Inhibition of Phosphorylation of Na+,K+-ATPase by Mutations Causing Familial Hemiplegic Migraine*

Background: Familial hemiplegic migraine type II (FHM2) is caused by mutations in the Na+,K+-ATPase α2-isoform. Results: Several FHM2 mutations inhibit phosphorylation or dephosphorylation. Conclusion: These mutations cause FHM2 by local and long range effects on the catalytic site and not by reducing the affinity for external K+. Significance: Insights into the pathophysiological mechanism of FHM2 and the molecular mechanism of the Na+,K+-ATPase have been obtained. The neurological disorder familial hemiplegic migraine type II (FHM2) is caused by mutations in the α2-isoform of the Na+,K+-ATPase. We have studied the partial reaction steps of the Na+,K+-pump cycle in nine FHM2 mutants retaining overall activity at a level still compatible with cell growth. Although it is believed that the pathophysiology of FHM2 results from reduced extracellular K+ clearance and/or changes in Na+ gradient-dependent transport processes in neuroglia, a reduced affinity for K+ or Na+ is not a general finding with the FHM2 mutants. Six of the FHM2 mutations markedly affect the maximal rate of phosphorylation from ATP leading to inhibition by intracellular K+, thereby likely compromising pump function under physiological conditions. In mutants R593W, V628M, and M731T, the defective phosphorylation is caused by local perturbations within the Rossmann fold, possibly interfering with the bending of the P-domain during phosphoryl transfer. In mutants V138A, T345A, and R834Q, long range effects reaching from as far away as the M2 transmembrane helix perturb the function of the catalytic site. Mutant E700K exhibits a reduced rate of E2P dephosphorylation without effect on phosphorylation from ATP. An extremely reduced vanadate affinity of this mutant indicates that the slow dephosphorylation reflects a destabilization of the phosphoryl transition state. This seems to be caused by insertion of the lysine between two other positively charged residues of the Rossmann fold. In mutants R202Q and T263M, effects on the A-domain structure are responsible for a reduced rate of the E1P to E2P transition.

Hemiplegic migraine is a severe subtype of migraine with aura associated with transient motor weakness and sensory as well as speech difficulties. It is autosomal dominantly inherited (familial hemiplegic migraine). Besides mutations in neuronal voltage-gated Ca 2ϩ and Na ϩ channels (FHM1 and FHM3), familial hemiplegic migraine has been associated with mutations in the ␣2-isoform of the Na ϩ ,K ϩ -ATPase (FHM2) (1). The Na ϩ and K ϩ gradients created across the cell membrane by the Na ϩ ,K ϩ -ATPase are of vital importance for cellular function and activities, including generation of action potentials and secondary active transport of ions, nutrients, and neurotransmitters. In the mammalian brain, three different isoforms of the catalytic ␣-subunit are expressed, ␣1, ␣2, and ␣3. The ␣2-isoform is mainly distributed in glial cells, whereas the ␣3-isoform is found in neurons and is absent in glial cells (2). So far, we know of more than 50 Na ϩ ,K ϩ -ATPase ␣2 mutations associated with hemiplegic migraine. It has been suggested that the pathophysiology results from an impaired clearance of extracellular K ϩ , producing a wide cortical depolarization (3,4). Moreover, the gradients of both Na ϩ and K ϩ are important for the reuptake of the excitatory transmitter glutamate from the synaptic cleft via the glutamate transporter, and impaired Na ϩ transport would also lead to a rise of intracellular Ca 2ϩ via the Na ϩ /Ca 2ϩ exchanger with secondary effects on Ca 2ϩ signaling (1,5,6). Cell survival studies of transfected cells, in which the wild type has been inhibited by ouabain, have demonstrated the inability of certain FHM2 mutants to sustain cell growth in accordance with the theory of haploinsufficiency, i.e. only the wild type allele encodes a functional enzyme (1,7). Other FHM2 mutants have nevertheless been found to retain transport function (4, 8 -11), and more functional mutants exist, as indicated by the present results. A crucial question is therefore whether and how their function deviates from that of the wild type. One study has reported a 2-fold reduced apparent K ϩ affinity for activation of the ATPase reaction of mutant T345A, which was suggested to give rise to a reduced rate of external K ϩ clearance in vivo (4).
