To suggest and develop intelligent strategies to comprehend the regulation of organelle formation, a deeper mechanistic interpretation requires more than just the involvement of proteins. Our approaches link the formation of endomembranes with both signalling and membrane physical properties. Hitherto, membrane morphology, local physical structure and signalling have not been well integrated. Our studies derive from a cross-disciplinary approach undertaken to determine the molecular mechanisms of nuclear envelope assembly in echinoderm and mammalian cells. Our findings have led to the demonstration of a direct role for phosphoinositides and their derivatives in nuclear membrane formation. We have shown that phosphoinositides and their derivatives, as well as acting as second messengers, are modulators of membrane morphology, and their modifying enzymes regulate nuclear envelope formation. In addition, we have shown that echinoderm eggs can be exploited as a milieu to directly study the roles of phospholipids in maintaining organelle shape. The use of the echinoderm egg is a significant step forward in obtaining direct information about membrane physical properties in situ rather than using simpler models which do not provide a complete mechanistic insight into the role of phospholipids in membrane dynamics.
In recent years, cell and developmental biologists have taken up the challenge of understanding in more detail the mechanisms behind the formation of various organismal and cellular membrane compartments. Most of this re-interest has resulted from the ability to apply biophysical tools, in situ or in vivo, to investigate the involvement of various classes of molecules in the formation and dynamics of endomembrane compartments [1–4]. The importance of these studies derives from the realization that the localized morphology of a membrane compartment affects the properties of signalling reactions. In fact, changes in membrane morphology and signalling reactions mutually affect one another [2,5]. Currently, the common understanding among researchers is that membranes are not just static scaffolds or barriers, but that their morphology and dynamics actually play a crucial role in signalling reactions [5,6].
However, grounds of controversy arise from deciding which classes of molecules are responsible for maintaining the proper morphology of subcellular compartments [1,4,5,7]. Most groups who have investigated the morphological properties of endomembranes have emphasized the role of membrane proteins or cytoskeletal reorganization in reshaping or formation of subcellular compartments [4,5,7,8]. In most recent papers, only brief mention of lipids has been made, mostly as general building blocks of membranes bilayers. Probably this understatement is due to the difficulty in manipulating the composition of phospholipids locally and acutely. Our research groups have overcome this difficulty by exploiting the rapalogue dimerization device [1,2]. Localized acute interventions demonstrate directly the active role of phospholipids in regulating and maintaining the correct morphology of endomembranes.
The aim of the present article is to summarize our published and unpublished data showing that proteins do not solely drive morphology and dynamics (fusion) of subcellular compartments which should instead be considered proteolipid events. We reiterate the important roles of phospholipids, specifically phosphoinositides and their derivatives, in the maintenance of the proper morphology of the endoplasmic reticulum (ER) and its subcompartment, the nuclear envelope (NE). We propose the concept of the principle of duality in phospholipids, as second messengers and regulators of endomembrane morphology.
Intervention of phosphoinositides as local modulators of membrane morphology and dynamics (fusion)
Cytoplasmic membranes of eukaryotic cells undergo fission and fusion, thus reorganizing cytoplasmic architecture and facilitating import, export and transport of various cargoes within the cell. Much of our knowledge of membrane fusion derives from genetic approaches in yeast, protein-free synthetic membrane fusion systems and cell-free systems for identifying various components of the fusion machinery [9,10], including soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptors (SNAREs) [11,12].
Cellular recognition of membrane identity via its local structure and organization is important to maintain appropriate function. The role of lipids in membrane fusion has received theoretical and experimental attention especially concerning effects of phospholipids exhibiting negative curvature [13–17].
Recently researchers have reformulated ‘SNARE models’ by studying the involvement of phosphoinositides and their derivatives such as diacylglycerol (DAG) in fusion [6,18–20]. Membrane lipids alter the energetics of fusion . Lipids of negative curvature such as DAG, cholesterol and phosphatidylethanolamine tend to disrupt bilayers and induce fusion [15,21–23]. DAG plays a role in membrane fusion and curvature during yeast vacuole formation . The involvement of phosphoinositides, specifically PtdInsP2 and PtdInsP, has also been shown in vacuolar fusion in yeast and in Drosophila photoreceptors [24–26]. Nevertheless, a detailed molecular mechanism of how higher phosphoinositides are involved in fusion events has not been elucidated.
