Kinetic Alterations due to a Missense Mutation in the Na,K-ATPase α2 Subunit Cause Familial Hemiplegic Migraine Type 2*

A number of missense mutations in the ATP1A2 gene, which encodes the Na,K-ATPase α2 subunit, have been identified in familial hemiplegic migraine with aura. Loss of function and haploinsufficiency have been the suggested mechanisms in mutants for which functional analysis has been reported. This paper describes a kinetic analysis of mutant T345A, recently identified in a detailed genetic analysis of a large Finnish family (Kaunisto, M. A., Harno, H., Vanmolkot, K. R., Gargus, J. J., Sun, G., Hamalainen, E., Liukkonen, E., Kallela, M., van den Maagdenberg, A. M., Frants, R. R., Farkkila, M., Palotie, A., and Wessman, M. (2004) Neurogenetics 5, 141–146). Introducing T345A into the conserved rat α2 enzyme does not alter cell growth or catalytic turnover but causes a substantial decrease in apparent K+ affinity (2-fold increase in K0.5(K+)). In view of the location of Thr-345 in the cytoplasmic stalk domain adjacent to transmembrane segment 4, the 2-fold increase in K0.5(K+) is probably due to T345A replacement altering K+ occlusion/deocclusion. Faster K+ deocclusion of the mutant via the E2(K) + ATP → E1·ATP + K+ partial reaction is evidenced in (i) a marked increase (300%) in K+ stimulation of Na-ATPase at micromolar ATP, (ii) a 4-fold decrease in KATP, and (iii) only a modest increase (∼3-fold) in I50 for vanadate, which was used as a probe of the steady state E1/E2 conformational equilibrium. We suggest that the decreased apparent K+ affinity is the basis for a reduced rate of extracellular K+ removal, which delays the recovery phase of nerve impulse transmission in the central nervous system and, thereby, the clinical picture of migraine with aura. This is the first demonstration of a mutation that leads to a disease associated with a kinetically altered but fully functional Na,K-ATPase, refining the molecular mechanism of pathogenesis in familial hemiplegic migraine.

Familial hemiplegic migraine (FHM) 1 is a rare autosomal dominant form of migraine with aura. This disorder is usually associated with hemiparesis and can be accompanied with clinical features ranging from ataxia to epileptic seizures. This genetically heterogeneous disease has been traced to at least two loci, FHM1 and FHM2. FHM1, which accounts for over 50% of all FHM families, has been traced to chromosome 19p13 and associated with missense mutations in the CACNA1A gene encoding the ␣1 subunit of the voltage-dependent neuronal (P/Q type) calcium channel. A recent breakthrough in migraine genetics is the discovery of missense mutations in the ATP1A2 gene on chromosome 1q23 that encodes the ␣2 isoform of the Na,K-ATPase. This finding gives strong support to the notion that FHM is caused by disruption of normal cation transport.
(For a recent update on familial hemiplegic migraine, see Ref. 1.) The Na,K-ATPase is an integral membrane protein complex that comprises a large catalytic ␣ subunit of ϳ110 kDa as well as a smaller, highly glycosylated ␤ subunit that ensures the proper folding and mooring of ␣ in the plasma membrane. This P-type ion pump catalyzes the ATP-driven exchange of intracellular Na ϩ for extracellular K ϩ ions across the plasma membrane of virtually all animal cells, with a stoichiometry of 3:2 (for recent reviews, see Refs. 2 and 3). Thus, the sodium pump is essential to the maintenance of the electrochemical alkali cation gradients that are tapped by ion channels in the propagation of action potentials. During the course of its catalytic cycle, this P-type ion pump undergoes phosphorylation and dephosphorylation of a conserved aspartate residue in the active site of its catalytic subunit. This ion pump also undergoes conformational transitions of dephospho-and phosphoenzyme, commonly referred to as E 1 7 E 2 and E 1 P 7 E 2 P transitions, respectively (see Scheme 1).
