Facts

  • Phosphoinositides are spatiotemporally enriched membrane lipids, and their turnover is dynamically, yet tightly, regulated.

  • Phosphoinositides were initially characterised as second messengers for major signalling pathways.

  • Phosphoinositides were later found to act as docking lipids to recruit and modulate actin cytoskeleton remodelling and membrane dynamics.

  • Involvement of phosphoinositides in cell death had been poorly defined, often overshadowed by their traditional role as second messengers and regulators of cytoskeleton-membrane interactions.

  • Recent reports support emerging importance of phosphoinositides in pyroptosis, necroptosis, host defense peptide-induced necrosis, NETosis and autophagic cell death.

Open questions

  • Considering the importance of phosphoinositides in membrane and cytoskeleton dynamics, what are their roles in morphological changes during cell death, such as apoptosis?

  • Do other lytic forms of cell death, including secondary necrosis, require phosphoinositides and what are the phosphoinositide-binding effectors?

  • Can we target phosphoinositides to modulate cell death execution for therapeutic development?

Introduction

Phosphoinositides are derived from phosphatidylinositol (PI), comprising a diacylglycerol moiety linked to a D-myo-inositol ring via a phosphodiester linkage (Fig 1a). The inositol hydroxyls at positions 3, 4 or 5 are reversibly conjugated with phosphate groups, resulting in monophosphorylated [PI(3)P, PI(4)P and PI(5)P], bisphosphorylated [PI(3,4)P2), PI(3,5)P2 and PI(4,5)P2] and trisphosphorylated [PI(3,4,5)P3] derivatives. These seven interconvertible phosphoinositide species are distinctly distributed at the cytoplasmic side of the plasma and subcellular organelle membranes where they dynamically participate in various distinct cellular processes (Fig. 1b). PI(4,5)P2 and PI(3,4,5)P3 are predominantly enriched at the inner leaflet, particularly accumulated at sites of active membrane-associated activities [1]. PI(4)P is located along Golgi-endosomal trafficking axis, on exocytic vesicles and also at the plasma membrane [2]. In contrast, 3-phosphorylated phosphoinositides are primarily associated with endosomal pathways: PI(3)P on the limiting membrane of early endosomes and intraluminal vesicles contained within multivescular bodies, PI(3,4)P2 on non-clathrin endocytic vesicles, and PI(3,5)P2 at late endolysosomal membranes [3, 4]. The precise location of PI(5)P pool(s) has not been comprehensively defined, although low levels of PI(5)P have been found in the nucleus, the plasma membrane, Golgi complex and sarco/endoplasmic reticulum [5, 6]. Under certain physiological and pathological circumstances, phosphoinositides can be synthesised at locations other than those aforementioned. For example, PI(4,5)P2 has also been observed at sites of membrane remodelling and ruffling, the nuclear mRNA splicing factor compartment (nuclear speckles), perinuclear vesicles, focal adhesion and epithelial cell-cell junctions [7,8,9,10].

Fig. 1
figure 1

Structure and cellular distribution of phosphoinositides. a Chemical structure of phosphatidylinositol, comprising phosphodiester-linked diacylglycerol and D-myo-inositol moieties. R1 and R2 represent any acyl (fatty acid) chain. Hydroxyl at positions 3, 4 and 5 of D-myo-inositol ring are readily phosphorylated, giving rise to seven phosphoinositides. b Localisation of phosphoinositides on plasma and organelle membrane. This distribution map is exemplary, only showing the cellular location where a particular phosphoinositide species is prominently found. CE clathrin-dependent endocytosis, NCE clathrin-independent endocytosis, NCV non-clathrin endocytic vesicles, EE early endosomes, RE recycling endosomes, MVB multivesicular bodies, LE late endosomes, L lysosome, ER endoplasmic reticulum, N nucleus, GA Golgi apparatus, SV secretory vesicles. Black arrow (→) indicates the progression of membrane trafficking pathway

Full size image

Phosphoinositide turnover at specific compartmentalised distributions is dynamically, but tightly, regulated by specific kinases and antagonistic phosphatases which add or remove phosphates from the 3, 4 or 5 positions of the myo-inositol ring. This critically allows for the rapid generation of product phosphoinositides or removal of the precursor phosphoinositides [1, 4, 11]. Typically, these enzymes are catalytically selective for specific phosphoinositide species, and generally exist as different isoforms with non-overlapping localisation. For instance, PI(4,5)P2 is synthesised chiefly by phosphorylation of PI(4)P by either of three PI(4)P 5-kinase (PIP5K) isoforms (α, β and γ) [7,8,9,10]. These PIP5Ks are mainly localised on the plasma membrane, although distinct additional subcellular distributions have also been observed. Together, spatially restricting and temporally governing the phosphoinositide-metabolising enzymes are critical for phosphoinositide metabolic turnover at dedicated membrane compartments and to a particular stimulation.

