Brain Ankyrin PURIFICATION OF A 72,000 M, SPECTRIN-BINDING DOMAIN*

Polypeptides of M, = 190,000-220,000 that cross- react with erythrocyte ankyrin were detected in immunoblots of membranes from pig lens, pig brain, and rat liver. The cross-reacting polypeptides from brain were cleaved by chymotrypsin to fragments of M, = 95,000 and 72,000 which are the same size as fragments obtained with erythrocyte ankyrin. The brain 72,000 M, fragment associated with erythrocyte spectrin, and the binding occurred at the same site as that of erythrocyte ankyrin 72,000 M. fragment since (a) brain 72,000 M,fragment was adsorbed to erythrocyte spectrin-agarose and (b) 12sI-labeled erythrocyte spectrin bound to brain 72,000 M. fragment following transfer of the fragment from a sodium dodecyl sulfate gel to nitrocellulose paper, and this binding was dis- placed by erythrocyte ankyrin 72,000 M, fragment. Brain series DEAE-cellulose followed The brain 72,000 M, was not erythrocytes since peptide maps and pig erythrocyte 72,000 M. fragments were distinct.

The amount of brain 72,000 M, fragment was estimated as 0.28% of membrane protein or 39 pmol/mg based on radioimmunoassay with 12SI-labeled brain fragment and antibody against erythrocyte ankyrin. Brain spectrin tetramer was present in about the same number of copies (30 pmol/mg of membrane protein) based on densitometry of Coomassie blue-stained sodium dodecyl sulfate gels. The binding site on brain spectrin for both brain and erythrocyte ankyrin 72,000 M, fragments was localized by electron microscopy to the midregion of spectrin tetramers about 90 nM from the near end and 110 nM from the far end. These studies demonstrate the presence in brain membranes of a protein closely related to erythrocyte ankyrin, and are consistent with a function of the brain ankyrin as a membrane attachment site for brain spectrin.
The human erythrocyte membrane is currently the best * This research was supported by National Institutes of Health Grant RO 1 AM29808. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.  understood system in terms of knowledge of the organization of its membrane and cytoskeletal proteins (see reviews by Branton et al., 1981;Bennett, 1982). The major integral membrane protein, which contains an anion channel, is associated on the cytoplasmic surface of the membrane with ankyrin. Ankyrin, in turn, is attached through a 72,000 M , domain to spectrin which is a flexible rod-shaped protein 180 nm in length composed of two subunits. The subunits of spectrin are aligned side-to-side to form heterodimers, and the dimers are assembled by head-to-head association to form tetramers. Spectrin tetramers bind at their ends to a protein named band 4.1 and to actin oligomers. Spectrin tetramers with associated actin and band 4.1 form a two-dimensional network that lines the inner surface of the plasma membrane and provides mechanical stability for the fragile lipid bilayer.
Analogues of erythrocyte membrane proteins are widely distributed in other cell types. Nonerythroid spectrin was initially identified on the basis of cross-reaction with antibody against erythrocyte spectrin (Goodman et at., 1981;Repasky et al., 1982;Bennett et al., 1982a;Burridge et al., 1982). Immunoreactive forms of band 4.1 have also been detected in other cells (Cohen et al., 1982). Brain spectrin has been purified and demonstrated to have properties quite similar to erythrocyte spectrin, including two subunits arranged as a tetramer with the morphology of a flexible rod 200 nm in length, and binding sites for actin, band 4.1 and ankyrin (Bennett et al., 1982a, Burridge et al., 1982Glenney et al., 1982;Burns et al., 1983;Lin et al., 1983). Subunits of brain spectrin tetramers are arranged the same as those of erythrocyte spectrin, and it is possible to prepare functional hybrids of subunits of brain and erythrocyte spectrin (Davis and Bennett, 1983).
