Metabolic Biotinylation as a Probe of Supramolecular Structure of the Triad Junction in Skeletal Muscle*

Excitation-contraction coupling in skeletal muscle involves conformational coupling between dihydropyridine receptors (DHPRs) in the plasma membrane and ryanodine receptors (RyRs) in the sarcoplasmic reticulum. However, it remains uncertain what regions, if any, of the two proteins interact with one another. Toward this end, it would be valuable to know the spatial interrelationships of DHPRs and RyRs within plasma membrane/sarcoplasmic reticulum junctions. Here we describe a new approach based on metabolic incorporation of biotin into targeted sites of the DHPR. To accomplish this, cDNAs were constructed with a biotin acceptor domain (BAD) fused to selected sites of the DHPR, with fluorescent protein (XFP) attached at a second site. All of the BAD-tagged constructs properly targeted to junctions (as indicted by small puncta of XFP) and were functional for excitation-contraction coupling. To determine whether the introduced BAD was biotinylated and accessible to avidin (∼60 kDa), myotubes were fixed, permeablized, and exposed to fluorescently labeled avidin. Upon expression in β1-null or dysgenic (α1S-null) myotubes, punctate avidin fluorescence co-localized with the XFP puncta for BAD attached to the β1a N- or C-terminals, or the α1S N-terminal or II-III loop. However, BAD fused to the α1S C-terminal was inaccessible to avidin in dysgenic myotubes (containing RyR1). In contrast, this site was accessible to avidin when the identical construct was expressed in dyspedic myotubes lacking RyR1. These results indicate that avidin has access to a number of sites of the DHPR within fully assembled (RyR1-containing) junctions, but not to the α1S C-terminal, which appears to be occluded by the presence of RyR1.

In skeletal muscle, two major proteins involved in excitationcontraction (EC) 1 coupling are the dihydropyridine receptor (DHPR), a voltage-gated calcium channel in the plasma mem-brane, and the ryanodine receptor (RyR), a calcium release channel in the sarcoplasmic reticulum (SR). Physically, the skeletal DHPRs are grouped in "tetrads," and each DHPR within a tetrad is located in exact correspondence to one of the four subunits of RyR1 (1,2). Functionally, a bidirectional interaction occurs in skeletal muscle between the DHPR and RyR1. Depolarization of the plasma membrane causes transmission of an orthograde signal from the DHPR (3,4) to the RyR; this in turn causes Ca 2ϩ release via RyR1 that does not require the entry of extracellular Ca 2ϩ (5,6). In addition to this orthograde signal, there is a retrograde signal whereby RyR1 increases the magnitude of the voltage-gated calcium current carried through the DHPR (L-type current, Ref. 7).
Despite a large number of studies involving functional analyses of intact cells and biochemical analyses of cell fractions and isolated proteins, the identity of the protein-protein links necessary for the functional and morphological coupling between DHPRs and RyRs in skeletal muscle remains unresolved. The preceding paper (8) described a new approach in which the fluorescent resonance energy transfer efficiency of a cyan fluorescent protein-yellow fluorescent protein (CFP-YFP) tandem was used as an indirect indicator of sites of possible proximity between the DHPR and RyR. Here we report on another novel approach in which the topology of the plasma membrane/SR junctions is probed by determining whether avidin can access biotin introduced at different sites of the DHPR. The site-specific introduction of biotin was based on the metabolic incorporation that normally occurs only for the small number of cellular enzymes that contain biotin as an essential co-factor. The biotin is incorporated into these enzymes by the catalytic action of biotin protein ligase, with the result that biotin is attached to a lysine contained within a biotin acceptor domain (BAD). In mammalian cells, metabolic biotinylation is effective for fusion proteins containing a minimal BAD (9). We constructed cDNAs that encoded a 70-or 97-residue BAD fused to sites of the DHPR, as well as a variant of green fluorescent protein (XFP) fused at a second site as an independent reporter of localization. It was then possible to determine whether the BAD had been biotinylated and was accessible for binding of fluorescently labeled avidin.
The XFP/BAD/DHPR fusions were found to be functional in EC coupling and to target to discrete foci indicative of membrane junctions. Moreover, effective, metabolic biotinylation occurred for BAD at the N-and C-terminals of the ␤ 1a and ␣ 1S subunits of the DHPR, as well as at an internal site within the ␣ 1S II-III loop. Nearly all of these sites were accessible to avidin, even though the DHPRs were present within fully assembled junctions. In contrast, the C-terminal of ␣ 1S was inaccessible to avidin in junctions that contained RyR1, but was accessible in junctions lacking RyR1. Thus, RyR1 appears to occlude access to the C-terminal of ␣ 1S but not to the other sites of the DHPR (␤ 1a N-and C-terminals, ␣ 1S N-terminal and II-III loop). However, a surprising number of sites within DH-PRs localized to fully assembled junctions are accessible to avidin, a molecule of substantial size.

