An N-terminal Sequence Targets and Tethers Na+ Pump α2 Subunits to Specialized Plasma Membrane Microdomains*

Sodium pumps (αβ dimers) with the α1 isoform of the catalytic (α) subunit are expressed in all cells. Additionally, most cells express Na+ pumps with a second α isoform. For example, astrocytes and arterial myocytes also express Na+ pumps with the α2 isoform. The α2 pumps localize to plasma membrane (PM) microdomains overlying “junctional” sarco-/endoplasmic reticulum (S/ER), but the α1 pumps are more uniformly distributed. To study α2 targeting, we expressed α1/α2 and α2/α1 chimeras and 1-90 and 1-120 amino acid N-terminal peptides in primary cultured mouse astrocytes. Immunocytochemistry revealed that α2/α1 (but not α1/α2) chimeras markedly reduced native α2 (i.e. were “dominant negatives”). N-terminal (1-120 and 1-90 amino acids) α2 (and α3), but not α1 peptides also targeted to the PM-S/ER junctions and were dominant negative for native α2 in astrocytes and arterial myocytes. Thus α2 and α3 have the same targeting sequence. Ca2+ (fura-2) signals in astrocytes expressing the 1-90 α2 peptide were comparable to signals in cells from α2 null mutants (i.e. functionally dominant negative): 1 μm ATP-evoked Ca2+ transients were augmented, and 100 nm ouabain-induced amplification was abolished. Amino acid substitutions in the 1-120 α1 and α2 constructs, and in full-length α1, revealed that Leu-27 and Ala-35 are essential for targeting/tethering the constructs to PM-S/ER junctions.

Because ␣1 and ␣2 are differently distributed in the PM of astrocytes and other cell types, the two isoforms must be targeted and tethered by different mechanisms. There is no information on the sorting of ␣2, but immunocytochemical and co-immunoprecipitation data indicate that it is tethered to a complex containing the cytoskeletal protein, ankyrin-B (13,16). In epithelia, ␣1 also is tethered to an ankyrin (12,17), probably ankyrin-G (18).
The sorting of Na ϩ pump ␣1 and the homologous H ϩ , K ϩ -ATPase ␣ subunit have been studied extensively in polarized intestinal and renal epithelia. Here, this Na ϩ pump localizes to the basolateral membrane, and the H ϩ , K ϩ -ATPase sorts to the apical membrane (19). The sorting signal apparently involves a region of transmembrane helix 4 (M4; see Fig. 1A) and the flanking extracellular and cytoplasmic domains (19).
The precise role of the small, highly glycosylated ␤ subunits is uncertain. Experiments on Na ϩ pump ␣1 indicate, however, that association with the ␤ subunit is required to chaperone ␣ to the PM, as well as for the functional maturation of the pump (20). The ␤1 subunit (three ␤ isoforms have been identified (2)) interacts with a highly conserved amino acid (aa) sequence, SYGQ, in the extracellular loop between M7 and M8 (21) (aa 896 -899 in mouse ␣1; see Fig. 1A). This sequence is present in most ␣ subunit isoforms from chickens (21) to mice and humans.
A sorting mechanism based on interaction with ␤ cannot explain the differences between ␣1 and ␣2 localization, because both ␣1 and ␣2 can co-assemble with either ␤1 or ␤2 (10,22,23). This promiscuous coassembly may be a consequence of the aforementioned conservation of the SYGC sequence in the ␣ isoforms.
In some epithelia, cycling of ␣1 subunits into the PM may be governed by protein kinase C-mediated phosphorylation, and several N-terminal poten-tial phosphorylation sites have been recognized (24,25). These sites are, however, identical in the mouse ␣1, ␣2, and ␣3 isoforms and, thus, cannot explain isoform-specific targeting.
Here we report that ␣1 and ␣2 Na ϩ pumps are sorted by different mechanisms in astrocytes. The ␣2 sorting signal is located in the N-terminal cytoplasmic segment of the molecule. This selective sorting and membrane insertion of ␣2 apparently does not depend upon phosphorylation or upon association with ␤.

EXPERIMENTAL PROCEDURES
Primary Cultured Mouse Astrocytes-Mice with null mutations in both Na ϩ pump ␣2 alleles (homozygous knockouts, ␣2 Ϫ/Ϫ or "KO") were generated by mating heterozygotes (15). Standard PCR methods were used for genotyping genomic DNA from fetus tails. Cortical protoplasmic type-1 astrocytes were prepared from WT and KO mouse fetuses on days 18 or 19 (E18 -19) as described (14,26). The cells were plated onto poly-L-lysine-coated 25-mm glass coverslips (ϳ50,000 cells/coverslip) for immunocytochemistry and Ca 2ϩ imaging, or onto 10-cm Petri dishes for immunoblots. Experiments were performed on subconfluent cultures on days 7-9 in vitro. All animal protocols were approved by our Institutional Animal Care and Use Committee.
Primary Cultured Rat Arterial Myocytes-Arterial myocytes were dissociated from the mesenteric arteries of male Sprague-Dawley rats (100 -150 g). The cells were plated onto 25-mm coverslips and were grown in primary culture for 4 -6 days as described (27).
To transfect the ␣1/␣2 and ␣2/␣1 chimera constructs into astrocytes, the cultured cells were treated with purified adenoviral vector solutions. A multiplicity of infection of 10 plaque forming units/cell induced ϳ80% transfection of cells on coverslips in 48 h. Transfection efficiency was even higher (Ͼ80 -90%) for cells in 10-cm dishes used for immunoblots. All cells were studied after 48 h.
