Ankyrin-B targets beta2-spectrin to an intracellular compartment in neonatal cardiomyocytes.

Ankyrin-B is a spectrin-binding protein that is required for localization of inositol 1,4,5-trisphosphate receptor and ryanodine receptor in neonatal cardiomyocytes. This work addresses the interaction between ankyrin-B and beta(2)-spectrin in these cells. Ankyrin-B and beta(2)-spectrin are colocalized in an intracellular striated compartment overlying the M-line and distinct from T-tubules, sarcoplasmic reticulum, Golgi, endoplasmic reticulum, lysosomes, and endosomes. Beta(2)-Spectrin is absent in ankyrin-B-null cardiomyocytes and is restored to a normal striated pattern by rescue with green fluorescent protein-220-kDa ankyrin-B. We identified two mutants (A1000P and DAR976AAA) located in the ZU5 domain which eliminate spectrin binding activity of ankyrin-B. Ankyrin-B mutants lacking spectrin binding activity are normally targeted but do not reestablish beta(2)-spectrin in ankyrin-B(+/-) cardiomyocytes. However, both mutant forms of ankyrin-B are still capable of restoring inositol 1,4,5-trisphosphate receptor localization and normal contraction frequency of cardiomyocytes. Therefore, direct binding of beta(2)-spectrin to ankyrin-B is required for the normal targeting of beta(2)-spectrin in neonatal cardiomyocytes. In contrast, ankyrin-B localization and function are independent of beta(2)-spectrin. In summary, this work demonstrates that interaction between members of the ankyrin and beta-spectrin families previously established in erythrocytes and axon initial segments also occurs in neonatal cardiomyocytes with ankyrin-B and beta(2)-spectrin. This work also establishes a functional hierarchy in which ankyrin-B determines the localization of beta(2)-spectrin and operates independently of beta(2)-spectrin in its role in organizing membrane-spanning proteins.

Ankyrins are a closely related family of membrane adaptors required for organizing diverse membrane-spanning proteins including ion channels and transporters and L1 CAM cell adhesion molecules in physiologically important membrane domains (1,2). Targeted knock-out of ankyrin-G in the mouse cerebellum results in loss of clustering of the voltage-gated sodium channel Na v 1.6 and L1 CAMs neurofascin and NrCAM at Purkinje neuron initial segments (3,4). Knock-down of ankyrin-G in cultured epithelial cells by small interfering RNA results in loss of the lateral membrane domain (5). Loss of ankyrin-B in mice results in deficiency of Na,K-ATPase, Na/Ca exchanger, and inositol 1,4,5-trisphosphate (InsP 3 ) 1 receptor located in a specialized microdomain of T-tubules in adult cardiomyocytes (6). Humans with loss-of-function mutations in ankyrin-B and mice heterozygous for a null mutation in ankyrin-B (ankyrin-B ϩ/Ϫ mice) display stress-induced cardiac arrhythmia and sudden cardiac death (6,7).
The identity of cellular pathway(s) involved in ankyrin-dependent organization of membrane-spanning proteins is an important and currently unresolved question. Clues may come from elucidating proteins that interact with ankyrins. Erythrocyte spectrin, which is assembled with actin in a two-dimensional network, was the first ankyrin-binding protein to be identified. Ankyrin was in fact discovered based on its role as the membrane attachment site for spectrin in erythrocyte membranes (8 -11). Mutations of erythrocyte ankyrin result in deficiency of spectrin and cause hereditary spherocytosis in humans and mice (12)(13)(14). Members of the spectrin family also collaborate with ankyrins in cells other than erythrocytes. ␤ IV -Spectrin colocalizes with ankyrin-G at axon initial segments and is lost in ankyrin-G knock-out mice (4). Conversely, ankyrin-G exhibits reduced levels at axon initial segments and nodes of Ranvier of ␤ IV -spectrin mutant mice (15).
This work addresses the interaction between ankyrin-B and ␤ 2 -spectrin in neonatal cardiomyocytes. Ankyrin-B is required for localization of InsP 3 receptor and ryanodine receptor in the sarcoplasmic reticulum of these cells based on studies with primary cultures from ankyrin-B mutant mice (6, 7, 16 -18). Ankyrin-B contains a 62-kDa spectrin binding domain that associates with high affinity (K D ϳ 25 nM) with ␤ 2 -spectrin (19). ␤ 2 -Spectrin spliceoforms are expressed in adult heart (20), although their localization with respect to ankyrin-B has not been evaluated. Here we report that ␤ 2 -spectrin is a physiological binding partner for ankyrin-B in neonatal cardiomyocytes. We also provide evidence for a functional hierarchy in which ankyrin-B determines the localization of ␤ 2 -spectrin and operates independently of ␤ 2 -spectrin in its role in organizing membrane-spanning proteins.