cDNA Constructs
Biotin Acceptor Domain-The BAD was extracted from the PinPoint Xa-1 expression vector (Promega, Madison, WI) containing the sequence encoding the PSTCD (Propionibacterium shermanii transcarboxylase domain). This PinPoint vector encodes a BAD of 123 amino acids in length with an approximate molecular mass of 13 kDa. The crystal structure of this biotin acceptor domain (Protein Data Bank code 1DCZ, Ref. 10) is shown in Fig. 1. The biotinylated lysine is located at position 89 of this sequence (highlighted in yellow). Fig. 2 summarizes the positions at which the BAD was incorporated into the ␣ 1S and ␤ 1a subunits of the skeletal muscle DHPR. The biotin acceptor domain in these constructs was either 70 or 97 amino acids in length. The cDNAs for the fusion proteins were constructed as follows, with restriction digests and sequencing used for verification.
GFP-␣ 1S -BAD-For this and all other C-terminal fusions (except for GFP-BAD, as described later), PCR mutagensis (QuickChange Kit, Stratagene, La Jolla, CA) was used to introduce two KpnI restriction sites into the PSTCD sequence of the PinPoint Xa-1 plasmid. One KpnI site was inserted directly before amino acid Glu 53 (by inserting GGGG-TACCC-5Ј to nucleotide G 157 , where 1 indicates the first nucleotide of the PSTCD coding sequence). The second KpnI site, preceded by a stop codon, was introduced after amino acid Gly 122 by inserting TAGGTAC-CGG-3Ј to nucleotide G 366 . The 219-base pair (bp) fragment yielded by KpnI digestion of this modified PinPoint plasmid was inserted into KpnI-digested GFP-␣ 1S long and GFP-␣ 1S short, which were constructed essentially as described by Papadopoulos et al. (8) for YFP-␣ 1S long and YFP-␣ 1S short. In these two constructs, the rabbit ␣ 1S sequence (3) was terminated, respectively, after either amino acid 1860 or 1667 (fulllength ␣ 1S has 1873 residues). The GFP-␣ 1S long or GFP-␣ 1S short was connected to the 70-residue BAD via a 4-or 3-residue linker, respectively.
XFP-␣ 1S (I-II) BAD-The 219-bp KpnI BAD fragment was ligated to KpnI-digested XFP-␣ 1S (I-II), which was constructed as described in Papadopoulos et al. (8), where the fluorescent protein moiety was obtained from pECFP-N1, pEGFP-N1, or pEYFP-N1 (BD Biosciences, Palo Alto, CA). This construct contained residues 1-671 of rabbit ␣ 1S (including repeats I and II together with a small portion of the II-III loop). A 2-residue linker connected XFP-␣ 1S (I-II) to the 70-residue BAD. This construct was co-expressed with the plasmid ␣ 1S (III-IV) (8), which encoded residues 686 -1860 of ␣ 1S (including a large part of the cytoplasmic II-III loop in addition to repeats III and IV).
XFP-␤ 1a -BAD-The 219-bp KpnI BAD fragment was ligated to KpnI-digested XFP-␤ 1a (constructed as described in Ref. 8). The 70-residue BAD was fused to the C-terminal of ␤ 1a via a one-residue linker. BAD-␣ 1S -YFP-To fuse the BAD to the N-terminal of ␣ 1S , the Pin-Point Xa-1 plasmid was digested with HindIII and SalI. This BAD sequence codes for residues 26 -133 of the original PSTCD sequence. The ␣ 1S short-YFP plasmid (8) was digested with XhoI and HindIII removing a small plasmid fragment before the ␣ 1S N-terminal. The resulting ␣ 1S plasmid and the BAD fragment were co-ligated producing BAD-␣ 1S short-YFP with a 16-residue linker connecting the 97-residue BAD to the N-terminal of ␣ 1S short-YFP.
BAD-␣ 1S (III-IV)-YFP-A 347-bp BAD fragment (encoding BAD residues 26 -128) was isolated from BAD-␣ 1S short-YFP by restriction cuts with SacI and NheI. The plasmid ␣ 1S (III-IV)-YFP (8) was digested with SacI and NheI, opening the plasmid before the N-terminal of ␣ 1S repeat III, and the BAD cDNA was inserted into this site. A 6-residue linker connected the 97-residue BAD to the N-terminal of ␣ 1S (III-IV)-YFP. This construct was co-expressed with the plasmid ␣ 1S (I-II) (8).
GFP-BAD-Using PCR mutagenesis, a KpnI restriction site was introduced into the PSTCD sequence of the PinPoint Xa-1 plasmid directly before amino acid Glu 53 (by inserting GGGGTACC-5Ј to nucleotide G 157 ). A stop codon followed by a KpnI site was introduced after amino acid Gly 122 by inserting TAGGTACCGG-3Ј to nucleotide G 366 . The modified plasmid was digested with KpnI, and the 218-bp fragment was inserted into the KpnI-cut multiple cloning site of pEGFP-C1 (BD Biosciences). A 17-residue linker connected the 70-residue BAD to the C-terminal of GFP.

