Flippase.

The magnesium adenosine triphosphate (Mg ²⁺ -ATP)-dependent aminophospholipid translocase, or “flippase,” is responsible for localization of PS in the inner leaflet by rapidly translocating it from the outer to inner leaflet. The half-time for translocation is about 5 minutes at 37°C. A similar activity exists in all mammalian red cells and probably in all cells. The flippase contains a critical cysteine, because it is inhibited by sulfhydryl-modifying reagents. It is also inhibited by the pseudosubstrates of P-type ATPases, such as vanadate, and by high concentrations of Ca ²⁺ .

Unfortunately, despite almost two decades of effort, the flippase has not been definitely identified. An assortment of phospholipid transporters have been characterized in cells and organelles from bacteria to humans, and some of these have been identified in red cell membranes, including ATP8A1. ATP8A1 is an Mg ²⁺ -ATPase first described in bovine chromaffin granules. It is a P4-type ATPase, has the size and characteristics of the red cell membrane flippase, and translocates fluorescent PS in reconstitution experiments. Yeasts lacking a homologous protein, Drs2, are defective in aminophospholipid transport. But, an accessory CDC50 family protein needed for proper expression of P4-type ATPases is not found in red cells. More definitive proof—such as a gene knockout—is needed to be sure ATP8A1 is the authentic red cell flippase in mature red cells.

Floppase.

The Mg ²⁺ -ATP–requiring phospholipid translocase, or “floppase,” is responsible for movement of phospholipids from the inner to outer leaflet of the bilayer and is less specific than the flippase. The outward lipid movement by the floppase is also much slower than the inward movement of PS by the flippase. Available evidence suggests the floppase is one of the family of ABC transporters, of which the multidrug resistance transporter (ABCB1 or MDR1 or P-glycoprotein) is the best known. The best candidate is ABCC1 (MRP1). Long-term (48-hour) inhibition of the activity in red cells leads to loss of PC and SM from the outer bilayer but no change in the distribution of PE or PS. Additionally, red cells from mice lacking the gene equivalent to ABCC1 can no longer translocate short-chain fluorescent phospholipid analogues of PC, SM, and PS from the inner to the outer bilayer. However, it is not yet certain that ABCC1 can transport full-length, native phospholipids, which is a required property of the physiologic floppase.

Scramblase.

The Mg ²⁺ -ATP–independent, Ca ²⁺ -dependent phospholipid “scramblase” causes a rapid, bidirectional movement of all phospholipids in the bilayer, resulting in the collapse of phospholipid asymmetry, which may occur in less than a minute under some circumstances. The exposure of PS on the cell surface causes red cells to adhere to each other and to endothelial cells and phagocytes and triggers clotting (see later).

After several frustrated attempts, it appears the scramblase has been cloned, or at least part of the scramblase. A protein called anoctamin 6 or TMEM16F (transmembrane protein 16F) is mutated and absent from patients with Scott syndrome, a very rare, autosomal recessive disorder characterized by decreased scramblase activity (see Chapter 30 ). As a consequence, activated platelets fail to expose sufficient PS on their surface to assemble prothrombinase and affected individuals have a bleeding tendency. Erythrocytes and other cells also lack scramblase, but without apparent consequence. However, although anoctamin 6 is responsible for Scott syndrome, it does not account for all scramblase activity. Knockout of anoctamin 6 in the mouse impairs but does not abolish scramblase activity, and other experiments show that there are at least two pathways of scramblase induction. In platelets, anoctamin 6 is not required for, but enhances, the scramblase pathway induced during apoptosis but is required by the pathway induced by thrombin and the rattlesnake venom convulxin. The latter pathway is the one that is defective in Scott syndrome. This may be explained by the fact that at least 5 of the 10 members of the anoctamin/TMEM16 family have lipid scramblase activity. More needs to be learned about which members are expressed in different cells and how they are triggered.

Phospholipid Asymmetry is Lost in Sickle and Thalassemic Red Cells.

Phospholipid flip-flop is increased and phospholipid asymmetry is lost in sickle cells, presumably owing to decreased phospholipid translocase activity and uncoupling of the lipid bilayer from the underlying skeleton in the sickle cell spicules. Consequently, subpopulation of circulating sickle red cells expose PS on their surface. There is large variability in the numbers of such cells among sickle cell patients (0.5% to >10%). A subpopulation of thalassemic red cells also expose PS on their surface, and phospholipid asymmetry is altered in diabetic red cells which, like sickle cells, tend to adhere to endothelial cells.

Externalized PS Triggers Clotting and Phagocytosis.

Cells sequester aminophospholipids on the cytoplasmic side of the plasma membrane because the exposed lipids, particularly PS, trigger the conversion of prothrombin to thrombin and promote intravascular clotting. Loss of phospholipid asymmetry also activates the alternate complement pathway.

In addition, PS is an “eat me” signal. Macrophages bind and ingest cells, or even liposomes, that have PS exposed on their outer surfaces. Exposure of PS is one of the earliest events in apoptosis, and the PS receptor seems to have evolved to scavenge dead or dying cells that have activated scramblase during apoptosis. Expelled erythroblast nuclei also display PS on their surface and are removed by this mechanism.

A protein called PSR was initially believed to be the PS receptor, however that proved incorrect. Three different PS receptors have been identified that appear to be real: Tim4, Bal1, and stabilin-2. Bal1 functions as the receptor for the ELMO/Dock/Rac pathway that directs the actin cytoskeleton to engulf apoptotic corpses. Stabilin-2 interacts with the intracellular adaptor GULP to activate Rac1 and coordinates with Bal1. Tim4 does not appear to signal but serves primarily as a tethering receptor. The interplay between these receptors and other parts of the phagocytic system is fascinating but complex, and its discussion is beyond the scope of this chapter.

Protein-Bound Lipids.

There is evidence that some transmembrane proteins interact with neighboring “boundary” lipids in a dynamic fashion whereas others bind lipids so tightly that they resist detergent extraction. Presumably the latter proteins bear high-affinity lipid-binding sites. In either case, but particularly when lipids are tightly bound, they must slow the next layer of lipids that they interact with and so on, so that integral proteins exert a kind of “crowd control” on the lipids in the vicinity.

The structure, and in some cases the function, of integral membrane proteins is dependent on the surrounding phospholipids. Membrane proteins can be parochial in their choice of lipid partners. For example, specific lipids are required for the enzymatic functions of the sodium and calcium pumps whereas glycophorin A (to be discussed), binds anionic (PS, PI) but not choline (PC, SM) phospholipids, and those lipids are relatively tightly bound. Positively charged amino acids are concentrated on the cytoplasmic side of the bilayer-spanning domains of membrane proteins, in part because their position determines the orientation of the protein. Although it is not well documented, anionic phospholipids probably cluster near these regions of positive charge. Surface-bound membrane proteins such as spectrin and protein 4.1R, which bind preferentially to anionic phospholipids, may also contribute to the nonrandom topography of inner membrane phospholipids. And, conversely, the lipids appear to contribute to the stability of the membrane skeleton.