Despite their low abundance, phosphoinositides and their metabolism are crucial to the precise spatiotemporal regulation of key cellular events, where phosphoinositides can serve as second messengers or modulate recruitment and/or activity of membrane proteins [1, 12]. Τhese include, but are not limited to, signal transduction, cytoskeleton reorganisation, membrane dynamics, membrane and vesicular trafficking (e.g. endocytosis, phagocytosis, macropinocytosis and exocytosis). Until recently, phosphoinositide involvement in the context of cell death was largely inferred from phosphoinositide-mediated cell signalling or based on their crucial role in regulating cytoskeleton and membrane remodelling.

Phosphoinositides in the context of cell death via classical activities

Second messengers in signal transduction

Phosphoinositides have been well-defined as second messengers of transmembrane signalling. PI(4,5)P2, in particular, is an essential intermediate for two important pathways: phosphoinositide-specific phospholipase C (PLC)/diacylglycerol (DAG)/inositol-1,4,5-trisphosphate (IP3) pathway and class I phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt) pathway (Fig. 2). Typically, stimulation of G-protein-coupled receptors and receptor tyrosine kinases leads to activation of PLC, that via its PI(4,5)P2-binding Pleckstrin homology (PH) domain, hydrolyses PI(4,5)P2 to generate DAG and IP3 and thus initiating two-armed intracellular signalling cascade. DAG remains at the plasma membrane to facilitate activation and membrane localisation of protein kinase C (PKC) family members, whereas cytosolic IP3 binds the IP3 receptor and activates a ligand-gated calcium channel on smooth ER surface, triggering calcium release from intracellular storage [13,14,15]. In contrast, the PI3K/Akt signalling cascade canonically begins with the phosphorylation of PI(4,5)P2 by PI3K to yield PI(3,4,5)P3, an essential effector of Akt signalling [16]. The resultant PI(3,4,5)P3 promotes translocation of phosphoinositide-dependent kinase 1 (PDK1) and Akt to the plasma membrane, leading to Akt phosphorylation of threonine-308 and its partial activation [16]. Complete Akt activation requires additional phosphorylation of serine-473, achievable by multiple proteins, including phosphoinositide-dependent kinase 2 (PDK2), mechanistic target of rapamycin complex (mTORC), integrin-linked kinase (ILK) or DNA-dependent protein kinase (DNA-PK) [17, 18]. Subsequently, downstream transduction of DAG/PKC, IP3/calcium and Akt signalling indispensably orchestrate many cellular and physiological processes including, but are not limited to, cell proliferation, cell growth and survival, cell division, cell migration, immune response, muscle contraction, memory and learning [19,20,21].

Fig. 2
figure 2

Phosphoinositides as a second messengers. PI(4,5)P2 mediates two major transmembrane signalling pathways: PLC-DAG-IP3 and PI3K-Akt. Activated receptor recruits and induce class I PI3K to phosphorylate PI(4,5)P2. The resultant PI(3,4,5)P3 then recruits Akt and its activator kinases PDK1 and PDK2 (or mTORC2, ILK) can also to the plasma membrane, leading to complete phosphorylation of Akt. Alternatively, Akt can also be non-canonically recruited by PI(3,4)P2. Subsequently, Akt regulates downstream effectors through its activated kinase activity. PLC, on the other hand, cleaves PI(4,5)P2 into DAG and IP3 upon stimulation. DAG remains at the plasma membrane to recruit and activate PKC to modulate its downstream effectors via its regulatory phosphorylation. IP3 translocates to ER to excite calcium channels, causing calcium flux into cytoplasm, initiating calcium signalling via calcium binding effectors. PI3K phosphoinositide 3-kinases, PLC phosphoinositide-specific phospholipase C, DAG diacylglycerol, IP3 inositol 1,4,5-trisphosphate, PKC protein kinase C, Akt protein kinase B, PDK1 phosphoinositide-dependent kinase 1, PDK2 phosphoinositide-independent kinase 2, ER endoplasmic reticulum, mTORC2 mechanistic target of rapamycin complex 2, ILK integrin-linked kinase (ILK)

