Mechanisms of band 3 oxidation and clustering in the phagocytosis of Plasmodium falciparum-infected erythrocytes

Abstract

Erythrocytes (RBCs) opsonized by IgG and complement are prevalently recognized and phagocytosed by complement receptor CR1. This mechanism, effective in senescent and damaged RBCs seems to be operative in ring-parasitized RBCs, since infection by Plasmodium falciparum induces stage-dependent binding of auto-antibodies and activated C3 to the RBC membrane. Later, parasite forms are also recognized by non-opsonic receptors, such as scavenger receptor CD36. Malaria parasites induce the oxidative formation of hemichromes which are the trigger for the auto-antigen development. Band 3 protein is the most plausible candidate of the RBC auto-antigen, induced by hemichromes. Auto-antigens isolated from trophozoites were found only in a high-molecular-weight protein aggregates not present in the normal RBC. The immunocomplex was purified by protein-A affinity chromatography, purified proteins digested by trypsin and analyzed by MALDI-TOF. Peptide mapping showed that the main antigen consisted of band 3 protein aggregates that also contained hemichromes, IgGs, complement factor 3 (C3), and traces of spectrin and glycophorin but no parasite proteins. Two cysteines located in the band 3 cytoplasmic domain were found to be particularly reactive to oxidants and mediated band 3 covalent dimerization via disulfide bonds. Thus, parasites promote oxidative alterations in the membrane of the host which lead to exposure of antigenic sites recognized by anti-band 3 auto-antibodies. Formation of band 3 clusters appears to be mediated by cytoplasmic binding of hemichromes and also by direct band 3 oxidation, whereby clustered, oxidized and antigenic band 3 was underglycosylated.     

Mechanisms of band 3 oxidation and clustering in the phagocytosis of Plasmodium falciparum-infected erythrocytes

Erythrocytes (RBCs) opsonized by IgG and complement are prevalently recognized and phagocytosed by complement receptor CR1. This mechanism, effective in senescent and damaged RBCs seems to be operative in ring-parasitized RBCs, since infection by Plasmodium falciparum induces stage-dependent binding of auto-antibodies and activated C3 to the RBC membrane. Later, parasite forms are also recognized by non-opsonic receptors, such as scavenger receptor CD36. Malaria parasites induce the oxidative formation of hemichromes which are the trigger for the auto-antigen development. Band 3 protein is the most plausible candidate of the RBC auto-antigen, induced by hemichromes. Auto-antigens isolated from trophozoites were found only in a high-molecular-weight protein aggregates not present in the normal RBC. The immunocomplex was purified by protein-A affinity chromatography, purified proteins digested by trypsin and analyzed by MALDI-TOF. Peptide mapping showed that the main antigen consisted of band 3 protein aggregates that also contained hemichromes, IgGs, complement factor 3 (C3), and traces of spectrin and glycophorin but no parasite proteins. Two cysteines located in the band 3 cytoplasmic domain were found to be particularly reactive to oxidants and mediated band 3 covalent dimerization via disulfide bonds. Thus, parasites promote oxidative alterations in the membrane of the host which lead to exposure of antigenic sites recognized by anti-band 3 auto-antibodies. Formation of band 3 clusters appears to be mediated by cytoplasmic binding of hemichromes and also by direct band 3 oxidation, whereby clustered, oxidized and antigenic band 3 was underglycosylated.     

Co-localization of the immunoreactivities of corticotropin-releasing factor and arginine vasotocin in the brain and pituitary system of the teleost Catostomus commersoni

Summary Using the label-fracture technique, an ultrastructural comparison was made of the number and distribution of wheat germ agglutinin (WGA)-binding sites between human normal and sickle red blood cells. The WGA was adsorbed to colloidal gold, and quantitative analysis of the electron micrographs revealed that more binding sites were present on the sickle erythrocytes than on the normal erythrocytes. Moreover, the sites were more clustered on the sickle red cells than on the normal red cells. Use of another lectin, Bandieraea simplicifolia-II, revealed that it did not bind to normal or sickle red cells. Because of the affinity of the WGA for sialic acid residues, it is probable that the WGA is binding to a transmembrane sialoglycoprotein, glycophorin A. The conformation and/or distribution of the glycophorin A molecules may be altered by the sickle hemoglobin that binds to the red cell membrane. Hence, as detected by WGA, new surface receptors, which could play a role in the adhesion of sickle cells to endothelium may be exposed.     

