Association between Human Erythrocyte Calmodulin and the Cytoplasmic Surface of Human Erythrocyte Membranes *

This report describes Ca2+-dependent binding of lZ51labeled calmodulin (‘“I-CaM) to erythrocyte membranes and identification of two new CaM-binding proteins. Erythrocyte CaM labeled with ‘251-Bolton Hunter reagent fully activated erythrocyte (Caz+ + Mg*+)-ATPase. ”‘I-CaM bound to CaM depleted membranes in a Ca2+-dependent manner with a K , of 6 X M Ca2+ and maximum binding at 4 X M Ca2+. Only the cytoplasmic surface of the membrane bound IZ5I-CaM. Binding was inhibited by unlabeled CaM and by trifluoperazine. Reduction of the free Ca2+ concentration or addition of trifluoperazine caused a slow reversal of binding. Nanomolar ”‘I-CaM required several hours to reach binding equilibrium, but the rate was much faster at higher concentrations. Scatchard plots of binding were curvilinear, and a class of high affinity sites was identified with a K D of 0.5 nM and estimated capacity of 400 sites per cell equivalent for inside-out vesicles (IOVs). The high affinity sites of IOVs most likely correspond to Ca2+ transporter since: (a) K, of activation of (Ca2+ + Mg2+)-ATPase and K D for binding were nearly identical, and (b) partial digestion of IOVs with a-chymotrypsin produced activation of the (Ca2+ + Mg2+)-ATPase with loss of the high affinity sites. ‘”1-CaM bound in solution to a class of binding proteins ( K D 55 nM, 7.3 pmol per mg of ghost protein) which were extracted from ghosts by low ionic strength incubation. Soluble binding proteins were covalently cross-linked to ’25J-CaM with Lomant’s reagent, and 2 bands of 8,000 and 40,000 M, (Mr of CaM subtracted) and spectrin dimer were observed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis autoradiography. The 8,000 and 40,000 M, proteins represent a previously unrecognized class of CaM-binding sites which may mediate unexplained Ca“+-induced effects in the erythrocyte.


Association between Calmodulin
and Erythrocyte Membranes 6259 washed again before lysis in 15 liters of 7.5 mM NaP04, 1 mM NaEGTA, 10 pg/ml of PMSF, 2 pg/ml of pepstatin A, pH 7.5, 0 "C. Solid NaCl was added to the lysate (to 0.15 M ) and 40 ml of DE52 cellulose (cycled in 7.5 mM NaP04, pH 7.5) was added and stirred for 30 min. The gel was washed white on a sintered glass filter with 0.15 M NaCl in buffer A (7.5 mM NaP04, 0.1 mM NaEGTA, 0.2 mM dithiothreitol, 1 mM NaN3, pH 7.5) and then poured into a column and eluted in two steps of buffer A (step 1 with 0.25 M NaCI, step 2 with 0.5 M NaC1). The second eluate was concentrated against polyethylene glycol 6000 flakes, dialyzed against 50 mM NaCl in buffer A, and applied to an AcA 54 Ultrogel column (2.6 X 65 cm) equilibrated with 50 mM NaCl in buffer A. The third A280 peak (1.6 X V,J was dialyzed against 0.15 M NaCl in buffer A and loaded onto a DE52 cellulose column (1 X 14 cm). The column was eluted with a 100-ml linear gradient of 0.15 to 0.6 M NaCl in buffer A. The first peak (eluted a t 0.26-0.28 M NaCl) had an absorption spectrum identical with CaM (7) and was >98% pure by SDS-PAGE. This was dialyzed against 50 mM NaC1, 5 mM NaP04, 1 mM NaN3, 10 p~ CaCI,, p H 7.5, and frozen at -20 "C in aliquots of 0.4 mg of protein/ml. The second peak (0.34-0.5 M NaCl) was probably nucleic acid.
