Lipid remodeling of GPI-anchored proteins and its function

Many proteins are attached to the cell surface via a conserved post-translational modification, the glycosylphosphatidylinositol (GPI) anchor. GPI-anchored proteins are functionally diverse, but one of their most striking features is their association with lipid microdomains, which consist mainly of sphingolipids and sterols. GPI-anchored proteins modulate various biological functions when they are incorporated into these specialized domains. The biosynthesis of GPI and its attachment to proteins occurs in the endoplasmic reticulum. The lipid moieties of GPI-anchored proteins are further modified during their transport to the cell surface, and these remodeling processes are essential for the association of proteins with lipid microdomains. Recently, several genes required for GPI lipid remodeling have been identified in yeast and mammalian cells. In this review, we describe the pathways for lipid remodeling of GPI-anchored proteins in yeast and mammalian cells, and discuss how lipid remodeling affects the association of GPI-anchored proteins with microdomains in cellular events.     

Glycosylphosphatidylinositol-anchored proteins: structure, function, and cleavage by phosphatidylinositol-specific phospholipase C.

A wide variety of proteins are tethered by a glycosylphosphatidylinositol (GPI) anchor to the extracellular face of eukaryotic plasma membranes, where they are involved in a number of functions ranging from enzymatic catalysis to adhesion. The exact function of the GPI anchor has been the subject of much speculation. It appears to act as an intracellular signal targeting proteins to the apical surface in polarized cells. GPI-anchored proteins are sorted into sphingolipid- and cholesterol-rich microdomains, known as lipid rafts, before transport to the membrane surface. Their localization in raft microdomains may explain the involvement of this class of proteins in signal transduction processes. Substantial evidence suggests that GPI-anchored proteins may interact closely with the bilayer surface, so that their functions may be modulated by the biophysical properties of the membrane. The presence of the anchor appears to impose conformational restraints, and its removal may alter the catalytic properties and structure of a GPI-anchored protein. Release of GPI-anchored proteins from the cell surface by specific phospholipases may play a key role in regulation of their surface expression and functional properties. Reconstitution of GPI-anchored proteins into bilayers of defined phospholipids provides a powerful tool with which to explore the interactions of these proteins with the membrane and investigate how bilayer properties modulate their structure, function, and cleavage by phospholipases.     

Sphingolipids are required for the stable membrane association of glycosylphosphatidylinositol-anchored proteins in yeast

Ongoing sphingolipid synthesis is specifically required in vivo for the endoplasmic reticulum (ER) to Golgi transport of glycosylphosphatidylinositol (GPI)-anchored proteins. However, the sphingolipid intermediates that are required for transport nor their role(s) have been identified. Using stereoisomers of dihydrosphingosine, together with specific inhibitors and a mutant defective for sphingolipid synthesis, we now show that ceramides and/or inositol sphingolipids are indispensable for GPI-anchored protein transport. Furthermore, in the absence of sphingolipid synthesis, a significant fraction of GPI-anchored proteins is no longer associated tightly with the ER membrane. The loose membrane association is neither because of the lack of a GPI-anchor nor because of prolonged ER retention of GPI-anchored proteins. These results indicate that ceramides and/or inositol sphingolipids are required to stabilize the association of GPI-anchored proteins with membranes. They could act either by direct involvement as membrane components or as substrates for the remodeling of GPI lipid moieties.     

Two different types of lipid moieties are present in glycophosphoinositol-anchored membrane proteins of Saccharomyces cerevisiae.

Numerous glycoproteins of Saccharomyces cerevisiae are anchored in the lipid bilayer by a glycophosphatidylinositol (GPI) anchor. Mild alkaline hydrolysis reveals that the lipid components of these anchors are heterogeneous in that both base-sensitive and base-resistant lipid moieties can be found on most proteins. The relative abundance of base-resistant lipid moieties is different for different proteins. Strong alkaline or acid hydrolysis of the mild base-resistant lipid component liberates C18-phytosphingosine indicating the presence of a ceramide. Two lines of evidence suggest that proteins are first attached to a base-sensitive GPI anchor, the lipid moiety of which subsequently gets exchanged for a base-resistant ceramide: (i) an early glycolipid intermediate of GPI biosynthesis only contains base-sensitive lipid moieties; (ii) after a pulse with [3H]myo-inositol the relative abundance of base-sensitive GPI anchors decreases significantly during chase. This decrease does not take place if GPI-anchored proteins are retained in the ER.     