EXPERIMENTAL PROCEDURES
Immunoblot Analysis-Quantitative Western blot analysis was performed using affinity-purified antibodies, 125 I-labeled protein A, and phosphorimaging (17).
Yeast Two-hybrid Assays-The yeast two-hybrid system was used to assay the interaction between ankyrin-B and ␤ 2 -spectrin. Full-length and truncated fragments of the 220-kDa ankyrin-B spectrin binding domain were PCR amplified and inserted into pAS2-1 (Clontech). The * This work was supported by the Howard Hughes Medical Institute and Johnson and Johnson. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. known ankyrin binding domain of ␤ 2 -spectrin was PCR amplified and inserted into pACT2. All constructs were completely sequenced and were free of mutations. Fusion protein expression in yeast was confirmed by Western blot analyses using commercially available binding domain and activation domain antibodies (Clontech).
Plasmids were cotransformed into yeast (AH109) and assayed using ADE2, HIS3, and lacZ selection (Clontech). Transformants that exhibited growth on ϪAde/His/Leu/Trp media were considered positive for interaction.
Random Mutagenesis-A mutant ankyrin-B spectrin binding domain library was created using GeneMorph PCR mutagenesis (Stratagene). The PCR protocol was optimized to obtain mutation frequency of 1 nucleotide mutation/ankyrin-B spectrin binding domain (ϳ1.75 kb). Mutagenized products were cloned into pAS2-1 vector, and clones were amplified to create a mutant library for yeast two-hybrid screening (ϳ4,000 clones). Yeast were cotransformed with mutant library and pACT2 ␤ 2 -spectrin ankyrin binding domain (residues 1563-2093) and plated on ϪLT media. Single colonies were replicated on ϪALT and ϪAHLT plates to identify cotransformants that did not interact with ␤ 2 -spectrin. Noninteracting ankyrin-B clones were analyzed by immunoblot to confirm absence of stop codons, and positive clones were then analyzed by PCR and sequencing. Clones with only one mutation were subcloned back into the bait plasmid, and loss-of-interaction was confirmed.
Immunofluorescence-Neonatal cardiomyocytes were washed with phosphate-buffered saline, pH 7.4, and fixed in warm 4% paraformaldehyde (37°C). Cells were blocked/permeabilized in phosphate-buffered saline containing 0.075% Triton X-100 and 3% fish oil gelatin (Sigma) and incubated in primary antibody overnight at 4°C. After washes (phosphate-buffered saline plus 0.1% Triton X-100), cells were incubated in secondary antibody (Alexa 488, 568, 633; Molecular Probes) for 8 h at 4°C and mounted using Vectashield (Vector) and no. 1 coverslips. Images were collected on Zeiss 510 Meta confocal microscope (100 power oil 1.45 NA (Zeiss), pinhole equals 1.0 Airy Disc) using Carl Zeiss Imaging software. All channels were collected on PMT3. Images were imported into Adobe Photoshop for cropping and linear contrast adjustment. Three-dimensional images were created using 0.18-m-thick Zscans and assembled using Volocity Software (Improvision).
Statistics-When appropriate, data were analyzed using a two-tailed Student's t test, and values less than p Ͻ 0.05 were considered significant. Values are expressed as the mean Ϯ S.D.