Partial characterization of the copolymerization reaction of erythrocyte membrane band 3 with hemichromes

Early intermediates in the denaturation of hemoglobin, termed hemichromes, have been found previously to associate with the cytoplasmic domain of erythrocyte membrane band 3 in a manner which rapidly propagates into an insoluble, macroscopic copolymer. Because this interaction is thought to force a redistribution of band 3 in situ, the properties of the copolymerization reaction were investigated in greater detail. The band 3-hemichrome coaggregate was found to be stabilized largely by ionic interactions since elevation of either ionic strength or pH led to dissolution of the complex. The pH dependence, however, shifted to a more alkaline pH with increasing hemichrome concentration, suggesting a strong linkage between band 3 or hemichrome protonation and copolymer formation. The stoichiometry of the copolymer was measured at five globin chains per band 3 chain whenever underivatized dimer-tetramer hemichrome mixtures were employed. However, cross-linking of the hemichromes at either the alpha or the beta chains to form the stabilized tetramer yielded a copolymer stoichiometry of approximately eight globin chains per band 3 chain, i.e., two hemichrome sites per band 3 subunit. While underivatized hemichromes exhibited both a fast and slow phase of copolymerization, the cross-link-stabilized tetrameric hemichromes displayed predominantly the fast phase kinetics. Naturally occurring disulfide cross-linked hemichromes also reacted more avidly with band 3 than their reduced counterparts; however, the copolymerization process also proceeded to completion with totally reduced components. It is concluded that copolymerization of band 3 with hemichromes should occur under normal cellular conditions and at an accelerated velocity when the intracellular reducing power is low.     

Immunocytochemical determination of membrane-bound IgG using physiologically aged and experimentally aged erythrocytes

The expression of IgG receptor sites at the band 3 protein is important for the recognition and elimination of aged and experimentally altered erythrocytes. Membrane bound IgG was detected in different erythrocyte preparations and microvesicles by means of electron microscopic procedures (protein A-gold-, protein A-gold-silver- and anti-ferritin-sandwich-technique) and light microscopic procedures (immunofluorescence). Physiologically "old", pronase and neuraminidase as well as diamide treated erythrocytes and microvesicles demonstrated significant IgG loading. An increased IgG binding of erythrocytes treated with phenylhydrazine was only evident when higher phenylhydrazine concentrations were used. Both, the alteration of the glycocalyx (conformational changes of the external segment of the glycophorins) and the alteration of the membrane skeleton lead to an unmasking of the IgG receptor site at band 3 proteins (transmembrane effect). The result is an overcritical loading of cells with IgG molecules which initiate the elimination of the erythrocytes by macrophages of the Reticulo-Histiocytic-System.     

Autologous immunoglobulin-G binding to erythrocytes subjected to cyclical deoxygenation in vitro

This study demonstrates that low-density metabolically replete HbSS erythrocytes suspended in heat-inactivated autologous plasma and subjected to 15 hr of cyclical deoxygenation (under nitrogen) bind significantly increased quantities of autologous IgG as compared with oxygenated paired samples. IgG binding to the erythrocyte surface was quantified by a nonequilibrium 125-iodinated protein A binding assay and by flow cytometry. Sickle cells deoxygenated 15 hr (37 degrees C) in the presence of 2 mM calcium bound 2.2 +/- 0.2 (mean +/- SD)-fold more IgG (p less than 0.01) than oxygenated paired samples. Sickle erythrocytes deoxygenated in 0.4 mM EDTA bound 1.7 +/- 0.3 (mean +/- SD)-fold more autologous IgG than oxygenated controls (p less than 0.05). Indirect immunofluorescence assays also demonstrated that the relative levels of autologous IgG bound to sickle cells after 15 hr cyclical deoxygenation in the presence or absence of calcium was increased as compared with IgG binding by oxygenated paired samples. After 3 hr of cyclical deoxygenation in the presence of 2 mM calcium sickle erythrocytes exhibited a 40-60% increase in IgG binding, as compared with 10-20% increased IgG binding by paired samples treated in EDTA. These findings demonstrate that repeated morphologic sickling will increase the IgG binding capacity of low-density sickle cells, and suggest that sickling-associated alterations of the cell surface will produce new binding sites recognized by autologous IgG. These studies also show that the sickling-induced increase in IgG binding may be slightly enhanced by the presence of extracellular calcium.     

