Aggregation of band 3 in hereditary ovalocytic red blood cell membranes Electron microscopy and protein rotational diffusion studies

Hereditary ovalocytosis occurs in South-east Asian populations in areas where malaria is endemic (Castelino et al., 1981). Ovalocytic erythrocytes are strongly resistant to invasion by malarial parasites in vitro (Kidson et al., 1981), though only slightly so in vivo (Cattani et al., 1987). It has recently been shown that resistance in vitro is largely a consequence of rapid ATP depletion in ovalocytes (Dluzewski et al., 1992). A particular feature of ovalocytes is their high degree of rigidity (Mohandes et al., 1984; Saul et al., 1984), which, surprisingly, appears to be a consequence of a mutation in band 3, the anion transporter of red cells, rather than a mutation in a cytoskeletal protein (Jarolim et al., 1991; Schofield et al., 1992a).

Band 3 in hereditary ovalocytes in fact has two mutations. Firstly, a point mutation occurs at residue 56 involving substitution of Lys by Glu; this mutation is also found in a common asymptomatic variant known as band 3 Memphis. Secondly, nine amino acids (residues 400-408) are deleted at the interface between the membrane-associated and cytoplasmic domains of band 3 (Jarolim et al., 1991; Schofield et al., 1992a; Mohandes et al., 1992).

The cytoplasmic domain of band 3 is linked to the erythrocyte cytoskeleton by ankyrin, which has binding sites for both spectrin and band 3 (Bennett and Stenbuck, 1979, 1980). Exactly how the mutations in ovalocytic band 3 increase rigidity is, however, unclear. Schofield et al. (1992a) found that ankyrin binding to ovalocytic band 3 is normal, though this was in disagreement with earlier studies by Liu et al. (1991). A powerful method of investigating band 3-cytoskeletal interactions is to measure the rotational mobility of band 3 (Nigg and Cherry, 1980). In normal erythrocytes transient dichroism and phosphorescence depolarization measurements with eosin-labelled band 3 indicate that almost half of the band 3 population has restricted rotational mobility arising from interactions with cytoskeletal proteins (Nigg and Cherry, 1980; Matayoshi and Jovin, 1991). This approach was recently employed to study Melanesian ovalocytes by Tilley et al. (1991), who demonstrated that there is a marked increase in the fraction of immobile band 3 in membranes from these cells.

Loss of band 3 mobility could arise from changes in cytoskeletal interactions or from aggregation of band 3 molecules in the membrane. Tilley et al. (1991) did not observe any aggregation by immunofluorescence microscopy but this does not rule out microaggregation below the resolution of optical microscopy. Here we have used freeze-fracture electron microscopy to investigate band 3 aggregation in ovalocytes and performed further studies of band 3 rotational mobility to determine the role of cytoskeletal interactions in immobilising band 3 in ovalocytic membranes.

Ovalocytes were obtained from 22-year-old Mauritian twins of Indian background, the same source as that employed in the studies of Schofield et al. (1992a,b). Control cells were either normal erythrocytes or erythrocytes from a subject with the common band 3 Memphis variant. Cells were taken by venipuncture and stored as whole blood in heparin at 4°C.

Fresh or stored cells were washed three times in 0.1 M potassium phosphate (K₂HPO₄/KH₂PO₄) buffer (pH 7.2) and then fixed in 2.5% (v/v) glutaraldehyde in the same buffer for 2 h at 4°C. Cells were then washed twice and stored at 4°C in the phosphate buffer containing 2% (w/v) NaN₃. Some ovalocytes were treated before fixation with PIGPA (50 mM pyruvate, 50 mM inosine, 100 mM glucose, 50 mM sodium phosphate and 150 mM sodium chloride) to reconstitute their ATP content (Valeri, 1974). For this purpose, 100 μl of packed red cells were suspended in 4.5 ml phosphate buffered saline (PBS) and 0.5 ml 10× PIGPA at 37°C for 1 h with intermittent shaking. Cells were washed twice in 0.1 M phosphate buffer before fixation.

For scanning electron microscopy, fixed cells were washed four times, then post-fixed with 1% osmium tetroxide in 0.1 M phosphate buffer for 1 h at 4°C. They were then passed as pellets through a series of acetone solutions to 100% acetone, and critical-point dried. The pellets were then broken up on double-sided sellotape on scanning electron microscope metal stubs, and gold coated. Cells were viewed at magnifications of ×13,000 to ×15,000.

For freeze fracture, fixed and washed red cells were cryoprotected with 25% glycerol in phosphate buffer for 24 h at 4°C, rapidly frozen by quenching in nitrogen slush and fractured and replicated in a Reichert-Jung cryofracture apparatus, using a Balzers specimen holder.

