Wunen, a Drosophila lipid phosphate phosphatase, is required for septate junction-mediated barrier function

Lipid phosphate phosphatases (LPPs) are integral membrane enzymes that regulate the levels of bioactive lipids such as sphingosine 1-phosphate and lysophosphatidic acid. The Drosophila LPPs Wunen (Wun) and Wunen-2 (Wun2) have a well-established role in regulating the survival and migration of germ cells. We now show that wun has an essential tissue-autonomous role in development of the trachea: the catalytic activity of Wun is required to maintain septate junction (SJ) paracellular barrier function, loss of which causes failure to accumulate crucial luminal components, suggesting a role for phospholipids in SJ function. We find that the integrity of the blood-brain barrier is also lost in wun mutants, indicating that loss of SJ function is not restricted to the tracheal system. Furthermore, by comparing the rescue ability of different LPP homologs we show that wun function in the trachea is distinct from its role in germ cell migration.     

Lysophosphatidic acid and sphingosine 1-phosphate biology: the role of lipid phosphate phosphatases

The biological actions of the lysolipid agonists sphingosine 1-phosphate and lysophosphatidic acid, in addition to other bioactive lipid phosphates such as phosphatidic acid and ceramide 1-phosphate, can be influenced by a family of lipid phosphate phosphatases (LPP), including LPP1, LPP2, LPP3, the Drosophila homologues Wunen (Wun) and Wunen2 (Wun2) and sphingosine 1-phosphate phosphatases 1 and 2 (SPP1, SPP2). This review describes the characteristic of these enzymes and their potential physiological roles in regulating intracellular and extracellular actions and amounts of these lipids in addition to the involvement of these phosphatases in development.     

Germ cell-autonomous Wunen2 is required for germline development in Drosophila embryos

In many animals, primordial germ cells (PGCs) migrate through the embryo towards the future gonad, a process guided by attractive and repulsive cues provided from surrounding somatic cells. In Drosophila, the two related lipid phosphate phosphatases (LPPs), Wunen (Wun) and Wun2, are thought to degrade extracellular substrates and to act redundantly in somatic cells to provide a repulsive environment to steer the migration of PGCs, or pole cells. Wun and Wun2 also affect the viability of pole cells, because overexpression of either one in somatic cells causes pole cell death. However, the means by which they regulate pole cell migration and survival remains elusive. We report that Wun2 has a maternal function required for the survival of pole cells during their migration to the gonad. Maternal wun2 RNA was found to be concentrated in pole cells and pole cell-specific expression of wun2 rescued the pole cell death phenotype of the maternal wun2 mutant, suggesting that wun2 activity in pole cells is required for their survival. Furthermore, we obtained genetic evidence that pole cell survival requires a proper balance of LPP activity in pole cells and somatic cells. We propose that Wun2 in pole cells competes with somatic Wun and Wun2 for a common lipid phosphate substrate, which is required by pole cells to produce their survival signal. In somatic cells, Wun and Wun2 may provide a repulsive environment for pole cell migration by depleting this extracellular substrate.     

Fly and mammalian lipid phosphate phosphatase isoforms differ in activity both in vitro and in vivo

Wunen (Wun), a homologue of a lipid phosphate phosphatase (LPP), has a crucial function in the migration and survival of primordial germ cells (PGCs) during Drosophila embryogenesis. Past work has indicated that the LPP isoforms may show functional redundancy in certain systems, and that they have broad-range lipid phosphatase activities in vitro, with little apparent specificity between them. We show here that there are marked differences in biochemical activity between fly Wun and mammalian LPPs, with Wun having a narrower activity range than has been reported for the mammalian LPPs. Furthermore, although it is active on a range of substrates in vitro, mouse Lpp1 has no activity on an endogenous Drosophila germ-cell-specific factor in vivo. Conversely, human LPP3 is active, resulting in aberrant migration and PGC death. These results show an absolute difference in bioactivity among LPP isoforms for the first time in a model organism and may point towards an underlying signalling system that is conserved between flies and humans.     

