The Na,K-ATPase α2 Isoform Is Expressed in Neurons, and Its Absence Disrupts Neuronal Activity in Newborn Mice*

Na,K-ATPase is an ion transporter that impacts neural and glial physiology by direct electrogenic activity and the modulation of ion gradients. Its three isoforms in brain have cell-type and development-specific expression patterns. Interestingly, our studies demonstrate that in late gestation, the α2 isoform is widely expressed in neurons, unlike in the adult brain, in which α2 has been shown to be expressed primarily in astrocytes. This unexpected distribution of α2 isoform expression in neurons is interesting in light of our examination of mice lacking the α2 isoform which fail to survive after birth. These animals showed no movement; however, defects in gross brain development, muscle contractility, neuromuscular transmission, and lung development were ruled out. Akinesia suggests a primary neuronal defect and electrophysiological recordings in the pre-Bötzinger complex, the brainstem breathing center, showed reduction of respiratory rhythm activity, with less regular and smaller population bursts. These data demonstrate that the Na,K-ATPase α2 isoform could be important in the modulation of neuronal activity in the neonate.

Na,K-ATPase is a plasma membrane enzyme necessary for maintaining the sodium and potassium ion gradients in the cell, and it drives the sodium-dependent transport of calcium and amino acids as well as the reuptake of neurotransmitters. The ion gradients generated by Na,K-ATPase are also used to regulate the volume of the cell and to support and modulate electrical activity through direct (electrogenic) and indirect effects on membrane potential. Na,K-ATPase is a heteromeric protein composed of an ␣ catalytic subunit that binds sodium and potassium ions, ATP, and cardiac glycosides, and ␤ and ␥ (FXYD) subunits that can modulate substrate affinity. There are different genes that code for multiple ␣, ␤, and ␥ isoforms. Four ␣ isoforms (␣1, ␣2, ␣3, and ␣4) have been identified, and all but ␣4 are expressed in the brain (1,2). Examination of the enzymatic properties of the ␣ and ␤ isoforms in different expression systems revealed that the ␣ isoforms have differences in substrate affinity and kinetic properties (3)(4)(5)(6)(7). In most adult mammals, the ␣2 isoform is expressed most abundantly in skeletal muscle and brain and in lower abundance in heart, adipocytes, and eye (8 -11). In situ hybridization performed on sections of embryonic days 9.5-16.5 mouse brain revealed that the Na,K-ATPase ␣2 isoform is expressed throughout most regions of the brain (12). In the adult brain it has been found in astrocytes, pia/arachnoid, and a few types of neurons (13)(14)(15).
To understand further the specific roles of individual Na,K-ATPase isoforms, we have analyzed mice in which the ␣2 isoform gene has been knocked out. The animals died shortly after birth but did not display obvious gross morphological defects in any tissue, including the brain. Lack of motor activity was significant, but muscle contractility was not found to be critically impaired. Consequently we investigated the cellular distribution of ␣2 in the newborn brain and the function of an intrinsic neuronal circuit that could contribute directly to immediate death: generation of the breathing rhythm.

EXPERIMENTAL PROCEDURES
Genotyping and Blood Analysis-Mice heterozygous for the Na,K-ATPase ␣2 isoform were generated as described previously (16). Heterozygous females were mated with heterozygous males, and the resulting offspring were genotyped by Southern blot as described previously (16). Blood was taken from decapitated mouse pups immediately after birth, collected in capillary tubes, and measured for carbon dioxide and oxygen content using a Chiron blood gas analyzer (Norwood, MA model 384). Blood glucose levels were analyzed using an Accudata GTS Glucose Test Station (Roche Molecular Biochemicals).
Lung Histology-Within 15-30 min after birth, newborn mouse pups were sacrificed, and the lungs were carefully removed and immersed in 10% formalin. The tissues were then embedded in paraffin and sectioned at 5 m. Sections were stained with hematoxylin and eosin, and digital pictures were taken using a microscope setting on ϫ10 magnification.
Microsome Preparation and Western Blot Analysis-Tissues from at least four embryonic day 18.5 pups of the same genotype (␣2 ϩ/ϩ , ␣2 ϩ/Ϫ , or ␣2 Ϫ/Ϫ ) were pooled and microsomes prepared as described (16). The pellet was resuspended in 1 mM imidazole, 1 mM EDTA, pH 7.4, then aliquoted and stored at Ϫ80°C. Protein concentration was determined using the BCA assay (Pierce Chemical Co.). The microsomal membranes were used for Western blot analysis. SDS-PAGE was performed as described (17). The Western blot procedure was performed as described (18).
