Anatomy of the red cell membrane skeleton: unanswered questions

Modern knowledge of the red cell plasma membrane and its membrane skeleton began with Marchesi and Steers’s identification of spectrin in 1968.¹  Prior to that year, almost the only thing known about membranes was that they contained a lipid bilayer.²  Indeed, there was a period in the 1960s when it was believed that red cell membranes contained only a single 22.5-kDa protein called “structural protein.”³  The fertile decade following the discovery of spectrin and the more or less coincident introduction of sodium dodecyl sulfate polyacrylamide gel electrophoresis as an analytic tool led to the isolation and characterization of several major red cell membrane proteins. The first simple model of the membrane skeleton appeared 36 years ago (Figure 1),⁴  and although it contained the core elements of the modern model, a great deal has been learned in the intervening decades (Figure 2 ⁵ ). Hundreds of individuals have contributed to our knowledge of the red cell membrane skeleton, but a few deserve special recognition, including Peter Agre, David Anstee, G. Vann Bennett, Daniel Branton, Lesley Bruce, Jean Delaunay, David Discher, Velia Fowler, Patrick Gallagher, Walter Gratzer, Philip Low, Vincent Marchesi, Narla Mohandas, Jon Morrow, Jeri Palek, Luanne Peters, David Speicher, Ted Steck, and Michael Tanner.

This review focuses on our current understanding of the anatomy of the membrane skeleton, particularly its spectrin-actin core, and on some of the key unanswered questions. Because of space limitations, many important aspects of the red cell membrane are not discussed. These include integral membrane proteins not attached to the skeleton, the lipid bilayer, membrane biogenesis, posttranslational modifications of the membrane, red cell aging, and mouse and human membrane diseases. For interested readers, a few other recent reviews are recommended.^6-10 

The red cell membrane contains ∼20 major proteins and at least 850 minor ones.¹¹  Some of the important ones are described in Table 1. The integral membrane proteins are organized into macromolecular complexes centered on band 3, the anion-exchange channel. Most of the peripheral membrane proteins form the membrane skeleton, a protein meshwork 40- to 90-nm thick that laminates the inner membrane surface. The skeleton is composed principally of spectrin, actin and its associated proteins (tropomyosin, tropomodulin, adducin, and dematin), protein 4.1R, and ankyrin.

Erythrocyte spectrin is a long, flexible, wormlike protein composed of 2 parallel chains (α- and β-spectrin) oriented in opposite directions. Each chain contains multiple spectrin-type repeats, with specialized functional domains at the “head” end for spectrin dimer-tetramer association and for erythrocyte ankyrin (also known as ankyrin-1 or ankyrin-R) binding, and domains at the “tail” end for binding to protein 4.1R, protein 4.2, short filaments of actin, and other proteins. The details of spectrin structure are shown in Figure 3.^12-26  On average, 6 spectrins bind per actin filament, leading to a pseudohexagonal arrangement.²⁷ 

Isolated α- and β-spectrin chains bind to each other electrostatically at a pair of nucleation sites formed by repeats near the spectrin tail (Figure 3A). After this initial connection, the rest of the α and β chains zip together, coiling around each other approximately every 4²⁸  repeats. The side-to-side interactions beyond the nucleation site are relatively weak,²⁹  allowing the 2 chains to slide past each other when the spectrin molecule flexes and extends during membrane deformation.

Spectrin α and β chains associate with each other at the head end. Basically, the complementary partial repeats (α0 and β17) at the end of each chain join to form a complete 3-helix repeat: α0/β17 (Figure 3A). In spectrin αβ dimers, the longer α chain folds back on itself, so the α0/β17 repeat interacts with repeat α4 and the α1 and α3 repeats contact each other (Figure 3A).¹³  In the α₂β₂-heterotetramers, there are 3 such side-to-side interactions between the longer α chains (Figure 3C). These lateral interactions, though weak, contribute to self-association. Higher oligomers (hexamers, octamers, and so forth) form by the same mechanism and can also be seen on the membrane.^27,30,31  Even though nearly 95% of the spectrin is in the tetramer or oligomer forms,³²  spectrin self-association is uniquely weak in the red cell.³³  Tetramers dissociate and reform under physiological conditions, and this is greatly accentuated when the membrane is distorted by shear forces.³⁴  This mechanism may be an evolutionary accommodation to permit the enormous distortions that the red cell undergoes in parts of the microvasculature.