At present, four isoforms of ␣ and three isoforms of ␤ have been described. Both ␣ and ␤ are distributed in a tissue-and development-dependent manner. In adult mammals, the ␣2 isoform is located principally in skeletal muscle, and the brain (in particular glial cells), and to a lesser extent in the heart, adipocytes, and the eye (see . We have previously shown that the ␣2 enzyme differs from the ubiquitous ␣1 subunit, primarily in the steady state of the E 1 /E 2 equilibrium. Thus, compared with ␣1, the E 1 /E 2 poise of ␣2 is shifted toward E 1 , and this shift is associated with an ϳ2-fold increase in apparent affinity for ATP, a 50% decrease in catalytic turnover, a 3-fold increase in the K ϩ deocclusion rate, and a 20-fold increase in the I 50 for vanadate inhibition of Na,K-ATPase (8 -11).
Since the recent original observation that ATP1A2 alleles are associated with FHM2 (12), an increasing number of mutant alleles of the ␣2 subunit gene have been identified in families segregating FHM2. However, there have been extremely limited functional studies probing the mechanism by which the proposed pathogenic alleles act. These include L764P and W887R (12), M731T and R689Q (13), and most recently T345A (14), as well as several putative candidates presently under investigation (21). 2 Thus far all alleles are missense, nearly all altering one amino acid in the large catalytic cytosolic loop between M4 and M5. Only the mutant W887R is found outside this domain in the extracellular loop between M7 and M8. No deletions, frameshifts, nonsense, or splice site alleles have been clearly identified. In the first two reported mutations, L764P and W887R, De Fusco et al. (12) implicate altered sodium pump activity and suggest that the disease is the result of haploinsufficiency because the mutant enzyme does not support the growth of cells in culture. Because the gene was not fully sequenced in this report, and because prior studies of 1q-linked FHM2 failed to reveal causal mutations in the gene (15), this functional attribute of the specific mutations was an essential link to ascribing pathogenicity. Although subsequent reports have included complete sequencing of at least all exons to ensure definition of a discrete mutation, none of the other reports of mutant alleles have included any functional analysis, leaving the link between the in vitro pump phenotype and pathogenicity insecure. In initial experiments, we tested the hypothesis that the in vitro growth phenotype would prove a robust means of discriminating the disease-associated alleles by introducing the five aforementioned disease-associated mutations (L764P, W887R, M731T, R689Q, and T345A) into the evolutionarily conserved (ϳ95% identity) rat ␣2 enzyme and assessing their abilities to support cell growth. Cells bearing mutants L764P and W887R failed to grow, but the other three alleles supported growth, suggesting that either many alleles were misclassified or that the critical aspect of altered pump function remained to be identified.
In this paper we describe a detailed kinetic analysis of the mutant T345A because the previous study (14) identified all of the common polymorphisms in the exons and introns of the gene, showing that only the T345A mutation co-segregated with the disorder in a large family in which clinical data were available on 28 family members; the disorder was not present in 132 healthy Finnish control subjects.

EXPERIMENTAL PROCEDURES
Mutagenesis, Transfection, and Cell Culture-The T345A, R689Q, M731T, L764P, and W887R mutations of the rat ␣2 cDNA were derived from the ouabain-insensitive rat ␣2* 3 cDNA developed by Jewell and Lingrel (16). Rat ␣2* cDNA was introduced into a modified pIBI shuttle vector, and mutations were introduced with the QuikChange mutagenesis kit (Stratagene). Clones containing the correct substitutions were identified by full-length sequencing of the mutated cDNA. The mutant cDNAs were then excised from the shuttle vector with HindIII and ligated into pcDNA3.1 (Invitrogen), and orientation was determined by restriction analysis.
Growth Curves-Cells were plated (5000 cells/well), grown in 24-well plates, trypsinized for 10 min at 37°C at the times indicated, and then counted using a hematocytometer.
Membrane Preparation and Enzyme Assays-NaI-treated microsomal membranes were prepared from the mutant cells as described previously (16,17). To determine maximal Na,K-ATPase activity (as in assays of catalytic turnover) and the effect of inorganic orthovanadate, assays were carried out with final concentrations of 1 mM ATP, 100 mM NaCl, 10 mM KCl, 3 mM MgSO 4 , 20 mM histidine (pH 7.4), 5 mM EGTA (pH 7.4), and 5 M ouabain (Sigma). Base-line hydrolysis was determined using 5 mM ouabain. Components of Na,K-ATPase activity were measured as the release of 32 P i from [␥-32 P]ATP as described previously (18). Catalytic turnover was estimated from the ratio of V max to EP max as in Ref. 11. Unless indicated otherwise, transport assays were carried out as described by Munzer et al. (19). Kinetic constants were determined by fitting the data to a simple 1-site Michaelis-Menten model (KЈ ATP ), or to a 2-site (K 0.5(K ϩ ) ) or 3-site (K 0.5(Na ϩ ) ) cooperative model, v ϭ V max [cat] n /(K ϩ [cat] n where n ϭ 2 or 3 for the cation (cat), either K ϩ or Na ϩ , respectively. Curve fitting was carried out using the Kaleidagraph computer program (Synergy). All measurements were carried out concurrently on mutant and control (wild-type) ␣2* enzymes, and each assay was performed in triplicate on at least two separate clones.
