Band 3 protein of the red cell membrane of the llama: crosslinking and cleavage of the cytoplasmic domain

Comparative studies were done on the cytoplasmic domain of the band 3 protein in the red cell membranes of the the human and the llama. Two approaches were used: crosslinking with o-phenanthroline/CuSO4, and cleavage with 2-nitro-5-thiocyanobenzoate. o-Phenanthroline/CuSO4 crosslinks the band 3 polypeptide chains in the human; in contrast band 3 in the llama is minimally crosslinked by this agent. 2-Nitro-5-thiocyanobenzoate cleaves band 3 in the human into a 23,000-dalton fragment; a similar fragment is not generated from the llama band 3. These studies show that the cysteine residue located 23,000 daltons from the N-terminus of band 3 in the human involved in these reactions is unavailable for crosslinking and cleavage in the llama. Species differences in the cytoplasmic domain of band 3 may contribute to the unusual resistance of llama red cells to osmotic, chemical and physically-induced deformation.     

Oligomeric structure and the anion transport function of human erythrocyte band 3 protein

Conclusions Evidence from many laboratories using several different techniques strongly suggests that, in the intact red cell, band 3 exists as dimers which can associate with other dimers to form tetramers. The kinetics of anion transport inhibition by stilbenedisulfonates indicate that irreversible inhibition of one subunit does not detectably affect anion transport by the other subunit. This does not imply that monomeric band 3 could necessarily transport anions; the native conformation of each subunit may require stabilizing interactions with another subunit, as indicated by the recent work of Boodhoo and Reithmeier [10]. A more detailed understanding of the structure of the band 3 dimer/tetramer will require information on which specific segments of the primary structure are involved in subunit-subunit contact. The combination of chemical cross-linking with proteolysis [136] is a promising approach to this problem.     

Alterations in pyridoxal 5'-phosphate inhibition of human erythrocyte anion transport associated with osmotic hemolysis and resealing

When pyridoxal 5'-phosphate (PLP) is covalently bound to band 3 protein in intact red blood cells and those cells are subjected to the osmotic hemolysis and resealing process, a significant reduction in the original PLP anion transport inhibitory potency occurs. We show that partial deinhibition is not due to the development of a second anion transport pathway in resealed ghosts. Rather, partial deinhibition arises from a hemolysis-induced conformational change in CH17 (17-kDa integral chymotryptic domain of band 3). This change causes the extracellular exposure of new transport inhibitory sites. Exposure of the new sites leads to a 2-fold increase in PLP labeling of CH17 in resealed ghosts compared with CH17 in intact red cells. The hemolysis and resealing process has no effect on the labeling of CH35 (35-kDa integral chymotryptic fragment of band 3). Double-labeling studies show restoration of transport inhibitory potency to near red cell levels when the newly exposed CH17 sites are labeled with PLP in resealed ghosts. The results support the view that CH17 contains PLP transport inhibitory sites. They show that a major conformational change occurs in band 3 with hemolysis.     

Binding of hemoglobin to red cell membranes with eosin-5-maleimide-labeled band 3: analysis of centrifugation and fluorescence data

We have studied the binding of hemoglobin to the red cell membrane by centrifugation and fluorescence methods. The intact red cell was labeled with eosin-5-maleimide (EM), which specifically reacts with lysine 430 of band 3. Even though this residue is not part of the cytoplasmic domain of band 3 (cdb3) associated with hemoglobin binding, fluorescence quenching was observed when hemoglobin bound to inside-out vesicles (IOVs). The use of fluorescence quenching to measure band 3 binding was quantitatively compared with the binding determined by centrifugation, which measures binding to band 3 and non-band 3 sites. For the centrifugation it was necessary to include the non-band 3 association constants determined from chymotrypsin-treated IOVs. The binding of hemoglobin to band 3 was interpreted in terms of the binding of two hemoglobin tetramers to each band 3 dimer. An anticooperative interaction associated with the conformational change produced when hemoglobin binds results in a 2.8-fold decrease in the intrinsic constant of (1.54 +/- 0.25) x 10(7) M(-1) for the binding of the second hemoglobin molecule. From the changes in lifetime produced by binding the first and second hemoglobin molecules, it was possible to show that the conformational change associated with binding the second hemoglobin molecule results in a decrease of the heme-eosin distance from 47.90 to 44.78 A. Reaction of cyanate with the alpha-amino group of hemoglobin (HbOCN) is shown to produce a very dramatic decrease in the binding of hemoglobin to both the band 3 and non-band 3 sites. The intrinsic constant for binding the first hemoglobin molecule to band 3 decreases by a factor of 29 to (5.34 +/- 0.15) x 10(5) M(-1). The anticooperative interaction is greater with the intrinsic constant decreasing by a factor of 3.8 for the binding of the second hemoglobin tetramer to band 3. In addition, the nature of the conformational change produced by binding hemoglobin is very different with the second HbOCN increasing the heme-eosin distance to 55.99 A. The utilization of eosin-5-maleimide-reacted red cell membrane to study hemoglobin binding makes it possible to directly study the binding to band 3. At the same time a sensitive probe of the conformational changes, which occur when hemoglobin binds to band 3, is provided.     

