PIP₂ signaling dictates the activatable state of a plethora of ion channels [2,14,15] (Fig. 1) with broad reaching cellular function. The first indication that a channel is PIP₂ dependent usually arises when a channel, excised from the plasma membrane (e.g. inside out patch), steadily decreases in conductance until the channel inactivates, this is known as “rundown” [2,16]. The excised patch lacks the cytosolic factors to maintain sufficient PIP₂ levels in the membrane to support ion channel function; hence the channels in the patch close. Adding ATP and Mg was shown to delay rundown [16]. Presumably, PIP₂ synthesizing enzymes are excised in the patch with the channels and that these enzyme utilize the ATP to replenish PIP₂ [2,16]. Adding back a soluble PIP₂ analog dioctanoyl PIP₂ (C8PIP₂) rescues activity [2,15] of many ion channel types [17–20]. In a second method, PIP₂ scavengers (e.g. polyamines or PIP₂ antibodies) are used to deplete or mask PIP₂ availability [21–23]. Polyamines are positively charged polymers that bind via avidity to the multiple negative charges of PIP₂. More complete descriptions of PIP₂ dependent ion channels and PIP₂ cellular function are reviewed by Suh and Hille [2,11], Xie [5], and McLaughlin [9]. Recently, a voltage sensitive phosphatase (Ci-VSP) was shown to provide direct control over PIP₂ signaling in the membrane [24–26]. When Ci-VSP is co transfected with K_ir [24–26], K_v7.1 [27], Ca_v2 [28,29], and TRP [30,31], channels are voltage dependent consistent with Ci-VSP regulation of PIP₂. This method provides better control of PIP₂; however, indirect effects of PIP₂ remain a possibility.

In order to directly show PIP₂ modulation, an ion channel can be purified and reconstituted (reinserted) into lipid vesicles with a known lipid composition. A lack of purified ion channels limited this technique, but recent advancements in membrane protein expression and purification [32,33] has overcome this problem for select channel types [34–38]. The nAChR was among the first channels to show direct dependence on a lipid for activation, phosphatidic acid (PA) [39]. Recently PIP₂ dependent channels were reconstituted into lipid vesicles and shown to respond directly to PIP₂ modulation, this includes GIRK [40,41], TRPV1 [42], TRPM8 [43], and K_ir2.1-2 [44] channels.

Despite robust channel modulation by indirect methods, absent a crystal structure, an understanding of the molecular action of PIP₂ and the precise binding site remained speculative. In 2011 an X-ray crystal structure complex of K_ir2.2 with PIP₂ revealed a PIP₂ binding site in the channel’s transmembrane domain [10] (Fig. 2). The glycerol backbone and 1′ phosphate of PIP₂ capped the first transmembrane spanning helix (TM1) of K_ir. Intimate coordination of the 5′ inositol phosphate in the distal end of the second transmembrane spanning helix (TM2) accounted for PIP₂ specificity. And a conformational change appeared to initiate or open the ion conduction pathway. Basic residues on a linker between the transmembrane domain and cytoplasmic domain directly contacted PIP₂, but distal basic residues proposed in the CTD [45] did not, rather they were buried and stabilized proper folding of the cytoplasmic domain structure [10]. Prior to the K_ir2.2/PIP₂ complex, structures of PIP₂/protein complexes were limited to soluble membrane localization domains, which lack a transmembrane domain and share few if any functional similarities with ion channels. A lack of appropriate structural examples and an understanding of how lipids and proteins interact in the plasma membrane hindered a complete mechanistic interpretation of PIP₂ data. Furthermore, early studies on the C-terminus of K_ir included residues that turned out to be in the TMD of K_ir and key to binding the 5′ inositol phosphate [6] (Fig. 2). Only with recent structural data has a model emerged where lipids bind to specific sites in the transmembrane domain of ion channels [10,46–49].

Taken together these finding suggest a ligand-gating theory of PIP₂ activation. In biochemistry, the term ligand refers to the reversible, specific, and dose dependent binding of a substance to a protein to form a complex. Ligands include small molecule drugs, hormones, peptides, and metabolites. Normally ligands stabilize at least two states, one bound and one unbound [50,51].

