Diverse Functional Consequences of Mutations in the Na+/K+-ATPase α2-Subunit Causing Familial Hemiplegic Migraine Type 2*

Mutations in ATP1A2, the gene coding for the Na+/K+-ATPase α2-subunit, are associated with both familial hemiplegic migraine and sporadic cases of hemiplegic migraine. In this study, we examined the functional properties of 11 ATP1A2 mutations associated with familial or sporadic hemiplegic migraine, including missense mutations (T263M, T376M, R383H, A606T, R763H, M829R, R834Q, R937P, and X1021R), a deletion mutant (del(K935-S940)ins(I)), and a frameshift mutation (S966fs). According to the Na+/K+-ATPase crystal structure, a subset of the mutated residues (Ala606, Arg763, Met829, and Arg834) is involved in important interdomain H-bond networks, and the C terminus of the enzyme, which is elongated by the X1021R mutation, has been implicated in voltage dependence and formation of a third Na+-binding site. Upon heterologous expression in Xenopus oocytes, the analysis of electrogenic transport properties, Rb+ uptake, and protein expression revealed pronounced and markedly diverse functional alterations in all ATP1A2 mutants. Abnormalities included a complete loss of function (T376M), impaired plasma membrane expression (del(K935-S940)ins(I) and S966fs), and altered apparent affinities for extracellular cations or reduced enzyme turnover (R383H, A606T, R763H, R834Q, and X1021R). In addition, changes in the voltage dependence of pump currents and the increased rate constants of the voltage jump-induced redistribution between E1P and E2P states were observed. Thus, mutations that disrupt distinct interdomain H-bond patterns can cause abnormal conformational flexibility and exert long range consequences on apparent cation affinities or voltage dependence. Of interest, the X1021R mutation severely impaired voltage dependence and kinetics of Na+-translocating partial reactions, corroborating the critical role of the C terminus of Na+/K+-ATPase in these processes.

Familial hemiplegic migraine (FHM) 3 is a rare autosomal dominant variant of migraine with aura. It is clinically characterized by the presence of transient motor weakness in addition to other cortical symptoms as part of the migraine aura. FHM has been associated with mutations in three different genes. The CACNA1A gene (FHM1) (1,2) and SCN1A gene (FHM3) (3,4) both encode neuronal voltage-gated ion channels, whereas the ATP1A2 gene (FHM2) encodes the ␣ 2 -subunit of the Na ϩ /K ϩ -ATPase (5). These observations suggest a key role for cation translocation in the pathophysiology of FHM and possibly in migraine in general (6).
The Na ϩ /K ϩ -ATPase utilizes the free energy of ATP hydrolysis to export three Na ϩ ions and import two K ϩ ions per ATP molecule according to the Post-Albers scheme (Fig. 1A) (7,8). It generates net current and maintains the transmembrane gradients of Na ϩ and K ϩ (9, 10), which are critical for numerous secondary active transporters or ion channels. Functional Na ϩ /K ϩ -ATPase consists of a large catalytic ␣-subunit and a smaller accessory ␤-subunit. Both subunit isoforms are expressed in a tissue-specific and developmentally specific manner (11). In the adult brain, the ␣ 2 -subunit is mainly located in glial cells (12)(13)(14)(15).
To date, 33 different alleles of ATP1A2 have been associated with FHM, and nine have been associated with SHM (5, 16 -27). Only five of them have been functionally examined. L764P and W887R led to nonfunctional proteins (5,28), whereas others resulted in partially active enzymes with decreased (T345M) or increased (R689Q and M731T) K ϩ affinities or reduced turnover rates (R689Q and M731T) (29,30). These findings suggested that broad experimental approaches are required to cover the diversity of functional disruptions caused by FHM2 mutations.
In this study we analyzed a set of 11 ATP1A2 alleles associated with familial or sporadic hemiplegic migraine. All of the mutations are located in highly conserved regions that are either core motifs of all P-type ATPases or conserved among all P 2 -type ATPases (31). The mutations were selected because of their location in distinct regions within the three-dimensional structure (Fig. 1B); T263M is one of three FHM2 mutations in the actuator (A) domain of the enzyme. T376M affects the core sequence SDKTGT and is two amino acids apart from Asp 374 , which forms a phosphorylated intermediate during the reaction cycle. R383H alters a residue within a hinge region between the nucleotide-binding (N) and the phosphorylation (P) domains. The remaining point mutations and del(K935-S940)ins(I) localize to a region, which is critical for interdomain interactions during the catalytic cycle and might also interfere with intracellular access of cations to their binding sites. The Ser 966 frameshift mutation adds 32 amino acids to the C-terminal end of helix M9 and therefore most likely disrupts helix M10. The X1021R mutation results in the addition of 28 amino acids to the C terminus of the enzyme, which according to the recently published crystal structure of the Na ϩ /K ϩ -ATPase represents an important regulatory element to optimize Na ϩ binding at a putative third binding site (32).