Membrane Lipid Rafts.

Lipid rafts are domains within the lipid bilayer that were initially defined by being insoluble in cold, nonionic detergents but have since been defined by multiple other techniques. They are present in all cells and have been implicated in a variety of sorting and signaling processes. They are rich in cholesterol, sphingomyelin, and glycosylphosphatidylinositol (GPI)-anchored proteins; have more ordered acyl side chains; and are thicker than nonraft bilayer domains. Cholesterol prefers the straighter and stiffer acyl chains of sphingomyelin to more disordered phospholipid side chains and is enriched in the rafts. The extra length of glycosphingolipid side chains increases the thickness of the rafts and promotes further segregation. Various measurements indicate that lipid rafts are dynamic, nanoscale (10 to 20 nm) structures that assemble reversibly into larger structures, often for functional reasons. Red cells lipid rafts are 100 to 300 nm and have an irregular shape.

Red cell lipid rafts contain a complex mixture of membrane proteins, including stomatin, flotillin-1 and flotillin-2, the Duffy protein receptor (DARC), heterotrimeric Gα _s GPI-anchored proteins such as CD55, CD58, and CD59, and at least some band 3 molecules. Proteins with myristate or palmitate side chains also tend to partition into the rafts, whereas proteins with prenyl side chains are excluded.

Because the glycolipid anchor of GPI-linked proteins is only embedded in the outer leaflet of the bilayer, GPI-linked proteins are much more laterally mobile than transmembrane proteins that interact with the membrane skeleton (discussed later). This high mobility may be important in recruiting complement regulatory proteins to sites of complement activation. Defective biosynthesis of the GPI anchor precludes attachment of GPI-linked proteins to the membrane, causing increased susceptibility to hemolysis by complement, as clinically manifest by paroxysmal nocturnal hemoglobinuria (see Chapter 14 ).

Depletion of cholesterol from the red cell membrane blocks raft formation and the ability of raft-associated proteins to form membrane microdomains. It appears there are different kinds of lipid rafts in the membrane, with different protein components and different fates, but this is not well defined. Although the function of the rafts in normal red cell physiology remains to be determined, they seem to be critical for the invasion of red cells by malarial parasites.

Purified Lipids Exist in Gel or Liquid Crystalline States.

Purified phospholipids exhibit discrete, liquid crystalline to gel phase transitions that are dependent on the length and degree of unsaturation of their acyl side chains. The presence of a double bond in the acyl chain increases the disorder and flexibility all along the hydrocarbon chain, especially toward the terminal methyl end. Below the temperature of the liquid crystalline to gel transition, acyl chains of purified lipids are extended in stiff, parallel, hexagonally packed arrays that are more nearly solid than liquid.

Cholesterol Creates an Intermediate State between Gel and Liquid Lipids.

Cholesterol buffers the extremes of the gel and liquid crystalline states. At temperatures above the gel-liquid phase transition, cholesterol partially immobilizes the acyl chains of phospholipids, particularly the proximal eight to ten carbons that are adjacent to the bulky steroid nucleus. Below this transition, cholesterol disrupts the ordering of the acyl chains, in effect preserving membrane fluidity. The net result is that cholesterol tends to abolish the gel to liquid crystalline transition and create a condition of intermediate fluidity in which the proximal ends of the hydrocarbon chains are extended and relatively rigid and the distal ends are disordered and fluid.

The red cell membrane contains relatively large amounts of cholesterol (cholesterol/phospholipid = 0.8 mol/mol) and is relatively less fluid than other plasma membranes. The inner bilayer is somewhat more fluid than the outer, probably because it contains more unsaturated fatty acids.

Mature Erythrocytes Cannot Synthesize Lipids.

Mature erythrocytes are unable to synthesize fatty acids, phospholipids, or cholesterol de novo and depend on lipid exchange and fatty acid acylation as mechanisms for phospholipid repair and renewal ( Fig. 15-4 ). Damaged chains are removed by phospholipase A ₂ , producing a lysophospholipid. Plasma-derived fatty acids are esterified to acyl-CoA by acyl-CoA synthetases, especially ACSL6, and transferred to the lysophospholipids by acyl-CoA: lysophospholipid acyltransferases. In theory this enzyme should facilitate the renewal of lost or damaged fatty acid side chains and should prevent the accumulation of deleterious lysophosphatides within the membrane; however, it is not certain that the phospholipases necessary to remove damaged fatty acids operate in the red cell.

Pathways for lipid exchange and fatty acid replacement in the red cell membrane.

Chol, Cholesterol; FA, fatty acid; Lyso-PC, lysophosphatidylcholine; PC, phosphatidylcholine; PS, phosphatidylserine; PE, phosphatidylethanolamine.

This renewal pathway, although limited, permits a slow replacement of membrane lipid components. Approximately 30 days are required before erythrocyte lipids reach equilibrium after a change in dietary fatty acids. Theoretically this could be pathologically significant in persons on unusual diets (e.g., infants receiving a high intake of medium-chain triglycerides); however, to date, no pathologic consequences have been reported.

Esterification Depletes Red Cell Cholesterol.

Red cell membrane cholesterol (unesterified) exchanges readily with the unesterified cholesterol in plasma lipoproteins (t _1/2 = 7 hr), where it is partially converted to esterified cholesterol by the action of lecithin:cholesterol acyltransferase (LCAT). Because the newly formed cholesteryl ester cannot return to the red cell membrane, LCAT catalyzes a unidirectional pathway (see Fig. 15-4 ) that depletes the membrane of cholesterol and decreases its surface area. Conversely, if LCAT is absent or inhibited, excess membrane cholesterol accumulates, expanding the membrane surface area and resulting in target cells.

The N-Terminal Sequence Binds Glycolytic Enzymes, Peroxiredoxin-2, and Deoxyhemoglobin and is an Oxygen-Regulated Membrane Metabolon.

The first 23 amino acids of band 3 contain two tandem repeats terminating in the sequences DDY ₈ ED and EEY ₂₁ ED. G3PD, phosphofructokinase (PFK), aldolase and deoxyhemoglobin all bind in this region. So does peroxiredoxin-2. Aldolase requires both sites to bind, G3PD binds separately to each, and PFK and deoxyhemoglobin bind to the more distal site. In confirmation, the red cells from a child with a mutant band 3 that lacks the first 11 amino acids (Bd3 Neapolis) contain no membrane bound aldolase.

