The Aldolase-binding Site of the Human Erythrocyte Membrane Is at the N H 2 Terminus of Band 3*

1,6-bisphosphate aldolase cytoplasmic

' J. D. Jenkins and T. L. Steck, unpublished data. bind rapidly and reversibly to the membrane in the intact cell.
The binding of glycolytic enzymes to membranes and other structural elements of the cell has been widely observed in vitro (20-22), but the specificity and physiologic significance of the phenomenon is ill defined. In the present study, a site of association for fructose 1,6-bisphosphate aldolase (EC 4.1.2.13) was found in the NHz-terminal region of band 3. The highly acidic amino acid composition of this region upp ports the hypothesis that band 3 associates electrostatically with glycolytic enzymes. A preliminary account of this research has been published (23).

EXPERIMENTAL PROCEDURES
Materials-Outdated units of blood from normal human donors were generously provided by the Blood Bank of the University of Chicago. From Sigma we obtained rabbit muscle aldolase (Grade I), the accessory components for its assay (6,7), and 2-mercaptoethanol. We obtained L-1-tosylamido-2-phenylethyl chloromethyl ketonetrypsin from Worthington and cyanogen bromide from Eastman. SP-Sephadex C-25, Sephadex G-50, and Sephadex G-10 were products of Pharmacia. Bio-gel P-10 and P-30 were obtained from Bio-Rad. Citraconic anhydride was obtained from Pierce. Chemicals were reagent grade or better from Fisher, Mallinckrodt, or Baker. 2-Nitro-5thiocyanobenzoate was synthesized as described (24).
Preparations-All procedures were at 0-5 "C, and all centrifugations were performed in a Sorvall SS-34 rotor at 15,000 rpm unless otherwise indicated. Ghosts were prepared as described (25), except as whole unit preparations (26). The 23,000-dalton fragment was generated by the S-cyanylation of alkali-stripped ghosts, precipitated by cold isopropanol, and extracted into water (26,27). The extract was concentrated by ultrafiltration (Amicon PM-10) to about 10 ml and subjected to preparative gel electrophoresis (28). Fractions containing the 23,000-dalton fragment were pooled, concentrated, precipitated with isopropanol, and the protein was freeze-dried. Typically, 15-20 mg of the purified 23,000-dalton fragment were isolated per unit of blood. Acetylation (29)"Samples were dispersed in 100 p1 of water, and an equal volume of a saturated solution of sodium acetate was added. The sample was cooled and stirred at 0 "C while 5 aliquots of acetic anhydride were added in the course of an hour to equal the weight of the peptide. The reaction mixture was desalted by gel filtration on Sephadex G-IO.
Peptides-Typically, the 23K2 fragment obtained from one unit of blood was dissolved in 5 ml of 0.1 M Tris-acetate buffer. pH 8.6. Solid guanidine hydrochloride was added to make a 6 M solution. Three hundred pl of citraconic anhydride were added in 20-pl aliquots over the course of an hour; the pH was maintained between 8.2 and 8.6 by the addition of 1 M sodium hydroxide (30). The solution was dialyzed The abbreviation used is: 23K, the 23,000-dalton fragment derived from the NHn-terminal region of band 3 by S-cyanylation. In naming peptide fragments, the following symbols are used: T, tryptic; C, citraconyl; CN, cyanogen bromide; A, acid cleaved. 11203 against 0.2 M ammonium bicarbonate. The citraconylated protein was incubated overnight at 37 "C in the same buffer with trypsin (1% of its weight). The digest was chromatographed on a column of Bio-Gel P-30 (3 x 42 cm) in 30% formic acid. The peptides were identified by absorbance at 280 nm and by ninhydrin reaction after alkaline hydrolysis. The largest tryptic peptide, TCl, emerged from the column at 85-105 ml volume in pure form, as determined by sodium dodecyl sulfate-urea and basic urea gel electrophoresis.
