PHOSPHATIDYLINOSITOL 4-PHOSPHATE; A MINOR LIPID WITH MULTIPLE PERSONALITIES

Abstract

Phosphorylated products of phosphatidylinositol (PI), named Diphosphoinositide (DPI) and triphosphoinositide (TPI) were identified long time ago and found to exhibit high turnover rates based on their rapid ³²P-phosphate labeling. The PI kinase activities that were responsible for their production were subsequently identified and found to be associated with different organelle membranes, including the plasma membrane. These activities were then linked with a certain group of cell surface receptors that activated phospholipase C enzymes to hydrolyze PI and used calcium or cGMP as a second messenger. This visionary concept was introduced in the seminal BBA review written by Robert Michell, exactly 50 years ago. The enzymology and functional diversity of PI 4-phosphate (PI4P) (the term that has replaced DPI) has since underwent an expansion that could not have been foreseen. In this review I will attempt to revisit this expansion with some historical reflections celebrating the 50^th anniversary of the Michell review.

Phospholipids show fast labeling with ³²P-phosphate; the early years

It was the pioneering studies of Jordi Folch in the 1940’s that identified diphosphoinositide (called DPI) in bovine brain as an ethanol-insoluble phospholipid fraction that contained phosphates and inositol in a molar ratio of 2:1 [1]. Folch showed that DPI was enriched in myelin associated with the protein he called “neurokeratin” [2]. During studies to determine the structure of DPI it was discovered that the brain “diphosphoinositide” fraction also contained triphosphoinositides [3]. This was also shown in parallel studies that also demonstrated that the “diphosphoinositide” fraction contained both a monophosphoinositide and a triphosphoinositide (TPI) in addition to DPI [4]. TPI was identified as diacyl-phosphoryl-inositol-diphosphate [5] and it was independently demonstrated that phosphatidylinositol contained an inositol ring linked to a diacylglycerol (DAG) backbone through a phosphodiester bond linked to the 1-position of the inositol ring [6, 7]. Detailed analysis of the inositol phosphates liberated from these different phosphoinositides assigned the positions of the phosphate groups on the inositol ring [8]. All these studies together have established that phosphoinositides contained an inositol ring linked via a phosphodiester bond to an sn-1,2 DAG backbone and are phosphorylated at the 4-position of the inositol ring in DPI, and on the 4- and 5-positions in TPI (Fig. 1). Beyond their identification, the interest in these lipids was raised by early studies, that showed rapid labeling of some phospholipids, phosphoinositides in particular, with ³²P-phosphate injected into guinea pigs [9, 10].

Figure 1.

(A)The chemical reaction catalyzed by phosphatidylinositol 4-kinases.

What a difference one phosphate makes! It is hard to comprehend in how many ways Mother Nature has taken advantage of this very simple reversible covalent modification. PI4P controls multiple pathways through direct interaction with proteins or as a precursor of PI(4,5)P₂ or other phosphoinositides. A selected and best characterized examples are listed in panel A.

(B)Schematic structural features of the mammalian PI4Ks

Numbers on the scale correspond to amino acids numbers.

DPI and TPI were phosphorylation products of PI, by kinase activities that were identified in various membrane fractions and quite importantly, also in red blood cell membranes [11–13]. It is interesting to note that, even at this early stage, PI kinase activities were found in subcellular membrane fractions additionally to the plasma membrane (PM) [13, 14], namely in the microsomes where the activity was massively increased in the presence of the detergent, cutscum, suggesting that more than one PI kinase enzyme was present in rat liver [15]. Nevertheless, the prominent presence of PI kinase activity in the PM and in red blood cell membranes have focused the attention to functions associated with the PM.

Secretagogues stimulate the turnover of phospholipids

In parallel studies started in the early 50’s by the Hokins, it was found that stimulation of pancreatic and brain slices with acetylcholine increased the ³²P-labeling of phospholipids that later have been identified as PI and phosphatidic acid (PA)(see original citations in a recollection by the Hokins [16]). The Hokins also found that acetylcholine stimulation, in fact, led to the hydrolysis of PI to inositol-1-phosphate and DAG, and that the phosphorylation of DAG by a kinase was behind the increased labeling of PA. Moreover, they showed that upon termination of the stimulus, the ³²P-label disappeared from PA and showed up in PI [17]. It was concluded that the phospholipase activities, described earlier in the Dawson [18, 19] and Hawthorne [20] laboratories, were responsible for the receptor-triggered hydrolysis of phosphoinositides, primarily PI. There, the concept of the “PI-cycle” was born. Many important observations that will not be detailed here and are covered in other reviews in this collection, led Michell to propose that the “PI-cycle” and the primary activation of phospholipase C (PLC) enzymes was part of an early receptor-triggered event that was characteristic of receptors that used Ca²⁺ or cGMP as second messengers [21]. Michell suggested that the primary lipid hydrolyzed by PLC was PI, leaving for a short time PI4P and PI(4,5)P₂ (“DPI” and “TPI”) in the PM in limbo (see [22]. However, even at that time, several reports hinted that PLCs also hydrolyzed the phosphorylated derivatives of PI [23–25]. Final consolidation of the receptor and second messenger hypothesis happened in the early 80’s when it was discovered that activated receptors, primary enhanced the activity of PLC enzymes leading to the hydrolysis of PI(4,5)P₂ [26, 27] [28, 29] producing inositol 1,4,5-trisphosphate (InsP₃) and DAG that both serve as intracellular 2^nd messengers [30]. These exciting developments draw much attention to phosphoinositides prompting a whole array of intense studies on PLC enzymes and PIP kinases that produced PI(4,5)P₂. PI kinases and PI4P were viewed as metabolic intermediates only subserving the receptor mediated inositol lipid cascade and hence, have received less attention.

