Phosphoinositide signaling in plant development

Biological membranes act as a barrier limiting the free exchange of molecules and information. They also serve as a primary contact site between intracellular and extracellular spaces and, in eukaryotic cells, between subcellular compartments. Proteins embedded in, or associated with, membranes enable membranes or membrane areas to serve particular functions (Nicolson, 2014). These functions include the controlled exchange of substance and information across membranes, but also the attachment of proteins and cytoskeletal structures to membranes, and the insertion or recycling of membrane components (Fig. 1A). Membranes are composed of a diverse array of lipids (van Meer et al., 2008). Owing to their structural complexity and various membrane systems, eukaryotic cells require an extra level of control of subcellular membrane traffic and have evolved to contain specialized classes of membrane lipids that exert regulatory influences on membrane-associated processes (Lee, 2004). Such regulatory lipids can exert their influence by acting as ligands for membrane-associated proteins (Eyster, 2007) or by influencing the biophysical properties (Lundbaek et al., 2010) of the membrane itself, or both (Fig. 1B).

Phosphoinositides (PIs) are an example of such regulatory membrane lipids found in eukaryotic membranes. Unlike the majority of membrane lipids, which serve a structural role, PIs are of only minor abundance and their dynamic formation occurs within narrow spatial and temporal limits. Although plant PIs were discovered two or three decades ago, their function long remained elusive. However, in recent years, PIs have been shown to have crucial roles in plant development and function (Boss and Im, 2012; Heilmann and Heilmann, 2015; Munnik and Nielsen, 2011; Thole and Nielsen, 2008). Here, I provide an overview of PI structure, biogenesis and modes of action, and I discuss the key functions of PIs in plants.

As in other eukaryotes, all plant PIs are formed via phosphorylation of the head group of the membrane phospholipid, phosphatidylinositol (PtdIns). The PtdIns head group is composed of D-myo-inositol, which is a cyclic polyol linked via a phosphodiester bond in the C1 position to the glycerin backbone of the lipid {Fig. 2A; this figure depicts the structure of the key phosphorylated PI, phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P₂]}. The hydroxyl groups in positions 3, 4 and 5 of the lipid head group are accessible for phosphorylation. There are notable differences in the complement of PI-based species found in plants and animals, and in plants only five of the seven PI species known from other eukaryotic model systems have been detected, namely the PtdIns monophosphates PtdIns3P, PtdIns4P and PtdIns5P, and the PtdIns bisphosphates PtdIns(3,5)P₂ and PtdIns(4,5)P₂ (Fig. 2B). The presence of PtdIns(3,4)P₂ or PtdIns(3,4,5)P₃ in plants has not been reliably confirmed.

PtdIns3P and PtdIns4P are formed from PtdIns by specific PI-kinases (Fig. 2B): the PI 3-kinase VPS34 and at least two subfamilies of PI 4-kinases, namely PI4Kα and PI4Kβ, respectively (Mueller-Roeber and Pical, 2002). To date, no PI-kinase has been demonstrated to be capable of generating PtdIns5P from PtdIns, leaving the biogenesis of this lipid unclear. PtdIns3P and PtdIns4P can be further phosphorylated by PIP-kinases, resulting in the formation of PtdIns(3,5)P₂ and PtdIns(4,5)P₂, respectively. The formation of PtdIns(3,5)P₂ from PtdIns3P is catalyzed by four PI3P 5-kinases that bear similarity to yeast Fab1, while the conversion of PtdIns4P to PtdIns(4,5)P₂ is carried out by PI4P 5-kinases, which are encoded in the Arabidopsis genome in subfamilies A, comprising isoforms PIP5K1-PIP5K9, and B, comprising isoforms PIP5K10 and PIP5K11 (Mueller-Roeber and Pical, 2002). As a side reaction, both these subfamilies also convert PtdIns3P to PtdIns(3,5)P₂. Some plant PI4P 5-kinases display particular domain structures and features not encountered in related enzymes from other model systems, suggesting that the plant enzymes contain additional mechanisms for regulation. Examples are the highly variable linker domains of PI4P 5-kinases of subfamily B in Arabidopsis, which do not resemble any known protein sequence and have been shown to be involved in the specification of downstream regulatory effects of PtdIns(4,5)P₂ formed by these enzymes (Stenzel et al., 2012). The same enzymes also contain multiple membrane occupation and recognition nexus (MORN) repeats not found in PI4P 5-kinases from other organisms (Mueller-Roeber and Pical, 2002).