Identification of Mutations Eliminating Spectrin
Binding Activity of Ankyrin-B-We needed ankyrin-B mutants lacking spectrin binding activity to evaluate the potential role of ␤ 2spectrin in cellular localization and function of ankyrin-B (see below). Two strategies were employed to accomplish this goal: first, identify the minimal spectrin binding domain and then evaluate alanine-scanning mutations of surface residues within this domain; second, use random mutagenesis to create loss-of-binding mutations. Both approaches utilized the yeast two-hybrid assay. We constructed full-length and truncated polypeptides of the ankyrin-B spectrin binding domain (residues 862-1443) fused to the GAL4 DNA binding domain (see Fig. 1B). We also fused the minimal ankyrin binding domain of ␤ 2 -spectrin (repeats 13-17, residues 1563-2093 (23)) with the GAL4 activation domain. As expected, the 62-kDa ankyrin-B domain interacts with ␤ 2 -spectrin residues 1563-2093 in yeast (Fig. 1B). Truncated ankyrin-B constructs were designed based on the previous finding that limited proteolysis of the spectrin binding domain of ankyrin-R yields a subdomain corresponding to the ZU5 domain plus an additional C-terminal sequence (24). ZU5 domains are a feature shared with the proteins ZO-1 and Unc5 (smart.embl-heidelberg.de). The predicted ZU5 domain did not have ␤ 2 -spectrin binding activity. However, the 55 residues C-terminal to the predicted ZU5 domain have homology with a stretch of Unc5 but are missing in ZO-1 (not shown). The predicted ZU5 domain plus 55 C-terminal residues had ␤ 2 -spectrin binding activity equivalent to full-length constructs, whereas constructs N-terminal and C-terminal to this stretch were inactive (Fig. 1B). These results define a 160amino acid ␤ 2 -spectrin binding sequence in the 62-kDa ankyrin-B domain.
Alanine-scanning mutagenesis of predicted surface residues on the minimal spectrin binding domain (Fig. 1C, construct D9, amino acids 966 -1125) was performed to identify mutants lacking spectrin binding activity. A total of 13 sets of residues selected for high probability of location on the surface of the folded domain (i.e. 2 or more consecutive charged amino acids) were changed to alanines (two to four alanine substitutions/ site; Fig. 1C). These mutants were cotransformed with ␤ 2spectrin in yeast. 11 of the 13 mutants did not affect the ankyrin-B/␤ 2 -spectrin interaction (Fig. 1C). However, two alanine mutants affected ankyrin-B interaction with ␤ 2spectrin. One mutant, DAR976AAA, completely abolished interaction with ␤ 2 -spectrin, whereas the second mutant, ENGD1070AAGA, reduced but did not eliminate the ankyrin-B/␤ 2 -spectrin interaction (we observed minimal growth on AHLT medium; Fig. 1C).
We also used random mutagenesis with yeast two-hybrid analysis to identify additional loss-of-binding mutations and to confirm independently the minimal spectrin binding domain. A mutant ankyrin-B spectrin binding domain library was created using PCR mutagenesis (see "Experimental Procedures"). Yeast cells were cotransformed with the mutagenized library and ␤ 2 -spectrin ankyrin binding domain (residues 1563-2093). Ankyrin-B mutants that did not interact with ␤ 2 -spectrin and produced full-length ankyrin-B spectrin binding domain by immunoblot analysis were sequenced completely. Clones with only one mutation were subcloned back into the bait plasmid, and loss-of-interaction was confirmed. A screen of ϳ4,000 clones identified one that did not interact with ␤ 2 -spectrin. This mutant contained a single base mutation (G2998C) resulting in an alanine-to-proline substitution at amino acid 1000 (A1000P; Fig. 1B). A1000P is located within the minimal spectrin binding construct identified by truncation analysis (Fig. 1B) and is very close to the DAR976AAA loss-of-binding mutant.
These experiments identified two mutations that eliminate spectrin binding activity of ankyrin-B and established a domain encompassing the predicted ZU5 domain as the minimal spectrin binding site of ankyrin-B. The 160-amino acid region, particularly residues surrounding DAR976, are highly conserved among the three human ankyrin gene products (R, B, G) and evolutionarily conserved in ankyrins from mouse to Drosophila and Caenorhabditis elegans (Fig. 1D), all of which associate with ␤-spectrin polypeptides. Therefore, the minimal ␤ 2 -spectrin binding domain of ankyrin-B likely is responsible for spectrin binding activity of other ankyrins. The ZU5 domains of Unc5 (Fig. 1D) and ZO-1 are highly divergent in this region, and these do not bind ␤ 2 -spectrin (not shown).