Diaphragm Immunohistochemistry-For whole mount diaphragm preparations, diaphragms from embryonic day 18.5 mice were prepared as described previously (19). Diaphragms were fixed in 2% paraformaldehyde in phosphate-buffered saline (PBS), 2 blocked in 0.1 M glycine in PBS, and then permeabilized in 2% bovine serum albumin, 1% Triton X-100 in PBS (TPBS) overnight. The muscles were then incubated with rabbit antibodies to synaptophysin (Zymed Laboratories, San Francisco) and then incubated simultaneously with fluorescein-conjugated donkey anti-rabbit IgG (Jackson Immunochemicals, West Grove, PA) and tetramethylrhodamine-conjugated ␣-bungarotoxin (Molecular Probes, Eugene, OR) in 2% bovine serum albumin in TPBS overnight at 4°C. After washing in TPBS, diaphragms were mounted on coverslips in glycerol-paraphenylenediamine to retard fading, viewed with epifluorescence and filters that were selective for rhodamine or fluorescein, and evaluated with an Axionplan2 microscope (Zeiss, Thornwood, NY). Images were captured with a digital camera (Hamamatsu, Bridgewater, NJ) and imaging software (QED Imaging, Pittsburgh, PA).
Diaphragm Contractility-Diaphragms with ribs attached were removed from embryos (day 18.5) and placed in Krebs solution containing (in mM) 118 NaCl, 4.7 KCl, 25 NaHCO 3 , 2.5 CaCl 2 , 1.2 MgSO 4 , 1.2 NaH 2 PO 4 , 0.026 EDTA, and 11 glucose, equilibrated with 95% CO 2 and 5%O 2 . The diaphragm was cut to obtain a muscle strip from the central region with ribs attached to the tendon at each end. Triangular clips were attached at each end of the muscle strip, and the muscle was held in the clip against the ribs. Muscles were mounted in a constant temperature, sealed chamber and fixed to a stainless steel post at one end with the clip and the other end fixed to an isometric force transducer (Kistler Morse, Redmond, WA). The muscle length was adjusted to produce a resting tension of 3 millinewtons. The muscles were electrically field stimulated using two platinum electrodes positioned along either side of the muscle. Supramaximal voltage and frequency were determined empirically using a series of short (1-s) tetani and subsequently increasing the voltage or the frequency for each tetanus. The stimuli employed capacitor discharges of equal but alternating polarity (60 Hz at 15 V) with three to five instances of tetani or twitches at a duration of 2 ms. Digital recordings of force production were obtained with the BioPac data acquisition system (BioPac System, Inc., Goleta, CA) and evaluated to determine maximal twitch tension.
In Situ Hybridization-A pregnant wild-type mouse of the same genetic background as the ␣2 Ϫ/Ϫ mice was euthanized by carbon dioxide inhalation, and the embryonic day 18.5 pups were removed, decapitated, and the neck and head were immediately perfused and fixed in 4% paraformaldehyde (w/v) in PBS overnight. The tissues were cryoprotected and embedded and then cryosectioned in 6 -8-m-thick sections. The sections were dried on slides and then postfixed, prehybridized, hybridized, and developed as described previously (20,21). Antisense and sense RNA probes were synthesized with 35 S-labeled rUTP from plasmids that contain either ␣1 or ␣2 Na,K-ATPase isoformspecific sequences (9).
Immunofluorescence-Slices used for immunofluorescence were prepared in an ice-cold artificial cerebrospinal fluid containing (mM) 118 NaCl, 3 KCl, 1.5 CaCl 2 , 1 MgCl 2 , 25 HEPES, and 30 D-glucose, pH 7.4, which was not bubbled with gas. Most slices were transferred immediately into fixative, consisting of 2% paraformaldehyde in periodatelysine buffer (PLP fixative) (22). A few slices were maintained in icecold artificial cerebrospinal fluid for up to 40 min before fixing. Slices were immersed in 30% sucrose in PBS overnight and then embedded and frozen in TBS tissue freezing medium (Triangle Biomedical Sciences, Durham, NC) in aluminum boats. Cryostat sections (12-14 m) were picked up on ProbeOn Plus positively charged microscope slides (Fisher Scientific) and stored at Ϫ20°C until use. Unstained slides were warmed to room temperature, and a PAP pen (Kiyota International, Elk Grove, IL) was used to draw a hydrophobic ring around the sections. Slides were rinsed in PBS for 5 min, transferred to 95°C 10 mM sodium citrate, pH 6.0, in a Coplin jar standing in a boiling water bath, and incubated for 20 min. This antigen retrieval method enhanced the detection and specificity of the stain. The Coplin jar containing the slides was then removed from the bath and allowed to cool for 20 min. Slides were reequilibrated in several changes of room temperature PBS over 30 min. For all subsequent incubations, the slides were laid flat in a dark moist box. The sections were covered (ϳ50 l/section) with 5% normal goat serum in PBS with 0.3% Triton X-100 (PBSt) for 1 h at room temperature. This blocking solution was removed with an aspirator, and primary antibody McB2 was applied (1:4 dilution). The specificity of McB2, a monoclonal antibody specific for the ␣2 isoform of Na,K-ATPase, has been described previously (23). The sections were incubated overnight at 4°C in the primary antibody, rinsed in PBS (three times at 10 min each time), and then incubated in Cy3-conjugated goat anti-mouse IgG (1:300; Accurate, Westbury, NY) in PBSt for 2 h. Finally, they were rinsed in PBS and coverslipped in Vectashield fluorescence mounting medium (Vector Laboratories, Burlingame, CA). Slides were examined and images were collected on a Nikon TE300 fluorescence microscope equipped with a Bio-Rad MRC 1024 scanning laser confocal system, version 3.2.
