Chapter 2:  Cell Membrane Structures and Functions

Cells are separated from their environment by lipid bilayers

The fundamental importance of lipids in membrane structure was established early in this century by demonstrations that positive correlations exist between cell membrane permeabilities to small nonelectrolytes and the oil/water partition coefficients of these molecules. Contemporary measurements of the electrical impedance of cell suspensions suggested that cells are surrounded by a hydrocarbon barrier, which was first estimated to be about 3.3 nm thick. It was originally thought that a membrane containing a lipid monolayer could account for these data. However, subsequent experiments established that the ratio of the area of a monolayer formed from erythrocyte membrane lipids to the surface area of these cells is nearly 2. These and other studies of the physical chemistry of lipids fortified the concept that a continuous lipid bilayer is a major component of cell membranes. This concept has received support from many other studies, including the interpretation of X-ray diffraction data obtained from intact cell membranes.

Forces acting between lipids and between lipids and proteins are primarily noncovalent, consisting of electrostatic, hydrogen-bonding and van der Waals' interactions. Although these are weak interactions relative to covalent bonds, their sum can produce very stable associations. Ionic and polar parts of molecules exposed to water will become hydrated. Substances dissolve in a solvent only if their molecules interact with the solvent more strongly than with each other. Complex molecules may have two or more surface domains that differ in polarity. In aqueous solution their apolar surfaces form an internal hydrophobic phase that minimizes their exposure to water and their more polar surfaces form an external hydrated phase. Molecules with segregated polar and nonpolar surfaces are termed amphipathic. These include most biological lipids and many proteins.

Amphipathic molecules form bilayered lamellar structures spontaneously if they have an appropriate geometry

Most of the major cell membrane lipids have a polar head, commonly a glycerophosphorylester moiety and a hydrocarbon tail, usually consisting of two esterified fatty acids (Chap. 3). The head groups can interact with water and aqueous phase solutes, whereas the nonpolar tails aggregate to form an internal phase.

Three principal phases with different structures are formed by phospholipids in the presence of water [1] (Fig. 2-2). Although the lamellar, or bilayer, structure is generally found in cell membranes, the two hexagonal phases probably occur during some membrane transformations. The importance of molecular geometry for bilayer stability is illustrated by the effects of the phospholipase A₂ component of many venoms: they remove one fatty acid from a phospholipid to produce lysophosphatides, which ultimately destabilize bilayers relative to hexagonal phase structures. In sufficient amounts, lysophosphatides disrupt cell membranes and lyse cells. Detergents are amphipathic molecules with similar abilities to transform lipid bilayers into water-soluble micelles. In contrast to the destabilizing effects of lysophosphatides and other detergents, cholesterol stabilizes bilayers by intercalating at the interface between head and tail regions of phospholipid so as to satisfy the bulk requirements for a planar geometry.

Multilamellar bilayer structures (Fig. 2-2) form spontaneously if small amounts of water are added to solid or liquid phase phospholipids. These can be dispersed in water to form vesicular structures called liposomes. These are often employed in studies of bilayer properties and may be combined with membrane proteins to reconstitute functional membrane systems. A valuable technique for studying the properties of proteins inserted into bilayers employs a single bilayer lamella, also termed a black lipid membrane, formed across a small aperture in a thin partition between two aqueous compartments. Because pristine lipid bilayers have very low ion conductivities, the modifications of ion-conducting properties produced by membrane proteins can be measured with great sensitivity (Chap. 6).

In aqueous systems, phospholipid structures may manifest either gel, that is, rigid, or liquid-crystalline, that is, two-dimensionally fluid, properties. In the case of pure phospholipids, these states interconvert at a well-defined transition temperature, T_c, that increases with alkyl chain length and decreases with introduction of unsaturation. In cell membranes, there is marked heterogeneity in both the polar and nonpolar domains of the bilayer. Alkyl chain heterogeneity and the presence of cholesterol maintain cell membrane bilayers in the fluid state over a broad temperature range. Bilayer fluidity causes membrane lipids and proteins to diffuse rapidly within the plane of the bilayer.