Reduction in Ankyrin-B Expression in Neonatal Cardiomyocytes Results in Abnormal Targeting and Expression of ␤ 2 -
Spectrin-We determined the predominant ␤ 2 -spectrin isoform(s) present in mouse neonatal cardiomyocytes using an affinity-purified ␤ 2 -spectrin antibody raised against spectrin repeats 4 -9 (21). This ␤ 2 -spectrin antibody does not cross-react with ␤ 1 -spectrin from red blood cell ghosts but does recognize a 274-kDa ␤ 2 -spectrin isoform in whole brain lysates ( Fig. 2A). Immunoblot analysis of wild-type neonatal mouse cardiomyocytes lysates revealed that 274-kDa ␤ 2 -spectrin is the major isoform in developing ventricular cardiomyocytes (Fig. 2B). We did not observe lower molecular mass ␤ 2 -spectrin isoforms in neonatal cardiomyocytes, including 240-or 180-kDa ␤ 2 -spectrin isoforms observed in adult heart (data not shown).
␤ 2 -Spectrin staining by immunofluorescence is nearly eliminated in ankyrin-B-null cardiomyocytes (the cell in this example is a cardiomyocyte as evidenced by ␣-actinin staining; Fig.  2E). Ankyrin-B in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes is localized in a chimeric pattern where distinct regions of the cell display nearly normal ankyrin-B localization, and other regions of the cell are nearly completely lacking ankyrin-B staining (Fig. 1F). ␤ 2 -Spectrin expression in ankyrin-B ϩ/Ϫ precisely mimics the subcellular chimeric distribution of ankyrin-B and is missing from the same areas of these cells lacking ankyrin-B (Fig. 1F). Loss of ankyrin-B in either ankyrin-B ϩ/Ϫ or ankyrin-B Ϫ/Ϫ cardiomyocytes does not affect the localization of other cardiomyocyte proteins marking the contractile apparatus (␣actinin), endoplasmic reticulum/sarcoplasmic reticulum (sarco-  966 -1070). B, 220-kDa ankyrin-B spectrin binding domain truncation constructs D1-D9 were tested for ability to bind ␤ 2 -spectrin by yeast twohybrid analysis. Positive interaction with ␤ 2 -spectrin is displayed on the AHLT selection plate (right). The minimal ␤ 2spectrin binding domain includes ankyrin-B residues 968 -1125 (construct D9). The predicted ZU5 domain in ankyrin-B (D8, residues 966 -1070) did not interact with ␤ 2 -spectrin. WT, wildtype; C, predicted surface residues of the minimal ␤ 2 -spectrin binding sequence were mutated to alanines and tested for ␤ 2 -spectrin binding. The Ala to Pro mutation (A1000P) was identified by random mutagenesis and the DAR976AAA mutation from alanine mutagenesis both blocked ␤ 2 -spectrin binding. The ENGD1067AAGA mutant showed reduced growth rate of yeast on AHLT selection consistent with a reduced affinity for ␤ 2 -spectrin. D, conserved sequence of the N terminus of the ankyrin-B ␤ 2spectrin binding site. DAR residues (boxed) critical for spectrin binding are conserved in ankyrins throughout evolution (hAnkB, human ankyrin-B; hAnkG, human ankyrin-G; hAnkR, human ankyrin-R; Unc-44, C. elegans ankyrin; dAnk1, Drosophila ankyrin 1). Human Unc-5 (hUnc-5) contains a ZU5 domain but does not interact with ␤ 2 -spectrin. plasmic reticulum calcium-ATPase 2 (SERCA2)), or nascent T-tubules (dihydropyridine receptor) (18).
We next tested whether abnormal localization of ␤ 2 -spectrin in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes could be restored to normal by expression of exogenous GFP-ankyrin-B. Reduction of ankyrin-B in ankyrin-B ϩ/Ϫ and ankyrin-B Ϫ/Ϫ neonatal cardiomyocytes leads to abnormal cardiomyocyte spontaneous contraction rates and abnormal localization and expression of ankyrin-B-associated proteins including InsP 3 receptor, ryanodine receptor, Na,K-ATPase, and Na/Ca exchanger (6,17,18). These abnormal phenotypes can be rescued by transfection of the cardiomyocytes with GFP-220-kDa ankyrin-B (6,17,18).
We expressed GFP-220-kDa ankyrin-B in 3-day-old ankyrin-B ϩ/Ϫ neonatal cardiomyocytes and 36 h later determined the localization of GFP-220-kDa ankyrin-B using an affinity-purified GFP antibody. GFP-220-kDa ankyrin-B was localized in a striated pattern similar to that of endogenous ankyrin-B (Fig.  3B). Transfection of GFP-220-kDa ankyrin-B rescues the expression and localization of ␤ 2 -spectrin in neonatal cardiomyocytes (Fig. 3B). These results demonstrate that ankyrin-B is required for ␤ 2 -spectrin targeting and expression in neonatal cardiomyocytes.