PI kinases draw some new attention

As it often happens, renewed interest in PI kinases came from a completely different line of research. Several laboratories have reported that gene products that are able to transform cells, profoundly increased the activity of lipid kinases, including PI kinases and DAG kinases and that these transforming factors were physically associated with PI kinase activities [31–34]. Work to characterize and isolate these kinases identified two PI kinases, named type-I and type-II enzymes that differed in their K_m for ATP and sensitivity to detergents or inhibitory effects of adenosine [35]. This study also showed that the PI kinase associated with the middle-T/pp60c-src complex and the PDGF receptor was the type-I activity. Another PI-kinase with a significantly larger size was isolated from bovine brain with properties different from the type-II activities and it was named type-III PI-kinase [36]. It was subsequently discovered that the Type-I PI kinase activity phosphorylated PI at the 3-position of the inositol ring [37] and so the PI 3-kinase (PI3K) field was born (see [38] for a historical perspective). That left the PI 4-kinases (PI4Ks) with the designation of type-II and type-III PI4Ks. Since transforming oncogenic proteins were associated with the PI3Ks whose activity was also controlled by receptor tyrosine kinases, including the insulin receptor, an enormous effort was directed to their characterization, purification and molecular cloning. These outstanding works are not the subject of this review and can be found in several excellent collections [38–40]. They are also covered in separate Chapters in this collection. Our focus is the type-II and type-III PI kinase enzymes that are the genuine PI4Ks. Since these enzymes were regarded less attractive, still believed to “only” make PI4P as a substrate for phosphatidylinositol 4-phosphate 5-kinases (PIP5Ks), few groups pursued their characterization and molecular identification.

PI 4-kinases are purified and cloned

While some early cell fractionation studies showed the presence of PI kinase (presumed PI 4-kinase) activities in subcellular fractions, such as the Golgi [41, 42], lysosomes [42, 43] and secretory granules [44–46], the primary motivation to purify PI4K enzyme(s) was to identify the activity that was responsible for the production of PI4P in the PM. Early efforts focused on the type-II PI4K activity, which was found abundant in many tissues as a tightly membrane-bound, 45–55 kDa size protein [47–49] and which was also present in the red blood cell membranes [50, 51]. Logically then, the type-II PI4K was presumed to be responsible for PM PI4P generation. As mentioned above, the type-III PI4K activity was described in cholate extracts of bovine brain, which was distinguished from the type-II enzyme by its larger size (~230 kDa), and its lower affinity to ATP and less sensitivity to inhibition by adenosine [36] but purification of this activity was not attempted at the time.

Our group was led to PI4Ks in the early 90’s through experiments in which we tested several Ca²⁺-Calmodulin (CaM) inhibitors on the production of inositol 1,3,4,5-tetrakisphosphate (InsP₄) in angiotensin II (AngII)-stimulated adrenocortical cells [52]. InsP₄ was produced from inositol 1,4,5-trisphosphate by a then recently discovered InsP₃-kinase [53], which was shown to be regulated by Ca²⁺-CaM [54]. Our hope was then that by targeting CaM we could inhibit this pathway and gain insights into the importance of InsP₄ as a signaling entity, a question that was hotly investigated at the time [55–57]. While none of the inhibitors have proven to be useful for this original purpose, one inhibitor, wortmannin [code named MS-54 in our inhibitor collection, also targeting myosin light chain kinase (MLCK)], had a strong Inhibitory effect on the sustained but not initial phase of InsP₃ formation [52]. This curious effect was eventually traced to wortmannin inhibiting PI4P levels and causing a rapid run-down of PI(4,5)P₂ levels upon stimulation, explaining the transient nature of the InsP₃ rise and the cytoplasmic Ca²⁺ signal [58]. Wortmannin was originally described as a PLC and PLD inhibitor [59, 60], which also inhibited MLCK [61]. As we just concluded that wortmannin inhibited the PI4K activity relevant to the receptor-regulated PI4P pool in the PM, two publications reported that wortmannin was a potent inhibitor of PI3Ks [62, 63]. Importantly, these studies also showed that wortmannin did not inhibit PI4Ks. This apparent discrepancy was resolved when we isolated a soluble PI4K activity from adrenal cortex that was inhibited by wortmannin (although less potently than PI3Ks) and we also confirmed that the membrane-bound (presumed type-II) PI4K activity was, indeed, completely resistant to the inhibitor. An important corollary of these findings was that contrary to the then prevailing views, the tightly membrane bound type-II PI4K was not the main source feeding the agonist-sensitive PM pools of PI4P and PI(4,5)P₂ [58]. This prompted us to characterize the soluble PI4K activity from bovine adrenal and brain and showed that the properties of this activity matched those of the brain type-III PI4K [64]. Upon purification, two activities could be clearly separated, one with an apparent molecular size of 110 kDa and another with a size of ~ 210 kDa, both of which we were eventually able to clone [65]. Notably, in parallel efforts the larger enzyme was also purified from bovine brain [66] and the cloned human enzyme successfully expressed [67]. Homology cloning also identified the larger enzyme, named 230 kDa [68] and the smaller enzyme, named 92 kDa enzyme [69] both from rat. The smaller human enzyme was also identified with homology cloning [70, 71]. An earlier study isolated a cDNA that coded a shorter, 92 kDa fragment of the large, 230 kDa PI4K enzyme, and based on its catalytic properties and inhibition by an antibody believed to be specific for type-II PI4Ks, erroneously classified it as a type-II enzyme [72]. With all these studies, the type-III PI4Ks have been defined as the large, 230 kDa enzyme named PI4KIIIα (since then called PI4KA) and the smaller, 92 kDa PI4KIIIβ (or PI4KB), both inhibited by wortmannin.