The binding of PIP₂ to K_ir has many features of a ligand. First, PIP₂ is in low abundance [9,12]. This requires that PIP₂ bind with high affinity to its targets to exert an effect. Second, PIP₂ binds reversibly to ion channels in a dose dependent manner [20,23]. Third, PIP₂ binds with specificity; for example, PI(4,5)P₂ activates K_ir2.1 and PI(3,4)P₂ inhibits the same channel [52]. This specificity is striking since the two lipids are chemical isomers and only differ in the position of the 5′ phosphate. Another anionic lipid, oleoyl-CoA, competitively and reversibly inhibits all K_ir’s [52] except K_atp, which is specifically activated by oleoyl-CoA [53,54]. Fourth, like neurotransmitter, PIP₂ is a dynamically regulated molecule [55,56]; a signaling cascade can rapidly change the concentration of PIP₂ to cause the channels to open or close [57–59]. And lastly, PIP₂ channel affinity determines channel function [60]. Mutations that allosterically decrease the affinity of PIP₂ cause disease (e.g., Andersen-Tawil syndrome) [45,61].

The ligand-like characteristics of PIP₂ binding to the entire family of inward rectifiers warrant classification of these channels as ligand-gated. The unique properties of lipids logically give rise to a lipid subclass suggested here “lipid-gated” ion channels.

PIP₂ was first speculated to induce ion channel activation by non-specific avidity of negatively charged phospholipid binding to clusters of basic amino acids in the C-terminus of channels [2,5,8]. Anionic lipids were thought to accumulate on the inner leaflet and non-specifically attract positively charged residues on the surface of K_ir’s cytoplasmic domain (CTD). The rational for the theory is sound and was based on data from K_atp (K_ir7.x) [21,62–64] and proteins like MARCKS [2,8]. However, in light of the PIP₂/K_ir complexes, the previous role of electrostatic theory appears inadequate for K_ir. The glycerol backbone of PIP₂ bound tightly to the transmembrane domain (TMD), and the inositol phosphates interacted with residues in or proximal to the TMD, not the CTD. The original influential lack of K_atp’s specificity is an anomaly among K_ir’s and appears to be an adaptation that allowed regulation by oleoyl-CoA [20] and not a mechanistic requirement as speculated. If non-specific anionic interactions regulate K_ir, the site of anion lipid binding are likely distal to the canonical PIP₂ site [65] or act synergistically with PIP₂ [44,66] by binding to one of the 4 canonical sites. The notion that the cytoplasmic domain is the binding site for PIP₂ and that PIP₂ localizes the CTD similar to a PH domain appears to be incorrect. The K_ir2.2 CTD did move toward the membrane and may reflect an evolutionary origin; but the primary mechanism appears to be an allosteric conformational change, not non-specific electrostatic attractions of the CTD to the membrane surface. The key PIP₂ binding interactions were confirmed in a complex of PIP₂ with GIRK2 [48] suggesting a common mechanism in related K_irs (Fig. 2B).

Voltage activated ion channels better exemplify non-specific electrostatic interaction. A well studied domain called the “voltage sensor domain” (VSD) senses and responds to changes in surface charge [33,47,67,68]. Conserved basic residues in the VSD electrostatically move towards the charge causing a conformational change that gates the channel. The charge is non-specific and can be applied by external current or by changing the charge of lipids in the plasma membrane. The latter was shown in recent bilayers studies where K_v responded symmetrically and non-specifically to anionic lipids [69]. The same study showed a distinct phosphatidic acid site in the cytoplasmic leaflet that specifically and dramatically affected K_v gating [69]. This suggests both ligand and electrostatic modes can operate in the same channel, however the structural determinants of the two are likely distinct. A similar arrangement exists in Ca_v2, which has a voltage sensor and a putative PIP₂ specific binding site [11,70].

Few other channels currently have sufficient molecular description to definitively discriminate the mechanism of action seen in K_ir and K_v. Many tetrameric channels exhibit a C-terminal charged cluster and varying degrees of specificity reminiscent of K_ir, including TRP [19,42,71–75], and P2X4 [76,77] (see table 1). Typically, these charges immediately follow or are located in the last transmembrane domain. Many other channels respond to PIP₂ in ways that parallel K_ir responses, including Ca_v [70], NMDA [78], K_v [27], P2X1-3 [79] channels (see also Fig. 1C), but it is unknown if the interactions are direct with the TMD or indirect through membrane charge or other proteins. Since numerous soluble domains use polybasic clusters to target to the plasma membrane [80], some yet undefined cytoplasmic domains could utilize a membrane surface charge as previously speculated [2,8]. Future structural studies will continue to reveal the details and breadth of electrostatic theory.