The glycolytic enzyme complex also binds to a site near the junction of the cytoplasmic and membrane domains of band 3 (amino acids 356 to 384), which lies close to the N-terminus in the folded protein. Enzymes in the glycolytic complex can be crosslinked to proteins in both the ankyrin complex and the actin junctional complex, which suggests the glycolytic enzymes attach to band 3 dimers in both complexes (see Fig. 15-6 ).

There are more than enough band 3 molecules to bind all the peroxiredoxin-2 (~200,000 copies), G3PD (~500,000 copies), aldolase (~20,000 copies), and PFK (~6,000 copies). About 65% of G3PD, 40% of aldolase, and 50% of PFK are membrane bound in the intact red cell. PGK may also attach to band 3, but the numbers of bound molecules are small (~230 copies).

Deoxyhemoglobin binding is well documented. Its affinity is weak (kD _{(deoxy Hb)} = 0.3 mM), but because the concentration of hemoglobin is so high in red cells, roughly half of the band 3 molecules are bound to deoxyhemoglobin under physiologic conditions. Hemichromes, a partially denatured form of hemoglobin, bind more avidly than hemoglobin and cause band 3 to cluster, which is believed to target the cell for phagocytosis (see later section, “ Membrane Changes Associated with Aging and Senescence of Red Cells ”).

G3PD, PFK, and aldolase are inhibited when attached to the membrane. In addition, lactic dehydrogenase (LDH) and pyruvate kinase (PK) also localize to the membrane, even though they do not bind to band 3. All five enzymes are displaced into the cytoplasm by an antibody to the N-terminus of band 3, by deoxyhemoglobin, or by p72 ^syk phosphorylation of tyrosines 8 and 21, which lie in the middle of the two binding sites. Similarly in mice, even though the acidic N-terminal domain of band 3 is not well conserved, glycolytic enzymes bind to the red cell membrane in a band 3–dependent manner and redistribute to the cytoplasm when hemoglobin is deoxygenated and bound to band 3 or band 3 is deleted.

The data suggest that many of the enzymes in the glycolytic pathway form a complex—or more accurately a series of complexes because the wide variation in numbers of enzyme molecules means they cannot form a single stoichiometric complex. The complexes (or “metabolons”) are bound to band 3, and their interaction can be regulated by tyrosine phosphorylation and the state of hemoglobin oxygenation, and perhaps by substrates and cofactors of the enzymes ( Fig. 15-8 ). The enzymes are in discrete clumps along the membrane, which also supports the idea of a metabolon (see Fig. 15-8, A ). In addition, when metabolic fluxes are measured in intact normal red cells, activating glycolysis by displacing the glycolytic enzymes from band 3 increases glycolytic fluxes by 45% and decreases utilization of the pentose phosphate pathway by 66%.

Glycolytic metabolon.

A, Fluorescent imaging of band 3 and aldolase on the left and phosphofructokinase (PFK) , lactate dehydrogenase (LDH) , and pyruvate kinase (PK) on the right. Note that band 3 is uniformly and exclusively in the membrane whereas the five enzymes are mostly attached to the membrane (some cytoplasmic staining, especially for LDH) and have a punctate appearance as though they are clustered. Glyceraldehyde-3-phosphate dehydrogenase (G3PD) stains similarly but is not shown. Aldolase, PFK, and G3PD bind to band 3; PK and LDH do not. However, all are displaced by a fragment of the band 3–binding site, by phosphorylation of the site, or by deoxyhemoglobin, plus all can be cross-linked to one another, indicating that all are bound in a complex or “metabolon.” B, Discrete, discontinuous submembrane pools of a fluorescent analogue of adenosine triphosphate (ATP) , synthesized by phosphoglycerate kinase. C, Schematic of the postulated glycolytic metabolon. The enzyme complex binds to the amino terminus of band 3 (right) and to a site near the junction between the cytoplasmic and membrane domains (not shown). When the complex is bound to band 3 (right), the enzymes are inactive. The complex can be displaced and activated (left) by phosphorylation of Tyr8 and Tyr21 on band 3 or by binding of deoxyhemoglobin. Some of the ATP produced is retained at the membrane and used by the Na ⁺ ,K ⁺ -ATPase and Ca ²⁺ -ATPase. F-6-P, Fructose-6-phosphate; Hb, hemoglobin.

( A , From Campanella ME, Chu H, Low PS: Assembly and regulation of a glycolytic enzyme complex on the human erythrocyte membrane. Proc Natl Acad Sci U S A 102:2402–2407, 2005; B , From Chu H, Puchulu-Campanella E, Galan JA, et al: Identification of cytoskeletal elements enclosing the ATP pools that fuel human red blood cell membrane cation pumps. Proc Natl Acad Sci U S A 109:12794–12799, 2012.)

The obvious reason for a complex of glycolytic enzymes would be to increase the efficiency of glycolysis by channeling glycolytic intermediates from one enzyme to the next (see Fig. 15-8, C ). The complex is conveniently located near band 3, which transports the phosphate needed to make ATP. Band 3 also transports phosphoenolpyruvate, pyruvate, and lactate, all substrates or products of the putative complex. The immediate modification of substrates on the inner membrane surface also prevents back-flux through transport channels, increasing the efficiency of transport.

There is strong evidence that human red cell membranes compartmentalize ATP (see Fig. 15-8, B ). The membrane ATP pool is in contact with β spectrin, ankyrin, band 3, and G3PD, and it fuels the Ca ²⁺ -ATPase and Na ⁺ ,K ⁺ -ATPase pumps and presumably the phospholipid flippase.

Interestingly, the existence of a membrane bound complex of glycolytic enzymes was postulated more than 45 years ago and was indirectly observed more than 35 years ago, but the interpretation was not generally accepted. The membrane ATP pool is another discredited old idea that has been revived.

The Cytoplasmic Domain Is Where the Membrane Skeleton Attaches.

Ankyrin links the major membrane skeletal protein spectrin to the plasma membrane of the red cell by binding to a pair of adjacent beta hairpin loops in the cytoplasmic domain of band 3 formed by amino acids 63-73 and 175-185 ( Fig. 15-9 ; see also Fig. 15-7 ). There is considerable evidence that ankyrin also interacts with the acidic N-terminal domain of band 3. First, blocking the N-terminus with an antibody inhibits ankyrin binding. Second, association of ankyrin with band 3 is blocked by phosphorylation of tyrosines 8 and 21, and, vice-versa, ankyrin binding prevents phosphorylation of these tyrosines. As mentioned earlier, deoxyhemoglobin also binds to the N-terminus. In doing so it displaces ankyrin and weakens the membrane. Stefanovic and colleagues postulate that this could improve blood flow during brief periods of deoxygenation but might promote unwanted membrane vesiculation during prolonged deoxygenation, as in sickle cell disease.

Structure of the band 3 cytoplasmic domain.