Even though the early purification efforts focused on the tightly membrane-bound type-II PI4K activity (see above), it took several more years to clone the type-II PI4Ks [73, 74], which also exist in two forms, PI4KIIα (currently named PI4K2A) and PI4KIIβ (PI4K2B) [75]. Both forms were found to be associated primarily with the endo-lysosomal compartments and the TGN although a fraction of PI4KIIβ was also found associated with the PM [76–79].

Yeast studies on PI4Ks have propelled the PI4K field

The first PI4K, in fact, was cloned from yeast [80] based on purification of the protein from soluble fractions and named Pik1 [81]. The same gene was also isolated in a library screen that used an antibody against a nuclear pore protein [82]. Both studies found that PIK1 is an essential gene in yeast. Another PI4K, named Stt4 was cloned from yeast in a screen for staurosporine hypersensitivity [83]. Pik1 is a ~110 kDa protein, while Stt4 is a ~220 kDa protein and sequence comparison revealed their similarity within their C-termini to the C-terminal segment of the already cloned p110 catalytic subunit of mammalian PI3Ks [84, 85] and the yeast PI3K, Vps34 [86]. Genetic inactivation of STT4 was not lethal in every yeast strain but required osmotic stabilizers for viability and its function was linked to the yeast protein kinase C (PKC1) pathway and cell wall biogenesis [83, 87]. The respective roles of Pik1 and Stt4 have been thoroughly analyzed in yeast and it was concluded that Pik1 primarily affects post-Golgi vesicular transport [88–90], while Stt4 primarily produces PI4P in the plasma membrane and also affects actin dynamics and the structure of vacuoles [89]. Nuclear shuttling of Pik1 was also described [91] and both the nuclear and the Golgi functions were found critical for viability [92]. Yeast with temperature-sensitive alleles of both pik1 and stt4, showed ~90% reduction in their PI4P levels at the non-permissive temperature (they show ~50% decrease with inactivation of either one alone) indicating that these two enzymes were the main sources of PI4P in yeast. Membrane bound PI4K activities in yeast had been also described earlier and partially purified [93] and subsequently cloned and named Lsb6 [94]. Lsb6 has low but detectable PI4K activity and it was suggested that it functions at the endocytic pathway [94, 95].

In terms of regulation, Pik1 was found to genetically and physically interact with the yeast frequenin (Frq1), a small Ca²⁺ binding protein, which stimulated its lipid kinase activity [96]. The interaction region between the two proteins have been mapped [97, 98]. It was also found that Arf1 was necessary for the Golgi localization and function of Pik1 [90, 99, 100] and that Arf1 was able to recruit Frq1-Pik1 complexes to liposomes that mimicked Golgi lipid composition if they also contained DAG [101]. For Stt4, a synthetic rescue screen identified a protein, named Sfk1 (Suppressor of Four Kinase 1) that could rescue the stt4^ts defect when overexpressed. It was concluded that Sfk1 facilitated the PM-localization of the Stt4 PI4K [87]. Subsequently, it was shown that the localization of Stt4 at the plasma membrane required two proteins, Ypp1 and Efr3 [102]. These studies clearly suggested that Stt4 is a main source of PM PI4P and that it required additional proteins for PM localization. However, a puzzling finding that was not easy to explain at the time was that the PI4P generated by the Stt4 PI4K was also controlled by the Sac1 PI4P phosphatase, which is an ER-localized protein [103]. A similarly puzzling finding was that elimination of the Sac1 phosphatase led to massive accumulation of PI4P but not its phosphorylation product, PI(4,5)P₂ [103–105]. It took additional years to understand these curious observations, which will be described below.

Identified downstream targets of PI4P in yeast included the small clathrin adaptor Gga2p protein [106] and the Arf-GEF, Sec7p, the latter also binding to the enzyme Pik1 to control clathrin coat recruitment at the late Golgi [107]. The Rab GEF, Sec2p protein was also described as a target of PI4P thereby regulating the switch between Sec2p and Sec15p during maturation of secretory vesicles [108]. Another regulatory target by the PI4P produced by Pik1 was the Drs2 phosphatidylserine flippase protein [109, 110], which was originally described as critical for secretory vesicle formation in the trans-Golgi in yeast [111].

Studies on PI4Ks in higher organisms follow suit

The yeast studies clearly demonstrated that PI4Ks do more than just provide PI4P to serve as a precursor for PI(4,5)P₂ in the plasma membrane. Studies that described the unique intracellular localization of the mammalian enzymes followed: PI4KIIIβ in the Golgi [112, 113], and the type-II PI4Ks at endosomes (see citations above) (Fig. 2A). Initial work suggested that PI4KIIIα was located to the ER [112]. The puzzling observation that collectively came out from these studies, however, was that none of the wortmannin-sensitive type-III PI4Ks could be clearly located at the PM.

Figure 2.

(A)Cellular localization of mammalian PI4Ks. The plasma membrane contains PI4KIIIα and a small fraction of the PI4KIIα and PI4KIIβ enzymes. PI4KIIIβ is found primarily in the Golgi but a fraction is found in the nucleus and perhaps it is also present in small amounts in the endo-lysosomal compartment (not indicated in the cartoon). The major site of action of PI4KIIα and -β is the endo-lysosomal compartment, but these enzymes also play a role in the TGN.

(B)Plasma membrane localization of PI4KIIIα in a dimeric form of a tetrameric complex formed by the proteins as indicated.