Functional importance of bilayer asymmetry. Although there is rapid diffusion within the plane of a bilayer, spontaneous transverse migration of phospholipids from one bilayer leaflet to another is rare. This allows the two leaflets of a cell membrane bilayer to have very different compositions (Chap. 3). Aminophospholipids are normally confined almost exclusively to the cytoplasmic leaflet, whereas most glycolipids and sphingolipids are in the extracytoplasmic leaflet. This is accomplished by an ATP-dependent process that “pumps” the head groups of aminophospholipids toward the cytoplasmic surface. A second ATP-dependent pumping process may be involved in maintaining the extracytoplasmic orientation of phosphatidylcholine (Chap. 5). High intracellular Ca²⁺, which can arise from any condition that depletes intracellular ATP, activates a “scramblase” protein, which catalyses random transverse lipid movements. The appearance of substantial amounts of phosphatidylserine on outer cell surfaces can initiate apoptosis and phagocytosis [2].

Insertion of lipids into bilayers. Most biosynthesis of membrane lipids occurs in the endoplasmic reticulum (ER) (Chap. 3). Glycosphingolipids can segregate laterally in the bilayer to form microdomains, or “rafts,” with cholesterol [3]. In apical membranes of epithelial cells, these domains are associated with the sites of formation of small vesicles, called “caveolae” (see below), which transport cholesterol and, perhaps, other lipids to plasma membranes [4].

Most bilayer phospholipids are physically constrained by association with integral membrane proteins

In addition to interacting with each other to form the bilayer, membrane lipids may interact to varying degrees with membrane proteins [5]. Some physical measurements, such as electron spin resonance, have indicated that the acyl moieties of lipids immediately surrounding integral membrane proteins are motionally restricted and reoriented relative to the bilayer. This “annulus” fraction can comprise 20 to 90% of the total membrane phospholipid. Because the annulus lipids appear to equilibrate rapidly, within microseconds, with the bulk membrane lipids in comparison with the time scale of most enzyme-catalyzed reactions, which occur in milliseconds, the significance of such interactions has been questioned. However, phosphatidylethanol-amine is now known to be a component of the crystalline structure of the membrane protein, cytochrome oxidase [6]. Some proteins, including certain integral membrane proteins, contain domains that can interact directly and strongly with phospholipids (see below).

Diffusional flow of water directly through lipid bilayers largely accounts for the water permeability of most cell membranes

The water permeability of ion channels is estimated to account for only about 1% of the total cell water permeability. Measurements of water permeability of bilayers of varying lipid composition have ranged from 2 to 1,000 × 10⁻⁵ cm²/sec [7]. Measurements of cell membrane water permeabilities are in the same range for many cell types. Erythrocytes are a known exception, with a water permeability of about 2 × 10⁻² cm². The high water permeability of the plasma membranes of erythrocytes, kidney epithelia, certain glia and other cells results from the presence of specific membrane proteins, designated aquaporins. These are the eukaryotic members of a large and widespread family of membrane channel proteins that select water and, in some cases, admit small neutral molecules such as urea and glycerol [8]. Aquaporin-2 mediates vasopressin-sensitive water transport [9]. Aquaporin-4 is expressed in high levels in certain glial and ependymal cells [10].

The head-group regions of phospholipid monolayers facilitate lateral diffusion of protons and possibly of other ions

In model systems, pH changes have been shown to be transmitted more rapidly along these interfaces than in bulk solution. This may have particular importance in mitochondrial ATP synthesis and other processes that depend on transmembrane proton gradients [11]. This high mobility may not be restricted to protons: nuclear magnetic resonance studies have shown that the exchange of metal cations among phospholipid head groups can also be more rapid than in free solution.

Integral proteins have transmembrane domains that insert directly into the lipid bilayer

These transmembrane domains consist predominantly of nonpolar amino acid residues and may traverse the bilayer once or several times. High-resolution structural information is available for only a few integral membrane proteins, primarily because it is difficult to obtain membrane protein crystals that are adequate for X-ray diffraction measurements. Consequently, much of our knowledge of integral membrane protein structure derives from the application of various topographical mapping techniques.

