons to make ArcIns3P. Some inositol monophosphate analogues, including one in which a nonhydrolysable phosphonomethyl group replaces the 3-phosphate of Ins3P, serve both as quite good inhibitors of this reaction and as alternative substrates [34].

display math(3)

At much the same time, another research group independently detected PtdIns3P as a lipid that transiently accumulates in response to osmotic stress in two actinobacterial species (Mycobacterium smegmatis and Corynebacterium glutamicum) [35]. They suggested that the transient kinetics of its accumulation might indicate that it is a ‘signalling’ lipid in some unspecified manner rather than a product of this novel biosynthetic pathway [35, 36], but the characteristics of the process look largely compatible with the latter interpretation.

There are a few organisms that must make their own Ins3P de novo if they are to be fully functional: a generous supply of exogenous Ins does not suffice [37]. Why this should be so is not known. One such is Mycobacterium tuberculosis: deletion of its MIPS gene suppresses the pathogen's virulence in a mouse model [38, 39]. Another is the eukaryotic pathogen Trypanosoma brucei, which loses viability even in inositol-replete cultures if either of the genes encoding MIPS and PtdIns synthase is disrupted [40, 41]. It is not clear in either organism how the intact cells functionally discriminate between imported free Ins and Ins made endogenously by MIPS.

It is tempting to speculate whether this ‘new’ pathway of PPIn synthesis that directly incorporates Ins3P – which might, of course, be the oldest such pathway! – occurs more widely than is yet appreciated, and whether it might provide an explanation for at least some such observations. Might other organisms, perhaps even eukaryotes, employ reactions akin to the ‘novel’ biosyntheses of PtdIns3P/ArcIns3P and PtdIns/ArcIns that Morii et al. [29, 33, 34] have revealed in archaeons and mycobacteria? I am not aware, for example, that anyone has ever experimentally determined whether any eukaryotic ‘PtdIns synthase’ might transfer phosphatidyl groups efficiently to the 1-hydroxyl group of Ins3P (making PtdIns3P).

PtdIns(3,5)P2, a late arrival on the PPIn scene

In 1989, Auger et al. [42] found an unexpected product in their PPIn kinase assays in vitro, and suggested that this might be PtdIns(3,5)P2, but only in 1996 was this ‘novel’ PPIn formally identified as a normal cell constituent in mammalian fibroblasts [43, 44] and in yeast and plants [44]. PtdIns(3,5)P2 was immediately of biological interest for two reasons: diverse eukaryotic cells make it in minute amounts; and its concentration in yeast increases dramatically during hyperosmotic stress [44]. The mechanisms underlying this response and a similar hyperosmotically activated response that was later discovered in adipocyte-differentiated 3T3-L1 cells [45] remain undetermined. The role of PtdIns(3,5)P2 in plant responses to salt and other osmotic stresses also remains ill-understood, but may involve direct interactions of PtdIns(3,5)P2 and/or PtdIns3P with the immunophilin molecule ROF1 [46].

It has since been established that most eukaryotic cells have the machinery needed to make PtdIns(3,5)P2, and that this lipid plays essential roles in the dynamic relationships between endosomes and lysosomes – and also between these organelles and other cell compartments [8, 10, 47-50]. Research is also focusing increasingly on the involvement of PtdIns(3,5)P2 and its effectors in initiating and regulating autophagy, which has recently become recognized as a pivotal process in maintaining organismal balance in the face of intermittent feeding (see [51-53] and the Supplementary Material to [48]).

Recent genetic studies in a variety of organisms, including mammals, have further highlighted the essential nature of PtdIns(3,5)P2. Mice and humans that lack elements of the synthetic machinery to make PtdIns(3,5)P2 sometimes die early in embryonic development in utero; or, if they remain able to make limited amounts of PtdIns(3,5)P2, they may be born alive, but suffer from severely defective nervous systems, skeletal deformities and a rapidly increasing number of other developmental faults that condemn them to brief and handicapped lives [54-57]. A very recent review gives an excellent summary of the systemic effects of genetically perturbing PtdIns(3,5)P2 synthesis and degradation in mammals [58].