Polyacrylamide Gel Electrophoresis and Western Blotting-Unless indicated otherwise, SDS-PAGE was carried out using 10% NuPage gels (Novex) with SDS/MES (4-morpholineethanesulfonic acid) running buffer. Running and sample buffers were prepared according to the manufacturer's instructions. Immunoblotting to detect ␣2 was performed with the ␣2-specific monoclonal antibody McB2, which was kindly provided by Dr. K. Sweadner.

RESULTS
In initial experiments, we cloned the full-length cDNA of the rat ␣2 subunit expressing the T345A, R689Q, M731T, L764P, and W887R substitutions. To distinguish the activity of these exogenous pumps from endogenous pumps in the recipient cell line, mutations Q116R and N127D had been introduced in the first extracellular loop to render them insensitive to ouabain. As first shown by Jewell and Lingrel (16), wild-type (WT) ␣2*-transfected cells bearing these mutations can be selected, isolated, and grown in medium that contains ouabain added at low concentration (1 M) to inhibit the endogenous recipient cell enzyme. Following transfection of T345A, R689Q, M731T, L764P, and W887R, neither L764P nor W887R supported the growth of the cells in 1 M ouabain, confirming the findings of De Fusco et al. (12). In contrast, cells transfected with mutants R689Q and M731T, like WT ␣2*, survived growth in ouabain, although at a markedly reduced rate in the case of R689Q (not shown).
Growth of the T345A mutant in ouabain was shown previously by Arguello et al. (20) in their mutagenesis study of oxygen containing residues of the Na,K-ATPase. 4 In fact, Fig. 1 shows that the growth rates of WT ␣2* and the T345A are similar. Analysis of protein expression by Western blotting (Fig. 1, inset) indicates that membranes isolated from WT and mutant cells have virtually the same amounts of ␣2 protein/ unit of activity. This indicates that the catalytic turnover is not altered by the T345A mutation. In view of the normal survival of T345A and the particularly clear identification of this mutation as a cause of FHM2 (see the Introduction), we carried out a series of kinetic assays to determine whether the Thr 3 Ala substitution alters the kinetic behavior of the ␣2 pump.
Na ϩ -, K ϩ -, and ATP-dependent Activation Profiles-The representative experiments in Fig. 2, A and B, show the membrane Na ϩ -and K ϩ -activation profiles of T345A compared with WT ␣2*. As shown in Fig. 2A, a difference in the Na ϩ -activation profile between the two enzymes could not be detected; K 0.5(Na ϩ ) is ϳ5 mM for both. In contrast, the T345A replacement resulted in a 2-fold increase in K 0.5(K ϩ ) (Fig. 2B) and an ϳ4-fold decrease in KЈ ATP (Fig. 2C). It is noteworthy that the decrease in apparent affinity for K ϩ seen in unsided membrane preparations is also seen under more physiological conditions prevailing in intact cells. As shown in Fig. 3, the apparent affinity for extracellular K ϩ is similarly decreased ϳ2-fold by the T345A mutation. Values of K 0.5(K ϩ ) , which are 0.63 and 1.13 mM for WT ␣2 and the T345A mutant, respectively (Fig. 3), were doubled when the flux experiments were carried out at 100 mM Na ϩ (experiment not shown).