In a few experiments, we transfected cells with plasmids containing normal or mutated, full-length ␣1 with a C-terminal FLAG tag and the Ca 2ϩreporter protein, GCaMP2 (28,29). A FLAG epitope was also, in some cases, inserted at Pro-120 in the first extracellular loop (Fig. 1A). The sitespecific mutations (see "Results") were generated with "QuikChange" (Stratagene, La Jolla, CA) using the manufacturer's directions, and the constructs were cloned into the pIRES2-DsRed2 vector (Clontech, Mountain View, CA) to facilitate visualization of transfected cells before immunostaining. These constructs were transfected using Lipofectamine 2000 (see next section). The transfected cells expressed the ␣1f/GCaMP2 construct and DsRed2 independently and simultaneously.
Lipofectamine 2000 (Invitrogen) was used to transfect the fusion protein vectors and truncated ␣ subunits into astrocytes and arterial myocytes. We used 4 g of DNA plus 8 l of Lipofectamine in 2.5 ml of culture medium per 25-mm coverslip and 24 g of DNA plus 60 l of Lipofectamine in 13 ml of culture medium per 10-cm dish. The cells were incubated for 48 h (5% CO 2 , 37°C) and then assayed for transgene expression by immunocytochemistry or GFP and Ca 2ϩ imaging. Approximately 25-35% of the cells on coverslips were transfected by this method; transfection efficiency was ϳ50% for cells cultured in 10-cm dishes and used for immunoblotting.
Immunoblot Analysis of Na ϩ Pump ␣ Subunit Isoforms and Constructs-Mouse WT and KO astrocytes were cultured in 10-cm dishes for 2 weeks. In some instances the cells were transfected with Na ϩ pump ␣ subunit constructs as described below. The cells were harvested, and the membrane proteins were prepared and analyzed by immunoblotting using published methods (13,14). Isoform-specific monoclonal or polyclonal antibodies raised against the Na ϩ pump ␣2 subunit (Fig. 1D) and anti-FLAG antibodies (Fig. 1E) were employed for these studies. For quantitative assessment (14), ␣ subunit bands were normalized with the glia-specific protein, GFAP (glial fibrillary acidic protein).
Cells on coverslips were imaged with a Nikon Diaphot inverted microscope (Nikon Corp., Melville, NY) equipped with a long working distance PlanApo 60X water immersion objective lens (numerical aperture 1.2) and a CoolSNAP charge-coupled device camera (Photometrics, Tucson, AZ). Illumination was provided by a diffraction grating-based Till Photonics Polychrome II illumination system (Applied Scientific Instrumentation, Inc., Eugene, OR). Images were acquired and analyzed with a Meta Imaging System (Universal Imaging, West Chester, PA).
Identification of Expressed Proteins-Na ϩ pump ␣ subunit isoformselective antibodies as well as the GFP-and the FLAG tags were used to detect the expressed protein constructs. Monoclonal antibodies directed against cytoplasmic N-terminal peptide sequences in ␣1 (McK1 (30)) and ␣2 (McB2 (31)) and polyclonal antibodies directed against peptide sequences in the large cytoplasmic loop between M4 and M5 (11) in ␣1 (NASE) and ␣2 (HERED) were employed. In non-permeabilized cells, the extracellular C-terminal FLAG tag of the 120-aa peptide (Fig. 1, A and C), or a FLAG tag inserted between amino acids 120 and 121 in the full-length protein, was used to determine insertion into the PM.
N-terminal GFP-tagged constructs were also expressed in astrocytes. In live imaging experiments, these constructs could be visualized FIGURE 1. Na ؉ pump transmembrane organization; ␣ subunit N-terminal sequences; fusion, chimera, and truncation constructs; and construct expression. A, diagram of the Na ϩ pump ␣ and ␤ subunits. B, comparison of mouse Na ϩ pump ␣1, ␣2, and ␣3 isoform and rabbit skeletal muscle SERCA1a N-terminal sequences. Numbering is based on the mature ␣1 protein sequence; the first five residues in ␣1 and ␣2, # Ϫ5 to # Ϫ1, are absent in the mature proteins. The ␣3 N terminus is displayed with Pro-7 aligned in all three isoforms. Helices 1 and 2 (aa in green) refer to published data (5,6); amino acids highlighted in boxes are identical in ␣2 and ␣3, but differ in ␣1. C, diagrams and nomenclature of Na ϩ pump ␣ subunit fusion protein, chimera, and truncation constructs. D, Western blots of ␣1/␣2 and ␣2/␣1 chimeras transfected into WT and ␣2 Ϫ/Ϫ (KO) astrocytes. Expression of the N-and C-terminal portions were detected with, respectively, two different ␣2-selective antibodies, McB2 and anti-HERED (panel A shows epitope location). All lanes were loaded with 5 g of protein except KO with ␣2(1-130)/␣1f and ␣2(1-333)/␣1f, which contained 18 g of protein. E, Western blots of N-terminal ␣2-truncation constructs identified with anti-FLAG antibody. All lanes were loaded with 10 g of protein.
directly. Thus, following fura-2 loading, Ca 2ϩ signaling in transfected and non-transfected cells could be compared on a single coverslip.
The ratio of fura-2 fluorescent emission (510 nm) at two excitation wavelengths (340 and 380 nm was used to calculate [Ca 2ϩ ] (14). GFP fluorescence was excited at 488 nm and emitted at 525 nm. The imaging system was based on a Nikon Eclipse 2000 inverted microscope equipped with a UV-Fluor 40ϫ (oil) objective lens and a Hamamatsu ORCA-ER charge-coupled device camera (Hamamatsu Photonics, Bridgewater, NJ). Illumination was provided by a Sutter DG-4 filter changer (Sutter Instruments, Novato, CA). Images were acquired and analyzed with a Meta Imaging System.