In comparison to the well-studied and highly specific enzymes of PI biosynthesis, there is only limited information about the enzymes involved in PI degradation (Fig. 2C). The Arabidopsis genome encodes a number of PI phosphatases and phospholipases (Mueller-Roeber and Pical, 2002). However, partly due to their low degree of substrate specificity, it has been difficult to assign particular catalytic or physiological functions to many of these candidates. Important families of PI phosphatases include suppressor of actin (SAC) phosphatases (Zhong and Ye, 2003), which are involved in the degradation of PtdIns bisphosphates at the plasma membrane or the tonoplast, and phosphatase and tensin homolog deleted on chromosome 10 (PTEN)-related enzymes, which mediate the degradation of 3-phosphorylated PtdIns monophosphates and PtdIns bisphosphates (Pribat et al., 2012). Further relevant phosphatases are enzymes of the Arabidopsis PI 5-phosphatase (5PTase) family, which exhibit varying specificities and may also act on soluble inositol polyphosphates (IPPs) (Gunesekera et al., 2007). PIs can also be degraded by PI-specific phospholipase C (PI-PLC) enzymes, which are encoded in the Arabidopsis genome as a family of nine isoforms (Mueller-Roeber and Pical, 2002). All plant PI-PLCs reported so far are activated by Ca²⁺ and display similarity to the human ζ-family of PLCs that cannot be activated by heterotrimeric G-proteins (Pokotylo et al., 2014).

The enzymes listed above together mediate the biosynthesis and degradation of PIs. PIs are constantly formed and degraded, resulting in rapid and dynamic turnover (reviewed by Heilmann, 2016). This high turnover contributes to the dynamic changes in PI levels that have been observed in numerous studies in different plant models. For instance, the levels of PtdIns(4,5)P₂ change in response to environmental stresses, including wounding (Mosblech et al., 2008), salt or osmotic stress (DeWald et al., 2001; Einspahr et al., 1988a,b; Heilmann et al., 1999, 2001; König et al., 2008, 2007; Pical et al., 1999) or heat (Mishkind et al., 2009). PI formation also changes transiently during gravitropic curvature (Perera et al., 1999, 2001), and it has also been shown that PI levels respond to exogenously applied auxin (Tejos et al., 2014) or salicylic acid (Krinke et al., 2007b). Notably, Arabidopsis PLC2 has been reported to act in the endoplasmic reticulum (ER) stress response (Kanehara et al., 2015). The modes of regulation underlying the dynamic interplay between PI biosynthesis and degradation in plants remain unclear. The enzymes catalyzing PI interconversions are expressed in different patterns (see Table 1), with some being expressed ubiquitously and at all developmental stages analyzed, whereas others are restricted in their expression patterns to certain organs or cell types (Heilmann, 2016). The transcription of some enzymes of PI metabolism is also induced upon perception of environmental stresses. For instance, in Arabidopsis transcription of the PLC1 gene is induced upon dehydration and salt stress (Hirayama et al., 1995) and that of PIP5K1 by exogenous auxin (Tejos et al., 2014). However, a review of publicly available transcriptomic data (Zimmermann et al., 2004) indicates that the transcription of the majority of PI-related genes does not change dramatically upon challenge by a wide range of stresses, suggesting that transcriptional control might in sum contribute only little to the dynamic changes in PI levels observed during plant stress responses (Heilmann, 2016). The control of PI dynamics might thus not take place at the transcriptional level, but rather via post-translational modifications such as phosphorylation (Westergren et al., 2001). However, to date the relevance of post-translational modifications of PI pathway enzymes during plant stress responses has not been addressed in detail.

Overall, plants contain a complex network of enzymes that mediate the dynamic formation and degradation of PIs. As in animals, plant PIs appear to be continuously formed and degraded with a high rate of turnover, resulting in a dynamic equilibrium that is easily perturbed when plants are exposed to stress conditions. This dynamic equilibrium enables fast and transient or localized signaling events to be mediated by PIs in response to developmental cues or stimulation.