Our experimental approaches and model systems have demonstrated the effects of polyphosphoinositdes on membrane dynamics and offer a physical and mechanistic explanation of how the phosphoinositides and their derivatives may remodel membrane morphology to promote fusion.
We have used the regulation of NE assembly as an example of membrane fusion. The NE is disassembled and reassembled each mitosis and correct re-formation is central to proper cell functioning. Accurate reformation of the NE is necessary for gene expression for all cells that undergo open mitosis.
Our groups have pioneered investigations on the role of lipids in NE formation in simpler model organisms, and we have now extended these investigations to mammalian cells. Briefly, we showed an important role for phospholipids in the regulation of NE assembly using formation of the male pronucleus in echinoderm eggs. This process is comparable with NE assembly in differentiated cells. Both in vivo and in vitro, a non-ER membrane fraction (MV1) enriched in phospholipase Cγ (PLCγ) and Src kinase with very high levels of phosphoinositides (60 mol%) is critical for NE formation .
This finding led to the hypothesis that the phospho-inositides may not only serve as transient signals, but also can affect membrane structural dynamics, thus emphasizing roles as both signalling and membrane-modifying molecules.
Structural and dynamic implications of phosphoinositides in membrane fusion
To better comprehend the physical effects of phosphoinositides on membrane fusion, we initiated studies to characterize their polymorphisms. To define the morphology and dynamics of polyphosphoinositides in natural membrane domains, we have used new ways of exploiting high-resolution solid-state NMR spectroscopy [28,29]. These studies showed that higher phosphorylated phosphoinositides enhance the curvature of the membrane by inducing a spontaneous positive curvature that increases with the degree of phosphorylation [PtdIns<PtdIns4P<PtdIns(4,5)P2<PtdIns(3,4,5)P3]. At pH 7, PtdInsP3 forms small unilamellar vesicles (SUVs) (50 nm in diameter), PtdInsP2 forms mainly large unilamellar vesicles (LUVs) (100–500 nm in diameter) and PtdIns forms multilamellar vesicles (MLVs). The combination of phase diagrams and EM images is illuminating as it suggests that, through increases in spontaneous positive curvature, the local accumulation of polyphosphoinositides favours the formation of fusion pores, the last step in membrane fusion, and once again indicates that phosphoinositides can locally deform membrane structure .
Our results set the stage for investigating the dynamics and structure of model membranes that mimic the phospholipid composition of the nuclear membrane precursor MV1. We used deuterium solid-state NMR methods and developed methods  to probe membrane dynamics of model membranes that mimic the composition of NE precursor membranes.
Structural and dynamic events occurring in the membrane core and at the membrane surface were monitored by solid-state deuterium and phosphorus NMR. Results are summarized as follows: the ‘MV1-like’ membranes with the same composition of PtdIns, PtdInsP and PtdInsP2 as MV1 have an unusually fluid membrane core. Localized high levels of phosphoinositides impose orientation dynamics on other lipids. It is under these circumstances that the polyphosphoinositides can induce a clustering effect, which may enhance the ‘bulging’ of a region of negative potential. These conditions are favourable for interaction with various phosphoinositide-dependent protein domains . In that case, the phosphoinositides, even during their transient formation, can deform the local structure of the membrane to enhance protein binding .
Regulation of membrane morphology by derivatives of phosphoinositides and negatively curved lipids
Our work in echinoderms and formation of the male pronucleus led us to postulate that, in mammalian cells, phosphoinositides are also required to reform the NE and that DAG is the main fusogenic lipid involved in this process. To test this, an original approach was undertaken using chemical biology, confocal microscopy, correlative light and electron microscopy (CLEM) and three-dimensional reconstructions to reveal the structure of NE and ER membranes upon acute modification of DAG.
Roles for DAG in initiating fusion of biological membranes had been demonstrated in cell-free assays . However, studies directly manipulating DAG in vivo had not. Indirect studies on lipins (lipid phosphatases that convert phsophatidic acid into DAG) have shown that the deletion of the yeast lipin gene results in irregular nuclei with expanded NEs  and down-regulation of lipin by RNAi in Caenorhabditis elegans results in NE disruption [8,31].