Expression of cDNA
Primary cultures of myotubes isolated from newborn dysgenic, dyspedic, or ␤ 1 -null mice were prepared as described previously (11). Myoblasts were plated on ECL-coated (Upstate Biotechnology, Lake Placid, NY) 35-mm plastic culture dishes or dishes with glass coverslip bottoms (MatTek, Ashland, MA) and grown for 6 -7 days in a humidified 37°C FIG. 1. Biotin acceptor domain constructs used to probe the ability of avidin to access sites within the skeletal DHPR. Left, the source of the BAD was the PinPoint Xa-1 vector, which contains the coding sequence for a domain of one subunit of transcarboxylase from P. shermanii (PSTCD), illustrated here in single letter code. The two different segments of the PSTCD used to introduce the BAD into DHPR fusion proteins are indicated by the numbered arrows. The color coding used is the same as that used for the crystal structure. Right, the crystal structure is illustrated as a ball-and-stick model, with the appended single-letter code indicating a lack of published structure. Because it is not part of the crystal structure, biotin (indicated in orange) has uncertain orientation with respect to the lysine (yellow) to which it is covalently attached, but the flexibility of the linkage suggests substantial mobility. Rose and aqua indicate the N-and C-terminals, respectively, of the 70-residue BAD sequence.
incubator with 5% CO 2 . Approximately 1 week after plating, myotubes were microinjected (12) in a single nucleus with one of the above cDNA constructs (5-100 ng/l). After injection, the cells were changed into a culture medium containing normal levels of biotin (1 M). For Western blot analyses, immortal dysgenic myotubes were transfected with cDNA constructs using LT-1 transfection reagent (Mirus, Madison, WI).

Electrically evoked contractions
Contractions were elicited by 10-ms, 55-100 V stimulus applied via an extracellular pipette placed near intact myotubes expressing constructs of interest. Images of these myotubes were acquired at a rate of 40 -50 Hz. The contractions were quantified by measuring the movement of an identifiable portion of a myotube across the visual field.

NeutrAvidin Staining
Two days after injection, myotubes were washed in PBS (calciumand magnesium-free), and then fixed with 4% paraformaldehyde in PBS for 20 min. The cells were then permeabilized with 0.1% Triton X-100 (Sigma) in PBS for 30 min and incubated in PBS blocking solution for 1 h. The cells were then exposed to NeutrAvidin-tetramethylrhodamine (here after referred to as "avidin"; Molecular Probes, Eugene, OR) or streptavidin-Cy3 (Jackson ImmunoResearch, West Grove, PA) (1:2000 -1:5000) in 0.1% Triton X-100/PBS blocking solution for 1 h in the dark. The cells were washed with 0.1% Triton X-100 in PBS followed by PBS.