PIs exert their functions via various mechanisms. For example, they can act as intact lipids, binding to target proteins, or they can influence the properties of the membrane in which they are embedded. Alternatively, they can act as precursors for the formation of soluble IPPs (Fig. 3).

Intact PIs are found predominantly at the cytosolic face of cellular membranes. It has also been reported that under certain conditions PIs may occur in the luminal/extracellular leaflets of vesicles or the plasma membrane (Gonorazky et al., 2012, 2008; Kale et al., 2010), although it is currently unclear how PIs would manifest in these locations. Within the membrane, PIs act as regulators of integral membrane proteins or mark membrane areas for the recruitment of peripheral membrane proteins. In this context, PIs act mainly through their head groups, which are characterized by their phosphorylation patterns and protrude from the plane of the membrane further into the cytosol than the head groups of other, structural lipids (Takenawa, 2010). These distinctive phosphorylated PI head groups can be recognized and bound by PI-binding protein domains, such as Pleckstrin homology (PH) domains, Fab1 YOTB Vac1 EEA1 (FYVE) domains and phagocytic oxidase (PX) domains (Lemmon, 2003). The interactions between PIs and such domains can result in the recruitment of particular proteins to membrane areas enriched in PIs.

A number of proteins regulated by PI binding have been identified; in some studies, these proteins have been termed PI-‘modulins' (McLaughlin et al., 2002). It should be noted that not all PI-binding domains of such modulins display a high degree of specificity towards a particular PI species (Fig. 3A). In fact, only a select few PI-binding domains have been found to be highly specific, including the PH domain of human PLCδ1, which preferentially binds PtdIns(4,5)P₂, and the PH domain of human four-phosphate adaptor protein 1 (FAPP1, also known as PLEKHA3), which preferentially binds PtdIns4P (Takenawa and Itoh, 2006). Nonetheless, even though most PI-binding proteins do not display exquisite specificity, protein recruitment to PIs does exhibit substantial specificity depending on the local abundance of certain PIs and on coincident binding to other (protein) factors and coincidence detection (Carlton and Cullen, 2005), a concept discussed in more detail below. It should also be noted that PI-binding domains, such as PH domains, are not well conserved with regards to primary sequence and are instead characterized by their three-dimensional fold. These domains are therefore not readily identified by mere sequence analysis. Extensive bioinformatics analyses, however, have revealed that the Arabidopsis genome encodes numerous proteins with recognizable PI-binding domains (van Leeuwen et al., 2004), including: the PI 3-kinase VPS34 (At1g60490), the PI 4-kinases α1 (At1g49340) and α2 (At1g51040), the 3-phosphoinositide-dependent protein kinase 1 (PDK1; At3g10540), the dynamin-related proteins 2A (DRP2A; At1g10290) and B (DRP2B; At1g59610) and the Rho-of-plants GTPase-activating proteins 1-3 (ROP-GAPs 1-3; At4g24580, At5g12150 and At5g19390), all of which contain a PH domain; FAB1A (At4g33240) and FAB1B (At3g14270), which contain a FYVE domain; and the sorting nexin-like protein 2 (SNX2B; At5g07120), which contains a PX domain (van Leeuwen et al., 2004). Interestingly, there are several instances in which proteins contain more than one PI-binding domain, for example the phospholipases D ζ1 (PLDζ1; At3g16785-90) and ζ1 (PLDζ2; At3g05630), which contain both a PH domain and a PX domain (van Leeuwen et al., 2004). It appears likely that the number of proteins with PI-binding domains is still underestimated, as PI-binding domains may form transiently upon dynamic interaction of protein partners, thus escaping detection via sequence annotation (van Rossum et al., 2005). Furthermore, despite the seeming abundance of PI-modulins, the direct experimental identification of PI-binding proteins has so far not yielded a substantial amount of data. A notable exception is the identification of numerous candidate binding partners for 3-phosphorylated PIs from Arabidopsis, which were determined by agarose-phosphatidylinositol-phosphate affinity chromatography and subsequent mass spectrometric analysis (Oxley et al., 2013).