Pure erythrocyte CaM was radiolabeled with "'I-Bolton Hunter reagent (25). One mCi of Iz5I-reagent (2200 Ci/mmol) in benzene was dried in the original vial with a stream of N,. Fifty-100 pl of CaM (4-40 pg in 40 mM NaP04, pH 8.1) were added for 90 min at 0 "C, diluted to 0.5 ml (with 0.5 mg of gelatin in 100 mM Hepes, 1 mM NaNs, 0.2 mM dithiothreitol and 50 p~ CaCl,), and dialyzed overnight at 2 "C against the same buffer. There was 30-55% incorporation of "' I into CaM (determined by 10% trichloroacetic acid precipitation of an aliquot of reaction mixture diluted with bovine albumin carrier), and specific activities ranged from 0.06 to 1.05 mol of '251 per mol of CaM. lZI-CaM was usually diluted to approximately 2 X cpm/ pg by adding unlabeled CaM immediately after the reaction (except when biological activity of I2'I-CaM was determined (Fig. 1)). Aliquots were frozen at -20 "C, and binding characteristics were unchanged even after several weeks.
White ghosts were prepared from fresh human blood (24)  [T-~*P]ATP (7700 cpm/nmol) (final concentration 0.4 mM) and MgCl2 (final concentration 0.8 mM) were added for an additional hour a t 24 "C. Trichloroacetic acid (1.0 ml, 5% w/v, 0 "C) and then Norit A (0.2 ml, 5% w/v) were added and free Pi in the supernatant was determined by counting Cerenkov radiation after centrifugation for 15 min at 4000 X g. Basal and CaMactivated (Ca2+ + Mg+)-ATPase activity were calculated after subtracting the amount of P, hydrolyzed in the absence of membranes.
Free '2'I-CaM concentrations were determined in parallel as described (see under "Methods"). The data are expressed as the average of duplicate determinations.
FIG. 2  in 7.5 mM NaP04, 1 mM NaEGTA, 10 pgjml of PMSF, pH 7.5, with a final wash and storage a t 0 "C in 10 mM Hepes, 1 mM NaN3, 0.1 mM dithiothreitol, pH 7.3. IOVs were prepared quantitatively from ghosts as described but omitting the dextran step (26) , and   IOVs were   washed and stored like ghosts. IOV cell equivalents were calculated   by comparing the band 3 content of IOVs and ghosts (determined by   scanning Coomassie blue stained SDS-PAGE slabs). Fresh ghosts contained "3.5 X 10"' mg of protein per cell equivalent and IOVs -2.6 x 10"". (Ca" + M?)-ATPase was assayed within 24 h and "'I-CaM binding within 3 days. 'Z'II-CaM binding was determined by incubating '2'II-CaM (0.25-200 nM) with ghosts or IOVs (10-60 pg of protein). The incubation was in 0.2 ml of 100 mM Hepes containing 0.25 mg/ml of gelatin (a neutral carrier which reduced nonspecific adsorption of subnanomolar CaM concentrations to plastic) and 2.50 mM NaEGTA with or without CaCI2 (providing a specific pCa2+ (27)), pH 7.30, in polystyrene tubes (12 X 75 mm) a t 24 "C. Bound and free '2'I-CaM were separated by layering 0.18 ml over a 0.2-ml sucrose (20% w/v) barrier containing the same buffer with or without CaCI2 in 400-pI hard polyethylene Eppendorf microtest tubes. The tubes were centrifuged for 30 min at 25,000 X g and frozen in dry ice. The t.ips were clipped off and assayed for "' I in a y counter. Ca"-independent binding was measured by including 2.50 mM NaEGTA without CaC12 in the incubation mixtures and sucrose barriers for each concentration of membrane or "'I-CaM, and the value (-0.5% of total cpm added) was subtracted from the corresponding CaC12-containing sample to yield Ca2+-dependent binding. Values were determined in duplicate and the range was within ?5%.
Protein was estimated by the method of Lowry et al. (28) using bovine serum albumin as a standard.