So, how is PtdIns(3,5)P2 made, how is its synthesis regulated, and what does it do in cells? Answers to the first two questions are becoming much clearer, and the list of its proven and likely functions is rapidly growing.

The synthesis and turnover of PtdIns(3,5)P2

Early on, it was realized that PtdIns(3,5)P2 is made by large and complex PtdIns3P 5-kinases (PI3P5Ks). The prototype of these is encoded by the FAB1 gene of the yeast Saccharomyces cerevisiae, and its loss causes remarkable enlargement of the yeast vacuole, pointing to some defect(s) in membrane trafficking. The aminoacid sequence of Fab1 was determined before PtdIns(3,5)P2 came on the biological scene, and suggested that it might encode a PtdIns phosphate kinase [59]. By combining enzyme assays in vitro with genetic complementation and mutation studies, it was established that Fab1 (or its homologue in Schizosaccharomyces pombe) is the PI3P5K responsible for all PtdIns(3,5)P2 synthesis in yeast [42, 60-62]. The Fab1-like kinase in mammals is known as PIKfyve [62-65], and for other organisms the choice of name has tended to be pretty random. These PI3P5Ks are also known as type III PIPkins; the former abbreviation summarizes their activity more explicitly, and will be used here.

It seems clear that similar PI3P5Ks make PtdIns(3,5)P2 in all eukaryotes, but there remains some uncertainty over whether these PI3P5Ks have other biologically relevant kinase activities, either as protein kinases [64] or as kinases that directly 5-phosphorylate PtdIns to form PtdIns 5-phosphate (PtdIns5P), an even less abundant PPIn that is attracting burgeoning interest (reviewed in [66]). Three recent papers concurred that much of the PtdIns5P found in mammalian cells is made either directly by PI3P5K or as an indirect result of its activity, but came to opposite conclusions about how this happens. One concluded that PI3P5K directly phosphorylates PtdIns to form PtdIns5P in intact cells [67], whereas the other two presented evidence that PI3P5K-dependent generation of PtdIns5P occurs mainly as a result of the 3-dephosphorylation of PI3P5K-synthesised PtdIns(3,5)P2 by the phosphatase MTMR3 myotubularin-related protein 3 [68, 69]. However, a fourth study recently concluded that PI3P5K plays no part, at least during H2O2-stimulated PtdIns5P production [70]. Whether or not all PtdIns5P is made by a single route therefore remains less than clear.

Most organisms have a single PI3P5K gene that encodes a large polypeptide, typically of 1500–2500 residues, of which rather more than 600 residues are in four well-conserved domains. All PI3P5Ks share a strikingly similar domain organization (for details, see Box 2 of [45]). The catalytic PtdIns phosphate kinase domain is near the C-terminus; there is usually a PtdIns3P-binding FYVE domain near the N-terminus; a CCT-chaperone-like domain is central; and somewhat C-terminal of the CCT domain is a well-conserved ‘cysteine-rich’ [71] or ‘PIPkinIII-unique’ [48] domain that is unlike any domain present in other proteins. A consensus aminoacid sequence of the latter domain, of unknown function, is an effective search term with which to identify PI3P5K genes in databases without pulling out other lipid kinases.

Similar enlargement of the yeast vacuole occurs when another yeast gene, VAC14, is ablated, and the phenotypic resemblance between FAB1 and VAC14 mutants led to the recognition that Vac14 plays some essential role in controlling PtdIns(3,5)P2 formation by PI3P5Ks: cells lacking Vac14 make little PtdIns(3,5)P2, and PtdIns(3,5)P2 synthesis is not stimulated normally following hyperosmotic challenge [72, 73]. A third gene, FIG4, encodes yeast's PtdIns(3,5)P2 5-phosphatase, and its deletion causes surprisingly similar effects [74]. These observations led to the conclusion that Vac14 (known as ArPIKfyve in mammalian cells) and Fig4 (sometimes called Sac3 in mammals) are both essential for physiological regulation of PI3P5K-catalysed synthesis and turnover of PtdIns(3,5)P2. In some fungi, the transmembrane vacuolar protein Vac7 is also necessary for this process [74]. However, Vac7 is only present in fungi and its role is still unexplained, so it will not be discussed further here.

