Chloride Binding to the Anion Transport Binding Sites of Band 3 A 35c1 NMR STUDY*

Band 3 is an integral membrane protein that ex- changes anions across the red cell membrane. Due to the abundance and the high turnover rate of the band 3 transport unit, the band 3 system is the most heavily used ion-transport system in a typical vertebrate or-ganism. Here we show that 36Cl NMR enables direct and specific observation of substrate C1- binding to band 3 transport sites, which are identified by a variety of criteria: (a) the sites are inhibited by 4,4’-dinitrostilbene-Z,2’-dlsulfonate, which is known to in- hibit competitively Cl- binding to band 3 transport sites; (b) the sites have affinities for 4,4‘-dinitrostil- bene-2,2’-disulfonate and C1- that are quantitatively similar to the known affinities of band 3 transport sites for these anions; and (c) the sites have relative affinities for C1-, HCO,, F-, and I- that are quantitatively similar to the known relative affinities of band 3 transport sites for these anions. The 36Cl NMR assay also reveals a class of low affinity C1- binding sites (KO >> 0.5 M) that are not affected by 4,4’-dinitrostilbene- 2,2’-disulfonate. These low affinity sites may be re-sponsible for the inhibition of band 3 catalyzed anion exchange that has been previously observed at high [Cl-J. In the following paper the 36Cl NMR assay is used to resolve the band 3 transport sites on opposite sides


M) that are not affected by 4,4'
-dinitrostilbene-2,2'-disulfonate. These low affinity sites may be responsible for the inhibition of band 3 catalyzed anion exchange that has been previously observed at high [Cl-J. In the following paper the 36Cl NMR assay is used to resolve the band 3 transport sites on opposite sides of the membrane, thereby enabling direct observation of the transmembrane recruitment of transport sites.
Band 3 is an integral membrane protein in human erythrocyte membranes. The protein consists of a single polypeptide chain (M, = 95,000) and although the catalytic unit is thought to be the monomer (l), the protein exists in the membrane as a dimer (2). Band 3 has at least two functions, one structural and the other physiological. The cytoplasmic portion of this protein contains the site to which the red cell cytoskeleton binds and thereby is anchored to the membrane (3,4). Completely unrelated to this structural function is the role of band 3 in the respiratory system, where it facilitates the transport of CO, by the bloodstream. As the central component of the Hamburger (or chloride) shift, band 3 exchanges HCO,, which is produced from C 0 2 and H20 by ll To whom correspondence should be addressed. carbonic anhydrase inside the cell, for C1-on the other side of the membrane (5)(6)(7). This exchange process allows the serum to carry the bulk of the dissolved CO, in the form of HCO,. The physiological importance of this exchange process is illustrated by the fact that the band 3 system is the most heavily used ion-transport system in a typical vertebrate animal such as man ("Appendix I").' We believe that an important current goal of membrane biochemistry should be to understand, in molecular terms, the anion transport event which occurs within this relatively simple and easily obtainable band 3 protein.
The mechanism of band 3-catalyzed anion exchange has been extensively studied in kinetic experiments; for a review, see Ref. 1. These kinetic studies have stimulated the development of a variety of models (1,(8)(9)(10)(11)(12)(13)(14)(15) for the exchange process, all of which postulate the existence of one or more transport sites that bind substrate anion during the transport event. The existence of transport sites is suggested by the saturation kinetics that are observed at high concentrations of substrate anion (1); however, saturation kinetics are also exhibited by ion channels that require single file passage of ions (16). Thus, we have attempted to observe directly transport sites using an assay for substrate (chloride) binding to band 3.