In the absence of direct and acute in vivo data on the role of lipid asymmetry in endomembrane morphology, it has been argued that cytoskeletal proteins, scaffolding proteins or integral membrane proteins rather than non-lamellar bilayer lipids are responsible for shaping of organelles, local membrane curvature and fusion . Investigation of the contribution of non-lamellar lipids in vivo has been obstructed by the difficulty of altering composition in specific subcellular compartments.
We used 1,3-DAG to isolate the physical effects of DAG on membrane dynamics and morphology of mammalian cells . Although 1,2-DAG plays a well-documented role as a second messenger in signalling pathways, its 1,3 isomer is not a second messenger, but has the same physical properties . To examine the in vivo participation of DAG in NE formation and in ER architecture, we used our modified rapalogue heterodimerization device  that can recruit lipid-modifying enzymes to specific endomembranes to rapidly deplete DAG. Lamin B receptor (LBR) was used to target the ER/NE compartment.
To acutely modify DAG, we targeted DAG kinase ε (DGKε) (Figure 1) or skeletal muscle- and kidney-enriched inositol phosphatase (SKIP) . DGKε phosphorylates DAG to phosphatidic acid, SKIP dephosphorylates PtdIns(4,5)P2 to PtdIns4P, reducing the physiological substrate for phosphatidylinositol-specific PLC to form DAG. We showed that DAG is essential for the complete formation of the NE in mammalian cells and normal ER morphology.
The rapalogue dimerization tool used for localized intervention
We explain the mechanisms of the role of DAG in NE assembly and ER morphology by taking into account theoretical considerations that suggest that membrane curvature and bending energy are fundamental to the shaping of organelles and to fusion events .
In vivo, the energetic costs for maintaining organelle shape or highly localized membrane curvature have been ascribed to a wide variety of proteins in order to stabilize curved structures such as ER tubules and small vesicles or to facilitate localized curvature at fusion stalks . However, our data suggest that lipids contribute to such changes in natural membranes.
We had shown previously that acutely depleting the phospholipids PtdIns3P and PtdIns(3,5)P2 (exhibiting spontaneous positive curvature) from early endosomes results in elongated tubules characterized by high cross-sectional membrane curvature . In a subsequent investigation, we showed that depleting levels of DAG (exhibiting spontaneous negative curvature) leads to loss of NE assembly and formation of multilamellar sheets of ER at the expense of tubules (Figure 2).
NE formation is disrupted in a dose-dependent manner in DAG-depleted mitotic cells
We suggest that the balance and asymmetry of lipids of positive and negative curvature affects the localized geometry in vivo, which cannot be attributed solely to proteins.
We showed that the NE phenotype is reversed by exogenous delivery of unsaturated 1,3-DAG by SUVs, although the multilamellar sheets remained (Figure 3). In control telophase cells, numerous ER tubules extend to the nuclear membrane , whereas in cells depleted of DAG, the multilamellar sheets have fewer tubular extensions connected to the NE. We suggest that to form highly curved ER tubules sufficient DAG is required.
1,2- and 1,3-DAG rescue the fragmented NE phenotype
To rescue the fragmented NE, additional proteins may not be required, as 1,3-DAG (the non-protein-binding isomer of 1,2-DAG) completely reverses the phenotype. Furthermore, NE formation does not require a completely normal ER.
Our results show two functions for diacylglycerol in vivo: a structural role in organelle morphology and a role in localized extreme membrane curvature required for fusion for which proteins alone are an insufficient explanation. Previous work and experiments described below with echinoderms demonstrate that conserved mechanisms operate from embryonic to differentiated cells across the deuterostome superphylum.
Role of spontaneously negatively curved lipids in maintaining ER morphology of unfertilized echinoderm eggs
We sought a cell type in which we could rapidly evaluate the effects of phospholipid content on organelle morphology. The ideal organelle is the ER which consists of relatively flat sheets connected and continuous with tubules that are highly curved in cross-section [4,34]. Our aim was to be able to rapidly alter in the cell the overall content of lipids which exhibit different negative spontaneous curvature without intentional alterations of proteins and to monitor continuously changes in the amount of sheets.