To fully understand the functional implications of the FHM2 mutations and the structural basis, it is necessary to reveal the mutational effects on the individual partial reaction steps of the pump cycle and the interaction with Na ϩ and K ϩ and to try to relate the observed effects to known structural features. We present here the functional consequences of nine Na ϩ ,K ϩ -ATPase ␣2 mutations, including T345A, which all have been found in patients exhibiting the characteristic symptoms of hemiplegic migraine (12)(13)(14)(15)(16)(17). The ␣-subunit carrying the mutations consists of 10 transmembrane helices M1-M10 (harboring the Na ϩ -and K ϩ -binding sites) and a cytoplasmic "head" made up of three subdomains as follows: A ("actuator"), N ("nucleotide binding"), and P ("phosphorylation") (18,19). Generally, FHM2 mutations are widely scattered over the Na ϩ ,K ϩ -ATPase protein, but with a preference for the P-domain (7). Four of the mutations investigated here are located in the P-domain (R593W, V628M, E700K, and M731T), two are in the A-domain (R202Q and T263M), one is in M2 (V138A), another is in M4 near the interface to the P-domain (T345A), and the last is in the L6 -7 loop between M6 and M7 close to the P-domain (R834Q) (Fig. 1). We assessed their functional impact in a panel of analyses of the partial reaction steps (cf. Scheme 1), using both steady-state and transient kinetic measurements, the latter allowing determination of the rates of the rapid phosphorylation and dephosphorylation reactions. The observed effects are analyzed in the context of information derived from the recently determined crystal structures of the Na ϩ ,K ϩ -ATPase (18,19).

EXPERIMENTAL PROCEDURES
FHM2 mutations were introduced into full-length cDNA encoding an ouabain-resistant version of the human ␣2-isoform (4). Mutants and wild type were expressed in COS-1 cells under ouabain selection pressure (20). The plasma membrane fraction was isolated and made leaky to allow access of incubation media from both sides of the membrane, and Na ϩ ,K ϩ -ATPase activity was studied at 37°C by following the liberation of P i (20). The catalytic turnover rate was calculated by relating the ATPase activity to the active site concentration determined by phosphorylation at 0°C in the presence of 150 mM Na ϩ and oligomycin (21,22). Measurements of phosphorylation and dephosphorylation were performed using either a manual mixing technique at 0°C or a QFM-5 quench-flow module (Bio-Logic Instruments, Claix, France), allowing transient kinetic studies at 25°C (21,22). To eliminate the contribution of the endogenous Na ϩ ,K ϩ -ATPase, ouabain was included in the reaction media (21). Data processing was performed using SigmaPlot (SPSS, Inc.) for linear regression analysis (21), and the results are reported as average values Ϯ S.E. (shown by error bars in figures when larger than the size of the symbols). The number of independent determinations is indicated by n in Tables 1 and 2. Structural figures were prepared using PyMOL.

RESULTS
Overall Function-The nine FHM2 mutations were introduced into the ouabain-resistant version of the human ␣2-isoform, and the mutants were expressed in mammalian COS-1 cells and subjected to ouabain-selection pressure (20), taking advantage of the Ͼ100-fold difference between the ouabain affinities of the exogenous Na ϩ ,K ϩ -ATPase and the endogenous COS cell enzyme. Like the wild type, all mutants were able FIGURE 1. Na ؉ ,K ؉ -ATPase structure with indication of the FHM2 mutations studied. The structure shown has Protein Data Bank code 2ZXE (E 2 [K 2 ] with bound MgF 4 2Ϫ as phosphate analog (19)). Color codes for P-, A-, N-, and M-domain are gray, green, yellow, and cyan, respectively. The bound K ϩ ions are shown as purple spheres and MgF 4 2Ϫ (MgF) in the catalytic site as one green sphere with four cyan spheres. The mutated residues (numbered according to the human ␣2-isoform) are highlighted as sticks colored according to the elements (carbon, gray; oxygen, red; nitrogen, blue; sulfur, yellow). SCHEME 1. Model of the reaction cycle of Na ؉ ,K ؉ -ATPase ("Post-Albers scheme"). Ligand binding and dissociation steps and conformational changes between E 1 /E 1 P and E 2 /E 2 P forms are indicated. Na ϩ binds from the cytoplasmic side to the E 1 form and activates phosphorylation from ATP bound with high affinity to E 1 . The E 1 P form is ADP-sensitive, i.e. able to donate the phosphoryl group back to ADP, forming ATP. The E 2 P form is K ϩ -sensitive, i.e. hydrolysis of E 2 P is activated by binding of K ϩ from the extracellular side. Occluded ions are shown in brackets. Subscripts c and e indicate the cytoplasmic and extracellular sides, respectively. The ATP molecule binding with low affinity to E 2 is shown boxed. Because ATP binds with higher affinity to E 1 than to E 2 , an increase in the apparent affinity for ATP upon mutation reflects a shift of the E 1 -E 2 distribution toward the E 1 form, whereas a decrease in the affinity for ATP reflects a shift toward the E 2 form.
to confer ouabain resistance in growth medium containing 5 M ouabain, indicating that the rate of transport of Na ϩ and K ϩ is sufficiently high to sustain cell growth. Determination of the ouabain dependence of Na ϩ ,K ϩ -ATPase activity showed 2-9fold reduction of the apparent affinity for ouabain for mutants V138A, T263M, T345A, R593W, V628M, M731T, and R834Q, wild type-like apparent affinity for R202Q, and 2-fold increased apparent affinity for E700K (Table 1). Hence, all the functional studies described below were carried out in the presence of ouabain to selectively inhibit the endogenous Na ϩ ,K ϩ -ATPase (K 0.5 Յ 1 M).
