Plasma Membrane-Cytoskeleton-Endoplasmic Reticulum Complexes in Neurons and Astrocytes*

The possibility that certain integral plasma membrane (PM) proteins involved in Ca2+ homeostasis form junctional units with adjacent endoplasmic reticulum (ER) in neurons and glia was explored using immunoprecipitation and immunocytochemistry. Rat brain membranes were solubilized with the mild, non-ionic detergent, IGEPAL CA-630. Na+/Ca2+ exchanger type 1 (NCX1), a key PM Ca2+ transporter, was immunoprecipitated from the detergent-soluble fraction. Several abundant PM proteins co-immunoprecipitated with NCX1, including the α2 and α3 isoforms of the Na+ pump catalytic (α) subunit, and the α2 subunit of the dihydropyridine receptor. The adaptor protein, ankyrin 2 (Ank 2), and the cytoskeletal proteins, α-fodrin and β-spectrin, also selectively co-immunoprecipitated with NCX1, as did the ER proteins, Ca2+ pump type 2 (SERCA 2), and inositol-trisphosphate receptor type 1 (IP3R-1). In contrast, a number of other abundant PMs, adaptors, and cytoskeletal proteins did not co-immunoprecipitate with NCX1, including the Na+ pump α1 isoform, PM Ca2+ pump type 1 (PMCA1), β-fodrin, and Ank 3. In reciprocal experiments, immunoprecipitation with antibodies to the Na+ pump α2 and α3 isoforms, but not α1, co-immunoprecipitated NCX1; the antibodies to α1 did, however, co-immunoprecipitate PMCA1. Antibodies to Ank 2, α-fodrin, β-spectrin and IP3R-1 all co-immunoprecipitated NCX1. Immunocytochemistry revealed partial co-localization of β-spectrin with NCX1, Na+ pump α3, and IP3R-1 in neurons and of α-fodrin with NCX1 and SERCA2 in astrocytes. The data support the idea that in neurons and glia PM microdomains containing NCX1 and Na+ pumps with α2 or α3 subunits form Ca2+ signaling complexes with underlying ER containing SERCA2 and IP3R-1. These PM and ER components appear to be linked through the cytoskeletal spectrin network, to which they are probably tethered by Ank 2.


INTRODUCTION
Cytosolic Ca 2+ plays a key role as a second messenger in all cells and is responsible for regulating numerous cellular processes simultaneously. Spatial as well as temporal control of the cytosolic Ca 2+ signals is therefore critical (1,2). Much of the "signal Ca 2+ " comes from intracellular stores, primarily in the endoplasmic reticulum (ER) 1 . Nevertheless, Ca 2+ signaling depends critically upon the coordination of Ca 2+ entry across the plasma membrane (PM), Ca 2+ release from the ER, Ca 2+ extrusion across the PM, and refilling of the ER stores by resequestration and by entry from the extracellular fluid (1,2). Clearly, the many specific mechanisms responsible for all these Ca 2+ movements must be precisely coordinated. One possibility is that several functionally inter-dependent mechanisms may be clustered and organized into specialized "Ca 2+ signaling complexes" (CaSCs). Indeed, recent results suggest that PM microdomains, adjacent "junctional" ER (jER), and the intervening, tiny volume of cytosol (units we have called "PLasmERosomes") (2), serve as CaSCs in many types of cells (3)(4)(5)(6)(7). Classic examples are the PM-sarcoplasmic reticulum (SR) junctions in skeletal and cardiac muscle (8). Structurally similar PM-SR or PM-ER junctions have been observed by electron microscopy in smooth muscle (9) and neurons (10,11), even though Ca 2+ signaling in these cells involves some different Ca 2+ transporters.
The Na + /Ca 2+ exchanger (NCX) is a Ca 2+ transporter that helps to control intracellular and jER Ca 2+ concentrations and cell signaling in many types of cells (12). All three isoforms 9 Immunocytochemistry. Primary cultured mouse neurons were washed with PBS at 22 o C and fixed for 10 min in ice-cold methanol. The fixed neurons on the coverslips were rinsed 3 times with PBS, permeabilized with 0.25% Triton X-100 in PBS, and rinsed 3 times more with PBS.
After blocking nonspecific binding with 10% BSA in PBS for 1 hr, the neurons were incubated overnight at 4 o C with primary antibodies diluted in Antibody Buffer (0.5 mM NaCl, 10 mM