Full size image

Other phosphoinositides, such as PI(3,4)P2 and PI(5)P, although historically considered as intermediate products of PI(3,4,5)P3 and PI(4,5)P2 metabolism, respectively, have also emerged to be second messengers themselves. Challenging the exclusive PI(3,4,5)P3/PI3K/Akt signalling axis, mounting evidence has suggested the importance of PI(3,4)P2 in directly recruiting Akt, hence executing a distinct arm of the PI3K pathway [22]. Understanding of PI(3,4,5)P3 and PI(3,4)P2 cross-talk in regulating, balancing and contextualising Akt activity, however, undoubtedly requires future investigation. PI-derived PI(5)P is reportedly also an important second messenger for different cellular processes, depending on its production pathway [23,24,25,26]. PI(5)P generated by direct phosphorylation of PI, via phosphoinositide 5-kinase PIKfyve catalytic activity, is required for antiviral immunity, via interferon regulatory factor 3 (IRF3)-TANK-binding kinase (TBK1) axis [24]. In addition, PI(5)P can also be generated ‘indirectly’ through a complex series of steps involving class III PI3K (generating PI(3)P from PI), then PIKfyve (converting PI(3)P to PI(3,5)P2) and finally phosphoinositide 3-phosphatase myotubularin-related protein-3 (MTMR3), dephosphorylating PI(3,5)P2 into PI(5)P). PI(5)P produced by this indirect pathway is involved in cell migration signalling (via PI(5)P-recruited and activated exchange factors T-cell lymphoma invasion and metastasis 1 (Tiam1) and Ras-related C3 botulinum toxin substrate (Rac1)) [23, 25, 26].

The role of phosphoinositides in cell death, survival and proliferation is often deduced from the abovementioned phosphoinositide-mediated signalling pathways. For instance, the PI3K-Akt signalling pathway critically confers protection against apoptosis and autophagy, hence essential for cell survival (Fig. 3). Activated Akt, through its kinase activity, induces either inhibitory or stimulatory phosphorylation of components of the apoptotic machinery. For example, the phosphorylated nuclear Forkhead transcription factors (FoxO/FH) are exported to cytosol and subsequently degraded via ubiquitin-proteasome-dependent pathway, resulting in the repression of FOX-activated pro-apoptotic gene expression [27, 28]. Akt also inhibits pro-apoptotic proteins, including p73-mediated apoptotic Yes-associated protein (YAP) [29] and caspase cascade-stimulating Bcl-2-associated death promoter (BAD) and caspase 9 [30]. Conversely, Akt-mediated phosphorylation positively regulates pro-survival/anti-apoptotic gene expression, via NF-κB induction (for c-Myb and Bcl-xL expression) [31] or cAMP response element binding protein (CREB) and CREB-binding protein (for Bcl-2 expression) [32, 33]. In addition, Akt indirectly inhibits p53-dependent apoptosis via apoptotic murine double minute 2 (Mdm2)-mediated degradation of p53 [34]. In addition, the PI3K-Akt signalling downregulates autophagy via activation of downstream Akt effector mTOR kinase which is a negative regulator of autophagy proteins [35].

Fig. 3
figure 3

Phosphoinositide-mediated signalling pathways in cell death and survival. Proteins involved in (pro-)apoptosis are in dark grey box, anti-apoptosis and survival in green box, autophagy in maroon box and NETosis in orange box. Legends for arrows/lines: kinase-catalysed PIP synthesis (curl arrow), PIP-mediated membrane recruitment (dashed arrow), direct inhibition (blunt-ended solid line), direct activation (solid arrow), stimulatory phosphorylation (blue arrow), inhibitory phosphorylation (red arrow)

Full size image

Intriguingly, despite the well-characterised pro-survival role, PI3K-Akt as well as PLC-PKC cascades are also two major signalling pathways driving NETosis (Fig. 3), a more recently described form of programmed neutrophil death that results in the formation of neutrophil extracellular traps (NETs). Phorbol myristate acetate or microbial infection (e.g. Leshmania amazonesis, Candida albicans) trigger the signal transduction through PI3K and PLC, leading to reactive oxygen species release or increased intracellular calcium [36,37,38]. These key events in turn activate myeloperoxidase, neutrophil elastase and protein-arginine deiminase type 4 to promote chromatin decondensation and ultimately forming NETs to ensnare invading pathogens [37]. The discovery of the phosphoinositide-mediated pathways in transducing NETotic death signal has added another layer of complexity into functional pleiotropy of phosphoinositides. It remains to be determined how these pathways fine tune the opposing outcomes in different cellular contexts.

Actin cytoskeleton and membrane dynamics

Tightly regulated and precisely coordinated changes in the cell membrane and cytoskeleton are central to many aspects of cell physiology, such as membrane-associated and actin-dependent motility, membrane trafficking, cell polarity, cell signalling, cell division, and other morphogenetic processes. The physical and mechanistic links between the actin network and cell membrane are essential for these critical changes, which includes membrane detachment/attachment, actin polymerisation/depolymerisation, membrane curvature and membrane protrusion [39]. Phosphoinositides, especially PI(4,5)P2 and PI(3,4,5)P3, associate with multiple signal transductions, actin-binding proteins and membrane-remodelling machineries, hence they spatiotemporally and interconnectedly control the organisation and dynamics of actin cytoskeleton and plasma membrane [11, 39].