Western Blot Analysis
Samples were run on SDS-PAGE gels (4 -20% precast Tris-HCl; Bio-Rad) in a Mini-PROTEAN II electrophoresis cell (Bio-Rad) according to the instructions. Protein was transferred to nitrocellulose using a Mini Trans-Blot electrophoretic transfer cell (Bio-Rad) at 100 V for 1 h. Blots for GFP detection were blocked in 3% milk, PBS and exposed to monoclonal GFP antibody (1:1,000 in 3% milk/PBS) (Chemicon, Temecula, CA) overnight at room temperature. The blot was washed with 0.1% Tween 20, PBS and then incubated with goat anti-mouse secondary antibody (1:10,000 in 3% milk/PBS) (Pierce, Rockford, IL) for 30 min. Blots for biotin detection were blocked using Startingblock (Pierce), and then incubated with streptavidin-poly-HRP (Pierce), diluted (1:10,000 in Poly-HRP Dilution buffer, Pierce) for 30 min. Both GFP and biotin detection blots were then washed with 0.1% Tween 20, PBS. Blots were developed with Super Signal West Femto detection kit (Pierce).

Confocal Microscopy
Fluorescence was analyzed using an Axiovert/LSM 510 META laserscanning confocal microscope (Zeiss, Thornwood, NY). Excitation and emission parameters for each fluorophore were set as follows: CFP, excitation at 458 and 458 nm dichroic, and emission using a 465-495-nm band pass filter (Chroma Technology Corp., Rockingham, VT); GFP, excitation at 488 and 488/543 nm dual dichroic, and emission with a 505-530-nm band pass filter; YFP, excitation 488 and 488/543 nm dual dichroic, and emission with a 505-530-nm band pass filter (to allow separation from rhodamine fluorescence); rhodamine, excitation with 543 and 488/543 nm dual dichroic and emission with a 560-nm long pass filter. Cells were viewed with a ϫ40 (1.3 NA) or ϫ63 (1.4 NA) oil immersion objective.