The amino acid backbone of the band 3 dimer is shown. One monomer is brown, the other green. Note how the monomers interlock in the center. The two loops that protrude from the surface of the molecule to form the ankyrin-binding site (amino acids 63 to 73 and 175 to 185) are yellow.

Ankyrin preferentially binds to band 3 “tetramers.” Whether band 3 spontaneously forms tetramers and then attracts ankyrin, stabilizing the tetramers, or whether ankyrin, which has two binding sites for band 3, independently binds two band 3 dimers (i.e., a “dimer of dimers”) is unresolved. The fact that a stable subpopulation of band 3 tetramers can be isolated from nonionic detergent extracts of ghosts argues for the former. The fact that at least one structure of the band 3/ankyrin complex precludes a tetramer argues for the latter. It is estimated that 40% of band 3 is bound to ankyrin as tetramers (120,000 tetramers) whereas 60% exists as dimers (360,000 dimers).

Protein 4.1R, which links the spectrin-actin–based skeleton to the plasma membrane in part via interactions with glycophorin C, binds to the cytoplasmic domain of band 3 at specific (I/L)RRRY sequences near the inner membrane surface (see Fig. 15-7 ). The interaction is important because zebrafish with a severe hemolytic anemia due to lack of band 3 can only be rescued by band 3 that contains IRRRY sites. Rescue is markedly reduced if mouse band 3 containing a mutation in either the LRRRY or IRRRY site is used and is nearly abolished if both sites are mutated. Because protein 4.1R and ankyrin compete for binding to band 3 and ankyrin binds exclusively to band 3 tetramers, it seems likely that protein 4.1R binds preferentially to dimeric band 3 (see Fig. 15-6 ).

Protein 4.2, a 77-kD peripheral membrane protein of the red cell, also interacts with the cytoplasmic domain of band 3. The binding site is not well defined. Red cells from patients with the cytoplasmic band 3 mutations Glu40→Lys, Gly130→Arg, and Pro327→Arg show some deficiency of protein 4.2, and the mutant band 3 molecules bind protein 4.2 poorly in vitro. Pro327 and Gly130 are not too close to each other in the crystal structure of band 3 so either there is more than one binding site or one of the mutations induces more global effects on band 3 structure. It is not clear whether protein 4.2 preferentially binds to the ankyrin-linked band 3 tetramers or to band 3 dimers, but, as discussed later in the section on protein 4.2, it probably binds to both. Because protein 4.2 binds to the carboxy-terminus (C-terminus) of α spectrin, near the spectrin-actin junction, it appears that band 3 may attach the skeleton to the lipid bilayer at both the head end of spectrin, where ankyrin binds, and at the tail end, where actin, proteins 4.1R and 4.2, and adducin bind (see Fig. 15-6 ). Whatever the arrangements, protein 4.2 depends on band 3 to be expressed in the membrane. Membranes from band 3 knockout mice completely lack 4.2, and membranes from 4.2 knockout mice are partially deficient in band 3.

Adducin, is an actin-binding protein that caps actin filaments and enhances the binding of spectrin to actin. It is discussed later in the chapter. Adducin binds to the cytoplasmic domain of band 3 via its extended, flexible tail domain (see Fig. 15-6 ). The tails of both α and β adducin bind band 3, but the free β-adducin tail domain disrupts red cell membrane morphology and stability more than the α-adducin tail, suggesting it is the more important interaction physiologically. Mice lacking all adducins have a mild spherocytic hemolytic anemia but have normal amounts of band 3 and spectrin.

Band 3 Tyrosine Phosphorylation Weakens Ankyrin Binding.

Band 3 is phosphorylated on tyrosine residues 8, 21, 359, and 904 by p72Syk kinase and to a less extent by Lyn kinase. This displaces the glycolytic enzyme complex, as noted earlier, but it also greatly weakens band 3/ankyrin binding, which in turn promotes band 3 diffusion within the lipid bilayer and red cell spiculation and vesiculation. The effect on the interactions of band 3 with its other binding partners is unknown. In the basal state, band 3 tyrosines are unphosphorylated, owing to the action of phosphatases, but they become phosphorylated after malaria invasion and by oxidative damage, such as occurs in hemoglobinopathies and glucose-6-phosphate dehydrogenase (G6PD) deficiency.

Flexible Link between the Cytoplasmic and Membrane Domains.

Kurtz and his colleagues have analyzed hundreds of high-resolution electron micrographs of single band 3 dimers and find that cytoplasmic and membrane domains orient differently in different molecules, strongly suggesting that the connecting region is flexible. The fact that loss of a portion of this linker in Southeast Asian ovalocytosis (see Chapter 16 ) results in very rigid red cell membranes suggests that the link may be important in regulating red cell elasticity.

Kidney Band 3.

Band 3 is also expressed in the basolateral membrane of the type A intercalated cells in the kidney collecting ducts where it collaborates in the excretion of H ⁺ and reclamation of HCO ₃ ⁻ . The kidney isoform lacks the first 65 amino acids and does not bind ankyrin.

The 52-kD Membrane Domain (Amino Acids 360 to 911) Contains an Anion Exchange Channel.

The anatomy of the membrane domain of band 3 has been an active area of interest, with the goal of research efforts focused on understanding how the membrane domain’s topology enables the 1.2 million channels in each red cell to exchange 10 to 30 billion HCO ₃ ⁻ or Cl ⁻ anions per second. The 6 nm long × 11 nm wide × 7 nm deep membrane domain contains numerous transmembrane and intramembrane helices, some of which are tilted in a V shape like they are in the bacterial ClC chloride channel. Multiple topology models of the membrane domain have been derived over the years using different techniques to see which amino acids are accessible from the exterior and interior of the cell. The latest and most plausible one is based on homology between band 3 and the ClC channel (see Fig. 15-7 ).

Despite considerable effort, we still do not know which transmembrane segments form the anion channel or exactly how the channel functions. Residues in helices D, F, H, I, N, and R are predicted to contribute to the transport process. The clustering of mutations responsible for various disorders of red cell membrane permeability between helices K and N suggests that region is also involved (see Fig. 16-35 ).

Carbonic anhydrase IV (CAIV), a GPI-linked protein, binds to the outside of the membrane domain, probably just beyond the glycosylation site (Asn642) in the protein loop that connects helices I and J (see Fig. 15-7 ).

Carbonic anhydrase II (CAII) may or may not bind to band 3, on the C-terminal extension at the inside surface of the channel (see Fig. 15-7 ). Assuming it does, there is evidence that it increases HCO ₃ ⁻ /Cl ⁻ exchange and helps create an acidic microdomain around band 3 during HCO ₃ ⁻ transport. The two carbonic anhydrases potentially form a CO ₂ /HCO ₃ ⁻ transport metabolon with band 3.