The catalytic turnover rate determined in the presence of 130 mM Na ϩ , 20 mM K ϩ , and a saturating concentration of MgATP was found reduced in all mutants, most severely for R593W, V628M, M731T, and R834Q, displaying less than one-third the turnover rate of the wild type. T263M, T345A, E700K, and V138A displayed a turnover rate around 50%, whereas R202Q showed ϳ20% reduction (Table 1).
K ϩ Interaction-A critical issue is whether the affinity for external K ϩ is lowered, thereby resulting in reduced ability to clear K ϩ from the extracellular space. The interaction with K ϩ was investigated by determining the K ϩ dependence of ATPase activity (Fig. 2). In the wild type, K ϩ at a submillimolar concentration activates ATP hydrolysis by binding at extracellularly facing sites of the E 2 P form, thereby stimulating dephosphorylation (cf. Scheme 1, Reaction 6). The majority of the mutants displayed a K 0.5 value for K ϩ activation similar to or lower (i.e. corresponding to higher affinity) than that determined for the wild type, with E700K being the only exception, exhibiting a slight 1.6-fold increase in the K 0.5 value for K ϩ activation relative to wild type ( Fig. 2 and Table  1). K ϩ concentrations above 15 mM inhibited the wild type somewhat, which is explained by K ϩ binding in competition with Na ϩ at the cytoplasmically facing sites of the enzyme in E 1 conformation, leading to conversion of E 1 back to the K ϩ -occluded E 2 state (cf. Scheme 1, Reaction 1) (23,24). Mutants V138A, T345A, R593W, V628M, M731T, and R834Q exhibited a more distinct inhibitory phase than the wild type, whereas no inhibition was observed for E700K at the K ϩ concentrations studied (Fig. 2).  Table 1.  (24): . The values determined for K 1 are indicated in the table. b The Na ϩ ,K ϩ -ATPase activity was determined at 37°C in the presence of 30 mM histidine buffer (pH 7.4), 130 mM NaCl, 3 mM ATP, 3 mM MgCl 2 , 1 mM EGTA, ouabain to inhibit the endogenous enzyme, and 20 mM KCl. The catalytic turnover rate was calculated as the ratio between the Na ϩ ,K ϩ -ATPase activity and the active site concentration (maximum phosphorylation from ͓␥-32 P͔ATP measured at 0°C in the presence of 150 mM NaCl and oligomycin) (21,22). c K 0.5 values for K ϩ activation, vanadate inhibition, and ATP activation were extracted from the data in Figs. 2, 8 and 9, respectively.
Because the variable inhibitory phase in the K ϩ dependence of ATPase activity inevitably will influence the apparent affinity for K ϩ activation corresponding to the rising phase, the activation by extracellular K ϩ was also examined more directly by determining the rate of E 2 P dephosphorylation ( Fig. 3 and Table 2). At a nonsaturating K ϩ concentration of 1 mM (Fig. 3, open symbols), T345A, R593W, V628M, M731T, and R834Q displayed a dephosphorylation rate constant almost identical to that of the wild type, indicating that these mutations did not weaken the interaction with extracellular K ϩ . R202Q exhibited a higher rate constant of dephosphorylation than wild type. By contrast, the dephosphorylation rate constant was reduced for V138A and E700K relative to wild type (1.3-and 2.5-fold, respectively). To determine whether this was related to a reduction in affinity for external K ϩ , similar experiments were performed at a saturating K ϩ concentration of 20 mM (Fig. 3, filled symbols). The calculated ratios between the dephosphorylation rate constants at 1 and 20 mM K ϩ are 0.29 Ϯ 0.02, 0.42 Ϯ 0.04, and 0.32 Ϯ 0.02 for V138A, E700K, and wild type, respectively. Hence, at 1 mM K ϩ , the saturation of the external sites is similar to wild type for V138A and is actually higher for E700K, indicating that the reduced dephosphorylation rate is caused by a lower V max of phosphoenzyme hydrolysis rather than a lower affinity for K ϩ .
Na ϩ Interaction-Phosphorylation from ATP is triggered when three Na ϩ ions have bound to intracellularly facing high affinity sites of the E 1 form and become occluded in the E 1 [Na 3 ] form (Scheme 1, Reaction 2) (25). To assess the Na ϩ affinity of the E 1 form of the mutants, we studied the Na ϩ dependence of phosphorylation from ATP in the absence of K ϩ and presence of oligomycin to support occlusion (Fig. 4). The mutants displayed affinities similar to that obtained for the wild type; only R834Q exhibited a more than 2-fold reduced apparent affinity for Na ϩ relative to wild type.