Structural remodeling of GPI anchors during biosynthesis and after attachment to proteins

Glycosylphosphatidylinositol (GPI) anchoring of proteins is a conserved post-translational modification in eukaryotes. In mammalian cells, approximately 150 proteins on the plasma membrane are attached to the cell surface by GPI anchors, which confer specific properties on proteins, such as association with membrane microdomains. The structures of lipid and glycan moieties on GPI anchors are remodeled during biosynthesis and after attachment to proteins. The remodeling processes are critical for transport and microdomain-association of GPI-anchored proteins. Here, we describe the structural remodeling of GPI anchors and genes required for the processes in mammals, yeast, and trypanosomes.     

Saccharomyces cerevisiae CWH43 is involved in the remodeling of the lipid moiety of GPI anchors to ceramides

The glycosylphosphatidylinositol (GPI)-anchored proteins are subjected to lipid remodeling during their biosynthesis. In the yeast Saccharomyces cerevisiae, the mature GPI-anchored proteins contain mainly ceramide or diacylglycerol with a saturated long-fatty acid, whereas conventional phosphatidylinositol (PI) used for GPI biosynthesis contains an unsaturated fatty acid. Here, we report that S. cerevisiae Cwh43p, whose N-terminal region contains a sequence homologous to mammalian PGAP2, is involved in the remodeling of the lipid moiety of GPI anchors to ceramides. In cwh43 disruptant cells, the PI moiety of the GPI-anchored protein contains a saturated long fatty acid and lyso-PI but not inositolphosphorylceramides, which are the main lipid moieties of GPI-anchored proteins from wild-type cells. Moreover, the C-terminal region of Cwh43p (Cwh43-C), which is not present in PGAP2, is essential for the ability to remodel GPI lipids to ceramides. The N-terminal region of Cwh43p (Cwh43-N) is associated with Cwh43-C, and it enhanced the lipid remodeling to ceramides by Cwh43-C. Our results also indicate that mouse FRAG1 and C130090K23, which are homologous to Cwh43-N and -C, respectively, share these activities.     

The yeast p24 complex regulates GPI-anchored protein transport and quality control by monitoring anchor remodeling

Glycosylphosphatidylinositol (GPI)-anchored proteins are secretory proteins that are attached to the cell surface of eukaryotic cells by a glycolipid moiety. Once GPI anchoring has occurred in the lumen of the endoplasmic reticulum (ER), the structure of the lipid part on the GPI anchor undergoes a remodeling process prior to ER exit. In this study, we provide evidence suggesting that the yeast p24 complex, through binding specifically to GPI-anchored proteins in an anchor-dependent manner, plays a dual role in their selective trafficking. First, the p24 complex promotes efficient ER exit of remodeled GPI-anchored proteins after concentration by connecting them with the COPII coat and thus facilitates their incorporation into vesicles. Second, it retrieves escaped, unremodeled GPI-anchored proteins from the Golgi to the ER in COPI vesicles. Therefore the p24 complex, by sensing the status of the GPI anchor, regulates GPI-anchored protein intracellular transport and coordinates this with correct anchor remodeling.     

Two endoplasmic reticulum (ER) membrane proteins that facilitate ER-to-Golgi transport of glycosylphosphatidylinositol-anchored proteins

Many eukaryotic cell surface proteins are anchored in the lipid bilayer through glycosylphosphatidylinositol (GPI). GPI anchors are covalently attached in the endoplasmic reticulum (ER). The modified proteins are then transported through the secretory pathway to the cell surface. We have identified two genes in Saccharomyces cerevisiae, LAG1 and a novel gene termed DGT1 (for "delayed GPI-anchored protein transport"), encoding structurally related proteins with multiple membrane-spanning domains. Both proteins are localized to the ER, as demonstrated by immunofluorescence microscopy. Deletion of either gene caused no detectable phenotype, whereas lag1Delta dgt1Delta cells displayed growth defects and a significant delay in ER-to-Golgi transport of GPI-anchored proteins, suggesting that LAG1 and DGT1 encode functionally redundant or overlapping proteins. The rate of GPI anchor attachment was not affected, nor was the transport rate of several non-GPI-anchored proteins. Consistent with a role of Lag1p and Dgt1p in GPI-anchored protein transport, lag1Delta dgt1Delta cells deposit abnormal, multilayered cell walls. Both proteins have significant sequence similarity to TRAM, a mammalian membrane protein thought to be involved in protein translocation across the ER membrane. In vivo translocation studies, however, did not detect any defects in protein translocation in lag1Delta dgt1Delta cells, suggesting that neither yeast gene plays a role in this process. Instead, we propose that Lag1p and Dgt1p facilitate efficient ER-to-Golgi transport of GPI-anchored proteins.     

