Protein 4.1R-dependent multiprotein complex:New insights into the structural organizationof the red blood cell membraneMarcela Salomao*, Xihui Zhang*, Yang Yang*, Soohee Lee†, John H. Hartwig‡, Joel Anne Chasis§, Narla Mohandas*,and Xiuli An*¶

*Red Cell Physiology Laboratory and †Membrane Biochemistry Laboratory, New York Blood Center, New York, NY 10065; ‡Brigham and Women’s Hospital,Harvard Medical School, Boston, MA 02115; and §Life Sciences Division, Lawrence Berkeley National Laboratory, University of California, Berkeley, CA 94720

Communicated by Joseph F. Hoffman, Yale University School of Medicine, New Haven, CT, April 3, 2008 (received for review January 4, 2008)

Protein 4.1R (4.1R) is a multifunctional component of the red cellmembrane. It forms a ternary complex with actin and spectrin, whichdefines the nodal junctions of the membrane-skeletal network, andits attachment to the transmembrane protein glycophorin C creates abridge between the protein network and the membrane bilayer. Wenow show that deletion of 4.1R in mouse red cells leads to a largediminution of actin accompanied by extensive loss of cytoskeletallattice structure, with formation of bare areas of membrane. Whereasband 3, the preponderant transmembrane constituent, and proteinsknown to be associated with it are present in normal or increasedamounts, glycophorin C is missing and XK, Duffy, and Rh are muchreduced in the 4.1R-deficient cells. The inference that these areassociated with 4.1R was borne out by the results of in vitro pull-down assays. Furthermore, whereas Western blot analysis showednormal levels of band 3 and Kell, flow cytometric analysis using anantibody against the extracellular region of band 3 or Kell revealedreduction of these two proteins, suggesting a conformational changeof band 3 and Kell epitopes. Taken together, we suggest that 4.1Rorganizes a macromolecular complex of skeletal and transmembraneproteins at the junctional node and that perturbation of this macro-molecular complex not only is responsible for the well characterizedmembrane instability but may also remodel the red cell surface.

macromolecular complex � cytoskeleton

An essential attribute of the red cell is its ability to undergoextensive and repeated deformations while maintaining struc-

tural integrity. The cell owes this mechanical resilience to themembrane-associated protein skeleton (1, 2). This has the form ofa lattice, made up of spectrin tetramers, formed by self-associationof �� spectrin heterodimers (3). The tetramers are attached at theirends to predominantly sixfold junctions consisting of short F-actinfilaments (protofilaments) and several actin-binding proteins, in-cluding 4.1R, protein 4.9 (dematin), adducin, tropomyosin, andtropomodulin (4). Defects or deficiency of components of thejunctional complexes, and especially of 4.1R, lead to instability ofthe network and consequently of the cell. This reveals itself inprogressive fragmentation in vivo (5).

In addition to the above-mentioned cytoskeletal proteins, anumber of transmembrane proteins that specify blood group anti-gens have also been purified and characterized biochemically.These include band 3, glycophorin A, glycophorin B, glycophorin C,RhAG, Rh, Duffy, Lu, LW, CD44, CD47, Kell, and XK (6). Thesetransmembrane proteins exhibit diverse functions. For example,band 3 functions as an anion exchanger. Rh/RhAG are probably gastransporters although there is some controversy regarding whetherthey transport ammonia or carbon dioxide (7, 8). Duffy serves as achemokine receptor and is also a receptor for the malarial parasitePlasmodium vivax (9, 10). Lu, LW, and CD44 are proteins that areinvolved in adhesive interactions (11). CD47 can function as amarker of self on erythrocytes by binding to the inhibitory receptorSIRP� (12). Kell possess endothin-3-converting enzyme activity(13), but the function of XK remains to be defined.

The membrane-skeletal network is coupled to the lipid bilayerthrough transmembrane proteins. One such linkage is generated byankyrin, which forms a bridge between spectrin and band 3 tet-ramers (14–17). Band 3 also contains a binding site for carbonicanhydrase II at its C-terminal cytoplasmic domain (18) and bindingsites for glycolytic enzymes, hemoglobin, and protein 4.2 at itsN-terminal cytoplasmic domain (19). In addition, there is a clearinteraction between glycophorin A (GPA) and band 3 (20–23). Theassociation of these proteins with band 3 forms the band-3-basedcomplex. In addition to the band 3 complex, studies using humanRh-null erythrocytes suggested the existence of the Rh proteincomplex comprising RhAG, Rh, CD47, LW, and GPB (24, 25).More recently, the finding that components of both the band 3complex and the Rh complex are absent or reduced in band-3-deficient erythrocytes led to the concept of a band-3-based mac-romolecular complex (26).