Statistical Analysis-Summarized fura-2 imaging ([Ca 2ϩ ]) data are presented as means Ϯ S.E.; n is the number of cells studied. Data comparisons were made with two-way analysis of variance.

RESULTS
The Na ϩ pump ␣2 isoform is localized to PM microdomains that overlie the ER, whereas the ␣1 isoform is more uniformly distributed in several different cell types (9,32). This differential distribution is readily apparent when wild-type (WT) astrocytes are double-labeled with ␣1 and ␣2 isoform-specific antibodies ( Fig. 2A). Based on this observation, our goal was to identify the one or more regions of the Na ϩ pump ␣2 subunit (Fig. 1, A and B) that are involved in targeting and tethering this subunit isoform to its appropriate PM location in astrocytes.
Detection of Native ␣ Subunit Isoforms and Transfected Constructs-To examine the difference between ␣1 and ␣2 targeting, we constructed ␣1/␣2 and ␣2/␣1 isoform chimeras and truncated segments ( Fig. 1, C-E). To facilitate identification, these proteins were constructed with an N-terminal-fused green fluorescent protein, GFP (G), and/or a C-terminal FLAG tag (f). As we shall see, these tags apparently did not interfere with the targeting, tethering, or function of the constructs. Transfected and native (endogenous) ␣ subunits also could be identified with antibodies raised against isoform-specific epitopes located at the N terminus or in the large cytoplasmic loop between trans-membrane helices 4 and 5 (Figs. 1A, 1D, and 2A) (11,30,31). These epitopes were especially useful for studies of transfected astrocytes from ␣2 Ϫ/Ϫ (KO) mice with no native ␣2, or for detecting repression of endogenous ␣2 expression.
Western blots probed with anti-␣2-selective antibodies (McB2 and HERED; see Fig. 1A) show the expressed proteins in WT and KO cells In both cases, the construct distribution pattern is reticular, similar to that of SERCA2b (H). The similarity between the stained structures in the left and right panels in H is readily apparent. K, expression of native ␣2 detected with anti-HERED in a non-transfected WT cell. The distribution of this epitope, too, is reticular. All scale bars ϭ 10 m. (Fig. 1D). Both antibodies detected ␣2 in non-transfected and fulllength ␣1f-transfected WT cells, but not in KO cells. Both antibodies also detected ␣2 in KO cells transfected with full-length ␣2f. When KO cells were transfected with ␣1(1-130)/␣2f or ␣1(1-333)/␣2f, anti-HERED, but not McB2 antibodies, cross-reacted with the membrane proteins (Fig. 1D). Conversely, when KO cells were transfected with ␣2(1-130)/␣1f or ␣2(1-333)/␣1f, McB2, but not anti-HERED, antibodies cross-reacted with the membrane proteins (Fig. 1D). These results are expected because of the different locations of the McB2 and anti-HERED antibody epitopes (Fig. 1A).
The ␣2/␣1 chimeras appeared to be expressed primarily in a reticular (honeycomb) pattern (e.g. Fig. 2H), similar to the distribution of native ␣2 (Fig. 2, A (panels a and c) and K (32)). This distribution resembled the distribution of SERCA2b in the same cells (Fig. 2H), but overlays (not shown) indicated a broader distribution of the ␣2/␣1 chimeras. This suggests that these chimeras are expressed, in part, in PM microdomains that overlie "junctional" (sub-PM) ER (32) or are confined within the ER. In contrast, the ␣1/␣2 chimeras appeared to be more uniformly distributed in the PM (Fig. 2, B and D), reminiscent of the distribution of native ␣1 (Fig. 2A, panel b (32).
These results raise the possibility that the ␣2/␣1 chimeras may be sorted and tethered to the normal ␣2 as well as ␣1 distribution sites. Accordingly, the ␣2 sorting sequence may be contained within the N-terminal 1-130 aa segment. In subsequent studies, the dominant negative assay with anti-HERED antibodies (Fig. 2G) was used to identify the ␣2 sorting sequence.
Expression of ␣2 N-terminal Segments-The fact that the N-terminal ␣2 chimeras were dominant negative for native ␣2 raised the possibility that the N-terminal segment, alone, might have a similar effect. To explore this possibility, the effects of expressing ␣1 and ␣2 N-terminal 1-90 and 1-120 aa peptides (Fig. 1, C and E) in WT cells were examined.
An even shorter construct, ␣2(1-90)f, with no transmembrane helices ( Fig. 1, A, C, and E) also was dominant negative for native ␣2 (Fig. 4B, panels a and aЈ, arrowheads). The implication is that a sequence within the N-terminal 90 amino acids preferentially binds to an appropriate "partner" and thereby displaces native ␣2, which is then retrieved and degraded.
As a control for the role of the N-terminal sequence in sorting, the entire initial cytosolic segment (i.e. the N terminus, through Gln-90, Fig.  1, A and B) was deleted from ␣2 (ϭ⌬N(1-90)␣2f). This peptide, which contains all ten transmembrane helices, also was expressed in WT astrocytes (detected with anti-FLAG antibodies). ⌬N(1-90)␣2f, failed to act as a dominant negative for the expression of native ␣2 (Fig. 4B,  panels b and bЈ, arrowhead). In this case, native ␣2 was detected with McB2 antibodies, because ⌬N(1-90)␣2f contains the HERED epitope but lacks the McB2 epitope (Fig. 1A). This is further evidence that only a portion of the N-terminal segment plays an essential role in targeting and tethering of ␣2; apparently, no other parts of the molecule are required.