Vac14/ArPIKfyve is a widely conserved eukaryotic protein of 700–900 residues, and is constructed almost entirely from ~20 repetitive HEAT protein domains [75]. HEAT repeats are commonly found in proteins that form ‘scaffolds’ upon which other proteins organize, and each HEAT domain is essentially an ~47-residue solenoid structure composed of two linked and intertwined antiparallel α-helices. In isolation, the Vac14 protein forms a very elongated homodimer that is held together by mutual interactions between the C-terminal portions of two Vac14 subunits. Monomeric Vac14 is nonfunctional, and interference with Vac14 dimerization – by mutation or by expression of inhibitory constructs – inhibits its function in the PI3P5K complex [76, 77].

Work from several laboratories [75-78] has built up a quite detailed picture of a novel type of multiprotein complex that is responsible for regulation of PtdIns(3,5)P2 synthesis and degradation. However, we still lack detailed structural information on the spatial relationships between the various domains in the large PI3P5K molecule and on the orientation of the contacts between the endolysosomal membrane, PI3P5K, the Vac14 dimer, and Fig4.

For correct regulation of PI3P5K activity, Fig 4 must first interact with the C-terminal half of the Vac14 dimer. To complete the core complex, PI3P5K makes two key contacts. It binds through its FYVE domain to PtdIns3P in the endosomal membrane, so mooring the assembled complex to membranes containing its substrate, and through its CCT domain to the N-terminal half of Vac14 in the preassembled Vac14–Fig4 complex: see [75] for a schematic depiction of this core complex. The PtdIns(3,5)P2 effector protein Atg18 (see below) is a fourth protein that may associate less tightly with the (Vac14)2–Fig4–PI3P5K complex. It inhibits the PI3P5K activity of the complex [75]. Catalytic capacities for the synthesis and 5-dephosphorylation of PtdIns(3,5)P2 are thus combined within a remarkable multiprotein assembly, with these activities also being regulated by a PtdIns(3,5)P2-dependent effector that regulates autophagy and controls PtdIns(3,5)P2-dependent retrograde vesicular trafficking from the endolysosomal compartment [79].

The functions of PtdIns(3,5)P2

Known functions for PtdIns(3,5)P2 have become remarkably numerous in recent years, and the outer orbit of Fig. 3 summarizes a number of these. The evidence that has implicated this very low-abundance PPIn in a particular function has usually involved one or more observations of the following types. The function is impaired when one (or more) of the three core components of the PtdIns(3,5)P2 turnover complex – PI3P5K, Vac14/ArPIKfyve, or Fig 4/Sac3 (or, additionally, Vac7 in yeast) – is compromised. The affected protein(s) may be mutationally disabled, expression may be downregulated by RNA interference treatment, or activity may inhibited by expression of an interfering protein construct. The function may also be abolished or reduced when cells are treated with the small-molecule PI3P5K inhibitor YM201636 [80]. As with any experiments using combinations of pharmacological inhibitors, mutant cells, and cells overexpressing active or interfering constructs, conclusions tend only to be secure when they are supported by at least two different types of evidence. This is true of most, although not all, of the examples given below.

Figure 3.