In addition to the aforementioned phosphoinositide signalling, the actin cytoskeleton remodelling and actin-binding proteins are also controlled by various master regulators, among which are small GTPases (Fig. 4). The Rho-family small GTPases (e.g. Rac1, cell division control protein 42 homologue Cdc42), are in turn regulated by PI(3,4,5)P3-recruited guanine nucleotide exchange factors (GEFs) in formation of focal adhesion complexes, extension of migratory pseudopods, or specialised actin filament assembly such as stress fibres [40, 41]. PI(3,4,5)P3, and the more rigorously characterised PI(4,5)P2, are also considered as direct-positive regulators of actin filament formation, by directly interacting with actin-binding proteins at cellular and subcellular organelle membranes in a local concentration-dependent manner [39, 40]. Binding to PI(4,5)P2 typically leads to conformational change-associated activation of proteins that are involved in promoting actin filament assembly such as nucleators Wiskott-Aldrich syndrome proteins (WASPs) and actin-related protein 2/3 complex (Arp 2/3), rulers, stabilisers and membrane anchoring. Conversely, the PI(4,5)P2 interaction can inhibit actin depolymerising factor ADF/cofilin, capping (CapZ), bundlers, crosslinkers and severing proteins as well as other actin disassembly inducers, such as profilin [39]. The PI(4,5)P2:actin-binding protein interactions are generally mediated through canonical PI(4,5)P2 binding domains, such as PH domain and band 4.1-ezrin-radixin-moesin (FERM), or defined basic amino acid-rich clusters/motifs on their sequences [42, 43]. For instance, ERM family proteins, which directly connect the actin cytoskeleton to the plasma membrane, characteristically contain an N-terminal FERM domain and C-terminal actin-binding site. An interaction between these two domains effectively establishes autoinhibition, rendering these proteins inactive. The subsequent binding of PI(4,5)P2 to the FERM domain leads to their activation by exposing the actin-binding site [44]. In contrast, gelsolin binds PI(4,5)P2 through two basic amino acid-rich regions which also overlap with actin-binding sites, resulting in the blocking of its actin binding, and thus actin-severing and filament end-capping activity [45].

Fig. 4
figure 4

Regulation of actin remodelling by phosphoinositides PI(3,4,5)P2, synthesised from PI(4,5)P2 upon receptor-mediated stimulation of PI3K via Ras GTPase, recruits GEFs to plasma membrane. GEFs then facilitate GTP loading on Rho family GTPases Cdc42 and Rac, leading to activation of actin nucleators WASPs and, subsequently, Arp2/3 complex. PI(4,5)P2 directly binds to and activates protein mediating membrane-cytoskeleton interaction (e.g. ERM proteins) or inhibits actin depolymerising factor ADF/cofilin, capping proteins (e.g. CapZ) and severing proteins (e.g. gelsolin). PI3K phosphoinositide 3-kinases, Ras Rat sarcoma GTPase, Cdc42 cell division control protein 42 homologue, Rac Ras-related C3 botulinum toxin substrate, GEF guanine nucleotide exchange factor, WASP Wiskott-Aldrich syndrome protein, Arp2/3 actin-related proteins 2/3 complex, ERM erzin, radixin and moesin, ADF actin depolymerising factor

Full size image

Dynamic membrane remodelling and generation of membrane curvatures are also mechanistically associated with phosphoinositide-mediated actin reorganisation. For example, mainly PI(4,5)P2 and PI(4)P) binding activity of certain actin-associated proteins, such as Bin-Amphiphysin-Rvs (BAR) domain superfamily, epsin/ATP180 NH2-terminal homology (ENTH/ANTH) domain family and dynamin, directly influences membrane remodelling including membrane protrusions or invaginations by manipulating phosphoinositide-rich membranes [46, 47]. The binding of BAR domain-containing proteins, such as amphiphysin 2, to phosphoinositides promotes their membrane binding and bending to create rigid scaffolds and affects the degree of intrinsic curvature and mechanistic variability [46,47,48,49]. Similarly, phosphoinositide binding via ENTH domains also enables membrane interactions of epsins as well as induce folding of distorted N-terminal extension into an amphipathic α-helix, an important structure for ENTH-mediated membrane bending/deformation [50, 51]. The interaction between dynamin and PI(4,5)P2 through its PH domain not only allows its membrane translocation, but also mediates membrane scission via PI(4,5)P2 clustering [52].

