Tryptophan 1093 is largely responsible for the slow off rate of calmodulin from plasma membrane Ca2+ pump 4b.

Tryptophan 1093 resides in the 28-residue calmodulin-binding/autoinhibitory domain of the plasma membrane Ca(2+) pump (PMCA). Previous studies with the isolated calmodulin-binding/autoinhibitory peptide from PMCA have shown that mutations of the tryptophan residue decrease the affinity of the peptide for calmodulin and its affinity as an inhibitor of proteolytically activated pump. In this study, the PMCA mutation in which tryptophan 1093 is converted to alanine (W1093A) was constructed in the full-length PMCA isoform 4b. The mutant pump was expressed in COS cells, and its steady state and pre-steady state kinetic properties were examined. The W1093A pump exhibited an increased basal activity in the absence of calmodulin, so the activation was approximately 2-fold (it is 10-fold in the wild type). The W1093A mutation also lowered the steady state affinity for calmodulin from K(0.5) of 9 nm for wild type to 144 nm (assayed at 700 nm free Ca(2+)). Pre-steady state measurements of the rate of activation by Ca(2+)-calmodulin revealed that the W1093A mutant responded 2.5-fold faster to calmodulin. In contrast to these relatively modest effects, the half-time of inactivation of the mutant was reduced by more than 2 orders of magnitude from 41 min to 7 s. We conclude that tryptophan 1093 does not play a substantial role in Ca(2+)-calmodulin recognition; rather it functions primarily to slow the inactivation of the calmodulin-activated pump.

tains the calmodulin-binding region of the molecule. A rise in intracellular Ca 2ϩ results in 4 Ca 2ϩ binding to calmodulin; the Ca 2ϩ -calmodulin complex then binds to the inhibitory domain, activating the pump. Activation results in an increase in the affinity of the pump for Ca 2ϩ at its transport site as well as an increase in its maximal turnover.
PMCA in mammals is encoded by four genes with additional diversity provided by alternate splicing at two or more sites (for a recent review of PMCA genes, splicing variants, and nomenclature, see Strehler and Zacharias (1)). PMCA isoforms differ among themselves in three ways: 1) basal activities in the absence of calmodulin, 2) rates of activation by Ca 2ϩ -calmodulin, and 3) rates of inactivation by calmodulin dissociation. These differences are caused by differences in the interactions between the inhibitory domain, calmodulin, and the catalytic core. The present study investigates the effect of changing one residue in the inhibitory domain upon these three properties.
These differences in PMCA responses are integrated into cellular physiology. A cell that utilizes PMCA to respond rapidly to a Ca 2ϩ spike requires an isoform that has a high basal activity, such as isoform 3f in heart muscle (2), and/or one that is activated rapidly by calmodulin, such as isoform 2a in inner ear hair cells. In contrast, a cell that requires a sustained Ca 2ϩ signal may utilize isoforms with low basal activity and slow activation by Ca 2ϩ -calmodulin, such as isoform 4b in Jurkat T lymphocytes (3). In addition to their response to a single Ca 2ϩ spike, PMCA isoforms also have a potentially wide range of responses to a sequence of spikes. An isoform such as 4b, with low basal activity, relatively slow calmodulin activation, but much slower inactivation will allow the first spike to develop for several seconds before becoming activated but may remain activated for several minutes after the spike is dissipated, thus responding much faster to the next Ca 2ϩ spike. We have termed this the "memory effect" of the pump (4), and it is thought to be a major driving force for the large number of PMCA isoforms since it is a response that cannot be regulated simply by differences in activity or expression levels.
Since much of the information about the role of the inhibitory region of PMCA has been obtained through the use of synthetic peptides (5, 6), a discussion of the data is necessary as background for the present study. The synthetic peptides have been referred to by names such as C28 with C representing the calmodulin-binding domain and the numerals indicating the length of the peptide beginning with leucine 1086 (Fig. 1). C28 is the full-length peptide containing all of the residues necessary for high affinity calmodulin binding (7). Peptide C28 inhibited the calmodulin-activated form of the pump by competing with the enzyme for calmodulin. The calculated dissociation constant for calmodulin was 0.1 nM (5), which was comparable to the constant obtained by direct binding experiments with dansyl-calmodulin.