Enzymatic analysis of lipid phosphate phosphatases

Lipid phosphate monoesters including phosphatidic acid, lysophosphatidic acid, sphingosine 1-phosphate and ceramide 1-phosphate are intermediates in phosho- and sphingo-lipid biosynthesis and also play important roles in intra- and extra-cellular signaling. Dephosphorylation of these lipids terminates their signaling actions and, in some cases, generates products with additional biological activities or metabolic fates. The key enzymes responsible for dephosphorylation of these lipid phosphate substrates are collectively termed lipid phosphate phosphatases (LPPs). They are integral membrane enzymes with a core domain of six transmembrane spanning alpha-helices linked by extramembrane loops. LPPs are oriented in the membrane with their N- and C-termini facing the cytoplasm. LPPs exhibit isoform and cell specific localization patterns being variably distributed between endomembrane compartments (primarily the endoplasmic reticulum and Golgi apparatus) and the plasma membrane. The active site of these enzymes is formed from residues within two of the extramembrane loops and faces the lumen of endomembrane compartments or, when localized to the plasma membrane, towards, the extracellular space. Biochemical, pharmacological, cell biological and genetic studies identify roles for LPPs in both intracellular lipid metabolism and the regulation of both intra- and extra-cellular signaling pathways that control cell growth, survival and migration. This article describes procedures for the expression of LPPs in insect and mammalian cells and their analysis by SDS-PAGE and Western blotting. The most straightforward way to determine LPP activity is to measure release of the substrate phosphate group. We described methods for the synthesis and purification of [(32)P]-labeled LPP substrates. We describe the use of both radiolabeled and fluorescent lipid substrates for the detection, quantitation and analysis of the enzymatic activities of the LPPs measured using intact or broken cell preparations as the source of enzyme.     

Lipid phosphate phosphatases and their roles in mammalian physiology and pathology

Lipid phosphate phosphatases (LPPs) are a group of enzymes that belong to a phosphatase/phosphotransferase family. Mammalian LPPs consist of three isoforms: LPP1, LPP2, and LPP3. They share highly conserved catalytic domains and catalyze the dephosphorylation of a variety of lipid phosphates, including phosphatidate, lysophosphatidate (LPA), sphingosine 1-phosphate (S1P), ceramide 1-phosphate, and diacylglycerol pyrophosphate. LPPs are integral membrane proteins, which are localized on plasma membranes with the active site on the outer leaflet. This enables the LPPs to degrade extracellular LPA and S1P, thereby attenuating their effects on the activation of surface receptors. LPP3 also exhibits noncatalytic effects at the cell surface. LPP expression on internal membranes, such as endoplasmic reticulum and Golgi, facilitates the metabolism of internal lipid phosphates, presumably on the luminal surface of these organelles. This action probably explains the signaling effects of the LPPs, which occur downstream of receptor activation. The three isoforms of LPPs show distinct and nonredundant effects in several physiological and pathological processes including embryo development, vascular function, and tumor progression. This review is intended to present an up-to-date understanding of the physiological and pathological consequences of changing the activities of the different LPPs, especially in relation to cell signaling by LPA and S1P.     

Pulmonary phosphatidic acid phosphatase and lipid phosphate phosphohydrolase

The lung contains two distinct forms of phosphatidic acid phosphatase (PAP). PAP1 is a cytosolic enzyme that is activated through fatty acid-induced translocation to the endoplasmic reticulum, where it converts phosphatidic acid (PA) to diacylglycerol (DAG) for the biosynthesis of phospholipids and neutral lipids. PAP1 is Mg(2+) dependent and sulfhydryl reagent sensitive. PAP2 is a six-transmembrane-domain integral protein localized to the plasma membrane. Because PAP2 degrades sphingosine-1-phosphate (S1P) and ceramide-1-phosphate in addition to PA and lyso-PA, it has been renamed lipid phosphate phosphohydrolase (LPP). LPP is Mg(2+) independent and sulfhydryl reagent insensitive. This review describes LPP isoforms found in the lung and their location in signaling platforms (rafts/caveolae). Pulmonary LPPs likely function in the phospholipase D pathway, thereby controlling surfactant secretion. Through lowering the levels of lyso-PA and S1P, which serve as agonists for endothelial differentiation gene receptors, LPPs regulate cell division, differentiation, apoptosis, and mobility. LPP activity could also influence transdifferentiation of alveolar type II to type I cells. It is considered likely that these lipid phosphohydrolases have critical roles in lung morphogenesis and in acute lung injury and repair.     