The TCl tryptic peptide was acid cleaved in 70% formic acid at 37 "C for 3 days. The cleavage products were resolved by chromatography on a Bio-Gel P-10 column (1.5 X 45 cm) in 30% formic acid and then on an SP-Sephadex column (1.5 X 10 cm) in 10% acetic acid.
Cyanogen bromide digestion was carried out overnight in 70% formic acid for 14-16 h at 37 "C. The CNBr reaction mixture was diluted with water and freeze-dried. The products were resolved by gel filtration on a Bio-Gel P-10 column (1.5 X 45 cm) equilibrated with 0.05 M ammonium bicarbonate or a Sephadex G-50 column (3 X 50 cm) in 10% acetic acid.
Amino acid analysis was performed following 20-h hydrolysis at 115 "C with 5.7 M hydrochloric acid containing 0.05% 2-mercaptoethanol and 1 mM phenol on a Beckman Model 121 M amino acid analyzer coupled to a System AA computing integrator (26).
Enzyme Assays and Analysis of Inhibitor Binding-Maximal aldolase activity (Le., in the presence of sufficient fructose 1,6-bisphosphate to overcome the inhibitory effects of any band 3 fragments present) was measured using minor modifications (6,7) of the method of Wu and Racker (31), as described for individual experiments. Inhibition of aldolase activity by membrane peptides was measured under conditions which favored enzyme-band 3 association (50 (LM substrate in 10 mM imidazole acetate, pH 7.0, etc.), as described in Ref. 7 and Figs. 1-3 below. Ki values and the percentage of maximal inhibition were determined by fitting the data to a simple partial competitive inhibition model, using a measured K , of 1.55 p~ to represent the dissociation constant of the substrate-aldolase complex and an estimated 1.2 X lo6 band 3-aldolase binding sites per ghost (1, 3, 7, 8). The best fits were obtained when the number of band 3 molecules bound per aldolase tetramer was taken as unity, the stoichiometry found for aldolase binding to band 3 both in intact ghosts 1.5-ml ghosts loaded with aldolase were incubated in 28.5 ml of 0.075% Triton X-100 in 10 m~ imidazole-HC1 (pH 7) for 20 min on ice. Following a 30-min centrifugation the supernatant fluid was collected; a portion was adjusted to 0.3 M NaCl by the addition of the solid. After a 1-h incubation on ice, aldolase activity was measured as a function of substrate concentration in the presence (0) and absence (0) of 0.3 M NaCl. In this experiment, 0.3 M NaCl shifted the apparent K,,, from 650 p~ to 50 p~. That this effect is attributed to the dissociation of the enzyme-band 3 complex is shown by the fact that, in the absence of membranes, 0.3 M NaCl increased the K , from 2 to 50 FM. Thus, the inhibitory potency of the membranes is even more profound than suggested in this experiment.
FDP, fructose 1,6-bisphosphate. Various soluble fragments of band 3 were prepared as described in the text and mixed with rabbit muscle aldolase (11 pmol) in 10 mM imidazole acetate, pH 7.0, containing 0.004% Triton X-100 (final). Aldolase activity was then assayed with 50 p~ fructose 1,6-bisphosphate. The curves were generated from the values given in Table I (a representative experiment). A, .