Because we were using an anti-mouse secondary antibody on mouse tissue, there was light nonspecific staining of large cells and blood vessels. We tried using different blocking solutions, secondary antibodies (different host species, and different fluorophores), and using immunohistochemistry with horseradish peroxidase/diaminobenzidine (not shown), but the nonspecific stain was always visible in control sections (treated only with secondary antibodies, no primary antibody). Nonetheless, the cellular nonspecific stain could be differentiated easily from positive stain because it was very light and was only seen in the cytoplasm. Positive ␣2 stain, on the other hand, was much brighter and was only seen on the plasma membrane.
Central Nervous System Electrophysiology-600 -700-m-thick transverse medullary slices were obtained from 31 embryonic (day 18.5) mice according to procedures for neonatal mice described in detail elsewhere (24). Tail samples were frozen for later genotyping. Slices used in physiology were prepared in an ice-cold artificial cerebrospinal fluid containing (in mM) 118 NaCl, 3 KCl, 1.5 CaCl 2 , 1 MgCl 2 , 25 NaHCO 3 , 1 NaH 2 PO 4 and 30 D-glucose and equilibrated with carbogen (95% O 2 and 5% CO 2 , pH 7.4). KCl was elevated to 8 mM over a span of 30 min before commencing recordings. All chemicals were obtained from Sigma. Extracellular population activity was recorded and integrated as described previously (25). The data were digitized with a Digidata acquisition board (Axon Instruments, Foster City, CA), stored on an IBM compatible PC with the software program Axotape (Axon), and analyzed offline using Igor Pro (WaveMetrics, Lake Oswego, OR) and Prism (GraphPad, San Diego, CA). All recordings had a signal:noise ratio sufficient for quantitative evaluation. Statistical comparisons among all three genotypes were performed using analysis of variance. Comparisons among groups were performed subsequently using the Tukey post-test. These post-tests sometimes yielded a significant difference between the ␣2 Ϫ/Ϫ group and only one of the ␣2 ϩ/Ϫ or wild-type groups. Because in no case did the ␣2 ϩ/Ϫ and wild-type groups differ significantly, we grouped the ␣2 ϩ/Ϫ and wild-type recordings together to perform the t tests and nonparametric tests reported in the text.

Na,K-ATPase ␣2 Ϫ/Ϫ Mice Appear Normal but Do Not
Breathe-Mice lacking the ␣2 isoform are born and display no gross anatomical or histological abnormalities. However, these mice appear limp and do not respond to pinch. By opening the chest cavity of newborn pups immediately after birth we established that the hearts from the ␣2 Ϫ/Ϫ pups were beating, indicating that the mice were alive when born. Several minutes after birth the wild-type and ␣2 ϩ/Ϫ mice breathe and turn pink, but the ␣2 Ϫ/Ϫ animals do not appear to breathe. Therefore we measured blood gas levels in newborn mouse pups within 15-30 min after birth. Both the wild-type and ␣2 ϩ/Ϫ mice showed normal levels of oxygen and carbon dioxide (26), whereas the ␣2 Ϫ/Ϫ newborn pups displayed very low oxygen and high carbon dioxide levels consistent with failure to breathe (Table I). Blood glucose levels and body weight were normal for the ␣2 Ϫ/Ϫ mice (Table I).
Lungs removed from pups ϳ15-30 min after birth as well as from day 18.5 embryos were fixed and stained with hematoxylin and eosin. As shown in Fig. 1, lungs from the ␣2 Ϫ/Ϫ mice appeared developmentally normal at both embryonic day 18.5 and at birth compared with wild-type. Saccules of the ␣2 Ϫ/Ϫ mice at embryonic day 18.5 were less expanded, however, consistent with a failure of normal prenatal breathing motions that exchange lung fluid with amniotic fluid and assist in the maturation of the lung (27). Postnatally, the saccules contained more cellular debris because of hemorrhage in the lungs, which most likely represents a secondary shock lesion in the dying pup. The ␣2 Ϫ/Ϫ lungs at birth also showed less postnatal dilation of saccules, confirming that the animals did not breathe after birth. In contrast, the lungs from the wild-type pups show dilated respiratory bronchioles and saccules consistent with partially expanded lungs, indicating there has been some breathing activity.