Rate of Phosphorylation-Because oligomycin stabilizes the Na ϩ -occluded E 1 [Na 3 ] form, thereby promoting phosphorylation (25,26) and blocking the E 1 P-E 2 P transition (Scheme 1, Reaction 4), the presence of oligomycin enables the phosphoenzyme to build up a maximum level. Table 2 lists the level of phosphoenzyme accumulated at 150 mM NaCl in the absence of oligomycin relative to the maximum level obtained in the presence of oligomycin (EP/EP oligo ). Mutants R593W, V628M, M731T, and R834Q displayed a markedly reduced EP/EP oligo ratio, and V138A and T345A showed a slight reduction relative to wild type, whereas T263M and E700K displayed higher EP/EP oligo ratios than the wild type. Such effects can be due to changes in either the rate of phosphorylation or the rate of dephosphorylation. In the presence of oligomycin, the level of phosphoenzyme built up was sufficiently high to allow rapid kinetic studies of the rate of phosphorylation at a millisecond time scale at 25°C, and thus the time course of phosphorylation in the presence of 2 M ATP was determined (Fig. 5). Mutants V138A, T345A, R593W, V628M, M731T, and R834Q all displayed a reduced rate constant relative to wild type, most pronounced for M731T, suggesting that either the phosphorylation reaction or the binding of ATP is affected in these mutants. To determine the maximal rate of phosphorylation and the apparent affinity of these mutants for ATP we performed the analysis at varying ATP concentrations. E700K was included, because the dephosphorylation studies (Fig. 3) indicated that the catalytic site might be defective in this mutant. In Fig. 6, the results are depicted as double-reciprocal plots of the initial phosphorylation rate per ATPase molecule as a function of the concentration of ATP. The linear dependence allows extraction of the maximal phosphorylation rate per ATPase molecule, V max , and the apparent affinity for ATP (the Michaelis constant K m ). M731T exhibited a most severe 6-fold reduction of V max , relative to wild type. Mutants V138A, T345A, R593W, V628M, and R834Q showed 2-3-fold reduction, whereas E700K was wild type-like ( Table 2). None of the mutations increased the  Table 2 with statistics). The wild type data are represented in each panel by dotted red and blue lines. 6 and Table 2), likely reflecting the marked reduction of V max , which leads to increased accumulation of nonphosphorylated enzyme with ATP bound. According to the equation K m ϭ (k Ϫ1 ϩ k 2 )/k 1 , where k 1 and k Ϫ1 are the respective rate constants for binding and dissociation of ATP, and k 2 equals V max (measured in units of rate per ATPase molecule), the K m value will decrease with k 2 , unless k Ϫ1 is much higher than the k 2 value, which is not the case here, because Na ϩ ,K ϩ -ATPase binds the nucleotide rather tightly (27). In an analogous way, the accumulation of E 1 explains the increased apparent affinity for Na ϩ in M731T (Table 2). E 1 P-E 2 P Distribution-The phosphorylated form of the Na ϩ ,K ϩ -ATPase exists in E 1 P and E 2 P states that can be distinguished by their different sensitivities to K ϩ and ADP (Scheme 1). To estimate the extent of each phosphoenzyme pool present at steady state, the dephosphorylation time course was examined following addition of ADP to the phosphoenzyme. A biexponential function could be fitted to the data, permitting extraction of a rapid and a slow decay component, reflecting the initial amounts of E 1 P and E 2 P, respectively (Fig. 7). Table 2 lists the E 1 P fraction. Notably, mutant T263M displayed a marked increase in the level of E 1 P (to 96% versus 44% for the wild type). The other mutants displayed a less pronounced shift of the E 1 P-E 2 P distribution toward the E 1 P form or were wild type-like.
Vanadate and ATP Affinities-Vanadate is an inhibitor that acts as analog of the phosphoryl group in the transition state during E 2 P hydrolysis and binds exclusively to the enzyme in the E 2 conformation (28). The properties of the E 2 state were investigated by determining the apparent affinity for vanadate inhibition of ATP hydrolysis (Fig. 8). All the mutants exhibited a reduced apparent affinity for vanadate (Table 1). For R593W, V628M, E700K, M731T, and R834Q, the affinity was extremely low (Ն30-fold reduced relative to wild type); in fact, no significant inhibition was seen for E700K within the vanadate concentration range examined (Fig. 8, right panel). A reduced sensitivity to vanadate inhibition can in principle arise from either a lowering of the intrinsic binding affinity of E 2 (possibly reflecting destabilization of the transition state of E 2 P hydrolysis) or from a shift of the E 1 -E 2 distribution away from the vanadate binding E 2 form in favor of E 1 . To examine the mutational effects on the E 1 -E 2 distribution, the ATP dependence of ATPase activity was also studied (Fig. 9). Because ATP binds with higher affinity to E 1 than to E 2 , an increase in the apparent affinity for ATP upon mutation reflects a shift of the E 1 -E 2 distribution toward the E 1 form, whereas a decrease in the affinity for ATP reflects a shift toward the E 2 form. The results summarized in Table 1 show a significant increase of the apparent ATP affinity for all the mutants (2-5-fold), indicating a shift toward the E 1 form. However, for R593W, V628M, E700K, M731T, and R834Q, the effect on ATP affinity was much less spectacular than the reduction of vanadate affinity, indicating that the latter arises not only from a shift in E 1 -E 2 distribution but also from a change in the intrinsic affinity of the E 2 form for vanadate.