Intact PIs can also influence a membrane directly by modifying its biophysical properties, inducing membrane curvature or stabilizing curved membrane areas (Lundbaek et al., 2010). Thus, PIs can both exert an intrinsic influence on the membrane and mediate the recruitment of proteins that stabilize certain membrane structures, although the relative contributions of these individual effects remain unclear.

Besides their roles as intact lipids, PIs can serve as precursors for PLC-mediated cleavage into diacylglycerol (DAG) and soluble IPPs (Gillaspy, 2013). In mammals, DAG acts as a potent activator of protein kinase C (Amadio et al., 2006), but this enzyme is missing in plants (Mueller-Roeber and Pical, 2002) and a signaling function for plant DAG has not been reported. Instead, DAG in plants is rapidly phosphorylated to phosphatidic acid (PtdOH), which has been shown to have a number of regulatory functions (Meijer and Munnik, 2003). PtdOH can be further phosphorylated to diacylglycerol pyrophosphate (DGPP), a lipid of unknown relevance (van Schooten et al., 2006). Soluble IPPs, such as inositol hexakisphosphate (InsP₆) or inositol pentakisphosphate (InsP₅), are present within the crystal structures of receptor proteins for auxin (Tan et al., 2007) or jasmonic acid (Sheard et al., 2010). However, the role of these IPPs in the perception of auxin or jasmonic acid as structural or possibly functional co-factors is currently unclear. It has been reported that Arabidopsis ipk1-1 mutants with decreased levels of InsP₆ and increased levels of InsP₅ display an enhanced wounding response and enhanced resistance to herbivory (Mosblech et al., 2011). More recently, other IPPs, such as the pyrophosphorylated IPPs (PP-IPPs) InsP₇ or InsP₈, have also been implicated in modulating the function of the jasmonic acid receptor (Laha et al., 2015), possibly suggesting a role for IPPs in phytohormone perception and the control of hormone-induced gene expression patterns. Other IPPs, such as inositol 1,4,5-trisphosphate (InsP₃), are also implicated in mediating Ca²⁺ responses, although an InsP₃-sensitive Ca²⁺ channel has not been reported in plants (Krinke et al., 2007a).

From the current data it is evident that plant PIs exert a number of different functions by interacting with a variety of binding partners. However, it is also clear that not all cellular processes that can in principle be controlled by PIs will manifest whenever PIs are formed. In fact, the physiological consequences of enhanced PI formation tend to be rather limited and highly specific, raising the question of how the specificity of PI functions is achieved (Heilmann and Heilmann, 2013). A key piece of information for understanding PI function is that these lipids do not display uniform distribution within membranes (Ischebeck et al., 2008, 2011, 2013; Kost et al., 1999; Kusano et al., 2008; Sousa et al., 2008; Tejos et al., 2014; van Leeuwen et al., 2007; Vermeer et al., 2009; Vincent et al., 2005). Instead, PIs are enriched in certain membrane domains, and it is likely that such microdomains play a part in mediating the specific effects of PIs (Fig. 4A). Plants form particular molecular species of PIs under certain conditions (e.g. when challenged by salt) that are characterized by an increased degree of unsaturation in their associated fatty acids (König et al., 2007). The degree to which lipids are unsaturated is known to influence their lateral mobility within the plane of the membrane and to guide their association with particular planar or curved membrane domains (Cho et al., 2006; Mukherjee et al., 1999). The degree of unsaturation of PI-associated fatty acids might, therefore, be an as yet underestimated factor governing the lateral mobility of PIs within membranes, with possible implications for their association with certain partner proteins and for defining the specificity of PI effects. The lateral distribution of PIs within a membrane is also likely to be influenced by other membrane lipids, such as sphingolipids and sterols (Furt et al., 2010).