A second membrane skeleton–bilayer link, consisting of a nexusamong 4.1R, p55, and the transmembrane glycophorin C (GPC), islocated at the network junctions (27–29). GPC and p55 are missingfrom 4.1R�/� mouse red cells (30) and are much reduced in human4.1R-deficient red cells (31, 32). These proteins, as well as sometransmembrane blood group proteins, Duffy, Lu, and CD44, andthe glucose transporter GLUT1 are found in normal or elevatedamounts in band-3-deficient red cells (26). The work described herewas undertaken to examine whether and to what extent 4.1R playsa part in the formation of membrane structures other than thenetwork junctions. The results, based on the study of 4.1R�/� mousered cells, have allowed us to identify a 4.1R-based macromolecularcomplex and to develop a more refined model of red cell membraneorganization.

ResultsSpecificity of Various Anti-Mouse Antibodies. To compare the ex-pression of red cell membrane proteins between wild-type and4.1R�/� cells, we first needed to generate a panel of variousantibodies against mouse transmembrane and cytoskeletal pro-teins. For transmembrane proteins we usually generate two anti-bodies, one against the extracellular region and one against thecytoplasmic part. The antigens used for antibody production arelisted in supporting information (SI) Table S1. All of the antibodiesare raised in rabbit with the exception of monoclonal anti-Kellantibody, which was generated in mice using red cell as antigen. Thespecificity of our antibodies was confirmed by Western blot analysis

Author contributions: S.L., J.A.C., N.M., and X.A. designed research; M.S., X.Z., Y.Y., S.L., andJ.H.H. performed research; M.S., S.L., J.H.H., J.A.C., N.M., and X.A. analyzed data; and M.S.,N.M., and X.A. wrote the paper.

The authors declare no conflict of interest.

¶To whom correspondence should be addressed at: Red Cell Physiology Laboratory, 310East 67th Street, New York, NY 10065. E-mail: xan@nybloodcenter.org.

This article contains supporting information online at www.pnas.org/cgi/content/full/0803225105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

8026–8031 � PNAS � June 10, 2008 � vol. 105 � no. 23 www.pnas.org�cgi�doi�10.1073�pnas.0803225105

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

using corresponding knockout mice as negative controls. All anti-bodies generated recognize the corresponding mouse proteins, andsome also recognize the cognate human proteins. Fig. S1 demon-strates the specificity of a representative set of antibodies againstmouse red cell proteins.

Analysis of Cytoskeletal Protein Components of 4.1R�/� Red Cells byWestern Blot. We have previously shown that the membranes ofmouse red cells lacking 4.1R have greatly impaired shear resistance(30). The same is true of human 4.1R-deficient cells (5). Althoughit is assumed that this is probably because of weakened interactionbetween spectrin and actin in the absence of 4.1R, the detailedmolecular basis is not clear. Thus, we first examined the expressionof skeletal proteins of red cells from 4.1R�/� and 4.1R�/� mice byWestern blotting. Protein loadings were standardized by applying tothe gels equal amounts of membranes as measured in terms ofcholesterol concentration. Fig. 1 shows that, in addition to theabsence of 4.1R and p55, the amount of actin in the membrane wasunexpectedly reduced by �50% in 4.1R-deficient red cells. Whenthe entire protein complement of whole red cells was examined, an�50% deficit of actin was again recorded (data not shown),showing that the missing actin was not in the cytosol. Whereas twojunction-associated proteins (tropomyosin and adducin) were in-creased, spectrin, tropomodulin, and dematin were unchanged. Theincrease of tropomyosin and adducin is probably due to reticulo-cytosis in 4.1R�/� mice because an increase is also noted in otheranemias with increased reticulocytosis such as sickle cell disease

and thalassemia (data not shown). Another major skeleton protein,ankyrin, was unchanged.

Structural Consequences of 4.1R Deletion. Having discovered thesignificant reduction of actin in 4.1R-deficient red cells, we thenexamined the structural consequences of 4.1R deletion. Staining offixed red cells with fluorescent phalloidin demonstrates that F-actinis indeed much sparser on the 4.1R�/� cell membranes (Fig. S2). Todetermine whether F-actin is reduced or redistributed, we quanti-tated the fluorescence levels (pixel intensity/unit area) and found an�30% reduction in phalloidin levels in 4.1R-null red cells whencompared with wild-type erythrocytes [knockout, 17,958 � 1,783;wild type, 25,479 � 1,858 (n � 8, P � 0.0000004)]. To establish howthe actin deficiency affects the structure of the network, weexamined the membranes by electron microscopy under negativestain. Fig. 2 demonstrates that the regularity of the lattice is grosslydisrupted, with large bare regions. It thus appears that many of thejunctions that are normally present are missing in the mutantmembranes.