Through the modulation of membrane and cytoskeleton dynamics, phosphoinositides could be involved in morphological changes of dying cells, such as membrane blebbing (formation of circular membrane bulges) [39, 53]. The membrane blebs are locally disrupted cortical cytoskeleton-plasma membrane interaction and/or increased internal hydrostatic pressure, modulable by PI(4,5)P2-regulated ERM family proteins [39, 53]. Hence, elevated plasma membrane PI(4,5)P2 concentration suppresses bleb formation by strengthening actin-membrane interaction, whilst its depletion causes membrane blebbing due to the loss of adhesion energy between plasma membrane and cytoskeleton [39]. Pseudomonas aeuginosa phospholipase-like cytotoxin ExoU hydrolyses PI(4,5)P2 acutely causes disrupted focal adhesion, disorganised cytoskeleton structure, membrane blebbing and eventually cell lysis due to loss of membrane integrity [54, 55]. Similar membrane-bleb-associated cell death was also reported for Gambierdiscus toxicus maitotoxin, allegedly due to its PI(4,5)P2 phosphodiesterase activity [56, 57]. Membrane blebbing is also a hallmark of apoptotic cell death, formed upon Rho kinase (ROCK1)-dependent actin-myosin contraction through phosphorylation of myosin light chain [53]. Cortical contraction-induced increase in cytoplasmic hydrostatic pressure and associated disruption of ERM-mediated membrane-cytoskeleton interaction are believed to be the driving forces of bleb formation during apoptosis [58,59,60]. However, the role of phosphoinositides, including PI(4,5)P2, in apoptotic bleb formation remains largely speculative. Furthermore, as apoptotic blebs go through cycles of formation, expansion and retraction, a dynamic machinery to modulate membrane-actin cortex network should be in place [60]. The trigger and progression of such machinery is still relatively unknown. As crucial players in regulating membrane and cytoskeleton dynamics, it will be interesting to decipher the importance of phosphoinositides in the apoptotic bleb-cycling machinery.

Emerging roles of phosphoinositides in directly facilitating cell death

Considering that phosphoinositides crucially mediate numerous cellular processes, one would expect they would have greater and a more direct influence on cell death than merely being second messenger lipids of PI3K-Akt signalling or modulating membrane and actin dynamics. However, only recently have phosphoinositides been recognised for their roles in regulating cell death. A number of impactful studies have reported their essential functions in mediating the execution of lytic cell death (pyroptosis, necroptosis and defensin-induced necrosis), autophagic cell death, and NETosis (Fig. 5), whilst a few others have limitedly addressed this in the context of apoptosis.

Fig. 5
figure 5

Emerging roles of phosphoinositides in cell death. a Defensin-induced necrosis. Defensins internalise into target cells, then bind to PI(4,5)P2 and oligomerise, leading to bleb-associated membrane permeabilisation and cell lysis. b Suicidal NETosis. PI(3,4,5)P3 binding of Src kinase-associated phosphoprotein-2 (Skap2) liberates its active conformation to recruit Wiskott-Aldrich syndrome protein (WASP) for activation of integrins and actin polymerisation, which triggers expulsion of neutrophil extracellular traps (NETs). c Necroptosis. Tumour necrosis factor (TNF) binds to TNF receptor to trigger the formation of the necrosome multiplex, which phosphorylates MLKL and induces its conformational change to expose a PI(4,5)P2-binding domain. Upon PI(4,5)P2 binding, MLKL is recruited to membrane and forms lytic pore to induce membrane rupture. d Autophagic cell death. Lethal autophagy, or autophagic cell death (ACD), occurs upon autophagy overactivation, glucose starvation or death-specific inducer (e.g. lipid-binding ApoL-1). Activation of autophagy effectors (e.g. WIPIs) via PI(3)P- or PI(5)P-mediated pathway leads to the formation of autophagosomes, which are fused with lysosomes, possibly via PI(3,5)P2-dependent manner to form autophagolysomes. Membrane rupture and accumulation of autophagolysosomes are some typical ACD characteristics. e Pyroptosis. Binding of pathogen-associated molecular patterns (PAMPs) to receptors activate the inflammasome to cleave the N-terminus of GSDMD (GSDMD-Nter) from its C-terminus. Through interaction with PI(4,5)P2, GSDMD-NT relocates to the plasma membrane and oligomerises to form lytic membrane pores, causing the release of interleukin 18 (IL-18) and IL-1b

Full size image

Autophagic cell death

Autophagy, also known as autophagocytosis, is a destructive mechanism to degrade and recycle unnecessary or dysfunctional cellular components [61, 62]. Autophagy is often characterised by the formation of cytoplasmic autophagic vesicles (autophagosomes) that sequester disposed constituents targeted for digestion in autolysosomes [63, 64]. Physiologically, the tight regulation of autophagy is required for proper cellular homoeostasis. Autophagy also provides a defense mechanism by killing ingested bacteria [65, 66]. Under certain developmental, pathophysiological (e.g. post-ischaemic injury, starvation) or chemical (e.g. vintexin treatment) conditions, autophagy can result in autophagic cell death (ACD) [67,68,69]. Typical characteristics of ACD include membrane rupture, numerous autophagolysosomes, and enlargement of major cellular organelles [70,71,72]. Mechanistically, ACD is the lethal outcome of disturbed autophagic homoeostasis via excessive self-consumption of cellular content, uncontrolled mitophagy and autosis (Na+/K+-ATP-dependent ACD) [68, 69].