Immunoprecipitation of the NCX1 from a detergent-soluble extract of rat brain membranes.
Ca 2+ homeostasis in most cells depends upon two PM Ca 2+ transporters, the PMCA (34) and the NCX (12). The NCX1 isoform, which is prevalent in neurons and glia (Introduction), appears to be confined to membrane microdomains that form junctional complexes with the adjacent ER (3,4). Moreover, NCX activity modulates ER-dependent Ca 2+ signaling (2). To learn if NCX1 forms a complex with elements of the ER, and if these junctional structures contain cytoskeletal structures capable of linking them together, we prepared immunoprecipitates of NCX1 from detergent-solubilized extracts of rat brain membranes. Immunoblots were performed on these IPs to identify other proteins that co-IP with, and may therefore associate with, NCX1 (Figs. 1 and 2).

[Figs. 1 & 2 near here]
Immunoprecipitation removed a large fraction (about 73%) of the NCX1 from the detergent extract. Fig. 1A shows that the density of the dominant NCX1 band (140 kDa in this immunoblot 2 ) in the post-IP, detergent-soluble supernatant (IP S) fraction was only about 27% of the band density in the original homogenate (Homog; n = 4 different IPs). For controls, the Dynabeads were coated with non-specific Mab (MOPC-21 = MOPC). No NCX1 bound to these beads. As a result, the NCX1 band in the post-IP, detergent-soluble fraction in this control sample was comparable in density to that in the homogenate.
Some PM proteins co-IP to a significant extent with NCX1 (Fig. 1B), notably the α2 and α3 (catalytic) subunits of the Na + pump and the α2 subunit of the L-type voltage-gated Ca 2+ channel (DHPR α2). TRPC-4 protein, a component of some store-operated channels (36), was 12 also observed in the IP, although to a lesser extent (not shown; see Table 1). In marked contrast, several other neuronal and glial PM proteins that also are prevalent in the detergent extract did not co-IP with NCX1 (Fig. 1B). These include the α1 isoform of the the Na + pump, PMCA1, Trk B (two isoforms at 95 and 145 kDa) and N-CAM (two isoforms at 140 and 180 kDa). These differences are exemplified by the relative densities of the protein bands detected in immunoblots of IPs generated with anti-NCX1. For example, the band densities of the α2 and α3 isoforms of the Na + pump and the α2 subunit of the DHPR in the post-IP supernatant (IP S) were reduced by about 24%, 63% and 49% (means of 2-3 IPs), respectively, compared to the band densities in the original detergent extract (Homog). In contrast, the Na + pump α1 isoform, PMCA1, Trk B and N-CAM were not significantly reduced. The fact that these proteins were prevalent in the detergent extract, but absent from the IPs generated with anti-NCX1, confirms the specificity of the interactions detected by co-IP, and serves as an additional control. These and our other co-IP data (Figs. 1-9) are summarized in Table 1.
Immunoblots also revealed that, in addition to PM proteins, the original detergent extract ( Fig. 2, Homog) contained several adaptor or anchoring proteins, cytoskeletal elements and ER membrane proteins. The IPs were examined for the presence of the adaptor proteins, ankyrins 2 and 3 (Ank 2 and Ank 3), and homers 1 and 2, the cytoskeletal proteins αand β-fodrin, and βspectrin, and the ER membrane proteins, SERCA2b and IP 3 Rs 1, 2 and 3. Ank 2, α-fodrin and β-spectrin, SERCA2b and IP 3 R-1 all co-immunoprecipitated with NCX1, but Ank 3, β-fodrin, the two homer proteins, and IP 3 Rs 2 and 3 did not ( Fig. 2 and Table 1). antibodies raised against α subunit isoforms 1-3 of the Na + pump (Figs. 3-5; Na + pumps are composed of an α and a β subunit). When antibodies raised against the α2 (Fig. 4) or α3 (Fig. 5) isoforms were bound to the Dynabeads and used for immunoprecipitations, both NCX1 and IP 3 R-1 were found in the IP pellet, as was Ank 2, but the PMCA1 band was weak (α2 IP) or absent (α3 IP). In contrast, IP with an antibody raised against the Na + pump α1 isoform did not co-IP NCX1 or IP 3 R-1, but did co-IP PMCA1 ( Fig. 3; note the stronger 134 kDa band after boiling 3 ; Ref. 34); weak Ank 2 and Ank 3 bands were also seen. These data are consistent with evidence that the Na + pumps with α2 or α3 subunits (i.e., those with high affinity for ouabain), but not those with α1 subunits, are functionally coupled to NCX1 and help to modulate Ca 2+ signaling (5,7). Moreover, as suggested by immunocytochemistry (4), PMCA1, like Na + pumps with α1 subunits, may be very widely distributed in the PM, but excluded from the microdomains that contain NCX1 and Na + pumps with α2 or α3 subunits.