Reduced transglutaminase-catalyzed cross-linking of exogenous amines to membrane proteins in sickle erythrocytes

In order to determine the capacity of sickle cells to undergo transglutaminase-catalyzed cross-linking of membrane proteins, human normal and sickle erythrocytes were incubated with [ring-2-14C]histamine in the presence of Ca2+ and ionophore A23187. The [14C]histamine incorporation into membrane components was observed in freshly prepared erythrocytes. Incorporation of radioactivity into spectrin and Band 3 membrane components was significantly (P less than 0.001) less in sickle erythrocytes than in normal cells. Transglutaminase deficiency was excluded by the finding of increased activity of this enzyme in sickle cells from patients with reticulocytosis. The incorporation of [3H]spermine into red cell membranes was also less in sickle erythrocytes than in normal cells under the same conditions of incubation used for [ring-2-14C]histamine. Sickle erythrocytes were more permeable to these amines than normal cells. It is proposed that the gamma-glutamyl sites of membrane proteins in sickle erythrocytes are less accessible for transglutaminase-catalyzed cross-linking to histamine and polyamines in vitro, perhaps due to prior in vivo activation of this enzyme by the increased calcium in sickle cells and/or shielding secondary to altered membrane organization.     

Adherence of Mycoplasma gallisepticum to glass.

Attachment of washed Mycoplasma gallisepticum cells to glass was quantified with organisms in which membrane lipids were labelled with 3H. Siliconization of the test tubes decreased attachment, while centrifugation increased it. Attachment increased with temperature, decreased with increasing pH and ionic strength of the attachment mixture, but was unaffected by Ca2+, Mg2+ and EDTA. This suggests that ionic bonds, but not salt bridges, participate in the attachment process. Glycophorin, the major receptor responsible for M. gallisepticum attachment to erythrocytes, partially inhibited the attachment of the organisms to glass. However, bovine serum albumin also decreased attachment. Extensive pretreatment of the organisms with trypsin decreased their ability to attach to glass by about 35 to 40%. Trypsin and pronase failed to detach the organisms already bound to glass, suggesting that external mycoplasma cell components, other than membrane proteins, also participate in attachment of the organisms to glass.     

Rotational diffusion of the erythrocyte integral membrane protein band 3: effect of hemichrome binding.

Human erythrocyte band 3 was covalently labeled within the integral membrane domain by incubating intact erythrocytes with the phosphorescent probe eosinyl-5-maleimide. The rotational diffusion of band 3 in membranes prepared from these labeled cells was measured using the technique of time-resolved phosphorescence anisotropy. Three rotational correlation times ranging from 16 to 3800 microseconds were observed, suggesting that band 3 exists in different aggregate states within the plane of the membrane. The oxidizing agent phenylhydrazine was used to induce hemichrome formation within intact erythrocytes. The immobilization of band 3 in membranes prepared from these erythrocytes suggests that the binding of hemichromes induces clustering of band 3. The addition of purified hemichromes to erythrocyte ghosts leads to a similar effect. We have also examined the mobility of the cytoplasmic domain of band 3. This region was labeled indirectly using a phosphorescently labeled antibody which binds to an epitope within the cytoplasmic domain. We observed very rapid motion of the cytoplasmic region of band 3, which was only partially restricted upon hemichrome binding. This suggests that the integral and cytoplasmic domains of band 3 may be independently mobile.     

Binding of anti-band 3 autoantibody to oxidatively damaged erythrocytes. Formation of senescent antigen on erythrocyte surface by an oxidative mechanism