Proteolysis of band 3 is enhanced by anti-Rho(D) binding to the red cell membrane

Surface radioiodinated human red cells were incubated with IgG fractions and the radioelectrophoretic profile of the ghost membranes determined. The patterns of RhO(D)-negative membranes exposed to anti-RhO(D) IgG and RhO(D)-positive membranes exposed to non-immune IgG fractions remained intact. Membranes of RhO(D)-positive membranes following incubation with anti-RhO(D) IgG showed a sharp reduction in the quantity of intact band 3, the main glycoprotein of the red cell membrane. This process was significantly abrogated in the presence of protease inhibitors. The results suggest a possible role for IgG binding in promoting the generation of band 3-derived fragments described by others as normal constituents of isolated ghosts.     

Affinity of hemoglobin for the cytoplasmic fragment of human erythrocyte membrane band 3. Equilibrium measurements at physiological pH using matrix-bound proteins: the effects of ionic strength, deoxygenation and of 2,3-diphosphoglycerate

The cytoplasmic fragment of band 3 protein isolated from the human erythrocyte membrane was linked to a CNBr-activated Sepharose matrix in an attempt to measure, in batch experiments, its equilibrium binding constant with oxy- and deoxyhemoglobin at physiological pH and ionic strength values and in the presence or the absence of 2,3-diphosphoglycerate. All the experiments were done at pH 7.2, and equilibrium constants were computed on the basis of one hemoglobin tetramer bound per monomer of fragment. In 10 mM-phosphate buffer, a dissociation constant KD = 2 X 10(-4)M was measured for oxyhemoglobin and was shown to increase to 8 X 10(-4)M in the presence of 50 mM-NaCl. Association could not be demonstrated at higher salt concentrations. Diphosphoglycerate-stripped deoxyhemoglobin was shown to associate more strongly with the cytoplasmic fragment of band 3. In 10 mM-bis-Tris (pH 7.2) and in the presence of 120 mM-NaCl, a dissociation constant KD = 4 X 10(-4)M was measured. Upon addition of increasing amounts of 2,3-diphosphoglycerate, the complex formed between deoxyhemoglobin and the cytoplasmic fragment of band 3 was dissociated. On the reasonable assumption that the hemoglobin binding site present on band 3 fragment was not modified upon linking the protein to the Sepharose matrix, the results indicated that diphosphoglycerate-stripped deoxyhemoglobin or partially liganded hemoglobin tetramers in the T state could bind band 3 inside the intact human red blood cell.     

Mode of interaction of phosphofructokinase with the erythrocyte membrane

Phosphofructokinase is known to associate with the human erythrocyte membrane both in vitro and in vivo. Such association activates the enzyme in vitro by relieving the allosteric inhibition imposed by ATP (Karadsheh, N.S., and Uyeda, K. (1977) J. Biol. Chem. 252, 7418-7420). We now demonstrate that ADP, ATP, and NADH, all of which are known to bind to the enzyme's adenine nucleotide activation site, are particularly potent in eluting the enzyme from the membrane. In addition, both inside-out red cell membrane vesicles and a 23-kDa fragment containing the amino terminus of the membrane protein, band 3, cause a slow, partial, and reversible inactivation of phosphofructokinase. The dependence of the residual phosphofructokinase activity on phosphofructokinase concentration demonstrates that inactivation occurs through the dissociation of active tetramers to inactive dimers. Dimers of phosphofructokinase associate with the membrane more avidly than tetramers. The kinetics of phosphofructokinase inactivation are consistent with the dissociation of tetramers in solution followed by the binding of dimers to the membrane. There is no indication of an association equilibrium between tetramers and dimers of phosphofructokinase bound to the membrane. Taken together, these results suggest that the amino-terminal segment of band 3 binds to the adenine nucleotide activation site, which is thought to be located in a cleft between the dimeric subunits of phosphofructokinase. As a result, band 3 not only rapidly activates the phosphofructokinase tetramer but also slowly inactivates the enzyme by preferentially binding its dissociated subunits.     