Transmembrane domains are usually α helices

The peptide bond is intrinsically polar and can form internal hydrogen bonds between carbonyl oxygens and amide nitrogens, or either of these may hydrate. Within the lipid bilayer, where water is essentially excluded, peptides usually adopt the configuration that maximizes their internal hydrogen bonding, which is an α helix. A length of α helix sufficient to span the usual width of a lipid bilayer requires 18 to 21 amino acid residues (Fig. 2-3). Because the surface properties of an α helix are determined by its side chains, a single helical segment that anchors a protein by insertion into the bilayer consists largely of hydrophobic residues. Integral membrane proteins with multiple transmembrane helices may have amphipathic amino acid sequences, with the more polar residues involved in helix-helix interactions, intramembrane channel formation and other interactions. Derivation of “hydrophobicity profiles” from protein sequence data often provides major insights about transmembrane protein topography.

Proteins with one transmembrane domain may have soluble domains at either or both surfaces

Cytochrome b5 has a single hydrophobic segment that forms a hairpin loop, which acts as an anchor to the cytoplasmic surface but is thought not to penetrate the bilayer totally. It is an example of a monotopic protein. Bitopic proteins are more common, having a single transmembrane helix, which, if oriented with the N-terminus on the extracytoplasmic surface, is classified as type I or, if on the cytoplasmic surface, as type II [12] (Fig. 2-4).

Bitopic membrane proteins are often involved in signal transduction. For example, some of the receptor-activated tyrosine kinases are bitopic (Chap. 25). Agonist occupation of an extracytoplasmic receptor domain can transmit structural changes via a single transmembrane segment to activate latent kinase activity in its cytoplasmic domain.

Ion channels, transport pumps and many receptor-effector complexes are polytopic. Their predominantly hydrophobic transmembrane segments are commonly interspersed with polar and helix-destabilizing residues. Such proteins that perform tasks within the bilayer frequently have amphipathic helices that interact to form the requisite functional structures.

Transmembrane helices are usually closely packed

Two examples of this are bacteriorhodopsin and the sarcoplasmic Ca²⁺ pump. Peptide bonds have substantial dipole moments that are transmitted to the ends of α helices. This circumstance would be expected to favor close packing of antiparallel helices and is, in fact, consistent with the observed disposition of helices in bacteriorhodopsin [13]. However, intersubunit packing in oligomeric proteins can involve interactions of extramembranous protein domains and may encompass some bilayer lipids.

The fluidity of the lipid bilayer permits dynamic interactions among membrane proteins

For example, the interactions of a neurotransmitter or hormone with its receptor can dissociate a “transducer” protein, which in turn will diffuse to interact with other effector proteins (Chap. 20). A given effector protein, such as adenylyl cyclase, may respond differently to different receptors because of mediation by different transducers. These dynamic interactions require rapid protein diffusion within the plane of the membrane bilayer. Receptor occupation can initiate extensive redistribution of membrane proteins, as exemplified by the clustering of membrane antigens consequent to binding bivalent antibodies [14].

In contrast to these examples of lateral mobility, the surface distribution of integral membrane proteins can be fixed by interactions with other proteins. Membranes may also be partitioned into local spatial domains by networks of cytoskeletal proteins. This partitioning may restrict the translational motion of enmeshed proteins and yet allow rapid rotational diffusion. Examples of such spatial localization include restriction of Na⁺ pumps to the basolateral domains of most epithelial cells, Na⁺ channels to nodes of Ranvier and nicotinic acetylcholine receptors to the postsynaptic membranes of neuromuscular junctions.

Mechanical functions of cells require interactions between integral membrane proteins and the cytoskeleton

These functions include cell motility, endo- and exocytosis, formation of cell junctions and regulation of cell shape. Several different families of membrane-associated proteins mediate specific interactions among integral membrane proteins, cytoskeletal proteins and contractile proteins. Many of these linker proteins consist largely of various combinations of conserved protein-association domains, which often occur in multiple variant copies (Table 2-1).

The spectrin-ankyrin network. In erythrocytes and most other cells, the major structural link of plasma membranes to the cytoskeleton is mediated by interactions between ankyrin and various integral membrane proteins, including Cl⁻/HCO₃⁻ antiporters, sodium ion pumps and voltage-dependent sodium ion channels (Table 2-1). Ankyrin also binds to the ˜100-nm, rod-shaped, antiparallel αβ heterodimers of spectrin and, thus, secures the cytoskeleton to the plasma membrane. Spectrin dimers self-associate to form tetramers and further to form a polygonal network parallel to the plasma membrane (Fig. 2-5D). Neurons contain both spectrin I, also termed erythroid, and spectrin II, also termed fodrin. Spectrin II is found throughout neurons, including axons, whereas spectrin I occurs only in the soma and dendrites. This spectrin network further binds to actin microfilaments and to numerous other ligands. These associations are probably dynamic. For example, phosphorylation of ankyrin can alter its affinity for spectrin. Spectrin II has binding sites for microtubules. The functions of the multiple protein-interaction domains of both spectrin and ankyrin have been as yet only partially defined (see Chap. 8).