Immunoreactive forms of erythrocyte ankyrin have been detected by radioimmunoassay in membranes and whole cells from a variety of tissues (Bennett, 1979). The polypeptides cross-reacting with erythrocyte ankyrin include microtubuleassociated proteins localized in the cytoplasm and in mitotic structures (Bennett and Davis, 1981;Bennett et al., 198213). In addition, nonerythroid cells contain membrane-associated polypeptides of M, -200,000 that cross-react with ankyrin (Bennett et al., 1982b). This report describes further studies with membrane-associated forms of ankyrin from brain, including purification of a 72,000 M , fragment that binds to spectrin. Brain ankyrin is present in approximately the same quantities as brain spectrin tetramers and is a logical candidate for a membrane-attachment site for brain spectrin. These studies provide additional support for the view that the organization of the erythrocyte membrane will have direct relevance for other cells. cinimide ester were from Calbiochem-Behring. Plastic thin layer sheets coated with 0.1-mm thick cellulose were from E. Merck. PMSF,' DFP, D m , pepstatin A, leupeptin, Triton X-100, and pancreatic trypsin inhibitor were from Sigma. Nitrocellulose paper and electrophoresis reagents were from Bio-Rad. Sucrose, urea, and ammonium sulfate were from Schwarz/Mann. a-Chymotrypsin (54 units/mg) was from Worthington. Cyanogen bromide-activated Sepharose 4B, Protein A, Protein A-Sepharose, and Sephacryl S-500 were from Pharmacia. Avidin-ferritin was from LKB. Pig erythrocyte spectrin was purified by chromatography on a Sephacryl S-500 column as described (Bennett, 1983). Ankyrin 72,000 M, fragment was purified from pig and human erythrocytes as described (Bennett, 1978). Affinity-purified rabbit antibody against erythrocyte ankyrin was prepared as described . Preimmune Ig was isolated by affinity chromatography on Protein A-Sepharose, using the same elution conditions as for immune antibody. Pig brains were obtained from a local slaughter house; tissue from the cerebral cortex was dissected free of connective tissue, washed with 0.25 M sucrose and frozen in liquid nitrogen. Frozen brain was stored at -100 "C and used within 6 weeks.
Methods-Tissues except for liver were homogenized with a Brinkman Polytron (large head) for 30-60 s at a setting of 5.5. Liver was disrupted in a Dounce homogenizer. SDS-polyacrylamide electrophoresis was performed on 3.5-17% exponential gradient slab gels with the buffers of Fairbanks et al. (1971). Protein was determined by the method of Bradford (1976) with bovine serum albumin as a standard. The of brain ankyrin 72,000 M, fragment was 9.0 based on protein determination by Bradford assay, and A m was used to estimate protein with purified preparations of fragment. Autoradiography was performed at -100 "C with X-Omat AR film (Kodak) and Cronex intensifier screens (DuPont). Proteins were radioiodinated with Na-'%I using chloramine-T as an oxidant (Hunter and Greenwood, 1962).
Immunoblot analysis was performed by electrophoretically transferring proteins from SDS gels to nitrocellulose paper (Towbin et al., 1979) using conditions as described . The nitrocellulose paper was incubated 15 min at 24 "C in immunoblot buffer (40 mg/ml of bovine serum albumin, 150 mM NaCI, 10 mM sodium phosphate, 1 mM Na EDTA, 1 mM NaN3, 0.2% (v/v) Triton X-100, pH 7.5) and then for 14 h at 4 "C with the same buffer and 0.5-1 pg/ml of antibody. The nitrocellulose paper was washed 5 times with buffer (no albumin) and once with 2 M urea, 0.1 M glycine, 1% Triton X-100. The bound antibody was labeled by incubation for 2 h at 4 "C with '261-labeled Protein A (0.5-1.5 Cilpmol; lo6 cpm/ml final concentration) in immunoblot buffer. The nitrocellulose was washed as before and dried and an autoradiogram was prepared.
Pig erythrocyte spectrin-agarose was prepared by addition of cyanogen bromide-activated Sepharose 4B to an equal volume of solution containing 2 mg/ml of spectrin, 25 mM sodium phosphate, 750 mM NaCI, pH 8. The suspension was mixed gently a t 4 "C for 2 h, and then poured into a column and washed with 1 liter of 1 M NaCI, 0.1 M glycine, 0.5% Triton X-100,l mM NaN3. The gel was then washed with 1 liter of 0.1 M NaCI, 10 mM sodium phosphate, 1 mM NaN3, and stored in this buffer. The affinity column was regenerated after use by washing with the high salt/Triton X-100 buffer and could be used twice.

Identification of Membrane-associated Polypeptides Crossreacting with Erythrocyte
Ankyrin-A polypeptide of M, = 190,000 cross-reacting with erythrocyte ankyrin has been identified in rat liver plasma membranes by the immunoblot technique (Bennett et al., 1982b). Membranes from pig brain, pig lens, and rat liver were analyzed by the same method ( Fig.  1). Lens membranes contain a major cross-reacting polypeptide of M, = 220,000 and many polypeptides of lower Mr. Brain   FIG. 1. Identification of polypeptides cross-reacting with erythrocyte ankyrin in membranes of pig lens, rat liver, and pig brain. Tissues were homogenized (see "Experimental Procedures") in 6 volumes of a solution containing 250 mM sucrose, 10 mM sodium phosphate, 2 mM sodium EGTA, 200 pg/ml of PMSF, 0.015% (v/v) DFP, 5 pg/ml of leupeptin, and 2 pg/ml of pepstatin A, pH 7.5. Liver plasma membrane sheets were isolated by upward flotation of the nuclear pellet fraction through 1.42 M sucrose dissolved in 2 mM sodium EGTA, 10 mM sodium phosphate, pH 7.5 (Bennett et al., 1982a). Crude membrane fractions from brain and lens were isolated by pelleting for 5 min at 1,500 X g to remove nuclei and then centrifuging for 30 min at 30,000 X g to pellet membranes. These particulate fractions were washed once, dissolved in SDS, and electrophoresed on a SDS-polyacrylamide gel stained with Coomassie blue (left): pig erythrocyte ghosts (lane I ) , pig lens membranes (lane 2). pig brain membranes (lane 3), and rat liver plasma membranes (lane 4). Proteins in a parallel gel were transferred electrophoretically to nitrocellulose paper to detect polypeptides cross-reacting with erythrocyte ankyrin (see "Experimental Procedures"). time required to isolate membranes. The M, = 215,000 and 190,000 polypeptides may be proteolytic products of the M, = 220,000 or may represent related isoforms with some common sequence but which are products of different genes. Liver plasma membranes contained a major polypeptide of M, = 190,000 and a fainter cross-reacting band at M, = 215,000. Control immunoblots with preimmune Ig labeled no detectable polypeptides (Fig. 1). Such controls with preimmune Ig were negative in other immunoblot experiments  and are not shown.