K ϩ Sensitivity of Na-ATPase Activity-Because K ϩ binding/ occlusion (K ext ϩ E 2 P 3 E 2 (K) ϩ P i ) precedes K ϩ deocclusion (E 2 (K) ϩ ATP 3 ATP⅐E 1 ϩ K cyt ), a decrease in the latter reaction should be evidenced in an increase in K ϩ stimulation of Na-ATPase at micromolar ATP concentration with an associated higher apparent affinity for ATP. This is indeed the case, as seen in Fig. 4; the associated decrease in KЈ ATP was shown above (Fig. 2C). At 1 M ATP, the maximal K ϩ stimulation of Na-ATPase of T345A is near 300% compared with 120% for ␣2*. It is noteworthy that this ϳ2.5-fold increase in the percent stimulation is close to that estimated from a simple Michaelis-Menten model, where the ϳ4-fold decrease in KЈ ATP of T345A compared with the WT ␣2* (Fig. 2C) should affect a similar increase in E 2 (K) 3 E 1 .
Sensitivity to Vanadate-To determine whether the relatively faster E 2 (K) 3 E 1 sequence of T345A results in a shift in steady-state E 1 /E 2 poise toward E 1 , we compared the vanadate sensitivities of the WT and T345A mutants. Inorganic orthovanadate is a transition-state analog of inorganic phosphate that binds to P-type ATPases in the E 2 conformation during steady state catalysis. Thus, Na,K-ATPase sensitivity to inhibition by vanadate is a measure of the proportion of enzyme in the E 2 state. In the representative experiment shown in Fig. 5, I 50 (M) values were 6.40 Ϯ 0.75 M and 20.09 Ϯ 0.40 M for ␣2* and ␣2T345A, respectively. Because this shift is relatively small, it is not surprising that the T345A mutation did not significantly affect catalytic turnover measured as either the ratio of V max /EP max (not shown) or the ratio of immunoreactive ␣2 protein/unit of activity (Fig. 1). DISCUSSION The identification of ATP1A2 as a causative gene in FHM2 is a major discovery in migraine genetics and has underscored the relevance of ion transport dysfunction to the pathophysiology of migraine. An increasing number of missense mutations in ATP1A2 that likely cause FHM2 are being identified, including L764P and W887R (12), R689Q and M731T (13), and most recently D718N, R763H, P979L (21), and T345A (14). Several others await validation (21). For at least two mutations (L764P and W887R), loss of function and haploinsufficiency are the suggested underlying defects based on a cell growth phenotype. However, because a myriad of critical physiological processes (from cell volume regulation to sodium gradient-dependent transport to the Nernst potentials for Na ϩ and K ϩ and their influence upon membrane potentials) all ultimately rely on normal pump function, the pathophysiological mechanism underlying FHM2 remains unresolved. In this study, we performed a detailed kinetic analysis of the T345A mutant pump cycle to reveal qualitative and quantitative changes associated with the disease allele, bringing focus to its reduced affinity for extracellular K ϩ . We also demonstrate that the growth phenotype may not be capable of discriminating many pathogenic alleles. It is clear that the growth rate of T345A-transfected HeLa cells is as high as WT control and that the T345A mutant enzyme is fully active, with no reduction in catalytic turnover (V max /EP max ).
The glial cells that abundantly express the ␣2 isoform of the sodium pump have the key role in determining the resting extracellular K ϩ concentration bathing neurons in the brain. This concentration directly alters E K ϩ and thereby the magnitude of neurons' after-hyperpolarizations and rates of repolarization and pacing (22,23). Accordingly, the lowered apparent affinity for extracellular K ϩ affected by the T345A mutation should slow removal of K ϩ from the extracellular space and thereby slow the recovery from neuronal excitation. It is noteworthy that the K 0.5(K ϩ ) values at a resting membrane potential of Ϫ70 mV may be even higher (ϳ35%) (see Fig. 5B in Ref. 24.) than in the depolarized state (the conditions of the present study with porous membranes). The change in K ϩ handling is thus a likely explanation for the altered cortical spreading depression in migraine associated with aura as hypothesized by Lauritzen (25). We suggest that the decreased removal of extracellular K ϩ during the recovery phase of nerve impulse transmission is probably the critical manifestation of altered pump function that leads to disease. This may arise because of either a decrease in pump number (L764P and W887R) or a decrease in apparent affinity for extracellular K ϩ (T345A). In an earlier analysis of the heterozygous ␣2 knock-out mouse, Moseley et al. (26) had similarly suggested that impaired clearance of extracellular K ϩ resulting from a loss of ␣2 function affects neuronal excitability in both neurons and glia. Under physiological conditions of high extracellular Na ϩ concentration, the apparent affinity of the T345A mutant for K ϩ is lowered sufficiently (see under "Results"). This delays the normal rate of recovery of extracellular K ϩ from values well above saturation to the resting level (3 mM) following action potential repolarization (27). It is less likely that the primary diseasecausing effect of the T345A mutation results from altered subcellular localization or delivery to the plasma membrane rather than the kinetic effect on K ϩ affinity. Our experiments with intact cells indicate that the mutation neither decreases cell growth nor diminishes cell surface expression as evidenced in V max of pump flux measured with intact cells (see legend to Fig. 3).