We examined tissues from ␣2 Ϫ/Ϫ mice to check for any alteration in Na,K-ATPase isoform expression. Western blot analysis of seven tissues from embryonic day 18.5 mice shows that in wild-type animals, the ␣2 isoform was detected with similar abundance in brain and diaphragm, and a faint signal was found in heart (Fig. 2). 3 In the ␣2 Ϫ/Ϫ pups, there did not appear to be a significant change in abundance of the ␣1 or ␣3 isoform in any tissue compensating for the loss of the ␣2 isoform.
Diaphragm Functional Analysis-Because the ␣2 isoform was expressed primarily in muscle and brain around the time of birth we examined these tissues further for abnormalities resulting from the absence of the ␣2 isoform. If the diaphragm were not functioning properly it could be caused by either a defect in the neuromuscular junction (NMJ) in which the signal from nerve to muscle is defective or the diaphragm muscle itself could be unable to contract. In agrin-deficient mice, for example, acetylcholine receptors are reduced in number and density at the NMJ, and these mice die at birth from an inability to breathe (19). We used rhodamine-labeled bungarotoxin to detect acetylcholine receptors as a marker for NMJ development. Synaptophysin, a synaptic vesicle-specific membrane protein, is expressed abundantly in nerve terminal synaptic vesicle boutons, and we used synaptophysin antibody and fluorescein isothiocyanate-labeled secondary antibody to detect this protein as a marker for the nerve terminal. Whole mount immunohistochemistry revealed normal NMJ development in ␣2 Ϫ/Ϫ mice (Fig. 3). We then tested whether the NMJ was functional by electrically stimulating the phrenic nerve. The diaphragm was able to contract, indicating that the synaptic connection between muscle and nerve was functional.
To test whether the absence of the ␣2 isoform in diaphragm altered contractility we developed a method of electrically stimulating and measuring isometric contractility in muscle preparations from day 18.5 embryos. Because of the small size of the diaphragm muscle as well as the presence of the attached ribs, accurate weights were difficult to obtain. Thus, assuming that the thickness of each diaphragm was the same in all preparations we normalized the tension data to muscle area (length times width) of the diaphragm strips. No significant differences in maximum twitch force of contraction were observed between wild-type and ␣2 Ϫ/Ϫ mice (Fig. 4). Two other normalization routines were evaluated: force normalized to length and force normalized to width. In all cases of normalization (Fig. 4A) as well as the raw tension data (Fig. 4B), a similar trend was observed, with the ␣2 Ϫ/Ϫ muscle producing a force within 10% of the wild-type muscle with no statistically significant difference (p Ͼ 0.05, Student's t test). Together, these results demonstrate that embryonic diaphragm muscle without the ␣2 isoform is able to contract both by direct electrical stimulation and by stimulation via the phrenic nerve with a force similar to that of wild-type. Therefore the brain, 3 It has been reported that ␣2 is expressed in alveolar cells when their phenotype changes from ATII-like to ATI-like in culture (46). However, there have been several reports that ␣2 mRNA is lacking in lung (9, 47-49), consistent with the absence of the protein reported here.  (8) FIG. 1. Hematoxylin and eosin staining of lungs of Na,K-ATPase ␣2 ؊/؊ and wild-type mice at embryonic day 18.5 and 15-30 min after birth. Note that the alveoli in the homozygous mice (␣2 Ϫ/Ϫ ) (B and D) are not as dilated compared with wild-type (wt) (A and C). Scale bar, 1 mm.
FIG. 2. The Na,K-ATPase ␣2 isoform is expressed in brain and diaphragm at embryonic day 18.5. The expression of each of the ␣ isoforms of Na,K-ATPase was examined in microsomes prepared from wild-type (wt) and Na,K-ATPase ␣2 homozygous (Ϫ/Ϫ) mice in brain (Br), heart (Ht), skeletal muscle (Sm), diaphragm (Dia), kidney (K), liver (Liv), and lung (Lu). 10 g of protein was loaded in each lane for each antibody blot except for the kidney sample, which contained 5 g of protein for the ␣1 blot.
which also showed expression of the ␣2 isoform at embyronic day 18.5, was examined further for physiological defects associated with the absence of the ␣2 isoform.