The polypeptides cross-reacting with erythrocyte ankyrin are especially sensitive to exogenous protease (see below) as well as tissue proteases. The gels presented here were the best of several experiments, and were obtained with membranes isolated rapidly and with protease inhibitors DFP, leupeptin, pepstatin A, PMSF, and EGTA (to inhibit Ca2+-dependent protease). The cross-reacting polypeptides were especially sensitive to a leupeptin-inhibited protease activity. Lens membranes exhibited multiple cross-reacting bands in spite of these precautions, and it is possible some degradation occurred in uiuo.
Brain membranes were chosen for further studies since these can be obtained in large quantities. A crude subcellular fractionation of brain indicated that the cross-reacting M, = 220,000 and 215,000 polypeptides were confined almost entirely to particulate fractions (Fig. 2). The cross-reacting polypeptides co-migrated with the major peak of membrane 1876 Brain Ankyrin C. BLUE Mr x10-= .

IMMUNOBLOT I 2 3 4 5 6
FIG. 2. Subcellular distribution of brain polypeptides crossreacting with erythrocyte ankyrin. Frozen pig brain was homogenized (see "Experimental Procedures") in ten volumes of 0.32 M sucrose, 2 mM sodium EGTA, 200 pg/ml of PMSF, 0.015% (v/v) DFP, 5 pg/ml of leupeptin, and 2 pg/ml of pepstatin A, and centrifuged at 900 X g for 10 min. The pellets from centrifugation of the 900 X g supernatant at 30,000 X g, and 200,000 X g, respectively, were resuspended to the original volume of homogenization buffer. The resuspended 30,000 X g pellet was layered over a 13-ml linear gradient of 15-60% sucrose dissolved in 2 mM sodium EGTA, 100 pg/ml of PMSF, pH 7.5, and centrifuged 4 h at 40,000 X g in a SW-41 rotor. Fractions containing the dense membrane peak (40-50% sucrose) and the lighter myelin membranes were pooled. The samples were dissolved in SDS and electrophoresed on an SDS-polyacrylamide gel protein when membranes were fractionated by isopycnic centrifugation on sucrose gradients, but were deficient in the myelin fractions (Fig. 2). The distribution of these bands paralleled approximately that of the brain spectrin doublet (M, = 260,000 and 265,000). It was difficult to compare exactly the amounts of spectrin and cross-reacting polypeptides because of the nonquantitative nature of immunoblots.
The cross-reacting polypeptides are associated tightly with membranes, since they were not extracted by repeated washes in 0.5 M NaCl which removed about 50% of the spectrin (Fig.  3). Similarly, the polypeptides were not solubilized by extraction of membrane at low ionic strength which also entracted spectrin (not shown).
