Phosphoinositides are a family of phospholipid messenger molecules that control various aspects of cell biology in part by interacting with and regulating downstream protein partners. Importantly, phosphoinositides are present in the nucleus. They form part of the nuclear envelope and are present within the nucleus in nuclear speckles, intra nuclear chromatin domains, the nuclear matrix and in chromatin. What their exact role is within these compartments is not completely clear, but the identification of nuclear specific proteins that contain phosphoinositide interaction domains suggest that they are important regulators of DNA topology, chromatin conformation and RNA maturation and export. The plant homeo domain (PHD) finger is a phosphoinositide binding motif that is largely present in nuclear proteins that regulate chromatin conformation. In the present study I outline how changes in the levels of the nuclear phosphoinositide PtdIns5P impact on muscle cell differentiation through the PHD finger of TAF3 (TAF, TATA box binding protein (TBP)-associated factor), which is a core component of a number of different basal transcription complexes.
Life changing days
Certain events change the rest of your days! In science new discoveries, your own if you are lucky, or more likely those from other scientists, enable greater understanding that helps to guide the following steps, whereas in life it can be a chance meeting, an interview or a new child. Sitting in a traffic jam humming La bella tartaruga (The beautiful turtle), your two year old daughter's favourite song, means things have changed. My struggles with the turtle started in 1987 when I was lucky enough to come for an interview to work as a postdoctoral researcher with Robin Irvine. The then guitar wielding (now a lute), turtle loving renaissance scientist with a penchant for a quick jibe or gag (why did god choose phosphates? Because Inositol. ARF ARF) turned out to be the best thing in science that could have happened to me. I am sure that is a sentiment held by many of the other young scientists that Robin selflessly mentored, encouraged and promoted. Life changing? Sure was!
So what is so special about the turtle? The turtle is a 3D mnemonic devised by Agranoff that represents the most stable chair conformation of myo-inositol with one axial and five equatorial hydroxy groups . The head of the turtle represents the axial hydroxy with the other limbs and tail the equatorial hydroxy groups (Figure 1). Conventionally the turtle is considered to be ‘right flippered’ and this hydroxyl group is labelled as 1. The counter clockwise D numbering then follows making the axial ‘turtle head’ number 2. There are 63 possible phosphomonoesters of myo-inositol with a further 15 being found with pyrophosphorylations at the 1, 3 or 5 positions. Leashing the turtle with a diacylglycerol (DAG) moiety through a phosphodiester bond on the right flipper (number 1) generates phosphatidylinositol (PtdIns) the precursor to seven phosphorylated derivatives (phosphoinositides) (Figure 1). These can be interconverted by a series of lipid kinases and phosphatases . Unlike the myo-inositol phosphates which are water soluble, PtdIns and its phosphorylated derivatives with their hydrophobic tails much prefer to swim within a membrane. At the heart of this series of turtles is PtdIns(4,5)P2. In response to various types of receptor activation PtdIns(4,5)P2 can be phosphorylated on the 3′ position to generate another new second messenger PtdIns(3,4,5)P3  or hydrolysed by a phospholipase C (PLC) to generate Ins(1,4,5)P3 and DAG. Ins(1,4,5)P3 acts as second messenger to regulate calcium release by binding to its receptor in the endoplasmic reticulum . DAG remains in the membrane and regulates a number of proteins, such as protein kinase C (PKC) by interacting with them and recruiting them to the membrane .