Na,K-ATPase ␣2 Isoform Is Expressed in Neurons at Embryonic Day 18.5-Previous reports on Na,K-ATPase ␣2 isoform expression have shown that it is expressed primarily in astrocytes of adults (for review, see Ref. 28); however, little data exist on the expression of ␣2 at the time of birth in mice. Therefore, as an initial step toward the evaluation of the ␣2 Ϫ/Ϫ mice, we determined the expression profile of the ␣2 isoform from embyronic day 18.5 wild-type mice by in situ hybridiza-tion and immunofluorescence analysis of mRNA and protein, respectively, to determine the cell type and the regions of the brain that express it. Fig. 5 shows in situ hybridization of sagittal brain sections in which signals for ␣1 and ␣2 isoforms can be compared. In Fig. 5, A and B, the choroid plexus is shown. This highly elaborated secretory epithelium that emerges from the ventricular lining shows strong hybridization for ␣1 isoform mRNA with the antisense probe, but little or no hybridization above background with the ␣2 antisense probe. In contrast, in Fig. 5D, it can be seen that the ␣2 isoform antisense probe showed extremely heavy hybridization over the pia mater, a tissue that is known to express the ␣2 isoform, whereas the ␣1 isoform antisense probe showed less (Fig. 5C). Neither the ␣1 isoform antisense nor the ␣2 isoform sense probes labeled the pia. These data validate the methods by confirming the known distribution of the ␣1 isoform in choroid plexus and the ␣2 isoform in pia. In the cortical layer, the ␣1 FIG. 4. Maximal twitch force of contraction in diaphragm of ␣2 ؊/؊ mice is similar to wild-type mice. Diaphragm muscle strips from embryonic day 18.5 mice were used for isometric force of contraction measurements. A, maximal twitch force was normalized to muscle length, muscle width, or to area (length times width) for both wild-type (wt) and ␣2 Ϫ/Ϫ (Ϫ/Ϫ) mice. B, maximal twitch force without normalization. Values are the means Ϯ S.E. (n ϭ 7 in the wild-type group, n ϭ 6 in the ␣2 Ϫ/Ϫ group).

FIG. 5.
In situ mRNA hybridization analysis revealed that both the Na,K-ATPase ␣1 and ␣2 isoforms are expressed throughout the brain in wild-type embryonic day 18.5 mice. Embryonic day 18.5 brain sections from wild-type mice were incubated with antisense ␣1 and ␣2 isoform-specific probes. A, C, and E represent darkfield photomicrographs of sections hybridized with the ␣1 antisense probe. B, D, and F represent sections hybridized with the ␣2 antisense probe. The ␣1 sense control is shown in G, and the ␣2 sense control is shown in H. A and B, sections through the hippocampus. A, the dentate gyrus shows uniform expression (bright white grains) of the ␣1 isoform, and the star designates robust expression for the ␣1 isoform but not the ␣2 isoform in the choroid plexus. The adjacent layers also show uniform expression but overall less intensity because of the cellularity differences. B, in contrast, the ␣2 pattern is punctate although not uniform across all cells in that region. The ependymal lining is also more intense (arrow) for the ␣2 isoform. C and D, sections through the cortex. C, the ␣1 signal shows a mostly uniform pattern throughout the cortex but no expression in the pia mater. D, in contrast to C, the ␣2 isoform pattern is much more nonuniform in the neuronal layers of the cortex because some neurons are more intense than others in this layer. The pia mater shows strong ␣2 isoform expression (arrow). E and F, brainstem near the ventral respiratory group; the inset shows a brightfield view of hematoxylin and eosin staining with the larger (neuronal) nuclei purple, and the hybridization signal is the small black grains. The ␣2 isoform shows stronger expression than the ␣1 isoform. Scale bar, 100 m.
probe showed a more uniform pattern of diffuse signal in all of the cells, small and large. In contrast, the ␣2 probe showed much more cell to cell variability, with some neuronal somas showing strong expression. Fig. 5, E and F, shows hybridization in a region of the brainstem just ventral and caudal to the position of the choroid plexus in the floor of the fourth ventricle. Similar results were seen deeper in the brainstem and also in the cerebral cortex. It can be seen that large diameter neural cell bodies were sometimes labeled for the ␣1 isoform but more heavily for the ␣2 isoform. There was also signal above background over regions between neurons, particularly for the ␣2 isoform, which is presumably in glia, which have a less localized cytoplasm. The sense probe controls (Fig. 5, G and H) showed scattered background grains that were not localized to any structure. Fig. 6 shows immunostain for the ␣2 isoform in wild type and ␣2 Ϫ/Ϫ mice. Unlike the sagittal sections used for in situ hybridization, these sections were from tissue slices like those used for electrophysiological recording below, i.e. brainstem cut at an angle that includes the cellular elements required for respiratory rhythm generation. The images shown in Fig. 6, A and B, were from the region of the pre-Bötzinger complex, but a similar stain was seen in most of the section. Stain appeared to be present in both neurons and glia, most prominently in stained somas and fine processes characteristic of neurons. Fig.  6B is a portion of Fig. 6A at higher magnification and with fewer stacked optical sections to show more cellular detail. Fig.  6C, which shows a section through the midline raphe, shows stain in bundles of fibers on either side of the midline raphe. In Fig. 6, D and E, are the controls, showing light stain of neurons and blood vessels with the anti-mouse secondary antibody used to detect the ␣2-specific antibody (Fig. 6D) and a lack of specific anti-␣2 stain in the ␣2 Ϫ/Ϫ mouse (Fig. 6E). These results show that neurons throughout different regions of the brain contain abundant levels of both mRNA and protein for the Na,K-ATPase ␣2 isoform at the time of birth.