Proteolysis of Brain Cross-reacting Polypeptides to a Membrane-associated Fragment of M, = 95,000 and a Spectrinbinding Fragment of M, = 72,000-Erythrocyte ankyrin contains two major protease-resistant domains, one of MI = 72,000 which contains the spectrin-binding site (Bennett, 1978) and another of M , = 95,000 (Bennett and Stenbuck, 1980). The 95,000 MI domain contains the anion channel binding site and remains associated with membranes after cleavage.' An important criterion for nonerythroid ankyrin is that these proteins should have a similar domain structure. The brain cross-reacting polypeptides fulfill this requirement, 3. Extraction of a 72,000 M. polypeptide cross-reacting with erythrocyte ankyrin following digestion of brain membranes with a-chymotrypsin. Brain membranes were processed in the same fashion as described in Fig. 5 except that all postnuclear centrifugation was performed in a 60 Ti rotor (50,000 rpm, 10 min). Briefly, membranes were washed first in 10 mM sodium phosphate, 1 mM sodium EGTA, 50 pg/ml of PMSF, pH 7.5, and then extracted twice (30 min, 4 "C) in the same buffer with 0.5 M NaCI. The membranes were resuspended in buffer without salt or PMSF and digested with 20 pg/ml of a-chymotrypsin for 1 h at 4 "C. Proteins in a parallel gel were transferred electrophoretically to nitrocellulose, and polypeptides cross-reacting with erythrocyte ankyrin were detected (see "Experimental Procedures"). since limited digestion with a-chymotrypsin degraded the MI = 220,000, 215,000, and 190,000 polypeptides to fragments of M, = 95,000 and 72,000 (Fig. 3). The 72,000 M, fragment was extracted from the digested membranes with 0.5 M NaCI, while the major portion of the 95,000 M, fragment remained membrane-bound (Fig. 3). The persistent binding of the 95,000 M, fragment while the 72,000 M, fragment was extracted suggests that the 95,000 M, fragment is primarily responsible for attachment of the intact polypeptide to the membrane.
Experiments in Fig. 4 demonstrate that the solubilized brain 72,000 M, fragment binds to erythrocyte spectrin at the same site as erythrocyte ankyrin 72,000 MI fragment. The brain fragment was adsorbed to an erythrocyte spectrin aftinity column and eluted onto a second column of hydroxylapatite (see below) resulting in substantial purification (Fig. 4).
The association of the fragment with the affinity colcmn most likely involved a direct association between the fragment and spectrin since the fragment transferred from SDS gels to nitrocellulose bound Iz5I-labeled erythrocyte spectrin under immunoblot conditions (Fig. 4). Several types of controls Binding of brain 72,000 M , ankyrin fragment to an erythrocyte spectrin affinity column, and binding of 1251-labeled erythrocyte spectrin to brain 72,000 M. fragment transferred to nitrocellulose from an SDS-polyacrylamide gel. Brain ankyrin 72,000 M, fragment was extracted from 200 g of membranes, and precipitated with ammonium sulfate as in Fig. 5. The fragment was partially purified by DEAEchromatography and eluted with a gradient of 10-300 mM NaCl at pH 7.5. The peak containing 72,000 M, fragment was identified by immunoblots (see "Experimental Procedures"), and applied to an erythrocyte spectrin-affinity column (12 ml gel; 1.5 mg of pig erythrocyte spectrin/ml of Sepharose 4B (see "Experimental Procedures") equilibrated with 0.1 M NaCl, 10 mM sodium phosphate, 1 mM sodium azide, 0.5 mM DTT. Fractions from the affinity column were monitored by Am and by immunoblotting to detect polypeptides cross-reacting with erythrocyte ankyrin. The affinity column was eluted with 400 ml of 0.5 M sodium bromide, 10 mM sodium phosphate, 1 mM sodium azide, 0.2 mM DTT directly onto a 2-ml hydroxylapatite column which adsorbed brain demonstrate that binding of spectrin to nitrocellulose-adsorbed fragment was specific. A polypeptide of 72,000 M, that was not adsorbed to the spectrin affinity column did not bind spectrin in the blot assay (Fig. 4). The 72,000 M, polypeptide eluted from the affinity column was much more active on blots than the starting material. Binding was displaced by excess unlabeled spectrin, indicating a limited number of sites. Finally, binding of spectrin was blocked by 100 nM erythrocyte 72,000 M, ankyrin fragment, which suggests that the brain and erythrocyte fragments bind to the same site on spectrin.
Purification of Brain Ankyrin 72,000 M, Fragment-The first step in purification of the fragment was to selectively extract this polypeptide from the membrane. Membranes were extracted prior to digestion with 0.5 M NaCl, which removed about 20% of the protein and little ankyrin (Fig. 3). The 72,000 M, fragment remained associated with membrane after the digestion with chymotrypsin, which permitted removal of protease and protein solubilized during proteolysis. The frag-ment then was extracted from digested membranes with 0.5 M NaCl, thus selecting for polypeptides released in high salt only after proteolysis. The extract of digested membranes contained about 80-9096 of the fragment and only 4% of the protein ( Table I).