[Figs. 3-5 near here]
The Na + pump α1 isoform is expressed in virtually all cells (37), and appears to be the "housekeeping" isoform of the catalytic subunit that maintains the low cytosolic Na + concentration (7). In addition, astrocytes express the α2 isoform whereas mature neurons express α3 (3,38,39). Thus, it is not surprising that IP of the detergent-soluble brain membrane fraction with an antibody to the α2 subunit of the Na + pump co-IPs Ank 2 and α-fodrin, both of which are present in astrocytes, but not β-spectrin, which is not ( Fig. 4; Ref. 26). In contrast, IP with an antibody to Na + pump α3 (Fig. 5) co-IPs β-spectrin 4 as well as Ank 2 and α-fodrin, all of which are expressed in neurons (26,27).

Immunoprecipitation of adaptor, cytoskeletal and ER proteins.
As shown above, the detergent extract from rat brain membranes contains various adaptor and cytoskeletal elements, some of which appear with PM and ER membrane proteins in the IP fractions generated with antibodies to NCX1. If, as we hypothesize, the adaptor and cytoskeletal proteins link the ER and PM proteins in large multi-molecular complexes, then IP of some of these adaptor and cytoskeletal proteins should co-IP PM and ER membrane proteins. IPs were prepared with antibodies to αfodrin, β-spectrin and Ank 2 to test this possibility (Figs. [6][7][8]. IP with antibodies to α-fodrin or β-spectrin both co-immunoprecipitated NCX1 and IP 3 R-1 as well as Ank 2. The IP generated with antibodies to α-fodrin also contained NCX1, the Na + pump α2 and α3 isoforms (not shown), β-spectrin and IP 3 R-1 (Fig. 6), and that generated with antibodies to β-spectrin also contained NCX1, α-fodrin, DHPR α2 and IP 3 R-1 (Fig. 7). The IP generated with antibodies to β-spectrin also exhibited a weak band of the α3 subunit of the Na + pump, but not of the α2 subunit (not shown). The IP generated by antibodies to Ank 2 contained NCX1, α-fodrin and βspectrin ( Fig. 8). Although we did not test all antibodies by IP, these results suggest that these adaptor and cytoskeletal proteins are present in complexes that contain both PM and ER proteins.

[Figs. 6-8 near here]
The fact that IP of certain PM or cytoskeletal proteins resulted in co-IP of SERCA2 (or 2b) and IP 3 R-1 implies that IP of these ER proteins should co-IP some PM and cytoskeletal proteins. This inference was confirmed: NCX1, Na + pump α3, Ank 2 (but not Ank 3), α-fodrin and β-spectrin all co-IP with IP 3 R-1 (Fig. 9).