In other sets of experiments the C28 peptide was added to a proteolytic fragment or to an expressed construct (ct120), each of which had the C-terminal residues of PMCA removed. Both the proteolytic fragment and ct120 were fully active in the absence of calmodulin since each lacked the autoinhibitory domain. When C28 was added to a truncated PMCA, it caused a decrease in the activity of the enzyme, and the inhibited state was restored. These data showed that the calmodulin-binding domain itself is an inhibitor. The effect of changes in tryptophan 1093 in such peptides has been examined by a series of investigations using peptide C28 in which the tryptophan was replaced. In direct binding experiments with dansyl-calmodulin, changing tryptophan 1093 to alanine increased the K d for calmodulin from Ͻ1 to 18 nM (6). In comparable experiments using C28 as a calmodulin antagonist, the mutation increased the K d from 0.1 to 5 nM. The tryptophan to alanine mutation also had a strong effect on the ability of the peptide to inhibit proteolyzed pump, increasing the IC 50 to Ͼ30 M compared with 3 M for the wild-type C28 peptide (5).
Small angle x-ray scattering experiments with calmodulin and the C20 peptide revealed that only the C-terminal domain of calmodulin bound, and calmodulin remained in an extended, noncollapsed conformation (8). This extended structure was also confirmed by an NMR solution structure of the C20-Ca 2ϩcalmodulin complex (9). However, a longer peptide that contained both presumed anchor residues (C24 peptide, Fig. 1) bound both lobes of calmodulin, and calmodulin assumed a collapsed, globular formation (8) typical of calmodulin complexes with peptides from other well studied enzymes such as myosin light chain kinase (10,11) and calmodulin kinase II (12). The NMR structure of C20 bound to calmodulin revealed that the residue with the most contacts to calmodulin was tryptophan 1093 (tryptophan 8 in C20), the presumed anchor residue. The tryptophan R-group was buried in calmodulin, making contacts with nine different residues of the C-terminal domain.
The old peptide results, while very interesting, were from experiments necessarily performed on molecules differing substantially from the full-length PMCA. In full-length PMCA, the calmodulin-binding domain is flanked by upstream and downstream sequences and interacts with the catalytic core of the pump. To relate these results to the active pump, we constructed the full-length W1093A mutation and examined its kinetic properties. This is the first report of any full-length PMCA mutation affecting calmodulin activation. The aims of this research were to determine the kinetic role of the Cterminal domain anchor on 1) basal activity (degree of autoinhibition), 2) rate of activation by calmodulin, and 3) rate of inactivation by calmodulin dissociation. Our results show that the main role of Trp-1093 is to slow the inactivation of the pump when calmodulin is removed from the solution. Since a slower inactivation contributes to a longer molecular memory of previous stimulations (4), our result gives an underpinning to an important physiological property.

MATERIALS AND METHODS
Construction of W1093A-Overlapping PCR was used to create a cassette with a silent SacII site at the nucleotide positions partially encoding phenylalanine 1094 to glycine 1096 in the center of the 28residue calmodulin binding motif of hPMCA4b. The front PCR product was amplified with the primer pair 5Ј-TGACAACATCAACACAGCCC-3Ј/5Ј-CCGCGGAACGCAGGATCTGGCCTCGGCG-3Ј. The back PCR product was amplified with the primer pair 5Ј-CCGCGGCCTGAACC-GTATCCAGAC-3Ј/5Ј-CTCGAGGGTACCTCAAACTGATGTCTCTAGG-3Ј. The silent SacII sites are underlined, and the residues to change the tryptophan codon to alanine are shown with double underlines. The template for the PCRs was full-length h4b in pMM2 (pMM2 was originally called pMT2-m (13)). The front PCR product contained a unique endogenous XbaI site near the 5Ј-end, while a unique KpnI site was added to the 3Ј-end of the back PCR product immediately after the stop codon. Two PCR products were separately cloned into the pCR Blunt II TOPO vector (Invitrogen) and ligated at the SacII site. The W1093Aencoding insert was excised from pCR Blunt II TOPO with XbaI and KpnI. The excised insert was then ligated to XbaI/KpnI-cut vector of full-length h4b in pMM2 to replace the wild-type insert.
Construction of W1093A Mutant Minus the C-terminal 92 Residues-A single PCR was carried out with 5Ј-TGACAACATCAACA-CAGCCC-3Ј/5Ј-GGTACCCTAGGAACTATGGAACGCTTT-3Ј. The reverse primer contains a stop codon (underlined) to truncate the protein immediately after serine 1113 and a KpnI site for cloning. The PCR product was first cloned into pCR Blunt II TOPO, then excised with XbaI and KpnI, and ligated to the vector fragment of an XbaI/KpnIdigested full-length h4b in pMM2. This produced a construct ending immediately after the 28-residue calmodulin binding motif. This construct was named ct92 W1093A (C-terminal mutant truncated of 92 residues).