Plastidic phosphatidic acid phosphatases identified in a distinct subfamily of lipid phosphate phosphatases with prokaryotic origin

Plastidic phosphatidic acid phosphatase (PAP) dephosphorylates phosphatidic acid to yield diacylglycerol, which is a precursor for galactolipids, a primary and indispensable component of photosynthetic membranes. Despite its functional importance, the molecular characteristics and phylogenetic origin of plastidic PAP were unknown because no potential homologs have been found. Here, we report the isolation and characterization of plastidic PAPs in Arabidopsis that belong to a distinct lipid phosphate phosphatase (LPP) subfamily with prokaryotic origin. Because no homolog of mammalian LPP was found in cyanobacteria, we sought an LPP ortholog in a more primitive organism, Chlorobium tepidum, and its homologs in cyanobacteria. Arabidopsis had five homologs of cyanobacterial LPP, three of which (LPP gamma, LPP epsilon 1, and LPP epsilon 2) localized to chloroplasts. Complementation of yeast Delta dpp1 Delta lpp1 Delta pah1 by plastidic LPPs rescued the relevant phenotype in vitro and in vivo, suggesting that they function as PAPs. Of the three LPPs, LPP gamma activity best resembled the native activity. The three plastidic LPPs were differentially expressed both in green and nongreen tissues, with LPP gamma expressed the highest in shoots. A knock-out mutant for LPP gamma could not be obtained, although a lpp epsilon 1 lpp epsilon 2 double knock-out showed no significant changes in lipid composition. However, lpp gamma homozygous mutant was isolated only under ectopic overexpression of LPP gamma, suggesting that loss of LPP gamma may cause lethal effect on plant viability. Thus, in Arabidopsis, there are three isoforms of plastidic PAP that belong to a distinct subfamily of LPP, and LPP gamma may be the primary plastidic PAP.     

Structural organization of mammalian lipid phosphate phosphatases: implications for signal transduction

This article describes the regulation of cell signaling by lipid phosphate phosphatases (LPPs) that control the conversion of bioactive lipid phosphates to their dephosphorylated counterparts. A structural model of the LPPs, that were previously called Type 2 phosphatidate phosphatases, is described. LPPs are characterized by having no Mg²⁺ requirement and their insensitivity to inhibition by N-ethylmaleimide. The LPPs have six putative transmembrane domains and three highly conserved domains that define a phosphatase superfamily. The conserved domains are juxtaposed to the proposed membrane spanning domains such that they probably form the active sites of the phosphatases. It is predicted that the active sites of the LPPs are exposed at the cell surface or on the luminal surface of intracellular organelles, such as Golgi or the endoplasmic reticulum, depending where various LPPs are expressed. LPPs could attenuate cell activation by dephosphorylating bioactive lipid phosphate esters such as phosphatidate, lysophosphatidate, sphingosine 1-phosphate and ceramide 1-phosphate. In so doing, the LPPs could generate alternative signals from diacylglycerol, sphingosine and ceramide. The LPPs might help to modulate cell signaling by the phospholipase D pathway. For example, phosphatidate generated within the cell by phospholipase D could be converted by an LPP to diacylglycerol. This should change the relative balance of signaling by these two lipids. Another possible function of the LPPs relates to the secretion of lysophosphatidate and sphingosine 1-phosphate by activated platelets and other cells. These exogenous lipids activate phospholipid growth factor receptors on the surface of cells. LPP activities could attenuate cell activation by lysophosphatidate and sphingosine 1-phosphate through their respective receptors.     

Differential localization of lipid phosphate phosphatases 1 and 3 to cell surface subdomains in polarized MDCK cells

Lipid phosphate phosphatases (LPPs) are integral membrane proteins with six transmembrane domains that act as ecto-enzymes dephosphorylating a variety of extracellular lipid phosphates. Using polarized MDCK cells stably expressing human LPP1 and LPP3, we found that LPP1 was located exclusively at the apical surface whereas LPP3 was distributed mostly in the basolateral subdomain. We identified a novel apical sorting signal at the N-terminus of LPP1 composed of F(2)DKTRL(7). In the case of LPP3, a dityrosine motif present in the second cytoplasmic portion was identified as basolateral targeting signal. Our work shows that LPP1 and LPP3 are equipped with distinct sorting signals that cause them to differentially localize to the apical vs. the basolateral subdomain, respectively.     