RESULTS
Radiolabeled binding proteins must retain biological activity if physiologic conclusions are to be made. CaM radiolabeled with 'Y51-Bolton Hunter reagent retained full ability to activate brain phosphodiesterase (29), although another report noted such preparations had reduced biological activity (22). The CaM in this report was purified from erythrocytes and radiolabeled with '"I-Bolton Hunter reagent to 1.05 mol of ' ' ' 1 per mol of CaM. 12'II-CaM and native CaM activated erythrocyte membrane (Ca2+ + Mg2+)-ATPase identically (Ka = 0.3 nM, V,,, = 4.5 X basal, Fig. 1). The erythrocyte ghosts in this study were prepared in buffers containing EGTA and retained less than 0.1% of native erythrocyte CaM (Table I). These ghosts were permeable to large molecules such as Ficoll

6260
Association between Calmodulin and Erythrocyte Membranes and did not reseal during the CaM binding assay (see under "Methods"). Characteristics of Ca2+-dependent 1251-CaM Binding to Erythrocyte Membranes-Ca"-dependent binding (see below) occurred at intracellular sites and increased linearly with increased concentrations of membranes. Intact erythrocytes failed to bind '251-CaM (Fig. 2) indicating that binding was restricted to the cytoplasmic membrane surface. Binding to ghosts and IOVs increased linearly up to 0.06 mg of ghost protein per assay, so all studies were conducted in the linear range. IOVs bound less lZ5I-CaM than ghosts suggesting a loss of binding sites during preparation (see below). lZ51-CaM binding depended upon the free Ca2+ concentration. Membranes bound negligible lZ5I-CaM at pCa2+ 8.0 but reached maximum at pCa2+ 6.4 (Fig. 3). Caz+-independent binding was subtracted from all data since it was judged to be nonspecific. Ca2+-independent binding did not saturate at increasing concentrations of lZ51-CaM (Fig. 8, A and B ) , was not reduced by trifluoperazine (Fig. 5 ) , and was not displaced by excess unlabeled CaM (Fig. 4). Furthermore, Ca2+-inde-  'T-CaM (14 nM, 45,000 cpm/pmol) was incubated for 2 h at 24 "C with ghosts (0.14 mg of membrane protein/ml) in 0.1 M Hepes, 0.25 mg/ml of gelatin with 2.50 mM NaEGTA, 2.376 mM CaC12, pH 7.30 (pCa2+ 6.00, 0), or with 7.50 mM NaEGTA, 2.376 CaC12, pH 7.30 (pCaZ+ 7.61, O), or with 2.50 mM NaEGTA, 2.376 mM CaC12, pH 7.30, but with addition of NaEGTA at 120 min to a final concentration of 7.50 mM (pea" 6.00 + 7.61, A). Aliquots were removed at various times thereafter and Ca2+-dependent binding was determined (see under "Methods"). pendent binding was not time dependent.
Membrane-bound lZ51-CaM was dissociated by lowering the free Ca2+ concentration or by adding trifluoperazine (Figs. 5 and 6). Maximum high affinity binding occurred at pCa2' 6.0. Subsequent addition of EGTA reduced the free Ca2+ concentration to pCa2+ 7.61, and binding was slowly reversed (Fig.  6). The reversal was biphasic on a semi-log scale with the first

Association between Calmodulin and
Erythrocyte Membranes 6261 TI,, -18 min and second Tl12 -80 min. Dissociation of membrane-bound Iz5I-CaM required 4-fold greater concentrations of trifluoperazine than required for inhibition of binding. The reversibility is further evidence that the binding is specific and does not represent trapping.
Binding was slow at low concentrations of '251-CaM (Fig.   7 ) . Binding of 1.5 nM "'I-CaM (a concentration near the K, for activation of (Ca2+ + M e ) -A T P a s e ) was still increasing slightly after 4 h of incubation. Slow association and slow dissociation indicate that the sites are in slow equilibrium with CaM. The slow off-rate could also influence the extent of extraction of CaM from membranes during ghost preparation (Table I). The on-rate was driven much faster at 60 nM "'I-CaM (Fig. 7). Erythrocytes contain M CaM (7), so binding may be extremely rapid in uiuo.