To assess the function of the various ATP1A2 alleles, we used the two-electrode voltage-clamp technique to analyze stationary and pre-steady state currents and measured Rb ϩ uptake and plasma membrane expression in Xenopus oocytes. Our findings reveal a broad spectrum of abnormalities, frequently involving changes in the voltage-dependent properties of the enzyme. Mutations that disrupt critical interdomain H-bond patterns, although at distance from the cation-binding pocket, exhibited long range consequences on cation affinities. The expansion of the C terminus of the enzyme by the X1021R mutation had a dramatic effect on the interaction with Na ϩ , supporting the critical role of this highly conserved structural element in voltage sensing and Na ϩ transport properties.

EXPERIMENTAL PROCEDURES
cDNA Constructs-Human Na ϩ /K ϩ -ATPase ␣ 2 -and ␤ 1 -subunit cDNAs were subcloned into a modified pCDNA3.1 vector, as described (28). Site-directed mutagenesis was carried out by recombinant PCR. All PCR-derived fragments were verified by sequencing (Eurofins MWG Operon, Ebersberg, Germany). To distinguish the activity of the expressed from the endogenous Na ϩ /K ϩ -ATPase, mutations Q116R and N127D were introduced to obtain an ouabain-resistant protein (IC 50 in millimolar range) (33). The ouabain-resistant wild type ␣ 2 -subunit is referred to as WT, and the respective mutants are labeled by their amino acid substitutions, e.g. R383H. Nucleotide sequences and nomenclature correspond to those reported in Refs. 5, 17, and 34).
cRNA Synthesis and Oocyte Treatment-cRNA synthesis was carried out with the T7 mMessage mMachine kit (Ambion, Austin, TX). The oocytes were obtained by partial ovarectomy from anesthetized Xenopus laevis females, followed by treatment with collagenase 1A (Sigma). Each oocyte was injected with 25 ng of ␣ 2 -subunit and 2.5 ng of ␤ 1 -subunit cRNAs and stored in ORI buffer (110 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 5 mM HEPES, pH 7.4) containing 50 mg/liter gentamycin at 18°C for 3-4 days. Preceding the experiments, intracellular [Na ϩ ] was elevated by 45 min of incubation in Na ϩ loading solution (110 mM NaCl, 2.5 mM sodium citrate, 5 mM MOPS, 5 mM Tris, pH 7.4) followed by an incubation of at least 30 min in Na buffer (100 mM NaCl, 1 mM CaCl 2 , 5 mM BaCl 2 , 5 mM NiCl 2 , and 2.5 mM MOPS, 2.5 mM Tris, pH 7.4). FIGURE 1. Mechanism and structure of the Na ؉ /K ؉ -ATPase. A, reaction cycle of the Na ϩ /K ϩ -ATPase. Yellow, reaction sequence studied by voltage pulses under high Na ϩ /K ϩ -free conditions. Red arrow, voltage-dependent reverse binding of extracellular Na ϩ . Blue arrow, conformational change E 1 P-E 2 P, which is rate-limiting for the forward reaction. B, structure of the Na ϩ /K ϩ -ATPase according to Protein Data Bank structure entry 3B8E with amino acids numbered according to human ATP1A2. Arg 763 resides within the M5 helix, Thr 263 resides in the A domain, Thr 376 is two residues downstream the phosphorylation site Asp 374 , and Arg 383 localizes to one of two "hinge" loops, which link the P and the N domains. Ala 606 is located at the tip of a short loop between an ␣-helix and a ␤-sheet in the P domain, close to Arg 763 and Met 829 in the L67 loop (see Fig. 6 for structural details). L67 also contains Arg 834 . Arg 937 resides in the M8-M9 loop. X1021R generates an extension of the C terminus (red ball) by 28 amino acids (RPHWKKNQAWKDGELWRCCGDG-DGEGWK), which is not depicted here.