Integral Membrane Proteins in 4.1R�/� Cells. It has been shown thatin both mouse and human band-3-deficient red cells the known orsurmised band-3-associated proteins, namely GPA, GPB, RhAG,Rh, CD47, and LW, are missing or greatly reduced. GPC, 4.1R, andp55, which are confined to the network junctions, are present innormal amounts, and so also are other transmembrane proteins(Duffy, Lu, GLUT1, LFA-3, and CD44) with no known cytoskel-eton interactions (26). We have used Western blots to compare theabundance of all of these proteins in 4.1R�/� and 4.1R�/� mousered cells. The outcome was that the amounts of band 3, GPA,RhAG, CD47, LW, and NHE1 (Na�–H� exchanger) were eitherunchanged or increased, and GPC was missing entirely, whereasXK, Duffy, and Rh were significantly reduced in the mutant cells(Fig. 3). Once again, the increase of GPA, LW, and NHE1 isprobably due to reticulocytosis in 4.1R�/� mice because theseproteins are also increased in other anemias such as sickle celldisease and thalassemia (data not shown). The implication is thatnot only GPC and p55, but also XK, Duffy, and Rh, are associateddirectly or indirectly with the 4.1R.

Surface Expression of Transmembrane Proteins in 4.1R�/� Cells. Wehave used flow cytometry with antibodies against extracellularepitopes of transmembrane proteins to refine our estimations oftheir relative concentrations in 4.1R�/� cells by Western blotting.As Fig. 4 shows, no GPC could be detected, and the concentrationof Duffy was reduced, whereas that of GPA was normal, in accord

Fig. 1. Immunoblots of membrane skeletal proteins in red cells of 4.1R�/�

and 4.1R�/� mice. Blots of SDS/PAGE of total membrane protein were probedwith antibodies against the indicated proteins. Note the absence of 4.1R, aswell as p55 in the 4.1R-deficient cells, the reduced actin concentration, and theelevated tropomyosin and adducin.

Fig. 2. Electron micrographs of membrane skeletons of red cells of 4.1R�/� and 4.1R�/� mice. (Left) Membrane skeleton of wild-type cells. (Right) Membraneskeleton of 4.1R-deficient cells. Note deficient membrane junctions in the mutant cells and large bare areas.

Salomao et al. PNAS � June 10, 2008 � vol. 105 � no. 23 � 8027

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

with the results from the immunoblots. By striking contrast, theproportions of band 3 and of Kell protein registered by flowcytometry were much lower than by Western blotting. The lowsurface expression of band 3 was also confirmed by labeling witheosin-5�-maleimide, a reagent commonly used for the diagnosis ofband 3 deficiency in human red cells (33). The results from age- andsex-matched mice are summarized in the Table 1. These resultssuggest that a fraction of band 3 and Kell may undergo a confor-mational change that masks the extracellular epitope in 4.1R-deficient red cells.

Binding of XK, Duffy, and Rh to 4.1R in Vitro. The finding that XK,Duffy, and Rh are reduced in 4.1R�/� red cells strongly impliestheir association with 4.1R. We used a pull-down assay to examinewhether they directly bind to 4.1R. As shown in Fig. 5, 4.1R bindsto the cytoplasmic domain of XK, Duffy, or Rh. As controls, we alsoexamined the binding of 4.1R to cytoplasmic domains of GPA,CD47, and Kell that are not reduced in 4.1R-deficient red cells andfound no binding under the same experimental condition (data notshown). Furthermore, the cytoplasmic domain of XK, Duffy, or Rhbinds specifically to the 30-kDa membrane-binding domain of 4.1R

but to neither of the other domains. The 30-kDa domain iscomposed of three lobes that bind to different transmembraneproteins (34). By examining subfragments of the 30-kDa domain wefound that the binding site for Rh is contained in lobe A and thatthe binding sites for XK and Duffy are both located in lobe B.

DiscussionIt has become clear that the simple model of red cell membraneorganization, which endured for so long based on an irreversiblyassembled membrane skeleton and a population of predominantlyfree-floating transmembrane proteins, is inadequate. It was discov-ered, for instance, that a number of transmembrane and cytosolicproteins are associated with the abundant anion channel proteinband 3, thereby forming a metabolon (26). We can now define twotypes of multiprotein complexes surrounding the connections be-tween the membrane skeleton and the bilayer. Fig. 6 summarizes inschematic form the probable interactions that go to make up whatwe shall term the band-3-based macromolecular complex and the4.1R-based macromolecular complex. The band-3-based macro-molecular complex is attached through ankyrin to spectrin tetram-ers at a point near their center. The band 3 in this complex is thoughtto be a tetramer (35). The remaining band 3 is dimeric (35), andsome or all of this fraction is surmised to bind to 4.1R and toanother junction component, adducin (36), to generate the 4.1R-based macromolecular complex, as depicted in Fig. 6 Right. Ac-cording to our evidence from present study, the 4.1R macromo-lecular complex comprises, besides the ternary spectrin–actin–4.1Rnexus, p55 and the transmembrane proteins GPC, Rh, Duffy, Kell,and XK. The last four are present in much smaller numbers thanGPC and 4.1R and are presumably associated with certain 4.1R–GPC complexes.