ACD strictly depends on autophagy machinery, which requires the production of PI(3)P by Vps34 kinase complexes, disruptions of which, demonstrated by genetic interference and pharmacological inhibitions, inhibit the autophagic pathway [73,74,75,76,77]. PI(3)P fundamentally provides docking platforms on the ER cytoplasmic membrane for autophagy effectors, and possibly mediates the fusion of autophagosomes and lysosomes to form autolysosomes [77]. Particularly in humans, PI(3)P is reported to specifically recruit Autophagy-linked FYVE protein (Alfy), WD-repeat domain phosphoinositide-interacting proteins (WIPIs) and FYVE domain-containing protein 1 (DFCP1) that sequesters target cargo to the autophagic machinery and forms an isolation compartment, eventually maturing into autophagosomes [78,79,80]. Similar to most phosphoinositide-regulating processes, PI(3)P turnover in autophagy by phosphatidylinositol 3-kinase Vps34-Beclin-1 and phosphatase Jumpy is critical for the initiation and termination of autophagy [74, 81]. Non-canonically, autophagy can be executed via Vps34-independent (hence PI(3)P-independent) pathways, mediated by PIKfyve-synthesised PI(5)P via known PI(3)P effectors, such as WIPI proteins. The non-canonical PI(5)P-dependent pathway can compensate the absence of PI(3)P, e.g. due to defective Vps34, or overrule PI(3)P in particular circumstances, such as glucose starvation [82, 83]. It also been reported that PI(3,5)P2 is required for completion of basal autophagy, particularly in mouse models of neurodegeneration. PI(3,5)P2-deficient mice exhibit neuronal loss and accumulation of autophagy intermediates, possibly due to the lack of proper PI(3,5)P2-governed autophagosome-lysosome fusion and/or lysosome function [83, 84].

A shift from basal autophagy to ACD has been linked with hyperactivation of Vps34-Beclin-1, overexpression of Beclin-1, downregulation of the molecular brakes of Beclin-1, hyperactivation of non-canonical autophagic pathways and certain death-specific factors (e.g. apolipoprotein 1 ApoL-1) [67, 68, 85]. ApoL-1, a BH3-only lipid-binding protein, can induce ACD, speculatively via disruption of phosphatidic acid-mTOR interaction and inhibition mitochondrial cardiolipin-associated apoptosis [85]. However, in this study, the protein-lipid overlay assay also indicated that ApoL-1 binds to phosphoinositides, particularly PI(3)P, PI(4)P, PI(5)P, PI(3,5)P2 at comparable affinity to PA and CL [85]. Whether the binding of ApoL-1 to phosphoinositides alters the lipid homoeostasis leading to ACD as suggested by the authors [85], or a more sophisticated phosphoinositide-associated mechanism of ApoL-1-induced ACD, requires further elucidation.

NETosis

Suicidal NETosis is a novel form of neutrophil cell death, characterised by the formation of NETs that follows DNA decondensation and plasma membrane rupture [86, 87]. NET release forms a novel host defense mechanism against invading microbial pathogens. Though the execution and regulation of NETosis are still not fully understood, it is emerging that PI(3,4,5)P3-mediated recruitment and activation of NETotic effectors, including Src kinase-associated phosphoprotein-2 (Skap2), WASPs and integrins [88]. While initial studies showed that Skap2 adopts an autoinhibitory conformation in steady-state neutrophils, upon stimulation, interactions with PI(3,4,5)P3 via its PH domain induces a switch to the active conformation [89]. Then, Skap2 can form a complex with WASPs and translocate to the inner leaflet of the plasma membrane, leading to the downstream activation of integrins [88, 89]. In Skap2−/− neutrophils reconstituted with a Skap2 R140M mutant, which has been previously shown to impair PI(3,4,5)P3 binding to the PH domain [89], both Skap2 and WASPs are unable to translocate to the plasma membrane and, consequentially, resulting in failed NETosis. These findings suggest the importance of direct PI(3,4,5)P3-mediated membrane recruitment in directing NETosis.

Pyroptosis

In order to respond to bacterial infection and danger signals, cells can undergo pyroptosis, a programmed cell death associated with membrane lysis and inflammatory cytokine release [90]. During pyroptosis, activated caspase-1 or caspase-11 cleaves the effector protein gasdermin D (GSDMD), liberating its N-terminal (GSDMD-Nter) from the autoinhibitory C-terminal domain (GSDMD-Cter) [91,92,93]. The GSDMD-Nter then relocates to the plasma membrane, through its preferential interaction with PI(4,5)P2. In fact, GSDMD-Nter harbours clusters of positively charged and hydrophobic residues, a common feature observed in PI(4,5)P2-binding motifs, within its α1 helix and the β1–β2 loop [94]. Upon PI(4,5)P2 binding, GSDMD-Nter promptly inserts into plasma membrane, oligomerises and forms large transmembrane pores to trigger inflammatory cytokine (interleukins 18 and 1β) release and, subsequently, membrane lysis [93, 95]. A recent time-lapse atomic force microscopy study has highlighted the crucial role of PI(4,5)P2 throughout GSDMD assembly progress, not only governing its membrane targeting step but also accelerating membrane insertion and ensuring correct pore topology [94].