The Transport Channel Exchanges Chloride and Bicarbonate Rapidly (~20,000 to 50,000 Times per Second per Channel), Facilitating CO ₂ Transport from the Tissues to the Lungs.

In the peripheral tissues, CO ₂ is either rapidly converted to HCO ₃ ⁻ at the surface of band 3 by CAIV and enters the red cell via band 3 or it diffuses across the red cell membrane and is then converted to HCO ₃ ⁻ by CAII. The H ⁺ by-product of the carbonic anhydrase reaction binds to hemoglobin and facilitates O2 release to the tissues (i.e., the Bohr effect). The whole process is reversed in the lungs ( Fig. 15-10 ). As the red cell circulates, intracellular HCO ₃ ⁻ is exchanged for extracellular chloride. As a result, HCO ₃ ⁻ is transported in the plasma as well as within the red cell, which increases CO ₂ transport from the tissues to the lungs by approximately 60%. The HCO ₃ ⁻ /Cl ⁻ exchange is believed to occur by a “ping-pong” mechanism, whereby an intracellular anion enters the transport channel and is translocated outward and released, with the channel remaining in the outward configuration until an extracellular anion enters and activates the reverse cycle.

Function of band 3 in O ₂ and CO ₂ transport.

A, Tissue CO ₂ is converted into HCO ₃ ⁻ and H ⁺ in the red cell by carbonic anhydrase. Band 3–mediated exchange of the HCO ₃ ⁻ for Cl ⁻ (Hamburger shift) allows HCO ₃ ⁻ to be carried in the plasma as well as the red cell and increases CO ₂ uptake by about 60%. H ⁺ binds to hemoglobin and facilitates O ₂ release (Bohr effect) and tissue oxygenation. B, These processes are reversed in the lungs, which promotes hemoglobin oxygenation and release of CO ₂ .

(Modified from Weith JO, Anderson OS, Brahm J, et al: Chloride-bicarbonate exchange in red blood cells: physiology of transport and chemical modification of binding sites. Philos Trans R Soc Lond [Biol] 299:383, 1982.)

Although HCO ₃ ⁻ and Cl ⁻ are the predominant physiologically relevant anions exchanged by band 3, the specificity of the channel is actually quite broad and larger anions such as sulfate, phosphate, pyruvate, and superoxide are also transported, albeit at much slower rates. Several potent and experimentally useful inhibitors of band 3 anion transport are also known, with the most commonly employed being stilbene disulfonates, such as 4,4′-diisothiocyanostilbene-2,2′-disulfonate (DIDS).

The Extracellular Surface of the Band 3 Membrane Domain is the Site of Specific Blood Group Antigens.

Asn642, located in the extracellular loop between helices I and J, is derivatized with a complex lactosaminoglycan that is variable in length and contains the I/i blood group antigens (see Fig. 15-7 ). In fetal cells, the carbohydrate chain is unbranched and has i specificity, whereas in adults there is a branched structure that yields I reactivity. Extracellular domains of band 3 also contain the antigens of the Diego and at least 13 other blood groups.

The Integral Membrane Protein Glycophorin A (GPA) and the Membrane Domain of Band 3 are Functionally Associated.

The Wright (Wr) blood group antigen is formed by the association of GPA and band 3. Glu658 in band 3 is required for normal expression of this blood group antigen, and an Ala65→Pro mutation in GPA causes aberrant Wright antigen expression. Furthermore, a monoclonal antibody directed against the Wr ^b antigen immunoprecipitates both proteins, and anti-GPA antibodies immobilize both GPA and band 3 in the membrane.

GPA also has a chaperone-like role in bringing band 3 to the plasma membrane. The coexpression of GPA and band 3 facilitates the expression of band 3, red cells lacking band 3 also lack GPA, and mutant band 3 molecules that are stuck in the endoplasmic reticulum can be rescued by coexpression of the mutant band 3 with GPA. GPA also enhances band 3 transport. SO ₄ ²⁻ transport is diminished 60% in the absence of GPA.

Defects in GPA, GPB, and GPE Give Rise to Nearly 40 Variant Phenotypes of the MNS Blood Group System.

Complete loss of GPA, GPB, or both, often by gene deletion from unequal recombination of the neighboring loci, gives rise to En(a−), (S−s−), and M ^k M ^k red cells, respectively ( Fig. 15-12 ). En(a−) red cells, for example, are completely deficient in GPA and have weak MN blood group antigens. The affected red cells compensate for a potential 60% loss of surface charge by increasing the glycosylation of band 3, such that the overall surface charge is only reduced by approximately 20%. Given the proximity and homology of GPA, GPB, and GPE genes, hybrid glycophorin molecules may be generated by recombination events during meiosis. In the MiV variant, for example, Lepore-like recombination yields a fusion gene containing the N-terminus of GPA and C-terminus of GPB. In the St ^a variant, anti-Lepore recombination produces the reverse hybrid (N-terminal GPB with C-terminal GPA), which is inserted between the normal GPA and GPB genes, neither of which is lost. Interestingly, none of the GPA and GPB variants produces detectable changes in erythropoiesis or in the shape, function, or life span of the affected red cells. Rare individuals with complete absence of GPA and GPB apparently exhibit no erythrocyte abnormalities. Similarly, GPA knockout mice have a normal phenotype except that they have 60% less O-linked oligosaccharides than normal red cells. Thus either GPA, GPB, and GPE do not serve any critical basal functions in red cells or such functions are subsumed by other protein(s), such as GPC or band 3, when GPA, GPB, and GPE are deficient. If GPA really has no nonredundant function, it is surprising that deletions and other mutations are not more common, because GPA is a major receptor for P. falciparum malaria (see later) and must be under intense evolutionary pressure.

Glycophorin variants.

Organization of the glycophorin A (GPA) , B (GPB) , and E (GPE) genes on chromosome 4 and probable deletions in the En(a−), S−s−U−, M ^k , Miltenburger type V (MiV) , and Stones (St ^a ) variants. The MiV and St ^a variants arise from unequal crossing over, analogous to the Lepore hemoglobins. In MiV red cells the normal GPB gene is replaced by a fusion gene (Lepore-type recombination). In St ^a erythrocytes the product of the reciprocal exchange is inserted between the GPA and GPB loci but no normal genes are lost (anti–Lepore-type recombination).

(Modified from Vignal A, London J, Rahuel C, et al: Promoter sequence and chromosomal organization of the genes encoding glycophorins A, B, and E. Gene 95:289–293, 1990.)

Variations in GPC and GPD Affect the Gerbich Antigen System.