Anion-proton cotransport through the human red blood cell band 3 protein. Role of glutamate 681.

The band 3 protein of the human red blood cell membrane contains a glutamate residue that must be protonated in order for divalent (SO4=) anion transport to take place at an appreciable rate. The carboxyl side chain on this glutamate residue can be converted to the primary alcohol by treatment of intact cells with Woodward's reagent K (N-ethyl-5-phenylisoxazolium 3'-sulfonate) followed by reductive cleavage with BH4-. Edman degradation of CNBr fragments from band 3 labeled in intact cells with Woodward's reagent K and [3H]BH4- showed that Glu681 is heavily labeled under conditions in which Cl- exchange is inhibited, SO4= exchange is accelerated, and Cl- conductance is accelerated. No other glutamate residue in band 3 is detectably labeled under the conditions of these experiments, as demonstrated either by Edman degradation or by the lack of label in major known proteolytic fragments. It is concluded that Glu681 is the binding site for the H+ that is transported with SO4= during band 3-catalyzed H+/SO4= cotransport. This residue is conserved among all species of red cell band 3 (AE1) as well as the related proteins AE2 and AE3. Glu681 is the first amino acid residue in band 3 which has been identified as a binding site for a transported substrate (H+). The functional characteristics of this residue suggest that it lies within the transport pathway and can be alternately exposed to the intracellular and extracellular media.     

Peptide heterogeneity in the band 3 protein of rabbit erythrocyte plasma membranes

We have examined the band 3 protein(s) of rabbit erythrocyte membranes by a combination of differential extraction and surface labeling methods. Only one major peptide was labeled when intact red cells were exposed to ¹²⁵I^? and lactoperoxidase; this coincided with band 3. When intact cells were exposed to galactose oxidase followed by [³H]borohydride, numerous surface glycoproteins were labeled, one of which clearly coincided with band 3. Differential extraction with lithium diiodosalicylate revealed one major band 3 glycoprotein which contained both the ¹²⁵I^? and ³H surface labels and three peptides which were unlabeled; these three peptides are apparently not exposed at the cell surface.     

αI-spectrin represents evolutionary optimization of spectrin for red blood cell deformability

Spectrin tetramers of the membranes of enucleated mammalian erythrocytes play a critical role in red blood cell survival in circulation. One of the spectrins, αI, emerged in mammals with enucleated red cells after duplication of the ancestral α-spectrin gene common to all animals. The neofunctionalized αI-spectrin has moderate affinity for βI-spectrin, whereas αII-spectrin, expressed in nonerythroid cells, retains ancestral characteristics and has a 10-fold higher affinity for βI-spectrin. It has been hypothesized that this adaptation allows for rapid make and break of tetramers to accommodate membrane deformation. We have tested this hypothesis by generating mice with high-affinity spectrin tetramers formed by exchanging the site of tetramer formation in αI-spectrin (segments R0 and R1) for that of αII-spectrin. Erythrocytes with αIIβI presented normal hematologic parameters yet showed increased thermostability, and their membranes were significantly less deformable; under low shear forces, they displayed tumbling behavior rather than tank treading. The membrane skeleton is more stable with αIIβI and shows significantly less remodeling under deformation than red cell membranes of wild-type mice. These data demonstrate that spectrin tetramers undergo remodeling in intact erythrocytes and that this is required for the normal deformability of the erythrocyte membrane. We conclude that αI-spectrin represents evolutionary optimization of tetramer formation: neither higher-affinity tetramers (as shown here) nor lower affinity (as seen in hemolytic disease) can support the membrane properties required for effective tissue oxygenation in circulation.     