The solubilized fragment was then purified by a one step procedure involving three different columns connected in series. The advantages of continuous chromatography are that loss of sample is minimized and the time for the procedure is reduced. The digest was applied first to a column of DE53cellulose under conditions where the fragment was not adsorbed. The effluent from the DE53 column ran directly onto an erythrocyte spectrin affinity column which adsorbed the fragment. The fragment was gradually eluted from the affinity column with a large volume of loading buffer and collected directly on a small column of hydroxylapatite. The hydroxylapatite column was then eluted with a gradient of phosphate, and the fragment was obtained in the peak fractions an erythrocyte spectrin affinity column, and hydroxylapatite. Frozen pig brain grey matter (250 g) was homogenized (see "Experimental Procedures") in 1.5 liters of 0.32 M sucrose, 2 mM sodium EGTA, 200 pg/ml of PMSF, pH 7.5. The homogenate was centrifuged at 900 X g for 10 min, and the 900 X g supernatant was centrifuged 30 min at 14,000 rpm (30,000 X g maximum) in a JA-14 rotor. The pellet was washed once in 10 mM sodium phosphate, 1 mM sodium EGTA, 50 pg/ml of PMSF, pH 7.5 (Buffer A), and then extracted 30 min at 4 "C in 0.5 M NaCl dissolved in Buffer A. The membranes were pelleted, washed twice in 0.5 M NaCI, and resuspended in 1.5 liters of Buffer A without PMSF. The membranes were digested 1 h at 4 "C with 20 pg/ml of a-chymotrypsin, and the enyme was quenched by addition of DFP (0.015% final concentration) and PMSF (200 pg/ml final concentration). The digested membranes were pelleted, resuspended in 750 ml of 0.5 M NaCl in Buffer A plus 0.5 mM DTT, and incubated 30 min at 4 "C. The suspension was centrifuged 1 h at 40,000 X g in a 45 Ti rotor and the supernatant containing the 72,000 M, fragment was collected. Protein was precipitated by addition of solid ammonium sulfate to give 60% saturation, and the precipitate was resuspended in 40 ml of column buffer (0.1 M NaCI, 10 mM sodium phosphate, 1 mM NaN3, 0.5 mM DTT). Following dialysis against column buffer, the suspension was centrifuged 50,000 rpm for 45 min in a 60 Ti rotor to remove aggregated protein. The supernatant was applied (3 ml/h) to a 20-ml DE53 column (equilibrated with 10 mM sodium phosphate, pH 7.5) connected directly to a 25-ml column of erythrocyte spectrin-agarose (2 mg of spectrin/ml). The 72,000 M, fragment did not bind to DEAE in 0.1 M salt, and passed to the affinity column where it was retarded. The combined DE53 and affinity columns were washed with column buffer until the A m was less than 0.05 (left panel; arrow marks end of wash). Then the DE53 column was removed and the affinity column was connected below to a 2-ml hydroxylapatite column. The 72,000 M, fragment was eluted from the affinity column with 500 ml of column buffer and the eluted fragment then adsorbed to the hydroxylapatite column. The hydroxylapatite column was then eluted (12 ml/h) at 24 "C with gradient of sodium phosphate (40 ml, 10-400 mM, pH 7.3) with 1 mM DTT, 1 mM NaN3. Fractions (1 ml) were collected and placed on ice, and protein was monitored by the method of Bradford (center p a n e l ) . Fractions were pooled and analyzed for 72,000 M, fragment by immunoblot. The peak of fragment (pool C) was dialyzed against 50 mM NaCI, 10 mM sodium phosphate, 1 mM NaN3, 0.5 mM DTT. Higher affinity spectrin binding proteins were eluted from the affinity column with 0.5 M NaBr onto a second hydroxylapatite column, and subsequently eluted from the hydroxylapatite column with 0.4 M sodium phosphate, pH 7.3. Fractions from the purification were analyzed on a SDS-polyacrylamide gel stained with Coomassie blue: erythrocyte 72,000 M, fragment (lane I ) , material irreversibly precipitated by ammonium sulfate (lane 2), starting sample for DE53/affinity columns (lane 3), protein adsorbed to DE53 column and eluted with 0.5 M NaCl ( l a n e 4), breakthrough fractions from DE53/affinity columns ( l a n e 5), other spectrin-binding protein eluted from the affinity column with 0.5 M NaBr (lane 6), and purified 72,000 M. fragment from the first hydroxylapatite column (lane 7). Proteins on a parallel gel were electrophoretically transferred to nitrocellulose and polypeptides cross-reacting with erythrocyte ankyrin were identified by immunoblot (see "Experimental Procedures" and Fig. 1) (right two panels).

Amount of protein in samples at various stages of purification of Counts of ferritin-labeled brain spectrin molecules from electron brain ankyrin fragment micrographs of brain spectrin plus avidin-ferritin in the presence and
Fraction" Protein absence of biotin-labeled ankyrin fragment from brain and erythrocytes mg From experiment described in Fig. 8.    (Fig. 5). Other spectrin-binding proteins were eluted from the spectrin affinity column with 0.5 M NaBr, and these proteins also could be concentrated on a second column of hydroxylapatite (Fig. 5). The purified fragment was about 80% pure based on densitometry of Coomassie blue-stained SDS gels, and the yield from 250 g of brain tissue was 300 pg (Table 11).