Chloride binding to protein binding sites has been studied in a large number of water-soluble protein systems using 35Cl NMR, and the theory and practice of this approach have been the subject of extensive reviews by Lindman and Forsen (17) and Forsen and Lindman (18). Rothstein and his co-workers (19) first showed that 35Cl NMR can be used to study the binding of chloride to band 3; we have improved upon this technique and extended its applications in the two papers presented here. In the first paper we use the 35Cl NMR chloride binding assay to observe two classes of chloride binding sites on leaky red cell membranes. One class is composed of low affinity sites that may include nonspecific chloride binding sites, while the other class is composed of high affinity chloride binding sites. We are able to identify the high affinity sites as band 3 transport sites by studying their affinity for a variety of anions: chloride, fluoride, iodide, and bicarbonate, as well as the inhibitor of anion exchange DNDS.' Surprisingly, we see no evidence of an inhibitory chloride binding site termed the modified site (20) which has Chloride Binding to Band 3 Transport Sites 6473 been thought to be present on band 3. In the second paper (21) we (a) resolve the transport sites into two populations on opposite membrane surfaces, (b) study the transmembrane recruitment of band 3 transport sites, and (c) discuss the implications of these studies for the mechanism of band 3 catalyzed chloride exchange.

MATERIALS AND METHODS
Reagents-4,4'-Dinitrostilbene-2,2'-disulfonic acid, disodium salt (Pfaltz and Bauer), was recrystallized one time as follows: 10 g of DNDS were dissolved in 200 ml of boiling H20, 100 ml of saturated NaCl in H20 (24 "C) were added, the suspension was cooled (4 "C) overnight, and the crystals were isolated and washed with 60% saturated NaCl in H,O (0 "C). When the crystals were redissolved in H20, they gave a single absorption maximum at 353 nm and the A3531 A310 ratio was 2.25, indicating pure or nearly pure trans-isomer (22).
Preparation of Ghost Membranes-Freshly outdated human blood (packed red cells) was a kind gift of the Los Angeles Chapter of the American Red Cross. Two units of any type were mixed and ghost membranes were prepared essentially as described previously (23,24). The following modifications were necessary to produce large quantities of leaky ghost membranes (-100 ml of pellet) which were not crushed by the forces of centrifugation (see Ref. 21). The entire preparation was carried out at 0 to 4 'C using a Sorvall GSA rotor, and the efficiency of all washes was maximized by filling the centrifuge bottles to maximum capacity. The packed red cells were aliquoted into six 250-ml centrifuge bottles and suspended in PBS (150 mM NaCl, 10 mM NaH2P04, pH to 8 with NaOH). The cells were pelleted by centrifugation a t 8,000 rpm (10,400 X g-) for 20 min, then the supernatant and buffy coat were removed by aspiration. The cells were washed twice more by resuspending each time in phosphatebuffered saline, then pelleting (10 min at 3,000 rpm or 1,500 X g , , ) , and then aspirating away the supernatant. Following the washes, the cells were lysed. The pellets were resuspended in 5P8(+) (or 5 mM NaH2P04, pH to 8 with NaOH, 130 PM dithiothreitol, and 10 p~ phenylmethylsulfonyl fluoride) then the membranes were pelleted by centrifugation at 11,000 rpm (19,700 X gma) for 20 min. The supernatant and the dense pellet underlying the membranes were removed by aspiration. This wash cycle in 5P8(+) was repeated six or seven times until the supernatant was colorless. The resulting leaky ghost membranes possess the typically observed range of shapes from biconcave to spherical, and few (-5%) crushed ghosts are produced (see Ref. 21). The membranes were used within 4 days and were stored at 4 "C.
NMR Sample Preparation-In all experiments samples were made on ice by diluting the ghost membrane pellet from the above preparation with an equal volume of ice-cold 2 X NMR buffer (see figure and table legends for final buffer compositions). Samples that were compared to each other, for instance those plotted in the same figure, were always made using aliquots of the same membrane suspension. Samples were always prepared and stored on ice and were assayed using 35Cl NMR the same day, within 10 h of preparation.