Membrane structural specializations. The spectrin-ankyrin network comprises a general form of membrane-organizing cytoskeleton within which a variety of membrane-cytoskeletal specializations are interspersed. Many of these are concerned with cell-cell or cell-matrix interactions (Chap. 7). The several morphological types of cell-cell junctions are associated with junction-specific structural and linking proteins. For example, tight junctions, also termed zona occludens, are constructed of the integral membrane protein occludin, which binds the linking proteins ZO-1 and ZO-2 [15]. These linking proteins are members of a large family, termed membrane-associated guanylyl kinase homologs (MAGUKs). The general structure of this family has, distributed from the N- to C-terminus, one or more PDZ-binding domains, a src homology 3 (SH3) domain (see Chap. 25) and a guanylyl kinase homolog domain (Table 2-1). Other members of the PDZ family are expressed in neurons at postsynaptic densities. One of these, PSD-95, contains two N-terminal PDZ domains that can bind to a motif, -E-S/T-D-V-, that occurs in N-methyl-D-aspartate (NMDA) receptors and in certain types of K⁺ channel. Multimeric clusters of these receptors or channels can be formed through disulfide cross-linking between cysteines of the N-terminal domains of PSD-95 molecules [16]. Different PDZ domains within a single linker protein can display different peptide motif selectivities. Accordingly, it has been suggested that a given linker protein may simultaneously bind to multiple different channels and receptors to produce complex clusters at various postsynaptic sites.

Certain transmembrane glycoproteins can mediate interactions between the cytoskeleton and the extracellular matrix

Many glycoproteins link sites on extracellular matrix proteins or on other cell surfaces with cytoskeletal proteins (Chap. 7). In some cases, the extracellular binding specificity is for sites found on matrix proteins, while the intracellular specificity is for a cytoskeletal protein, such as talin, which may further interact with an intermediate filament protein (Fig. 2-4) (see Chap. 8). Integrins are a major family of transmembrane receptors with these properties [17]. Their adhesion to extracellular ligands can be up- or downregulated by cytoplasmic signals and, thus, function in cell migration, cell aggregation and other intercellular interactions. Conversely, their interactions with cytoskeletal components and protein kinases can be modulated by extracellular ligands [18].

Neural cell adhesion molecules (NCAMs) belong to a widely distributed family of cell-surface glycoproteins that have extracellular domains structurally related to the immunoglobulins (see Chap. 7). They can be homotypic; that is, they can bind to each other, but they can also interact with heparin in the matrix. Differential splicing of their mRNAs can result in the expression of different polypeptides from a single NCAM gene. Two of these are transmembrane glycoproteins with identical extracellular domains and differing cytoplasmic domains. A third NCAM is not a transmembrane protein and does not participate in transmembrane signaling because it is wholly extracellular and anchored to the membrane only through a covalent attachment involving glycosylphosphatidylinositol (Fig. 2-4) (Chap. 3). The N-terminal extracellular domains of NCAMs are heavily glycosylated, and their adhesive properties can be suppressed by further addition of long polysialic acid chains (Chaps. 7 and 28).

Cadherins are Ca²⁺-dependent, homotypic adhesion proteins that may be largely responsible for the preferential adhesion of similar cell types [19]. They associate intracellularly with actin microfilaments at adherens junctions by means of linker proteins called catenins and to the intermediate filaments α-actinin and vinculin at desmosomes by means of other linker proteins called desmoplakins.

Covalently attached lipids often participate in binding proteins to membranes

Myristate can be added cotranslationally to the N-terminal glycine of a number of peripheral proteins, thus participating in binding them to the cytoplasmic membrane surface. The catalytic subunit of cAMP-dependent protein kinase, calcineurin B and NADH-cytochrome b5 reductase are myristoylated proteins [20] (Fig. 2-4).