The immunocytochemical data in Figs. 2, 4A, and 4B were obtained in permeabilized cells, and expression of the constructs in the surface membrane could not be specifically assessed. To overcome this difficulty, immunocytochemistry on permeabilized and non-permeabilized cells was compared (Fig. 4, A-C). Here, we took advantage of the fact that the FLAG tag at the C termini of the 1-120 aa constructs are located in an extracellular domain that normally forms the loop between M1 and M2 (Fig. 1A).
An important question is whether these dominant negative effects are limited to astrocytes. Mouse artery myocytes also express ␣1 and ␣2, but not ␣3 Na ϩ pumps (34,35); the ␣2 Na ϩ pumps, but not ␣1, appear to play a critical role in the long term control of blood pressure and in hypertension (35)(36)(37). Therefore, we also expressed the 1-120 aa ␣1 and ␣2 peptides in primary cultured mouse mesenteric artery myocytes. Fig. 5 shows that ␣2(1-120)f (Fig. 5C, panels b and bЈ, arrowheads), but not ␣1(1-120)f (Fig. 5B, panels b and bЈ), was also dominant negative for native ␣2 expression in arterial myocytes. Neither peptide affected ␣1 expression (Fig. 5, B (panels a and aЈ) and C (panels a and aЈ)). Thus, this mechanism for targeting and tethering ␣2 subunits applies to other tissues as well.
agonist dose-response curves. This is, indeed, the case, as illustrated by the comparison of Ca 2ϩ transient responses to ATP in WT and ␣2 KO cells (Fig. 6A).
The dominant negative ␣2 constructs should have effects similar to ␣2 KO or low dose ouabain. This is exemplified in WT astrocytes trans-fected with the G␣2(1-90) construct (Fig. 1C). Fig. 6B (panel a) shows a field with two representative astrocytes: The cell on the right was transfected (indicated by the presence of GFP), whereas the cell on the left was not (Fig. 6B, panel b). ATP (0.1 M) induced a small Ca 2ϩ transient (measured with fura-2) in the non-transfected cell (Fig. 6B, panel c, top FIGURE 5. Expression of truncated ␣1 and ␣2 constructs in permeabilized rat mesenteric artery myocytes. A, non-transfected myocytes: cross-reactivity with anti-NASE (a) and anti-HERED (b) antibodies indicates that these cells express both ␣1 and ␣2 Na ϩ pumps. B, myocytes transfected with ␣1(1-120)f: both transfected cells (detected with anti-FLAG antibodies) and nontransfected cells express native ␣2 (bЈ) as well as ␣1 (aЈ). C, myocytes transfected with ␣2(1-120)f: transfected cells (a and b) express ␣1 (aЈ), but are dominant negative for ␣2 (bЈ, arrowhead). All scale bars ϭ 10 m. . This effect is similar to that observed with low dose (100 -500 nM) ouabain in WT astrocytes (33). These low doses of ouabain block only ␣2 Na ϩ pumps in WT rodent astrocytes. Data are from 10 coverslips (5 WT and 5 KO cells; 2-3 coverslips/mouse). In B: a, two adjacent cells from a preparation treated with Lipofectamine 2000 and G␣2(1-90). Scale bar ϭ 30 m. b, only the cell on the right was transfected, as indicated by the GFP fluorescence. c, data recorded from red and green boxed areas in B, panel a. The non-transfected cell (red record) exhibited a small cytosolic Ca 2ϩ transient in response to 0.1 M ATP; the response to this dose of ATP was greatly augmented by prior exposure to 100 nM ouabain for 1 min. The transfected cell (green record), in which ␣2 should be down-regulated (Fig. 4A, panel dЈ), exhibited a large Ca 2ϩ transient in response to 0.1 M ATP and no augmentation by 100 nM ouabain (as if the ␣2 was already inhibited). Data are representative of results from seven coverslips; each coverslip was transfected separately. C, dose-response curve showing the effect of G␣2(1-90) transfection on the response to ATP. As is true of low dose ouabain and ␣2 KO, G␣2(1-90) augments the Ca 2ϩ transients induced by ATP (p Ͻ 0.01 by two-way analysis of variance; n ϭ 43 WT cells and 20 cells transfected with G␣2(1-90)). Data are from six coverslips (three/mouse); each coverslip was transfected separately. D, restoration of the low dose ouabain effect in an ␣2 KO cell transfected with full-length G␣2f. a, two nearby ␣2 KO cells on a coverslip. The cell on the left was transfected with G␣2f, as indicated by the GFP fluorescence (b). Scale bar (b) ϭ 30 m. c, data recorded from red and green boxed areas in D, panel a. The non-transfected cell (red record) exhibited a relatively large response to 0.1 M ATP that was not augmented by 100 nM ouabain. The transfected cell (green record) exhibited a smaller response to 0.1 M ATP, but considerable augmentation when the ATP was applied 1 min after 100 nM ouabain. Data are representative results from six coverslips (three/mouse); each coverslip was transfected separately. panel). When this cell was treated with 100 nM ouabain, the response to 0.1 M ATP was greatly augmented. In contrast, the transfected cell exhibited a large Ca 2ϩ transient in response to 0.1 M ATP (as if ␣2 had already been inhibited), and the response was not augmented by 100 nM ouabain (Fig. 6B, panel c, bottom panel). Summarized data (Fig. 6C) indicate that the response to ATP was significantly augmented in cells transfected with G␣2(1-90), in which native (functional) ␣2 expression was, presumably, markedly reduced. Also, the usual augmentation of the ATP-evoked response by 100 nM ouabain in WT cells was attenuated in transfected cells (Fig. 6B, panel c, bottom panel versus top panel). This is expected in cells in which expression of the high ouabain affinity ␣2 receptor is reduced. (Recall that in rodents, ␣1 has a very low affinity and should not respond to 100 nM ouabain (2,14,35).) Thus, the effects observed in G␣2(1-90)-transfected cells were comparable to the results observed in ouabain-treated (i.e. ␣2-inhibited) cells or in astrocytes from ␣2 Ϫ/Ϫ mice. Clearly, the reduced ␣2 function in the cells expressing G␣2(1-90) reflects the dominant negative effect of the nonfunctional, N-terminal ␣2 peptide on ␣2 expression.