ATP1A2 Mutations in Familial Hemiplegic Migraine
Electrophysiology-Currents were recorded using a Turbotec 10CX amplifier (NPI Instruments, Tamm, Germany) and pClamp 7 software (Axon Instruments, Union City, CA) at room temperature (21-23°C). The solutions used were: Na buffer (see above), K buffer (100 mM KCl, 5 mM BaCl 2 , 5 mM NiCl 2 , 1 mM CaCl 2 , 2.5 mM MOPS, 2.5 mM Tris, pH 7.4). Sodium-rich buffers containing distinct K ϩ concentrations were prepared by mixing appropriate amounts of Na buffer and K buffer (resulting in (100 Ϫ x) mM [Na ϩ ] and x mM [K ϩ ], where x is a number between 0 and 10). The [K ϩ ] ext -dependent currents were calculated by subtracting currents measured in K ϩ -free Na buffer from the currents measured in presence of a distinct K ϩ concentration. Current amplitudes were corrected for rundown, if necessary. All of the buffers contained 10 M ouabain to inhibit the endogenous Na ϩ /K ϩ -ATPase.
To determine the voltage dependence of stationary currents, the cells were subjected to voltage pulses. Starting from Ϫ30-mV holding potential, the cells were clamped to test potentials between Ϫ180 and ϩ60 mV (in ϩ20-mV increments), followed by a step back to Ϫ30 mV. This protocol was applied first (a) in Na buffer, then (b) during perfusion with Na-based buffers containing certain [K ϩ ], and (c) after washout in Na buffer again. If currents measured during periods a and c did not differ, the stationary currents were calculated from the difference between corresponding currents measured in periods b and c. All of the currents within one experiment were normalized to the amplitude at 10 mM K ϩ and 0 mV. Voltagedependent K 0.5 values for the stimulation of stationary pump currents by extracellular K ϩ were determined using fits of the following Hill equation, to the normalized currents at a given membrane potential (K 0.5 is the concentration of half-maximal activity, and n H is the Hill coefficient). The Hill parameters from the fits were between 1 and 1.5.
Analysis of Transient Currents-The major electrogenic event within the catalytic cycle is the extracellular release or reverse binding of Na ϩ . At high Na ϩ and extracellularly K ϩ -free conditions, dephosphorylation is slow, and the enzyme is limited to carrying out electrogenic Na ϩ /Na ϩ exchange (35). Voltage pulses can then shift the enzyme between E 1 P and E 2 P states (Fig. 1A), resulting in transient currents, which yield kinetic information about extracellular Na ϩ binding/release and the voltage dependence of the distribution between E 1 P and E 2 P states. Pre-steady state currents under Na ϩ /Na ϩ exchange conditions were measured as the difference between currents measured first in Na buffer and then in Na buffer containing 10 mM ouabain. The data were fitted monoexponentially disregarding the first 3-5 ms to exclude capacitive artifacts. The charge Q transported during a transient current was determined as the integral of fitted currents, extrapolated to the time of onset of voltage pulses. The resulting Q-V curves were fitted with a Boltzmann function, where Q max and Q min are the saturation values of Q(V), V 0.5 is the half-maximal voltage, z q is the fractional charge, F is the Faraday constant, R is the molar gas constant, T is the temperature, and V is the transmembrane potential. A positively shifted V 0.5 value of a mutant compared with WT is indicative of an increased apparent affinity for extracellular Na ϩ , because less negative potentials are needed to drive Na ϩ movement in conjunction with the E 2 P-E 1 P conformational change. An equivalent interpretation is an apparent stabilization of the E 1 P (or destabilization of the E 2 P) conformational state. Turnover Numbers-The turnover number was calculated by dividing the stationary current (at 10 mM K ϩ and Ϫ30 mV) by the total moved charge Q tot ϭ Q max Ϫ Q min (from fits of Q-V curves with a Boltzmann function), both measured on the same oocyte. For the ATP1A2 WT-expressing cell in Fig. 4D, the stationary current and Q tot were 286 nA and 22.6 nC, respectively, resulting in a turnover of 12.7 s Ϫ1 .
Rb ϩ Uptake Measurements-Uptake of Rb ϩ into oocytes was measured by atomic absorption spectroscopy using an AAnalyst TM 800 spectrometer (PerkinElmer Life Sciences) equipped with a Rubidium hollow cathode lamp (Photron, Melbourne, Australia) and a transversely heated graphite furnace. After Na ϩ loading, the oocytes were incubated for 3 min at 21-22°C in Na buffer ϩ 1 mM RbCl containing either 10 M or 10 mM ouabain. After three washes in Rb ϩ -free Na buffer and one in distilled H 2 O, individual cells were homogenized in 1 ml of distilled H 2 O. From these homogenates, 20-l samples were injected into the graphite furnace for analysis.