Direct Interaction with Ankyrin-B Is Required for ␤ 2 -Spectrin Targeting in Neonatal Cardiomyocytes-We used the ankyrin-B ϩ/Ϫ cardiomyocyte rescue assay and ankyrin-B mutants lacking ␤ 2 -spectrin binding to determine whether direct interaction between ankyrin-B and ␤ 2 -spectrin is required for ␤ 2spectrin targeting in neonatal cardiomyocytes. GFP-ankyrin-B EDE1082AAA, an ankyrin-B mutant in the spectrin binding region which does not affect binding to ␤ 2 -spectrin (Fig. 1C), is localized as wild-type GFP-220-kDa ankyrin-B and also rescues ␤ 2 -spectrin localization in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes (Fig. 4B). In contrast, GFP-220-kDa ankyrin-B mutants that lack ␤ 2 -spectrin binding (DAR976AAA and A1000P) did not rescue the localization of ␤ 2 -spectrin, even though they were normally expressed and correctly targeted in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes (Fig. 4, C and D). Therefore, direct binding of ␤ 2 -spectrin to ankyrin-B is required for the normal targeting of ␤ 2 -spectrin in neonatal cardiomyocytes. Moreover, localization of ankyrin-B in cardiomyocytes occurs independently of ␤ 2 -spectrin.

Ankyrin-B-dependent Targeting of the InsP 3 Receptor Is
Independent of ␤ 2 -Spectrin Activity-The role of ␤ 2 -spectrin in ankyrin-B-dependent targeting of the InsP 3 receptor was determined by rescue experiments using GFP-220-kDa ankyrin-B mutants lacking ␤ 2 -spectrin binding activity. Abnor-

FIG. 2. Reduction in ankyrin-B expression
in ankyrin-B ؉/؊ and ankyrin-B ؊/؊ neonatal cardiomyocytes results in decreased expression and abnormal targeting of ␤ 2 -spectrin. A, ␤ 2 -spectrin antibody does not recognize red blood cell ghost ␤ 1 -spectrin but does recognize brain ␤ 2 -spectrin. Coomassie blue (C.B.)-stained gel (left) and immunoblot (IB) (right) of total lysates from red blood cell ghost (G) and brain (B) are shown. B, quantitative immunoblot analysis of 220-kDa ankyrin-B and ␤ 2 -spectrin expression in neonatal cardiomyocytes from wild-type and ankyrin-B ϩ/Ϫ neonatal cardiomyocytes. C, protein levels were measured using 125 I-labeled protein A and phosphorimaging. Levels of ␣-actinin were monitored as a loading control. WT, wild-type. D-F, localization of ankyrin-B and ␤ 2 -spectrin in neonatal cardiomyocytes derived from wild-type, ankyrin-B Ϫ/Ϫ , and ankyrin-B ϩ/Ϫ mice. Reduction in ankyrin-B expression in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes results in abnormal organization of ␤ 2 -spectrin. The scale bar equals 10 m. Ankyrin-B Ϫ/Ϫ neonatal cardiomyocytes were labeled with ␣-actinin antibody to confirm that the cell was a cardiomyocyte. mal localization and expression of InsP 3 receptor in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes are rescued by expression of GFP-220-kDa ankyrin-B (Fig. 5, A-C) (6,17). GFP-220-kDa ankyrin-B DAR976AAA or A1000P mutants that do not bind ␤ 2 -spectrin do restore normal InsP 3 receptor expression and localization (Fig. 5, D and E). In contrast, GFP-220-kDa ankyrin-B EDE1082AAA is ineffective in restoring the localization of InsP 3 receptor in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes (InsP 3 receptors are localized in small puncta throughout the cell), even though this mutant has full ␤ 2 -spectrin binding activity (Fig. 1C) and rescues localization of ␤ 2 -spectrin in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes (Fig. 4B). These results demonstrate that ankyrin-B-dependent activity in targeting InsP 3 receptors in neonatal cardiomyocytes is independent of ␤ 2 -spectrin. These results also demonstrate that neither the A1000P nor DAR976AAA mutations affect the folding of fulllength ankyrin-B because these mutants are normally targeted and have full activity in rescuing InsP 3 receptor localization.