It appears unlikely that any of the known integral membraneproteins normally exist as free-floating monomeric chains. Theevolutionary advantage of sequestering these blood group proteinsas captives within large complexes may be to prevent them from

Fig. 3. Immunoblots of transmembrane proteins in red cells of 4.1R�/� and 4.1R�/� mice. Blots of SDS/PAGE of total membrane protein were probed withantibodies against the indicated proteins. Note the disappearance of GPC and diminution of Rh, XK, and Duffy proteins in the 4.1R-deficient mice and the increaseof GPA, LW, and NHE1 expression.

Fig. 4. Flow cytometric analysis of red cell membrane proteins of 4.1R�/� and4.1R�/� mice. The ordinate measures the number of cells displaying thefluorescent intensity given by the abscissa. Black lines, 4.1R�/�; gray lines,4.1R�/�; dotted lines, negative control.

Table 1. Expression of blood group antigen-carrying proteinsas assessed by flow cytometry

Protein

Mean fluorescenceKnockout %of wild type P valueWild type (n) Knockout (n)

TER119 2,791 � 161 (6) 2,703 � 355 (6) 97 0.601GPC 2,800 � 137 (10) 250 � 147 (10) 8 0.00023Duffy 1,746 � 207 (11) 1,183 � 295 (12) 68 0.00003Kell 2,405 � 243 (14) 2,020 � 265 (14) 84 0.00049Band 3 1,108 � 260 (8) 534 � 200 (9) 48 0.00026EOM 8,721 � 224 (5) 5,970 � 1,178 (5) 68 0.002

8028 � www.pnas.org�cgi�doi�10.1073�pnas.0803225105 Salomao et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

clustering upon encountering an adhesive surface and therebyforming tight, possibly irreversible, attachments.

The breakdown of membrane-skeletal organization in 4.1R-deficient mouse red cells is striking, although not necessarilyunexpected. The stability of the ternary complex of spectrin,F-actin, and 4.1R is high (Kd � 10�15 M) (2) whereas the binarycomplex of spectrin and F-actin is weak (Kd � 10�5 M) (37).Therefore, one may expect that the actin protofilaments mightdissociate or disproportionate into longer filaments and cause thejunctions to disintegrate. This must be assumed to apply to human4.1R-deficient red cells, which are associated with elliptocytosis andanemia (38–40). Indeed the skeleton network was also found to bemarkedly disrupted in human 4.1R-deficient red cells (41). It isinteresting to note that the 4.1R-deficient mice exhibited a muchmore severe hemolytic anemia than 4.1R-deficient human patients.One possible reason is that the nature of 4.1 deficiency betweenmice and humans is different. Whereas in humans the high-molecular-mass (135 kDa) isoform of 4.1R appears early in eryth-ropoiesis (42), in both the normal and the 4.1R-deficient case, to bereplaced by the mature (80 kDa) isoform in normals, the 4.1Rknockout mice produce neither isotype (30). The 135-kDa formmay suffice to ensure normal assembly of the protofilaments, whichare then at least partially stabilized by tropomyosin (43, 44),spectrin, and actin-capping proteins. The mechanism(s) by whichlack of 4.1R results in reduced actin is not clear, but one possibilityis that a part of the actin in the protofilaments dissociates and isproteolytically degraded in the cytosol because of failure of appro-priate actin assembly in the immature cells.

The finding of normal levels of band 3 and Kell by Western blotanalysis but reduced levels by flow cytometric analysis suggestspossible conformational change of band 3 and Kell epitopes inintact membranes. Because band 3 tetramers are present at theband 3–ankrin-based macrocomplex, we speculate that the confor-mation of band 3 dimers that are present at the 4.1R-basedmacrocomplex is probably affected by the absence of 4.1R. Theconformation of Kell is probably due to reduction of XK in theabsence of 4.1R. Another interesting finding is that RhAG levelsremain constant while Rh levels decline in 4.1R�/� cells. Thisfinding may seem contradictory with the well accepted notion thatRh and RhAG exist as tetramers in mature red cells. However, it

should be noted that we recently found that Rh and RhAG trafficdifferently in the erythroblasts (our unpublished data) and as suchthat these two proteins may exist independent of each other duringerythroid development.

In summary, our results throw light on the basis of red cellmembrane skeleton stability and enlarge our understanding of theorganization of the membrane as a whole.

Materials and MethodsMaterials. A SulfoLink kit was purchased from Pierce. A Cholesterol Quantifica-tion Kit was from BioVision. pMAL vector, MBP resin, monoclonal anti-MBPantibody, T4 ligase, and restriction enzymes were obtained from New EnglandBioLabs. Top Pfu polymerase and BL21 (DE3) bacteria were from Stratagene,reduced-form glutathione and isopropyl �-D-thiogalactopyranoside were fromSigma, proteinase inhibitor mixture set II was from Calbiochem, a DC ProteinAssay Kit, SDS/PAGE, and electrophoresis reagents were from Bio-Rad, and aSuperSignal West Pico chemiluminescence detection kit reagent was from Pierce.Alexa Fluor 633-conjugated phalloidin was purchased from Invitrogen–Molecular Probes. HRP-conjugated anti-mouse IgG and HRP-conjugated anti-rabbit IgG were from Jackson ImmunoResearch Laboratories. Biotin-labeledsynthetic peptides corresponding to the cytoplasmic tail of XK, Duffy, or Rh werefrom Genemed.