Isolation of Membrane Fractions from Oocytes-Plasma membranes were prepared as described (36), with some modifications; 8 -12 oocytes were rotated for 30 min at 4°C in Eppendorf cups filled with MBSS buffer (20 mM MES, 80 mM NaCl, pH 6.0) ϩ 1% colloidal silica (Ludox Cl, Sigma), followed by two washes with MBSS. Subsequently, the cells were rotated for 30 min at 4°C in MBSS ϩ 0.1% polyacrylic acid (Sigma) and then washed twice with MBS buffer (88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO 3 , 10 mM HEPES, 0.82 mM MgSO 4 , 0.33 mM Ca(NO 3 ) 2 , 0.41 mM CaCl 2 , pH 7.4). After careful removal of MBS, the oocytes were homogenized in 1500 l of HbA solution (20 mM Tris, 5 mM MgCl 2 , 5 mM NaH 2 PO 4 , 1 mM EDTA, 80 mM sucrose, pH 7.4) containing Complete protease inhibitor (Roche Applied Science) and centrifuged for 30 s at 10 ϫ g (4°C). The top 1300 l of these samples, which contained total cellular membranes, were removed and subjected to a total membrane preparation (see below). With the remaining 200 l, which contained the silica beads with bound plasma membranes, the addition of 1000 l of HbA, centrifugation, and removal of supernatant were repeated three times (at 10 ϫ g, 20 ϫ g, and 40 ϫ g). After the last low speed centrifugation, plasma membranes were pelleted by centrifugation for 30 min at 16,000 ϫ g (4°C).
For the isolation of total membranes, the supernatant from the first centrifugation step (see previous paragraph) was centrifuged for 3 min at 2,000 ϫ g and 4°C to remove yolk platelets.
Subsequently, the membranes were isolated by centrifugation for 30 min at 16,000 ϫ g (4°C).
All of the samples were solubilized in SDS-PAGE sample buffer (4 l/oocyte). The aliquots corresponding to the content of one or two oocytes were separated by 10% SDS-PAGE and blotted on polyvinylidene difluoride membranes (Milipore, Bedford, MA). The ␣ 2 -subunits of Na ϩ /K ϩ -ATPase were detected with the specific polyclonal antibody AB9094 (Chemicon, Temecula, CA). Structural
To investigate whether the differences in transport activity were due to different expression levels or plasma membrane targeting, we carried out Western blots of simultaneously prepared plasma membrane and total membrane protein fractions. As shown in Fig. 2C total protein expression of mutant alleles was not different from WT, which also suggests intact ␣-␤ subunit interactions during protein folding and membrane insertion (39). However, analysis of plasma membrane fractions revealed strongly reduced levels of mutants S966fs and del(K935-S940)ins(I) (Fig. 2C), indicating that impaired targeting to the cell surface accounts for the absence of pump currents in oocytes expressing these mutants.
Dependence of Stationary Pump Currents on Membrane Potential and [K ϩ ] ext -We next determined the K 0.5 values for extracellular K ϩ for mutants that showed sufficiently high pump currents at Ϫ30 mV (Table 1). Apparent K ϩ affinities were increased for R763H and X1021R but decreased for A606T (p Ͼ 0.05). Examination of the voltage dependence of pump currents revealed altered [K ϩ ]-dependent I-V curves for A606T, R763H, and R834Q compared with WT (Fig. 3), and the voltage dependence of K 0.5 values for A606T and R763H was markedly changed (Fig. 3, insets). At 10 mM [K ϩ ] ext , stationary currents of WT ATP1A2 were in saturation, and the corresponding I-V curve increased steadily with voltage ( Fig. 3A). At [K ϩ ] ext Ͻ 5 mM, however, bell-shaped I-V curves were obtained with a maximum at ϩ20 mV. At subsaturating [K ϩ ] ext the elec-FIGURE 2. Activity and expression of ATP1A2 constructs. A, mean stationary currents at 10 mM K ϩ and Ϫ30 mV from oocytes expressing WT and mutant ATP1A2 alleles. This method could only be applied for the determination of pumping activity for mutants, which consistently yielded currents larger than 10 nA, in other cases (marked n.d.) currents in response to 10 mM K ϩ did not exceed background current fluctuations. The data were obtained from more than 30 oocyte batches (means Ϯ S.E., numbers of cells are indicated). B, residual ion transport activity as assessed by ouabain-sensitive Rb ϩ uptake, which was measured as the difference between mean uptake values measured in presence of 10 M and 10 mM ouabain, respectively, from four batches of cells (n ϭ 14 -30 cells; n.d., not determined). Uninjected oocytes served as controls. The data were normalized to WT ATP1A2 measured at 10 M ouabain (Experiment 1, 49 pmol/oocyte/min; Experiment 2, 130 pmol/ oocyte/min; Experiment 3, 38 pmol/oocyte/min; Experiment 4, 61 pmol/oocyte/min). C, representative Western blots of protein samples from plasma membrane (PM) and total membrane (TM) preparations of oocytes expressing various ATP1A2 constructs. The equivalent of one (left panels) or two (right panels) oocytes was loaded per lane. Uninjected oocytes served as controls. The experiments were repeated three times, with identical results. trogenic binding of extracellular K ϩ is responsible for the decrease of the stationary currents at positive potentials (40). For A606T the maxima of the corresponding I-V curves were shifted to ϩ50 mV, and bell-shaped curves were only obtained at [K ϩ ] ext Ͻ 2 mM (Fig. 3B). In contrast, for R763H and R834Q bell-shaped I-V curves were even found at saturating extracellular [K ϩ ] (Fig. 3, C and D), indicating an abnormal decrease of enzyme turnover at positive potentials.