Lipids are sometimes viewed as co-factors. Before discussing PIP₂ as a cofactor I must first define a cofactor and distinguish it from a ligand. The term cofactor stems from enzymology and generally refers to a permanent organic compound or metal that is required for the enzyme to function. A cofactor normally derives its function by remaining bound to a protein. In contrast, a ligand derives its function by binding and dissociating from its partner protein. Lipids have always existed in cells and it is reasonable to assume that some lipids may bind as cofactors. A crystal structure of K_v in a lipid like environment revealed phospholipid binding sites near the voltage sensor and some of these appear to be lipid cofactors [47]. In other words they facilitate the proper organization of the channel but at present they do not appear to initiate a change in the channel state by dynamic regulation of the lipid.

In a speculative role, PIP₂ was proposed to act as a ‘coincidence detector’ in order to facilitate transport of an inactive channel [2,15,81,82]. A nascent channel in the endoplasmic reticule (ER), where PIP₂ is scarce, remains inactive until it arrives, at the plasma membrane where an abundance of PIP₂ constitutively activates the ion channel. This fits well a definition of cofactor in the resting state. Directly demonstrating the physiological contribution remains a challenge since PIP₂ is dynamically regulated [2]. For example PLC hydrolysis of PIP₂ in the plasma membrane inhibits K_ir [58,59], a function also consistent with ligand-like properties.

In another speculative role, PIP₂ might function as a cofactor in sensing protons. The pka’s of inositol phosphates are around 6.5 and 6.9, an optimal range for sensing physiological changes in proton concentration [83]. The lipid could remain bound and simply supply the metal phosphate as a proton sensing cofactor. Ions interacting with lipids were recently shown to regulate a receptor [84]. Acid sensing ion channels (ASIC) are likely candidates for such a mechanism since they bind PIP₂ and sense protons. Alternatively, PIP₂ may serve as a proton sensitive ligand. An atomic structure is known for ASIC [38] but the role of PIP₂ in channel activation requires further investigation.

Perhaps one reason for a slow adaptation of a “lipid-gating” model for PIP₂ is the fact that the prototypical PIP₂ gated channel K_ir is active during the resting state of excitable cells. These channels are often considered “constitutively active” leak channels. While it is true they allow potassium out of the cell during the resting state, acetylcholine stimulation of M1 muscarinic receptor inactivates K_ir [58,59]. An early study on high affinity K_ir2.1 in oocytes showed resistant to ACh inactivation [60], but later studies in mammalian cells demonstrated robust and complete inhibition of K_ir2.1 through activation of M1 receptor [59]. Thus neurotransmitter induced closure of K_ir potassium channels is presumably synergistic with the opening of calcium, sodium and voltage-gated channels, and should result in a stronger action potential or sustained excitability.

Phosphoinositides distributes heterogeneously in the plasma membrane [85–87]. Hydrophobicity causes lipids to partition (see Fig. 5). Saturated lipid chains partition into cholesterol rich lipid rafts, often referred to as detergent resistant membranes (DRMs). Lipids with unsaturation partition into the liquid disordered phase (L_d). Mass spec of resting cells indicate that PIP₂ is comprised of a polyunsaturated fatty acyl chain [88–90] and localizes in the L_d region of the membrane [87]. Quantitative studies of PIP₂ suggest close to 85% of PIP₂ is polyunsaturated and 70% comprised of an arachidonyl acyl chain [90]. In contrast, PIP₃ is primarily comprised of saturated or monounsaturated lipid acyl chains [89]. Strikingly, arachidonyl PIP₃ was not detected in quiescent cells [89]. Based on standard lipid partitioning, the saturated PIP₃ is likely located in cholesterol rafts. In agreement with this arrangement, PI3 Kinase (the enzyme that generates PIP₃ from PIP₂) localizes to lipid rafts [91]. Taken together, these data indicate an acyl chain based localization of PIPs in the plasma membrane. Figure 4 shows a hypothetical layout of the quiescent cell based on available, but limited, mass spec, super resolution imaging, and localization studies [87–90].