For experiments with added DNDS, a sufficiently large volume of ghosts in NMR buffer was aliquoted to give identical samples, then an appropriate volume of inhibitor (in H20) stock solution was added to each sample to give the desired final inhibitor concentration. The same total volume (50 pl/ml of sample) of inhibitor stock plus H20 was added to all the related samples so that they were identical except (when appropriate) for the inhibitor concentration. Due to the lightsensitive nature of DNDS, the DNDS stock solution was stored in darkness and was used only if it satisfied the conditions A353/A310 z 2.20 (or, more than 95% trans-isomer, the cis-isomer being inactive (22)). Also, for routine assays excess DNDS (1.0 mM total, which gives -0.9 mM unbound) was used so that full inhibition of the DNDS-sensitive line broadening was ensured even if some degradation occurred.
Samples containing different amounts of the anions chloride, bicarbonate, fluoride, or iodide (Figs. 5 and 6) were prepared by diluting 1 volume of membranes with 1 volume of a different 2 X NMR buffer stock for each anion concentration. The ionic strength was held constant in all samples by including a sufficient amount of citric acid (pH to 8.0 with NaOH). Thus, enough citrate was added to make the ionic strength the same as that of the sample containing the highest concentration (500 mM, Fig. 5, or 200 mM, Fig. 6) of the varying anion. No citrate was added to the latter sample.
35C1 NMR Spectroscopy-The spectra were obtained using one of two NMR spectrometers: a JEOL FX-90 (35Cl resonance frequency is 8.8 MHz) or a Varian XL-200 e5Cl resonance frequency is 19.6 MHz). The standard parameters for spectral acquisition were as follows. The spectral width was 1000 Hz, containing 256 data points and centered on the solution chloride peak in the 35Cl NMR spectrum. Using 5-or 10-mm sample tubes, from 1000 to 3000 pulses were accumulated for up to 6.4 min (3000 pulses) at 3 "C without sample spinning. An extra linebroadening of precisely 10.0 Hz was added to all samples during data processing to improve signal/noise, and the number of Fourier transform points was zero filled to over 8000 to smooth the spectrum (see Ref. 25). The central 500 Hz of the spectrum was plotted and the line width of the 35Cl peak at half-height hand measured. For all experiments the same acquisition parameters were used for samples that were compared to each other, for instance, those plotted in the same figure.
NMR Sample Analysis-After NMR spectra were obtained, samples were stored overnight at 4 "C before chemical analysis. Total ghost protein was determined using the modified (26) Lowry protein assay (27) that has been developed for use with membrane samples. The concentration of nonmembrane-bound DNDS was determined essentially as described elsewhere (28, 29). The membranes were pelleted by centrifugation for 45 min at 15,000 rpm (29,000 X g, , , J in a Sorvall 83-34 rotor, 3-ml tubes. The optical absorbance of the supernatant was measured and corrected for the background absorbance present in supernatants containing no inhibitor. The DNDS concentration was calculated using the molar extinction coefficient Statistics-All confidence limits for means and for best fit (nonlin-t363 = 3.0 X lo4 (22). ear least squares) parameters are given as ? one S.D. for n 2 3.

RESULTS
The 35Cl-NMR Resonance-For chloride in aqueous solution, the chloride 35Cl NMR spectrum (here termed the 35Cl-NMR spectrum) contains a single resonance of Lorentzian shape ( Figs. 1 and 2). Within experimental error, the shape of the observed resonance remains Lorentzian when leaky ghost membranes are added ( Figs. 1 and 2). Note, however, that the line width at half-height of the resonance is increased by the membranes (Fig. 1). This line width increase is due to the presence of membrane-bound chloride binding sites. In the following, the line width increase will be used as an assay for these sites; but first it is necessary to (a) examine the characteristics of the 35Cl-NMR spectrum and ( b ) demonstrate the validity of the binding site assay in the leaky ghost system. lution, chloride ions in the intracellular solution, and chloride ions bound to the membranes. The Lorentzian shape of the observed 35Cl-resonance indicates that the resonance stems from a homogeneous population of chloride ions. This homogeneous population contains 100 5 2% of the total number of chloride ions in the sample (from integration of the spectra used to generate Fig. 2). Thus, the observed population is large compared to the population of chloride bound to macromolecules (total protein 510 PM, observed chloride 250 2 5 mM), and it follows that the bulk of the observed population are solution choride ions. Both the intracellular and extracellular populations of solution chloride contribute to the observed 35Cl-resonance. The ratio of intracellu1ar:extracellular chloride ions in these samples is -25:75 (calculated assuming 100 pm3 total volume and 6 X mg of total protein/ghost membrane). The integration data indicate that the majority of chloride ions in both compartments are visible. The fact that these two populations give rise to a 35Cl-resonance due to a homogeneous population of chloride ions is not surprising, since large holes exist in these leaky ghost membranes ((30); see also Ref. 21). These holes should allow rapid exchange of chloride ions between the intra-and extracellular compartments such that the intra-and extracellular solution chloride ions are chemically equivalent on the NMR timescale. Thus, the 35Cl-NMR resonance of samples that contain ghost membranes is best described as the spectrum of a homogeneous population of solution chloride ions whose line width is perturbed (increased) by the presence of chloride binding sites.