Voltage-dependent K 0.5 values of WT ATP1A2 (inset in Fig.  3A) follow a characteristic U-shaped curve with a minimum at 0 mV. The K 0.5 curve for A606T was positively shifted with a minimum at ϩ50 mV, resulting in increased K 0.5 values below 0 mV compared with WT. In contrast, K 0.5 values for R763H were significantly reduced at all potentials, and those of R834Q were slightly smaller than for WT between ϩ20 and Ϫ40 mV. Other mutants did not differ from WT in this set of experiments (data not shown). Because of elevated background con-ductances and current drifts in oocytes expressing X1021R, sufficiently stable recordings for repetitive voltage pulse experiments were not feasible.
Kinetics and Voltage Dependence of Transient Currents-To further investigate the properties of electrogenic Na ϩ transport, we performed voltage pulse experiments under extracellularly high [Na ϩ ], K ϩ -free conditions. Fig. 4B shows representative ouabain-sensitive transient currents from an oocyte expressing WT ATP1A2. The apparent rate constants were ϳ80 s Ϫ1 at positive potentials and exhibited a characteristic increase at negative potentials (Fig. 4B). Fig. 4D shows the voltage-dependent distribution of translocated charge (Q-V curve) for the WT ATP1A2-expressing cell from Fig. 4B.
The mean values for the fractional charge z q and the midpoint potential V 0.5 from Boltzmann fits of Q-V curves for the different ATP1A2 mutants are listed in Table 1. The C-terminal missense mutation X1021R led to most striking abnormalities in the behavior of transient currents under electrogenic Na ϩ /Na ϩ exchange conditions. Contrasting with WT (Fig. 4, B and C), mutant X1021R showed an increase of apparent rate constants with positive potentials (Fig. 4, E and F). Moreover, in all X1021R-expressing cells the Q-V distribution was nearly linear over the range of potentials investigated (Fig. 4G). Thus, the saturation values Q max and Q min from fits of a Boltzmann function were not well defined, and the strongly reduced z q value of ϳ0.3, which was consistently derived from the data, likely represents only an upper limit. Therefore, the voltage dependence of charge movement is very shallow compared with WT, implying that electrogenic Na ϩ translocation is severely altered for X1021R.  The remaining constructs showed z q values similar to WT. However, various mutants exhibited de-or hyperpolarizingly shifted V 0.5 values (Table 1 and Fig. 5), indicating changed apparent Na ϩ affinities. Apparent rate constants from transient currents for R383H were generally slower than WT values (Fig.  5A) and saturated at ϳ250 s Ϫ1 at potentials below Ϫ120 mV, indicating impaired flexibility during the E 1 P to E 2 P conformational transition. In contrast, significantly increased rate constants were found for A606T, R763H, M829R, and R834Q ( Fig.  5, C, E, G, and J). The voltage dependence of rate constants for A606T (Fig. 5C) exhibited a shift to depolarizing potentials compared with WT, whereas the values for R763H (Fig. 5E), M829R (Fig. 5G), and R834Q (Fig. 5J) were elevated over the whole range of potentials tested, which emerges as a general feature for the group of mutations within the contact region between helix M5, the P domain, and the L67 loop.
Turnover Numbers-As listed in Table 1, R383H, R763H, and R834Q exhibited significantly reduced turnover numbers (p Ͼ 0.005) compared with WT, and also the A606T value was slightly smaller. A strongly reduced transport capacity was also found for X1021R. Here again, the obtained value (3.5 s Ϫ1 ) only represents an upper limit, because smaller values would result if the linear progression of the Q-V curve would extend over a larger voltage range.