The sea urchin egg has many advantages for studying the role of lipids in organelle morphology. The unfertilized egg has completed meiosis, contains a haploid female pronucleus and exists in a relatively dormant state until fertilization when the cell cycle resumes and many aspects of metabolism such as protein synthesis and respiration increase substantially . Its cytoplasm is filled with an extensive endoplasmic reticular network. Although the ER is capable of a rapid reorganization by fragmentation and reformation within 2–8 min of fertilization, in the unfertilized egg it is continuous and relatively static . Cells are large, abundant and uniform, and their health and viability can be tested by fertilization and subsequent embryonic development. Total phospholipid content remains constant in the unfertilized egg and embryo up to the blastula stage . Eggs can incorporate exogenous choline, ethanolamine and inositol. Phosphorus turnover in phosphoinositides occurs in the egg, but de novo synthesis of PtdIns is limited, although activated, after fertilization . Overall phospholipid synthesis is not significantly stimulated by fertilization .
Changes in ER morphology are straightforward to monitor using confocal microscopy. Effects of inadvertent egg activation can be eliminated by monitoring the lack of cortical granule exocytosis or progression of the female nucleus into mitosis of the first cell cycle. Cells can be loaded with exogenous phospholipids by exposure to SUVs of different lipid content. Internal phospholipid composition can be altered by microinjection of various lipid-modifying enzymes. We have reported previously that injections of either DGK (converting DAG into phosphatidic acid) or Syn1 [which depletes PtdIns(4,5)P2, the immediate endogenous precursor to DAG] result in a dose-dependent accumulation and stacking of ER sheets . Thus the egg can serve as an in vivo system allowing titration of the effects upon organelle shape of altering lipid content.
Eggs were labelled by microinjection of fluorescent DiIC18 in Wesson oil which partitions almost exclusively to the ER . The effects of injection of DGK were then evaluated. We tested pre-incubation of eggs with 20:80 mol% 1,3-DAG/phosphatidylcholine, 20:80 mol% phosphatidylethanolamine/phosphatidylcholine and 100% phosphatidylcholine SUVs. The spontaneous negative curvature of phosphatidylethanolamine is less than half that of DAG; and that of phosphatidylcholine less than 10% .
We also developed an image analysis protocol to quantify sheet accumulation. Sequential time images were imported into Volocity 6.3 (PerkinElmer, Improvision) for analysis. Since sheet regions appear to be more intense than tubular regions, presumably due to their stacking , we were able to partition the ER signal from each confocal section into sheet and non-sheet areas. As seen in Figure 4, the selected regions closely correspond to sheets and avoid the tubules.
Image analysis of ER sheet formation
Images were taken at the same level (near the equator) for any given egg. For comparison of trends across time, the sum of the cross-sectional area of each section assigned to sheets was divided by the cross-sectional area of the entire egg at each time point.
As indicated in Figure 5, the ER sheet content in control eggs remains virtually constant for >1 h. Microinjection of 250 μg/ml DGK causes a substantial increase in sheet content (Figures 4, middle panel, and 5a). Eggs pre-treated with 1,3-DAG-containing SUVs do not show an increase, and typically undergo a slight decrease in sheet content (Figure 5a). Pre-treatment with phosphatidylethanolamine-containing SUVs also prevents the increase in sheets.
Pre-treatment of echinoderm eggs with DAG or phosphatidylethanolamine SUVs prevents the effects of DGK
To test the ability of phosphatidylcholine, a lipid exhibiting almost no spontaneous negative curvature, eggs were exposed in the same way to SUVs made entirely of phosphatidylcholine and injected with only 100 μg/ml DGK. Nonetheless, phosphatidylcholine did not prevent accumulation of sheets (Figure 5b), whereas 1,3-DAG-containing SUVs effectively suppressed sheet accumulation (Figure 5b).
To follow trends of sheet formation or loss, fractional areas of sheets were normalized to the maximum of the time series for a given egg and plotted against relative time after injection. These graphs sensitively track trends in sheet formation or loss in a given egg (Figure 6). In control eggs, there are only minor changes in the pattern as regions of ER move in and out of the confocal plane (Figure 6a, light green line). However, when 250 μg/ml DGK was injected, sheets accumulated almost immediately after injection and progressively over the first 1 h (cyan). If the eggs were pre-loaded using 1,3-DAG-containing SUVs (red lines) or phosphatidylethanolamine-containing SUVs (dark green lines) sheet accumulation was either prevented or slowed respectively.