Further Analysis of the Observed 35Cl-Resonance-Two characteristics of the observed 35Cl-NMR spectrum deserve further explanation. First, the shape of the observed spectrum is simple, despite the fact that the 35Cl nucleus ( S = "2) actually gives rise to three distinct NMR transitions. Second, the increase in line width caused by ghost membranes is of interest because this effect forms the basis of the assay for chloride binding sites used here.
Both the shape and the width of the 35Cl-spectrum are largely controlled by the effect of the quadrupolar interaction on the three 35Cl-NMR transitions. The 35Cl nucleus possesses an electric quadrupole moment, which interacts with the electric field gradient in the nucleus. The magnitude of this field gradient is large when the chloride ion's electron cloud is polarized by an asymmetrical ligand environment. However, the effect of the field gradient on the 35Cl-NMR spectrum is decreased when the chloride-ligand complex tumbles sufficiently rapidly to partially or completely randomize the direction of the field gradient.
The tumbling of hydrated chloride in aqueous solution is unrestricted, essentially isotropic and rapid. Under these conditions the quadrupolar interaction is averaged to zero and the three 35Cl-NMR transitions have identical resonance frequencies and line widths so that they sum to give a Lorentzian absorption ( Figs. 1 and 2). In contrast, a large quadrupolar interaction occurs when chloride binds to a slowly tumbling, asymmetric binding site on a macromolecule. The large quadrupolar interaction causes differences in the resonance frequencies and line widths of the three transitions so that the spectrum is no longer Lorentzian. However, in the experiments presented here, free chloride in solution is present in large molar excess relative to chloride bound to macromolecules. Thus, the solution chloride spectrum with its Lorentzian shape dominates the observed spectrum, even in the presence of ghost membranes ( Figs. 1 and 2).
The increase in the line width of the 35Cl-NMR resonance in the presence of ghost membranes is due to the exchange of chloride between solution and chloride binding sites associated with the membranes (Fig. 1). Due to the quadrupolar interaction and to shifts of resonance frequencies that occur upon binding, the line width of a 35Cl-NMR transition is typically over lo4 times larger for chloride bound to a macromolecule than for chloride in solution (17, 18). As a result, when chloride exchanges sufficiently rapidly between binding sites and solution, the observed line width is larger than that of pure solution chloride. In "Appendix 11" we present a detailed analysis of the observed 35Cl-NMR resonance and of the physical processes that cause the line width increase.
The Information Contained in the 35Cl-Line Broadening-The line width increase (or line broadening) contains a variety of information about the sites that give rise to it, as shown in the following simple theoretical analysis. The 35Cl-line broadening (6) is defined as.