[Figs. 9 near here]
Immunocytochemistry. Published immunocytochemical data reveal that most of the NCX1 and Na + pumps with α2 or α3 subunits are confined to PM microdomains that overlie ER in glia and neurons (3,4). Only a fraction of the ER markers co-localized with the overlying PM microdomains (4), however, but this is expected because only a small fraction of the ER in these cells closely associates with adjacent PM.
The relationship between the cytoskeleton and the ER or PM microdomains can be expected to be even more complex because the cytoskeleton should be widely distributed under the PM as well as in the cytosol (25,26,40,41). Nevertheless, if there are specific associations between elements of the cytoskeleton and some ER and PM microdomains, as implied by our co-IP data, immunocytochemistry should reveal significant areas of overlap (co-localization) with the cytoskeletal elements. This is exemplified by the similar (though not identical) distribution of β-spectrin and NCX1 at the surface of the cell body in a primary cultured mouse neuron (Fig.   10A,B; arrows indicate examples of co-labeled structures). The partial co-localization of βspectrin with α3 Na + pumps near the surface of a neuron is illustrated in Fig. 10C,D. Similarly, the punctate distribution and partial co-localization of β-spectrin with the ER membrane protein, IP 3 R-1, was detected near the neuron surface (Fig. 10E,F). In contrast there was no overlap of the β-spectrin stain with that of IP 3 R-1 in the perinuclear region where much of the ER (and IP 3 R) is located (Fig. 10E,F; asterisks). Labeling was specific, as control antibodies failed to label similar structures at a comparable level (Fig. 10G,H). These results suggest that at least some of the complexes we have examined in co-IP experiments are located at, or immediately adjacent to, the neuronal cell surface, and not in intracellular compartments. They are consistent with our hypothesis that the junctional membrane complexes in neurons contain cytoskeletal proteins of the spectrin family as well as integral proteins of the PM and ER. Astrocytes express the cytoskeletal protein, α-fodrin, but not β-spectrin (26). Therefore, we compared the distribution of the NCX1 and SERCA2 isoforms with that of α-fodrin in primary cultured rat astrocytes. As Fig. 11A,B shows, there is striking similarity in the distribution of the NCX1, which exhibits a reticular pattern in these cells, and α-fodrin.
SERCA2 is also distributed in a reticular pattern in these cells, and partially co-localizes with the α-fodrin (Fig. 11C,D; Fig. 11E,F is a control that confirms the specificity of labeling). These results, too, are consistent with the presence of junctional membrane complexes containing elements of the spectrin-based cytoskeleton at or near the astrocyte surface.  2+ signaling. This study shows that the PM Na + /Ca 2+ exchanger, NCX1, in neurons and glia, is a component of large, multi-molecular complexes that include other PM and ER transporters with which NCX1 is functionally associated. The complexes also includes cytoskeletal elements as well as adaptor proteins that, presumably, link the PM and ER transport proteins to the cytoskeleton. Taken together, the efficacy of co-IP of these PM proteins, the results of the reciprocal co-IP experiments, and the immunocytochemical evidence of co-localization, all demonstrate that both neurons and astrocytes have specialized PM-jER complexes. Also, recent immunocytochemical data indicate that TRPC-1 is confined to PM regions that overlie the ER in astrocytes (17). These results support our hypothesis that the functional interrelationship involving NCX1, Na + pumps with α2 or α3 subunits, SOCs and the jER (2, 5, 7) is maintained by structural proteins that keep these transporters in close proximity at the PM-ER junctions. The data are consistent with the model diagrammed in Fig. 12. Here, the cytoskeletal elements (including α-fodrin and, although we did not examine it, β-fodrin in astrocytes, and both α-fodrin and β-spectrin in neurons) at the PM-ER junctions form a scaffold to which relevant PM and ER transport proteins involved in Ca 2+ signaling are tethered by Ank 2, and perhaps other adaptor proteins.

Organization of PM-ER junctions involved in Ca
[ Fig. 12

near here]
The quantitative data for the PM proteins may be relevant. We observed that large fractions of the Na + pump α2 and α3 subunits co-IP efficiently (> 25 %) with NCX1; TRPC4 also co-IPs, but less efficiently. This agrees with the immunocytochemical evidence (3,14,17) that these particular transporters are apparently confined to PM microdomains that overlie jER. Na + pumps with α1 subunits also are very prevalent in neurons and astrocytes. Nevertheless, the by guest on July 15, 2017 http://www.jbc.org/ Downloaded from IP data suggest that, consistent with earlier immunocytochemical results (3,4) and functional data (7), the α1 isoform may be excluded from the PM microdomains at these junctions.
Similarly, two PM Ca 2+ transporters, PMCA1 and NCX1, are prevalent in the brain membranes. NCX1 is confined to the PM microdomains at PM-jER junctions, however, whereas PMCA1 is more uniformly distributed but does not co-IP with NCX1 and may be excluded from these microdomains (4).
Smaller fractions of the adaptor, cytoskeletal and ER transport proteins co-IP with NCX1, and cytoskeletal and ER proteins only partially co-localize with NCX1. This is anticipated because the cytoskeleton is distributed widely at the PM, but only a small fraction of the ER is located at PM-ER junctions. The lower yields of some co-immunoprecipitated proteins also may be caused by differences in the stability of different spectrin-based membrane complexes (42,43), which also may be affected by the binding of particular antibodies. Furthermore, neurons and glia may express different alternatively-spliced forms of the cytoskeletal proteins (25,26,31), only some of which are designed to participate in junctional complexes. Spectrin at other regions of the PM is likely to be involved in stabilizing the membrane and in immobilizing different classes of proteins into membrane domains with compositions and functions that are distinct from those of the junctional complexes considered here (e.g., at nodes of Ranvier and synaptic densities). The links between spectrin and the integral membrane proteins in these other domains are probably mediated by anchoring proteins other than Ank 2 [e.g., Ank 1, Ank 3 (25) or homer proteins (44)] or by direct binding to spectrin (45).
Our results suggest that these transport proteins associate exclusively with the Ank 2 isoform of ankyrin, in agreement with other reports (25,27), even though Ank 1 and Ank 3 also are expressed in excitable cells (25,27). The structural basis of this selectivity remains to be by guest on July 15, 2017 http://www.jbc.org/ Downloaded from elucidated, but our data and those of Mohler and colleagues (46) are consistent with the idea that Ank 2 is selectively associated with integral membrane proteins at PM-jER complexes.
Furthermore, while our results are consistent with the presence of PM-jER complexes that are linked together into microdomains by spectrin and Ank 2, we still do not know how particular isoforms of PM transporters (e.g., the α2 and α3 isoforms of the Na + pump, but not α1) are targeted to these domains.
Two mechanisms have been proposed to explain the organization of PM microdomains in which functionally-coupled PM proteins are physically linked. One is the formation of cholesterol-, sphingomyelin-and glycosphingolipid-enriched "lipid rafts," in which certain PM proteins may be concentrated (47), perhaps because of unique physico-chemical interactions between the proteins and lipids (48). Our data, however, support the hypothesis that the cytoskeleton and cytoskeletal adaptor proteins link the PM and ER proteins together to form specialized signaling complexes. Thus, the highly specific organization of multiple ER and PM proteins at the PM-ER junctions likely depends on the specificity of protein-protein interactions.