An additional aspect contributing to the specificity of PI effects is the concept of coincidence detection (Carlton and Cullen, 2005). Although a number of plant proteins have been demonstrated to bind PIs, as discussed above, it is known from other model systems that there can be additional factors guiding the effects of PIs (Balla, 2005; Várnai et al., 2005). Regulation by coincidence detection (Fig. 4B) suggests that PI binding is necessary for the regulation of certain processes but might not be sufficient by itself, and other factors must coincide for an effect to manifest. Thus, in addition to the binding of a PI, a membrane-associated protein partner, for example, must also be bound and only membrane areas containing both recruitment factors will be targeted by modulins that require coincidence detection to function. An interesting aspect of coincidence detection involves the possible role of proteins that potentiate PI functionality by facilitating their interaction with certain target proteins. For instance, recent evidence in Arabidopsis indicates that Sec14-like lipid transfer proteins aid PI-protein interaction in a specific manner (Bankaitis et al., 2010; Ghosh et al., 2015; Schaaf et al., 2008; Vincent et al., 2005). As a consequence, membrane areas that contain Sec14-like proteins will exhibit potentiated PI effects, possibly similar to the effects of localized PI overproduction.

In recent years, PIs have emerged as central regulators of plant function and development, and a number of important and severe plant phenotypes have been reported when plant PI metabolism is perturbed (Carland and Nelson, 2009, 2004; Chen et al., 2008; Ercetin et al., 2008; Golani et al., 2013; Gunesekera et al., 2007; Ischebeck et al., 2008, 2011, 2013; Kaye et al., 2011; Kusano et al., 2008; Löfke et al., 2008; Mei et al., 2012; Novakova et al., 2014; Parker et al., 2000; Preuss et al., 2006; Rodriguez-Villalon et al., 2015; Sousa et al., 2008; Stenzel et al., 2008; Tejos et al., 2014; Wang et al., 2012; Williams et al., 2005; Zhong et al., 2004, 2005; Zhong and Ye, 2004). The perturbation of PI signaling components in Arabidopsis, for example, gives rise to a variety of phenotypes affecting multiple tissue types and aspects of plant development (summarized in Table 1). A number of studies utilizing model systems other than Arabidopsis have also highlighted a role for PIs in plant development and function. For instance, the overexpression of PI-synthase in maize results in altered drought tolerance (Liu et al., 2013). Work on petunia, tobacco and the moss Physcomitrella patens supports an important role for PIs in the control of polar tip growth (Dowd et al., 2006; Helling et al., 2006; Saavedra et al., 2011, 2015). Although some of the reported phenotypes are severe, the molecular mechanisms underlying PI functions in plants remain mostly unclear.

In the absence of detailed information on the molecular targets of PIs in plants, the interpretation of reported phenotypes poses a challenge. Based on the available data, the phenotypes can be categorized according to the defects observed, enabling a first approximation of how the elements of the plant PI network might be interlinked at the cellular level (Fig. 5). For example, phenotypes resulting from the perturbation of PI-synthases (Löfke et al., 2008), PI4Kβ1/PI4Kβ2 (Preuss et al., 2006), PIP5K1-PIP5K6 (Ischebeck et al., 2008, 2013; Kusano et al., 2008; Sousa et al., 2008; Stenzel et al., 2008; Tejos et al., 2014; Zhao et al., 2010) or of SAC1 (Zhong et al., 2005) or SAC7 (Thole and Nielsen, 2008) are all related to functional changes in membrane trafficking between the Golgi and the plasma membrane. Based on the subcellular localization of the individual enzymes, it appears that an increasing degree of phosphorylation [PtdIns→PtdIns4P→PtdIns(4,5)P₂] is associated with the progressive delivery or retrieval of vesicles to and from the plasma membrane. Another category of phenotypes indicates the control of cytoskeletal dynamics by PIs. For instance, the perturbation of PIP5K2 (Stenzel et al., 2012) or PIP5K10 and PIP5K11 (Ischebeck et al., 2011) results in altered actin dynamics and the loss of cellular polarity. Yet a different category of phenotypes indicates that 3-phosphorylated PIs, which are perturbed by altered levels of VPS34 (Lee et al., 2008a,b), Fab1-like proteins (Bak et al., 2013) or SAC2-SAC5 (Novakova et al., 2014), might be involved in endomembrane trafficking and the control of vacuolar functions. The overall scheme suggests that PIs are involved in controlling the central machinery for membrane trafficking and protein sorting in plants. As a consequence, PIs have an impact on several key cellular processes, such as the establishment of cell polarity (Ischebeck et al., 2008, 2011, 2013; Kusano et al., 2008; Mei et al., 2012; Preuss et al., 2006; Sousa et al., 2008; Stenzel et al., 2008; Tejos et al., 2014) and cell wall deposition (Krishnamoorthy et al., 2014; Zhong et al., 2004, 2005). These processes influence embryo development, vascular development and, ultimately, plant growth.