(C)The PI4P-driven counter-transport pathway as exemplified by the OSBP-mediated cholesterol/PI4P exchange cycle between the ER and the Golgi compartment. The FFAT-motif of the OSBP protein interacts with the MSP domain of the ER-localized VAP proteins, whereas Golgi anchoring of OSBP is mediated by its PH domain binding to PI4P together with Arf1 (the latter not illustrated for simplicity). ORD represents the lipid cargo binding domain of the molecule that can either bind cholesterol (yellow cones) or PI4P (green-headed lipid). Similar principles operate in other contact sites formed by the ER with various other organelles transporting different phospholipids in exchange for PI4P.

Studies on PI4KIIIβ

The important role of PI4KIIIβ in the Golgi in mammalian cells, was first reported in studies, which showed that PI4KIIIβ activity was controlled by Arf1, and that the kinase was important to maintain the organization of the Golgi [113] (Fig. 2A). Studies seeking to find effector proteins that could recognize the lipid, PI4P, especially in the Golgi, identified GOLPH3, an oncogenic protein that connects Golgi to the actin cytoskeleton [114, 115]. Other studies revealed that the pleckstrin homology (PH) domains of several soluble lipid transport proteins recognized PI4P, in the context of Arf1 [116–120]. The true meaning of this connection between lipid transport proteins and PI4P was, again, only recognized later as will be discussed below. These, and many other studies collectively consolidated the important Golgi function of the PI4KIIIβ protein [see [110, 121] and a separate Chapter in this collection for more comprehensive coverage of this topic]. The regulation of PI4KIIIβ by frequenin (named Neuronal Calcium Sensor-1, NCS-1 in mammals) was also explored in mammalian cells. Several reports suggested that NCS-1 and PI4KIIIβ functions converge on Golgi and post Golgi trafficking, especially in cells with regulated secretion [122–126] but direct stimulation by NCS-1 of the kinase activity of PI4KIIIβ was shown only in a few studies [99, 127] and was even questionable [128]. This might be explained by the limited conservation within the primary sequences of the yeast Pik1 and PI4KIIIβ in the region where yeast Pik1 interacts with Frq1. Instead, the role of phosphorylation of PI4KIIIβ by the Golgi-localized protein kinase D (PKD) was identified as an important regulator of the activity and localization of the enzyme. PKD-mediated phsophorylation also keeps the enzyme in an active form through interaction with 14-3-3 proteins [129, 130]. Curiously though, PKD-mediated phosphorylation did not seem to affect the previously described [131] nuclear shuttling of the enzyme [130].

More recent studies found additional proteins that interact with PI4KIIIβ, namely c10orf76/ARMH3 and ACBD3, which may aid or regulate the interaction of the kinase with the Golgi compartment. ARMH3 was shown to recruit PI4KIIIβ to the Golgi and play a role in the activation of the antiviral STING pathway [132, 133] and to regulate ceramide transport between the ER and the distal Golgi compartment [134]. Association of ARMH3 with Arl5 and its effect on activating PI4KIIIβ in the Trans-Golgi has also been recently described [135] as were the structural features of the interaction of PI4KIIIβ with ACBD3 [136] and ARMH3 [137].

Studies on PI4KIIIα

Given the primary function of PI4KIIIβ in the Golgi, the question remained as to which PI4K enzyme was responsible for generating the PI4P pool of the PM. At this point, important methodological advances helped us better understand the relative roles of the various PI4Ks in generating the different PI4P pools associated with organelles. Determining the distribution of inositol lipids in intact cells was attempted using protein domains that can specifically recognize the various phosphorylated forms of PIP lipids [reviewed in [138, 139]]. Other efforts went into refining the methods of antibody-based immune-fluorescence detection of phosphoinositides [140, 141]. These novel tools complemented with EM techniques [142], together helped define the inositol lipid signatures of organelles [143]. Studies that used pharmacological, genetic and imaging approaches suggested that PI4KIIIα was the enzyme that produced most of the PI4P in the PM [144, 145], but how the enzyme localized to the PM remained enigmatic. Here, again, the yeast studies have proven to be instructive, as the mammalian homologues of the yeast Ypp1 (TTC7A/B) and Efr3 (EFR3A/B) proteins were identified as necessary molecular partners for localization of PI4KIIIα to the PM and for generation of PI4P [146] (Fig. 2B). Subsequently, a third protein, Fam126 (again, A and B forms in mammalian cells but no homologue in yeast) was shown to be part of the ternary PI4KIIIα complex [147], which was then structurally characterized in subsequent studies [148–150]. The yeast studies that identified Sfk1 as a regulator of the Stt4 kinase also prompted studies that sought to identify mammalian counterparts. One study showed that out of the 5 mammalian genes that encode homologues of Sfk1, TMEM150A was able to associate with the PI4KIIIα enzyme and compete for its association with the Efr3/TTC7/Fam126 complex [151]. While these studies have identified molecular components that helped recruit the enzyme to the PM, regulation of its enzymatic activity remains poorly understood. It has been reported that PKC can activate PI4P production in the PM [152, 153], and it was also shown that specific isoform of calcineurin, CNAβ1could affect the phosphorylation status of FAM126A and hence may control the activity of the kinase [154]. More studies are warranted to explore regulatory mechanisms for the PI4KIIIα enzyme complex.

It must be mentioned that the Arf1 guanine nucleotide exchange factor, Gbf1 was shown to be recruited to the Golgi by PI4P [155, 156] and the PI4K responsible was identified as PI4KIIIα [156]. The presence and function of PI4KIIIα in the Golgi has not been widely claimed (but also see [157]), and more studies will be required to clarify if PI4KIIIα indeed works in the Golgi and if so, what recruitment mechanism(s) keep the enzyme in that location.