Incubation of human erythrocytes oxidized by iron catalysts, ADP/Fe3+ or xanthine/xanthine oxidase/Fe3+, with autologous IgG resulted in IgG binding as detected by enzyme immunoassay using protein A-beta-galactosidase conjugate. The binding of autologous IgG to ADP/Fe3(+)-treated erythrocytes maximized when the cells were treated with 1.8:0.1 mM ADP/Fe3+, and declined when treated above this concentration, suggesting that autologous IgG binds to moderately but not to excessively oxidized erythrocytes. The antibody involved in the binding was anti-Band 3, the autoantibody known to bind to aged erythrocytes, because isolated anti-Band 3 bound to the oxidized cells, but anti-Band 3-depleted autologous IgG did not. In addition, purified Band 3 inhibited the autologous IgG binding. Anti-alpha-galactosyl IgG, another natural antibody which has been reported to bind to aged erythrocytes, did not bind to the oxidized cells. Oxidation of membrane lipids, SH-groups of membrane proteins, and Hb of these cells was slight, but the cells contained an increased amount of membrane-bound native Hb, indicating that the oxidized cell membrane has an altered property. alpha-Tocopherol prevented the lipid oxidation and the subsequent IgG binding. Reduction of the oxidized erythrocytes with dithiothreitol resulted in a loss of the IgG binding. These results suggest that anti-Band 3 binding sites (Band 3 senescent antigen) are formed on moderately oxidized erythrocytes as a result of oxidation of membrane protein SH-groups which can be mediated by the membrane lipid oxidation and that formation of the anti-Band 3 binding sites on the oxidized cells is an essentially reversible membrane event which is linked to oxidation and restoration of the protein SH-groups.     

Clustering of integral membrane proteins of the human erythrocyte membrane stimulates autologous IgG binding, complement deposition, and phagocytosis.

Damaged or old erythrocytes are cleared rapidly from circulation. Because several common biochemical lesions can induce the clustering of integral membrane proteins, we have proposed that formation of microscopic protein aggregates in the membrane might constitute a cell surface marker that promotes removal of the defective/senescent cells. We demonstrate here that treatments that cluster integral membrane proteins in erythrocytes (1 mM ZnCl2, 1 mM acridine orange, and 0.35 microM melittin) induce autologous IgG binding, complement fixation, and phagocytosis by human monocytes in vitro. Removal of the clustering agents prior to incubation in autologous serum or cross-linking of cell surface proteins before addition of clustering agents prohibited the above response, while cross-linking after treatment with the clustering agents preserved the response even if the clustering agents were later removed. Furthermore, subsequent reversal of the chemical cross-link maintaining the clustered distribution also reversed the induction of IgG binding, complement deposition, and phagocytosis. Finally, by deleting or inactivating different steps in the phagocytosis pathway, the chronology of steps was shown to be: (i) integral protein clustering, (ii) IgG binding, (iii) complement deposition, and (iv) phagocytosis.     

Electron spin resonance study of the role of nitrosyl-heme in the activation of guanylate cyclase by nitrosoguanidine and related agonists.

One major component of lens plasma membrane is a glycoprotein that SDS-polyacrylamide gel electrophoresis shows to possess an apparent molecular weight of 26,000. When this protein is solubilized in low ionic strength buffers containing SDS, and heated to 100° for 1 to 3 min prior to electrophoresis, conversion into high molecular weight aggregate results. The heat lability of this protein is greatly enhanced if it solubilized and heated in buffers containing 0.1 M NaCl. At this ionic strength, incubation for 3 h at 38° results in conversion of 20% of the protein into high melecular weight aggregates. Most other membrane proteins isolated from lens membrane are insensitive to heat treatment. It is concluded that temperature and ionic strength must be recorded and controlled carefully when using SDS-polyacrylamide gel electrophoresis to study this membrane protein.     

SANS studies of interacting hemoglobin in intact erythrocytes.

Small angle neutron scattering (SANS) was used to investigate interaction forces between hemoglobin (Hb) molecules contained within human red cells. The scattering separately attributable to cell membranes and intracellular Hb was identified. A series of D2O-H2O contrast variation measurements were made in order to establish conditions for which scattering from the cell membrane is minimized (approximately 15% D2O). Measurements then were performed to examine changes in intermolecular Hb interactions occurring when the cells are contracted or swollen by varying the ionic strength of the suspension buffer. The scattering cross-sections were fitted to structure factors computed by a mean spherical approximation, and molecular parameters thereby extracted. Oxygenation studies on normal cells were performed, and results contrasted with those of similar studies of erythrocytes obtained from sickle cell disease patients.     

Altered Membrane Structure and Surface Potential in Homozygous Hemoglobin C Erythrocytes

Background

Hemoglobin C differs from normal hemoglobin A by a glutamate-to-lysine substitution at position 6 of beta globin and is oxidatively unstable. Compared to homozygous AA erythrocytes, homozygous CC erythrocytes contain higher levels of membrane-associated hemichromes and more extensively clustered band 3 proteins. These findings suggest that CC erythrocytes have a different membrane matrix than AA erythrocytes.