Na,K-ATPase ␣2 Ϫ/Ϫ Mice Display Altered Neural Firing in the Respiratory Network-The electrochemical ion gradients generated by Na,K-ATPase are essential for the electrical excitability of cells. Because the ␣2 isoform was expressed abundantly in neurons of wild-type mice, we tested the possibility that the ␣2 isoform may be required for the integrated function of an essential neural circuit: the generation of respiratory rhythm. The respiratory center of the brain was examined in embryonic day 18.5 mice. The respiratory rhythm network in the normal brain is well established before birth, and breathing motions occur in utero. At birth, the brain is already transmitting the appropriate signals for the newborn to breathe on its own. Respiratory neural activity was recorded in vitro from the ventral respiratory group using transverse medullary slices obtained from day 18.5 embryos from wild-type, ␣2 ϩ/Ϫ , and ␣2 Ϫ/Ϫ mice. This slice preparation has been described previously for neonatal and juvenile mice and rats, and it generates respiratory rhythms corresponding to both normal breathing ("eupnea") and sighs ("augmented breaths") (25). The outcome was that rhythmic activity showed abnormal properties.
Spontaneous neural activity was observed in all genotypes including the ␣2 Ϫ/Ϫ mice, indicating that the absence of the Na,K-ATPase ␣2 isoform does not result in a complete loss of neuronal function. Rhythmic activity was observed in the ventral respiratory group in 26 of 31 preparations examined. Of the five preparations from which activity could not be recorded, two were from ␣2 Ϫ/Ϫ , three from wild-type animals, and none from ␣2 ϩ/Ϫ . Of the 26 active slices, 19 produced two clearly distinct patterns of activity (Fig. 7, A and B). Larger amplitude bursts occurring at intervals of about 40 s to several minutes represent fictive sighs, whereas smaller amplitude bursts oc- Representative tracings of population activity recordings from transverse slices of embryonic day 18.5 brain containing the ventral respiratory group, ␣2 Ϫ/Ϫ (A) and wild-type (B) are shown. The raw data appear on the lower traces and integrated data on the upper traces. C, regularity of the respiratory rhythm, as defined by the eupnea period, was lower in the ␣2 Ϫ/Ϫ mice. D, eupnea amplitude was lower in the ␣2 Ϫ/Ϫ mice. Asterisks designate the sigh peaks, and the shorter peaks represent the eupnea. (n ϭ 9 for the ␣2 ϩ/Ϫ and wild-type groups, and n ϭ 10 for the ␣2 Ϫ/Ϫ group).
curring at intervals of a few seconds represent fictive eupnea as identified previously in neonatal preparations of outbred CD-1 mice (25). There was no correlation between genotype and the presence or absence of rhythmic activity in the slice (p ϭ 0.2793, 2 ϫ 3 chi square; p ϭ 1.000 by Fisher's exact test with ϩ/Ϫ and ϩ/ϩ pooled).
Although the absence of the Na,K-ATPase ␣2 isoform did not result in a complete loss of neural activity, the regularity and the amplitude of the respiratory rhythm of normal breathing (eupnea) were affected profoundly (Table ⌸, Fig. 7). Table ⌸ shows that neither the burst duration nor the mean frequency was significantly different (p ϭ 0.1659 and p ϭ 0.8225, respectively). However, the regularity of the eupnea rhythm was significantly lower in the ␣2 Ϫ/Ϫ mice compared with ␣2 ϩ/Ϫ and wild-type. This is reflected by the coefficient of variation of cycle periods of eupnea (designated as period CV in Table ⌸ and Fig. 7C) from 0.43 for the ␣2 ϩ/ϩ and ␣2 ϩ/Ϫ group to 0.66 for the ␣2 Ϫ/Ϫ group (p ϭ 0.0011). Note that in this case, a higher number means lower regularity of rhythm. We also examined the amplitude of the eupnea rhythm. Generally, the amplitude of the integrated population burst is not directly comparable between preparations because of differences in recording quality and the lack of an absolute measurement scale. However, sighs apparently reflect a maximal activation of inspiratory cells within the respiratory network (25), and this enabled us to assess the amplitude of the eupneic burst as a percentage of the sigh amplitude for each individual recording. The amplitude of the eupneic burst was significantly lower in slices obtained from ␣2 Ϫ/Ϫ embryos (38%) than in slices from ␣2 ϩ/Ϫ and wildtype animals (59%) (p ϭ 0.0056) ( Table ⌸ and Fig. 7, A and D). This appears to be a major factor leading to the failure of the animals to breathe effectively. DISCUSSION Mice deficient in the Na,K-ATPase ␣2 isoform are born but die soon after birth. By gross examination all organs appear normal; however, the mice are akinesic. This study shows for the first time that the ␣2 isoform is highly expressed in neurons throughout the brain at the time of birth, with lower levels of expression in glia. This would not have been predicted from the literature. A previous study showed that ␣2 mRNA is present quite early in brain development, but without cellular resolution it could not be certain whether the signal was in neurons or glia (12). In another study, embryonic stem cells and aggregates of undifferentiated cells expressed ␣1 and ␣2 isoform mRNA, but only ␣1 protein. In vitro neuronal induction of embryonic stem cells was accompanied by induction of the ␣3 isoform at late stages of differentiation, but ␣2 isoform mRNA or protein was not induced at any stage (29). Embryonic fetal forebrain examined after several days in culture either as aggregates or as plated cells expressed the ␣3 isoform in neurons and ␣2 isoform in glia, as assessed either by selective elimination of cell types by toxic agents (30) or by immunofluorescence microscopy (31). Different results have been reported for cultures of hippocampal neurons: one paper reported find-ing only the ␣1 and ␣3 isoforms by immunofluorescence (32), whereas two papers reported finding the ␣2 and ␣3 isoforms or all three isoforms by Western blots (33,34), although the latter result could have reflected contamination with astrocytes (35). Cerebellar granule neurons have been reported by Soga et al. (36) to express all three isoforms in culture, although the presence of the ␣2 isoform was found exclusively in cerebellar astrocytes in prior studies both in situ and in granule neuron cultures (13,14,28), raising the possibility that the commercial antibodies used by Soga et al. were cross-reactive. Inspection of the published stain for the ␣2 isoform in the developing mouse retina, however, suggests that it may have been present at birth in the nascent inner nuclear layer, but subsequently lost over the next few days (37).