Spectrin is a highly flexible molecule, but it is not certain how it achieves its flexibility because crystal structures of tandem repeats show a continuous helix across the junction between repeats (Figure 3B). There is evidence that the helical linker may bend in certain directions at the junction³⁵  and that helices within some repeats may “melt” and rearrange, shortening the repeat.³⁵  In addition, some of the repeats are unstable at physiological temperatures (starred repeats in Figure 3A)²³  and may partially unfold when red cells deform, which would contribute to flexibility.³⁶ 

Red cells contain short, double-helical filaments of nonmuscle or β-actin, termed “protofilaments” (Figure 2). These lie roughly parallel to the membrane plane (±20°) and are randomly oriented.³⁷  There are ∼30 000 to 40 000 protofilaments per red cell, with 6 to 8 actin monomers in each of the 2 strands.

Not much is known about how these unique actin filaments form, but some hints are emerging from defects in actin assembly. Mice that lack both Rac1 and Rac2 GTPases,³⁸  or lack Hem-1, a member of the Wiskott-Aldrich syndrome verprolin-homologous (WAVE) protein complex that regulates F-actin polymerization,³⁹  develop hemolytic anemias with misshaped red cells and clumped or irregular skeletons. This finding suggests that Rac GTPases and Hem-1 help organize or regulate the red cell membrane skeleton, acting through pathways that stimulate actin polymerization in many cell types. It is unclear whether such regulation occurs dynamically in the mature red cell or just during erythropoiesis.

Erythrocyte tropomyosin (TM) is a long rodlike dimer of α- and γ-TM isoforms (hTM5b and hTM5, respectively).¹⁰  One TM dimer binds to each of the 2 strands in the actin protofilament (Figure 2). Binding is magnesium ion dependent. The red cell TM isoforms also bind to tropomodulin (see below). Red cell membranes that lack TM are markedly more fragile than normal,⁴⁰  and only endogenous TM, not the longer muscle isoforms, can restore stability. Red cell TM is just long enough (∼34 nm) to cover the estimated 7 actin monomers in 1 strand of the F-actin protofilament, which supports the idea that TM both serves as a molecular ruler (along with actin-capping proteins) in assembling the filament⁴¹  and strengthens the filament after assembly.

In yeast, formins (a family of proteins that nucleate actin filaments) direct which TM isoform associates with an actin filament.⁴²  It will be interesting to see whether that is also true in erythroid precursors and whether formins play a role in protofilament formation.

Tropomodulin 1 (TMod1) has 2 functions: it caps the pointed or slow-growing end of actin filaments (Figure 2) and binds TM, which greatly strengthens actin capping.⁴³  There are 30 000 copies per red cell, or 1 per actin protofilament. One TMod1 binds to both actin strands and both TM dimers in the actin protofilament.⁴³  Mice that lack TMod1 in their red cells have variable actin protofilament lengths and irregular holes in the membrane skeleton.⁴⁴  These mice also have increased tropomodulin 3 in their red cells, which may partially compensate. Tropomodulin 3 is a protein that plays multiple roles in definitive erythropoiesis, erythroblast binding to so-called nurse macrophages, and enucleation,⁴⁵  but is not normally present in mature red cells.

Protein 4.1R promotes spectrin-actin binding and helps attach the membrane skeleton to the membrane. The first of these functions is especially important. Under physiological conditions, erythrocyte spectrin binds very weakly to F-actin. Protein 4.1R binds to both spectrin and actin and generates a high-affinity ternary complex.⁴⁶  The cofactor activity is contained in a 10-kDa spectrin-actin binding domain (SABD) in the middle of the molecule⁴⁷  (Figure 4^48,49 ). The actin-binding site in the SABD is straddled by a 2-part spectrin-binding site in a single linear peptide (Figure 4C).⁴⁹  Red cells that lack protein 4.1R are very fragile, but their membrane strength is normalized by the SABD peptide.⁵⁰  It is unfortunate that the 3-dimensional structure of the SABD within protein 4.1R is unknown. Solving the complete structure of this important protein is a major need in the field.

The other important 4.1R domain is the N-terminal, globular membrane binding domain. It binds glycophorin C, protein p55, band 3, and calmodulin in separate, well-defined areas⁴⁸  (Figure 4B). These interactions are discussed briefly in “The actin junctional complex.”

Adducin is a complex multifunctional protein containing an α subunit, and either a β subunit or, less often, a related γ subunit.⁵¹  The adducin subunits contain a globular head domain and an extended, flexible tail.⁵¹  The predominant form is probably the αβ heterodimer.⁵²  There are 30 000 copies of the dimer per red cell.