Band 3 was labelled with eosin-5-maleimide (Molecular Probes Inc) in intact cells from which ghosts were prepared as previously described (Nigg and Cherry, 1979). For studies with ghosts depleted of cytoskeletal proteins, band 6 was first removed by washing twice with PBS (pH 7.5). Removal of spectrin and actin by low salt incubation and subsequent removal of ankyrin and band 4.1 by high salt incubation were performed essentially as described by Nigg and Cherry (1980). Removal of the band 3 cytoplasmic domain by mild proteolysis with trypsin was also performed according to Nigg and Cherry (1980). Membranes depleted of band 6, spectrin, actin, ankyrin and band 4.1 are hereafter referred to as ‘stripped’ ghosts.

Protein was determined by the method of Lowry et al. (1951) using BSA as a standard after solubilising ghosts in 1% SDS, whilst the eosin concentration was determined by measuring absorbance at 531 nm. The eosin:band 3 molar ratio was calculated assuming band 3 accounted for 25% of the total membrane protein and taking the extinction coefficient for bound eosin to be 83000 M−1 cm−1 (Cherry et al., 1976). The ratio was found to be approximately 0.9:1 in all samples.

by a non-linear least squares analysis. With some sets of data, particularly for ovalocytic membranes, inclusion of the second exponential term was not justified by the data and this term was therefore omitted in the curve fitting. The coefficient r3 is related to the fraction of molecules that are immobile on the time scale of the experiment. For purpose of comparison, r₃ was calculated as a percentage of the initial anisotropy. Data analysis and interpretation are further discussed by Nigg and Cherry (1979) and by Matayoshi and Jovin (1991).

All samples were flushed with argon prior to measurement to obviate quenching of the eosin triplet state by oxygen. The eosin concentration was typically 1-2 μM and all experiments were performed at 37°C.

The results of transient dichroism measurements with eosinlabelled ovalocytic membranes are shown in Fig. 1. In Fig. 1a it can be seen that removal of spectrin and actin by low salt incubation has very little effect on the anisotropy decay. In membranes stripped of spectrin, actin, ankyrin and band 4.1, however, there is a marked increase in the decay of anisotropy indicating increased band 3 mobility (Fig. 1b). A similar result is obtained when the cytoplasmic domain of band 3 is removed by proteolysis with trypsin (Fig. 1c). For comparison, a corresponding set of anisotropy decays for normal membranes are shown in Fig. 1d-f.

To facilitate comparison, the values of r3 calculated as a percentage of initial anisotropy are shown in Table 1 for controls, stripped and trypsin-treated membranes from three different samples. Corresponding values previously obtained from normal cells are also included. It can be seen that r3 is significantly decreased in both ovalocytes and normal cells by stripping of cytoskeletal protein or by proteolysis of band 3 but that values of r3 are always higher for ovalocytes than for normal cells under the same conditions. The value of r3 for band 3 Memphis was similar to that of normal band 3 (data not shown). The triplet lifetime of the eosin probe was essentially the same for ovalocytic and normal band 3 under all conditions investigated.

The surfaces of these cells were noticeably more irregular than normal erythrocytes, as seen with both scanning electron microscopy (Fig. 2A,B) and low power electron micrographs of freeze-fracture replicas (Fig. 2C), with shallow depressions and low elevations, some cells being more irregular than others. At higher magnifications replicas of normal erythrocytes (Fig. 3A) and of Memphis variants (Fig. 3B) were identical with each other, with randomly clustered and non-clustered intramembranous particles (IMP) on P faces. In untreated ovalocytes, some P face IMPs were distributed as in normal cells, but there were also numerous linearly arranged IMPs, forming straight or slightly curved single lines of 10-15 particles in single rows (Fig. 3C-F); the sizes and shapes of the IMPs did not differ significantly from those of normal red cells. The same results were obtained in two experiments, one of which was performed with fresh cells and the other with cells stored for eight days.

PIGPA-treatment of ovalocytes did not alter their structure appreciably, except for a tendency for their surfaces to be rather more regular. In particular, ATP recovery in stored ovalocytes does not affect the formation of the linear arrays (Fig. 3D).

In agreement with Tilley et al. (1991), we find that labelling of band 3 by eosin-5-maleimide in ovalocytes is similar to normal cells with molar labelling ratio eosin:band 3 of close to 1:1. This is in contrast to labelling with H2-DIDS (diisothiocyanato-dihidrostilbene disulphonate) where Schofield et al. (1992b) found little more than half of band 3 was labelled in ovalocytes, the label being associated with normal band 3 in these heterozygous cells. Although both eosin-5-maleimide and H2-DIDS are potent inhibitors of the anion transport function of band 3 and occupy partially overlapping sites, they attach covalently to different amino acids in the band 3 sequence (Cobb and Beth, 1990).

Fig. 1 and Table 1 show that band 3 is strongly immobilised in ovalocytes compared with normal cells. This is also in agreement with Tilley et al. (1991) whose value for the immobile fraction of band 3 in ovalocytes corresponds to an r3 of 81%. It should be noted that although these heterozygous cells contain about 60% normal band 3 (Schofield et al., 1992), dimers or tetramers of band 3, which are believed to be the predominant forms of band 3 in the membrane, will mostly contain at least one copy of the mutant band 3.