Methodology and Findings

We found that AA and CC erythrocytes differ in their membrane lipid composition, and that a subset of CC erythrocytes expresses increased levels of externalized phosphatidylserine. Detergent membrane analyses for raft marker proteins indicated that CC erythrocyte membranes are more resistant to detergent solubilization. These data suggest that membrane raft organization is modified in CC erythrocytes. In addition, the average zeta potential (a measure of surface electrochemical potential) of CC erythrocytes was ≈2 mV lower than that of AA erythrocytes, indicating that substantial rearrangements occur in the membrane matrix of CC erythrocytes. We were able to recapitulate this low zeta potential phenotype in AA erythrocytes by treating them with NaNO₂ to oxidize hemoglobin A molecules and increase levels of membrane-associated hemichromes.

Conclusion

Our data support the possibility that increased hemichrome deposition and altered lipid composition induce molecular rearrangements in CC erythrocyte membranes, resulting in a unique membrane structure.     

Identification of a spectrin-like protein in Sertoli cells

Sertoli cells prepared from rats ages 15 and 25 days were shown to contain a spectrin-like protein. Indirect immunofluorescence with monospecific antimouse erythrocyte immunoglobulin G (IgG) and with monospecific antimouse brain spectrin IgG revealed specific staining in Sertoli cells. Both antibodies precipitated two spectrin-like peptides of 240,000 and 235,000 daltons from cells solubilized with octyl glucoside. Proteins from Sertoli cell membranes were separated by electrophoresis on polyacrylamide gels containing sodium dodecyl sulfate and electrophoretically transferred to nitrocellulose membrane. Incubation of nitrocellulose membrane with either of the two antibodies, followed by horseradish peroxidase conjugated to second antibody, revealed only the larger, or alpha, spectrin subunit (Western blots). Both antibodies were used to provide immunoautoradiographic identification of the spectrin-like protein. In this procedure, spectrin and Sertoli cell membranes were shown to compete with [125I]-labeled spectrin from mouse erythrocytes for binding to antimouse erythrocyte spectrin IgG. Finally, two-dimensional proteolytic mapping of the 240,000- and 235,000-dalton peptides demonstrated limited spot homology with rat erythrocyte spectrin. However, subcellular fractions from Sertoli cells all contained a spectrin-like protein showing high homology from fraction to fraction. It is concluded that Sertoli cells contain a spectrin-like protein that is seen in cell fractions prepared by centrifugation, i.e., mitochondria, microsomes, nuclei, cytoplasm, and plasma membranes. Although homology with spectrin from erythrocytes or brain is not seen in peptide maps, the alpha subunit shares antigenic determinants with spectrin from erythrocytes. The beta subunit is believed to be precipitated by antispectrin as the result of binding to the alpha subunit, since the beta subunit shows no detectable antigenic homology with that of spectrin.     

Binding of autologous IgG to human red blood cells before and after ATP-depletion. Selective exposure of binding sites (autoantigens) on spectrin-free vesicles

Binding of autologous IgG to fresh, ATP-depleted red blood cells as well as to spectrin-free vesicles was studied by a non-equilibrium binding assay using 125I-iodinated protein A from Staphylococcus aureus. IgG binding was 14-times higher to spectrin-free vesicles than to ATP-maintaining red blood cells and 4-times higher than to ATP-depleted erythrocytes from which these vesicles were released. Protein A binding to vesicles that were released from washed and nutrient-deprived erythrocytes, was dependent on added autologous IgG. However, spectrin-free vesicles that were spontaneously released from erythrocytes conserved in whole blood, bound similar amounts of protein A with or without added autologous IgG (0.45-0.55 ng/micrograms band 3 protein). These findings demonstrate that opsonization of spectrin-free vesicles by autologous IgG occurs not only in the test tube, but also under blood blank conditions. The binding characteristics of IgG to spectrin-free vesicles are indicative of a natural autoantibody rather than an unspecific binding of autologous IgG. The preferential binding of IgG to spectrin-free vesicles implies a selective exposure of corresponding autoantigens in membrane regions that have lost cytoskeletal anchorage and bud off.     