As suggested by the number of copies, adducin caps the fast-growing or “barbed” end of the actin protofilaments in red cells (Figure 2).⁵²  It also recruits spectrin to neighboring sites on actin,⁵³  which enhances the capping ∼10-fold.⁵⁰  These interrelated functions require a domain at the end of the adducin tail.⁵⁰  Both functions are regulated by calmodulin and by several protein kinases.⁵⁴  The effects of phosphorylation are complex, and it is not clear whether they are physiologically important.⁵⁵  Indeed, the importance of adducin itself is uncertain, because mice that lack the protein have only mild hemolysis and spherocytosis.⁵⁶ 

Adducin also helps attach the membrane skeleton to the lipid bilayer through interactions with the anion-exchange channel (band 3)⁵⁷  and glucose transporter type 1 (Figure 2).⁵⁸  The band 3 binding site is located in the tails of α- and β-adducin.⁵⁷  Whether adducin binds dimers or tetramers of band 3 and whether it binds 1 or 2 of each is unknown.

Dematin is an actin-binding protein with 2 actin binding sites: one in its headpiece and one in its tail.⁵⁹  The recombinant protein is a monomer,⁵⁹  but the native protein may be a trimer. Dematin was first identified by its ability to bundle actin filaments into cables,⁶⁰  but in mature red cells, where there are no such cables, dematin has 2 functions. First, unmodified dematin binds to spectrin and enhances spectrin-actin binding.⁶¹  This function is lost when dematin is phosphorylated by protein kinase A.^59,61  Second, dematin binds to the glucose transporter⁵⁸  and helps attach actin to the membrane.

Surprisingly little is known about this central question. Actin and protein 4.1 bind to 2 tandem CH domains at the N-terminal (or tail) end of β-spectrin (Figure 3A). The CH1 domain is constitutively active but the CH2 domain is incipient.¹⁸  It is unmasked by removal of a blocking helix or by phosphatidylinositol-4,5-bisphosphate, a lipid involved in many signaling pathways.¹⁸  It is not known how, or even if, phosphatidylinositol-4,5-bisphosphate regulates the CH2 domain in the native red cell.

The neighboring calmodulin-like EF hand domain of α-spectrin is also important. Deletion of the last 13 amino acids of the domain in a minispectrin or in the severely hemolytic sph^1J mouse nearly ablates actin binding.²¹ 

The relationships among the EF hand domain, the CH domains, and F-actin are not known in spectrin but probably resemble those in better studied spectrin family members such as α-actinin,^25,26,62  utrophin,⁶³  dystrophin, fimbrin, and plectin (Figure 3C). The problem is that these proteins bind actin in different ways. Sometimes the CH domains are in contact with each other (closed position), and sometimes they are extended and separated from each other (open position). In some analyses, only CH1 binds to actin²⁶ ; in others, both CH1 and CH2 bind.^25,62  The EF hand domain is always in contact with the CH domains and not in contact with actin, so it presumably regulates the CH domains. The fact that α-actinin CH domains adopt both open and closed positions²⁵  and utrophin also has 2 modes of binding⁶³  suggests that the CH domains in spectrin might also do so and that this might be regulatory (convert weak binding to strong binding, for instance). Perhaps protein 4.1R and/or the EF domain act by converting the CH domains to a high-affinity conformation. Spectrin could occupy up to 2 actins in the actin filament, given the stoichiometry (6 spectrins per ∼14 actin monomers; see Table 1), so a variety of arrangements is possible. In any event, these are clearly key questions that need to be answered going forward.

Another obvious problem that is rarely mentioned is that spectrin is in a planar array, but the spectrin binding sites on the helical actin protofilament must be in a helical array. Spectrin does not hang down into the cytoplasm or stick up into the lipid bilayer, so if all the spectrin binding sites on actin are saturated, either the actin-binding end of spectrin must be flexible enough to curve around the actin filament, or spectrin must be able to bind to actin in more than 1 way.