Mapping of senescent cell antigen on brain anion exchanger protein (AE) isoforms using HPLC and fast atom bombardment ionization mass spectrometry (FAB-MS)

Molecular recognition of senescent cells involves oxidation of a crucial membrane protein leading to generation of a neoantigen, called 'senescent cell antigen' (SCA), and binding of physiologic autoantibodies. These IgG autoantibodies trigger macrophage removal of the cell prior to its lysis at a time when anion transport has decreased but the membrane is still grossly intact. The neoantigen SCA is generated by oxidation of a major anion transport protein called band 3 or anion exchange protein. In this study, we use IgG physiologic autoantibodies from senescent red cells to isolate SCA from brain, and HPLC and fast atom bombardment ionization mass spectrometry (FAB-MS) to compare brain SCA to band 3. HPLC peptide maps of band 3 and SCA showed substantial homology, suggesting that SCA is a subset of band 3, and includes an estimated >/=45% of the band 3 molecule. FAB-MS results indicate that residues matching all three band 3 isoforms (AE1, AE2 and AE3) are detected in SCA fractions. These findings suggest that other isoforms of band 3 may undergo the same aging changes that AE1 on red blood cells undergoes to generate SCA. This provides confirmation that SCA is on non-erythroid cell types. Implications of these studies to the generation of neoantigens by oxidation and their recognition by autoantibodies to them are discussed.     

Membrane effects of imidoesters in hereditary stomatocytosis

The marked increase in cation (Na⁺, K⁺) permeability that results in swollen, cup-shaped red cells in the hereditary stomatocytosis syndrome can be corrected in vitro with a bifunctional crosslinking reagent, dimethyl adipimidate (DMA). ⁴⁵Ca influx in intact RBC, ⁴⁵Ca efflux in red ghosts, and ⁴⁵Ca retention in red ghosts are normal and not influenced by DMA. Endocytosis in resealed red ghosts is strikingly impaired but becomes normal if cells are first treated with 2 mM DMA. Protein kinase mediated phosphorylation of membrane proteins by AT³²P–only 20–40% of normal control values in both short-term (5 min) and more extended (60 min) incubations–is not improved by DMA. After reaction of ¹⁴C-DMA with stomatocytes, radiolabel is found associated with phosphatidyl serine and phosphatidyl ethanolamine and is also widely distributed among membrane proteins. Cation permeability of stomatocytes is corrected at DMA concentrations (1 mM) that result in barely detectable crosslinking of aminophospholipids or proteins, suggesting that either crosslinking of a minor component present in only small quantities or intramolecular (rather than intermolecular) crosslinking is responsible for the permeability effects. DMA, whose maximal crosslinking dimension is 7.3–9 Å, is the most effective bifunctional imidoester of those tested. Shorter (dimethyl malonimidate) or longer (dimethylsuberimidate) reagents are either less effective than DMA or totally without effect.     

The role of calpain in the selective increased phosphorylation of the anion-transport protein in red cell of hypertensive subjects

The phosphorylation of the anion-transport protein (band 3) is selectively increased in human red cell membrane, following exposure of intact cells to ionophore and micromolar calcium. The phosphorylation is catalyzed by a membrane associated protein kinase distinct from either protein kinase C or Ca2+/calmodulin dependent protein kinase. We show that the increase in phosphorylation of band 3 is abolished if red cells had been pre-loaded with an inhibitor of calpain or with an anticalpain monoclonal antibody. Our findings suggest that calpain activity may control, both at a functional and at a structural level, the activity of this important transmembrane protein through the modulation of its susceptibility as a substrate of membrane bound protein kinase(s). Based on previous observations indicating the presence in erythrocytes from hypertensive patients of an uncontrolled intracellular calpain-mediated proteolytic system accompanied by an increased phosphorylation of band 3 protein(s), we suggest that our results may shed light on the type of molecular alteration which is associated with the hypertensive state.     

Unique microtubules in luteal cells from superovulated rats

Luteal cells of immature female rats treated with gonadotropins contain microtubules with a number of interesting features. Many of the microtubules of these cells are arranged in bundles in which they are separated one from another by strands of material (i-MT bands) of unknown composition. The microtubules within the bundles assume a hexagonal packing pattern with i-MT bands between any two microtubules. The bundle microtubules and their i-MT bands are further connected via crosslinking filaments: pattern obtained from densitometer scans (measuring the arrangement of the crosslinking filaments) suggest that the filaments may represent microtubule-associated proteins. The complex arrangement of the microtubules within a bundle does not appear to extend for the entire length of the individual microtubules, and occasionally one sees profiles of single microtubules fanning out from the ends of the bundle: whether the same microtubules are regrouped at some other point in the cell is not known. Structures similar to the i-MT band and the crosslinking filaments have also been observed connecting microtubules to segments of the luteal cell plasma membrane: in these instances the i-MT-like band is found between the longitudinally sectioned microtubule and the membrane, with filaments connecting the two structures via the intermediate band. It is of interest that the microtubules of these luteal cells are not sensitive to treatment with antimicrotubule drugs and we suggest that the complex bundling arrangement provides their unusual stability.     