Mice. The generation of 4.1R knockout mice has been described previously (30).The mice were backcrossed onto a C57BL/6 background and have been inbred for�20 generations. All of the mice were maintained at the animal facility of theNew York Blood Center under specific pathogen-free conditions according toinstitutional guidelines. Animal protocols were reviewed and approved by theInstitutional Animal Care and Use Committee. Band 3 knockout mice and Duffyknockout mice were kindly provided by Luanne Peters (The Jackson Laboratory)and Asok Chaudhuri (the New York Blood Center), respectively. Blood sampleswere obtained by tail or cardiac puncture of anesthetized mice.

Generation of Antibodies. Antibodies against mouse transmembrane proteinsGPC, band 3, Rh, RhAG, XK, Kell, Duffy, LW, and CD47 were raised in rabbit byusing synthetic peptides as antigens. With the exception of LW and Duffy, twoantibodies (one against the extracellular region and one against the cytoplasmicregion)againsteachproteinweremade.Theantibodieswereaffinity-purifiedbyusing specific peptides immobilized to SulfoLink Coupling Gel, and the specificityof the antibodies was examined by Western blotting using the correspondingknockout mouse red cells as negative controls.

Phalloidin Staining of Red Cells. Freshly drawn blood was washed three times inPBG buffer (PBS/0.1% BSA/10 mM glucose). A total of 4 � 107 cells were fixed

Fig. 5. Direct interaction of 4.1R with Rh, XK, and Duffy. The binding of 4.1R 80-kDa and its functional domains to cytoplasmic tails of XK, Duffy, and Rh wasassessed by streptavidin pull-down assay. The binding of 4.1R, GST-tagged domains of 4.1R, and MBP-tagged subdomains of 30-kDa to the cytoplasmic domainsof XK, Duffy, and Rh was detected by using anti-4.1R antibody, anti-GST antibody, and anti-MBP antibody, respectively. Note the binding of all three proteinsto 30-kDa, the binding of Rh to lobe A, and the binding of XK and Duffy to lobe B.

Salomao et al. PNAS � June 10, 2008 � vol. 105 � no. 23 � 8029

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

in 100 �l of 0.5% acrolein–PBG for 5 min at room temperature and resuspendedin PBG buffer to a final concentration of 107 cells per milliliter. The cells wereplated on poly-L-lysine-treated coverslips, permeabilized with 0.05% TritonX-100–PBS for 10 seconds, and washed three times with 0.1 M glycine–PBS.Staining with 1 unit of Alexa Fluor 633–phalloidin was carried out at roomtemperature for 40 min. Coverslips were washed four times with PBS, and slideswere mounted. Images were obtained with a Zeiss LSM META 510 ConfocalMicroscope.

Electron Microscopy. Mouse erythrocytes were attached to polyL-lysine-coatedcoverslips by centrifugation at 200 � g for 5 min at room temperature. Erythro-cytes were permeabilized by using 0.5% Triton X-100 in PHEM buffer (60 mMPHEM/25 mM Hepes/2 mM MgCl2/10 mM EGTA/1 �M phallacidin) containing0.05% glutaraldehyde for 2 min. Membrane skeletons were rapidly washed withPHEM buffer without fixative and then fixed with 1% glutaraldehyde in PHEMbuffer for 10 min at 37°C. They were washed into distilled water and rapidlyfrozen, freeze-dried at �90°C, and metal-cast with 1.4 nm of platinum at 45°Cwith rotation and 5 nm of carbon at 90°C without rotation (Cressington CFE-60Freeze Fracture Machine). Metal casts were separated from the coverslips byusing 25% hydrofluoric acid, washed with distilled water, picked up with 200-mesh formvar-coated copper grids, and photographed at 100 kV in a JEOL1200-EX electron microscope at 80-kV accelerating voltage.

Preparation of Red Cell Membranes. RBCs were washed three times in PBS. Whiteghostswerepreparedby lysisofRBCs in ice-cold5T5Kbuffer (5mMKCl/5mMTris,

pH 7.4/0.1 mM DFP) in the presence of 1 mM MgCl2 followed by three washes in35 volumes of the same buffer. Membrane cholesterol content was measured byusing a Cholesterol Quantification Kit according to the manufacturer’s protocol.Protein concentrations were measured with the Bradford method using the DCProtein Assay Kit.