Nuclear phosphoinositides are dynamically regulated during cell differentiation
Turtle's swimming in the nucleus
The nucleus is a double membrane bounded organelle with the outer membrane continuous with the endoplasmic reticulum. Within nuclei are discrete domains called speckles that vary in size with the largest being the nucleolus. These speckles domains have specialized functions, rather like organelles within a cell, however unlike organelles they do not have limiting membranes that separate them from the rest of the nucleus . They do however have phosphoinositides specifically localized to them. Early studies demonstrated the presence of PtdIns in nuclei and subsequent studies demonstrated that given radiolabelled ATP, isolated intact nuclei or isolated nuclear envelopes were able to synthesize labelled phosphatidic acid (PtdOH), PtdIns4P and PtdIns(4,5)P2 . Four years later Lucio Cocco in collaboration with Robin Irvine discovered that the labelling of nuclear phosphoinositides was different in nuclei isolated from control cells compared with differentiated cells (Figure 1B) . This paper stimulated the idea that nuclear phosphoinositides might function as a signalling system that responds to stimuli and transduces signals to impact on nuclear functions. While this labelling strategy is simple and useful it does not address how much lipid is actually present in the nucleus or what changes have occurred to deliver the differences in nuclear phosphoinositide labelling. Measurement of the mass of nuclear PtdIns(4,5)P2 suggested that 10% of the total cellular PtdIns(4,5)P2 is present in the nucleus, but that its mass level does not change under the conditions that change its labelling . A simple hypothesis is that there are multiple pools of nuclear PtdIns(4,5)P2 but that only a small pool, which is rapidly labelled with 32P-ATP changes during differentiation. In fact approximately 30% of total nuclear PtdIns(4,5)P2 is present in the nuclear envelope with the rest being present in different nuclear compartments including nuclear speckle-like structures, the nucleolus, the nuclear matrix and chromatin [10–13]. We now know that within the nucleus and within different nuclear structures there are the enzymes to generate PtdIns(4,5)P2 from PtdIns and to degrade it by dephosphorylation or by hydrolysis [10,13]. Changes in nuclear phosphoinositides have been observed in response to differentiation, cell cycle progression , growth factor and stress signalling [15,16] and in vivo during regeneration after partial hepatectomy . How exactly phosphoinositides are regulated or how they are presented within the nucleus is not clear. Membrane bilayer structures have not been observed within the nucleus and so we presume that phosphoinositides are presented either in micelles or in proteolipid structures. Some orphan nuclear receptors are able to bind the fatty acid tails of phosphoinositides and present them for further modifications. Binding of phosphoinositide to the orphan nuclear receptors changes both the ability of: (1) phosphoinositide modulating enzymes to use them as substrates and (2) the orphan receptors to act as transcriptional regulators [18,19]. As PtdIns(4,5)P2 has been found to be localized to many different types of speckle structures we suggest that there are likely to be multiple ways to maintain these different pools of PtdIns(4,5)P2. For example syntenin2 contains as PDZ (post synaptic density protein (PSD95), drosophila disc large tumor suppressor (Dlg1), and zonula occludens-1 protein (zo-1)) domain which can interact with PtdIns(4,5)P2 and siRNAi-mediated depletion results in a loss of PtdIns(4,5)P2 at nuclear speckles . How phosphoinositides enter the nucleus is also unclear. We presume that PtdIns is the lipid that is brought to the nucleus and then modified to generate the other phosphoinositides. PtdIns could track around from the endoplasmic reticulum into the inner nuclear membrane and then be grabbed by nuclear proteins such as the orphan receptors. Alternatively there may be transport mechanisms that bring PtdIns directly into the nucleus. Protein lipid transfer proteins might be good candidates to mediate this transport .
Given that nuclei can generate PtdIns(4,5)P2 does it act as a messenger lipid? In response to IGF1 stimulation the mass of PtdIns(4,5)P2 is decreased within the nucleus whereas the mass of DAG is increased . The simplest interpretation of the data is that IGF1 stimulates a nuclear PLC to hydrolyse PtdIns(4,5)P2 [22,23]. Further studies showed that MAP kinase, activated by IGF1 signalling, phosphorylates and activates PLCβ1 in the nucleus . The presence of PLCβ1 in the nucleus is without doubt, however, the extent to which it is exclusively nuclear has been the subject of some debate . Other PLC isoforms have also been shown to be present in the nucleus, although their role in this compartment is not completely clear . The increase in nuclear DAG sequesters and activates PKC in the nucleus leading to the phosphorylation of an array of nuclear protein substrates . Ins(1,4,5)P3, the other product of PLC-mediated hydrolysis of PtdIns(4,5)P2 can bind to its cognate receptor which is present on the inner nuclear membrane. Ins(1,4,5)P3 binding is thought to release calcium directly into the nucleus, which is essential for transcriptional regulation downstream of calcium-dependent calmodulin kinase . Ins(1,4,5)P3 can also be phosphorylated to generate highly phosphorylated inositol derivatives some of which regulate chromatin remodelling, histone modifications, transcription and mRNA export [29,30]. Many of the enzymes that generate these highly phosphorylated inositols are as would be expected localized in the nucleus . PtdIns(4,5)P2 can also be converted to PtdIns(3,4,5)P3 by the action of a nuclear PI3K in response to DNA damage and prevention of the production of nuclear PtdIns(3,4,5)P3 appears to attenuate downstream DNA damage repair pathways . It should be stated that the above description of the regulation and function of nuclear phosphoinositides is an amalgamation of data generated from many different cell types and model organisms. Data coupling upstream stimuli to changes in nuclear phosphoinositides and downstream nuclear outputs is at best weak and incomplete. What is required is a concerted effort to define completely different nuclear phosphoinositide pathways from beginning to end. Still given the above caveat it is clear that phosphoinositides are present in the nucleus, change in response to various stimuli and impact on important nuclear functions, many of which are disrupted in human disease.