Normalized data showing trends of sheet formation
At lower levels of DGK (100 μg/ml), 1,3-DAG-containing SUVs not only prevented the increase of sheet content for at least 90 min, but also decreased it (Figure 6b, red lines) whereas phosphatidylcholine SUVs (purple) allowed sheet accumulation. The latter eggs developed grossly distorted patterns of ER by 2 h similar to DGK-injected eggs not exposed to SUVs.
These experiments indicate that microinjection into relatively dormant egg cells of DGK (which converts DAG into phosphatidic acid and thus lowers the amount of highly negative spontaneous curvature lipids available in the cell) can drastically alter ER morphology towards sheet formation. The accumulation of sheets can be prevented by pre-loading eggs with SUVs containing negative-curvature phospholipids (1,3-DAG or phosphatidylethanolamine), but not phosphatidylcholine. Phosphatidylethanolamine was less effective than DAG, which showed some signs of decreasing sheet content.
These experiments suggest again that an insufficiency of lipids of negative curvature caused by depletion of endogenous DAG by DGK is responsible for the alteration of ER morphology towards sheets.
Introduction of an image analysis routine to track trends of sheet accumulation which can be followed quantitatively allows for titration of the egg with amount of added enzyme or composition and concentration of SUVs making further detailed analysis of intrinsic lipid curvature, time of exposure and reversibility in a living cell now possible.
To propose the concept of phospholipids, specifically the phosphoinositides and their derivatives, as both signalling molecules as well as modifiers of membrane morphology is classically a contradiction in terms.
However, our recent results described in the present article have led us to make such a bold statement as to suggest a picturesque analogy to the wave-particle duality of photons, where the particle aspect has a similitude to transient signalling properties of phosphoinositides and the wave aspect a resemblance to the less transient properties of phosphoinositides as modulators of endomembrane structure.
One matter is certain: neutral lipids such as DAG, a derivative of the hydrolysis of PtdIns(4,5)P2, can no longer be regarded as just second messengers. Their abundant presence at the ER and Golgi do have a significant role in maintaining the proper ER and NE architecture. This role is conserved from echinoderms to mammals.
The same type of principle may also apply to the even more transiently formed phosphoinositide PtdIns(3,4,5)P3. As we have shown using high-resolution solid-state NMR spectroscopy, elevated levels of PtdIns(3,4,5)P3 create disordered domains which favour protein interactions. These high levels are not just a property of echinoderm membranes, but they have also been observed and studied in motility of mammalian neutrophils . Although the accumulation of PtdIns(3,4,5)P3 at the leading edge of these cells is fast, this may not exclude the idea that, at the same time that PtdIns(3,4,5)P3 is an effector of binding proteins, it also facilitates binding of proteins to membranes by affecting the local structure of the membranes.
To enable a more profound understanding of the dual properties of phosphoinositides as regulators of membrane morphology and dynamics, we need to develop further the current analytical and high-resolution imaging tools (advanced EM) that can quantify and localize simultaneously the formation and, more appropriately, the turnover of phosphoinositides in endomembranes.
Membrane Morphology and Function: A Biochemical Society Focused Meeting held at Hotel del Camerlengo, Fara San Martino, Abruzzo, Italy, 5–8 May 2014. Organized and Edited by Banafshé Larijani [IKERBASQUE, Basque Foundation for Science and Unidad de Biofísica (CSIC-UPV/EHU), University of the Basque Country, Spain] and Marco Falasca (Barts and The London School of Medicine and Dentistry, U.K.)
correlative light and electron microscopy
lamin B receptor
skeletal muscle- and kidney-enriched inositol phosphatase
soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor
small unilamellar vesicle
We thank Tina M.C. Hobday, Vanessa Zhendre and Charles Ray and Erick J. Dufourc for their fruitful collaborations and contributions in the research described in this article.
We thank Amherst College SURF (Summer Undergraduate Research Fellowship) programme and Cancer Research UK.