It is important to know if the N-and C-terminal fusion peptides, i.e. GFP and FLAG, alter ␣2 function. This was tested by transfecting G␣2f (with the complete ␣2 sequence; Fig. 1C) into KO astrocytes. Fig. 6D, panel a, shows two representative KO astrocytes. The one on the left was transfected and expressed G␣2f, as indicted by the GFP fluorescence in panel b. The response of the non-transfected cell to 0.1 M ATP was not augmented by 100 nM ouabain (Fig. 6D, panel c, top panel). On the other hand, the transfected cell exhibited a greatly augmented response to 0.1 M ATP when treated with 100 nM ouabain (Fig. 6D,  panel c, bottom panel). This indicates that G␣2f was able to "rescue" the low dose ouabain effect in ␣2 KO cells. Because ouabain is hydrophilic, the response to ouabain implies that G␣2f was expressed in the PM. Thus, the N-terminal GFP and the C-terminal FLAG tag apparently do not interfere with normal Na ϩ pump sorting and function in the G␣2f construct.
Expression of ␣3 N-terminal Segments-There is evidence that the expression of Na ϩ pumps with ␣2 and ␣3 subunits is mutually exclusive. For example, neonatal rat cardiac muscle expresses ␣1 and ␣3, whereas ␣1 and ␣2 are present in adult cardiac muscle (39). In neurons, the Na ϩ pump ␣3 isoform is expressed with a distribution comparable to that of ␣2 in astrocytes: i.e. ␣3 is confined to PM microdomains that overlie the sub-PM ER, and it co-localizes with SERCA (9,32). Neuronal ␣3 and astrocyte ␣2 both co-immunoprecipitate comparable cytoskeletal and ER proteins (13). Moreover, the N-terminal sequences of ␣2 and ␣3 are very similar (Fig. 1B). For these several reasons, we speculated that the ␣3 isoform might have the same sorting signal as ␣2, but a different tissue-specific promoter. Therefore, the effect of transfection with ␣3(1-120)f was tested on the expression of ␣2 in astrocytes. As shown in Fig. 7A, panels a-c, ␣3(1-120)f is expressed in astrocytes, and is distributed in a reticular pattern similar to that of SERCA2b (panels c and cЈ insets, and see Fig. 7B, panel a inset). Indeed, the ␣3 N-terminal segment, like its ␣2 counterpart, markedly reduced native ␣2, but not ␣1 expression (Fig. 7A, panel bЈ versus aЈ). Also, ␣3(1-120)f experiments on non-permeabilized cells demonstrate that ␣3(1-120)f is expressed in the PM (Fig. 7B, panel a). The similar distribution patterns of ␣3(1-120)f and SERCA2b (Fig. 7A, panel c and cЈ), therefore, indicates that they are expressed in different but adjacent membranes. The implication is that the targeting and tethering mechanisms for ␣2 and ␣3 are identical or very similar.
Alteration of Targeting by Site-specific Mutations in the N Termini of ␣1 and ␣2 Subunits-This similarity between the effects of the ␣2 and ␣3 N-terminal constructs, led us to contrast their amino acid sequences with that of the ␣1 N terminus. At only three positions are the ␣2 and ␣3 N termini identical, and different from ␣1: residues 27, 35, and 64 (using the numbering for mature ␣1, Fig. 1B). To test the possibility that these specific amino acid residues (aa) play a critical role in ␣2 targeting and tethering, each of the three residues in the ␣2(1-120)f construct was mutated, one at a time, to the corresponding aa in ␣1, i.e. L27M, A35S, and Q64A. Astrocytes were then transfected with these mutated constructs (e.g. ␣2(1-120, L27M)f). The results are clear: mutations L27M and A35S (Fig. 8A, panels b and c, arrowheads in lower panels), but not Q64A (Fig. 8A, panel d, arrowhead), abolished ("knocked out") the dominant negative activity of the N-terminal ␣2 construct. Thus, both Leu-27 and Ala-35 are essential for ␣2 targeting.
The complementary experiment was then performed: the aforementioned three ␣2/␣3 amino acids were mutated, one at a time and together, into ␣1(1-120)f. When these constructs were then transfected into astrocytes (Figs. 8B and 9), only the two ␣1(1-120)f peptides that contained both the M27L and S35A mutations were dominant negative for ␣2 (Fig. 8B, panels d and f, arrowheads in lower panels); the A64Q mutation was not needed (Fig. 8B, panel e). Furthermore, the two ␣1(1-120)f constructs with both M27L and S35A were sorted to the PM as indicated by the detection of the FLAG epitope (Fig. 9B, panels a and b), but not the SERCA2b epitope (Fig. 9B, panel aЈ), in non-permeabilized cells (Fig. 9B shows the three-mutation construct). Also, like native, full-length ␣2 and ␣2(1-120)f (Fig. 4A, panel e), ␣1(1-120, M27L, S35A, and A64Q)f distributed in the PM in a reticular pattern and co-localized with SERCA2b in the underlying ER (detected following permeabilization: Fig. 9A (panel c) and B (panel d)). The introduction of dominant negative activity into ␣1(1-120)f with the mutations, M27L and S35A, and the sorting to the PM microdomains that overlie adjacent (junctional) ER, confirm the key role of Leu-27 and Ala-35 in the targeting and tethering of ␣2.