Comparison of p25 presequence peptide and melittin. Red blood cell haemolysis and band 3 aggregation.

The 25 residue presequence (p25) for subunit IV of yeast cytochrome oxidase had previously been shown to possess structural and behavioural characteristics in common with the bee venom polypeptide, melittin. The present study extends the results of leakage experiments on model-membrane systems to the haemolysis of human erythrocytes, which both peptides are shown to accomplish in a manner sensitive to membrane potential. In addition, the laser flash-induced transient dichroism technique for measuring protein rotational diffusion has been used to show that both peptides aggregate band 3, the major integral membrane protein of the erythrocyte. Aggregation cannot be reversed by high ionic strength; this serves to differentiate these peptides from other positively charged species such as polylysine that aggregate band 3 at low ionic strength. These results suggest that aggregation of membrane proteins may possibly prove to be a feature of the interaction of p25 signal peptide with mitochondrial membranes.     

Enzymatic removal of oxidized protein aggregates from erythrocyte membranes

Erythrocytes oxidized or aged in the circulation undergo membrane protein aggregation and anti-band 3 autoantibody binding to the cell surface. When human erythrocytes were mildly oxidized in vitro with 0.1 mM Fe(III) at 37 degrees C for 3 h, the aggregation of nonionic detergent C(12)E(8)-insoluble membrane protein and the binding of anti-band 3 IgG to the cell surface were increased. Incubation of membranes isolated from the oxidized cells increased the amount of protein aggregates by 5-fold after 6 h, while incubation for a further 12 h sharply decreased the amount of aggregates. In the presence of diisopropyl fluorophosphate (DFP), however, the increased amount of aggregates was maintained in the subsequent incubation. Western blot analysis of the aggregates using rabbit anti-band 3 showed that band 3 protein aggregates increased in the initial stage of incubation and decreased upon subsequent incubation, whereas the increased band 3 protein aggregates did not subsequently decrease when membranes were incubated in the presence of DFP. Incubation of the oxidized cells at 37 degrees C for 18 h caused reduction of the membrane protein aggregates and the (125)I-anti-band 3 IgG binding to the cell surface, while incubation in the presence of DFP did not cause these reductions. The results suggest that the oxidation-induced cell membrane protein aggregates were probably removed by 80-kDa serine protease, namely, oxidized protein hydrolase (OPH), in the oxidized cell membranes [Fujino et al. (1998) Biochim. Biophys. Acta 1374, 47-54; (1998) J. Biochem. 124, 1077-1085; (2000) Biochim. Biophys. Acta 1478, 102-112], and as a result the increased anti-band 3 binding to the cell surface was reduced.     

Combined deletion of mouse dematin-headpiece and beta-adducin exerts a novel effect on the spectrin-actin junctions leading to erythrocyte fragility and hemolytic anemia

Dematin and adducin are actin-binding proteins of the erythrocyte "junctional complex." Individually, they exert modest effects on erythrocyte shape and membrane stability, and their homologues are expressed widely in non-erythroid cells. Here we report generation and characterization of double knock-out mice lacking beta-adducin and the headpiece domain of dematin. The combined mutations result in altered erythrocyte morphology, increased membrane instability, and severe hemolysis. Peripheral blood analysis shows evidence of severe hemolytic anemia with reduced number of erythrocytes/hematocrit/hemoglobin and an approximately 12-fold increase in the number of circulating reticulocytes. The presence of a variety of misshapen and fragmented erythrocytes correlates with increased osmotic fragility and reduced in vivo life span. Despite the apparently normal protein composition of the mutant erythrocyte membrane, the retention of the spectrin-actin complex in the membrane under low ionic strength conditions is significantly reduced by the double mutation. Atomic force microscopy reveals an increase in grain size and a decrease in filament number of the mutant membrane cytoskeleton, although the volume parameter is similar to wild type erythrocytes. Aggregated, disassembled, and irregular features are visualized in the mutant membrane, consistent with the presence of large protein aggregates. Importantly, purified dematin binds to the stripped inside-out vesicles in a saturable manner, and dematin-membrane binding is abolished upon pretreatment of membrane vesicles with trypsin. Together, these results reveal an essential role of dematin and adducin in the maintenance of erythrocyte shape and membrane stability, and they suggest that the dematin-membrane interaction could link the junctional complex to the plasma membrane in erythroid cells.