Analysis of Membrane-binding Affinities and Capacities-Binding of '"I-CaM to ghosts and IOVs was measured as a function of CaM concentration (Fig. 8, A and B). Scatchard plots were curvilinear at equilibrium (Fig. 8C) indicating either multiple independent sites or negatively cooperative associations at a single site. Negative cooperativity affecting a single class of sites is unlikely since the high affinity sites were selectively removed by proteolytic digestion (Fig. 10, inset), and three different binding proteins have been identified (see below). High affinity binding sites were resolved from lower affinity sites with a reiterative qonlinear two-site fitting program (31). The capacity estimated for ghosts was 4.7 pmol/mg of membrane protein (-1000 high affinity sites per cell) and for IOVs 2.4 pmol/mg (-400 high affinity sites per cell equivalent) assuming K I~ = 0.3 nM. These values are -30% smaller than estimates made by linear extrapolation from the high affinity slope in Fig. 8C (ghosts, 6.4 pmol/mg and IOVs, 3.4 pmol/mg). High affinity binding was measured in more detail with several concentrations of lZ5I-CaM below 1 nM to determine the K,, more accurately (Fig. 9B). Double reciprocal binding plots for both ghosts and IOVs indicated that the high affinity binding KD = 0.5 nM, and this value is essentially identical with the K, = 0.3 nM for CaM activation of (Ca'+ + M$+)-ATPase measured under identical conditions (Fig. 9A).
The IOVs contained only half as many CaM-binding sites Soluble binding sites were identified in the low ionic strength extract (Fig. 11, see below). It is technically difficult to quantitatively correlate binding of lZ5I-CaM to membrane sites with binding to soluble sites, and it is likely that the soluble binding estimates are too low. The soluble extract, however, contained very little (Ca2+ + Mg2+)-ATPase activity (data not shown). It is unlikely that CaM binding sites are sequestered in right-side-out vesicles since the methods used here remove >95% of all spectrin from ghosts yielding >85% inside-out vesicles (26). The reduction in CaM binding sites is probably not due to damage of sites since neither repeated freezing and thawing nor prolonged storage at 0 "C reduced binding (data not shown). (Caz+ + MF)-ATPase activity, however, is much more labile with continuous loss of (Ca2+ + Mg2+)-ATPase activity even when chilled at 0 "C and abrupt loss of activity after exposure to sulfhydryl reactants (data not shown). The Ca2+ transporter is an integral membrane protein extractable only with detergents (19) and remains in  the IOV membranes after low ionic strength extraction. These conditions most likely remove a different class of CaM-binding proteins and also remove (or damage) a different class of (Ca'+ + Mg")-ATPase which is inactive in solution.
The high affinity binding sites remaining on IOVs most likely represent binding of "'I-CaM directly to the Ca" transporter. Estimates of the KD and K, were nearly identical (Fig. 9). IOVs were estimated to retain -400 nonextractable high affinity binding sites per cell equivalent which is the number of (Ca2+ + M$+)-ATPase copies per erythrocyte estimated from studies of phosphorylated intermediates (32). This value is much lower than estimates based on turnover number (33), photoaffinity labeling (34), or direct binding under very different conditions (23). (ea2+ + Mg"+)-ATPase was thought to be proteolytically activated by removal of CaM binding sites of the enzyme (20,35,36). Mild a-chymotrypsin digestion of IOVs produced activation of the (Ca2+ + Mg")-ATPase with loss of additional CaM stimulation and loss of most high affinity CaM binding sites but sparing of the low affinity sites (Fig. 10). It is unlikely that the 12'II-CaM was damaged by persistent traces of a-chymotrypsin since the supernatant (unbound I2'I-CaM) subsequently bound well to other membranes (not shown). Interestingly, other CaMsensitive enzymes are activated in the absence of CaM by partial proteolysis (phosphodiesterase (37), phosphorylase b kinase (38), and myosin light chain kinase (39)) suggesting that CaM regulates other enzymes by a similar manner.