We show in this study by in situ hybridization that large diameter neuronal cell bodies expressed the ␣2 isoform at all levels in the brain, and diffuse signal characteristic of glia was also observed. By immunofluorescence, most of the ␣2 isoform stain had a reticular pattern, with outlining of presumptive neuronal somas in places and many fine processes characteristic of neurons. At this age, the brainstem lacks the bundles of ascending and descending myelinated axons and well developed neuronal nuclei of the adult, but the distribution of stain clearly differs from the purely astrocytic pattern for the ␣2 isoform in the adult rat brainstem (13). Taken together with the published evidence, it seems likely that the ␣2 isoform is expressed transiently in many central nervous system neurons during development but that with phenotypic maturation either in vivo or in culture, the expression is lost. At the time of birth it is expressed mainly in neurons, whereas in adult it is expressed mainly in astrocytes, which proliferate and differentiate mostly postnatally.
The lack of ␣2 isoform expression in mice results in a gross defect in that the ␣2 Ϫ/Ϫ mice are unresponsive to pinch and do not breathe. Although the ␣2 Ϫ/Ϫ mice may have widespread functional disturbances throughout the brain, we were able to document the effect of the absence of ␣2 on the integration of neuronal activity in the brainstem respiratory center. We observed some basic neuronal activity in the ␣2 Ϫ/Ϫ mice; however, the fictive eupnea breathing rhythm recorded in vitro was disrupted significantly in the ␣2 Ϫ/Ϫ mice compared with ␣2 ϩ/Ϫ and wild-type littermates. Population bursts were both lower in amplitude and much less regular, and in some ␣2 Ϫ/Ϫ animals this was altered to such an extent that this rhythm was nearly absent (Fig. 7). To obtain rhythmic activity in slice preparations it is necessary to raise the extracellular potassium concentration (25). Thus it is conceivable that the centrally generated rhythm would be even weaker in vivo and insufficient to support a normal motor output under physiological conditions. Interestingly, the fictive sigh rhythm was still prominent, despite its apparent origination from the same neural network (25). This suggests that sighs alone may not be sufficient to sustain life, even in the relatively hypoxia-resistant newborn. Alternatively, the apparent failure of the knockouts to take even one breath, based on the lack of alveolar expansion, suggests that the first breath of life is not a sigh, but might instead be a gasp. Bursts corresponding to gasps are manifested in slices but are distinct from sighs in that gasps are found only in anoxia and are not intermixed with eupneic breaths (25). The decreased regularity and the decreased amplitude of the eupnea respiratory rhythm, combined with an apparent lack of breathing in the intact animal, indicates that the population level activity we observed in slices is essentially failed central breaths. The individual cells retain some bursting ability but do not synchronize sufficiently, the amplitude being a measure of synchronous activity, to generate a regular rhythm or to give rise to respiratory movements. Fictive sighs are observed in the brain slice preparations of ␣2 Ϫ/Ϫ animals. However, the data we present here show that the likelihood of observing fictive sighs was actually higher in ␣2 Ϫ/Ϫ mice (10 of 12, or 83%) compared with ␣2 ϩ/Ϫ mice (4 of 8, or 50%) or wild-type mice (5 of 11, or 45%). The genesis of sighs might be stimulated by the failure of the eupneic rhythm in the knockouts; perhaps the eupneic rhythm ordinarily exerts a negative modulatory influence on the frequency of sighs. Because the ␣2 isoform was found to be expressed throughout neurons in the brain of day 18.5 embryos, it is quite possible that the altered neuronal activity we observed in respiratory center neurons may be representative of a more global defect in neurons throughout the rest of the brain in the ␣2 Ϫ/Ϫ mice. This aberrant neuronal activity could be a result of one or more functional disturbances. A specific role for Na,K-ATPase has been proposed to be in clearance of potassium from the extracellular space to prevent depolarization of neurons during high neuronal activity (38,39). It was proposed that glial Na,K-ATPase (␣2 isoform) contributed to the initial fast uptake of extracellular potassium K, whereas the axonal Na,K-ATPase (␣3 isoform) participated in the slower poststimulus recovery from elevated extracellular potassium (39). We would predict that in our homozygous knockout mice, the absence of the ␣2 isoform might result in delayed removal of extracellular potassium in both neurons and glia and thus affect neuronal excitability.