Biological functions of PtdIns(3,5)P2, and PtdIns(3,5)P2-responsive effectors that may mediate some of these cell functions. This is an expanded version of Fig. 2 from [47]. The outer orbit of square boxes summarizes cell functions in which PtdIns(3,5)P2 has been implicated, and the inner orbit identifies PtdIns(3,5)P2 effectors that may be involved in mediating these functions. The endosomal–lysosomal system is the main site of PtdIns(3,5)P2 formation and function (see Fig. 1 and the text), and the green boxes depict processes that occur within that domain of the cell. Many or all of the other processes appear to be initiated by effectors that interact with PtdIns(3,5)P2 in the endolysosomal system, but in a manner that leads to consequences in the cytosol (brown), at the plasma membrane (blue), or in the nucleus (purple). Pre-2011 references to many of these processes are listed in [50], and more recent references are given in the text and/or the Figure. Evidence from mammalian genetic diseases and experimental mutants is summarized in [58]. CFTR, cystic fibrosis transmembrane conductance regulator; EEAT4, excitatory amino acid transporter-4; EIAV, equine infectious anaemia virus; GLUT4, glucose transporter type 4; MVB, multivesicular body; PM, plasma membrane; TPC, two-pore Na+ channel; TRPML, transient receptor potential mucolipin.

In some cases, putative protein effectors that are regulated by binding to PtdIns(3,5)P2 have been identified as participating in a particular function. When these are known (or strongly suspected), they are identified in the inner orbit of Fig. 3. Many of the relevant observations are admirably summarized in [50], where relevant references can be found, and herein I mostly focus on information published since that 2011 review.

Early on, it was realized that the formation and metabolism of PtdIns(3,5)P2 revolves around late endosomes and lysosomes (and the vacuole in yeast), and that the functions of PtdIns(3,5)P2 are focused primarily on supporting cell functions in which these organelles play key roles. The PtdIns(3,5)P2-dependent functions that occur primarily within this cell domain are colour-coded green in the outer orbit of Fig. 3. Many, or perhaps all, of the other PtdIns(3,5)P2-dependent cell functions seem also to be initiated by effectors that interact with PtdIns(3,5)P2 somewhere in the endolysosomal system, but through mechanisms that then have consequences elsewhere in the cell. These are also colour-coded in brown (targeted into the cytosol), blue (to the plasma membrane), or purple (into the nucleus).

β-Propellers that bind phosphoinositides (PROPPINs) in trafficking and autophagy

Atg18 (also known as Svp1) was the first clearly identified PtdIns(3,5)P2 effector, and is the prototype of the PROPPIN protein family, the mammalian homologues of which are also known as WD-40 proteins interacting with PPIn (WIPIs). High-affinity binding of PtdIns(3,5)P2 to the β-propellers of PROPPINs involves a highly conserved ψRRG motif between two of the predicted β-propeller ‘blades’, and this Atg18–PtdIns(3,5)P2 interaction is essential to support retrograde membrane traffic from the vacuole to the Golgi. Abolition of this traffic by lack of either PtdIns(3,5)P2 or Atg18 provides at least a partial explanation for vacuolar enlargement in yeast mutants [79]. Atg18 is also essential for correct regulation of PI3P5K activity (see above) and for autophagy.

Atg18 is functionally the best understood of the three yeast PROPPINs, so it is slightly unfortunate that the three high resolution PROPPIN structures so far obtained have all defined PtdIns3P binding to homologues of S. cerevisiae Hsv2 [81–83]. Hsv2 is localised at endosomes in a PtdIns3P-dependent manner, suggesting some function there [84] but yeast cells from which it is deleted remain fairly normal, so its function remains to be determined. These and other studies make it clear that the PROPPIN β-propeller accommodates a pair of PPIn-binding sites, and that elsewhere on the β-propeller there are sometimes additional binding sites through which the PROPPINs interact with other functional partners, e.g. with Atg2, one of the proteins with which it collaborates in the initiation of autophagy [82, 85]. Intriguingly, WIPI-1 also plays an essential role in the maturation of melanosomes, an endosome-initiated process that is distinct from autophagy [86, 87].

Endolysosomal ionic regulation

One of the earliest manifestations of PtdIns(3,5)P2 deficiency to be recognized was failure to acidify the yeast vacuole [61], a molecular explanation for which is likely to lie in failed regulation of the proton-translocating vacuolar V-ATPase. Similar events may be involved in PtdIns(3,5)P2-dependent control of the acidification of the stomatal vacuole in abscisic acid-controlled water conservation in plant leaves [88].