Solubilized CaM-binding Sites-It is clear that when IOVs were prepared from ghosts, binding sites were removed (Figs. 2,3,8, and 9). Spectrin binds CaM with a KO = 2.8 X M (16), but the binding sites removed during preparation of IOVs were of much higher affinity. These sites were not destroyed since a significant number of sites were recovered in the extract. Binding of "'I-CaM in solution was measured by a modified gel filtration method (Fig. 11 (40)). The peak in the upper panel represents "'1-CaM excluded from a pre-  (Ca" + M e ) -A T P a s e activity was calculated from free P, determinations (see Fig. 1).
viously equilibrated column due to Ca"-dependent interaction with soluble binding sites and was not detected in the absence of ea2+ (lower panel). The affinity of the association was estimated by separation of bound and unbound '251-CaM by gel filtration over a range of "'I-CaM concentrations (Fig.  12). Scatchard plots were curvilinear and tangential extrapolation along each of three regions suggests that different solubilized binding sites exist with most points falling along tangent Y ( K D = 55 nM, N = 7.3 pmol/mg based upon the original membrane protein). There also appeared to be a very small number of higher affinity sites (slope X ) and another class of sites (2) which did not approach saturation at 150 nM '"I-labeled CaM. Solubilized binding sites were identified by covalent crosslinking to "'I-CaM and SDS-PAGE autoradiography (Fig.  13). "'I-CaM has been shown to interact directly with calci-

Association between Calmodulin and
Erythrocyte Membranes 6263 20 40 60 Fraction no.   extrapolations (x, y , and z ) . Protein concentration refers to mg of protein of the original ghosts from which the extract was made.  50% (lanes 6-7), so the M, = 8,000 and 40,000 proteins may correspond to class Y sites (Fig. 12). Some radioactivity appeared on top of the lanes. This consisted of ""I-CaM bound to spectrin dimer (M, = 460,000) and large "'I-CaM aggregates which were separated on more porous gels (not shown). The low affinity large capacity sites (class 2, Fig. 12) probably correspond to spec-

6264
Association between Calmodulin and Erythrocyte Membranes trin, since purified spectrin dimer will cross-link to '251-CaM in the presence of Ca2+ (not shown). Neither M , = 8,000 nor 40,000 protein was found free in the cytosol (not shown). Little M , = 40,000 protein remained on IOVs while about half of the M , = 8,000 band remained (Fig. 13, lane 10). Ifestimates of K D from Fig. 12 apply to IOVs, it is likely that these contribute to the lower affinity sites (Fig. 8C). It is unlikely that the M , = 8,000 and 40,000 proteins are degradation products of the Ca2+ transporter since they appear without variation under a variety of extraction conditions, and there is no evidence of proteolytic degradation of ankyrin (42) or protein 4.1 (not shown). Also, there are too many copies of these proteins for them to be derived from the Ca2+ transporter, and their appearance is not accompanied by activation of (Ca'+ + M$*)-ATPase. Cross-linking of "'I-CaM to the M , -150,000 Ca'+ transporter was inefficient under these conditions but has been accomplished with a photoaffinity label (34). Altogether these observations are most consistent, with the high affinity sites on IOVs corresponding to the Ca2+ transporter and the M , =  Membrane binding of lZ5I-CaM was very slow at concentrations near 1 nM where binding to the Ca2+ transporter is predominant and required several hours to reach equilibrium ( Fig. 7). CaM activation of (Ca2+ + M$+)-ATPase (44) and binding of lZ51-CaM to erythrocyte ghosts (21,22) were both interpreted as positively cooperative interactions. Both phenomena might be explained by incomplete binding at the lowest CaM concentrations, for neither were observed in this study when sufficiently long incubations were employed. However, both were observed after short incubations (data not shown). CaM at 1 nM activated (Ca2+ + Mg+)-ATPase after a lag period, but this was eliminated by preincubating membranes with CaM (45). The binding rate, as expected for Photoactivated affinity cross-linkers such as azido-CaM react quickly and randomly with a variety of carbon-hydrogen bonds, whereas the chemical cross-linker, Lomant's reagent, is longer lived and specifically reacts with nucleophiles such as amines. Therefore, studies performed by these different methods of cross-linking may not produce identical results. a bimolecular reaction, was driven much faster at higher concentrations of CaM (Fig. 7). Experimental observations of high affinity interactions require unphysiologic dilutions of CaM (lo-' M), and nonequilibrium experiments are vulnerable to artifact resembling positive cooperativity due to the slow rate of binding. Also, extraction of native CaM from erythrocyte ghosts may be incomplete due to slow reversal of binding. It was found that the ghosts and IOVs used in this study retained (0.1% of basal erythrocyte CaM (Table I), while a 10-to 20-fold higher level of residual CaM was reported with high basal (Ca2+ + Mg2')-ATPase activity (46).