Another function of Na,K-ATPase in the brain is that the glutamate transporter works in concert with the ␣3 or ␣2 Na,K-ATPase isoform to clear glutamate from the extracellular space (40 -42). Glutamate is transported in a sodium-and potassium-dependent fashion utilizing the sodium and potassium gradients set by Na,K-ATPase. It has been shown that inhibition of Na,K-ATPase by ouabain will reduce the amplitude of neuronal compound action potentials (41), and this diminished neuronal activity was shown to be a result of toxicity mediated in part by glutamate release through reverse sodium-dependent glutamate transport. Interestingly, a recent report by Rozzo et al. (43) demonstrated that the spontaneous rhythmicity of network spinal cord neurons is disrupted with Na,K-ATPase inhibition by strophanthidin, a cardiotonic steroid related to ouabain (43). Specifically they found a smaller peak amplitude as well as irregular bursting intervals, which supports the present data in which we observed the same phenomena in the respiratory rhythm center of the ␣2 Ϫ/Ϫ mice. They further showed this was a glutamate-mediated response because the altered rhythmicity could be blocked using a glutamate receptor antagonist. The concentration of strophanthidin used by Rozzo et al. (4 M) would inhibit both the Na,K-ATPase ␣2 and ␣3 isoforms (43). Therefore it is intriguing to speculate that perhaps the ␣2 isoform works with the glutamate receptor to modulate neuronal excitability. To test the specific contribution by the ␣2 isoform in maintaining neuronal excitability versus the contribution from the ␣3 isoform, how-ever, would require the specific inhibition of the ␣3 isoform either by pharmacological means (no specific inhibitor exists yet) or by genetic knockout (that animal has not been produced).
Na,K-ATPase has been proposed to regulate intracellular calcium via the Na/Ca exchanger. It has been proposed that the Na,K-ATPase ␣2 isoform and the Na/Ca exchanger are colocalized in microdomains in the plasma membrane of astrocytes and mesenteric artery smooth muscle cells to work as a functional unit (34). During neuronal activity, sodium must be pumped out of the cell to maintain excitability. If the capacity of the neuron to pump out intracellular sodium is reduced, intracellular sodium would rise, which then would change the driving force for Na/Ca exchange, resulting in elevated [Ca 2ϩ ] i . A recent report by Golovina et al. (44) supports this idea in which it was shown that in our ␣2 Ϫ/Ϫ mice, astrocytes have elevated levels of intracellular calcium as well as elevated stores of calcium in the endoplasmic reticulum. We show in this study that in wild-type animals the ␣2 isoform is expressed more abundantly in neurons than in astrocytes at birth. We would predict even higher levels of intracellular calcium in neurons than that shown for astrocytes in the ␣2 Ϫ/Ϫ mice (44). Tang et al. (45) have reported that a rise in intracellular calcium caused by inhibition of the Na/Ca exchanger can enhance neurotransmitter release in chromaffin cells. It is possible that a similar mechanism could occur in the central nervous system of our ␣2 Ϫ/Ϫ mice in which glutamate release would be enhanced as a result of Na/Ca exchange reduction. As discussed above, glutamate has been shown to alter neuronal activity as a secondary response to inhibition of Na,K-ATPase. In sum, it is clear from the present work that the absence of the ␣2 isoform results in altered integration of neuronal activity, and further work will be required to sort out whether the mechanism entails neurotransmitters.
Last, it is also possible that the ␣2 isoform directly provides a significant contribution to the total Na,K-ATPase activity in the cell and that with its removal, the activity of the other Na,K-ATPase isoforms is simply not sufficient to handle the demand for ion transport in the cell, resulting in membrane depolarization. Because respiratory rhythm generation was not abolished, some basic synaptic and neuronal membrane properties were still intact. Given that the absence of the Na,K-ATPase ␣2 isoform could lead to a number of different functional defects as proposed above, this work provides a foundation to investigate further specific roles for the ␣2 isoform in maintaining neuronal excitability. The present work provides a new area of study of the ␣2 isoform in neurons as this cell type was not recognized previously to express this Na,K-ATPase isoform.