A-A,
CN1. An alignment of the peptides is shown in Fig. 4. and in Triton X-100 solution (6, 7). The standard deviation of the fractional inhibition observed ranged from 8% for band 3 to 22% for TClAl in Fig. 3; the Ki values obtained are probably reliable only to within a factor of two (Table I). Enzyme Displacement Assay-Ghosts were freed of endogenous glycolytic enzymes by washing in 0.15 M ammonium bicarbonate (6-8) followed by washing in 10 m~ imidazole acetate (pH 7.0). Rabbit muscle aldolase freshly dialyzed in the same buffer was added to the ghosts so that it exceeded their band 3 content by 20% (mole/mol). Band 3 was estimated from coincidence-corrected Coulter counts of ghosts, assuming 1.2 X lo6 band 3 binding sites/ghost (3,6,7). Aldolase concentration was calculated by assuming an absorbance at 280 nm of 9.24 for a 1% solution and M, = 160,000 (32). After a 30-60 min e, 0.1 -ml of 72-mM fructose 1,6-bisphosphate. equilibration, the aldolase-loaded ghosts were washed twice in 10 mM TABLE I Aldolase binding and inhibition by fragments of band 3 Left, ghosts or band 3 fragments were tested for their potency and extent of aldolase inhibition as described uhder "Experimental Procedures." The K, and plateau values listed here were those used to generate the curves represented in Fig. 3. Some of the plateau values were derived from very high inhibitor levels omitted from Fig. 3. Right, an independent estimate of inhibitor binding was obtained by the competitive displacement of aldolase from ghosts (see under "Experimental Procedures"). An alignment of the fragments is shown in Fig. 4.

Aldolase inhibition assay Displacement assav Peptide
Ghost band 3 23,000 dalton TC1 TClTl TClAl TClA2 " The affinity of the band 3 in ghosts for aldolase was determined to be Kd = 1.8 X 10"' M, against which the displacement assays were compared.
imidazole acetate (pH 7.0). Into a 400-pl conical polyethylene microcentrifuge tube was mixed (in a final volume of 100 p1 of 10 mM imidazole acetate at pH 7.0) aldolase-loaded ghosts containing 70 pm of band 3, 1 mg/ml of bovine serum albumin, 1 m M dithiothreitol, 0.1 m M EDTA, and up to 4 nmol of inhibitor (quantitated by amino acid analysis). After a 20-25 h incubation on ice, the ghosts were sedimented at 5600 rpm for 35 min in a swinging bucket HS-4 rotor in a Sorvall RC3 centrifuge. Ten to 25 p1 of supernatant fluid were assayed for maximal aldolase activity in 1 ml of 10 mM imidazole acetate (pH 7.0) containing 0.1 m NADH, 20 pg/ml of auxiliary enzymes (6, 7, 31), and 5 m M fructose 1,6-bisphosphate (sufficient to reverse inhibitor binding completely). Control runs lacked inhibitor (to determine background) or centrifugation (to determine input). From the amount of aldolase released, the measured concentrations of band 3, aldolase, and inhibitor present, and the Kd for the aldolase-ghost band 3 complex (independently determined by a Scatchard analysis to be 1.8 X 10"' M), Kd values were calculated from a simple competitive equilibrium-binding model. The calculation assumed a stoichiometry of one band 3 or inhibitor per aldolase tetramer. The standard deviation of the relative Kd values determined by the displacement assay ranged from 9% for fragment TC1 to 19% for TClA1; the Kd values obtained are probably reliable only to within a factor of two (Table I).

RESULTS
The Assay ofdldolase Inhibition-Membranes from erythrocytes (7) and other tissues (22) reversibly inhibit the catalytic activity of mammalian aldolases. Band 3 provides the exclusive binding site for this enzyme in red cell membranes (6, 7). The inhibition of aldolase activity thus affords an assay for the identification of its binding site among fragments of band 3.
Unsealed ghosts, Triton X-100 extracts of ghosts, and the cytoplasmic pole of band 3 all inhibited aldolase activity immediately upon addition to the assay cuvette (Ref. 33 and Fig. 1). This effect was rapidly reversed by conditions which elute the enzyme from the membrane, millimolar levels of substrate or high ionic strength (Fig. 1). The form of the inhibition appeared superficially to be competitive, in that a mutual antagonism between substrate and ghosts, reversed at high ionic strength, was manifest (Fig. 2). However, we shall demonstrate below that a more complex mechanism is likely.