Decreased activity of ␣2 effected by either haploinsufficiency or reduced apparent K ϩ affinity may alter the Na ϩ gradient. As proposed by James et al. (28), if the ␣2 pump and Na ϩ /Ca 2ϩ exchanger are co-localized in microdomains of the same cell (astrocyte), a localized rise in cell Na ϩ (caused by the decreased K 0.5(K ϩ ) ) would increase intracellular Ca 2ϩ because of decreased Na ϩ /Ca 2ϩ exchange activity. Increased cell Ca 2ϩ may, in turn, affect Ca 2ϩ -dependent processes in regions critical to the pathogenesis of migraine (cf. Ref. 1).
The T345A replacement likely decreases apparent K ϩ affinity by directly affecting the cation binding pocket. Although the cation binding sites of the sodium pump are located within the transmembrane segments, earlier studies, such as our H,K/ Na,K-ATPase chimera experiments, showed that a portion of the M4-M5 loop juxtaposed to the stalk segment leading to M4affected cation selectivity (29). Furthermore, based on the alignment of the amino acid sequence of the Na,K-ATPase with the homologous sarco(endo)plasmic reticulum Ca 2ϩ -ATPase pump and with a known crystal structure, Thr-345 is located in stalk segment 4 within three residues of the plasma membrane (30,31), at least in the E 1 conformational state. In addition, M4 has an important role in the formation of the cation binding pocket and undergoes a notable rearrangement during the course of large cytoplasmic domain movements that accompany rearrangement of the transmembrane helices. As the enzyme shifts from E 1 to E 2 , M4 moves a distance equivalent to one turn of an ␣-helix toward the extracellular space, consequently pulling Thr-345 into the lipid bilayer. (For a review of structural changes in conformational shifts, see Ref. 32.) Homology modeling of the Na,K-ATPase onto the structure of the sarco-(endo)plasmic reticulum Ca 2ϩ -ATPase pump has revealed that the residues involved in K ϩ ligation are similar to those involved in Ca 2ϩ binding (33). For the Na,K-ATPase, the residues of M4 involved in the formation of "site II" are similar for Na ϩ and K ϩ , with the exception of Glu-332. Although Glu-332 directly coordinates Na ϩ , it may only coordinate K ϩ via a water molecule. It is attractive to hypothesize that increased hydrophobicity created by the T345A replacement may interfere with the normal displacement of M4, thereby weakening the coordination of K ϩ ions at site II, consequently decreasing the intrinsic binding affinity for extracellular K ϩ .
There is an important difference in the nature of the functional change affected by T345A and mutations whose primary effect is a shift in steady state E 1 /E 2 poise in favor of E 1 . One example is an E233K substitution in ␣1 (34). Whereas E233K and T345A replacements result in similar changes in KЈ ATP and K ϩ sensitivity to Na ϩ -ATPase measured at micromolar ATP, E233K has no effect on K 0.5(K ϩ ) but markedly shifts the E 1 /E 2 poise toward E 1 (as seen in the 250-fold increase in I 50 for vanadate) and significantly decreases catalytic turnover. The T345A substitution, on the other hand, increases K 0.5(K ϩ ) but does not alter catalytic turnover and minimally shifts the E 1 /E 2 poise as evidenced in the relatively small (ϳ3-fold) effect of vanadate. Our analysis strongly supports the notion that the T345A mutation primarily alters intrinsic interactions of K ϩ with its binding pocket.
Here we report that the T345A missense FHM2 mutation leads to a fully active but functionally altered pump with reduced affinity for K ϩ . Although this is a firm, physiological foundation for understanding how this mutation may lead to disease, the challenge for the future is to integrate the consequences of functional pump lesions with the consequences of the large number of FHM1 mutations now recognized in the CACNA1A calcium channel gene, both of which produce a nearly identical phenotype in humans.