Necroptosis

Necroptosis is another inflammatory and lytic form of programmed cell death, occurring upon tumour necrosis factor receptor stimulation and caspase inhibition [96, 97]. Unlike pyroptosis, necroptosis is however distinctively characterised by caspase independence, membrane damage and release of inflammatory damage-associated molecular patterns [96, 97]. The GSDMD counterpart for necroptosis is mixed lineage kinase-like (MLKL) pseudokinase, structurally consisting of a N-terminal bundle (NB) domain fused by a brace region to a C-terminal pseudokinase domain [96, 97]. During necroptosis, MLKL is phosphorylated by a multiprotein complex, called necrosome, leading to its transition from dormant monomeric to brace-mediated and activated oligomeric confirmation [98]. The MLKL oligomer is initially recruited to the plasma membrane through its low-affinity PI(4,5)P2 binding provided by clusters of basic residues within NB domain [99, 100]. Subsequent robust membrane association progresses as higher-affinity PI(4,5)P2 binding sites are exposed and displace the brace [100]. Mutation of basic residues at the putative binding sites suggests PI(4,5)P2 interaction is particularly essential for the lytic activity. However, it remains to determine how PI(4,5)P2-bound MLKL causes membrane rupture, possibly by its membrane destabilisation, pore formation or osmosis imbalance induced by its cation channel-like activity [98,99,100,101].

Defensin-induced necrosis

In addition to pyroptosis and necroptosis, defensin-induced necrosis represents a parallelly novel form of PI(4,5)P2-dependent lytic cell death. Defensins are small cationic peptides of host innate immunity displaying broad-spectrum antimicrobial and immunomodulatory activities [102]. Additionally, recent studies have reported on membranolytic effect of ornamental tobacco defensin NaD1 [103], tomato defensin TPP3 [104] and human β-defensin 3 (HBD-3) [105] via PI(4,5)P2-mediated membrane permeabilisation. At sub-micromolar concentrations, these defensins promptly enter mammalian cell followed by binding to PI(4,5)P2 at inner membrane leaflet, thus destabilising the plasma membrane, causing bulge-like membrane structures (blebs) and eventually cell lysis. Supported by site-directed mutagenesis and crystallographic studies, the PI(4,5)P2 interaction was mapped to analogous cysteine-flanked, β-strand-spanning, cationic clusters (33HCSKILRR40 in NaD1, 37HCSKLQRK42 in TPP3 and 32KCSTRGRK39 in HBD-3), which resemble the K/R-X(3,7)-K-X-K/R-K/R motif found in nuclear PI(4,5)P2-binding proteins [106]. Strikingly, NaD1 can form an arch-shaped oligomer with PI(4,5)P2, stabilised by an extensive network of protein-protein and protein-lipid interactions. The oligomerisation propensity may increase NaD1 avidity for PI(4,5)P2, and resultant tight NaD1:PI(4,5)P2 complex would efficiently sequester PI(4,5)P2, hence disrupting plasma membrane [103]. Interestingly, the membranolytic defensins show a great specificity towards tumour cells, likely due to certain morphological changes of plasma membranes upon tumour transformation to influence robust growth, motility, invasion and metastasis, as opposed to normal cells. These may include increased negatively charged phospholipid, increased membrane surface area and possibly increased phosphoinositide levels [105]. More recently, the oligomeric defensin:PI(4,5)P2 complex has also been reported for human β-defensin 2 (HBD-2). Though structurally and mechanistically different to the abovementioned defensins, the HBD-2:PI(4,5)P2 complex is also be crucial for C. albicans membrane permeabilisation, directly related to its antifungal property [107]. These findings accentuate an interspecies conserved PI(4,5)P2-dependent mechanism among membrane-targeting innate immune molecules, mediating the necrosis of altered-self (such as tumour cells) and invading pathogens.

Apoptosis and secondary necrosis

Some early studies demonstrated a direct, yet opposing, action of phosphoinositides in mediating apoptosis, a non-lytic, non-inflammatory, caspase-dependent programmed cell death. PI(4,5)P2 acts as a direct inhibitor of apoptosis initiator caspases 8 and 9 and their common effector caspase 3, independently of PI3K and Akt [108]. Increased PI(4,5)P2 levels by PIP5Kα overexpression exacerbates apoptotic suppression; and PIP5Kα is actually cleaved by caspase 3 in vitro at a consensus cleavage site, mutation of which prevents PIP5Kα inactivation and enhances apoptosis in vivo [108]. In contrast, PI(4,5)P2, in a concentration-dependent fashion, essentially activates cation channel P2X7 current upon ATP stimulation, and indispensably induces P2X7-mediated cell death, particularly apoptosis, in endothelial cells, T cells and macrophages [109]. It is uncertain why such discrepancy exists for the role of phosphoinositides in apoptosis, although cell-type dependency could be part of the explanation. To this point, however, the direct involvement of phosphoinositides in apoptosis remains largely underexplored despite apoptosis being the most well-studied cell death program.