PLasmERosomes as Ca 2+ signaling complexes.
Earlier studies established that the Na + pump (20) and NCX1 (21) both bind to ankyrin. Indeed, sucrose gradient sedimentation of Triton X-20 50) and the α1 subunit of the Na + pump (51). Other studies demonstrated that proteins of the ER, specifically SERCAs, IP 3 Rs and ryanodine receptors, also interact with ankyrins (23,24,27,52,53), consistent with our findings. Indeed, co-IP of IP 3 R-3 with the α1 subunit of the Na + pump was recently described in renal epithelial cells (54); pre-treatment of the cells with cytochalasin D, which depolymerizes the actin cytoskeleton, abolished this co-IP.
The present study expands on these ideas and demonstrates their more general applicability. The IP results verify immunocytochemical and functional evidence that, in astrocytes and neurons, as in other cell types (1-5, 7, 17), NCX1, Na + pumps with α2 or α3 subunits and SOCs cluster at PM-ER junctions. These new data suggest that certain PM microdomains are linked, through Ank 2 and a network of spectrin, to the adjacent jER in neurons and astrocytes. If our model (Fig. 12) is correct, both ankyrin and spectrin are critical for both the structure and function of these PM-jER units.
Ankyrins may target a variety of membrane proteins to physiologically appropriate sites (25) and can serve as a bridge between ion transporters in the PM and the cytoskeleton (55).
Indeed, a null mutation in one copy of the Ank 2 gene reduces expression of Na + pump α subunits, the NCX, and the IP 3 R in the heart (44). Haploinsufficiency of Ank 2 can cause cardiac arrhythmias and sudden death in man and mouse, probably as a result of partial functional uncoupling of the PM-SR junctions (44). Ank 2 knockout mice, which die within three weeks of birth, exhibit a variety of structural defects in the brain including hypoplasia of the corpus callosum and pyramidal tracts, and dilation of the cerebral ventricles (56). This is consistent with our hypothesis that the PM-jER (or jSR) complexes play essential roles in Ca 2+ homeostasis and Ca 2+ signaling (2). Disruption of these units would be expected affect numerous processes including cell development.