There is also a growing awareness that PtdIns(3,5)P2 in the vacuolar membrane is key to the correct regulation of a variety of selective ion channels therein, and this seems likely to be a vein of work that will be richly worked in the next few years. For example, the transient receptor potential mucolipin channels that regulate Ca2+ release from vacuoles are a second set of vacuolar ion flux regulators that are under control by PtdIns(3,5)P2 in yeast [89, 90]. The unrelated two-pore Na+ channels constitute a third such group [90, 91].

Protein traffic to and from the plasma membrane

Interference with insulin regulation of the exocytosis of glucose transporter type 4 to the plasma membrane was one of the first responses to perturbation of PtdIns(3,5)P2 synthesis to be described [92, 93], and recent studies have shown that targeted ablation of PI3P5K from skeletal muscle in vivo has much more profound consequences – including adiposity and profound insulin resistance – than simply interference with the regulation of muscle glucose uptake [94].

Other observations have since implicated PtdIns(3,5)P2 in numerous other processes by which proteins are trafficked to or from the plasma membrane (see also [50]).

The first is a failure in the cellular organization of polarized epithelia. In a mouse model of embryonic PI3P5K deletion, embryos died at ~8.5 day, apparently as a result of undernutrition. In such embryos, the visceral endoderm normally harvests maternal nutrients by endocytosis to support the embryo [95], but this epithelium did not become correctly organized in the PI3P5K knockout [96]. When the same defect was selectively targeted to the intestinal epithelium, that tissue failed to develop correctly, and so did not absorb nutrients efficiently: the baby animals died within a few weeks, apparently of malnutrition. The underlying problem appeared to be failure of the components of the apical membrane to be correctly targetted to the brush border pole of the cells, so this usually exquisite structure became so morphologically disorganized that digestion and nutrient absorption were ineffective [96]. There was also profound inflammatory infiltration of the aberrant intestinal mucosa, and the authors suggested that the condition of these animals – triggered solely by a lack of intestinal mucosal PI3P5K – resembles human Crohn's disease more closely than any previous animal model [96].

Another defect of membrane delivery to the plasma membrane was seen in tip growth in the moss Physcomitrella when either the entire actin regulator formin-II was deleted or its PtdIns(3,5)P2-binding PTEN domain was disabled [97]. The need for PtdIns(3,5)P2 for the proper assembly and release of some viruses offers a third such situation [98]. The stimulatory influence of PtdIns(3,5)P2-derived PtdIns5P in cell motility represents yet another effect on a system that depends on actin-based motility [69].

In a number of these recent studies, it has not been clear exactly where the functional imbalance lies – in failure of protein delivery to the plasma membrane, in its misdirection to the wrong place, or in its retrieval back into the cell. For example, PtdIns(3,5)P2 appears be implicated in plasma membrane trafficking of both the GluA1 and GluA2 subunits of the ionotropic 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid (AMPA) receptor in neurones – but perhaps in different ways. In hippocampal slices, activation of PI3P5K activity as a result of its phosphorylation by the protein kinase SGK3 enhanced the sensitivity of the hippocampal neurones to stimulation. At least in part, this seemed to be because PtdIns(3,5)P2 is needed for trafficking of the GluA1 subunit to the plasma membrane [99]. However, and in apparent contrast, neurones lacking Vac14 – and so depleted of PtdIns(3,5)P2 – generated miniature endplate potentials that were much larger than normal. This seemed to be because these cells displayed a larger than normal number of excitatory AMPA receptors on their surfaces; it appeared that internalization of the GluA2 AMPA receptor subunit was slower in the cells that lacked Vac14 [100].