Mutations in the extracellular domain of GPC and GPD alter the antigenicity of red cells (Gerbich antigen system) but do not affect cellular mechanical properties. In contrast, in the Leach phenotype there is complete loss of GPC and GPD secondary to deletion of exons 3 and 4 or a frameshift mutation. Affected individuals do not have significant hemolysis, but their red cells are slightly elliptocytic and exhibit decreased membrane deformability and mechanical stability, probably because GPC-deficient cells are partially deficient in protein 4.1R and p55. Individuals with homozygous protein 4.1R deficiency have a more profound deficiency of GPC (~70%) in their elliptocytic red cells, underscoring that the cellular content of each component of the ternary complex depends on the expression level of the others.

The Spectrin Family in Humans is Encoded by at Least Seven Distinct Genes.

Two genes encoding for α spectrin and five genes encoding for β spectrin have been characterized. The αI-spectrin gene (SPTA) is located on chromosome 1q22-q23 adjacent to the Duffy blood group and encodes the erythroid-spectrin α chain. The αII-spectrin (also called α-fodrin) gene (SPTAN1) encodes an α-spectrin homologue found in nearly all cells except mature red cells. The βI-spectrin gene (SPTB) is located on chromosome 14q23-q24.2 and encodes the β subunit of erythroid spectrin, designated βI, isoform 1 (or βIΣ1). The βI gene promoter region contains specific GATA1 and CACCC-related protein binding sites that account for specific, high-level expression in erythroid cells. Differential 3′-processing of the βI-spectrin pre-mRNA generates an alternative, longer βI-spectrin subunit (βI, isoform 2 or βIΣ2) that is found in brain and muscle. The βII-spectrin (also called β-fodrin) gene (SPTBN1) encodes the generally expressed β subunit of spectrin, which shares 60% amino acid identity with βI spectrin. Thus erythrocyte spectrin is composed of αI and βI (isoform 1) chains and the widely expressed spectrin (called fodrin, tissue spectrin or brain spectrin) contains αII and βII subunits. Muscle spectrin is believed to contain predominantly αII and βI (isoform 2) subunits.

βIII and βIV spectrins are predominantly expressed in the central nervous system. Gene mutations cause neurologic diseases. A truncated isoform of βIV spectrin is associated with a subnuclear structure (promyelocytic leukemia bodies) and may be part of a nuclear scaffold. βV spectrin is mostly in retinal rods and cones and in gastric epithelial cells. Other spectrin-related proteins—dystrophin, utrophin, actinins, plakins, and dystonin among others—are not relevant to red blood cells.

Spectrins Consist Mostly of Tandem 106-Amino Acid Repeats.

Each spectrin chain is composed of a series of approximately 106-amino acid repeats. Each repeat is formed by three α-helices that are folded into a triple helical bundle. The α and β spectrins contain 20 and 16 full repeats, respectively. An SH3 domain within the α chain is conventionally called repeat α10, although it is not a true spectrin repeat (see Fig. 15-13, A ). As will be discussed, the α0 and β17 repeats are complementary partial repeats that join to form a full repeat and link spectrins together. The primary sequence of each repeat exhibits extensive heptad (seven-residue) symmetry, with conserved hydrophobic residues situated in the first and fourth positions of each heptad motif (designated as positions a and d in Fig. 15-14, A ). Each triple helical spectrin repeat is approximately 5 nm long and 2 nm wide (see Fig. 15-14, B ) and is rotated 60 degrees (right-handed) relative to the neighboring repeats. Details of the structure are presented in Figure 15-14 and in catalogs of spectrin repeat structures.

Structure of spectrin repeats.

A, Consensus sequence of a typical 106-amino acid, human erythrocyte spectrin repeat phased according to crystallographic data, which identifies three α helices per repeat. Less-conserved amino acids are marked with a dash. The amino acids in the repeat are designated “a” through “g.” The distance from one “a” to the next corresponds to two turns of the α helix. Residues “a” and “d” are the major contact points between helices. They tend to be hydrophobic and are conserved in most spectrins. B, Model of a single typical repeat. The positions of the nearly invariant tryptophans at A17 (Trp at amino acid 17 of the repeat) and at C15 (Trp at position 90) are shown. They interact with surrounding residues, including B18 (repeat amino acid 54), at the central junction where the A helix crosses over from B to C. C, Cross-section of a typical repeat. Note that the A, B, and C helices are in a triangular array and that the “a” and “d” residues lie on one face of each helix. The side chains of these amino acids are usually hydrophobic (Φ) and interact with each other to stabilize the triple helical configuration. Salt bonds between the typically polar amino acids at positions “e” and “g” also help attach the BC and AC helices to each other. D, Interconnection of two adjacent repeats. The A and C helices are colinear, forming one long helix. The distal repeat of each pair is rotated 60 degrees (right-handed). Note that the B helix of the proximal repeat overlaps the A helix of the following repeat. Interactions in the overlap region among conserved hydrophobic residues, such as C29, B4, and B5 (repeat amino acids 104, 40, and 41, respectively), and the mostly hydrophobic residues at A2 (repeat amino acid 2) probably stabilize the connection and limit mobility at the repeat junction.

Although spectrin is often depicted as a highly flexible molecule, the repeats themselves appear rigid and the crystal structures of multiple repeats do not show significant flexibility at the junction between repeats (see Fig. 15-14, D ). However, the stability of individual repeats varies and some are unstable at physiologic temperatures, which would introduce flexibility. Two of the less stably folded repeats, α13-α14 and β8-β9, are adjacent to each other in the spectrin molecule (see Fig. 15-13, A ) and may form a hinge. Measurements of the force required to unfold a spectrin repeat show that the repeats function as elastic springs. When intact red cells deform, certain spectrin repeats partially unfold, which presumably contributes to the elasticity.

Side-to-Side Interactions Between α and β Chains Results in Zipperlike Dimerization.

Nucleation sites located at repeats 18 to 21 of the α chain and 1 to 4 of the β chain are implicated in triggering the zipping up of α and β chains (see Fig. 15-13, A ). The key interactions are between the A and B helices of β spectrin repeat 1 and the C and A helices of α spectrin repeat 21 and between the C and A helices of β2 and the A and B helices of α20. These helices form complementary charged surfaces that draw the two chains together. After the initial tight association of these nucleation sites, a conformational change is propagated, causing the remainder of the α and β chains to pair together. The two chains twist around each other with a periodicity of a complete turn approximately every four to eight repeats. The side-to-side interactions beyond the nucleation site are relatively weak, which may be important in allowing the α and β chains to slide past each other when the spectrin molecule flexes in the plane of the membrane. A common polymorphism of the α-spectrin allele, α ^LELY , encodes α chains that lack one of the nucleation sites and therefore do not form stable heterodimers. This clinically relevant allele is discussed in under “Hereditary Elliptocytosis” in Chapter 16 .

Spectrin Tetramerization Occurs in Head-to-Head Fashion by Association of the Partial Repeats at the Ends of α- and β Spectrin and Side-to-Side Association of the First Three α-Spectrin Repeats.