DISCUSSION
The nine FHM2 mutations studied here, which are found in various regions of the Na ϩ ,K ϩ -ATPase ␣2-isoform, lead to a functionally altered but active enzyme capable of sustaining cell viability. Our results do not lend support to the hypothesis that (n ϭ 5) (n ϭ 3) (n ϭ 9) (n ϭ 9) (n ϭ 6) (n ϭ 6) a Extracted from the data in Fig. 4. b Ratio between phosphorylation levels without (EP) and with oligomycin (EP oligo ). Phosphorylation was carried out for 10 s at 0°C in 20 mM Tris (pH 7.5), 3 mM MgCl 2 , 2 M ͓␥-32 P͔ATP, 10 M ouabain, 150 mM NaCl, without and with 20 g of oligomycin/ml. c Extracted from the data in Figs. 5 and 6. The kinetic parameters were not determined (ND) for R202Q and T263M, because the phosphorylation rate at 2 M ATP was close to wild type. d Extracted from the data in Fig. 7. e Extracted from the data in Fig. 3. For T263M, the E 2 P dephosphorylation rate constant was not determined (ND), because the high extent of accumulation of E 1 P, even at 25°C, precluded such a determination. The dephosphorylation rate at 20 mM K ϩ was determined only when the dephosphorylation at 1 mM K ϩ was slower than that of the wild type. a reduced affinity for external K ϩ , causing a selective disturbance of K ϩ clearance, is an obligatory part of the mechanism underlying the disease (4). Neither can a defective Na ϩ interaction generally account for the pathophysiology of the disease, as only R834Q exhibited a significant reduction of the Na ϩ affinity. This is contrary to the neurological disorder rapid-onset dystonia parkinsonism, caused by mutation of the ␣3-isoform of Na ϩ ,K ϩ -ATPase, where all mutants characterized so far display markedly reduced Na ϩ affinity (21, 29, 30). At saturating Na ϩ and K ϩ concentrations, the FHM2 mutants studied here   Table 2. Oligomycin was added to optimize phosphorylation and prevent dephosphorylation, thereby minimizing effects on apparent Na ϩ affinity of variation of the phosphorylation and dephosphorylation rates. Each line represents the best fit of the equation EP ϭ EP max ⅐[Na ϩ ] n /(K 0.5 n ϩ [Na ϩ ] n ), giving the indicated K 0.5 values (also listed in Table 2 with statistics). The data corresponding to wild type are represented in each panel by a dotted blue line. display a reduced catalytic turnover rate relative to wild type. A reduced turnover rate has previously been reported for mutants T263M, T345A, M731T, and R834Q, but the responsible change(s) in partial reaction step(s) were not elucidated (4,9,10). Our rapid kinetic studies of the phosphorylation from ATP identify a reduced V max of phosphorylation as a major factor contributing to the reduction of the catalytic turnover rate of mutants V138A, T345A, R593W, V628M, M731T, and R834Q. These mutants displayed decreases in the V max of phosphory-lation of 2-6-fold ( Fig. 6 and Table 2). To be able to carry out transient kinetic measurements, oligomycin was added to promote phosphorylation by stabilizing the Na ϩ -occluded E 1 [Na 3 ] form and blocking the E 1 P[Na 3 ] 3 E 2 P transition. In the absence of oligomycin, the relative level of phosphoenzyme at steady state was in fact very low for R593W, V628M, M731T, and R834Q (Table 2, "EP/EP oligo "), suggesting that under physiological conditions without oligomycin the phosphorylation rate is even more dramatically affected than judged from the kinetic data obtained with oligomycin.
The low phosphorylation rate also accounts for the enhanced K ϩ inhibition of ATPase activity at high K ϩ concentrations seen particularly for R593W, V628M, M731T, and R834Q (Fig.  2). Hence, the low phosphorylation rate leads to increased availability of E 1 at steady state and consequently to enhanced K ϩ competition with Na ϩ at the intracellular E 1 sites. In the living cell with cytoplasmic concentrations of ϳ150 mM K ϩ and only ϳ10 mM Na ϩ , the enhanced K ϩ competition with Na ϩ may be highly relevant as a factor contributing to compromise pump function.