Studies on PI4KIIα and PI4KIIβ

Clathrin adaptor proteins were described as PI4P effectors in experiments that studied the functions of type-II PI4Ks in mammalian cells. It was shown that the tetrameric adaptors, AP-1 [78] and AP-3 [158, 159] as well as the monomeric adaptor GGA proteins [160] were recruited to the TGN by PI4P specifically made by PI4KIIα. PI4KIIα has also been found important for Fz signaling, [161] and at several steps in the endocytic pathway determining cargo sorting [76, 162–164], late endosome/lysosome function [165] including lysosome damage-repair [166], autophagosome-lysosome fusion [167, 168] and lysosome reformation [169]. Regulation of PI4KIIα enzymatic activity has not been fully explored, but one study showed that palmitoylation of the protein and its activity was controlled by cholesterol [170]. Palmitoylation is a rapid and reversible lipid modification of proteins that controls their association with membranes and determine their intracellular distribution, as well as their activities [171]. There are numerous S-palmitoyl transferase enzymes, and it is not known which of these are involved in PI4KIIα palmitoylation. However, palmitoylation increases the affinity of proteins to cholesterol-rich membrane domains [172]. Accordingly, PI4KIIα was found to have its highest activity in a membrane compartment enriched in the valosin-containing AAA+ATPase protein, VCP [173]. Less information is available on the PI4K IIβ enzyme, which is loosely membrane associated, and which is recruited to the PM by Rac-GTP [77]. The structures of PI4KIIα [174, 175] and PI4KIIβ [176] were solved (Fig. 2C), which help efforts to identify inhibitors for these enzymes (see below).

Studies on PI4K in other organisms

Important studies have been also performed on PI4Ks in lower organisms and in plants. One of the firsts of these was a report on a Drosophila mutant, called Four Wheel Drive (Fwd) that showed a defect in germline cytokinesis and its cause was mapped to the fly PI4KIIIβ gene [177]. The fact that this mutant only affected male flies and only their spermatogonia, was quite puzzling given the essentiality of the yeast homologue, Pik1. As clarified in later studies, this difference is due to the overlapping functions of Fwd and PI4KIIα in Drosophila. Type-II PI4K in Drosophila (there is only one form, called dPi4KIIα) controls trafficking of secretory granules in the glue-gland [178]. The size of secretory granules in the larval salivary gland is also controlled by the Rab GTPases, Rab1 and Rab11 [179]. Curiously, Rab11 associates with Fwd and with PI4KIIIβ in mammalian cells regardless of its GDP/GTP bound states [180], but it does not interact with the type-II PI4K. Structural studies on the human PI4KIIIβ also described interactions with Rab11 and Rab11 effectors [181]. Important functions of PI4KIIIα described in Drosophila, included the regulation of Hippo signaling [182], epithelial cell polarity in the egg-chamber [183], and cell viability during wing development [184]. PI4KIIIα was also shown to be the critical PI 4-kinase in the Drosophila eye, responsible for the generation of the PI4P pool in the PM in photoreceptors where the enzyme also works with its partners Efr3 and TTC7 [185].

Zebrafish studies showed that PI4KIIIα was critical for fish development and in the pectoral fin its role was linked to PI(3,4,5)P₃ generation and FGF signaling [186]. Inhibition of PI4KIIIβ in zebrafish, causes defects in ciliogenesis in the otic vesicle and malformation of the vestibular apparatus [187, 188]. As for type-II PI4Ks, zebrafish embryos depleted of PI4KIIβ do not develop pectoral fins due to impaired Wnt signaling [189].

In Arabidopsis, most of PI4P is located in the PM [190] and PI4P appears to assume many functions that are controlled by PM PI(4,5)P₂ in mammalian cells [191]. This PM pool of PI4P is generated by the Arabidopsis PI4Kα1 enzyme, which is essential in plants and is anchored to nanodomains of the PM as part of a four-subunit complex containing the NO POLLEN GERMINATION (NPG), HYCCIN-CONTAINING (HYC), and EFR3 OF PLANTS (EFOP) protein families in addition to the kinase [192]. These proteins are the plant homologues of the mammalian TTC7/FAM126/EFR3 proteins and feature the enormous evolutionary conservation of this molecular architecture. Arabidopsis also contains two genes, PI4Kβ1 and PI4Kβ2, encoding the homologues of mammalian PI4KIIIβ These proteins work in the TGN and early endosomes and control the secretory pathway in root hairs [193, 194].

PI4Ks are essential host factors for certain picornaviruses

All these studies on PI4Ks taught us a lot about the cellular functions of these proteins, but their relevance to human diseases had been largely missing. This has almost instantly changed in 2009, when several parallel studies reported that Hepatitis C virus replication required PI4KIIIα [195–199]. Other picornaviruses, such as polio- and coxsackie-virus, and even some strains of Hepatitis C, depended on PI4KIIIβ for their replication [200–202]. These developments have prompted several Pharma to identify small molecule inhibitors specifically targeting the two forms of type-III PI4Ks [203–208] and also to generate mice with floxed alleles of PI4KIIIα or PI4KIIIβ [205, 209]. Genetic studies showed that knockout of either PI4KIIIα [146] or PI4KIIIβ [210] is embryonic lethal in mice, and that inactivation of PI4KIIIα in adult mice causes severe gastrointestinal pathology [205, 209] that causes early death preventing assessment of other defects that could manifest later. These studies made it clear that targeting these essential enzymes to combat virus infections may not be feasible through systemic administration.

As mentioned above, ACBD3 and c10orf76/ARMH3 can interact with PI4KIIIβ, and these proteins were found as important host factors for viral replication [211–215]. Mutation of residues in either ARMH3 or PI4KIIIβ that interrupt their association [137] helped assessing the importance of the interaction for viral replication [137]. More on the connection between PI4P and viral replication can be found in a recent review [215].