In addition to their cytosolic effects on trafficking and membrane recruitment, emerging evidence indicates that PIs may also have a direct or indirect influence on nuclear function in plants. PI reporters (see Box 1) have been observed in the nuclei of plant cells (Mishkind et al., 2009), as have fluorescently tagged variants of enzymes involved in PI metabolism (Ischebeck et al., 2011, 2013; Lou et al., 2007; Tejos et al., 2014). However, so far it is unclear whether these observations have biological relevance or if they simply reflect fluorescence patterns ensuing after fluorescent fusion proteins have been degraded. The expression of human PI4P 5-kinase in plant cells resulted in changes to the cell cycle control machinery and histone modifications (Dieck et al., 2012b), but it is unclear whether this finding has relevance for endogenous plant PI-based processes. Further information comes from the analysis of factors involved in the control of stem cell populations in plant meristems. For instance, the protein phosphatase POLTERGEIST, an enzyme that functionally links plasma membrane perception of CLAVATA peptides and nuclear regulation, has been shown to bind PtdIns4P (Gagne and Clark, 2010). This binding to PtdIns4P results in the activation of phosphatase activity in vitro, suggesting that PtdIns4P might be involved in the signaling pathways controlling stem cell maintenance. So far, there is only very limited information about the effects of PIs on nuclear function in plants, and a role for PIs in plant nuclei (Dieck et al., 2012a), as has been proposed for PIs in animal cells, is currently unclear.

The perturbation of PI metabolism in plants results in severe phenotypes, highlighting the fact that PIs play central roles in plant development and plant adaptation to the environment. The profound influences of PIs on diverse cellular processes indicate that the plant PI network is an important field for future studies. Our current understanding of the plant PI network is still limited and the precise molecular targets of PI action in plants are largely unknown. A massive amount of data has already accumulated on the plant PI system, but to date it seems that not all dots have been connected and the available information has not been used to the fullest extent. For instance, numerous proteins have been proposed to contain PI-binding domains, making these proteins prime candidates for regulation by PIs (van Leeuwen et al., 2004). However, only a minority of these potential PI-modulins have been experimentally tested for PI binding by in vitro lipid overlay assays (Munnik and Wierzchowiecka, 2013) or liposome binding tests (Julkowska et al., 2013), leaving the field open for further scrutiny. Whereas many of the potential PI-modulins are proteins of unknown function, others are well annotated, linking potential PI binding to relevant cellular processes. In combination with the review of PI-dependent plant phenotypes related to membrane trafficking or cytoskeletal control, this information might indicate future avenues of research to elucidate how and when PIs are formed and what their respective modulins in each context might be. The study of the molecular mechanisms by which PIs act will require a combination of complementary in vitro and in vivo data to demonstrate a precise mode of action. In particular, information on the specificity and affinity of PI-protein binding is currently missing. Moreover, the study of PI-protein binding should expand to include other lipid classes known to influence membrane biophysics, such as sphingolipids and sterols, because PIs are unlikely to act individually in a cell but rather in the context of membrane areas defined by a mixture of different lipids (van Meer et al., 2008). In this sense, it will be a major goal of future studies to determine the coincident properties of mixed membrane microdomains with regard to protein recruitment and binding. On the in vivo side, plant researchers have started using fluorescent probes for the detection of membrane lipid-ordered domains (Frescatada-Rosa et al., 2014; Gerbeau-Pissot et al., 2016), which might prove a powerful tool to link in vitro and in vivo studies. Considering the current status of sequence information and the range of available in vitro and in vivo methods, I am confident that the coming years will bring substantial advances in the field of plant PI signaling.

The author apologizes to all authors whose relevant contributions might not have been cited owing to space restrictions and wishes to thank Dr Mareike Heilmann and Dr Praveen Krishnamoorthy (both Martin-Luther-University Halle-Wittenberg) for helpful discussion.