For R593W, V628M, E700K, M731T, and R834Q, the apparent affinity for vanadate was lowered to an extent that could not be accounted for by the observed shift of the conformational equilibrium toward E 1 (Fig. 8 and Table 1). Because vanadate binds at the phosphorylation site of the E 2 form as an analog of the phosphoryl group in the transition state between E 2 and E 2 P, the extraordinary low affinity for vanadate might be related to a perturbation of the phosphorylation site in E 2 and/or the transition state. For the mutants showing both a reduced V max of phosphorylation from ATP in E 1 and an extraordinary low affinity for vanadate (R593W, V628M, M731T, and R834Q), the underlying structural changes in E 1 and E 2 forms might be similar. By contrast E700K did not show a reduced phosphorylation rate with ATP, and for this mutant the phosphorylation site therefore seems not to be perturbed in E 1 . Besides its extraordinary lack of sensitivity to vanadate, E700K was characterized by a relatively low maximal rate of dephosphorylation of E 2 P (42 s Ϫ1 versus 132 s Ϫ1 for wild type), which probably causes the reduced catalytic turnover rate of E700K. Both the disruption of vanadate binding and the reduced dephosphorylation rate can be explained by assuming that the transition state of E 2 P dephosphorylation is destabilized. It is furthermore of note that E700K exhibited an increased apparent affinity for ouabain, and even though E700K did not show reduced affinity for K ϩ activation of dephosphorylation ( Fig. 3 and Table 2), a slight 1.6fold reduction of apparent K ϩ affinity in activation of ATPase activity was noted ( Fig. 2 and Table 1). Both of these effects might result from accumulation at steady state of the E 2 P[Na 2 ] phosphoenzyme form, which is an intermediate in the E 1 P[Na 3 ] 3 E 2 P transition (31) exhibiting a particularly high affinity for cardiotonic steroids ("E*P" cf. Refs 32 and 33). Thus, the evidence indicates that mutation E700K is most disturbing in the part of the pump cycle involving E 2 P and E 2 P-like states.
From the crystal structure it is known how the residues studied here are positioned in the E 2 [K 2 ] state with bound MgF 4 2Ϫ as a phosphate analog (18,19), and although crystal structures are static snapshots without the flexibility of the native enzyme, they are useful as a basis for analysis of the mechanisms under- FIGURE 7. Distribution of the phosphoenzyme between E 1 P and E 2 P. Phosphorylation was carried out for 10 s at 0°C in the presence of 2 M [␥-32 P]ATP, 20 mM NaCl, 130 mM choline chloride, 20 mM Tris (pH 7.5), 3 mM MgCl 2 , 1 mM EGTA, and 10 M ouabain. Dephosphorylation was followed by addition of a chase solution producing final concentrations of 2.5 mM ADP and 1 mM unlabeled ATP. The dephosphorylation reaction was terminated by acid quenching after the indicated time intervals. A bi-exponential decay function was fitted to the data. The rate constant corresponding to the rapid phase, reflecting the ADP reaction with E 1 P, was set to 2 s Ϫ1 . The fraction of E 1 P phosphoenzyme obtained from this analysis is indicated in the panel (also listed in Table 2 with statistics). The wild type data are represented in each panel by a dotted line.
lying the observed mutational effects. Arg 593 , Val 628 , Glu 700 , and Met 731 are located in the Rossmann fold, which is a central feature of the P-domain, consisting of a seven-stranded parallel ␤-sheet associated with seven short ␣-helices (P1-P7) as indicated in Fig. 10. Asp 374 (phosphorylation site) and residues involved in Mg 2ϩ binding, Thr 376 , Asp 714 , and Asp 718 , are located centrally in the Rossmann fold (Fig. 10). The phosphorylation of Asp 374 is triggered by an approach between the Nand P-domains associated with a bending of the P-domain. This conformational change is the result of transmission of the effect of Mg 2ϩ binding to the P7 helix of the Rossmann fold (34,35), meaning that the loop between ␤6 and P7, where Met 731 is located (Figs. 10 and 11), has to be strained when Mg 2ϩ binds to Asp 714 . The role of Met 731 may thus be to reduce the freedom of movement of this loop, thereby ensuring the strain. In fact the side chain of Met 731 is flanked on one side by Asp 714 , Thr 378 , and Arg 593 (Fig. 10B), and the introduction of the threonine side chain in M731T likely pushes Asp 714 , thus preventing a proper bending of the P-domain. It is interesting to note that M731T and R593W are the mutations giving the largest reduction of phosphorylation rate observed in this study (cf. Table 2). Considering that the Met 731 and Arg 593 side chains are located within a distance of only 4 -5 Å from each other in the crystal structure (Fig. 10, A and B), it is possible that the bulky tryptophan in R593W actually interferes by disturbing Met 731 . Furthermore, Arg 593 is within hydrogen bonding distance of the backbone carbonyls of Gly 377 and Thr 378 , which are located in the same loop as Asp 374 and Thr 376 (Fig. 10, A and B). The clash resulting from insertion of the tryptophan might therefore also disturb phosphorylation by shifting the positions of Asp 374 and Thr 376 . It is noteworthy that mutations T376M and T378N have been found in familial hemiplegic migraine patients and were reported as loss of function mutations based on cell survival studies (14, 36 -38).
The valine replaced in mutant V628M is positioned at the end of the P3 helix in the Rossmann fold. It is part of a hydrophobic/van der Waals interaction network (Fig. 10, A and C), and the interaction with Pro 592 and Val 596 at the N-terminal end of the P2 helix may contribute to stabilize the positions of helices P2 and P3. The V628M mutation might lead to a shift of the position of Pro 592 , thereby disturbing the interactions of the adjacent Arg 593 with Gly 377 and Thr 378 mentioned above.