Electron micrographs of purified spectrin tetramers⁶⁴  or spectrin filaments in fully stretched skeletons²⁷  (Figure 5A,⁶⁵ ) show a spaghetti-like molecule with an end-to-end length of up to 200 nm. This configuration is how spectrin is usually depicted in membrane models. However, the filaments of native, unstretched skeletons are much more compact (Figure 5B⁶⁶ ), and simple calculations show that the average distance between actin filaments (ie, the length of a spectrin tetramer) is only 60 to 70 nm in vivo.^6,67  This length corresponds roughly to that of presumed spectrin filaments in native skeletons.^30,31,66  It is not known how spectrin folds in this resting state. The best evidence comes from visually enhanced electron micrographs of spectrin in partially expanded skeletons (Figure 5C-D), which suggest that α- and β-spectrin are coiled about each other in a 2-start helix with ∼10 turns per spectrin tetramer²⁸  (Figure 5E). Some micrographs show relatively straight spectrin filaments²⁸ ; others show some kinking.³¹  This difference may reflect whether the filaments are under tension. In this model, spectrin molecules extend and contract along the helical axis by varying the pitch and diameter of the double strand (Figure 5E).²⁸  This springlike behavior is presumably critical to the elastic behavior of the membrane.

However, there is evidence for other models that propose that spectrin molecules adopt a random wormlike configuration,⁶⁸  or that helical segments of the repeats melt and extend under tension,³⁵  or simply unfold when stressed.³⁶  The fact that previously occluded cysteines can be labeled when red cells are deformed by shear is evidence that some repeats unfold and that spectrin may even detach from ankyrin or actin when stretched.⁶⁹  Because the structure of spectrin affects its mechanical properties, as well as the spatial relationships of all the other proteins that bind to it, defining its structure in vivo, at rest, and when deformed is a high priority.

Erythrocyte ankyrin links spectrin to a complex of band 3 and other proteins in the lipid bilayer. The protein has 3 domains⁷⁰  (Figure 6,^70-73 ). The membrane domain is composed entirely of ankyrin repeats. It is a remarkably open, spiral structure, somewhat like a twisted sickle, that encircles and binds band 3 tetramers (Figure 6B,⁷⁴ ). Whether band 3 spontaneously forms tetramers and then attracts ankyrin, stabilizing the tetramers, or whether ankyrin, which has 2 binding sites for band 3 (Figure 6A), independently binds 2 band 3 dimers (ie, a “dimer of dimers”), which then form a tetramer, is unresolved. The fact that a stable subpopulation of band 3 tetramers can be isolated from nonionic detergent extracts of ghosts⁷⁵  argues for the former. But the structure of a band 3 tetramer binding to the full membrane domain has not been solved, and at least 1 recent paper argues that band 3 dimers must bind independently,⁷⁶  so the question remains.

The ankyrin-spectrin binding site is well defined. The ZU5A subdomain of ankyrin binds to a special site created by the junction of the 14th and 15th repeats within β-spectrin (Figure 6C). Only 1 ankyrin binds per tetramer, perhaps for steric reasons. Ankyrin binding promotes spectrin tetramer and oligomer formation by ∼10-fold, and vice versa.¹⁶  Ankyrin attachment to band 3 also strengthens spectrin self-association.⁷⁷  Presumably, both of these mechanisms help regulate the relatively tenuous self-association interaction.

Erythrocyte band 3, or officially SLC4A1, is the major red cell membrane protein, with ∼1.2 million copies per cell. Functionally, it is 2 proteins⁷⁸ : (1) an N-terminal cytoplasmic, peripheral membrane protein that is a key attachment site for the membrane skeleton, glycolytic enzymes, and deoxyhemoglobin; and (2) a C-terminal integral membrane protein that forms the red cell anion-exchange channel and aids carbon dioxide transport (Figure 7 ^78-86 ).

The glycolytic enzymes form a metabolic complex (metabolon) extending from phosphofructokinase through lactic dehydrogenase. Three of the enzymes bind to the N-terminus of band 3: phosphofructokinase, aldolase, and glyceraldehyde-3-phosphate dehydrogenase. The others bind indirectly as part of the complex.⁸⁰  The enzymes are inactive when bound⁸⁰  but are displaced and activated either by deoxyhemoglobin, which also binds to the N-terminus,⁸¹  or by phosphorylation of 2 tyrosines within the binding sites⁸⁰  (Figure 7). The enzymes⁸⁰  and their adenosine triphosphate product⁸⁷  are located in discrete clumps along the membrane, supporting the idea of a physical metabolon. Conditions that displace the enzymes from the membrane activate glycolytic fluxes in intact red cells by 45%, which supports the idea of a functional metabalon.⁸⁸ 

As noted earlier, ankyrin binds to band 3 tetramers. The acidic N-terminus and 2 loops on the surface of band 3 are involved in the binding⁶⁹  (Figure 7). As with the glycolytic metabolon, ankyrin is displaced from band 3 by deoxyhemoglobin.⁸²  The affinity of deoxyhemoglobin is weak, but the concentration of hemoglobin is so high in red cells that approximately half of the band 3 molecules are bound to deoxyhemoglobin when red cells are deoxygenated.⁸²  Loosened ankyrin constraints could improve blood flow in hypoxic areas, but with the risk that red cells subject to prolonged deoxygenation, such as those trapped in the spleen, might suffer membrane vesiculation. (Could this be the long-sought mechanism for splenic conditioning⁸⁹  in hereditary spherocytosis?).