The dependence of aldolase inhibition on band 3 concentration in preparations of ghosts freshly dissolved in Triton X-100 is shown in the top curue of Fig. 3A ( O " --W . The data were well fitted to a hyperbola by assuming a simple binding isotherm, a 1:l stoichiometry (6,7), and the parameters shown in Table I (namely, a Ki of 1 X 10"' M and 9 0 % inhibition at saturation).
An Aldolase-Inhibitory 23,000-dalton Fragment from Band 3-A water-soluble 23K fragment was cleaved by scyanylation from the NH2-terminal end of band 3 and purified (26, 27). Its inhibitory effect on aldolase activity varied somewhat among several preparations; representative results are summarked in Fig. 3A (o"--o) and in Table I. In this representative experiment, the avidity of this fragment for aldolase was approximately one-fih that of band 3; like band 3, aldolase activity was never completely inhibited when saturated with the fragment.
Tryptic Subfragments from 23,000-dalton Fragment-The 23K fragment was citraconylated, digested with trypsin, and the cleavage products resolved by gel fitration on Bio-Gel P-30. The largest subfragment, TC1, was recovered from column chromatography in pure form (see under "Experimental Procedures"). The electrophoretic mobility of the peptide in a basic urea gel system was found to be nearly that of the tracking dye, bromphenol blue. The acidic nature of TC1 was c o n f i e d by its amino acid composition; 34 of its 75 residues were Asx or Glx (Table 11). Furthermore, Edman degradation revealed that 19 of 24 Glx residues were Glu and 8 of 10 Asx were Asp. The affinity of T C l for aldolase was approximately 15% that of band 3, approximately equal to that of the 23K fragment (Table I). However, unlike parent band 3 and the 23K species, the TC1 fragment only inhibited aldolase by 77% at saturation (i.e. using 670 pmol, lo3 times the Kt value; see None of the remaining tryptic fragments isolated by P-30 chromatography inhibited aldolase significantly. The TC1 fragment was decitraconylated by mild acid hydrolysis and redigested with trypsin. A 56-residue subfragment, TClT1, was isolated. It was relatively more acidic than TC1 in that two basic residues were missing (Table 11). It was just as potent an inhibitor of aldolase as TC1 itself (Table I).
Two Acid-cleavage Fragments-The TC1 fragment was also acid cleaved in 70% formic acid at 37 "C for 3 days. Two products were recovered, TClAl (23 residues) and TClA2 (52 residues), which summed up to the composition of TC1 (Table  11). The more acidic TClAl fragment was four times more potent than TClA2.
However, at saturation, only 37% of aldolase activity was inhibited by each of these fragments ( Fig. 3B and Table I).
Cyanogen Bromide Fragments-The T C l T l fragment was further cleaved with cyanogen bromide, as described under "Experimental Procedures." Three peptides were isolated (TClTlCN1, 2 and 3). One of the fragments, TClTlCN2, showed weak aldolase inhibitory activity (Table I). However, its aldolase inhibitory activity increased significantly following acetylation ( Fig. 3A and Table I). Its potency was less than the parent fragment, TC1, but at saturation, it inhibited aldolase activity by 85% (Table I). The composition of TClTlCN2 is given in Table 11.
The 23K fragment was also digested with cyanogen bromide. The largest fragment generated, CN1, contained -125 residues. Its amino acid composition (Table 11) is very similar to the composition of a 13,500-dalton CNBr peptide isolated by Drickamer (34). CN1 represents the COOH-terminal twothirds of the 23K fragment. As shown in Fig. 3B (A-A) and Table I, it did not interact significantly with aldolase.