The new pharmacological tools and the floxed mice greatly facilitated further research on PI4Ks allowing, for the first time, to test their importance in complex biological settings. Schwann-cell specific knockout of PI4KIIIα [216] or PI4KIIIβ [217] caused myelination defects in the peripheral nerves of mice with distinctively different outcomes: deletion of PI4KIIIα yielded a more severe phenotype compared to those caused by deletion of PI4KIIIβ. Postnatal development and maintenance of pituitary gonadotrophs was also shows to require PI4KIIIα [218]. An earlier study characterized a gene-trapped PI4KIIα mice and showed that the mice developed late-stage spinocerebellar degeneration. These studies highlighted the importance of PI4Ks in the central nervous system [219, 220].

PI4P gradients drive lipid transport at membrane contact sites

One of the most exciting developments in the last decade related to PI4P function was the discovery that PI4P gradients drive the non-vesicular transport of other lipids using a counter-transport mechanism (Fig. 2C). The roots of these discoveries, again, reach back to yeast studies. For several years, experiments on the yeast Osh4/Kes1 protein have focused on the question of how Kes1 deletion restores the viability of yeast strains that are defective in the function of the yeast PI/PC transfer protein, Sec14 [221, 222]. It was recognized that Kes1 showed homology to the mammalian OSBP proteins, and it was localized to the Golgi through binding to the Golgi pool of PI4P generated by Pik1 (and not Stt4) and it negatively regulated the secretory function of the Golgi complex [222]. The ability of Kes1/Osh4 to transport sterols has then become the focus of subsequent studies, as was its regulation by phosphoinositides.

The structure of Osh4/Kes1 was solved with sterol in the binding pocket and it was shown that PI(4,5)P₂, but not PI4P, regulated Kes1 sterol transport activity [223]. Yet, another study showed that Kes1 inhibited vesicular transport from the Golgi by inhibiting the Pik1-generated PI4P pool in the Golgi apparatus and concluded that Kes1 limits both the availability and the level of Golgi PI4P [224]. Most of these observations fell in place when it was shown that Osh4/Kes1was able to bind and transport both sterol and PI4P and that these two cargoes competed for the same lipid binding cavity [225]. This study also established that Osh4/Kes1 works by exchanging sterols for PI4P between lipid bilayers. This seminal study was followed by a series of experiments that established the concept that PI4P gradients between the Golgi complex and the ER set up by PI4KIIIβ in the Golgi and the Sac1 phosphatase in the ER, drives the transport of cholesterol out of the ER to the Golgi mediated by the OSBP protein [226, 227]. Other non-vesicular lipid transporters with connection to Golgi PI4P were also described, although they do not use the PI4P as a counter-transport entity. They include the ceramide transfer protein, CERT [118] and the glycosyl-ceramide transfer protein, FAPP2 [228], both of which possesses a PH domain that recognizes PI4P along with Arf1 in the Golgi.

That lipid and PI4P exchange occurs at contacts between the PM and the ER was also first hinted in yeast, where the Osh3 protein was found to allow the ER-localized Sac1 phosphatase to access the PI4P pool generated by the Stt4 PI4K in the PM [229]. While this study concluded that Osh3 sensed PI4P and allowed Sac1 to act in trans accessing PM PI4P, it was a critical step in establishing the connection between PM PI4P, Osh3 and Sac1. This study also explained the earlier finding that the PI4P pool generated by the Stt4 kinase is controlled by the ER-bound Sac1 phosphatase [103]. Subsequent studies then described other yeast OSBP homologues (Osh proteins) capable of transporting phosphatidylserine (PS) [230] in exchange for PM-localized PI4P [231]. In mammalian cells, the OSBP-related ORP5 and ORP8 proteins were described as PS transporters working at PM-ER contacts, where they use PM PI4P generated by PI4KIIIα [232]. It has also been suggested that some ORP, such as ORP5/8 and ORP2 can exchange PI(4,5)P₂ in exchange for PS and cholesterol, respectively [233, 234]. While PI(4,5)P₂ transport by ORPs was not confirmed in other studies, it was shown the PI(4,5)P₂ can determine the efficiency of PS transport by the ORP8 protein [235]. Other ORPs can also transport PI4P and possibly use PI4P gradients, such as ORP3, which works at PM-ER contacts, is regulated by PKC-mediated phosphorylation [236, 237], and controls focal adhesion dynamics [238, 239]. The PI4P-driven transport of lipids has been implicated in other processes such lipid droplet biogenesis utilizing ORP5 to transport PS [240, 241] and in lysosomal repair where ORP9/10/1 complex transports PS [166] and ORP1L transports cholesterol [242]. These studies by identifying a central role of PI4P in the control of non-vesicular lipid transport of various classes of lipids, explained the essential role of PI4Ks in supporting viral replication of picorna viruses; all of these viruses need to create an optimal lipid composition of their replication organelle [201].