We have further investigated the origin of this immobilisation by removing cytoskeletal proteins from the membrane. Removal of spectrin and actin has at most a rather small effect on band 3 rotational mobility in ovalocytes, which is within the range observed with normal cells (Fig. 1a). This result rules out a more extensive linkage of band 3 to spectrin (via ankyrin) or the formation of a more rigid link as the explanation of reduced rotational mobility in ovalocytes.

After additional removal of ankyrin and band 4.1, band 3 rotational mobility is markedly increased in ovalocytes, as it is in normal ghosts (Fig. 1b). However, mobility is still significantly less than in normal cells under the same conditions. A similar result is obtained when the cytoskeletal linkage is broken by dissociating the cytoplasmic domain of band 3 with trypsin. A difference between ovalocytic and normal ghosts is that trypsin produces a comparable effect to stripping of cytoskeletal proteins in ovalocytic membranes, whereas the trypsin treatment is always more effective in normal membranes. It is nevertheless clear that the reduced mobility of band 3 in ovalocytic membranes compared with normal membranes persists after removal of ankyrin and band 4.1 or after cleavage of band 3 and must therefore be a property of the transmembrane domain of band 3. This in turn indicates that the 9 amino acid deletion that resides in the membrane-bound domain is responsible, rather than the point mutation at residue 56, which is in the cleavable cytoplasmic domain. The fact that band 3 Memphis, which has the same point mutation but not the deletion, has normal rotational mobility points to the same conclusion.

The simplest explanation of the rotational mobility data is that ovalocytic band 3 has an increased tendency to aggregation compared with normal band 3. Rotational diffusion measurements are particularly sensitive to the size of the rotating particle (Saffman and Delbrück, 1975) and hence readily detect microaggregation of membrane proteins. It should be noted that even normal band 3 self-associates into slowly-rotating or immobile aggregates under a variety of experimental conditions (Cherry, 1992). As in normal membranes, additional constraints are imposed in ovalocytes by interactions involving cytoskeletal proteins and the cytoplasmic domain of band 3. The precise nature of these constraints is still not fully understood (Wyatt and Cherry, 1992).

Freeze-fracture electron microscopy was employed in order to further investigate ultrastructural differencies between ovalocytes and normal cells. These experiments yielded the striking observation that the IMPs (which consist principally of band 3) often form linear clusters in ovalocytes (Fig. 3C-F). Such clusters are not observed in normal cells or Memphis variants (Fig. 3A,B). The long linear clusters, which are typically 10-15 particles long, would be expected to have a very low rotational mobility (about 2 orders of magnitude less than single particles). There are also many smaller clusters. Although it is difficult to quantitatively relate the freeze-fracture and rotational diffusion data, the existence of linear clusters is in excellent qualitative agreement with the transient dichroism data and strongly supports the band 3 aggregation hypothesis.

A further morphological characteristic of ovalocytes is a marked though somewhat variable tendency to surface irregularity (Fig. 2). Otherwise particle-free membrane patches were not different in frequency or size compared with normal cells or Memphis variants. It was possible, however, that some clustering had occurred in these latter two types of cells prior to fixation as the IMPS were less evenly spaced than in some published freeze-fracture electron micrographs.

As rather little is known about the detailed structure of band 3, the mechanism whereby deletion of residues 400-408 enhances band 3 aggregation is largely a matter of speculation. These residues are largely hydrophobic and are inferred to form part of the first transmembrane helix of the membrane domain, which is postulated to start at residue 403 (Tanner et al., 1988). The loss of these residues might seriously perturb the structure of the protein in this region, either by partial unfolding of the helix or recruitment of adjacent more polar residues into the transmembrane helix in order to bridge the gap left by the deleted sequence. In either case there would be unfavourable interactions with the hydrophobic interior of the membrane, which might be alleviated by protein association. On symmetry grounds, the formation of linear aggregates suggests that the interacting oligomers are dimers or possibly linear tetramers.

It is not difficult to envisage how band 3 clustering could affect the mechanical properties of the cell. A linear array of 10-15 IMPs would probably contain several attachment points for the cytoskeleton that could inhibit expansion of the spectrin lattice, which is believed to occur when red cells deform. The likely distortion of the lattice to accommodate the band 3 arrays could also cause the surface irregularities seen by electron microscopy.

Thanks are due to John Manston of the Electron Microscope Unit at Queen Mary-Westfield College and Tony Braine at the King’s College London Electron Microsopy Unit for preparing the freeze-fracture specimens, and to Ken Brady of the Electron Microscope Unit at UMDS for preparation of scanning electron microscope material. We also thank Dr Walter Gratzer, King’s College London, for his interest and encouragement.

This work was supported in part by the UNDP/World Bank/World Health Organization Special Programme for Research and Training in Tropical Diseases and in part by the SERC.