DISCUSSION
This study provides a detailed electrophysiological and biochemical characterization of a comprehensive subset of ATP1A2 alleles implicated in FHM and SHM (17,34). A broad range of functional abnormalities was observed that can be subdivided into two classes. The first group of mutations is characterized by a severe reduction or loss of catalytic activity or by an impairment of plasma membrane delivery. For these mutations rather straightforward interpretations regarding the underlying structural disruptions can be given. The second group of mutants exhibits more subtle functional impairments, such as changes in apparent cation affinities, voltage dependence, or kinetics, which frequently occur in parallel to a reduction in turnover numbers. These mutations are in the C terminus (X1021R) or in regions involved in important interdomain interactions (31,32). We will discuss this latter group in more detail regarding the relationship between structural and functional disruptions. Of note, all mutant alleles impaired the ability of ATP1A2 alleles to sustain cell viability in ouabain survival assays, suggesting a loss-of-function effect also on a cellular level (data not shown), in agreement with earlier studies (5,28,30).

Mutations Affecting Catalytic Activity or Plasma Membrane
Expression-Pump currents and Rb ϩ uptake were very small or completely abolished for mutants T263M, T376M, R937P, del(K935-S940)ins(I), and S966fs (Fig. 2). Because the amount of total and plasma membrane protein was normal for T263M, T376M, and R937P, the most likely explanation for the observed loss of function is a decreased or abolished catalytic turnover. For del(K935-S940)ins(I) and S966fs, a severe reduction of plasma membrane protein compared with WT was found (Fig. 2C). The Ser 966 frameshift mutation disrupts the C-terminal sequence of the enzyme starting from the extracellular end of helix M9. Although this is the first time that reduced plasma membrane trafficking or protein stability was found in conjunction with FHM2 mutations, it is evident that a deletion or a frameshift mutation may result in premature degradation caused by improper protein folding and by interference with cellular quality control mechanisms.
A rather obvious explanation can also be provided for mutation T376M. Because Thr 376 resides in the signature sequence SDKTGT and is involved in Mg 2ϩ coordination at the catalytic site, the T376M mutant represents a completely dysfunctional enzyme. R937P, like del(K935-S940)ins(I), affects the loop between transmembrane segments M8 and M9. Despite normal protein levels, ion pumping activity was very small for R937P (Fig. 2, B and C). In the crystal structure of the Na ϩ /K ϩ -ATPase (32), Arg 937 has been implicated in correct positioning of the C terminus of the enzyme by forming H-bonds to Tyr 1019 and one of the terminal oxygen atoms of the C terminus (Fig.  6A). The latter is further stabilized against M5 by H-bonds of Tyr 1020 and the other C-terminal oxygen with Lys 770 . This arrangement was suggested to be important for correct formation of a putative third Na ϩ -binding site (32). Furthermore, it was proposed that Arg 937 is part of a C-terminal cluster of arginines, which might be involved in voltage sensitivity, similar to the voltage sensor arginines in voltage-gated cation channels (32). Because replacement of Arg 937 by proline can cause backbone distortions and disrupt H-bond patterns, positioning of the Na ϩ /K ϩ -ATPase C terminus is possibly disturbed. Thus, the M8-M9 loop appears to be critical in preserving the native enzyme structure and in coordinated interdomain interactions during the catalytic cycle.
Somewhat more subtle are the consequences of the T263M mutation, which is one of three FHM mutations identified in the A domain so far and affects a conserved residue within the third core motif of P-type ATPases (31).
Functional consequences were severe, because currents and transport capacity in Rb ϩ uptake experiments were reduced to below 10% of WT (Fig. 2). Because plasma membrane expression was normal, the decrease in pumping activity is most likely due to a drastic reduction in turnover. In the Na ϩ /K ϩ -ATPase crystal structure (32) Thr 263 resides in the intracellular M2-M3 loop at the C-terminal end of a short ␣-helix, at which helix M3 drags to move the A domain during the E 1 -E 2 conformational change. Inspection of the crystal structures of the homologous Ca 2ϩ -ATPase SERCA reveals that during this conformational transition the ␣-helical pattern around the corresponding SERCA-Thr 230 is partially broken (37,38). If a similar mechanism would apply for the Na ϩ /K ϩ -ATPase, the bulkier methionine side chain could sterically disrupt the local structure including the hydrogen bonding pattern around the mutated residue and thus impair the coupling of helix M3 movement to the A domain. The dramatic reduction in activity highlights the critical role of the respective core motif for function of the Na ϩ /K ϩ -ATPase.
A similar interpretation can be provided for R383H, which was characterized by a 48% decrease in catalytic turnover. Arg 383 is highly conserved within the fifth core motif of P-type ATPases and resides in a hinge region (homologous to SERCA-Gln 360 within the TNQMS loop motif), which flexibly links the N domain to the P domain (Fig. 1B). Because the N domain undergoes considerable movements toward the P domain during the catalytic cycle, the R383H mutation apparently interferes with the flexibility during conformational changes. This interpretation is supported by the observed reduction of rate constants during the voltage jump-induced E 1 P 7 E 2 P conformational transition and the reduced turnover number. The additionally observed shift of the Q-V curve derived from transient currents by about ϩ26 mV indicates an effect on conformational equilibria, suggesting a relative stabilization of the E 1 P conformation at physiological potentials. Of interest, the recently reported FHM mutation R593W (16) localizes to the second hinge between N and P domain. Largely reduced cell survival has been reported for this variant, emphasizing the critical role of these hinges for conformational flexibility and thus pumping rate.