An Aldolase Displacement Assay-An independent assess- Values are in residues per mol of protein (rounded to nearest integer). Met was determined as homoserine in the CNBr peptides. Trp and Cys were determined only in the 23K fragment. ment of the binding of the various band 3 fragments to aldolase was provided by their ability to displace the enzyme from ghosts. Unsealed ghosts were freed of their endogenous glycolytic enzymes and loaded to near saturation with freshly dialyzed rabbit muscle aldolase (6). Varied amounts of watersoluble fragments of band 3 were then equilibrated with the ghosts on ice overnight. Alternatively, the band 3 fragments were first added to depleted ghosts followed by a subsaturating dose of rabbit muscle aldolase. A critical test of the method was that in both procedures, a similar plateau of aldolase displacement was reached for a given equilibrium mixture.
The data were treated as in other competitive displacement assays except that we solved for the potency of fragments relative to that of ghost band 3 rather than the amount of the competitor. Included in Table I is a comparison of the binding potency of various band 3 fragments determined by the aldolase inhibition and displacement techniques described. The validity of both methods is a f f i e d by the close agreement between them for the several peptide species; the average difference between pairs of values was 18%.

DISCUSSION
This study has delineated two features of the interaction of aldolase with the membrane; the inhibition of its catalytic activity and its locus of binding to band 3.
Previous studies on the binding of aldolase to intact or dissolved ghosts revealed a stoichiometry of one enzyme molecule per band 3 polypeptide (6, 7). This result signifies that, by binding to one of its protomers, a band 3 polypeptide can inhibit an entire aldolase tetramer. Since aldolase contains four identical, catalytically active subunits per tetramer (35), its nearly complete inhibition upon binding to a single band 3 polypeptide is incompatible with a simple competitive mechanism. The observed stoichiometry, furthermore, argues against a mechanism of inhibition involving dissociation of the tetramer into inactive subunits, since this would require multiple band 3 ligands per tetramer. We also discount the possibility that a single 40,000-dalton cytoplasmic pole of band 3 (3) could bind to all four 40,000-dalton subunits of aldolase, which are arranged with 2-fold symmetry around three mutually perpendicular axes (36).
Simple competition is also incompatible with the fact that the presence of up to 1200 pmol of band 3 never completely inhibited 11 pmol of aldolase. This effect is even more dramatic in the case of certain small peptide fragments where inhibition at saturation was less than 40% (Table I and Fig.  3B). Thus, although band 3 and the substrate displace one another from aldolase (Fig. 2) and may both bind to the same (active) site, the data suggest the formation of partially catalytically active ternary complexes of inhibitor, enzyme, and substrate. That such ternary complexes exist was previously inferred from the observation that the residual aldolase activity found in the presence of saturating levels of ghosts sedimented with the membranes even while the aldolase reaction was in progress (7). (This fiding rules out, furthermore, the possibility that some aldolase molecules resist inhibition by failing to bind band 3.) That the saturation of the aldolase with band 3 or its fragments caused less than full inhibition also signifies that the binding of multiple band 3 polypeptides to a single aldolase tetramer is curtailed; otherwise the inhibition would approach completion as all four subunits became occupied. (It seems unlikely that band 3 binding to the first of four identical aldolase subunits is inhibitory while its binding to the others is not.) The binding of a single band 3 peptide to one subunit thus appears to reduce the affinity of the remaining protomers for band 3, independent of their catalytic activity. This behavior suggests a high degree of negative cooperativity among aldolase subunits. Mammalian aldolases have not been found to exhibit cooperative kinetics (35). However, Penhoet and Rutter (37) showed that specific noncross-reacting antibodies directed against homomeric aldolases A4, B4, or C4 were capable of inhibiting entirely the activity of heteromeric hybrids containing only one antibody-sensitive protomer; this finding suggests an interaction among subunits similar to that observed here. The most parsimonious hypothesis consistent with these findings is that a band 3 peptide can bind to any of four identical sites on the aldolase tetramer and thereby reduce both the catalytic activity and the affinity of the remaining three subunits for the inhibitor.