The fragment was purified about 400-fold with a yield of only about 2% based on estimates of 2.8 pg of fragment/mg of membrane protein (see below). Major losses occurred due to the tendency of the fragment to adsorb to surfaces, especially in early stages of purification. A major concern was that the 72,000 M, fragment might be derived from contaminating erythrocytes. This possibility was excluded by the fact that peptide maps of the brain and pig erythrocyte fragments are distinct with very few common peptides (Fig. 6). The peptide maps share some similarities in that the number of peptides are nearly the same, and that the pattern is similar. However, if peptides of each fragment are co-electrophoresed, it is clear that these are not identical peptides.
An initial problem in these studies was that some proteins in the crude digest, including fragment, formed a precipitate when dialyzed against low ionic strength (less than 10 mM NaCl). The precipitate was enriched in a polypeptide of M, = 130,000 cross-reacting with brain spectrin, 72,000 M, ankyrin fragment, and a polypeptide at 43,000 M, that may be actin. The requirement for salt during dialysis prevented conventional application of DEAE-chromatography since the fragment eluted from DE53-cellulose at 50-70 mM NaCl.
Estimate of the Amount of 72,000 M, Fragment in Brain Membranes-The pure '*'I-labeled fragment was used as a ligand in a radioimmunoassay to estimate the amount of 72,000 M, fragment and presumably ankyrin in brain tissue (Fig. 7). Binding of '2'I-labeled brain fragment was measured to antibody against erythrocyte ankyrin, and this binding was displaced in a parallel fashion by membrane protein solubilized in SDS to ensure accessibility to the antibody and by SDS-denatured fragment. The effects of SDS in the assay were minimal, and SDS at equivalent concentrations was present in control samples. Comparison of displacement of binding by fragment and membranes indicates the presence of fragment at about 0.28% of the total protein, or 39 pmol/ mg of membrane protein. Brain spectrin tetramer was estimated in the same membranes to be present at 30 pmol/mg based on densitometry of Coomassie blue-stained gels. A significant difficulty in performing these measurements was that anti-erythrocyte ankyrin Ig bound brain fragment with a relatively low affinity of 16 nM (not shown). Thus, the assay was relatively insensitive and required substantial amounts of brain membrane protein. High concentrations of membrane protein caused nonspecific interference with the assay, and for this reason the displacement curve in Fig. 7 was not extended above 70 pg/ml of membrane protein. More accurate measurements will require high affinity antibody raised against brain fragment. It is likely that the estimate of brain fragment from Fig. 7 is approximately correct since a similar value of 0.2% of the membrane protein as 72,000 M, fragment was obtained previously in a different assay (Bennett, 1979). The measurements in this earlier study were made by comparison of displacement of binding of '251-labeled erythrocyte 72,000 M, fragment to antibody against erythrocyte 72,000 M, fragment by rat erythrocytes and rat brain membranes.
Localization of the Binding Site on Brain Spectrin for Brain and Erythrocyte 72,000 M, Fragments-Binding of erythrocyte and brain ankyrin fragments to brain spectrin was visualized by rotary shadowing of spectrin molecules incubated with biotin-labeled fragments and then avidin-ferritin (Fig.  8). Biotin was coupled to the fragments using biotin-N-hydroxysuccinimide ester, and the reaction was monitored by the amount of fragment sedimented with avidin-ferritin. By this criterion, at least 90% of both fragments were conjugated to biotin. Electron micrographs of rotary-shadowed replicas of brain spectrin tetramer incubated with biotin-labeled brain ankyrin fragment and avidin-ferritin in a 1:l:l molar ratio demonstrated ferritin-labeling of 16% of spectrin molecules at a site in the midregion. This site was 90 f 4 nM from the near end and 110 f 5 nM from the far end (Fig. 8). Samples with erythrocyte ankyrin fragment exhibited labeling of 13% of spectrin molecules in the same midregion site. In addition to labeling at the midregion of spectrin, ferritin also was ob-  . fragment (B), and a mixture of peptides from both erythrocyte and brain fragments (C). Pig erythrocyte 72,000 M, fragment and brain 72,000 M, fragment (5 pg) were denatured in 0.05% (w/v) SDS, radiolabeled with 1 mCi of '"I using chloramine-?' as an oxidant, and electrophoresed on a SDS-polyacrylamide gel. The '251-labeled fragments were localized by staining with Coomassie blue, cut from the gel, and incubated 3 h at 37 "C in 50 mM ammonium acetate, 1 mM NaN3, 50 pg/ml achymotrypsin followed by a 15-h incubation with an additional 50 pg/ml of enzyme. The digest was lyophilized and analyzed by electrophoresis (horizontal dimension) and chromatography (uertical dimension) as described (Elder et al., 1977;Davis and Bennett, 1982). Estimate of amount of ankyrin in brain membranes by displacement of binding of '261-labeled brain 72,000 M. fragment to anti-erythrocyte ankyrin Ig. Pig brain membranes (10 mg/ml) (Fig. 1) and purified brain 72,000 M, fragment (100 pg/ ml) were denatured in 1% (w/v) SDS, 5 mM Na EDTA, 20 mM DTT, 10 mM Tris, 50 pg/ml of PMSF, pH 7.5. Portions of these samples were added to tubes containing, in a final volume of 0.2 ml, 0.9 pg/ ml of either anti-erythrocyte ankyrin Ig or nonimmune Ig, 1 mg/ml of bovine serum albumin, 150 mM NaCI, 10 mM sodium phosphate, 1 mM Na EDTA, 1 mM NaN3, 2 pg/ml of pancreatic trypsin inhibitor, 1% (v/v) Triton X-100, 0.2 pl of Protein A-bearing staphylococci, and '9-labeled brain ankyrin fragment (0.47 pmol, 1.4 X lo6 cpm/ pmol). The incubation was continued for 3 h a t 4 "C and then the samples were diluted with 3 ml of 0.5 M NaBr, 1 M urea, 0.1 M glycine, 1% Triton X-100. The staphylococci with adsorbed immune complexes were pelleted (10 min, 2000 X g) and assayed for '%I. The data are presented as per cent binding in the presence of equivalent amounts of SDS but without other additions, and are corrected for nonspecific binding by subtracting the value obtained with nonimmune Ig at each concentration of protein tested. served in 6% of molecules incubated with brain fragment at the ends of spectrin tetramers, and in 6% of the molecules at other sites (Table 11). Control samples with avidin-ferritin but no biotin-fragments exhibited labeling of 6% of spectrin molecules approximately in the midregion, 5% of molecules at the ends, and 3% at other regions along spectrin molecules.
Ferritin-labeling of the midregion of spectrin incubated with biotin-labeled brain fragment, was specific by several criteria. The site was labeled at a 3-fold lower frequency when fragments were omitted. When labeling at the midregion, ends, and other regions of the spectrin molecules was corrected for nonspecific labeling in the absence of fragments, the midregion site exhibited 10-fold more labeling than at the ends and 3-fold more labeling than all other sites on the molecule combined (Table 11).

DISCUSSION
This report describes identification of membrane-associated polypeptides of M , = 190,000-220,000 in brain, lens, and liver that cross-react with erythrocyte ankyrin. The crossreacting polypeptides in brain, and most likely other tissues as well, are closely related to erythrocyte ankyrin in structure and function. The brain polypeptides are digested by mild proteolysis to domains of M , = 95,000 and 72,000 (Fig. 3) which are the same size as fragments obtained by digestion of erythrocyte ankyrin (Bennett, 1978;Bennett and Stenbuck, 1980). The brain 72,000 M , fragment binds to erythrocyte spectrin at the same site as erythrocyte 72,000 M , fragment ( Fig. 4) and was purified by affinity chromatography on erythrocyte spectrin-agarose (Fig. 5). Purified brain 72,000 M , fragment was distinct from erythrocyte 72,000 M , fragment by peptide maps (Fig. 6), but both fragments bound to brain spectrin tetramer at a site in the midregion about 90 nM from the nearest end (Fig. 8). The ankyrin binding site .,_., " _ . . .  FIG. 8. Localization by electron microscopy of binding sites on brain spectrin for brain and erythrocyte ankyrin 72,000 M. fragments. Biotinyl derivatives of brain and human erythrocyte ankyrin 72,000 M, fragments were prepared by reaction for 1 h at 4 "C of 5 p~ biotin N-hydroxysuccinimide ester with 1 p~ fragment in 25 mM sodium phosphate buffer, pH 8, followed by dialysis against 50 mM NaCI, 10 mM sodium phosphate, 1 mM NaN3, 0.5 mM DTT. Brain spectrin and biotin-labeled fragments were incubated at 5 X lo-' M of each protein for 14 h at 4 "c in 90 mM NaCI, 10 mM sodium phosphate, 1 mM NaN3, 0.5 mM DTT, followed by incubation for 1 h at 4 "C with avidin-ferritin at a final concentration of 5 X

M.
Control samples were run with brain spectrin and avidin-ferritin in the absence of fragments. Samples were diluted to 10"' M of brain spectrin in 0.1 M ammonium formate, 30% (v/v) glycerol, 1 mM NaN3, pH 7, and within 2 min sprayed onto freshly cleaved mica. The mica sheets were dried under vacuum at 24 "C and rotary-shadowed with platinum at an angle of 6" followed by carbon (Fowler and Erickson, 1979;Shotton et al., 1979). Fields are shown of brain spectrin and avidin-ferritin in the presence ( A ) and absence ( B ) of biotin-labeled brain fragment. Selected molecules are shown below at higher magnification. See Table I1 for quantitation of the per cent of spectrin molecules labeled and location of the label in the middle region. on erythrocyte spectrin has been localized to a similar region on erythrocyte spectrin (Tyler et al., 1979). The amount of brain 72,000 M , fragment was estimated by radioimmunoassay to be 0.28% of the membrane protein or 39 pmol/mg, which is about the same as the number of brain spectrin tetramers (30 pmol/mg of membrane protein).