Mutations of Residues Involved in H-bond Networks at the Interface between the L67, P, and TM Domains-A606T, R763H, M829R, and R834Q affect highly conserved residues involved in interdomain H-bond networks of P-type ATPases. Apart from a reduction of the turnover number, which was observed for most mutations, the main findings for this group of mutations involve increased rate constants of transient currents (Fig. 5, C, G, E, and J). In addition, subtle changes in other voltage-dependent properties or cation affinities were found. In the Na ϩ /K ϩ -ATPase crystal structure (32) the distinct hydrogen bonding networks between Ala 606 , Arg 763 , Met 829 , and Gly 365 (Fig. 6B) stabilize the relative positions of the L67 loop (Met 829 ), the P domain (Ala 606 and backbone oxygen of Gly 365 ), and helix M5 (Arg 763 ). In addition, Arg 834 forms H-bonds to Asp 839 (L67) and Ser 366 (P domain) (Fig. 6C). Such H-bond networks have already been pointed out between SERCA residues Ala 617 , Arg 751 , and Met 817 (37,38), which correspond to ATP1A2 residues Ala 606 , Arg 763 , and Met 829 , respectively. During the conformational change from the SERCA E 1 (Ca 2ϩ ) structure (Protein Data Bank entry 1SU4 (37)) to the E 2 (TG) structure (Protein Data Bank entry 1IWO (38)), H-bonds are slightly rearranged but stay intact between Arg 751 and Ala 617 or Met 817 , indicating that the local structure needs to be rather rigidly stabilized during catalysis. Further C-terminally, SERCA residue Arg 822 (corresponding to ATP1A2 Arg 834 ) is hydrogenbonded via a water molecule to a conserved Glu residue within the first ␣-helix of the P domain (37). Thus, it is likely that ATP1A2 residues Ala 606 , Arg 763 , Met 829 , and Arg 834 are critically involved in structural stabilization of this interdomain interface region during the catalytic cycle. It emerges as a general feature that mutations that are expected to disrupt these H-bonding patterns cause an abnormally high flexibility during the E 1 P to E 2 P conformational transition.
A major finding of this study was that mutations of various residues within a hydrogen-bonded network of amino acids between various enzyme domains do not lead to a complete inactivation of the Na ϩ pump but induce changes in the kinetics of the E 1 P to E 2 P conformational change and of voltage-dependent properties, accompanied by complex long range effects on cation interactions. Because the interdomain interaction region is located at some distance from the cation-binding pocket, the observed changes in cation affinities cannot be explained by structural disruptions within the immediate vicinity of the mutated residues. However, the abnormally increased rate constants of the voltage jump-induced redistribution between E 1 P and E 2 P states can directly affect the poise of the voltage-dependent distribution between these two states. Indeed, shifted Q-V curves were predominantly observed that directly indicate changes of the voltage-dependent distribution between E 1 P and E 2 P. Therefore, in accordance with previous suggestions by Jørgensen et al. (41), it can be argued that the observed changes in the apparent affinities for cations occur secondary to shifts in conformational equilibria. However, the observed long range effects are complex, because shifts in the dynamic E 1 P-E 2 P equilibrium do not consistently correlate with cation affinity changes: On one hand, for A606T K ϩ affinity was decreased, but the Q-V curve was shifted to positive potentials, in line with an increased Na ϩ affinity or a relative stabilization of the E 1 P state. On the other hand, for R763H an increased K ϩ affinity was measured, but the Q-V curve was not shifted; thus, Na ϩ affinity or the poise of the equilibrium between E 1 P and E 2 P states was not changed. In contrast, mutations M829R and R834Q left the apparent K ϩ affinity unchanged, but from the observed shifts of the Q-V curves either a decrease (M829R) or an increase (R834Q) of the apparent affinity for extracellular Na ϩ could be inferred. Therefore, the observed affinity changes are most likely due to a complex superposition of, first, direct long range structural effects on the configuration of cation coordinating residues and, second, changes in kinetics that induce shifts in conformational equilibria. The consistently observed elevation of the apparent rate constants from transient currents for mutants A606T, R763H, M829R, and R834Q indicates an abnormally high flexibility during the E 1 P-E 2 P conformational change, in line with a destabilizing effect of the respective mutations. Although all these changes can explain impairment of Na ϩ /K ϩ -ATPase function under physiological conditions, it will be challenging to consistently reconcile the observed functional abnormalities with structural considerations or predictions from structural modeling.