21
The consequences of Ank 2 deficiency are only one example of the broad functional implications of our findings. Different PM (and ER) transport proteins may be included within the PM microdomains (and the jER or jSR regions) in different cell types, or even in different parts of individual cells such as neurons. In other words, a single cell may have different junctional complexes with distinct, but specific, functions (2,6,26). Thus, the particular cytoskeletal elements to which the PM proteins are linked also may influence function in ways that are not yet resolved. For example, β-spectrin is found in neuronal cell bodies, growth cones and dendrites, but not in axons, which contain β-fodrin (26,57). Our results suggest that most of the junctional complexes we isolated by IP with antibodies to NCX1 arise from the former compartments, as they contain only β-spectrin. Thus, these PM-β-spectrin-ER junctional units likely help regulate postsynaptic, but not presynaptic, Ca 2+ signaling and Ca 2+ sensitivity and, hence, neuronal plasticity. Clearly, further exploration of these units should greatly enhance our understanding of these complex phenomena. Indeed, we have tested only a selected group of antibodies in this study. For example, NCX2 and NCX3, as well as the K + -dependent Na + /Ca 2+ exchanger, NCKX2, are also present in neuronal PM (13,14,58), but there is little or no overlap in the distribution of NCX2 and NCX3 (13,14). It is possible that these other Na + /Ca 2+ exchangers may be confined to PM-ER junctional units distinct from those that contain NCX1.
It will be challenging to learn just how heterogeneous the junctional complexes in different tissues are, and to elucidate their respective contributions to cellular physiology and plasticity. 2 NCX1 protein has a molecular weight of 120 kDa. Following boiling, however, the protein sometimes runs as a 140 kDa band (35). 3 The PMCA1 antibody detects bands at both ∼130 kDa and ∼134 kDa, that correspond to the PMCA1a and PMCA1b isoforms (34). The lower molecular weight form was most prevalent in the homogenate (e.g. , Figs 1 and 3). When the immunoprecipitated proteins were extracted by boiling the beads, however, the density of the higher molecular weight band increased, and the lower molecular weight band density decreased (Fig. 3); similar results were obtained when the homogenate was boiled (not shown).

NCX1.
A. Immunoprecipitate was generated and probed with anti-NCX1 Mab (R 3 F 1 ; "IP Ab"). Control beads were prepared with mouse IgG1κ (MOPC; IP Ab). Lanes labeled "Homog" (detergent soluble homogenate), "IP S" (IP supernatant) and "Con S" (control IP supernatant) were loaded with equal volumes. Lane "IP P" (pellet eluted from NCX1-IP beads) and "Con P" (pellet eluted  shows that Ank 2, α-fodrin, β-spectrin, SERCA2b and IP 3 R-1 selectively co-IP with NCX1 whereas Ank 3, homers 1 and 2 and β-fodrin do not.  The gels were probed with the blot Abs indicated (abbreviations as in Fig. 2). The results are representative of data from 2 IPs. The figure shows that NCX1, Ank 2, α-fodrin and IP 3 R-1, but not β-spectrin, selectively co-IP with the α2 subunit of the Na + pump; a weak PMCA1 band is also seen (contrast Fig. 3).  The gels were probed with the blot Abs indicated (abbreviations as in Fig. 2). The results are representative of data from 2 IPs. The figure shows that NCX1, Ank 2, α-fodrin, β-spectrin and IP 3 R-1 all co-IP with the α3 subunit of the Na + pump, but PMCA1 does not.   Fig. 2). The results are representative of data from 2 IPs.
The figure shows that NCX1, Ank 2, α-fodrin, DHPR α2 and IP 3 R-1 all co-IP with β-spectrin.   Fig. 2). The results are representative of data from 2 IPs. The figure shows that NCX1, Na + pump α3, Ank 2 (but not Ank 3), α-fodrin and β-spectrin all co-IP with IP 3 R-1. and double labeled for immunofluorescence with affinity-purified, subunit-specific chicken antibodies to erythroid β-spectrin (B,D,F) and with antibodies to NCX1 (A), the α3 subunit of the Na + pump (C), or IP 3 R-1 (E), followed by species-specific secondary antibodies (see Methods). Cells were imaged at their dorsal surfaces, away from the cover slip (A-D,G,H), or through the cell (E,F) at a level that includes the perinuclear cytoplasm (asterisks). Labeling of NCX1, the α3 subunit of the Na + pump, and IP 3 R-1 all partially overlap with that of β-spectrin (arrows), suggesting that these integral proteins of the PM (A,C) and ER (E) are present at structures that also contain β-spectrin. Controls, processed similarly, but with non-immune mouse (MOPC, panel G) and chicken (H) antibodies, failed to show colabeling, indicating that the results in panels A-F are not due to non-specific labeling. Bar, 5 µm. affinity-purified, subunit-specific rabbit antibodies to brain α-fodrin (B,D), followed by speciesspecific secondary antibodies (see Methods). Cells were imaged at thin regions, to minimize background. Labeling of NCX1, and SERCA2 both partially overlapped with that of α-fodrin (arrows), suggesting that these integral proteins of the PM (A) and ER (C) are present at structures that also contain α-fodrin. Controls, processed similarly but with non-immune mouse (MOPC, panel E) and rabbit (F) antibodies, failed to show colabeling, indicating that the results in panels A-D are not due to non-specific labeling. Bars, 20 µm.