Delivery of PI to PI kinases is a critically important process

Although only indirectly linked to the PI4K research field, it may be helpful to touch upon the question of how PI is delivered to the PI4K enzymes from its site of synthesis in the ER. It has been established by early studies that phospholipids, including PI is synthesized in the microsomal fraction representing the ER [243, 244]. This, together with the recognition that PI hydrolysis by PLC during receptor activation occurs in the PM followed by increased PI resynthesis in the ER, made Michell already postulate in his 1975 Review that PI and PA should be exchanged between the PM and the ER during the accelerated “PI-cycle” upon receptor stimulation [21]. The logical assumption was then that the previously described PI transfer proteins (PITPs) [245, 246] assumed this function. PITPs will be covered in detail in separate Chapters in this collection, here we only mention that it was only in the early 2010’s that this lipid exchange process was identified. This started with the recognition in the Cockcroft laboratory that RdgBβ (PITPNC1), a relatively poorly characterized member of the PITP family can transfer phosphatidic acid (PA) between liposomes in vitro [247]. RdgBβ belongs to the so-called Class II PITPs that are multidomain proteins containing a PITP domain in their N-termini and several other domains (recently reviewed in [248]. However, RdgBβ only contains the lipid binding PITP domain, which shows higher homology to the PITP domains of the Class II than to the Class I PITPs [248]. The important report about RdgBβ being a potential PA transporter was quickly followed by studies that showed that the Class II PITP, Nir2 (Nir2 and Nir3 are Class II PITPs) is localized to ER-PM contact sites during receptor stimulation and helps maintain the signaling pool of PI(4,5)P₂ through delivery of PI from the ER to the PM [249, 250]. Our parallel studies showed that Nir2 also served as a PA transporter allowing PA transport from the PM to the ER during stimulation and suggested that Nir2 essentially functions as a PI/PA counter transporter that becomes engaged during the accelerated “PI-cycle” [251], just as envisioned by Michell in 1975. Independently, this mechanism was also found operational in the Drosophila eye where the RdgBα protein acts as the PI transfer protein in photo-signal transduction [252].

Testing the contribution of the Class I PITPs to PI delivery to PI kinases in intact cells have been more difficult as the two members that belong to this group, PITPα (PITPNA) and PITPβ (PITPNB), most likely serve redundant functions. While genetic manipulation of these PITPs in whole organisms created unique phenotypes [253–256], it is not easy to delineate how these complex phenotypes are linked to defective PI transfer functions. Our recent studies using a potent inhibitor targeting specifically Class I PITPs [257] allowed us to test their contribution to the maintenance of PI4P pools in different organelles in cultured cells [258]. What these studies showed was that Class I PITPs play a critical role in maintaining resting PI4P levels in the PM but also contribute to PM PI delivery during receptor activation, a function which becomes critical when the Nir2 protein is inactivated [258]. There is a lot more literature on the PITPs in mammalian cells, especially their functions in the Golgi, and on their functional homologues, the Sec14 proteins is yeast. These are discussed in great details in other Chapters of this collection and will not be further elaborated on, here.

However, it is hard not to see the parallels between the PI-cycle harnessing a PI/PA exchange between the ER and the PM by the Nir2/Nir3 proteins during receptor activation and another PI-cycle that occurs between the PM and the ER, mediated by the ORP proteins that exchange PI4P for PS (and perhaps other lipids) driven by the PI4P gradient that is set up by the PI4K and Sac1 enzyme pair positioned in the two opposing membranes. While the former does not have an equivalent in yeast, the latter is evolutionary highly conserved, suggesting that it is a more ancient mechanism.

PI4Ks and drug development

As mentioned earlier, several important human pathogen picornaviruses rely on type-III PI4Ks for their replication, prompting efforts to develop PI4K inhibitors to combat these diseases. Selective PI4KIIIα inhibitors have been developed in several pharmaceutical companies [203–205, 259] but their toxicity in mice [209] prevented their progression to human studies. The first inhibitor that was selective for PI4KIIIβ among PI4Ks, PIK-93, was identified during an extensive characterization of PI3K inhibitors [260], but it also targeted PI3Kγ and Vps34. Further derivatization of PIK-93, however, yielded very potent and selective PI4KIIIβ inhibitors [207]. Other studies have also identified highly specific and potent PI4KIIIβ inhibitors as potent antiviral agents [261]. Enviroxime is an interesting case, as it was initially identified as a potent agent against rhinoviruses [262] and went into clinical trials that showed no real benefits [263]. Only later studies found that enviroxime and enviroxime-like compounds targeted PI4KIIIβ, and that they also exerted potent anti-poliovirus activity [264]. While a large variety of these enviroxime-like aminothiazole compounds showed good correlation between antiviral potency and PI4KIIIβ inhibition, their acute toxicity in mice made them unlikely to be useful for human therapy through systemic administration [265]. Since viral replication also requires the lipid transfer proteins, such as OSBP or Nir2 that work downstream of the PI4Ks [266, 267] as well as adapter proteins that regulate PI4KIIIβ [136, 137, 211, 213], targeting the lipid transfer proteins rather than the PI4Ks themselves may hold more promise in combating these viruses [268, 269].

While targeting the host PI4KIIIβ may not be a feasible therapeutic strategy in humans, targeting the PI4K enzyme of cellular parasites holds significantly more promise. This is particularly true for malaria, as the Plasmodium falciparum (pf)PI4K was shown to be important for the life cycle of the parasite [270]. The Plasmodium falciparum enzyme is closer to the human PI4KIIIβ than to PI4KIIIα, but different enough from the human enzymes to exploit the difference in drug development [271]. One of the pfPI4K inhibitors, MMV390048, has already underwent clinical trials [272, 273].

Inhibitors for type-II PI4Ks have been scarce as a strong case for their potential benefits in humans in any specific disease has yet to be presented. A compound, named PI-273 was shown to inhibit the growth of MCF-7 breast cancer cells in culture or in xenografts through inhibition of PI4KIIα by competitive binding to its substrate binding site for PI [274]. Our group also has performed a high-throughput screen in collaboration with NCATS to identify small molecule inhibitors for PI4KIIα, and characterized a compound, named NCO2 that showed some promise to inhibit the enzyme. Unfortunately, multiple rounds of modifications of the NCO2 scaffold failed to yield more potent derivatives, and hence was not further pursued [275]. While we continue our efforts to find suitable PI4KIIα inhibitors, wider involvement of pharma or other research groups is yet to be manifested. Recent studies have shown that targeting PI4KIIα in lung cancer could be beneficial [276] and other studies found that high PI4KIIα predicts a poor prognosis of adenocarcinoma in colon [277]. It was also shown in Drosophila models that mutations in vesicle-associated membrane protein-associated protein, VAPB that cause amyotrophic lateral sclerosis, lead to increased PI4P levels in the late-endosome lysosome compartment causing lysosome dysfunction and defects in autophagy. Inhibition of PI4Ks was found to be beneficial to restore these functions [278]. More studies are warranted to see if this enzyme is a worthy therapeutical target, but these latest studies have increased interest in developing PI4KIIα inhibitors [279, 280].