Here ( A V~,~)~ is the line width (at half-height) of free chloride in solution, obtained using a blank sample, and Av11, is the observed line width when binding sites are present. In the presence of a heterogeneous population of independent sites the line broadening may be written ("Appendix II"): Here [Cl;] is the total (stoichiometric) chloride concentration, [XiC1] is the concentration of chloride bound to the ith type of site, and ai is a proportionality constant. Two fundamental assumptions have been made during the derivation of Equation 2 ("Appendix 11," also Refs. 31 and 32). First, it is assumed that the free chloride ions in solution are in vast molar excess relative to the bound chloride ions. This assumption is justified in the experiments presented here because the protein concentration is small: the ratio of band 3,

Chloride Binding to Band
3 Transport Sites 6475 the most abundant polypeptide in the ghost membrane (331, to the total chloride concentration is always Second, it is assumed that the only significant pathway available to a bound chloride ion is return to the solution. Here chloride bound to band 3 can undergo translocation as well as dissociation, but the dissociation rate is very fast3 relative to the translocation rate (turnover rate 400/s at 0 "C ( 2 2 ) ) ; thus, the use of Equation 2 is valid in the system at hand.
An important feature of Equation 2 is that the observed line broadening is the sum of the additive contributions from the different types of sites: where ai ( will give rise to a square hyperbola on a plot of line broadening versus [Cl-I", while on the same plot the curve due to a low affinity site (KDi >> [Cl-1) collapses to a straight line of zero slope. Consequently, binding sites can be operationally defined as high or low affinity sites on the basis of their behavior in this type of plot. Experimental Justification of the 35Cl-Line Broadening Assay for Chloride Binding Sites-The preceding theoretical analysis of the chloride binding assay is supported experimentally by the data of Fig. 3, which demonstrates that at constant reversibly to the extracellular band 3 anion transport sites in a manner that prevents chloride binding (1, 28, 36).
A saturating concentration (or, a concentration sufficient to yield maximal inhibition, see Fig. 4) of DNDS partially reduces the ghost line broadening (Fig. 3); the line broadening of DNDS-saturated membranes is 4.41 f 0.04 Hz/mg/ml of total protein, which is 36% less than the value for DNDS-free ghosts of 6.87 f 0.04 Hz/mg/ml. Thus, at least two types of chloride binding sites exist on ghost membranes. The sites that give rise to the DNDS-sensitive component of the line broadening are termed the DNDS-sensitive sites, while the remaining sites are DNDS-insensitive. The additivity of line broadening enables isolation of the DNDS-sensitive line broadening by subtraction of the DNDS-insensitive line broadening from the total line broadening. This DNDS subtraction technique will often be used in future presentations of data.

DNDS Binds to Band 3 Transport Sites and Thereby Inhibits the 35Cl-Line Broadening-The
line broadening of the DNDS-sensitive sites is inhibited by DNDS binding to a class of inhibitory binding sites that are identical, since the DNDS binding is well described by a single apparent dissociation constant (Fig. 4). This apparent KD is quantitatively similar to the known apparent KO for DNDS binding to band 3 transport sites in the presence of 250 mM chloride ( Table I). Note that because of the presence of chloride, which competes  with DNDS for binding (data not shown; also Ref. 28), the apparent KO is larger than the true KO for DNDS binding to the inhibitory sites ("Appendix 111"). DNDS is known to bind to the extracellular band 3 transport site with a stoichiometry of one molecule of DNDS/band 3 transport unit (28). When DNDS binds to this site, it inhibits the line broadening due to one or more chloride binding sites (Figs. 3 and 4). The affected sites could include: transport sites on band 3, other chloride binding sites on band 3, or chloride binding sites on a protein(s) other than band 3 that are allosterically coupled to the DNDS binding sites. One parameter which is useful in the identification of the DNDSsensitive sites is the affinity of these sites for chloride.
The DNDS-sensitive Sites Are High Affinity Chloride Binding Sites-Both high affinity and low affinity chloride binding sites are observed in the leaky ghost membrane system (Fig.  5A). In the presence of a saturating concentration of DNDS, the line broadening due to ghost membranes is well approximated by a best fit straight line of zero slope (Fig. 5A, lower  curue). Thus, the DNDS-insensitive sites can be operationally defined as low affinity sites that satisfy KDz >> 0.5 M. (The plot could not be extended to chloride concentrations greater than 0.5 M because DNDS becomes increasingly insoluble a t such chloride concentrations.) In the absence of DNDS, high affinity sites appear (Fig. 5A, upper curve). The resulting line broadening is the sum of the contributions from both the DNDS-sensitive and insensitive sites. This line broadening is well approximated by a best fit curve calculated for ( a ) a homogeneous set of high affinity binding sites described by Equation 8, plus ( b ) a set of low affinity binding sites described by the best fit straight line obtained for the DNDScontaining samples. The data indicate that saturation with DNDS completely inhibits the line broadening of the high affinity sites but has no effect on the low affinity sites.