Structural organization of mammalian lipid phosphate phosphatases: implications for signal transduction.

This article describes the regulation of cell signaling by lipid phosphate phosphatases (LPPs) that control the conversion of bioactive lipid phosphates to their dephosphorylated counterparts. A structural model of the LPPs, that were previously called Type 2 phosphatidate phosphatases, is described. LPPs are characterized by having no Mg(2+) requirement and their insensitivity to inhibition by N-ethylmaleimide. The LPPs have six putative transmembrane domains and three highly conserved domains that define a phosphatase superfamily. The conserved domains are juxtaposed to the proposed membrane spanning domains such that they probably form the active sites of the phosphatases. It is predicted that the active sites of the LPPs are exposed at the cell surface or on the luminal surface of intracellular organelles, such as Golgi or the endoplasmic reticulum, depending where various LPPs are expressed. LPPs could attenuate cell activation by dephosphorylating bioactive lipid phosphate esters such as phosphatidate, lysophosphatidate, sphingosine 1-phosphate and ceramide 1-phosphate. In so doing, the LPPs could generate alternative signals from diacylglycerol, sphingosine and ceramide. The LPPs might help to modulate cell signaling by the phospholipase D pathway. For example, phosphatidate generated within the cell by phospholipase D could be converted by an LPP to diacylglycerol. This should change the relative balance of signaling by these two lipids. Another possible function of the LPPs relates to the secretion of lysophosphatidate and sphingosine 1-phosphate by activated platelets and other cells. These exogenous lipids activate phospholipid growth factor receptors on the surface of cells. LPP activities could attenuate cell activation by lysophosphatidate and sphingosine 1-phosphate through their respective receptors.     

Cloning and Characterization of Three Eimeria tenella Lipid Phosphate Phosphatases

Although lipid phosphate phosphatases (LPPs) play an important role in cellular signaling in addition to lipid biosynthesis, little is thus far known about parasite LPPs. In this study, we characterized three Eimeria tenella cDNA clones encoding LPP named EtLPP1, EtLPP2 and EtLPP3. Key structural features previously described in LPPs, including the three conserved domains proposed as catalytic sites, a single conserved N-glycosylation site, and putative transmembrane domains were discovered in the three resulting EtLPP amino acid sequences. Expression of His6-tagged EtLPP1, -2, and -3 in HEK293 cells produced immunoreactive proteins with variable molecular sizes, suggesting the presence of multiple forms of each of the three EtLPPs. The two faster-migrating protein bands below each of the three EtLPP proteins were found to be very similar to the porcine 35-kDa LPP enzyme in their molecular size and the extent of their N-glycosylation, suggesting that the three EtLPPs are partially N-glycosylated. Kinetic analyses of the activity of the three enzymes against PA, LPA, C1P and S1P showed that Km values for each of the substrates were (in μM) 284, 46, 28, and 22 for EtLPP1; 369, 179, 237, and 52 for EtLPP2; and 355, 83, and 260 for EtLPP3. However, EtLPP3 showed negligible activity on S1P. These results confirmed that the three EtLPPs have broad substrate specificity. The results also indicated that despite structural similarities, the three EtLPPs may play distinct functions through their different models of substrate preference. Furthermore, particularly high expression levels of the three EtLPP genes were detected in the sporozoite stage of the E. tenella life cycle (p<0.001), suggesting that their encoded proteins might play an important biological function in the sporozoite stage.     

Regulation of lysophosphatidate signaling by autotaxin and lipid phosphate phosphatases with respect to tumor progression, angiogenesis, metastasis and chemo-resistance