Expression of Wild-type and Mutated Full-length ␣1 Constructs-We also tested full-length ␣1 with both the M27L and S35A mutations on ␣2 expression. This construct should contain the normal ␣1 targeting The ␣3 isoform of the Na ϩ pump ␣ subunit, which is highly homologous to both the ␣1 and ␣2 isoforms (Fig. 1B), is normally expressed in neurons, but not in astrocytes (8,10,30). Nevertheless, as shown here, the truncated construct, ␣3(1-120)f, can be expressed in transfected WT astrocytes and detected with anti-FLAG antibodies (a-c). This construct was dominant negative for native ␣2 (bЈ, arrowheads), but not native ␣1 (aЈ). The reticular distribution of ␣3(1-120)f (c, inset) parallels that of SERCA2b (cЈ, inset). B, non-permeabilized cells: the cross-reactivity with anti-FLAG antibodies (a) implies that the ␣3(1-120)f construct is inserted into the PM, with the FLAG epitope in the extracellular domain. Inset (enlargement of the boxed area) shows the reticular (honeycomb) distribution of ␣3(1-120)f. Lack of cross-reactivity with anti-NASE (aЈ) or anti-SERCA2b (b), which have intracellular epitopes, indicates that these cells are not permeabilized. All scale bars ϭ 10 m. sequence (see "Discussion") as well as the ␣2 targeting sequence. In this case, we used an ␣1f construct with a C-terminal Ca 2ϩ -sensitive fusion protein (GCaMP2) that is expressed in the PM with the same broad distribution as native ␣1 (29). Cells transfected with the construct containing the WT ␣1 sequence (detected with anti-FLAG antibodies), like non-transfected cells, expressed the HERED (␣2) epitope (Fig. 10A); i.e. the normal ␣1 sequence was not dominant negative for ␣2. In contrast, cells transfected with mutated full-length ␣1 constructs containing the ␣2 amino acids, Leu-27 and Ala-35, did not express the HERED epitope (arrowheads in Fig.  10, B and C, panel a). Thus, even the full-length ␣1 with the M27L and S35A mutations was dominant negative for native ␣2.
In some experiments, a FLAG epitope was inserted into the mutated, full-length ␣1 at Pro-120 in the extracellular loop between transmembrane helices M1 and M2. This epitope cross-reacted with anti-FLAG antibodies in non-permeabilized, transfected cells (Fig. 10, C (panel b)  and D (panels a and bi)). In contrast, the intracellular NASE (␣1)   (a and b), but not the SERCA2b epitope (aЈ), is accessible in these non-permeabilized cells. Following cross-reactivity with anti-FLAG antibodies (b), some coverslips were treated with Brij 58 to permeabilize the cells ("Experimental Procedures"); these cells were then cross-reacted with anti-SERCA2b antibodies (bЈ) to identify the ER and test for co-localization. Insets in A (panel a) and B (panels a and b) (enlargements of the boxed areas) indicate that this mutated ␣1 construct is distributed in a reticular pattern similar to that of SERCA2b (A, panel aЈ and B, panel bЈ). Panels c and cЈ are pseudocolor images (green ϭ anti-FLAG; red ϭ anti-SERCA2b) of enlarged boxes from b and bЈ, respectively; the yellow and yellow/green areas in the overlay (d) indicate regions of overlap between the two epitopes. All scale bars ϭ 10 m. epitope, present in all cells (e.g. Figs. 4A (panels aЈ and cЈ), 5A (panel a), and 5C (panel aЈ)) and included in this mutated ␣1 construct (Fig. 1A), was not detected in the non-permeabilized cells (Fig. 10C, panel b). Thus, this full-length construct, ␣1f 120 (M27L,S35A)f/GCaMP2, was expressed in the PM.
The intracellular SERCA2b epitope also did not cross-react with anti-SERCA2b antibodies in non-permeabilized cells (Fig. 10D, panel a), but it did cross-react after permeabilization with Brij 58 (Fig. 10D, panel bii). The SERCA2b stain distributed in a typical reticular pattern, as expected for this ER-specific label (see Figs. 4A (panel eЈ) and 9A). The extracellular FLAG stain was not limited to this reticular pattern but appeared to be more widely distributed in all cells in which a distinct reticular SERCA2b distribution was observed (Fig. 10D, panel bi, shows an example). This differs from all the truncated ␣2 and mutated, truncated ␣1 constructs that were dominant negative for native ␣2 (Figs. 4, 5, 8, and 9): their distribution invariably mimicked the subjacent ER pattern (e.g. Figs. 4A, panel e, and 9A). The more ubiquitous PM distribution as well as the dominant negative effect of ␣1f 120 (M27L,S35A)f/ GCaMP2 can be explained if this construct contains both the ␣1 and ␣2 sorting signals.