Heterologous Expression of PMCA in COS Cells-COS cells were transfected with the constructs using LipofectAMINE as described previously (14).
Isolation of Microsomes-Crude microsomal membranes from COS cells were prepared as described previously (15).
Ca 2ϩ Transport Activity-Ca 2ϩ transport was measured by the incorporation of 45 Ca 2ϩ into inside-out vesicles prepared from COS cells that transiently expressed PMCA. The conditions of the assay were 25 mM TES triethanolamine, pH 7.2, 7 mM MgCl 2 , 40 mM KH 2 PO 4 / K 2 HPO 4 , pH 7.2, 5 mM NaN 3 , 160 mM KCl, 5 ng/ml oligomycin, 400 nM thapsigargin, Ϯ544 nM calmodulin, 100 M total CaCl 2 , and EGTA was varied to obtain the desired free Ca 2ϩ concentration. The reaction was initiated with 6 mM ATP and allowed to proceed for 5 min at 37°C. The vesicles were then filtered, washed, and counted to measure 45 Ca 2ϩ incorporation.
Ca 2ϩ Dependence of Transport-This was measured as described above for Ca 2ϩ transport plus or minus 2 M calmodulin.
Calmodulin Dependence of Transport-This was measured at 0.7 M free Ca 2ϩ in the same assay conditions as described above. The data were plotted as fractional activation is the activity in the absence of calmodulin, V max is the activity in the presence of saturating calmodulin, and v is the activity at each plotted calmodulin concentration.
Solubilization of PMCA for ATPase Assays-COS cell membrane preparations were solubilized and reconstituted as described previously (16). Briefly, COS cell membrane preparations containing 100 -200 g of protein were pelleted by centrifugation in a microcentrifuge tube and resuspended in 80 l of solubilization buffer containing 60 mM TES triethanolamine, pH 7.2, 240 mM KCl, 10 mM MgCl 2 , 400 M EGTA, 10 mM NaN 3 , 2 mM dithiothreitol, 1 mM ouabain, 8 g/ml oligomycin, 400 nM thapsigargin, 4 g/ml aprotinin, 1 g/ml leupeptin, and 0.5% Triton X-100 at 4°C. After 4 min, 320 l of dilution buffer (same as solubilization buffer except 0.5 mg/ml phosphatidylcholine in place of Triton X-100) was added to the tube, and then 200 mg of Bio-Beads SM-2 were added. The tube was placed on a gel rocker for 1 h at 4°C, the bulk of the Bio-Beads was removed by centrifugation, and the few residual beads were removed by filtration with a 0.45-m spin filter.
Pre-steady State Rate of Activation by Calmodulin-The assay medium contained 30 mM TES triethanolamine, pH 7.2, 120 mM KCl, 5 mM MgCl 2 , 2.5 mM ATP, 5 mM NaN 3 , 1 mM dithiothreitol, 0.5 mM ouabain, 200 nM thapsigargin, 4 g/ml oligomycin, 2 g/ml aprotinin, 0.5 g/ml leupeptin, 200 M EGTA, enough CaCl 2 to obtain 0.5 M free Ca 2ϩ , Ϯ235 nM calmodulin, 0.2 mM MESG, and the coupled enzyme purine FIG. 1. Peptide and protein sequences used or discussed in this study. Residue numbering at the top is for human plasma membrane Ca 2ϩ pump isoform 4b. Only the 28-residue calmodulin-binding domain of the pump is shown in detail; where the sequences continue further, three periods are inserted. The predicted anchor residues for the two domains of calmodulin are highlighted in gray.
nucleoside phosphorylase at 1 unit/ml. In the presence of purine nucleoside phosphorylase, MESG is phosphorolyzed by inorganic phosphate to produce ribose-1-phosphate and 2-amino-6-mercapto-7-methylpurine. The time course of phosphorolysis of MESG was monitored on a Beckman DU 70 spectrophotometer as an increase in absorbance at 360 nm at 37°C.
Inactivation by Calmodulin Removal-Experiments to measure the rate of inactivation upon calmodulin removal were conducted in the same assay medium as for calmodulin activation. The enzyme was allowed to reach steady state at 0.5 M free Ca 2ϩ and 235 nM calmodulin, and then 10 M calmodulin-binding peptide RRKWQKT-GHAVRAIGRLSS from chicken smooth muscle myosin light chain kinase (17) was added to the reaction.