Knowing your turtle's friends illuminates the pathway
Localizing proteins to a membrane increases their effective concentration. If binding partners or enzymes and substrates are also membrane bound then this co-localization strongly increases the chances of two proteins meeting each other and engaging in productive interactions. Thus translocation to the membrane is an activation step even in the absence of changing other parameters such as conformation. Nearly all of the members of the phosphoinositides family induce target recruitment as a mechanism to transduce upstream signals into downstream pathway activation. Recruitment of proteins to phosphoinositides is mediated by a series of specific protein domains which include plekstrin homology domains (PH), FYVE domains and PX domains . Productive interactions with phosphoinositides can also be mediated by lysine/arginine-rich patches . In all it is thought that over a thousand different proteins within the cell are able to interact with phosphoinositides. Clearly defining who your turtle interacts with indicates how, when and where it initiates downstream signalling. Most of the known phosphoinositide interacting domains are present in proteins that are enriched in the cytoplasm, with the exception of lysine/arginine-rich patches that are often present in nuclear proteins. In a proteomic study, a strong enrichment of lysine/arginine-rich patches in nuclear proteins that interact with PtdIns(4,5)P2 was observed . These included proteins involved in DNA topology, transcription and in mRNA splicing. Previous studies also showed that PtdIns(4,5)P2 interacts with a lysine-rich region of histone H1. Histone H1 acts as a nucleosomal linker, compacts nucleosomes and inhibits RNA polymerase-mediated transcription in in vitro transcription extracts . The addition of PtdIns(4,5)P2 attenuates this inhibition thereby increasing transcription. Whether this occurs in vivo to impact transcriptional output is not clear. Local remodelling of nucleosomes is also important in transcription and PtdIns(4,5)P2 appears to regulate chromatin recruitment of the nucleosome remodelling BAF (BRG1- or HRBM-associated factors) complex . PtdIns(4,5)P2 also regulates splicing and selective mRNA polyadenylation . However, in most of the cases cited above little is known about how exactly phosphoinositides interact with and regulate their target protein. The discovery that nuclear phosphoinositides interact with and regulate proteins that contain plant homeo domain (PHD) fingers changed everything .
In order to understand the relevance of PHD fingers and turtles, we should perhaps paddle into the murky waters of chromatin regulation. DNA is wrapped around a histone octamer (two copies each of four subunits, histone H2a,H2b, H3 and H4) to form the nucleosome. Histone tails are excluded from the inner face of the nucleosome, are highly accessible and can be modified by ubiquitination, phosphorylation, methylation and acetylation among others . The combination of nucleosomal DNA, DNA-binding proteins and variable histone tail modifications are thought to generate different chromatin structures that are associated with variable transcriptional outputs. For example tri-methylation of histone H3 at lysine 9 (h3K9me3) is associated with transcriptionally inactive genes, whereas trimethylation of H3K4 is associated with increased gene transcription (for an in depth review see ). Local chromatin accessibility is important for transcription and appears to be regulated by both DNA methylation and histone tail modifications. Histone tail modifications are recognized by ‘reader’ type proteins which have domains that interact with the specific modification. For example PHD fingers and TUDOR (from the tudor protein in Drosophila) domains can interact with histones tails that are methylated . These reader proteins are usually part of or can recruit large protein complexes that act to repress or activate downstream transcription or other nuclear functions (interpret histone tail modification). For example TAF3 (TAF, TATA box binding protein (TBP)-associated factor) is a core component of the basal transcription complex (TFIID) that binds to promoters to help position RNA polymerase correctly before the start of transcription. TAF3 contains a PHD finger that binds to histone H3 methylated at lysine 4 (H3K4me3)  and enhances the ability of TFIID to regulate transcription  (Figure 2).