Ankyrin and band 3 are part of the multiprotein ankyrin complex (Figure 2). From the amounts of each protein in the red cell (Table 1), the complex is estimated to contain 1 ankyrin, 1 band 3 tetramer, 2 glycophorin A or B dimers or heterodimers, 2 protein 4.2 molecules, and 1 Rh complex (a trimer of RhAG combined with RhD and RhCE). CD47, the Landsteiner-Wiener blood group antigen, and the proteins in the glycolytic metabolon are also components, but are present in less than stoichiometric amounts (Table 1) and so must contribute only to a subset of complexes. One wonders how many such subsets there are. Do they have constant composition or are they in a dynamic equilibrium? And are the rare complexes that contain CD47 and Landsteiner-Wiener blood group antigen localized to specific membrane subdomains?

As detailed in Table 1 and Figure 2, there are multiple interactions among proteins in the ankyrin complex. Some of these are critical; for example, red cells lacking band 3 also lack protein 4.2 and glycophorin A,⁹⁰  and red cells lacking protein 4.2 nearly lack CD47.⁹¹  A consequence of this promiscuity is that missense mutations that interfere with a single interaction within the ankyrin complex have little effect. Hereditary spherocytosis is caused by defects in spectrin, ankyrin, band 3, or protein 4.2 that impair ankyrin complex formation, and almost all the known mutations diminish the concentration of these proteins rather than block their binding functions.

The actin protofilament and its associated proteins are also attached to the membrane. For many years, this attachment was believed to occur exclusively via a ternary complex of protein 4.1R, p55, and glycophorin C or D. Each of these proteins interacts with the other 2, and the interaction sites are well mapped.^48,92,93  However, if this was an important skeleton attachment and the sole attachment at the actin-binding end of spectrin, it was never clear why patients with complete absence of glycophorin C and D had, at most, very mild hereditary elliptocytosis.

During the past few years, it has become clear that the actin junctional complex is more complex and, like the ankyrin complex, is centered on band 3 (Figure 2). First, protein 4.1R and ankyrin compete for binding to band 3,⁹⁴  which suggests that they bind to separate band 3 populations. The interaction with protein 4.1R is known to be important because zebrafish lacking band 3 can be rescued by expression of mouse band 3 but only if the mouse band 3 contains the protein 4.1R binding sites.⁹⁵  Second, protein 4.2, which requires band 3 for incorporation into the membrane, binds to the EF hand domain at the actin-binding end of spectrin. Third, adducin binds to band 3,⁵⁷  which definitely localizes a subpopulation of band 3 to the neighborhood of actin (Figure 2). The latter interaction is important because the membrane fragments when the connection is severed.⁵⁷ 

The stoichiometry of the actin junctional complex is uncertain. On average, 6 spectrins and 6 proteins 4.1R bind per actin (Figure 8,⁹⁶ ), and there is enough band 3, glycophorin A, glycophorin C/D, and probably enough glucose transporter 1 and stomatin to supply 6 complexes per actin filament (Table 1). However, there is only 1 copy of adducin per actin filament and only 3 copies of dematin, if it is a monomer as recent studies indicate,⁵⁹  or one copy if it is a trimer. Similarly, whereas protein 4.1R, glycophorin C/D, and protein p55 are usually pictured as a 3-part complex, there is only enough p55 for 1 copy per actin filament if it is a dimer, as some data suggest.^9 7  And there is only ∼1 copy of protein 4.2 available if 240 000 copies (1 per band 3 dimer) are tied up in the ankyrin complex. The actin junctional complex also contains enzymes in the glycolytic metabolon,⁷⁹  as well as the blood group proteins Kx/Kell and DARC/Duffy,⁹⁸  although the amounts of these proteins are insufficient for even 1 copy per complex.