The apparent dissociation constant for chloride binding to the high affinity sites can be determined from the best fit curve in Fig. 5A. Alternatively, the inverse of the DNDSsensitive line broadening can be plotted as a function of the chloride concentration (Fig. 5 B ) . This type of plot yields a straight line for a homogeneous set of chloride binding sites  HCO, (b), and 290 f 30 mM for F-(c). The buffer contained the indicated concentration of NaA as well as 100 mM NaCI, 2.5 mM NaH2P04, 20% D20, pH to 8.0 with NaOH. Citric acid (pH to 8.0 with NaOH) was added so that the ionic strength in all samples was the same as that of the [A-] = 200 mM sample. Spectral parameters: 8.8 MHz, 3 "C and standard assay parameters (see text). that are described by Equation 8. The extracted best fit apparent KDi, along with the results of seven similar experiments at 19.6 or 8.8 MHz, together yield an average of 80 & 30 mM for the apparent KDi for the chloride binding to the DNDS-sensitive sites. This value is quantitatively similar to the known apparent KD for chloride binding to band 3 transport sites (Table I).
Surprisingly, no chloride binding sites are observed that have a KD = 300 mM, which is the predicted chloride affinity for a site termed the modifier site (1,20). The existence of a modifier site on band 3 has been proposed to explain the inhibition of chloride self exchange across the membrane which occurs a t high [Cl-1. Such inhibition has been thought to result from chloride binding to the modifier site. Yet no evidence of this chloride binding site is seen in experiments presented here.
The High Affinity Sites Are Band 3 Transport Sites-In order to verify the conclusion that the DNDS-sensitive sites are transport sites and are uncontaminated with modifier sites, we have studied the binding of anions other than chloride to ghost membranes. If an anion (A-) competes with chloride for binding to an anion binding site, then the 35Cl-NMR assay can be used to study A-binding. In such an experiment [Cl-] is held constant, while [A-] is varied. The resulting data (Fig. 6, for the DNDS-sensitive sites) allow determination of the apparent K D for Abinding at the given [Cl-] (see "Appendix 111"). The best fit theoretical curves in Fig. 6 were calculated using the assumption that the DNDSsensitive sites are a class of identical sites. The reasonable fit of these curves indicates that the assumption of homogeneity is consistent with the data.
The apparent KD values for bicarbonate, fluoride, chloride, and iodide binding to the DNDS-sensitive sites are given in Table 11. Also given are the known apparent KD values for binding of these anions both to band 3 transport sites and to the hypothetical modifier sites. Note that the apparent KD values for the DNDS-sensitive sites were obtained in the presence of competing chloride (100 mM), while the other values were obtained in the absence of competing chloride. Two patterns emerge in the data of Table 11 values of the DNDS-sensitive sites are smaller than the apparent K D values predicted for modifier sites, indicating that the DNDS-sensitive sites have higher affinities for anions than those expected for modifier sites. ( b ) The relative affinities (I > HCOs >> C1 > F) exhibited by the DNDSsensitive sites are essentially the same as those exhibited by band 3 transport sites but are different from those expected for modifier sites. Together the data of Table I1 suggest that the modifier sites do not contribute significantly to the DNDS-sensitive line broadening. Instead the DNDS-sensitive sites behave like a homogeneous class of band 3 transport sites.

DISCUSSION
The 35Cl-NMR line broadening assay reveals two classes of chloride binding sites associated with leaky ghost membranes. The first class consists of low affinity binding sites that are unaffected by DNDS. The second class consists of band 3 transport sites. The line broadening of these transport sites is completely inhibited by a saturating concentration of DNDS. Such line broadening inhibition is consistent with the previous observation that DNDS competes with chloride for binding to the extracellular band 3 transport site (1, 28, 36).