Fatty acids, most commonly palmitate, can link as thioesters to a cysteine residue that is usually located near a membrane-binding domain. Both integral membrane proteins, such as rhodopsin and transferrin receptor, and membrane-associated proteins, such as ankyrin and vinculin, may be acylated. A number of proteins can be post-translationally prenylated [21]. One synthetic pathway for prenyl anchors involves precursor proteins with a C-terminal sequence, CXXX. A C₂₀ acyl group from geranylgeranyl pyrophosphate is transferred to the cysteine sulfhydryl. The three terminal amino acids are then cleaved, and finally, a methyl group is added to the newly exposed cysteine α-carboxyl. Prenylated proteins include many signal transducers of the small G protein class and γ subunits of heterotrimeric G proteins (see Chap. 20). Proteins can be anchored to the external bilayer leaflet by covalent linkage to complex glycosylated phosphoinositides [22]. Glycosylphosphatidylinositol (GPI)-anchored proteins include alkaline phosphatase, 5′-nucleotidase, one form of acetylcholinesterase and one form of NCAM.

Membrane associations can occur by selective protein binding to lipid head groups

One example is spectrin, which binds to cytoplasmically oriented phosphatidylinositol-4,5-bisphosphate by means of a pleckstrin-homology (PH) domain [23] (Table 2-1) (Chap. 25). Several enzymes and structural proteins become membrane-bound in response to Ca²⁺ activation. These include protein kinase C (PKC), phospholipase A₂ and synaptotagmin.

Allosteric regulation of the hydrophobicity of protein-binding surfaces frequently occurs. One of the best known cases is the Ca²⁺-dependent binding of calmodulin to other proteins (Chap. 23). Annexins are a family of proteins that exhibit Ca²⁺-dependent associations with cell membranes through direct interaction with phospholipids, and conversely, interactions with phospholipids increase their affinities for Ca²⁺ [24].

Nascent membrane proteins must be inserted through the bilayer and transported to their destinations

The information that targets a polypeptide to the ER membrane is contained in a segment near the N-terminus called a signal sequence. These sequences are highly variable but include a hydrophobic segment of nine or more residues bracketed by basic residues at the N-terminus and a mixture of acidic and basic residues at the C-terminus. Membrane proteins possess additional, predominantly hydrophobic segments, termed “topogenic sequences,” that determine their primary membrane topologies, that is, the number of times they traverse the bilayer [25].

The predominant pathway for targeting proteins to the ER in animals begins with interaction of a signal-recognition particle (SRP) with the nascent signal sequence as it emerges from the mRNA-ribosome complex (Fig. 2-6). SRP is an 11S ribonucleoprotein consisting of six different peptides and one 7S RNA. Translation is arrested if SRP binds to the complex in the absence of ER membranes. Elongation of the nascent peptide can proceed after the ribosome-bound SRP interacts with the SRP receptor or docking protein, a component of the ER membrane. This is followed by SRP dissociation from the ribosome and insertion of the signal sequence into the ER membrane, permitting the mRNA translation to continue. GTP binding and hydrolysis are required at this step. Both SRP and the SRP receptor have GTP-binding domains, but present evidence implicates binding to the receptor at this stage [26]. Once a conjunction of the ribosomal complex with the ER membrane is effected, the growing peptide passes through the membrane. This mechanism is called cotranslational insertion.

Secreted proteins are synthesized as “pro-proteins” with amino-terminal signal sequences. After the rest of the peptide has been exported, the signal sequence is cleaved from the secreted product by a signal peptidase, itself a membrane protein with its active site within the ER lumen (see Chap. 18). Membrane proteins that have a single anchoring segment near the NH₂-terminus may insert by a similar mechanism involving an uncleaved signal sequence (Fig. 2-6). Membrane proteins that have a single anchor near their COOH-terminus require a stop-transfer sequence to form a permanent transmembrane segment and, finally, cleavage of the initial signal peptide.

Many membrane proteins contain two or more topogenic sequences. An example is opsin, which traverses the membrane seven times (Chap. 47). Since it has four intralumenal domains, as many as four stop-transfer sequences may be required. By the use of selectively deleted cDNA and subsequent translation of the corresponding RNA transcripts, the experimental indications are that, in fact, the first and sixth transmembrane segments of opsin contain the stop-transfer sequences hypothesized to be necessary for proper membrane insertion [27].