Detection of Cl- binding to band 3 by double-quantum-filtered 35Cl nuclear magnetic resonance.

We have applied double-quantum-filtered (DQF) NMR of 35Cl to study binding of Cl- to external sites on intact red blood cells, including the outward-facing anion transport sites of band 3, an integral membrane protein. A DQF 35Cl NMR signal was observed in cell suspensions containing 150 mM KCl, but the DQF signal can be totally eliminated by adding 500 microM 4,4'-dinitrostilbene-2,2'-disulfonate (DNDS), an inhibitor that interferes with Cl- binding to the band 3 transport site. Therefore, it seems that only the binding of Cl- to transport sites of band 3 can give rise to a 35Cl DQF signal from red blood cell suspensions. In accordance with this concept, analysis of the single quantum free induction decay (FID) revealed that signals from buffer and DNDS-treated cells were fitted with a single exponential function, whereas the FID signals of untreated control cells were biexponential. The DQF signal remained after the cells were treated with eosin-5-maleimide (EM), a noncompetitive inhibitor of chloride exchange. This result supports previous reports that EM does not block the external chloride binding site. The band 3-dependent DQF signal is shown to be caused at least in part by nonisotropic motions of Cl- in the transport site, resulting in incompletely averaged quadrupolar couplings.     

Hemichrome binding to band 3: nucleation of Heinz bodies on the erythrocyte membrane

Hemichromes, the precursors of red cell Heinz bodies, were prepared by treatment of native hemoglobin with phenylhydrazine, and their interaction with the cytoplasmic surface of the human erythrocyte membrane was studied. Binding of hemichromes to leaky red cell ghosts was found to be biphasic, exhibiting both high-affinity and low-affinity sites. The high-affinity sites were shown to be located on the cytoplasmic domain of band 3, since (i) glyceraldehyde-3-phosphate dehydrogenase, a known ligand of band 3, competes with the hemichromes for their binding sites, (ii) removal of the cytoplasmic domain of band 3 by proteolytic cleavage causes loss of the high-affinity sites, and (iii) the isolated cytoplasmic domain of band 3 interacts tightly with hemichromes, rapidly forming a pH-dependent, water-insoluble copolymer upon mixing in aqueous solution. Since the copolymer of hemichromes with the cytoplasmic domain of band 3 was readily isolatable, a partial characterization of its properties was conducted. The copolymer was shown to be of defined stoichiometry, containing approximately 2.5 hemichrome tetramers (or approximately 5 hemichrome dimers) per band 3 dimer, regardless of the ratio of hemichrome:band 3 in the initial reaction solution. The copolymer was found to be of macroscopic dimensions, generating particles which could be easily visualized without use of a microscope. The coprecipitation was also highly selective for hemichromes, since, in mixed solutions with native hemoglobin, only hemichrome was observed in the isolated pellet. Furthermore, no precipitate was ever observed upon mixing the cytoplasmic domain of band 3 with oxyhemoglobin, deoxyhemoglobin, (carbonmonoxy) hemoglobin, or methemoglobin.(ABSTRACT TRUNCATED AT 250 WORDS)     

Biosynthesis of erythrocyte membrane protein band 3 in DMSO-induced friend erythroleukemia cells

The major integral membrane protein of red blood cells, the mouse equivalent of human band 3, was purified and used to raise a specific antiserum. The murine protein resembles its human counterpart in several of its properties, including susceptibility to digestion by chymotrypsin added to intact cells and an ability to bind to concanavalin A. The synthesis of ³⁵S-labeled band 3 was detected in Friend erythroleukemia cells treated with DMSO by immuneprecipitation followed by SDS gel electrophoresis and fluorography. Induction with DMSO led to a greater than tenfold increase in the synthesis of band 3 and maximal synthesis was reached 3 to 4 days after the beginning of induction.     

Glycophorin A facilitates the expression of human band 3-mediated anion transport in Xenopus oocytes.