PI4K mutations are detected in human patients

The first case of human mutation mapped to a PI4K gene was described in a family where three inviable fetuses had been diagnosed with perisylvian polymicrogyria and cerebellar hypoplasia. Whole exome sequencing identified compound heterozygous variants in PI4KA gene that rendered the enzyme catalytically inactive [281]. However, even before the connection to PI4KIIIα was known, FAM126A mutations were described causing hypomyelination and congenital cataract in five affected families [282]. Similar cases now have been documented in several additional families [283] and the pathology reproduced in fly models [284]. Mutations of TTC7A, another component of the PI4KIIIα tetrameric complex, was also found in human cases of combined immunodeficiency and intestinal atresia and in other intestinal pathologies [285–288]. As pointed out in previous paragraphs, inactivation of pi4ka in adult mice also causes severe gastrointestinal pathology [205, 209]. Since TTC7A but not TTC7B mutations were found in such human patients, it appears that TTC7A and not TTC7B is the major partner of PI4KIIIα in the small intestine in humans. The immunodeficiency associated with PI4KA mutations has recently been found to be primarily due to B-cell deficiency and hypogammaglobulinemia [289]. It has now been recognized that a large variety of PI4KA mutations can produce a variety of presentations in patients that invariably include neurological disease in combination with intestinal and immunological disorders of different degree of severity [290–292]. Many of these PI4KA mutations do not severely affect the catalytic activity of the PI4KIIIα enzyme but impair its ability to interact with the other proteins forming the tetrameric complex. Structural studies mapping the interaction surfaces between the various protein components of the PI4KIIIα complex have been extremely helpful to understand how patient mutations may affect the assembly and function of the PI4KIIIα enzyme [147–150, 293].

Heterozygous missense mutations in the PI4KB gene were described causing nonsyndromic sensorineural deafness and inner ear malformation in humans [294], but these do not fully eliminate the enzymatic activity of the protein. On the other hand, families have been identified more recently with mutations that eliminate functional PI4KIIα enzymes [295, 296], and these patients suffer from intellectual disability, epilepsy, myoclonus, and dyskinetic encephalopathy.

Concluding remarks

This overview was designed to illustrate the evolution of the PI4P field as perceived by the author who was fortunate to witness and to a small extent contribute to the development of this research field. It is, therefore, not a comprehensive review in the conventional sense having some subjective elements and probably being incomplete in its coverage of all aspects of PI4K research. The intention was to demonstrate that the early foundational studies that led Bob Michell to write his BBA review in 1975 also were the starting points to a vast amount of work that followed and unveiled the larger universe of PI4P way beyond its role of being a precursor for second messengers in the PM. It is worth pointing out a few lessons learned as this field evolved: First, the enormous impact of research on model organisms, especially yeast, that often paved the way for discoveries on more complex mammalian systems. Second, the unpredictability of how seemingly disparate approaches combined with serendipity led to profoundly new discoveries. Third, how human curiosity and motivation to understand biological systems even without much consideration to how it may benefit human health, eventually yields knowledge of great translational value. Last, but not least, the power of methodological advances in propelling research and opening new opportunities to gain a deeper understanding of the systems in question. Examples include methods of lipid extraction and separation of newly identified lipids on TLC plates or by HPLC, or the advances in molecular biology tools as well as microscopy techniques and lipid mass spectrometry. In this spirit, future advances in high resolution lipid mass spectrometry, which would allow to map lipids with their fatty acyl side chain composition with subcellular resolution is highly anticipated. It is also worth pointing out that small molecule inhibitors have a great advantage over genetic silencing of enzymes, as they do not allow the cell or organism to adapt and rewire their metabolism to accommodate the lack of the specific metabolic pathway. Even with the caveats of potential off target effects, chemical inhibitors have been quite effective in treatments of human disease conditions.

One cannot help but speculate, how and why PI phosphorylation evolutionary popped up in organisms parallel to the appearance of internal membranes and organelles even before tyrosine kinases appeared. While several lipid kinases acquired their different isoforms through gene duplications in higher organisms, there are already three distinct PI4K genes present in yeast, which are highly conserved in higher eukaryotes including plants and humans. PI is certainly a suitable scaffold to accept several phosphates on its inositol ring and perhaps served initially as a way of storing phosphates before adapting these molecules as useful signaling entities. PI4P is a prominent representative holding the key to understand these questions.

Fifty years have passed since the 1975 BBA Review. The article collection in this jubilee issue is a testament to how this 1975 review article inspired the field, which has since expanded in unimaginable scale and depth. We can be certain that this process will continue. This collection also serves as a reminder for the new generation of scientists to appreciate the work that has laid the foundation of important discoveries and will continue to inspire the work of many that will eventually improve human health outcomes.

Highlights:

In this article, Balla gives a personal account on the enormous expansion of the phosphatidylinositol 4-kinase research field since the publication of the famous BBA Review written by Robert Michell 50 years ago.

Abbreviations:

DAG

diacylglycerol

PA

phosphatidic acid

PI¹

phosphatidylinositol

PI4P

phosphatidylinositol 4-phosphate

PI(4,5)P₂

phosphatidylinositol 4,5-bisphosphate

PI4K

phosphatidylinositol 4-kinase

PLC

phospholipase C

PM

plasma membrane

PS

phosphatidylserine