The data presented here indicate that DNDS can be used to isolate that part of the total ghost line broadening which is associated with band 3 transport sites. Actually, DNDS is not very specific for these sites. For instance, we have found that DNDS reduces the line broadening associated with chlo-ride binding to hemoglobin (data not shown). However, the molar ratio of band 3:hemoglobin in ghosts prepared at pH 8.0 is at least 101 (39). This molar ratio would result in a corresponding ratio of DNDS-sensitive line broadenings of a t least 1OO:l (data not shown), which is a negligible interference. Another protein which could contribute to the DNDS-sensitive line broadening is the erythrocyte Ca*+-ATPase, which is inhibited by 4,4'-diisothiocyanostilbene-2,2'-disulfonate, a structural analogue of DNDS. Again, band 3 is present in great molar excess to this protein; there are -100 times more copies of band 3 per membrane (40). These examples stress that in the unpurified ghost membrane system, the DNDSsensitive line broadening could be contaminated with contributions from proteins other than band 3. Fortunately, band 3 is the most abundant polypeptide in ghost membranes (33), so these contaminating contributions could be negligible. This appears to be the case, since the DNDS-sensitive sites are homogeneous with respect to their affinity for a variety of anions (chloride, bicarbonate, fluoride, iodide, and DNDS). Moreover, the apparent affinities that are observed for chloride and DNDS are quantitatively similar to those previously measured for band 3 transport sites. The relative apparent affinities that are observed for the inorganic anions (I > HCO, > > C1> F) are essentially the same as those previously measured for band 3 transport sites. Thus, we believe that the bulk of the sites that make up the DNDS-sensitive class are in fact band 3 transport sites.
An unexpected conclusion of the 35Cl-NMR experiments is that the line broadening assay reveals no evidence for the existence of the proposed inhibitory anion binding site on band 3, which has been termed the modifier site (20). It is possible that this site exists but is invisible to the line broadening assay; a site will become invisible when the quantity ai is Equation 5 tends to zero. This condition occurs when the exchange of chloride between the binding site and solution is sufficiently slow or when the symmetry of the binding site is tetrahedral or higher.
Actually, we prefer a different explanation in which an inhibitory chloride binding site of the predicted (KO = 300 mM) chloride affinity does not exist. Instead inhibition results from nonspecific chloride binding to the low affinity (KD >> 0.5 M ) chloride binding sites that are revealed by the line broadening assay. In this model, bound chloride builds up in the vicinity of the low affinity sites as the chloride concentration increases. The local density of bound chloride ion would in turn inhibit the transport process. The step in the transport cycle which is inhibited is the chloride transmembrane translocation step rather than the chloride binding step, since no inhibition of chloride binding has been observed at chloride concentrations up to 500 mM (Fig. 5). Previous observations indicate that the onset of transport inhibition is rapid as the chloride concentration is increased above 250 mM (411, suggesting that the mechanism of inhibition is highly cooperative. A cooperative process which could give rise to the observed self inhibition of anion exchange is membrane protein aggregation. Such aggregation has been observed in the erythrocyte system at large [NaCl-] (42). Perhaps in the aggregated state, band 3 cannot undergo a conformational change which is necessary for the transmembrane translocation of bound anion. If aggregation of erythrocyte membrane proteins indeed causes the modifier effect, then a comparison of different anions should show a correlation between the ability of an anion to induce inhibition of band 3 and its ability to induce aggregation of erythrocyte membrane proteins.
In this paper we have used the 35Cl-line broadening assay to observe band 3 transport sites. However, certain questions concerning these sites have not yet been addressed. For instance, do transport sites on both sides of the membrane contribute to the line broadening? If sites on both surfaces are indeed observed, how could DNDS inhibit transport sites on both membrane surfaces when it is thought to bind only to the external transport site? These questions are considered in the following paper.