Evidence from clinical, animal and cell culture studies demonstrates that increased autotaxin (ATX) expression is responsible for enhancing tumor progression, cell migration, metastases, angiogenesis and chemo-resistance. These effects depend mainly on the rapid formation of lysophosphatidate (LPA) by ATX. Circulating LPA has a half-life of about 3 min in mice and it is degraded by the ecto-activities of lipid phosphate phosphatases (LPPs). These enzymes also hydrolyze extracellular sphingosine 1-phosphate (S1P), a potent signal for cell division, survival and angiogenesis. Many aggressive tumor cells express high ATX levels and low LPP activities. This favors the formation of locally high LPA and S1P concentrations. Furthermore, LPPs attenuate signaling downstream of the activation of G-protein coupled receptors and receptor tyrosine kinases. Therefore, we propose that the low expression of LPPs in many tumor cells makes them hypersensitive to growth promoting and survival signals that are provided by LPA, S1P, platelet-derived growth factor (PDGF) and epidermal growth factor (EGF). One of the key signaling pathways in this respect appears to be activation of phospholipase D (PLD) and phosphatidate (PA) production. This is required for the transactivations of the EGFR and PDGFR and also for LPA-induced cell migration. PA also increases the activities of ERK, mTOR, myc and sphingosine kinase-1 (SK-1), which provide individual signals for cells division, survival, chemo-resistance and angiogenesis. This review focuses on the balance of signaling by bioactive lipids including LPA, phosphatidylinositol 3,4,5-trisphosphate, PA and S1P versus the action of ceramides. We will discuss how these lipid mediators interact to produce an aggressive neoplastic phenotype.     

Lipid phosphate phosphatases 1 and 3 are localized in distinct lipid rafts

Lipid phosphate phosphatases (LPPs), integral membrane proteins with six transmembrane domains, dephosphorylate a variety of extracellular lipid phosphates. Although LPP3 is already known to bind to Triton X-100-insoluble rafts, we here report that LPP1 is also associated with lipid rafts distinct from those harboring LPP3. We found that LPP1 was Triton X-100-soluble, but CHAPS-insoluble in LNCaP cells endogenously expressing LPP1 and several LPP1 cDNA-transfected cells including NIH3T3 fibroblasts. In addition to the non-ionic detergent insolubility, LPP1 further possessed several properties formulated for raft-localizing proteins as follows: first, the CHAPS-insolubility was resistant to the actin-disrupting drug cytochalasin D; second, the CHAPS-insoluble LPP1 floated in an Optiprep density gradient; third, the CHAPS insolubility of LPP1 was lost by cholesterol depletion; and finally, the subcellular distribution pattern of LPP1 exclusively overlapped with that of a raft marker, cholera toxin B subunit. Interestingly, confocal microscopic analysis showed that LPP1 was distributed to membrane compartments distinct from those of LPP3. Analysis using various LPP1/LPP3 chimeras revealed that their first extracellular regions determine the different Triton X-100 solubilities. These results indicate that LPP1 and LPP3 are distributed in distinct lipid rafts that may provide unique microenvironments defining their non-redundant physiological functions.     

Tetracyclines increase lipid phosphate phosphatase expression on plasma membranes and turnover of plasma lysophosphatidate

Extracellular lysophosphatidate and sphingosine 1-phosphate (S1P) are important bioactive lipids, which signal through G-protein-coupled receptors to stimulate cell growth and survival. The lysophosphatidate and S1P signals are terminated partly by degradation through three broad-specificity lipid phosphate phosphatases (LPPs) on the cell surface. Significantly, the expression of LPP1 and LPP3 is decreased in many cancers, and this increases the impact of lysophosphatidate and S1P signaling. However, relatively little is known about the physiological or pharmacological regulation of the expression of the different LPPs. We now show that treating several malignant and nonmalignant cell lines with 1 μg/ml tetracycline, doxycycline, or minocycline significantly increased the extracellular degradation of lysophosphatidate. S1P degradation was also increased in cells that expressed high LPP3 activity. These results depended on an increase in the stabilities of the three LPPs and increased expression on the plasma membrane. We tested the physiological significance of these results and showed that treating rats with doxycycline accelerated the clearance of lysophosphatidate, but not S1P, from the circulation. However, administering 100 mg/kg/day doxycycline to mice decreased plasma concentrations of lysophosphatidate and S1P. This study demonstrates a completely new property of tetracyclines in increasing the plasma membrane expression of the LPPs.     