The “self-association” sequences required for tetramerization are located at the N-terminus of the α-spectrin chain and the C-terminus of the β-spectrin chain. This is called the “head” end of spectrin and is opposite the “tail” end where binding to actin occurs and where the nucleation site is found (see Fig. 15-13, A ). The specific structure of the spectrin self-association site has been extensively analyzed. The free C helix at the N-terminus of α spectrin (α0) associates with the free A and B helices at the C-terminus of β spectrin (β17) to form a stable triple helical bundle repeat (α0/β17). In this manner, spectrin heterodimers associate to form tetramers and higher-order oligomers in head-to-head fashion (see Fig. 15-13, A ).

In isolated spectrin heterodimers the partial repeats at the ends of the α and β chains are bound to each other. The longer α chain folds back on itself so that the noncovalent α0/β17 repeat lines up and interacts with repeat α4 and so that repeats α3 and α1 interact with each other. In the heterotetramers, there are three such side-to-side interactions between repeats in the two α-spectrins (α and α′): α1-α′3, α2-α′2, and α3-α′1 (see Fig. 15-13, B ). These additional lateral interactions, although weak, contribute to self-association. For dimers to associate into tetramers or vice versa, the bonds involved in self-association must first open. This is the rate-limiting step in dimer-to-tetramer interconversion.

The formation of spectrin tetramers and higher-order oligomers is critical to maintaining the mechanical strength of the membrane skeleton. Mutations in α or β spectrin that interfere with self-association weaken the membrane skeleton. Such defects are by far the most common cause of hereditary elliptocytosis and hereditary pyropoikilocytosis (see Chapter 16 ).

Spectrin Exists Predominantly as Tetramers in Vivo.

Physiologic ionic strength and lower temperatures (25°C) favor tetramer formation; low ionic strength and physiologic temperatures favor dissociation to dimers. The self-association of spectrin can be studied in purified spectrin and by manipulating red cell ghosts to alter the degree of spectrin oligomerization. In red cell ghosts, driving the equilibrium toward dimerization produces exceedingly fragile structures, underscoring the importance of spectrin tetramerization in membrane skeletal stability. At 4°C, the dimer-tetramer equilibrium is nearly kinetically frozen owing to its high activation energy. Thus spectrin eluted from ghosts at that temperature retains the degree of oligomerization that existed in vivo. In such eluates, approximately 5% of the spectrin is heterodimers, approximately 50% exists as heterotetramers, and the remainder are divided among higher-order spectrin oligomers and high-molecular-weight complexes of spectrin, actin, protein 4.1R, and dematin.

Spectrin Molecules are Condensed in Vivo.

Electron micrographs of purified spectrin tetramers or spectrin filaments in fully stretched skeletons show the molecule has an end-to-end length of 200 nm. But, as discussed later, simple calculations show that spectrin tetramers can only have a length of 62 to 72 nm in vivo. This correlates with the observed length of spectrin filaments in unstretched and partially stretched skeletons. It is not known exactly how spectrin is folded in this physiologic filament. The best morphologic evidence is that the α and β subunits form a double helix and are twisted about each other along a relatively straight common axis ( Fig. 15-15 ). Spectrin molecules extend and contract along this axis by varying the pitch and diameter of the double strand. This springlike behavior serves as a “molecular shock absorber” and is presumably critical to the elastic behavior of the membrane. Native spectrin tetramer has approximately 10 turns with a pitch of 7 nm and a diameter of 5.9 nm. This corresponds to 1 turn every four spectrin repeats. Structural analyses of the spectrin self-association site also show that the chains coil around each other, but with a periodicity of about 1 turn per eight repeats (see Fig. 15-13, B ). Other structural models suggest there is some bending between spectrin repeats or that helices within some repeats may “melt” and rearrange, shortening the repeat. Whatever the mechanism, it is clear that the ends of a spectrin molecule are normally only about one-third as far apart as they appear in stretched skeletons.

Model of spectrin structure in situ on the membrane.

A, Enhanced images of spectrin molecules captured from negatively stained electron micrographs of variably expanded membrane skeletons. In general, spectrin molecules were relatively straight but of variable lengths and on image enhancement showed a coiled appearance. B, Right-handed double helical model derived from the images in A . The pitch of the helix varies from 10.4 nm (1 nm = 10Å) to 16.6 nm in the images shown, and the diameter varies from 5.2 nm to 3.6 nm. Like a spring, as spectrin extends, the pitch of the helix lengthens and the diameter narrows. The average pitch and diameter in unstretched skeletons are calculated to be 7 nm and 5.9 nm, respectively. This corresponds to about five turns per spectrin dimer and four repeats per turn for a length of 35 nm per spectrin dimer. This is far less than the length of full extended spectrin dimers (100 nm) and almost exactly to the length of spectrin dimers from calculations based on the number of actin junctions (40,000 in a red cell surface of 140 µm ² ) and from direct measurements in unstretched skeletons.

(From McGough AM, Josephs R: On the structure of erythrocyte spectrin in partially expanded membrane skeletons. Proc Natl Acad Sci U S A 87:5208-5212, 1990.)

Spectrin Contains Important Functional Domains Involved in Protein-Protein Interactions.

In addition to the critical sequences involved in spectrin nucleation and self-association, the α and β chains interface with other proteins, including ankyrin, proteins 4.1R and 4.2, actin, adducin, and dematin, via defined structural domains (see Fig. 15-13, A ).

Binding Sites for F-Actin, Protein 4.1R, Adducin, and Dematin.

The N-terminal 301 amino acid of β spectrin binds protein 4.1R and actin and shares sequence homology with several actin-binding proteins, including calponin, α actinin, dystrophin, filamin, ABP 120, adducin, and fimbrin. There are two of these “calponin homology domains” (CH1 and CH2) in tandem in β spectrin. It appears that protein 4.1R and actin bind to each of these domains. Binding to CH2, the more C-terminal CH domain, is normally inhibited by an α helix at its N-terminus, which may serve a regulatory role (see Fig. 15-13, A ). PIP ₂ increases attachment of protein 4.1R to the CH domains.

Adducin binds to the N-terminal actin-binding domain and the first two repeats of β spectrin. Because mutations within the last few repeats of α spectrin markedly reduce membrane adducin content, the C-terminus of α spectrin probably also contributes to adducin binding.

Dematin, but not phosphorylated dematin, binds somewhere in the tail region of spectrin and facilitates binding of spectrin to F-actin. The exact binding site is unknown.

The tail ends of spectrin tetramers and oligomers bind to short, double-helical protofilaments of F-actin. The interaction is greatly enhanced by protein 4.1R. Approximately six spectrins can be accommodated on each actin protofilament, leading to the characteristic quasi-hexagonal organization of the membrane skeleton ( Fig. 15-16 ). The spectrin/actin/protein 4.1R junctional complexes contain other proteins, including adducin, dematin, tropomyosin, tropomodulin, and p55. These are discussed in the sections that follow.