RESULTS
Steady State Transport Assays-W1093A was constructed in PMCA isoform h4b and transiently expressed in COS cells. COS cell membrane preparations (inside-out vesicles) were assayed for ATP-dependent incorporation of 45 Ca 2ϩ . Fig. 2 shows that the full-length W1093A mutant has about 50% basal activity (in the absence of calmodulin), while the wild type has 15%. The W1093A mutant was also constructed in a truncated isoform 4b lacking the C-terminal 92 residues (ct92). ct92 ends with the last residue of the 28-residue calmodulinbinding domain and therefore lacks some downstream inhibitory determinants (18). Because of this, wild-type ct92 is also stimulated only 2-fold by calmodulin as was also previously shown (7) (Fig. 2). The combination of truncation and W1093A mutation resulted in a pump with 70% of maximal activity in the absence of calmodulin. These experiments demonstrated that the W1093A mutation yielded a pump that was able to bind calmodulin and retained about half of the normal autoinhibitory properties of the calmodulin binding motif. The experiment combining truncation and the mutant showed that the inhibitory determinants destroyed by the two means were different since the inhibitory effects were additive. The higher basal activity of the W1093A mutants suggests a weaker interaction between the calmodulin-binding domain and the catalytic core.
Steady State Dependence of Ca 2ϩ Transport on Calmodulin-Since it was demonstrated that the W1093A mutant was indeed activated by calmodulin, kinetic experiments were performed to examine the steady state affinity for Ca 2ϩ -calmodulin. These curves were performed at 0.7 M free Ca 2ϩ . Fig. 3 shows that the calmodulin affinity of W1093A was much lower than the wild type. The K 0.5 for calmodulin was 144 nM for W1093A compared with 9 nM for the wild-type PMCA.
Steady State Ca 2ϩ Transport Curves-The curves in Fig. 4 were performed with either 0 or 2 M added calmodulin. Increasing Ca 2ϩ results in more Ca 2ϩ at the transport site and more of the activator Ca 2ϩ -calmodulin complex. W1093A showed a significant decrease in Ca 2ϩ affinity relative to wild type in the presence of 2 M total calmodulin. The K 0.5 for Ca 2ϩ was 0.35 Ϯ 0.03 M for the W1093A mutant compared with 0.19 Ϯ 0.01 M for the wild-type h4b. Since the activating species of PMCA is the Ca 2ϩ -calmodulin complex, the change in calmodulin affinity of the W1093A mutant was also reflected in a change in K 0.5 for Ca 2ϩ at constant calmodulin. These curves show that the differences in basal rates between the mutant and the wild type shown in Fig. 2   has shown that the wild-type h4b has a slow activation rate (half-time of about 1 min) and an extremely slow inactivation rate (half-time of more than 20 min) (16). For comparison, we did similar studies on the W1093A mutant. To perform the pre-steady state activation experiments (Fig. 5), wild-type or W1093A pump was allowed to reach steady state in the presence of 500 nM Ca 2ϩ , then 235 nM calmodulin was added, and PMCA was allowed to proceed to a new, activated steady state. The activation rate constant (k act ) was then obtained from the change in slope that occurred upon addition of calmodulin using the equation in the legend to Fig. 5. Wildtype h4b was activated by 235 nM calmodulin with a k for the fitted curve of 0.0245 s Ϫ1 , while the k obtained for W1093A was larger at 0.0574 s Ϫ1 . The values of k yield k act of 1.043 ϫ 10 5 M Ϫ1 s Ϫ1 for h4b and 2.443 ϫ 10 5 M Ϫ1 s Ϫ1 for W1093A. We used the total concentration of calmodulin (235 nM) for the calculation of k act . The actual concentration of the activator Ca 4 2ϩ -calmodulin would be dependent on the macroscopic constants for Ca 2ϩ binding to calmodulin (19) as well as cooperative events initiated by the binding of calmodulin with less than 4 Ca 2ϩ .
Pre-steady State Inactivation Measurements-The next experiment to examine the kinetic properties associated with tryptophan 1093 was to measure the rate of enzyme inactiva-tion by calmodulin removal. Since the steady state affinity of W1093A for calmodulin was lower than that of the wild-type h4b and the rate of activation by calmodulin was faster, it could be assumed that the rate of inactivation would be much faster for W1093A than wild-type h4b. Inactivation experiments were performed by allowing the enzyme to reach steady state at 500 nM Ca 2ϩ and 235 nM calmodulin, and then 10 M competing calmodulin-binding peptide from myosin light chain kinase was mixed with the reaction. The myosin light chain kinase peptide itself has no direct effect on PMCA but will sequester any calmodulin that dissociates from PMCA, and the vast excess of peptide should prevent any significant rebinding of calmodulin to PMCA. Fig. 6A shows the results of the inactivation experiment with wild-type h4b. The value obtained for k inact for h4b was 2.8 ϫ 10 Ϫ4 s Ϫ1 . The corresponding half-time for the reaction is 41 min. Fig. 6B shows that the rate of inactivation of the W1093A mutant was much faster than the wild type with a k inact of 0.1 s Ϫ1 and a corresponding half-time of 7 s.   (Ϫk inact t)). The apparent rate constant for inactivation (k inact ) was 0.00028 Ϯ 0.00002 s Ϫ1 for h4b and 0.100 Ϯ 0.004 s Ϫ1 for W1093A. These values of k inact correspond to half-times of 2483 and 7 s, respectively. The data shown are representative of at least three experiments, and the constants are given as mean Ϯ S.E. Coefficients of determination (r 2 ) for the fitted curves were all greater than 0.9998. The dotted straight lines in the figures are lines fitted to the minimum rate achieved at long times and are shown to allow a visual estimate of the curvature.