Not all cases of membrane protein insertion seem to conform to this cotranslational model. Ribosomal synthesis of some integral membrane proteins does not require formation of an SRP-membrane complex. For example, the Ca²⁺ pump ATPase contains stop-transfer sequences, and although SRP is required for membrane insertion, the absence of membranes containing the docking protein does not arrest its synthesis [28]. Other integral membrane proteins, such as cytochrome b5, can be synthesized on free ribosomes and subsequently insert into membranes in the absence of SRP.

To account for all possible configurations of integral membrane proteins, it was postulated that translocator proteins in the ER membrane interact with topogenic sequences to form channels that allow hydrophilic protein segments to traverse the bilayer. Combinations of stop-transfer sequences, cleaved signal sequences and appropriate responses by the translocator proteins to these signals account for most of the observed transmembrane dispositions of polypeptide segments [29].

Channel-forming proteins with the requisite properties have been detected in rough ER: 250 pS conductance events are induced in these membranes when protein synthesis is aborted with puromycin [30]. Because puromycin causes nascent peptides to dissociate from ribosomes, ribosomes lacking such peptides can combine with translocator proteins to form and open channels in the ER. These channels remain closed in the absence of ribosomes and do not conduct ions while they are occupied by ribosomes that are generating nascent peptides (Fig. 2-7). The peptide translocator channels are formed from a heterotrimeric complex [31].

“Molecular chaperones” are frequently required to mediate correct protein folding

It was long held that nascent polypeptides assume their “native” conformations spontaneously as they emerge from the ribosome. In its strict form, this model implied that the genetic information that specifies a primary sequence completely defines the native protein conformation. This concept has been revised because of the discovery of auxiliary proteins, “molecular chaperones,” that regulate polypeptide folding [32]. The SRP and other proteins that recognize topogenic sequences are examples of this functional class.

Numerous chaperones are resident within the ER lumen. For example, calnexin is an ER-resident chaperone that complexes selectively with certain partially glycosylated proteins, including the GLUT-I glucose transporter and the α and β subunits of the nicotinic acetylcholine receptor. BiP is another ER-resident chaperone. BiP interacts with newly synthesized γ subunits of nicotinic acetylcholine receptors [33] and binds ATP. Mature nicotinic receptors are pentamers of four different subunits, α₂βγδ, in which the two α subunits are not adjacent (Chap. 11). BiP will bind to αγ and αδ complexes but not to the mature receptor. This binding apparently assists in the correct assembly of this and probably many other oligomeric membrane proteins. The peptide-binding site of BiP involves seven adjacent residues. The site specificity is rather broad and similar to the composition of the interior of folded proteins, the composition of which is not very different from that of many transmembrane segments. Peptide binding causes hydrolysis of bound ATP, which causes release of the peptide.

Newly synthesized plasma membrane proteins travel from the endoplasmic reticulum through a succession of Golgi compartments

During this time they may undergo post-translational modifications such as glycosylation, finally fusing with pre-existing plasma membrane from the trans-Golgi network [34] (Fig. 2-8). Each transit event appears to be initiated by a set of “coat proteins,” also termed “COPs” or “coatomers,” which bind to a patch of the cytoplasmic surface of a Golgi membrane to form a “coated pit.” Coated pits transform into vesicles. Scission of the vesicles from one Golgi compartment is followed by their fusion with the membrane of the next compartment.

Subunits of the intra-Golgi vesicle coat complex, termed α-, β-, γ- and δ-COP, have been identified. Among these, α-COP has similarities to clathrin heavy chain (see below) and β-COP has similarities to the clathrin-associated adaptins. One hypothesis is that the process mediated by this complex consists of a nonselective “bulk flow” of most of the membrane proteins and soluble components through successive Golgi compartments. These proteins are progressively subjected to various covalent modifications by enzymes that reside in different Golgi compartments. Thus, both the ER and the various Golgi compartments contain characteristic resident proteins. Although the “default” mechanism of vesicular traffic outward from the ER is nonselective, there are “active” and selective mechanisms for retaining the resident proteins, such as retrograde, or “salvage,” transport from Golgi to ER. A C-terminal signature sequence, Lys-Asp-Glu-Leu or “KDEL,” occurs in ER-resident proteins and is recognized by the salvage mechanism.