Depletion of PIP4K2B stimulates myogenic differentiation
The PHD finger is a cross braced zinc finger that is present predominantly in nuclear proteins which regulate chromatin. There are 218 human proteins that are annotated to contain a PHD finger. These proteins function as histone adaptors, histone tail writers (deposit histone tail modifications such as methylases), erasers (remove histone tail modifications such as demethylases) and readers. PHD fingers can interact with non-methylated and methylated lysine residues and in rare cases with acetylated lysine residues of histone tails . Complex histone tail modifications can also modulate how PHD proteins bind to histones. For example the interaction of TAF3 with H3K4me3 is strongly attenuated if the adjacent threonine is also phosphorylated. Threonine phosphorylation occurs during mitosis and is important to inhibit transcription during this phase of the cell cycle . To add to the complexity many proteins contain more than one PHD finger with each finger binding differently modified histone tails. For example, in the demethylase KDM5A (lysine (K)-specific demethylase 5A) one of the PHD fingers recognizes H3K4me3, whereas the other recognizes non-methylated histone tails. Critically PHD fingers also interact with phosphoinositides. In a recent study 17 out of 32 different PHD fingers tested interacted with phosphoinositides . Phosphoinositide interaction was not associated with any other function of PHD fingers such as H3K4me3 interaction and was observed with PHD fingers derived from proteins that function as writers, erasers and readers . These data suggest that nuclear Phosphoinositides are likely to play a significant role in regulating chromatin conformation and transcriptional output.
Nuclear turtles and TAF3 promote muscle differentiation
PIP4K2B phosphorylates PtdIns5P to generate PtdIns(4,5)P2 and in doing so regulates the levels of both lipids [47,48]. PIP4K2B resides in both the nucleus and the cytoplasm, although in some cells it appears to be predominantly nuclear [49,50]. PIP4K2B knockout is synthetically lethal with p53 knockout  and in tumour cells reducing PIP4K activity induces cell death [51,52]. PIP4K2B knockout in mice prevents diet induced obesity, and increases insulin action in muscle enabling better systemic glucose control as mice age . Interestingly PIP4K2B is highly expressed in muscle tissue. This prompted us to investigate the role of PIP4K2B in muscle differentiation. In the C2C12 cell model, growth factor deprivation induces cells to exit the cell cycle, differentiate into muscle cells and fuse to form rudimentary myotubes. Growing C2C12 cells express the master myogenic transcription factor MYOD and during differentiation, up-regulate myogenin (MYOG), a transcription factor that drives irreversible myogenic differentiation. In vivo, quiescent satellite muscle stem cells situated beneath the basolateral lamina of muscle are important for muscle repair and undergo activation and replication to generate an actively proliferating progenitor cell in response to muscle damage. These progenitor myoblast cells migrate and fuse with existing myotubes. Quiescent satellite cells do not express MYOD but up-regulate it during activation. MYOD expressing cells can then either return to a quiescent stem like cell (high PAX7 expression) or up-regulate myogenin to differentiate irreversibly. C2C12 cells therefore functionally resemble the activated satellite myoblast cell population (Figure 2A). Knockdown of PIP4K2B in C2C12 cells strongly enhances myogenic differentiation increasing the expression of both early (MYOG) and late myogenic genes (myosin heavy chain MYH). The increase in the expression of these genes drives enhanced myotube formation (Figure 2B). Increased expression of MYOG and MYH is a consequence of increased transcription. PIP4K2B is present within nuclear speckles in control myoblasts but as cells differentiate, PIP4K2B becomes excluded from the nucleus. This translocation of PIP4K2B out of the nucleus is the first demonstration that nuclear localization of PIP4K2B is regulated and is correlated with an increase in the levels of its substrate PtdIns5P in the nucleus. Given that knockdown of PIP4K2B increases transcription of early and late myogenic genes we searched for a nuclear PtdIns5P-dependent downstream target that directly regulates myogenic transcription. The basal transcription complex (TFIID) component TAF3 contains a PHD finger that we found interacts strongly with phosphoinositides. Surprisingly during myogenic differentiation