SDS/PAGE and Western Blot Analysis. Aliquots of RBC ghosts, matched forcholesterol content, were separated by 10% SDS/PAGE. The proteins weretransferred to a nitrocellulose membrane. After blocking for 1 h in blockingbuffer (10 mM Tris, pH 7.4/150 mM NaCl/0.5% Tween 20/5% nonfat driedmilk powder), the blot was probed for 1 h with the desired primaryantibodies. After several washes, the blot was incubated with anti-rabbitor anti-mouse IgG coupled to HRP and developed with the SuperSignalWest Pico chemiluminescence detection kit. All steps were performed atroom temperature.

Flow Cytometry Analysis. RBCs from wild-type and 4.1R knockout mice werewashed three times in PBS supplemented with 0.1% BSA (PBS-BSA). A totalof 1 � 106 cells were incubated for 20 min on ice with the appropriateamount of antibodies against the extracellular region of transmembraneproteins GPC, band 3, GPA, Kell, and Duffy. After washing two times inPBS-BSA, the cells were incubated with Alexa Fluor 488-conjugated sec-ondary antibody (Molecular Probes) for 20 min on ice. Cells were washed

Fig. 6. Schematic representation of two types of multiprotein complexes in the red cell membrane. (Left) Protein complex attached to spectrin near the center ofthe tetramer (dimer–dimer interaction site). Tetrameric band 3 is bound to ankyrin, which is bound to spectrin. The membrane skeletal protein 4.2 has binding sitesfor band 3 and for ankyrin. Transmembrane glycoproteins GPA, Rh, and RhAG bind to band 3, and CD47 and LW associate with Rh/RhAG. The two cytoplasmic domainsofband3containbindingsites for solubleproteins, theshortC-terminaldomainforCAII, the largeN-terminaldomainfordeoxyhemoglobinandforglycolyticenzymes,aldolase, phosphofructokinase (PFK), and glyceraldehyde 3�-phosphate dehydrogenase (GAPDH). (Right) Protein complex at membrane skeletal junctions. Thejunctions contain the ternary complex of spectrin, F-actin, and 4.1R, as well as the actin-binding proteins tropomyosin, tropomodulin, adducin, and dematin. 4.1R entersinto an additional ternary interaction with the transmembrane protein GPC and p55 and is taken also to bind to band 3, in the form of a dimer, which also carries GPA.Rh, Kell, and XK also have binding sites on 4.1R. Note, however, that the copy numbers of all transmembrane proteins except GPA and GPC are low and therefore willnot be present on all complexes.

8030 � www.pnas.org�cgi�doi�10.1073�pnas.0803225105 Salomao et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

three times in cold PBS-BSA before analysis on a Becton Dickinson FACS-Canto. Preimmune IgGs were used as negative control. The mean fluores-cence intensity was used as a measure of antibody binding. Eosin-5�-maleimide staining was performed as described previously (45) with fewmodifications. Briefly, 5 �l of whole blood was incubated with 25 �l of PBScontaining 0.5 mg/ml EMA for 1 h at room temperature. Cells were washedfour times with PBS-BSA before flow cytometry analysis.

Preparation of Recombinant Proteins. The full-length 4.1R 80-kDa, GST-tagged 30-kDa, 16-kDa, 10-kDa, and 22/24-kDa domains of 4.1R wereconstructed and purified as described previously (46, 47). MBP-tagged lobeA, lobe B, and lobe C of the 30-kDa domain were subcloned into pMAL-p2xvector using EcoRI and SalI upstream and downstream, respectively. cDNAencoding the desired sequences was transformed into the BL21 bacterialstrain. Expression was induced by 0.1 mM isopropyl �-D-thiogalactopyr-anoside at 4°C overnight. MBP-tagged recombinant proteins were purifiedon an amylose resin affinity column. Proteins were dialyzed against bindingbuffer (10 mM Tris�HCl, pH 7.4/150 mM NaCl). Protein concentrations were

determined spectrophotometrically using extinction coefficients calcu-lated from the tryptophan and tyrosine contents (48).

Pull-Down Assay. To measure the binding of 4.1R 80-kDa and its functionaldomains and subdomains of the 30-kDa domain to cytoplasmic tails of XK, Duffy,and Rh, 4.1R or its domains were incubated with the biotin-labeled syntheticpeptide at room temperature for 1 h. Streptavidin beads were added to thereaction mixture, incubated for 10 min, pelleted, washed, and eluted with 10%SDS. The pellet was analyzed by SDS/PAGE, followed by transfer to nitrocellulosemembrane and the subsequent Western blotting. The binding of 4.1R to thecytoplasmic domains of XK, Duffy, and Rh was detected by using anti-4.1Rantibody. Similarly, the binding of GST-tagged 30-kDa, 16-kDa, 10-kDa and22/24-kDa domains of 4.1R was detected by using anti-GST antibody, and thebindingofMBP-taggedlobeA, lobeb,andlobeCwasdetectedbyusinganti-MBPantibody.