DISCUSSION
From the data for k act and k inact of ATPase experiments it is possible to calculate dissociation constants for calmodulin and compare them to the apparent K d values obtained from steady state Ca 2ϩ transport. The calculated K d for h4b from the presteady state experiments (performed at 500 nM Ca 2ϩ ) is ϳ3 nM. This was somewhat less than the apparent K d of 9 nM obtained for the steady state experiments (performed at 700 nM Ca 2ϩ ). The calculated K d for W1093A is ϳ407 nM compared with an apparent K d of 144 nM. The agreement of these numbers is quite good considering that phosphatidylcholine membranes were used for the ATPase measurement, and native COS cell membrane vesicles were used for the Ca 2ϩ transport assays. We consider the K d values obtained from the pre-steady state data to be more accurate because PMCA is enriched about 10-fold by the reconstitution, and the Ca 2ϩ concentration is better controlled with the buffer system used for the pre-steady state assays.
The data presented here show that the W1093A mutant differs from the wild type in three ways. The first is that the mutant has a less effective interaction of the calmodulin-binding domain with the catalytic core so that it retained only half of the normal autoinhibitory properties. The second is that the rate of activation by calmodulin is faster for the mutant, and the third is that the mutant exhibits a much faster dissociation of the Ca 2ϩ -calmodulin complex.
The basal activity in the absence of calmodulin is controlled by the interaction of the calmodulin-binding domain with the catalytic core. The degree of autoinhibition is also affected by the downstream regulatory region (7). In the absence of calmodulin, an equilibrium exists between closed and open conformations (in the closed conformation, the calmodulin-binding domain interacts with the catalytic core, while in the open conformation it does not). The catalytic core and calmodulin may be viewed as competing for the open conformation of the calmodulin-binding domain. Changing tryptophan 1093 to alanine increased the basal activity of the pump (in the absence of calmodulin); in the wild type the basal activity was 15% of the maximal calmodulin-stimulated activity, while in the mutant it was 50%. This indicates that the mutant pump was shifted more toward the open conformation, and thus the calmodulin-binding domain was more accessible to calmodulin. As was mentioned in the discussion of Fig. 2 (see "Results"), the tryptophan side chain was responsible for ϳ50% of the autoinhibitory properties of the calmodulin-binding domain. This confirms what had been seen with the peptide studies that show that tryptophan 1093 has a vital role in the autoinhibitory interaction as well as being an important anchor for calmodulin binding.
The displacement of PMCA toward the open state by the mutation should also make the calmodulin-binding domain of the mutant more accessible to Ca 2ϩ -calmodulin. This accessi-bility may be reflected in an acceleration in the rate of activation by Ca 2ϩ -calmodulin. Such an effect was observed for the W1093A mutation, which increased the rate of activation ϳ2.5fold. The effect of opening the pump is probably adequate to account entirely for the activation rate increase without needing to invoke any effect of tryptophan 1093 in calmodulin recognition.
In contrast with the small effect of the mutant on activation, the mutant greatly accelerates the inactivation (350-fold). This large effect indicates that tryptophan 1093 has a major role in slowing dissociation of Ca 2ϩ -calmodulin from the pump.
The small role of tryptophan 1093 in the initial interaction with calmodulin is rather surprising because of the dominant role of this tryptophan in stabilizing the fully formed complex with calmodulin. The NMR solution structure showed the tryptophan interacting with nine residues of calmodulin compared with only three calmodulin residues interacting with the next most important residues in the calmodulin-binding domain of the pump (9). These nine interactions evidently contribute very strongly to the stabilization of the pump-calmodulin complex and hinder its dissociation. However, the initial binding of calmodulin must involve recognition of a number of important features among which the tryptophan is not a major contributor.