Negatively stained electron micrographs of red cell membrane skeletons that have been stretched during preparation to reveal details of their structure.

A, Low magnification to emphasize the ordered, netlike structure. B, High-magnification view and schematic showing the predominantly hexagonal organization of the skeleton. C, Detail of the 37-nm long F-actin protofilaments. D, Location of various skeletal elements is shown in B . Sp4, spectrin tetramer; Sp6, spectrin hexamer.

(From Palek J, Sahr KE: Mutations of the red blood cell membrane proteins: from clinical evaluation to detection of the underlying genetic defect. Blood 80:308, 1992; and Byers T, Branton D: Visualization of the protein associations in the erythrocyte membrane skeleton. Proc Natl Acad Sci U S A 82:6153, 1985.)

EF Hands: Ca ²⁺ – and Protein 4.2–Binding Sites and Role in Spectrin-Actin Binding.

The C-terminus of α spectrin, like α actinin, contains two pairs of EF hands (see Fig. 15-13 ), which are structures that bind and mediate the regulatory effects of calcium. The N-terminal EF hands (EF _1,2 ) are structurally similar to calmodulin and bind Ca ²⁺ with low affinity. Ca ²⁺ binding to EF ₁ triggers a conformational change that allows EF ₂ to bind Ca ²⁺ . The C-terminal EF hands (EF _3,4 ) are Ca ²⁺ insensitive and similar to those in α actinin.

The EF hand domain binds protein 4.2 with high affinity (kD = 3 ×10 ⁻⁷ M). The binding capacity is augmented by micromolar concentrations of Ca ²⁺ , and the Ca ²⁺ effect is blocked by calmodulin, which also binds to the domain. Because protein 4.2 binds to band 3, this adds to the evidence that a portion of band 3 is located near the spectrin-actin junctions and serves as a secondary site for attaching the membrane skeleton to the lipid bilayer. It is not known whether protein 4.2 actually binds to the EF hands in vivo and, if so, what fraction of the protein is located there and what fraction is located in the ankyrin complex (see Fig. 15-6 ).

The relationships of the EF hand domain, the calponin homology domain, and F-actin are not known in spectrin but are likely to resemble those in α actinin ( Fig. 15-17 ). That is, the EF hands probably abut the actin/protein 4.1R–binding CH domains and may help regulate actin binding. Previously the EF hands were believed to be impotent in erythroid spectrin, although they were known to bind Ca ²⁺ and affect the function of αIIβII spectrin (fodrin). However, deletion of just the C-terminal 13 amino acids in the EF domain of the sph ^1J mouse leads to a severe, spherocytic hemolytic anemia and failure of the mutant spectrin to assemble into the membrane skeleton. Also, PfEMP3, a protein from P. falciparum that appears on the red cell membrane late in the parasite life cycle, binds to α spectrin in the EF hand domain region and disrupts the spectrin/actin/protein 4.1R complex, perhaps as a precursor to release of new parasites into the circulation. These observations show that the EF hand domain is functionally important, and direct evidence supports this conclusion. A minispectrin composed of the actin binding and EF hand domains and the four adjacent spectrin repeats in each chain that form the nucleation site binds to F-actin with the aid of protein 4.1R, as expected, but this binding is blocked if the EF hands are missing, or even if just the last 13 amino acids of α spectrin are deleted, as in the sph ^1J mouse.

Structural model, roughly to scale, of the tail end of spectrin deduced from the structure of α actinin.

Top, View from within the plane of the membrane skeleton. Bottom, View looking down on the membrane. CH1 and CH2 are the actin- and protein 4.1R–binding sites in the N-terminal, actin-binding domain of β spectrin. EF _1,2 and EF _3,4 are the N-terminal and C-terminal pairs of EF hands, respectively, in the EF domain, which lies at the C-terminal end of α spectrin. The intimate relationship between the EF and CH domains and recent evidence that the EF domain is required for optimal spectrin-actin binding suggest the EF hands may regulate actin binding.

Defects in Actin Assembly.

Not much is known about how the unique actin protofilaments of red cells are assembled, but some hints are emerging from defects in actin assembly. Red cells contain multiple members of the Rho family of guanosine triphosphatases (GTPases), including Rac1, Rac2, Cdc42, and RhoA, RhoG, and RhoH. Rac GTPases are activated by extracellular signals via guanine nucleotide exchange factors, which replace guanosine diphosphate (GDP) with GTP on Rac. GTP-activated Rac then interacts with P21-activated kinase, a diaphanous homologue (DIAPH3), and an assembly of proteins known as the Wiskott-Aldrich syndrome verprolin-homologous protein (WAVE) complex, stimulating ARP2/3 to initiate actin polymerization in many cell types. Erythrocytes probably also utilize this pathway because they contain Rac, RhoA, and all the members of the WAVE complex.

Mice that lack both Rac1 and Rac2 develop a microcytic hemolytic anemia with marked anisocytosis and poikilocytosis. Membrane skeletal junctions are aggregated, and there is pronounced irregularity of the spectrin meshwork. These changes are accompanied by decreased cellular deformability, an increase in the actin-to-spectrin ratio, and increased phosphorylation (Ser724) of adducin, an F-actin capping protein. Actin and phosphorylated adducin are more easily extracted by Triton X-100 from Rac1(−/−)/Rac2(−/−) red cells, indicating weaker association to the cytoskeleton.

Hematopoietic protein-1 (Hem-1) is a hematopoietic cell specific member of the WAVE complex. Mice lacking Hem-1 resemble the Rac1/Rac2 double knockout mice, with a moderate microcytic, hemolytic anemia, marked spherocytosis and poikilocytosis, clumped F-actin filaments, and decreased amounts of all membrane skeletal proteins. They also show increased expression of Ser724 on adducin.

Osteopontin phosphorylates or activates multiple proteins in human erythroblasts, including Rac-1 GTPase and adducin. Osteopontin knockout mice are anemic and also show defects in F-actin filaments.

The results indicate that Rac GTPases regulate the organization of the red cell membrane skeleton. It is unclear whether this occurs dynamically in the mature red cell or just during erythropoiesis.

Malaria.

P. falciparum generates a host-derived actin cytoskeleton within the red cell that appears to be an extension of the normal membrane skeleton. It functions in the transport of vesicles between the parasite and the red cell membrane. This cytoskeleton is abnormal in erythrocytes containing hemoglobins S or C, two hemoglobins that protect against malaria. One possible mechanism is that hemoglobin oxidation products, which are enriched in hemoglobin S and C red cells, inhibit actin polymerization in vitro.