Expansion of the C Terminus by Mutation X1021R-In accordance with the suggestions of Morth et al. (32), who proposed a significant influence of the Na ϩ /K ϩ -ATPase C terminus on voltage sensitivity and cation coordination, several striking functional abnormalities were observed for this mutant. A strong increase in the apparent affinity for extracellular K ϩ was determined, paralleled by a decreased apparent affinity for Na ϩ , as derived from the negatively shifted Q-V curve. In addition, the turnover number was reduced by 70% compared with WT. The relatively high Rb ϩ uptake (ϳ75% of WT), which was measured at 1 mM [Rb ϩ ], can be explained by the increased K ϩ affinity of X1021R, so that the transport rate for the mutant construct is close to saturation, whereas turnover is only about half-maximal for the WT enzyme under these conditions. Furthermore, voltage dependence and kinetics of transient currents were completely different from WT, because the apparent rate constants increased with positive potentials. Such an inverted behavior has only been described for a mutation (D808E) that explicitly affects one of the cation-coordinating residues within the Na ϩ /K ϩ -ATPase cation-binding pocket (39,42). To account for such a voltage dependence of rate constants, one of the reaction steps within Na ϩ forward transport (i.e. Na ϩ binding from intracellular, E 1 P-E 2 P conformational change, or extracellular Na ϩ release) must be voltage-dependent for X1021R (43). This is in contrast to WT, in which reverse binding of extracellular Na ϩ underlies voltage dependence (Fig.  1A). Also in contrast to WT, the voltage dependence of charge movement (z q from Q-V curve) was much weaker and thus electrogenicity of Na ϩ transport was drastically reduced. In summary, our data suggest that severe changes in the properties of electrogenic Na ϩ transport underlie X1021R behavior and support the notion that the C terminus of Na ϩ /K ϩ -ATPase critically affects voltage sensitivity and interaction with Na ϩ . At any rate the observed alterations severely affect the ability of the protein to maintain the physiologically required gradients for Na ϩ and K ϩ .
Functional Defects of ATP1A2 Alleles and Migraine-How can these findings be reconciled with the occurrence of migraine attacks in patients with FHM and SHM? Most of the aura symptoms in migraine are generated by cortical spreading depression (CSD) or CSD-like events that are characterized by a spreading front of excitation (depolarization), followed by long lasting depression (hyperpolarization) (44 -47). All of the mutations investigated here resulted in a partial or complete loss of activity of the Na ϩ /K ϩ -ATPase, yet because of diverse functional abnormalities. Loss of function of the Na ϩ /K ϩ pump may lead to neuronal (and glial) depolarization through various ways. First, because of the critical role of K ϩ channels for repolarization, a reduction of the K ϩ gradient might destabilize the membrane resting potential and thus reduce the activation threshold. Second, breakdown of the Na ϩ gradient might affect removal of the excitatory neurotransmitter glutamate from the synaptic cleft, which is driven by the astrocytic Na ϩ /glutamate transporter. Elevated levels of extracellular glutamate have been shown to trigger CSD. Third, impairment of the Na ϩ gradient might cause elevated intracellular Ca 2ϩ , augmented Ca 2ϩ influx into the endoplasmatic reticulum, which in turn leads to increased Ca 2ϩ signaling (48). This is because the Na ϩ gradient is the energy source for the Na ϩ /Ca 2ϩ exchanger. Together, these changes might contribute to the occurrence of CSD and CSD-like events in patients with FHM and SHM.
Conclusions-Our study identifies reduced or altered activity of the Na ϩ /K ϩ -ATPase as the common denominator associated with ATP1A2 mutations in familial and sporadic hemiplegic migraine. Among the variety of molecular phenotypes, most of the observed changes involve voltage-dependent properties that are critically important for this electrically active enzyme within a physiological context. Our results demonstrate that mutations of residues that are involved in relative stabilization of different enzyme domains during conformational changes result in abnormally high rate constants during the E 1 P-E 2 P conformational change and evoke intricate long range consequences on cation interactions. The complexity of these long range effects is illustrated by the finding that shifts in the E 1 P-E 2 P equilibria do not consistently correlate with changes in apparent cation affinities. Furthermore, our observations on X1021R strongly support conclusions by Morth et al. (32), who outlined the distinct impact of the C terminus on cation coordination, electrogenicity, and voltage dependence of cation transport. This hallmark of the Na ϩ /K ϩ -ATPase structure deserves further elucidation by systematic structure-function studies.