Defective lipid remodeling of GPI anchors in peroxisomal disorders, Zellweger syndrome, and rhizomelic chondrodysplasia punctata

Many cell surface proteins in mammalian cells are anchored to the plasma membrane via glycosylphosphatidylinositol (GPI). The predominant form of mammalian GPI contains 1-alkyl-2-acyl phosphatidylinositol (PI), which is generated by lipid remodeling from diacyl PI. The conversion of diacyl PI to 1-alkyl-2-acyl PI occurs in the ER at the third intermediate in the GPI biosynthetic pathway. This lipid remodeling requires the alkyl-phospholipid biosynthetic pathway in peroxisome. Indeed, cells defective in dihydroxyacetone phosphate acyltransferase (DHAP-AT) or alkyl-DHAP synthase express only the diacyl form of GPI-anchored proteins. A defect in the alkyl-phospholipid biosynthetic pathway causes a peroxisomal disorder, rhizomelic chondrodysplasia punctata (RCDP), and defective biogenesis of peroxisomes causes Zellweger syndrome, both of which are lethal genetic diseases with multiple clinical phenotypes such as psychomotor defects, mental retardation, and skeletal abnormalities. Here, we report that GPI lipid remodeling is defective in cells from patients with Zellweger syndrome having mutations in the peroxisomal biogenesis factors PEX5, PEX16, and PEX19 and in cells from patients with RCDP types 1, 2, and 3 caused by mutations in PEX7, DHAP-AT, and alkyl-DHAP synthase, respectively. Absence of the 1-alkyl-2-acyl form of GPI-anchored proteins might account for some of the complex phenotypes of these two major peroxisomal disorders.     

Requirement of PIG-F and PIG-O for transferring phosphoethanolamine to the third mannose in glycosylphosphatidylinositol

Many eukaryotic proteins are anchored by glycosylphosphatidylinositol (GPI) to the cell surface membrane. The GPI anchor is linked to proteins by an amide bond formed between the carboxyl terminus and phosphoethanolamine attached to the third mannose. Here, we report the roles of two mammalian genes involved in transfer of phosphoethanolamine to the third mannose in GPI. We cloned a mouse gene termed Pig-o that encodes a 1101-amino acid PIG-O protein bearing regions conserved in various phosphodiesterases. Pig-o knockout F9 embryonal carcinoma cells expressed very little GPI-anchored proteins and accumulated the same major GPI intermediate as the mouse class F mutant cell, which is defective in transferring phosphoethanolamine to the third mannose due to mutant Pig-f gene. PIG-O and PIG-F proteins associate with each other, and the stability of PIG-O was dependent upon PIG-F. However, the class F cell is completely deficient in the surface expression of GPI-anchored proteins. A minor GPI intermediate seen in Pig-o knockout but not class F cells had more than three mannoses with phosphoethanolamines on the first and third mannoses, suggesting that this GPI may account for the low expression of GPI-anchored proteins. Therefore, mammalian cells have redundant activities in transferring phosphoethanolamine to the third mannose, both of which require PIG-F.     

Novel applications for glycosylphosphatidylinositol-anchored proteins in pharmaceutical and industrial biotechnology