Control of central metabolic pathways

The pivotal protein kinase mammalian target of rapamycin complex 1 (mTORC1) is one of the key cell regulators that is controlled by insulin, by amino acid supply and by other nutritional triggers in 3T3-L1 adipocytes. These stimuli also increase PtdIns(3,5)P2 levels in these cells, raising the possibility that the two sets of events might be linked [101]. To test this idea, the consequences of RNA interference knockdown of the various components of the multiprotein complex that makes and degrades PtdIns(3,5)P2 (see above) were examined. This prevented insulin from stimulating ribosomal protein S6 kinase, a response in which activation of mTORC1 is an essential intermediate, and also interfered with the translocation of activated mTORC1 to the plasma membrane. The interaction between PtdIns(3,5)P2 and mTORC1 seems to be mediated through its binding to the β-propeller portion of the mTORC1 regulatory subunit Raptor, and this seems, unexpectedly, to be a PtdIns(3,5)P2-dependent event that is located at the plasma membrane in 3T3-L1 adipocytes [101].

Finally, it seems that endosomal PtdIns(3,5)P2 can, through direct interaction with the transcriptional regulator Tup1, also instigate cytoplasmic events that initiate a series of steps that culminate in regulation of the expression of many nuclear genes involved in core cell processes. For the Cyc8–Tup1 corepressor complex to be converted into the three-protein coactivator complex Cti6–Cyc8–Tup1, both Tup1 and Cti6 have to bind to endolysosomal PtdIns(3,5)P2. This ‘converted’ complex, after translocation into the nucleus, then becomes an activator at the GAL1 promoter [102]. The more general metabolic importance of this PtdIns(3,5)P2-dependent regulation of the gene control status of the Cyc8–Tup1 corepressor complex was further emphasized when it was shown that the PtdIns(3,5)P2-dependent Cti6–Cyc8–Tup1 coactivator complex similarly controls the expression of several enzymes – including fructose 1,6-bisphosphatase and isocitrate lyase – that are central to metabolic switching from glycolysis to gluconeogenesis [103].


Many years ago, we thought that there were only two PPIns with monoesterified phosphates attached to their Ins1P headgroup, and that these were special constituents of the nervous system. However, we now know that PPIns are far more diverse and are present in all eukaryotes, that each of these quite similar PPIns has distinctive cell functions, and that each is functionally versatile. When we find a ‘new’ PPIn, as happened with PtdIns(3,5)P2 in the 1990s, it is a curiosity for a few years, and then it starts to collect functions at an ever-increasing pace. Now, one of these PPIns has turned up in archaeons and another in a few bacteria. None of us had the foresight to see much of this coming, and no doubt the same will be true of future revelations.

Note in proof

A paper published since the acceptance of this article [108] has made two substantial advances. First, it identified formation of PtdIns(3,5)P2 as an essential step in the chain of events that links the activation of Toll-like receptors (TLRs) by microbial products to the development of the TH1 and TH17 subsets of T lymphocytes that are involved in the aetiology of numerous autoimmune diseases. Secondly, it provides the PtdIns(3,5)P2 research community with apilimod (also known as STA-5326), an inhibitor of PtdIns(3,5)P2 synthesis that is structurally related to YM201636 [80] and MF-4 [109], the inhibitors that were previously available, but is much more potent and selective.

Apilimod is a recently developed small molecule drug lead that, by a previously unknown mechanism, inhibits both production by immune cells of the related cytokines IL-12 and IL-23 and the resulting development of TH1 and TH17 cells [110,111]. This inhibition lies downstream of the activation of TLRs, for example by bacterial lipopolysaccharide. By using a combination of an affinity ligand and proteomic analysis to identify proteins with which apilimod associated reversibly, Cai et al. [108] homed in on only three molecules. Remarkably, these were PIKfyve/Fab1, ArPIKfyve/Vac14, and Sac3/Fig4, the three core components of the conserved complex that is responsible for PtdIns(3,5)P2 synthesis. Moreover, nanomolar concentrations of apilimod inhibited PtdIns(3,5)P2 synthesis both in vitro and in intact HeLa cells. Other experiments in the same study reinforced this evidence that activation of PtdIns(3,5)P2 formation is necessary for the apilimod-sensitive transcriptional activation of IL-12 and IL-23 production downstream of TLR activation.