Sphingosine-1-phosphate and lipid phosphohydrolases

Sphingosine-1-phosphate (S1P) is a bioactive sphingolipid that acts as both an extracellular ligand for the endothelial differentiation gene-1 (EDG-1) G-protein coupled receptor (GPCR) family and as an intracellular messenger. Cellular levels of S1P are low and tightly regulated in a spatial-temporal manner not only by sphingosine kinase (SPHK) but also by degradation catalyzed by S1P lyase, specific S1P phosphohydrolases, and by general lipid phosphate phosphohydrolases (LPPs). LPPs are characterized as magnesium-independent, insensitive to inhibition by N-ethylmaleimide (NEM) and possessing broad substrate specificity with a variety of phosphorylated lipids, including S1P, phosphatidic acid (PA), and lysophosphatidic acid (LPA). LPPs contain three highly conserved domains that define a phosphohydrolase superfamily. Recently, several specific S1P phosphohydrolases have been identified in yeast and mammalian cells. Phylogenetic and biochemical analyses indicate that these enzymes constitute a new subset of the LPP family. As further evidence, S1P phosphohydrolases exhibit high specificity for phosphorylated sphingoid bases. Enforced expression of S1P phosphohydrolase alters the cellular levels of sphingolipid metabolites in yeast and mammalian cells, increasing sphingosine and ceramide, bioactive sphingolipids that often have opposing biological actions to S1P. By regulating the cellular ratio between ceramide/sphingosine and S1P, S1P phosphohydrolase is poised to be a critical factor in cell survival/cell death decisions. Indeed, expression of S1P phosphohydrolase in mammalian cells increases apoptosis, whereas deletion of S1P phosphohydrolases in yeast correlates with resistance to heat stress. In this review, we discuss the role of phosphohydrolases in the metabolism of S1P and how turnover of S1P can regulate sphingolipid metabolites signaling.     

The importance of ceramide headgroup for lipid localisation in skin lipid models

The stratum corneum's lipid matrix is a critical for the skin's barrier function and is primarily composed of ceramides (CERs), cholesterol (CHOL) and free fatty acids (FFAs). The lipids form a long periodicity phase (LPP), a unique trilayer unit cell structure. An enzyme driven pathway is implemented to synthesize these key lipids. If these enzymes are down- or upregulated as in inflammatory diseases, the final lipid composition is affected often altering the barrier function. In this study, we mimicked down regulation of enzymes involved in the synthesis of the sphingosine and CER amide bond. In a LPP lipid model, we substituted CER N-(tetracosanoyl)-sphingosine (CER NS) with either i) FFA C24 and free sphingosine, to simulate the loss of the CER amide bond, or ii) with FFA C24 and C18 to simulate the loss of the sphingosine headgroup. Our study shows the lipids in the LPP would not phase separate until at least 25% of the CER NS is substituted keeping the lateral packing and conformational ordering unaltered. Neutron diffraction studies showed that free sphingosine chains localized at the outer layers of the unit cell, while the remaining CER NS head group was concentrated in the inner headgroup layers. However, when FFA C18 was inserted, CER NS was dispersed throughout the LPP, resulting in an even distribution between the inner and outer water layers. The presented results highlight the importance of the CER NS headgroup structure and its interaction in combination with the carbon chain invariability for optimal lipid arrangement.     

Involvement of Lysophosphatidic Acid,Sphingosine 1-Phosphate and Ceramide 1-Phosphate in the Metabolization of Phosphatidic Acid by Lipid Phosphate Phosphatases in Bovine Rod Outer Segments

The aim of the present research was to evaluate the generation of [2-3H]diacylglycerol ([2-3H]DAG) from [2-3H]-Phosphatidic acid ([2-3H]PA) by lipid phosphate phosphatases (LPPs) at different concentrations of lysophosphatidic acid (LPA), sphingosine 1-phosphate (S1P), and ceramide 1-phosphate (C1P) in purified ROS obtained from dark-adapted retinas (DROS) or light-adapted retinas (BLROS) as well as in ROS membrane preparations depleted of soluble and peripheral proteins. Western blot analysis revealed the presence of LPP3 exclusively in all membrane preparations. Immunoblots of entire ROS and depleted ROS did not show dark-light differences in LPP3 levels. LPPs activities were diminished by 53% in BLROS with respect to DROS. The major competitive effect on PA hydrolysis was exerted by LPA and S1P in DROS and by C1P in BLROS. LPPs activities in depleted ROS were similar to the activity observed in entire DROS and BLROS, respectively. LPA, S1P and C1P competed at different extent in depleted DROS and BLROS. Sphingosine and ceramide inhibited LPPs activities in entire and depleted DROS. Ceramide also inhibited LPPs activities in entire and in depleted BLROS. Our findings are indicative of a different degree of competition between PA and LPA, S1P and C1P by LPPs depending on the illumination state of the retina.