These experiments provide strong evidence for the widespread presence of proteins closely related to ankyrin in cell membranes. The criteria that seem appropriate at this point for a nonerythroid ankyrin are as follows: 1) cross-reactivity with erythrocyte ankyrin; 2) Mr of -200,000; 3) association with membranes; 4) protease-resistant domains of M , = 95,000 and 72,000; 5) binding of the 72,000 Mr domain to spectrin; 6) localization of the binding site on spectrin to the midregion of spectrin tetramers; 7) presence in approximately equivalent amounts as spectrin tetramer. Microtubule-associated proteins of M , = 370,000 have been identified in brain that cross-react with ankyrin (Bennett and Davis, 1981;Bennett et d., 1982b). The microtubule-associated proteins appear to not be as closely related to ankyrin in terms of antigenic sites or in other structural and functional features. The microtubule-associated proteins should be viewed as perhaps related to ankyrin in terms of a common evolutionary origin, but presently as a distinct group of proteins. The membrane-associated ankyrin proteins are similar enough to erythrocyte ankyrin to be considered members of a common family of proteins. It is likely that additional isoforms of ankyrin will be discovered in different tissues or during differentiation of the same cell, by analogy with various forms of brain and muscle spectrin (Nelson and Lazarides, 1983). Nomenclature may become complicated, but at the present time these proteins can be referred to based on the tissue or cell of origin, e.g. hepatocyte ankyrin, brain ankyrin, etc.
Brain spectrin and ankyrin are not identical to their analogues in erythrocytes. One difference already evident is that the affinity of ankyrin-spectrin binding is lower for brain proteins, with a Kd of 0.5-1 ~L M rather than a Kd of 0.02-0.05 FM observed for erythrocyte proteins3 (not shown). The average concentration of brain ankyrin and spectrin is about 1 ~L M in brain tissue (estimated on the basis of 30 pmol/mg of membrane protein, 30 mg of membrane protein/g of tissue), and local concentrations of spectrin and ankyrin on membrane surfaces are most likely 10-20-fold higher. Thus, the concentrations of spectrin and ankyrin on the membrane are sufficiently high for a major portion of these proteins to exist as a spectrin-ankyrin complex. However, the relatively low affinity compared to erythrocyte membranes suggests that the spectrin-ankyrin associations may be more dynamic than in erythrocytes and may be subject to regulation. Another important difference between membranes from brain and mammalian erythrocytes is that brain membranes contain tubulin (Bhattacharyya and Wolff, 1975) and most likely also have binding sites for tubulin (Bernier-Valentin et al., 1983). Erythrocyte ankyrin binds to microtubules assembled from pure brain tubulin (Bennett and Davis, 1981), and it is possible that brain ankyrin also has a binding site for microtubules and is complexed with tubulin on the membrane.
Ankyrin is the major membrane attachment site for spectrin in erythrocytes, but in brain and other tissues, spectrin may have additional membrane linkages. Preliminary measurements of association of '251-labeled spectrin to brain membranes indicated saturable, high affinity binding that persisted after extraction of brain 72,000 M, fragment and was not displaced by fragment (unpublished data). Furthermore, spectrin-binding polypeptides unrelated to the 72,000 M , fragment were recovered from the spectrin affinity column during Binding of fragment and brain spectrin was measured with brain fragment labeled with 'Z51-Bolton Hunter reagent; spectrin-bound fragment was separated from free fragment by immunoprecipitation with antibody against spectrin as described (Bennett and Stenbuck, 1980). purification of the fragment (Fig. 5). In fact, there presently is no direct evidence that brain ankyrin links spectrin to the membrane, although all of the available data is consistent with such an association. Elucidation of the membrane associations of brain spectrin clearly will be more complex than such studies in erythrocytes, and will be the subject of future work.
An important extension of the present studies will be to determine the protein(s) that links brain ankyrin to the membrane. If the analogy with erythrocyte membranes continues to be relevant, then the ankyrin-binding protein will be an integral membrane protein and also contain an ion channel. This putative protein may also associate with other integral proteins, as is the case with the erythrocyte anion channel and other membrane proteins in erythrocytes. The existence of an integral membrane protein or family of such proteins capable of lateral associations in the membrane and of binding to cytoskeletal proteins could explain how membrane proteins are restricted in their motion and localized at specialized regions on the cell surface.