Glycosylphosphatidylinositol (GPI)-anchored proteins have been regarded as typical cell surface proteins found in most eukaryotic cells from yeast to man. They are embedded in the outer plasma membrane leaflet via a carboxy-terminally linked complex glycolipid GPI structure. The amphiphilic nature of the GPI anchor, its compatibility with the function of the attached protein moiety and the capability of GPI-anchored proteins for spontaneous insertion into and transfer between artificial and cellular membranes initially suggested their potential for biotechnological applications. However, these expectations have been hardly fulfilled so far. Recent developments fuel novel hopes with regard to: (i) Automated online expression, extraction and purification of therapeutic proteins as GPI-anchored proteins based on their preferred accumulation in plasma membrane lipid rafts, (ii) multiplex custom-made protein chips based on GPI-anchored cell wall proteins in yeast, (iii) biomaterials and biosensors with films consisting of sets of distinct GPI-anchored binding-proteins or enzymes for sequential or combinatorial catalysis, and (iv) transport of therapeutic proteins across or into relevant tissue cells, e.g., enterocytes or adipocytes. Latter expectations are based on the demonstrated translocation of GPI-anchored proteins from plasma membrane lipid rafts to cytoplasmic lipid droplets and eventually further into microvesicles which upon release from donor cells transfer their GPI-anchored proteins to acceptor cells. The value of these technologies, which are all based on the interaction of GPI-anchored proteins with membranes and surfaces, for the engineering, production and targeted delivery of biomolecules for a huge variety of therapeutic and biotechnological purposes should become apparent in the near future.     

Human PIG-U and yeast Cdc91p are the fifth subunit of GPI transamidase that attaches GPI-anchors to proteins

Many eukaryotic proteins are anchored to the cell surface via glycosylphosphatidylinositol (GPI), which is posttranslationally attached to the carboxyl-terminus by GPI transamidase. The mammalian GPI transamidase is a complex of at least four subunits, GPI8, GAA1, PIG-S, and PIG-T. Here, we report Chinese hamster ovary cells representing a new complementation group of GPI-anchored protein-deficient mutants, class U. The class U cells accumulated mature and immature GPI and did not have in vitro GPI transamidase activity. We cloned the gene responsible, termed PIG-U, that encoded a 435-amino-acid hydrophobic protein. The GPI transamidase complex affinity-purified from cells expressing epitope-tagged-GPI8 contained PIG-U and four other known components. Cells lacking PIG-U formed complexes of the four other components normally but had no ability to cleave the GPI attachment signal peptide. Saccharomyces cerevisiae Cdc91p, with 28% amino acid identity to PIG-U, partially restored GPI-anchored proteins on the surface of class U cells. PIG-U and Cdc91p have a functionally important short region with similarity to a region conserved in long-chain fatty acid elongases. Taken together, PIG-U and the yeast orthologue Cdc91p are the fifth component of GPI transamidase that may be involved in the recognition of either the GPI attachment signal or the lipid portion of GPI.     

Diagnosis of paroxysmal nocturnal hemoglobinuria by fluorescent clostridium septicum alpha toxin

Paroxysmal nocturnal hemoglobinuria (PNH), a hematopoietic stem cell disorder, is caused by the loss of glycosylphosphatidylinositol (GPI)-anchored proteins on the cell membrane. PNH can be simply diagnosed by flow cytometry using monoclonal antibodies against GPI-anchored proteins or fluorescent-tagged aerolysin, a bacterial toxin that binds GPI anchored proteins. Clostridium septicum alpha toxin is homologous to aerolysin and specifically binds GPI-anchored proteins. Previously, we found that an alpha toxin m45 mutant with two amino acid changes, S189C/S238C, lost cytotoxicity but still possessed binding activity for GPI-anchored proteins. To use this mutant toxin as a diagnostic probe in flow cytometry, we constructed the EGFP-AT(m45) expression vector, comprising a S189C/S238C alpha toxin mutant with EGFP and His tags at the N and C termini, respectively. The recombinant EGFP-AT(m45) was easily purified using single-step affinity chromatography against His tag from Escherichia coli. EGFP-AT(m45) bound to CHO and HeLa cells in a similar manner to monoclonal antibodies against GPI-anchored proteins or aerolysin. In whole blood from a PNH patient, GPI-deficient granulocytes could be differentiated by EGFP-AT(m45) using the same procedure as that employed with commercially available monoclonal antibodies. Therefore, nontoxic EGFP-conjugated C. septicum alpha toxin could be used clinically for PNH diagnosis.     

Cross-talk between caveolae and glycosylphosphatidylinositol-rich domains.