Famously, Gq coupled GPCRs (guanine nucleotide coupled receptors) hydrolyze PIP₂ through phospholipase C (PLC) activation. G protein mediated PIP₂ hydrolysis was known more than 30 years ago [92]. However, most cell biologist viewed (and many still do) PIP₂ as little more than a substrate for second messenger signaling [93]. This view is inadequate for K_ir channels; PIP₂ must also be viewed as an ion channel activator [3,6] or agonist. Hence, hydrolysis of PIP₂ by M1 muscarinic receptors should be viewed as a direct regulatory mechanism to deplete agonist. PIP₂ hydrolysis inactivates both high and low affinity K_ir channels [58,59]. Downstream modulation of K_ir by phosphatases and kinases appear secondary to this direct PIP₂ regulation [6,94], a rational also supported by the central and highly conserved role of PIP₂ in channel activation as described above (2.5). PLC regulation of Ca_v [70], K_ir [95], HCN [96], K_v7 [27], K_2P [97], and TRP [98,99] channels (among others) is well-documented.

In addition to PLC, GPCR signaling activates phospholipase D [100] (PLD). PLD produces PA and free choline. PA has emerged as an important signaling lipid [101]. PA and PIP₂ appear to synergistically activate K_ir [44] and K_2P [102] channels, in contrast the nAChR [39] and some K_v [69] respond specifically to PA and not PIP₂. A third important class of lipases phospholipase A2 (PLA₂) also exhibits GPCR regulation [103]. PLA₂ hydrolyzes arachidonyl-lipids creating lysophospholipids and arachidonic acid. Downstream and second messenger signaling are well studied for PLA₂ and PLC and include the arachidonic cascade and IP₃ second messenger signaling respectively. In comparison the upstream role of the intact bioactive arachidonyl-phospholipids and PIP₂ is much less understood. Nonetheless, the added role of PIP₂ in directly gating ion channels solidifies a direct rout for GPCR regulation of ion channels independent of downstream kinases and calcium signaling [6,94].

Several ion channels bind G-proteins directly, this role is widely accepted for the G-protein regulated inward rectifiers (GIRK/K_ir3.x) and N-type calcium channels (Ca_v2) [104]. A trimeric complex of GIRK with Gβγ (a G-protein) and PIP₂ revealed the GIRK/Gβγ interface [40]. And biochemical studies suggest that Gβγ is important for increasing binding of PIP₂ to GIRK [41]. The precise mechanism by which Gβγ enhances PIP₂ activation needs further clarification.

Lipases localize with ion channels to increase the speed and specificity of PIP₂ channel gating [98,105]. For example rhodopsin activated PLC hydrolyzes PIP₂ opening TRPL channels. Colocalization of PLC with TRPL [98] allows for a fast 20ms response time [106]. And TRPM7 directly binds PLC to locally affect channel activation [94]. PLC functionally colocalized with NMDA receptors [78] and the IP3 receptor co-localizes with PLC to regulate calcium release [107]. PLD lipases directly localize to ion channels, including TRPM8 [108] and TREK-1 [109]. There are many subtypes of lipases; their diverse regulation and specific localization satisfies cells with the needed diversity for signaling.

The partitioning of PIPs and their modifying enzymes appears primed to deliver dynamic cell signaling. During a signaling event, G-proteins control PIP kinases, lipases, and phosphatases, to degrade PIP₂ signaling. This signaling generates lipid degradation products (Fig. 5B). For example, it was shown PLC activation generates arachidonyl-diacyl-glycerol [90]. And PLA₂ activation removes the arachidonyl sn2 acyl chain generating lysoPIP₂ [110]. In order to return to a resting state, degradation products need to be removed from the membrane and PIP₂ resynthesized.

Endocytosis recycles lipid micro domains and lipid rafts after signaling [111]. The late endosome and ER feed back into PIP₂ signaling. This postsynaptic lipid reuptake would then reset the membrane for another signaling event analogous to presynaptic neurotransmitter reuptake. Further studies are needed to understand the temporal and spatial regulation of PIP₂ in vivo in particular during a signaling event. However, signaling lipids are known to control the ion channel desensitization [19], voltage dependence [75], and recovery from inactivation [94], and these events correlate with ion channel rundown. Lipid regulated desensitization may prove to be a central function of many channel types. Much more data is needed to build a complete picture.