ACKNOWLEDGMENTS. This work was supported in part by National Institutesof Health Grants DK26263, DK32094, HL31579, HL78826, and HL075716.

1. Mohandas N, Evans E (1994) Mechanical properties of the red cell membrane in relationto molecular structure and genetic defects. Annu Rev Biophys Biomol Struct 23:787–818.

2. Discher DE, Carl P (2001) New insights into red cell network structure, elasticity, andspectrin unfolding—a current review. Cell Mol Biol Lett 6:593–606.

3. Shotton DM, Burke BE, Branton D (1979) The molecular structure of human erythrocytespectrin. Biophysical and electron microscopic studies. J Mol Biol 131:303–329.

4. Bennett V, Baines AJ (2001) Spectrin and ankyrin-based pathways: Metazoan inven-tions for integrating cells into tissues. Physiol Rev 81:1353–1392.

5. Takakuwa Y, Tchernia G, Rossi M, Benabadji M, Mohandas N (1986) Restoration ofnormal membrane stability to unstable protein 4.1-deficient erythrocyte membranesby incorporation of purified protein 4.1. J Clin Invest 78:80–85.

6. Reid ME, Mohandas N (2004) Red blood cell blood group antigens: Structure andfunction. Semin Hematol 41:93–117.

7. Soupene E, et al. (2002) Rhesus expression in a green alga is regulated by CO2. Proc NatlAcad Sci USA 99:7769–7773.

8. Soupene E, Lee H, Kustu S (2002) Ammonium/methylammonium transport (Amt) proteinsfacilitate diffusion of NH3 bidirectionally. Proc Natl Acad Sci USA 99:3926–3931.

9. Chaudhuri A, et al. (1994) Expression of the Duffy antigen in K562 cells. Evidence thatit is the human erythrocyte chemokine receptor. J Biol Chem 269:7835–7838.

10. Chaudhuri A, et al. (1989) Purification and characterization of an erythrocyte mem-brane protein complex carrying Duffy blood group antigenicity. Possible receptor forPlasmodium vivax and Plasmodium knowlesi malaria parasite. J Biol Chem 264:13770–13774.

11. Parsons SF, Spring FA, Chasis JA, Anstee DJ (1999) Erythroid cell adhesion moleculesLutheran and LW in health and disease. Baillieres Best Pract Res Clin Haematol12:729–745.

12. Oldenborg PA, et al. (2000) Role of CD47 as a marker of self on red blood cells. Science288:2051–2054.

13. Lee S, et al. (1999) Proteolytic processing of big endothelin-3 by the kell blood groupprotein. Blood 94:1440–1450.

14. Thevenin BJ, Willardson BM, Low PS (1989) The redox state of cysteines 201 and 317 ofthe erythrocyte anion exchanger is critical for ankyrin binding. J Biol Chem 264:15886–15892.

15. Davis L, Lux SE, Bennett V (1989) Mapping the ankyrin-binding site of the humanerythrocyte anion exchanger. J Biol Chem 264:9665–9672.

16. Willardson BM, et al. (1989) Localization of the ankyrin-binding site on erythrocytemembrane protein, band 3. J Biol Chem 264:15893–15899.

17. Michaely P, Bennett V (1995) The ANK repeats of erythrocyte ankyrin form two distinctbut cooperative binding sites for the erythrocyte anion exchanger. J Biol Chem270:22050–22057.

18. Sterling D, Reithmeier RAF, Casey JR (2001) A transport metabolon: Functional inter-action of carbonic anhydrase II and chloride/bicarbonate exchangers. J Biol Chem276:47886–47894.

19. Zhang D, Kiyatkin A, Bolin JT, Low PS (2000) Crystallographic structure and functionalinterpretation of the cytoplasmic domain of erythrocyte membrane band 3. Blood96:2925–2933.

20. Telen MJ, Chasis JA (1990) Relationship of the human erythrocyte Wrb antigen to aninteraction between glycophorin A and band 3. Blood 76:842–848.

21. Knowles DW, Chasis JA, Evans EA, Mohandas N (1994) Cooperative action betweenband 3 and glycophorin A in human erythrocytes: Immobilization of band 3 induced byantibodies to glycophorin A. Biophys J 66:1726–1732.

22. Hassoun H, et al. (1998) Complete deficiency of glycophorin A in red blood cells frommice with targeted inactivation of the band 3 (AE1) gene. Blood 91:2146–2151.

23. Auffray I, et al. (2001) Glycophorin A dimerization and band 3 interaction duringerythroid membrane biogenesis: In vivo studies in human glycophorin A transgenicmice. Blood 97:2872–2878.

24. Daniels GL, et al. (1999) Terminology for red cell surface antigens. ISBT Working PartyOslo Report. International Society of Blood Transfusion. Vox Sang 77:52–57.

25. Avent ND, Reid ME (2000) The Rh blood group system: A review. Blood 95:375–387.

26. Bruce LJ, et al. (2003) A band 3-based macrocomplex of integral and peripheralproteins in the RBC membrane. Blood 101:4180–4188.