Therefore, it is uncertain at the moment whether the actin junctional complex contains only a single band 3 dimer, interacting with just 1 or 2 of the 6 spectrin/protein 4.1R dyads, or whether it is a larger complex containing 3 to 6 band 3 dimers that interact with all the spectrins and proteins 4.1R. Both possibilities are depicted in Figure 8. Although the data are not conclusive, recent evidence favors a larger complex.⁹⁶  If so, it is not clear whether all or just a subset of band 3 molecules bind to adducin.

In either case, as shown in Figure 8, an interesting and underappreciated fact is that the ankyrin and actin junctional complexes are quite close to each other in situ on the membrane and must often collide, especially during red cell deformation but perhaps even when the cell is at rest. This knowledge raises the interesting possibility that proteins like band 3, protein 4.1R, protein 4.2, and adducin, which have binding partners in both complexes, may sometimes switch allegiances. Because some proteins compete for the same binding sites (eg, as noted, protein 4.1R displaces ankyrin from band 3),⁹⁴  this process could be a regulatory one.

Of the ∼1.2 million band 3 monomers per red cell, 40% are tetramers bound to ankyrin.⁹⁶  Depending on whether the actin junctional complex contains an average of 1 or 6 band 3 dimers, an additional 7% to 40% of the band 3 dimers are located there. Recent experiments comparing normal mouse erythrocytes to α-adducin-deficient erythrocytes suggest that roughly 33% of the band 3 molecules are immobilized by adducin, which favors a larger complex.⁹⁶  The remaining band 3 dimers (25%-30%) are believed to be diffusing freely in the lipid bilayer, constrained for the most part by the boundaries of the spectrin corrals, which average <100 nm.⁹⁶  These percentages are remarkably similar to those of freely diffusing band 3 dimers (25%-29%) measured in normal red cells using different methods.^96,99,100 

It is important to remember that some of the minor proteins found in band 3 complexes can exist only in a subset of the complexes (probably multiple different subsets) and that the lipid bilayer is heterogeneous, with areas like the lipid rafts where certain proteins concentrate. Therefore, the band 3 complexes must be more heterogeneous than these 3 states of band 3 imply. Analyses of band 3 mobilities support this conclusion.⁹⁵ 

A great deal has been learned about the structure of the red cell membrane and the membrane skeleton in the 36 years since the first model was published (Figure 1). I sometimes hear people say that the red cell membrane is no longer an active area of hematology research, but I disagree. We have the broad outlines of how the membrane is organized (Figure 2) but know few of the details, and what we do know is about a static structure. Many questions remain (Table 2). We do not know how the skeleton is constructed during erythropoiesis. We know very little about the dynamics of the skeleton: how labile individual bonds are, how the skeleton responds to capillary deformation, how the skeleton is disassembled and reassembled during invasion by malaria or other parasites, and so forth. And as discussed earlier, we do not know how spectrin is folded in the intact skeleton or how the proteins that form the actin junctional complex are organized. Or how spectrin binds to actin. We do not know whether band 3–binding proteins like protein 4.1R, protein 4.2, and adducin switch from one band 3 complex to another, given the apparent proximity of the complexes at each end of spectrin (Figure 8). We do not understand why the same repeats in different spectrins are so similar if they just serve as spacers. Do they have binding or mechanical functions that we do not yet know about? We do not know what roles posttranslational changes like phosphorylation, fatty acylation, methylation, glycation, hydroxylation, and ubiquitination play, or what roles calcium ion–binding and adenosine triphosphate–binding play. And we do not understand the interactions between the membrane skeleton and the overlying lipids, which will likely have regulatory as well as structural functions. This is only a partial list of questions; some others are itemized in Table 2. But I hope my point is clear. Because the red cell membrane skeleton is the paradigm for studies of spectrin-based membrane skeletons in all cells, and because we now know that spectrin and other membrane skeletal proteins invest internal organelles, some transport vesicles, and plasma membranes, and even play a role in nuclear organization, it is more important than ever to understand the structure and function of the red cell membrane skeleton.

The author regrets that important papers from many talented individuals who have made significant contributions to red cell research could not be discussed or cited due to space limitations. The author gratefully acknowledges the camaraderie and friendship of a large number of colleagues in the red cell membrane field who, over many years, have enriched his career and aided his research.

Contribution: S.E.L. wrote the review and designed the figures and tables.

Conflict-of-interest disclosure: The author declares no competing financial interests.

Correspondence: Samuel E. Lux, Boston Children’s Hospital, 300 Longwood Ave, Hunnewell 260.1, Boston, MA 02115; e-mail: lux@enders.tch.harvard.edu.