Lipids localize topically by leaflets generating a lipid signal. For example phosphatidylserine (PS) is found on the inner leaflet of the plasma membrane. Enzymes known as flippases and floppases move lipids between leaflets [112]. PS signals by flipping outside the cell [113]. PS is negatively charged and movement outside the cell has the ability to change the membrane surface charge from negative inside to negative outside. Recently, asymmetric changes to the charge of lipids in a bilayer dramatically shifted the voltage midpoint potential of a K_v channel[69]. Hence lipids may “flip” as a rapid mechanism to impose a lipid induced change on the cell membrane potential, a mechanism that would have likely preceded a synapse.

In a separate mechanism, lipid acyl transferases (LAT) could signal to ion channels by changing the unsaturation of a lipid acyl chain. LAT enzymes add acyl chains to lipids or move acyl chains between existing lipids [114]. If a LAT enzyme swaps an arachidonyl acyl chain with a saturated one, the signaling lipid would most likely translocate to a lipid raft (Fig. 5). This may simply sequester the signal away from the ion channel by moving the lipid into or out of a lipid micro domain. Alternatively, the translocation could make the lipid available to other modifying enzymes that would then deplete the signal from the membrane.

Or, lipid acyl chains may directly contribute to gating of an ion channel. The acyl chains contain chemical diversity and putative specificity could determine the affinity of the lipid for the channel or cause a specific conformational change that gates the channel. Hydrophobic sites for lipid acyl chains affect PIP₂ activation of Ca_v2.2 [11].

The four identical binding sites in K_ir are positioned for PIP₂ cooperatively and allosteric competition. Tetrameric channels engineered to have only one binding pocket indicated that one PIP₂ molecule is sufficient to activate the channel [66]. In wild type channels with four binding sites, PIP₂ in combination with PA, PG, or PS, dramatically increased channel conductance. However, absent PIP₂, these lipids failed to activate K_ir [44]. A structure of K_ir with PA bound showed PA binding to the canonical PIP₂ site [10]; a site also compatible with PG and PS. In biochemical studies, oleoyl-CoA, an endogenous inhibitor, also competes directly with PIP₂ [52]. Taken together, these studies suggest that in K_ir the lipid binding site is always occupied, and K_ir integrates the sum total of the lipid environment in a cooperative way. At least one site must be occupied by PIP₂; the remaining three canonical sites appear to be available to exert cooperative activation or inhibition through a rigid conformational change [10] in the CTD. Thus additional PIP₂ binding events are poised to activate K_ir with increasing affinity consistent with electrophysiology recordings [66].

Lastly, the relative abundance of diet-derived fatty acids may affect the levels of PIP₂ signaling in the plasma membrane. Cells appear to incorporate the relative amounts of saturated and unsaturated fats into their cell membranes (phospholipids) [115]. It is tempting to speculate that diets with excess saturated fat would lead to saturated PIP₂ signaling, which most likely favors PIP₃ signaling. Diets with large amounts of polyunsaturated fats (PUFAs) would lead to more arachidonyl-PIP₂ and more PIP₂ signaling. This may account for the positive affect of dietary PUFAs on heart arrhythmias and insulin resistance since PIP₂ channels (including K_ir) are central to both these diseases. Consistent, with this model, loss of PIP₂ channel activation is associated with the disease states [61]. Similar speculation could be made of chronic pain and perhaps some cancers. Understanding PIP₂ acylation may shed light on these important medical problems.

Better methods are needed for assaying lipid interactions with ion channels. Most studies rely on crude pharmacological shifts of PIP₂ concentrations in biological membranes; this is inadequate. Varying the concentration of lipids in a liposome is a good step in the right direction. Normally one describes a ligand in terms of an on and off rate. Certainly lipids have an affinity for ion channels, but we lack the methodology for effectively measuring lipid channel interactions. Better quantitative lipid binding assays are needed. New mass spec techniques will likely allow for quantitative measurements of lipids in vivo and in vitro. And there is no doubt lipidomics will continue to find ways to improve the quantitative, temporal, and spatial identification of lipids in a membrane.

The plasma membrane holds thousands of lipids with functions that remain largely a mystery [116]. A catalog of lipid signals appears poised to exert exquisite regulation on membrane proteins perhaps rivaled only by protein phosphorylation. Certainly the phosphodiester bonds in lipids are equally suited for rapid signaling. And lipid acyl chains may be as diverse in function as they are in chemistry. Recognizing low abundant phospholipid signaling molecules as potential ligands for membrane proteins reveals a vast pool of putative effector ligands for cellular signaling.