Ankyrin-B and ␤ 2 -Spectrin are Colocalized in an Intracellular Striated Compartment Distinct from Known Membrane
Structures-We compared the localization of ankyrin-B and ␤ 2 -spectrin in neonatal cardiomyocytes using confocal microscopy. ␤ 2 -Spectrin and ankyrin-B are colocalized in a transverse striated network in wild-type neonatal cardiomyocytes immu-

FIG. 4. Expression and localization of ␤ 2 -spectrin require direct interaction with 220-kDa ankyrin-B.
A, GFP-220-kDa ankyrin-B rescues abnormal localization of ␤ 2 -spectrin in ankyrin-B ϩ/Ϫ cardiomyocytes. B, GFP-220-kDa ankyrin-B EDE1082AAA has full-␤ 2spectrin binding activity and rescues ␤ 2 -spectrin localization in ankyrin-B ϩ/Ϫ neonatal cardiomyocytes. C and D, expression of GFP-220-kDa ankyrin-B DAR976AAA (C) or A1000P (D) (both lack ␤ 2spectrin binding activity) does not rescue abnormal ␤ 2 -spectrin targeting in ankyrin-B ϩ/Ϫ cardiomyocytes even though both mutants are normally expressed and localized. Localization of GFP-ankyrin-B was assessed using affinity-purified GFP antibody, and ␤ 2 -spectrin localization was determined using affinity-purified ␤ 2 -spectrin antibody. The scale bar for all images equals 10 m. nolabeled with affinity-purified antibodies (Fig. 7A). The localization of ␤ 2 -spectrin and ankyrin-B was intracellular as assessed by three-dimensional reconstructions of a Z-stack of consecutive confocal scans obtained at intervals of 0.18 m (see "Experimental Procedures"). ␤ 2 -Spectrin overlaps with ankyrin-B over the M-line near the sarcolemma but also intracellularly throughout the cell (see Fig. 7, B and C; in C the image is rotated to observe intracellular colocalization). Additionally, we identified longitudinal ankyrin-B staining connecting M-line staining that was not coincident with ␤ 2 -spectrin labeling (see red channel in Fig. 7C). Evidence that the striated network is intracellular is that a marker for the sarcolemma (Na/Ca exchanger) exhibits minimal overlap with ␤ 2 -spectrin or ankyrin-B (Supplemental Fig. 1). Moreover, ␤ 2 -spectrin and ankyrin-B exhibit precise juxtaposition in three-dimensional images with intracellular organelles (see below).
The finding that ␤ 2 -spectrin requires ankyrin-B for cellular targeting in cardiomyocytes was not anticipated. Studies of assembly of the erythrocyte skeleton have concluded that ankyrin associates with a preassembled spectrin skeleton (25). In addition, ␤-spectrins associate with phosphatidylinositol lipids through their pleckstrin homology domains (26). Moreover ␤ 2 -spectrin has ankyrin-independent membrane binding sites (19,(27)(28)(29) and also associates with ␣-catenin (30). The current findings do not exclude the possibility that ankyrin-independent interactions of ␤ 2 -spectrin with proteins and phospholipids also contribute to its localization and function. It will be of Note that InsP 3 receptor localization is rescued by two GFP-220-kDa ankyrin-B mutants that do not associate with ␤ 2 -spectrin. Ankyrin-B ϩ/Ϫ cardiomyocytes expressing GFP-220-kDa ankyrin-B EDE1082AAA at similar levels display abnormal localization and expression of InsP 3 receptors (small puncta throughout the cell).
interest to evaluate localization of mutant forms of ␤ 2 -spectrin lacking these activities.
A second surprising conclusion of this work is that ␤ 2 -spectrin and ankyrin-B are in an intracellular compartment com-pletely distinct from Golgi. Polypeptides immunoreactive with ankyrin and spectrin have been observed associated with Golgi at the level of immunofluorescence (31)(32)(33). However, the source of the spectrin-related Golgi staining recently has been Sevenday cardiomyocytes cultures were immunostained with antibodies to A, ankyrin-B (red) and ␣-actinin (green); B, ankyrin-B (red) and dihydropyridine receptor (green); C, ␤ 2 -spectrin (red) and InsP 3 receptor (green); and D, ␤ 2 -spectrin (green) and ryanodine receptor (red). Images are three-dimensional reconstructions of 10 -20 Z-scans through the cardiomyocyte at 0.18-m intervals.