The glutamate replaced in E700K is located in the P5 helix of the Rossmann fold. In the crystal structure of the Na ϩ ,K ϩ -ATPase E 2 [K 2 ] state with bound MgF 4 2Ϫ , Glu 700 is positioned between two positively charged residues, Arg 704 (at the end of the P5 helix) and Lys 724 (P6 helix), with a distance of 4.6 Å to each (Fig. 10A). Moreover, Lys 724 is within hydrogen bonding distance to Gln 703 in the P5 helix (2.7 Å). It can be imagined that the extra positive charge introduced between Arg 704 and Lys 724 by E700K leads to a repulsion between the three positive charges disturbing the positioning of P5 and P6. Such a disturbance could lead to a shift of ␤5 and the loop between ␤5 and P6 containing Asp 714 and Asp 718 critical to Mg 2ϩ binding (Fig.  10A). We wondered whether Glu 700 and either Lys 724 or Arg 704 might actually be closer to each other in the vanadate-bound E 2 state than the 4.6 Å in the E 2 [K 2 ] crystal structure. The Na ϩ ,K ϩ -ATPase residues Glu 700 , Lys 724 , and Gln 703 are conserved as Glu 689 , Lys 713 , and Gln 692 in the Ca 2ϩ -ATPase, which has been crystallized not only in the MgF 4 2Ϫ -bound E 2 form  Table 1.  Table 1.
similar to the Na ϩ ,K ϩ -ATPase crystal form but also in several other E 2 states, including the AlF 4 Ϫ -bound form. The latter is considered the state closest to the E 2 form with vanadate bound, as AlF 4 Ϫ and vanadate are both believed to mimic the trigonal bipyramidal structure of the penta-coordinated phosphate in the transition state of E 2 P dephosphorylation (39,40). Interestingly, among the AlF 4 Ϫ -bound E 2 crystal structures of the Ca 2ϩ -ATPase, 4 out of 7 have a distance between Glu 689 and Lys 713 Ͻ4 Å, i.e. short enough to indicate the presence of a salt link between these residues in the E 2 P transition state. In contrast, in the MgF 4 2Ϫ -bound Ca 2ϩ -ATPase structures, the corresponding distance is 4 -5 Å, consistent with the distance in the MgF 4 2Ϫ -bound form of Na ϩ ,K ϩ -ATPase (cf. supplemental Table S1). If this scenario for the E 2 P transition state were extrapolated to the Na ϩ ,K ϩ -ATPase, it would explain the very strong destabilization of the vanadate-bound state of E700K (Fig. 8). If in addition Glu 700 and Lys 724 were further apart in the E 1 [Na 3 ] state, the lack of effect of the mutation on the phosphorylation from ATP would be understandable. The Ca 2ϩ -ATPase E 1 structures do not provide a clear answer to this question, but it is noteworthy that 4 out of 8 E 1 structures show a distance larger than 4 Å (cf. supplemental Table S1).
T345A in the cytoplasmic extension of M4 and R834Q in the L6 -7 loop are both located at the boundary between the P-domain and the transmembrane region, and the reason that these mutations also reduce the phosphorylation rate significantly may be the participation of Thr 345 and Arg 834 in interaction networks involving helices connected with the Rossmann fold (Fig. 11). Thr 345 might participate in van der Waals interactions with Val 362 at the N-terminal end of the P-domain helix P1, from which the central ␤-strand of the Rossmann fold (␤-strand "1" in Fig. 11) leads to the phosphorylation site with Asp 374 . Interestingly, mutation V362E has also been found in patients with FHM2 (41). Depending on the actual rotational state of the Thr 345 side chain, the hydroxyl group might form a hydrogen bond with Glu 761 of M5. Mutation R834Q seems to disrupt bonds from P1-helix residues Glu 363 and Ser 367 (C-terminal end of the P1-helix) to the L6 -7 loop (19). R834Q furthermore disrupts the bond between Arg 834 and Glu 285 of M3 (Fig. 11), which may explain the reduction of Na ϩ affinity caused by R834Q, since mutation of Glu 285 has been shown to reduce Na ϩ affinity, possibly a consequence of the involvement of Glu 285 in control of the cytoplasmic entrance pathway for Na ϩ (22). Surprisingly, the mutation V138A also had significant impact on the phosphorylation rate, even though Val 138 is positioned in the transmembrane segment M2 far from the phosphorylation site, thus indicating a long range effect, which might be exerted through interference with the hydrophobic/van der Waals interactions between M1 and M2 (cf. Fig. 12A). These interactions seem to allow M1 to close the Na ϩ binding pocket through contact with the ion binding Glu 332 of M4 (42), a conformational change believed to be propagated to the phosphorylation site in a yet undefined way and result in phosphoryl transfer (35). Indeed, a reduced rate of phosphorylation was previously observed following replacement of M1 residues Leu 94 and Gly 97 (42,43).