DISCUSSION
N-terminal Constructs of ␣2 Are Dominant Negatives-Na ϩ pumps with ␣2 subunits in astrocytes are localized to PM microdo-mains that overlie junctional ER (32). Here, they are part of a complex that includes the Na ϩ /Ca 2ϩ exchanger and certain cytoskeletal and ER proteins (13). By functionally coupling with the Na ϩ /Ca 2ϩ exchanger and SERCA, these Na ϩ pumps help regulate Ca 2ϩ homeostasis and Ca 2ϩ signaling in astrocytes (14). This was confirmed by functional studies (Fig. 6), which show that knockdown of ␣2 by a dominant negative construct augments Ca 2ϩ signaling. To elucidate, further, the central role of the ␣2 Na ϩ pumps in astrocyte function, we examined the mechanism of targeting and tethering these Na ϩ pumps to their appropriate location in the PM.
One method that has been employed successfully to study differences in the targeting of membrane proteins with sequence homologies involves the use of chimeras (19). Application of this method to Na ϩ pump ␣1 and ␣2 isoforms not only indicated that the ␣2 N terminus was important for targeting but, unexpectedly, also revealed that chimeras with an ␣2 N-terminal were dominant negative for native ␣2. We then employed this "dominant negative assay" to identify two amino acids, Leu-27 and Ala-35, in the N-terminal segment of the Na ϩ pump ␣2 subunit that are essential for the targeting and tethering of this subunit to its proper location in the PM. The ␣1 isoform in these cells is targeted by different but as yet unresolved mechanisms.
The observations, that ␣1(1-120, M27L, S35A Ϯ A64Q) sorted to the PM microdomains that overlie junctional ER, co-localized with SERCA and was dominant negative for native ␣2 (Fig. 9), confirm that the ␣2 N FIGURE 10. Expression of native ␣2 in WT astrocytes transfected with WT and mutated fulllength ␣1 constructs. A and B, astrocytes were transfected with: ␣1f/GCaMP2 containing the fulllength, native ␣1 sequence (A), or ␣1(M27L,S35A)f/ GCaMP2, in which Met-27 and Ser-35 were mutated to the corresponding ␣2 amino acids, Leu and Ala, respectively (B). The transfected cells, detected with anti-FLAG antibodies, did not express the HERED epitope (arrowheads in B); i.e. ␣2 expression was knocked down. C and D, astrocytes were transfected with ␣1f 120 (M27L,S35A)f/GCaMP2, in which a (second) FLAG epitope was inserted between Pro-120 and Pro-121 in the extracellular loop between M1 and M2 (see Fig. 1A). In C: a, permeabilized cells. The transfected cells, detected with anti-FLAG antibodies, did not express the HERED epitope (arrowheads); i.e. ␣2 expression was knocked down. C (panel b) and D (panels a and bi): non-permeabilized cells. Crossreactivity with anti-FLAG, but not anti-NASE (␣1; C, panel b) or anti-SERCA 2b (D, panel a) antibodies indicate that the construct was expressed in the PM. Nuclei were stained with 4Ј,6Ј-diamidino-2-phenylindole (DAPI) to identify the non-permeabilized cells. D, panel b. Non-permeabilized cells were crossreacted with anti-FLAG antibodies (D, panel bi) to identify the constructs expressed in the PM. The cells were subsequently permeabilized with Brij 58 ("Experimental Procedures") and were then cross-reacted with anti-SERCA2b antibodies (D, panel bii) to identify the ER. The insets and color overlay indicate that there is some overlap between the FLAG and SERCA2b labels, but the FLAG tag (green; bi*) is more widely distributed than the SERCA2b label (bii*). All scale bars ϭ 10 m.
terminus contains all of the targeting information for this isoform. These mutated, truncated ␣1 constructs were apparently excluded from the remainder of the PM (Fig. 9A, panels b and bЈ). In contrast, the full-length ␣1f 120 (M27L, S35A)f/GCaMP2, which also was dominant negative for native ␣2, should contain the ␣1 as well as ␣2 sorting signals, because the ␣1 sorting signal is not located at the N terminus (19). Thus, this construct apparently sorts to all areas of the PM (Fig. 10D,  panel bi). A similar explanation may account for the finding that the ␣2/␣1 chimeras did not strictly overlap with SERCA2b even though the distribution of the chimeras appeared to be predominantly reticular (e.g. Fig. 2, H and J).
The fate of the displaced native ␣2, following transfection with the dominant negative gene constructs, and whether new, full-length ␣2 is then synthesized, are unknown. The decline of cross-reactivity with anti-HERED antibodies implies that the native ␣2 is synthesized and either retained within the Golgi-ER network and degraded or is transported to the PM but is rapidly retrieved and degraded. The ␣2 N-terminal constructs are overexpressed and should be present in relatively high concentration. Thus, it seems reasonable to speculate that, by "mass action," they displace native ␣2 from their normal binding (tethering) sites on the cytoskeleton. One possibility is that ␣2 (or the N-terminal construct) is assembled with its "target protein" within the Golgi-ER network. The target protein (Fig. 11) may then chaperone ␣2 (or the construct) to its correct location at the PM-S/ER junction. The observations with ␣2(1-90)f (Fig. 4, B and C) demonstrate that PM insertion is not required for the dominant negative effect. We cannot be certain, however, whether this construct is exported from the ER or resides on the binding site at the PM-ER junction normally occupied by native ␣2.
Because all of our constructs were tagged with a C-terminal FLAG, and/or an N-terminal GFP fusion protein, it is important to ask whether these tags affected targeting and tethering or function. The evidence indicates that neither tag interfered with the normal distribution of the ␣2 constructs as illustrated by their dominant negative effects and correct localization (e.g. see Figs. 2H, 6B, 6D, and 10, B-D). Also, the ability of full-length G␣2f to "rescue" normal ouabain-sensitive ␣2 Na ϩ pump activity in KO cells (Fig. 6D) indicates that neither tag abolishes Na ϩ pump function, although we cannot rule out an alteration in kinetics.