identified as nesprin-1␤/syne-1B (34), a large protein with spectrin repeats but not a member of the ␤-spectrin family (35). ␤ 3 -Spectrin has been associated with markers for the Golgi (36). In addition, 119-kDa ankyrin-G is localized in a subcellular pattern that overlaps with Golgi markers in subconfluent Madin-Darby canine kidney cells (37). It will be of interest to evaluate localization of ␤ 3 -spectrin and 119-kDa ankyrin-G using the three-dimensional rendering and high resolution confocal microscopy methods employed in this work.
A third unexpected finding was that ␤ 2 -spectrin is not required for ankyrin-B-dependent activity in targeting InsP 3 receptors in cardiomyocytes. Spectrin and ankyrin function collaboratively in stabilizing the plasma membrane of erythrocytes (12)(13)(14). Moreover, the phenotype of ␤ IV spectrindeficient mice is similar although less severe than ankyrin-G mutant mice (3,4,15). It is important to emphasize that our results were obtained using cultured neonatal cardiomyocytes and may not necessarily extend to neonatal heart tissue or to adult cardiomyocytes. It is of interest in this regard that loss of the sole ␤ 2 -spectrin ortholog in C. elegans is compatible with normal embryonic morphogenesis but not with survival after hatching (38,39). Moreover, mice homozygous for a null mutation in ␤ 2 -spectrin display cardiac abnormalities as well as other defects and die in midgestation, consistent with a critical role for ␤ 2 -spectrin in the heart (40). One possible role for ␤ 2 -spectrin in cardiomyocytes would be to provide a mechanical protection for the ankyrin-B compartment during cardiac contractions. Such a stabilizing function would not necessarily be important under conditions of tissue culture but could be critical in the context of an actively beating heart. This work defines the minimal spectrin binding domain as a sequence highly conserved in ankyrin polypeptides and encompassing a predicted ZU5 domain (smart.embl-heidelberg.de). This domain is also present in the tight junction protein ZO-1 and Unc5 netrin receptor (41,42), although the level of homol-ogy is low, and these proteins do not have spectrin binding activity (data not shown). Interestingly, the computer-predicted ZU5 domain in ankyrin-B (residues 966 -1070) did not interact with ␤ 2 -spectrin. However, full-spectrin binding occurred when we extended the boundaries of the ZU5 domain to 966 -1125 based on a similar sequence flanking the ZU5 domain of Unc5. Ankyrins associate with ␤ 1 -and ␤ 2 -spectrins with 2-3-fold different affinities (19,43), which raises the question of whether the minimal spectrin binding domain retains the ability to distinguish between spectrin isoforms. It is of interest in this regard that deletion of the N-terminal residues of the ankyrin-R spectrin binding domain results in a 20-fold loss of affinity from 10 to 200 nM (24), which would still register as a positive in yeast two-hybrid assays. It will be important to compare binding affinities of the minimal and full-length spectrin binding domains in biochemical assays.
In summary, this work demonstrates that interaction between members of the ankyrin and ␤-spectrin families previously established in erythrocytes and axon initial segments also occurs in neonatal cardiomyocytes with ankyrin-B and ␤ 2 -spectrin. This work also establishes a functional hierarchy in which ankyrin-B determines the localization of ␤ 2 -spectrin and operates independently of ␤ 2 -spectrin in its role in organizing membrane-spanning proteins. The supportive downstream role for ␤ 2 -spectrin in neonatal cardiomyocytes is consistent with results in C. elegans, where loss of ␤-spectrin is compatible with normal embryonic development. Such a "division of labor" between ankyrin and spectrin makes teleological sense considering the high level of diversity in protein interactions of ankyrin caused by ANK repeats and the properties of spectrin that allow formation of two-dimensional networks with actin. In contrast to erythrocytes and neurons, where ankyrins and spectrins are attached to the plasma membrane, the ankyrin-B-␤ 2 -spectrin complex of cardiomyocytes is associated with intracellular membranes. The ankyrin-B-␤ 2 -spectrin compartment is distinct from known organelles or striated membrane compartments at the level of light microscopy. Elucidation of the composition and functions of the ankyrin-B-␤ 2spectrin compartment as well as its evolutionary connection with the plasma membrane will be important goals for future research.