27. Marfatia SM, Leu RA, Branton D, Chishti AH (1995) Identification of the protein 4.1binding interface on glycophorin C and p55, a homologue of the Drosophila discs-largetumor suppressor protein. J Biol Chem 270:715–719.

28. Marfatia SM, Morais-Cabral JH, Kim AC, Byron O, Chishti AH (1997) The PDZ domain ofhuman erythrocyte p55 mediates its binding to the cytoplasmic carboxyl terminus ofglycophorin C. Analysis of the binding interface by in vitro mutagenesis. J Biol Chem272:24191–24197.

29. Nunomura W, Takakuwa Y, Parra M, Conboy J, Mohandas N (2000) Regulation ofprotein 4.1R, p55, and glycophorin C ternary complex in human erythrocyte mem-brane. J Biol Chem 275:24540–24546.

30. Shi ZT, et al. (1999) Protein 4.1R-deficient mice are viable but have erythroid membraneskeleton abnormalities. J Clin Invest 103:331–340.

31. Reid ME, Takakuwa Y, Conboy J, Tchernia G, Mohandas N (1990) Glycophorin C contentof human erythrocyte membrane is regulated by protein 4.1. Blood 75:2229–2234.

32. Lorenzo F, et al. (1993) Severe poikilocytosis associated with a de novo alpha 28Arg3Cys mutation in spectrin. Br J Haematol 83:152–157.

33. Stoya G, Baumann E, Junker U, Hermann J, Linss W (1997) Flow cytometric analysis ofband 3 protein of human erythrocytes. Acta Histochem 99:29–36.

34. Han BG, Nunomura W, Takakuwa Y, Mohandas N, Jap BK (2000) Protein 4.1R coredomain structure and insights into regulation of cytoskeletal organization. Nat StructBiol 7:871–875.

35. Yi SJ, et al. (1997) Red cell membranes of ankyrin-deficient nb/nb mice lack band 3tetramers but contain normal membrane skeletons. Biochemistry 36:9596–9604.

36. Allenspach EJ, et al. (2001) ERM-dependent movement of CD43 defines a novel proteincomplex distal to the immunological synapse. Immunity 15:739–750.

37. Morris MB, Lux SE (1995) Characterization of the binary interaction between humanerythrocyte protein 4.1 and actin. Eur J Biochem 231:644–650.

38. Tchernia G, Mohandas N, Shohet S (1981) Deficiency of skeletal membrane proteinband 4.1 in homozygous hereditary elliptocytosis. Implications for erythrocyte mem-brane stability. J Clin Invest 68:454–460.

39. Palek J, Lux SE (1983) Red cell membrane skeletal defects in hereditary and acquiredhemolytic anemias. Semin Hematol 20:189–224.

40. Conboy J, Mohandas N, Tchernia G, Kan YW (1986) Molecular basis of hereditaryelliptocytosis due to protein 4.1 deficiency. N Engl J Med 315:680–685.

41. Yawata A, Kanzaki A, Gilsanz F, Delaunay J, Yawata Y (1997) A markedly disruptedskeletal network with abnormally distributed intramembrane particles in completeprotein 4.1-deficient red blood cells (allele 4.1 Madrid): Implications regarding a criticalrole of protein 4.1 in maintenance of the integrity of the red blood cell membrane.Blood 90:2471–2481.

42. Chasis JA, et al. (1996) Differential use of protein 4.1 translation initiation sites duringerythropoiesis: Implications for a mutation-induced stage-specific deficiency of pro-tein 4.1 during erythroid development. Blood 87:5324–5331.

43. Fowler VM (1996) Regulation of actin filament length in erythrocytes and striatedmuscle. Curr Opin Cell Biol 8:86–96.

44. An X, Salomao M, Guo X, Gratzer W, Mohandas N (2007) Tropomyosin modulateserythrocyte membrane stability. Blood 109:1284–1288.

45. King MJ, et al. (2000) Rapid flow cytometric test for the diagnosis of membranecytoskeleton-associated haemolytic anaemia. Br J Haematol 111:924–933.

46. An XL, et al. (2001) Structural and functional characterization of protein 4.1R-phosphatidylserine interaction: potential role in 4.1R sorting within cells. J Biol Chem276:35778–35785.

47. Nunomura W, et al. (1997) Regulation of CD44-protein 4.1 interaction by Ca2� andcalmodulin. Implications for modulation of CD44-ankyrin interaction. J Biol Chem272:30322–30328.

48. Perkins SJ (1986) Protein volumes and hydration effects. The calculations of partialspecific volumes, neutron scattering matchpoints and 280-nm absorption coeffi-cients for proteins and glycoproteins from amino acid sequences. Eur J Biochem157:169 –180.

Salomao et al. PNAS � June 10, 2008 � vol. 105 � no. 23 � 8031

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0