Recent breakthroughs in apoptosis demonstrate that apoptosis is associated with three finely tuned downstream events with distinct  membrane morphologies including: (i) the previously discussed membrane blebbing, (ii) formation of thin apoptotic membrane protrusion causing bleb separation and (iii) protrusion fragmentation to form membrane-bound extracellular vesicles, called apoptotic bodies [110,111,112]. The entire process, collectively known as apoptotic cell disassembly, is an important physiological or pathophysiological phenomenon downstream of apoptosis. Resultant apoptotic bodies mediate efficient phagocytic removal of apoptotic cells as well as aid intercellular communication [110]. In line with the suggestion earlier, the role of membrane-actin dynamics-regulating phosphoinositides, particularly PI(4,5)P2 and PI(3,4,5)P3, in mediating morphological changes during apoptotic cell disassembly definitely require further investigation.

Intriguingly, in the events of impaired clearance, apoptotic cells can progress into secondary necrosis due to progressive loss of plasma membrane integrity. Though long regarded as an unregulated progress, recent reports suggest caspase 3-mediated cleavage of gasdermin E (GSDME, also known as DFNA5) predisposing apoptosis to membrane rupture via similar mechanisms to MLKL and GSDMD [113, 114]. GSDME knockout apoptotic cells fail to lyse, but instead extensively fragment to form apoptotic bodies. Once cleaved N-terminal fragment of GSDME translocates to the plasma membrane and induces membrane permeabilisation [113, 114]. Though detailed membrane recruitment and interaction of GSDME is unknown, based on its relatedness to GSDMD, one can logically propose a role for phosphoinositides in GSDME-induced secondary necrosis.

Exploiting phosphoinosititdes to therapeutically modulate cell death

With phosphoinositides emerging to critically integrate into an array of cell death pathways, developing phosphoinositide-targeting therapeutics could be useful in modulating cell death pathways, whether it be an inhibiting or promoting cell death. Phosphoinositide-binding cell death effectors (such as the defensins, GSDMD and MLKL) or related molecular mimetics could be exploited in the future to combat a range of diseases related to both microbial pathogens and cancer by inducing cell death through phosphoinositide interactions. This could prove an effective way of overcoming the ability of some cells, particularly in cancer, to evade programmed cell death, as direct treatment with effector molecule would enable bypass of the upstream signalling pathways.

In contrast, the opposite effect could be achieved by targeting phosphoinositides with therapeutics designed to prevent interactions with cell death effector molecules. This approach would be relevant in disease settings associated with unwanted cell death and inflammation (such as in autoimmune diseases), as reducing cell death could be beneficial. For example, in multiple sclerosis, where unwarranted pyroptosis contributes to neuroinflammation and demylenation, preventing the translocation of GSDMD-Nter to the plasma membrane by blocking PI(4,5)P2 could be useful in treating the disease [115]. It is therefore possible that other cell death-associated diseases, e.g. neurodegenerative diseases, can be limited by blocking autophagic cell death can using PI(3)P and PI(5)P-sequestering agents.

Concluding remarks

The last few years have seen a remarkable increase in reports of phosphoinositide-facilitated cell death, emphasising the importance of phosphoinositides in many important aspects of life and death within the cell. Direct PI(3)P, PI(3,5)P3, PI(4,5)P2 and PI(3,4,5)P3 function in mediating autophagic, lytic and NETotic forms of cell death essentially contribute to the homoeostasis, development and host defense mechanisms in many disease settings by preventing pathogen/tumour cell growth and alerting host immune response. These seminal findings however require additional in vivo evidence to strengthen the emerging importance of phosphoinositides in the context of cell death. Future investigations should also further delineate the death effector-phosphoinositide interactions and how such interactions molecularly execute different forms of cell death. It is worth studying how these pathways differentially are employed and/or cross-talk in different immune cells and against various infections/altered-self. There may also other cell death settings possibly recruiting phosphoinositides, particularly PI(4,5)P2, for their enactment including apoptotic cell disassembly, secondary necrosis and necrosis by other lytic host defense peptides and molecules such as venoms. Taken together, phosphoinositides, in particular PI(4,5)P2, may have potential as novel targets for combatting infection and cancer. The ability of PI(4,5)P2 to orchestrate lytic cell death through the recruitment and activation of death effectors is indeed an attractive feature to manipulate cell death for therapeutic benefit.