Most mammalian cells have in their plasma membrane at least two types of lipid microdomains, non-invaginated lipid rafts and caveolae. Glycosylphosphatidylinositol (GPI)-anchored proteins constitute a class of proteins that are enriched in rafts but not caveolae at steady state. We have analyzed the effects of abolishing GPI biosynthesis on rafts, caveolae, and cholesterol levels. GPI-deficient cells were obtained by screening for resistance to the pore-forming toxin aerolysin, which uses this class of proteins as receptors. Despite the absence of GPI-anchored proteins, mutant cells still contained lipid rafts, indicating that GPI-anchored proteins are not crucial structural elements of these domains. Interestingly, the caveolae-specific membrane proteins, caveolin-1 and 2, were up-regulated in GPI-deficient cells, in contrast to flotillin-1 and GM1, which were expressed at normal levels. Additionally, the number of surface caveolae was increased. This effect was specific since recovery of GPI biosynthesis by gene recomplementation restored caveolin expression and the number of surface caveolae to wild type levels. The inverse correlation between the expression of GPI-anchored proteins and caveolin-1 was confirmed by the observation that overexpression of caveolin-1 in wild type cells led to a decrease in the expression of GPI-anchored proteins. In cells lacking caveolae, the absence of GPI-anchored proteins caused an increase in cholesterol levels, suggesting a possible role of GPI-anchored proteins in cholesterol homeostasis, which in some cells, such as Chinese hamster ovary cells, can be compensated by caveolin up-regulation.     

PER1 is required for GPI-phospholipase A2 activity and involved in lipid remodeling of GPI-anchored proteins

Glycosylphoshatidylinositol (GPI) anchors are remodeled during their transport to the cell surface. Newly synthesized proteins are transferred to a GPI anchor, consisting of diacylglycerol with conventional C16 and C18 fatty acids, whereas the lipid moiety in mature GPI-anchored proteins is exchanged to either diacylglycerol containing a C26:0 fatty acid in the sn-2 position or ceramide in Saccharomyces cerevisiae. Here, we report on PER1, a gene encoding a protein that is required for the GPI remodeling pathway. We found that GPI-anchored proteins could not associate with the detergent-resistant membranes in per1Delta cells. In addition, the mutant cells had a defect in the lipid remodeling from normal phosphatidylinositol (PI) to a C26 fatty acid-containing PI in the GPI anchor. In vitro analysis showed that PER1 is required for the production of lyso-GPI, suggesting that Per1p possesses or regulates the GPI-phospholipase A2 activity. We also found that human PERLD1 is a functional homologue of PER1. Our results demonstrate for the first time that PER1 encodes an evolutionary conserved component of the GPI anchor remodeling pathway, highlighting the close connection between the lipid remodeling of GPI and raft association of GPI-anchored proteins.     

New mutant Chinese hamster ovary cell representing an unknown gene for attachment of glycosylphosphatidylinositol to proteins

Aerolysin, a secreted bacterial toxin from Aeromonas hydrophila, binds to glycosylphosphatidylinositol (GPI)-anchored protein and kills the cells by forming pores. Both GPI and N-glycan moieties of GPI-anchored proteins are involved in efficient binding of aerolysin. We isolated various Chinese hamster ovary (CHO) mutant cells resistant to aerolysin. Among them, CHOPA41.3 mutant cells showed several-fold decreased expression of GPI-anchored proteins. After transfection of N-acetylglucosamine transferase I (GnT1) cDNA, aerolysin was efficiently bound to the cells, indicating that the resistance against aerolysin in this cells was mainly ascribed to the defect of N-glycan maturation. CHOPA41.3 cells also accumulated GPI intermediates lacking ethanolamine phosphate modification on the first mannose. After stable transfection of PIG-N cDNA encoding GPI-ethanolamine phosphate transferase1, a profile of accumulated GPI intermediates became similar to that of GPI transamidase mutant cells. It indicated, therefore, that CHOPA41.3 cells are defective in GnT1, ethanolamine phosphate modification of the first mannose, and attachment of GPI to proteins. The GPI accumulation in CHOPA41.3 cells carrying PIG-N cDNA was not normalized after transfection with cDNAs of all known components in GPI transamidase complex. Microsomes from CHOPA41.3 cells had normal GPI transamidase activity. Taken together, there is an unknown gene required for efficient attachment of GPI to proteins.     