We have provisionally ordered the fragments capable of inhibiting aldolase activity as shown in Fig. 4. It is evident from a comparison of this figure with Table I that (3); its putative sequence is Ac-Met-Glu-Glu (38). The 23K fragment and subfragments TC1, TClTI, and TClAl are also unreactive to Edman degradation and/or yield a blocked single homoserine residue fohwing cyanogen bromide digestion, placing them at the amino terminus of band 3. According to carboxypeptidase digestion, CN1 has the COOH-terminal sequence of the 23K fragment, while TClT3 and TClA2 share the COOHterminal sequence of TCI. TCITlCN3 has the same COOH-terminal sequence as TClT1, The junction of TCITICN2 and TClTlCN3 is contained in the NH2-terminal sequence of TClA2. The acid-cleaved Asp-Pro bond between TClAl and TClA2 appears in the TClTlCN2 peptide. The overlap between TCITlCNl and TClTlCN2 is contained in TClAl.
could have resulted from its exposure to dodecyl sulfate, isopropanol precipitation, or another harsh treatment during its preparation. The potency of TClAl and the other fragments containing the NHZ-terminal portion of band 3 was equal to that of the 23K fragment itself. None of several fragments originating beyond residue 31 was inhibitory.
The most notable feature of the NHz-terminal region of band 3 is its highly acidic composition. There are 5 Asp and 11 Glu among the first 31 residues. Furthermore, this region is devoid of basic residues and has a blocked a-amino terminus ( Table 1 1 and Refs,3 and 38). It is likely that the aldolase inhibitory capacity of band 3 is dependent on the strongly anionic character of this region. Even the presence of free aamino groups, introduced by peptide bond cleavage, may reduce the interaction of these fragments with aldolase, in that N-acetylation increased the potency of the TClTlCN2 peptide (Table I).
Aldolase-binding activity apparently does not require conservation of native conformation, since the variety of denaturing treatments used in the preparation of the peptides did not abolish their activity { e g . dodecyl sulfate, citraconylation, concentrated acetic and formic acids). The binding site of the enzyme may well induce the appropriate conformation in the peptide. The polyanionic character of the NH2-tenninal 30 residues suggests that specific folding may not occur, particularly at low ionic strength, without charge neutralization within the binding site on the enzyme.
The binding site for band 3 on aldolase is likely to be its active center. Band 3 is specifically displaced by fructose 1,6bisphosphate (Figs. 1 and 2) and by competitive inhibitors of the enzyme, such as ATP and 2,3-diphosphoglycerate ( 7 ) . The inhibitory peptides are strongly anionic while the catalytic site for fructose 1,6-bisphosphate is strongly cationic (35); the binding reaction is demonstrably electrostatic in nature (Refs. 6 and 7 and Fig. 1). An analogous case can be made for glyceraldehyde-3-P dehydrogenase which also is electrostatically bound to an anionic site within the cytoplasmic pole of band 3 (8, 13) and displaces aldolase from the membrane (33,39). Pyridine nucleotides rather than fructose lB-bisphosphate reverse the association between band 3 and glyceral-dehyde-3-P dehydrogenase (8, 12). We thus view the NH2terminal portion of band 3 as a flexible polyanionic finger on an arm projecting beyond the submembrane reticulum into the cytoplasmic space, a plausible disposition of the membrane-binding site for soluble glycolytic enzymes.
Two forms of negative cooperativity were seen in the interaction of band 3 with aldolase, an enzyme generally considered to lack interprotomer interactions. These were manifested in the inhibition of binding of multiple band 3 peptides and in the inhibition of the catalytic activity of the unliganded enzyme subunits, These effects speak against a simple or nonspecitk interaction. It is conceivable that it would be highly detrimental for band 3 molecules to aggregate in the membrane. The observed negative cooperativity could serve to suppress the cross-linking of multivalent band 3 particles (presumably dimers (3, 8, 26, 40) or tetramers (41,42)) by tetravalent enzymes. The physiologic significance of the inhibition of catalytic activity upon membrane binding, however, remains obscure.