A different set of coat proteins, the clathrin complexes, function between the trans-Golgi network and its target membranes [35] (Fig. 2-9). Clathrin complexes consist of three molecules of clathrin heavy chain, which form a “triskelion” by joining together near their C-termini. Clathrin light chains associate with the proximal domains of the heavy chains. Each heavy chain is divided into proximal and distal domains by a hinge region. The clathrin triskelions self-assemble into extensive lattices by interacting at their heavy-chain N-termini. However, the interaction of these lattices with membrane proteins is mediated by another set of proteins, the adaptins. The combination of lattice forming by clathrin and protein selection by adaptins produces a concentration of targeted integral membranes within “coated pits” in both trans-Golgi and plasma membranes. Different adaptins reside in different membranes. For example, the adaptin AP-2 is involved in the assembly of coated pits on the plasma membrane that occurs during endocytosis, whereas AP-1 participates in the corresponding assembly on the trans-Golgi network that occurs during secretion.

Fusion of vesicles with their target membranes involves disassembly of the coat proteins. In the case of clathrin, this is mediated by an “uncoating ATPase,” identical to hsp70, a constitutive member of the heat-shock protein family, which includes several molecular chaperones. Each step of vesicular transit from the Golgi to the target membrane appears to involve both ATPase and GTPase activities.

Little is known about the mechanisms that direct membrane proteins to their ultimate targets. In both epithelial cells and neurons, membrane proteins with GPI anchors move almost exclusively to the apical plasma membrane, the neuronal analog of which is the axon [36]. Some proteins that are retained in the basolateral membranes of epithelial cells and, equivalently, in the soma and dendrites of neurons contain a cytoplasmic tyrosine-containing β-turn similar to the endocytotic signal sequence for coated vesicles (Fig. 2-9).

In addition to the clathrin-coated vesicle pathway of secretion and endocytosis, there is another pathway involving uncoated vesicles, or caveolae, that contain high concentrations of sphingolipids, gangliosides, cholesterol and the cholesterol-binding protein caveolin. Other proteins that are reported to occur in caveolae include heterotrimeric and monomeric G proteins, inositol 1,4,5-trisphosphate (IP₃) receptors and tyrosine kinases. GPI-anchored proteins, originally considered to be associated with caveolae [3], have been shown in several instances to reside in distinct membrane domains and to be internalized via conventional endosomes [37].

Neurons have special forms of vesicular transport. Some of the vesicles that bud from the trans-Golgi network are carried by fast axoplasmic transport (Chap. 28) to targets in nerve processes. Voltage-dependent exocytosis of neurotransmitters, triggered by the presynaptic influx of Ca²⁺, may occur within 200 μsec of the stimulus. Maintenance of adequate supplies of the membrane constituents that package these rapidly secreted neurotransmitters is facilitated by recycling synaptic vesicle components from the presynaptic plasma membranes (Chap. 9).

Some membrane proteins can be selectively tagged by ubiquitin for recycling or degradation

Ubiquitin (Ub) is a 76-residue polypeptide that can be covalently linked via its C-terminal carboxyl to lysines on various proteins to target them for further processing. In certain cases, the Ub tags are elongated by esterification of additional Ub molecules.

Many normal membrane proteins are subject to regulatory downregulation. For example, some receptors are endocytosed following ligand binding. A pathway for some of these proteins involves ubiquitination followed by endocytosis and lysosomal degradation (see Chap. 46). However, endocytosed membrane proteins may also be de-ubiquitinated and recycled to the plasma membrane. Various forms of stress can also initiate selective ubiquitination, leading to rapid endocytosis and degradation of certain membrane proteins. Defective membrane proteins, such as the cystic fibrosis transmembrane conductance regulator (CFTR) protein, are ubiquitinated and degraded by the proteosome pathway within the ER. It is likely that proteins enter this pathway because of improper folding. Three enzymes involved in ubiquitination occur in several differentially expressed isoforms. Different isoforms may direct molecules to different pathways. There is indirect evidence for this from studies of yeast in which the tag directing proteins to proteosomes involves poly-Ub with linkage to lys-46 of the proximal Ub. In contrast, the activation of endocytosis of a membrane protein, maltose permease, is via either a single Ub or, more effectively, poly-Ub with linkage to lys-63 of the proximal Ub [38].