Binding of Iz5I-CaM to erythrocyte membranes increased as free Ca2+ rose from pCa2+ 8 to pCa2+ 6.4. Erythrocyte cytosolic free Ca2+ concentrations were thought to be -1O"j M (9), but free Ca" is difficult to measure. Nondisruptive introduction of an intracellular chelator has shown the resting erythrocyte free Ca2+ to be approximately 2 X IO-' M (47). Physiological shear stresses have been found to greatly enhance Ca2+ influx (48). Therefore, it is likely that the Ca2+ transporter must respond to a sudden influx of Ca2+ during turbulent arterial flow, pump out Ca2+ until the free concentration is lo-' M, and then switch off. Ca2+ is considered an essential intracellular signal (49), and it is likely that the shear related influx of Ca2+ produces other CaM-mediated physiologic effects, perhaps a reversible contraction of the membrane skeleton mediated by the M , = 8,000 or 40,000 CaM-binding proteins. A temporary contraction should help the cell survive rapid flow related stress and is probably distinct from the pathological Ca2+ effects produced by M Ca2+ introduced with ionophores. The high affinity binding of CaM to membrane Ca2+ transporter would also be expected to rise dramatically as free Ca2+ rises above pCa2+ 8.0 (Fig. 3) and would fall off the membrane as the free Ca2+ is reduced (Fig. 6). The ATPase activity of the Ca2+ transporter, however, is negligible below pCa2+ 7.0 and rises to maximum activity near pCa2+ 5 (20). Thus there appears to be a discrepancy between the free Ca2+ range required for high affinity CaM binding (pCa'+ 8 -6.4) and the concentration range required for activation of (Ca2+ + M F ) -A T P a s e (pCa2+ 7 -

5.5).
The discrepancy in Ca2+ requirements for membrane binding and (Ca2+ + M e ) -A T P a s e activation suggests that two steps are involved. CaM is known to have four different Ca2+ binding sites with micromolar affinities which fill in a preferred sequence, and probably all sites need not be filled in order for the complex to activate some enzymes (50). At submicromolar Ca2+ concentrations it is possible that CaM occupied by a single Ca2+ ion could bind to the (Ca" + M$+)-ATPase which could shift it to a potentially activated form, and a second step would be required for final activation. Perhaps CaM occupied by only one Ca2+ ion will bind to the enzyme but the CaM must be occupied by 2 or 3 additional Ca2+ ions in order for it to completely activate the enzyme. Alternatively, once CaM has bound to the regulator site on the enzyme, additional Ca2+ ions may activate the enzyme directly by binding to the catalytic site of the enzyme as substrate. This hypothesis is likely since partial proteolysis removes the CaM binding regulator site of the enzyme. The digested enzyme is no longer dependent upon CaM but is still dependent upon free Ca2+ very much like the CaM-activated enzyme (20,36). The KO of CaM for Ca2' and the KIM of Ca2+ transporter for Ca2+ are both in the micromolar range which is consistent with the Ca2+ concentration being rate limiting for both steps.
Measurement of "'I-CaM binding to erythrocyte ghosts and IOVs may be useful in evaluating clinical disorders such as Duchenne muscular dystrophy (51,52) or sickle cell anemia