The effects of human red cell glycophorin A (GPA) on the expression of the human erythrocyte anion transporter (band 3, AE1) has been examined in Xenopus oocytes. The coexpression of GPA with band 3 increased stilbene disulfonate-sensitive chloride transport into the oocytes. The effect of GPA was particularly noticeable at low band 3 concentrations and less marked at high band 3 cRNA concentrations. The enhancement of chloride transport was specific to GPA and was not observed when either glycophorin B or glycophorin C was coexpressed with band 3. Immunoprecipitations of whole oocyte homogenates showed the amount of band 3 synthesized was not affected by GPA at subsaturating cRNA concentrations. More band 3 was detected at the oocyte surface by immunoprecipitation when GPA was also expressed. Chymotrypsin treatment of intact oocytes was also used to assess surface band 3 and greater cleavage of band 3 by chymotrypsin was observed when GPA was present. Band 3 synthesis and assembly into canine pancreatic microsomes in the reticulocyte cell-free translation system was not altered by cotranslation of GPA. We suggest that GPA facilitates the translocation of band 3 to the plasma membrane at some point during band 3 biosynthesis in Xenopus oocytes. However, GPA is not essential for the expression of band 3 in red cells, since GPA-deficient individuals have apparently normal levels of band 3. Other GPA-independent mechanisms must also allow translocation of band 3 to the surface membrane in erythroid cells and oocytes. GPA may affect the rate of accumulation of band 3 at the cell surface, rather than the final level in the plasma membrane.     

Dynamic association of band 3 with triton shells in human erythrocyte ghosts

To test a possibility that free band 3 and ankyrin-linked band 3 are exchanged in situ, band 3 was labeled with 125I, using intact red blood cells and lactoperoxidase. The cytoplasmic surface of this labeled band 3 was considered to be intact. When Triton shells were incubated with Triton supernatants prepared from 125I-labeled intact erythrocytes at 37 degrees C in the presence of Mg-ATP under isotonic conditions, the incorporation of free 125I-labeled band 3 to shells was observed. This incorporation was affected by the presence of Triton X-100 in the incubation mixture, and significantly decreased when the content of Triton X-100 was less than 0.04% (v/v). On the other hand, ankyrin-linked 125I-labeled band 3 was released when shells prepared from 125I-labeled intact erythrocytes were incubated with the Triton supernatants at 37 degrees C under the same condition as when free 125I-labeled band 3 incorporation was observed. These results strongly suggest that free and ankyrin-linked band 3 exchanged with each other in the presence of Triton X-100. A water-soluble 43 kDa fragment of band 3 inhibited the incorporation of free 125I-labeled band 3 to the shells and also inhibited the Mg-ATP-dependent shape change of ghosts in the absence of Triton X-100. Both of these inhibitory effects remained, even after 10 min of heat treatment at 100 degrees C, but drastically decreased by treatment with trypsin. Our results strongly suggest that a dynamic exchange of the free band 3 for ankyrin-linked band 3 may occur in intact erythrocytes, and it may even contribute to the shape change of erythrocytes.     

Ultrastructure of the human erythrocyte cytoskeleton and its attachment to the membrane

We attached paraformaldehyde-fixed human erythrocyte ghosts to coated coverslips and sheared them to expose the cytoskeleton. Quick-freeze, deep-etch, rotary-replication, or tannic acid/osmium fixation and plastic embedding revealed the cytoskeleton as a dense network of intersecting straight filaments. Previous negative stain studies on spread skeletons found 5-6 spectrin tetramers intersecting at each actin oligomer, with an estimated 250 such intersections/microns 2 of membrane. In contrast, we found 3-4 filaments at each intersection and approximately 400 intersections/microns 2 of membrane. Immunogold labeling verified that the filaments were spectrin, but their lengths (29-37 nm) were approximately one-third that of extended spectrin dimers. The length and diameter of the filaments were sufficient to accommodate spectrin dimers, but not spectrin tetramers. Our results suggest that, in situ, spectrin dimers may associate as hexamers and octamers, rather than tetramers. We present several explanations that can reconcile our observations on intact cytoskeletons with previous reports on spread material. Extracting sheared ghosts with solutions of low ionic strength removed the cytoskeleton to reveal projections from the cytoplasmic surface of the membrane. These projections contained band 3, as shown by immunogold labeling, and they aggregated to a similar extent as intramembrane particles (IMP) when the cytoskeleton was removed, suggesting a direct relationship between these structures. Quantification indicated a stoichiometry of 2 IMP for each cytoplasmic projection. Cytoplasmic projections presumably contain other proteins besides band 3 since further treatment with high ionic strength solutions extracts peripheral proteins and reduces the diameter of projections by approximately 3 nm.