Sphingomyelin chain length influences the distribution of GPI-anchored proteins in rafts in supported lipid bilayers

Glycosyl-phosphatidylinositol (GPI)-anchored proteins are enriched in cholesterol- and sphingolipid-rich lipid rafts within the membrane. Rafts are known to have roles in cellular organization and function, but little is understood about the factors controlling the distribution of proteins in rafts. We have used atomic force microscopy to directly visualize proteins in supported lipid bilayers composed of equimolar sphingomyelin, dioleoyl-sn-glycero-3-phosphocholine and cholesterol. The transmembrane anchored angiotensin converting enzyme (TM-ACE) was excluded from the liquid ordered raft domains. Replacement of the transmembrane and cytoplasmic domains of TM-ACE with a GPI anchor (GPI-ACE) promoted the association of the protein with rafts in the bilayers formed with brain sphingomyelin (mainly C18:0). Association with the rafts did not occur if the shorter chain egg sphingomyelin (mainly C16:0) was used. The distribution of GPI-anchored proteins in supported lipid bilayers was investigated further using membrane dipeptidase (MDP) whose GPI anchor contains distearoyl phosphatidylinositol. MDP was also excluded from rafts when egg sphingomyelin was used but associated with raft domains formed using brain sphingomyelin. The effect of sphingomyelin chain length on the distribution of GPI-anchored proteins in rafts was verified using synthetic palmitoyl or stearoyl sphingomyelin. Both GPI-ACE and MDP only associated with the longer chain stearoyl sphingomyelin rafts. These data obtained using supported lipid bilayers provide the first direct evidence that the nature of the membrane-anchoring domain influences the association of a protein with lipid rafts and that acyl chain length hydrophobic mismatch influences the distribution of GPI-anchored proteins in rafts.     

Leishmania mexicana mutants lacking glycosylphosphatidylinositol (GPI):protein transamidase provide insights into the biosynthesis and functions of GPI-anchored proteins

The major surface proteins of the parasitic protozoon Leishmania mexicana are anchored to the plasma membrane by glycosylphosphatidylinositol (GPI) anchors. We have cloned the L. mexicana GPI8 gene that encodes the catalytic component of the GPI:protein transamidase complex that adds GPI anchors to nascent cell surface proteins in the endoplasmic reticulum. Mutants lacking GPI8 (DeltaGPI8) do not express detectable levels of GPI-anchored proteins and accumulate two putative protein-anchor precursors. However, the synthesis and cellular levels of other non-protein-linked GPIs, including lipophosphoglycan and a major class of free GPIs, are not affected in the DeltaGPI8 mutant. Significantly, the DeltaGPI8 mutant displays normal growth in liquid culture, is capable of differentiating into replicating amastigotes within macrophages in vitro, and is infective to mice. These data suggest that GPI-anchored surface proteins are not essential to L. mexicana for its entry into and survival within mammalian host cells in vitro or in vivo and provide further support for the notion that free GPIs are essential for parasite growth.     

Disruption of pioneer growth cone guidance in vivo by removal of glycosyl-phosphatidylinositol-anchored cell surface proteins.

Cell surface proteins anchored to membranes via covalently attached glycosyl-phosphatidylinositol (GPI) have been implicated in neuronal adhesion, promotion of neurite outgrowth and directed cell migration. Treatment of grasshopper embryos with bacterial phosphatidylinositol-specific phospholipase C (PI-PLC), an enzyme that cleaves the GPI anchor, often induced disruptions in the highly stereotyped migrations of peripheral pioneer growth cones and afferent neuron cell bodies. In distal limb regions of embryos treated with PI-PLC at early stages of pioneer axon outgrowth, growth cones lost their proximal orientation toward the central nervous system (CNS) and turned distally. Pioneer growth cones in treated limbs also failed to make a characteristic ventral turn along the trochanter-coxa (Tr-Cx) segment boundary, and instead continued to grow proximally across the boundary. Treatment at an earlier stage of development caused pre-axonogenesis Cx1 neurons to abandon their normal circumferential migration and reorient toward the CNS. None of these abnormal phenotypes were observed in limbs of untreated embryos or embryos exposed to other phospholipases that do not release GPI-anchored proteins. Incubation of embryos with PI-PLC effectively removed immunoreactivity for fasciclin I, a GPI-anchored protein expressed on a subset of neuronal surfaces. These results suggest that cell surface GPI-anchored proteins are involved in pioneer growth cone guidance and in pre-axonogenesis migration of neurons in the grasshopper limb bud in vivo.