many TFIID components, including the core TATA-binding protein TBP1, are down regulated and replaced by a simplified complex consisting of just TAF3 and TRF3, a TBP-related protein (Figure 2) . TAF3 is essential for myogenic differentiation in C2C12 cells and is a potential downstream target for increased nuclear PtdIns5P. In order to demonstrate a direct role for nuclear phosphoinositides in regulating transcription through the PHD finger of TAF3, we investigated how PtdIns5P interacts with TAF3. Although no structural data are available to show how PtdIns5P might interact with a PHD finger, alignment of the seventeen PHD fingers that interact with phosphoinositides suggested the importance of a lysine/arginine polybasic region downstream of the PHD finger. The polybasic region is not sufficient for PtdIns5P binding and requires intact Zinc binding of the PHD finger. Mutation of lysine residues within the PBR eliminated phosphoinositide interaction but maintained interaction with H3K4me3 (KK-TAF3) suggesting that the overall structure of the PHD finger was maintained in this mutant. This mutant was then used to investigate the role of PtdIns5P binding to TAF3 in the regulation of muscle specific gene transcription in vivo. C2C12 cells depleted of endogenous TAF3 were reconstituted with either wild type or mutant KK-TAF3 and then PIP4K2B was knocked down. The data clearly demonstrate that the expression of a subset of PIP4K2B-regulated myogenic genes requires the intact phosphoinositide interacting PHD finger of TAF3 . Similar results were observed for PtdIns5P-mediated regulation of gene expression by the PHD finger containing protein ING2 (inhibitor of growth protein 2) under conditions of stress .
To demonstrate that phosphoinositide interaction with TAF3 is important in an organismal context, we turned to zebrafish. The phosphoinositide interacting region is completely conserved in zebrafish TAF3. We therefore depleted endogenous zebrafish TAF3 and reconstituted with either wild-type TAF3 or the KK-TAF3mutant. TAF3 depletion led to severe disruption in muscle architecture, which could be rescued by the overexpression of wild type TAF3, but importantly not by the expression of KK-TAF3 .
Nuclear turtles: future perspectives
The possibility that nuclear PtdIns5P or other nuclear phosphoinositides impact on muscle cell differentiation is exciting from both an aging and disease perspective. During aging the satellite cell compartment becomes progressively worse at being able to regenerate and maintain functional muscle. Interestingly aging induced decrease in satellite cell activity can be reversed by changing the extracellular environment  as well as by inhibiting the accumulation of the senescence gene p16INK4A . These data suggest that intervention therapies could be developed to maintain muscle function during aging. Defining exactly how a small molecule like PtdIns5P could allosterically regulate TAF3 to increase muscle specific gene expression might enable the development of a small molecule regulator to this site. From a biology perspective we do not understand how nuclear PtdIns5P selectively increases the expression of MYOG and MYH through its interaction with the PHD finger of TAF3 but does not change the expression of MYOD even through all three expressions are TAF3 dependent. This suggests that simple activation of TAF3 by interaction with PtdIns5P is unlikely to account for how nuclear PtdIns5P functions (Figure 3A). More likely, nuclear PtdIns5P could constitute a genome organization platform such that only genes within or close to the platform are regulated by both phosphoinositides and TAF3 (Figure 3B). If this is the case nuclear phosphoinositides might be responsible for the formation of distinct genomic territories that impact on transcription and cell fate decisions.
Models to explain how nuclear PtdIns5P might regulate TAF3
I thank Robin Irvine for giving me the opportunity to work with him and to enter into the field of phosphoinositide signalling. I also thank Yvette Stijf-Bultsma and Zahid Shah and all other past and present members of my laboratories. Sand Turtles were made and generously provided by Sofia Mili Divecha.
This work was supprted by The Dutch Cancer Society, and Cancer Research UK; and the University of Southampton and the European Union (Marie curie experienced research fellowship) [grant number PIEF-GA-2013-625639].
Signalling 2015: Cellular Functions of Phosphoinositides and Inositol Phosphates: Held at Robinson College, University of Cambridge, Cambridge, U.K., 1–4 September 2015.