For R202Q and T263M, the phosphorylation rate was wild type-like, and the reason for the reduced catalytic turnover rate seems to be a slow conversion of E 1 P to E 2 P, in particular T263M showed a conspicuous accumulation of E 1 P. Furthermore, both R202Q and T263M appeared to shift the E 1 -E 2 distribution of the dephosphoenzyme in favor of E 1 . These mutations are both associated with the A-domain, which undergoes drastic structural rearrangements during the E 1 -E 2 and E 1 P-E 2 P transitions (35,44). Arg 202 is positioned in a ␤-strand of the A-domain, where it appears to form a hydrogen bond with the backbone carbonyl oxygen of Pro 227 (2.7 Å distance, Fig. 12B). This bond could be important for stabilization of the loop con-taining the conserved TGES motif, and hence for interaction of TGES with the P-domain in E 2 and E 2 P. Thr 263 is located in the AM3 linker segment connecting the A-domain to M3 and appears to be involved in stabilization of the kinked ␣-helix present in the AM3 linker in E 2 , but not in E 1 (see Fig. 12, C and D). Hence, in E 2 the side chain hydroxyl group of the threonine seems to form a hydrogen bond with the backbone amide nitrogen of Gly 266 (2.8 Å distance). This structural arrangement is likely destabilized by the T263M substitution due to the bulkiness of the methionine side chain. The ability of the E 1 conformation to better accommodate the methionine, due to a looser structure compared with the helical arrangement in E 2 (Fig.  12D), explains that the conformational equilibrium is shifted toward E 1 in the mutant. In line with this interpretation, mutation of Gly 266 to alanine, disturbing the kinked helix arrangement, was previously shown to cause accumulation of E 1 (45).
To summarize, if disturbance of K ϩ clearance by glial cells is the reason for development of FHM2, this disturbance must be attributed to a low maximum turnover rate of the Na ϩ ,K ϩ -ATPase and not to a reduced affinity for external K ϩ . In several of the FHM2 mutants studied here, the function of the catalytic FIGURE 11. Side view of the membrane interface of the P-domain. The structure used is the same as in Fig. 1 with the same color codes for P-, A-, N-, and M-domain, and the phosphate analog MgF 4 2Ϫ (MgF) shown as one green and four cyan spheres. Side chains corresponding to mutated residues and interaction partners (numbered according to the human ␣2-isoform) are highlighted as sticks, as are key residues in the catalytic site. Atoms are colored according to the elements (carbon, gray; oxygen, red; nitrogen, blue; sulfur, yellow). P-domain helices P1-P3 and P5-P7 as well as the interposed ␤-strands (yellow numbers) and transmembrane helices M3-M5 are indicated. Note the location of the P1 helix with Val 362 interacting with Thr 345 , and Glu 363 and Ser 367 interacting with Arg 834 . Glu 761 (M5) may also interact with Thr 345 , depending on the actual rotational state of Thr 345 . In addition, Glu 285 (M3) interacts with Arg 834 . FIGURE 12. Structural relationships of residues altered in mutants V138A, R202Q, and T263M. The structure used in A-C is the same as in Fig. 1 with the same color codes for P-, A-, N-, and M-domain. Side chains corresponding to mutated residues and interaction partners (numbered according to the human ␣2-isoform) are highlighted as sticks. Atoms are colored according to the elements (carbon, gray; oxygen, red; nitrogen, blue). A, Val 138 in M2 interacts with M1 residues Leu 94 and Trp 103 . B, interaction between Arg 202 and Pro 227 may stabilize the loop containing the conserved TGES motif with Glu 219 . C, Thr 263 is located in the AM3 linker segment connecting the A-domain to the M3 transmembrane segment (cf. Fig. 1). In the E 2 and E 2 P conformations, part of the AM3 linker is coiled up, forming an ␣-helical arrangement, which is partially unwound at Thr 263 , with the two parts of the ␣-helix being almost orthogonal to each other. The side chain hydroxyl group of the threonine is within hydrogen bonding distance to the backbone amide nitrogen of Gly 266 . D, homology model of the Na ϩ ,K ϩ -ATPase in E 1 based on the structure of the closely related Ca 2ϩ -ATPase in E 1 (Protein Data Bank code 1T5T). The helical structure of the AM3 linker is much more unwound than in E 2 conformations of Na ϩ ,K ϩ -ATPase and Ca 2ϩ -ATPase, with less strict requirement for the bond between Thr 263 and Gly 266 . site in phosphorylation (V138A, T345A, R593W, V628M, M731T, and R834Q) or dephosphorylation (E700K) is affected, involving local effects on the catalytic assembly as well as helices connected with the Rossmann fold, and long range effects transmitted from as far away as the membrane domain. The last two mutations (R202Q and T263M) affect the maximum turnover rate by destabilizing the A-domain in the E 2 /E 2 P conformations.