Structure and Function of the ␣2 N-terminal Domain-Critical issues about ␣2 targeting and tethering concern the precise structure of the N-terminal and the protein(s) to which this terminal is tethered. The crystal structure of the Na ϩ pump ␣ subunit has not yet been determined, but the structure of the related P-type ATPase, SERCA, is known (6,40). Moreover, there are extensive sequence, domain, and mechanistic similarities between the Na ϩ pump ␣ subunit and SERCA. Indeed, homology modeling based on the SERCA structure has been very useful (5).
At the N terminus, the Na ϩ pump ␣1 subunit is 34 amino acids longer than SERCA1a (the skeletal muscle isoform). Moreover, sequence homology begins with His-39 (using ␣1 numbering) and the tetrapeptide, HKLS in ␣1 and ␣2, HKMS in ␣3, and HSKS in SERCA1a (Fig. 1B). The N terminus of SERCA contains two short helical segments that have homologous regions in ␣1, at amino acids 42-50 (helix 1) and 60 -70 (helix 2). Even the homologous 9-aa "break" between these helices has 33-44% sequence identity. These helices and the intervening break form part of the "activator" (A) domain in SERCA and, presumably, in the Na ϩ pump ␣ subunit. During the transport cycle, the A domain rotates into and out of the pocket between the nucleotidebinding (N) and phosphorylation (P) domains (Fig. 11). This rotation influences the opening and closing of the ion gates, because the A domain moves M1 (aa 91-112 in ␣1), to which it is attached by a 20-aa stalk (aa 71-90). It is the movement of M1, when the A domain pulls on the stalk, that apparently opens and closes the ion gates. This has been directly demonstrated for the Ca 2ϩ gates in SERCA (40) and likely also occurs in the Na ϩ pump ␣ subunit.
These homologies and presumed functional similarities between SERCA and the Na ϩ pump ␣ subunit suggest that N-terminal amino acids 39 -90 in the Na ϩ pump ␣ subunit are not involved in targeting and tethering. Indeed, this helps to explain why the ␣1-␣2 differences at aa 64 (Gln versus Ala within helix 2 in the A domain) does not affect targeting and tethering. Thus, it is not surprising that control of ␣2 targeting/tethering resides in a portion of the N-terminal first 38 amino acids. The fact that the critical Leu-27 and Ala-35 are 8 residues apart raises the possibility that these two amino acids are two turns apart on another helical segment in the Na ϩ pump that is lacking in SERCA. Our data demonstrate that these two residues are critically important in targeting/tethering ␣2 to its appropriate locus on the cytoskeleton. Fig.  11 shows a model that illustrates this relationship.
In epithelia, ankyrin binding to the ␣1 Na ϩ pump is required to "assemble and stabilize" this isoform at its location in the basolateral PM (41). This isoform apparently binds to both red cell ankyrin (ankyrin-R, a product of the Ank1 gene) and ankyrin-G (a product of the Ank3 gene (18,42)). The ankyrin attachment sites are located on both the M2-M3 and M4 -M5 cytoplasmic loops (18). Morrow and colleagues (41) suggest that ankyrin binds to a loop formed by ␣ subunit residues 152-158 (MESKFKNM) in the M2-M3 cytoplasmic domain. This sequence is extremely well conserved in vertebrate species (human, rat, mouse, dog, pig, and frog) and isoforms (identical in ␣1 and ␣3, and a conservative E-D substitution in ␣2). However, ␣2, in astrocytes, and ␣3, in neurons, complex with ankyrin-B, and not ankyrin-G (13). Moreover, as discussed above, in these cells ␣2 and ␣3 are localized to PM microdomains at PM-ER junctions, which is different from the distribution of ␣1 in these cells. And, as shown here, ␣2 (and, perhaps ␣3) is (are) targeted and tethered by an N-terminal sequence located in first 33 amino acids (␣2 is 2 amino acids shorter, and ␣3 is 10 amino acids shorter than ␣1; FIGURE 11. Block model of the ␣2 Na ؉ pump based on the crystal structure of SERCA1a. The model is drawn according to the structure of the calcium-bound form of SERCA1a (Fig. 2) (6). N, P, and A are the nuclear, phosphorylation, and activator cytoplasmic domains, respectively. A is attached to helix M1, one of the ten transmembrane helices in the M region. The red N-terminal segment, which is not homologous with SERCA (Fig. 1B), presumably contains a short helix that includes the critical Leu-27 and Ala-35 residues. This region may bind to the anchoring target protein presumably ankyrin-B. During Na ϩ pump activation, following phosphorylation of the P domain, the N and P domains separate, opening a space between them so that the A domain can rotate (broken arrow). This conformational change pulls the transmembrane helix M1 toward the cytoplasm and thereby opens the gate to permit Na ϩ ions to leave their binding sites within M, and exit to the extracellular fluid. see Fig. 1B). Whether this N-terminal sequence is an ankyrin-B attachment site remains to be determined.
In conclusion, our data and this analysis indicate that the 9-residue sequence, starting at position 27 (␣1 numbering), LDELKKEVA, targets and tethers ␣2 to its appropriate location in the PM. The sequence is identical in mouse, rat, and human ␣2. The ␣3 sequence in all three species differs by the conservative substitution of an Asp for the Glu at position 29.