RESULTS
As a tool for probing the supramolecular architecture of DHPRs and RyRs assembled into plasma membrane/SR junctions, we introduced a BAD at different sites of the DHPR (Fig.  2) to determine whether a 60-kDa avidin molecule could access those sites. BAD fusion proteins were expressed in myotubes from dysgenic mice (lacking functional ␣ 1S channels), dyspedic mice (lacking RyR1), or ␤ 1 -null mice (lacking ␤ 1 subunits), which were cultured in the presence of normal biotin levels (1 m). Previous studies have shown that a shortened 70-amino acid PSTCD is sufficient for endogenous enzymatic biotinylation in mammalian cells (9), and our fusion constructs included BADs of either 70 or 97 residues in length. As a test for endogenous biotinylation activity, we overexpressed a GFP-BAD fusion protein in dysgenic myotubes. The presence of biotinylated proteins was then assayed by fixing and permeabilizing the cells, exposing them to streptavidin-Cy3, and im-aging with confocal microscopy. Cells displaying strong green fluorescence also displayed bright red fluorescence (Fig. 3), whereas cells without detectable green fluorescence showed only a low level of background, red fluorescence (not shown). Additionally, within cells expressing GFP-BAD, there was a strong correlation between the regions displaying higher intensities of green and those displaying higher intensities of red fluorescence (Fig. 3), with one important exception. Specifically, the nuclei displayed stronger GFP fluorescence than Cy3 fluorescence, indicating that streptavidin-Cy3 has restricted access to the inside of nuclei. As a second confirmation of endogenous biotinylation, Western blot analysis of GFP-BAD expression in transiently transfected immortal dysgenic myotubes showed a ϳ37-kDa biotinylated protein when probed with anti-GFP antibody or streptavidin-Poly-HRP (Fig. 3). Thus, histochemistry and Western blot analysis demonstrated that the metabolic biotinylation pathway in myotubes is able to effectively biotinylate the BAD fusion proteins.
Previous studies have shown that ␣ 1S and ␤ 1a subunits contained in junctionally targeted DHPRs are clustered in punctate foci, typically near the cell surface (13). Such punctate foci of XFP fluorescence were observed for all of the constructs examined here in which XFP and BAD were fused to sites of ␣ 1S or ␤ 1a (Figs. [5][6][7][8][9][10]. Moreover, all of these constructs were able to support EC coupling, as indicated by electrically evoked contractions. In particular, evoked contractions were observed (  (20/27). Moreover, these same constructs mediated Ca 2ϩ currents that did not differ dramatically from those of the wild-type constructs expressed in myotubes (14). Thus, the incorporation of a BAD did not appear to interfere with the targeting or essential function of the ␣ 1S or ␤ 1a fusion proteins. and YFP-␤ 1a -BAD. To determine whether BAD on the C-terminal had been biotinylated and was accessible to avidin, myotubes were fixed, permeabilized, and then exposed to avidinrhodamine. Fig. 5 shows that red fluorescent foci were both present and co-localized with the yellow foci, as was observed in 28/35 dysgenic myotubes examined. A similar pattern of colocalizing red and yellow foci was also present when XFP-␤ 1a -BAD was expressed in ␤ 1 -null myotubes (9/9 cells). Thus, a large avidin molecule has access to a site located very near to the C-terminal of ␤ 1a .
In endogenous biotin-containing proteins, the BAD is usually located near the C-terminal. Nonetheless, effective biotinylation was achieved when a BAD was fused to the N-terminal of the ␤ 1a subunit, as is indicated by the occurrence of co-localizing red and yellow fluorescent puncta in ␤ 1 -null myotubes expressing BAD-␤ 1a -YFP ( Fig. 6; 18/18 cells) and in dysgenic myotubes co-expressing both BAD-␤ 1a -YFP and unlabeled ␣ 1S (12/12 cells). Moreover, coincident red and yellow foci were also observed in dysgenic myotubes expressing BAD-␣ 1S -YFP ( Fig.  7; 19/19 cells). Thus, avidin has access to sites near both the N- FIG. 3. Endogenous biotinylation of GFP-BAD in skeletal myotubes. Left, a dysgenic myotube expressing GFP-BAD displays both GFP fluorescence and Cy3streptavidin staining, which have corresponding variations in intensity except intra-nuclearly (arrows) where GFP fluorescence exceeds Cy3 fluorescence (evident in the overlay). Thus, it appears that the BAD sequence is effectively biotinylated and that avidin has access to biotinylated GFP within the myotube. Scale bar ϭ 10 m. Right, Western blot analysis of GFP-BAD expression in cell lysates from transiently transfected dysgenic myotubes. When blots were probed with anti-GFP or streptavidin-Poly-HRP, a ϳ37-kDa band was visible in both suggesting that the protein both contained GFP and was biotinylated. HRP, horseradish peroxidase.  5. The C-terminal of ␤ 1a is biotinylated and accessible to avidin within fully assembled junctions. YFP-␤ 1a -BAD coexpressed with unlabeled ␣ 1S in a dysgenic myotube produced colocalized fluorescent foci of YFP (left) avidin-rhodamine (center), as indicated by orange in the overlay (right). Because ␤ 1a is freely diffusible in the absence of ␣ 1S , these fluorescent foci likely represent the insertion of ␤ 1a into fully assembled junctions that occur between the plasma membrane and SR and contain the full complement of junctional proteins including ␣ 1S and RyR1. Scale bar ϭ 5 m. and C-terminals of ␤ 1a and the N-terminal of ␣ 1S within fully assembled junctions.
An important goal of the present work was to probe the environment of the ␣ 1S (II-III) cytoplasmic loop. Because the results described above showed that either N-or C-terminal BAD could be effectively biotinylated, the initial approach was to separate ␣ 1S into two fragments (one containing repeats I-II with the proximal portion of the loop and the other containing the distal portion of the loop together with repeats III-IV). These constructs allowed attachment of a BAD to either the C-terminal (residue 671) of the proximal II-III loop, "XFP-␣ 1S (I-II)-BAD," or the N-terminal (residue 686) of the distal II-III loop, "BAD-␣ 1S (III-IV)-YFP" (residues 672-685 were omitted). These BAD containing constructs were then expressed with the appropriate complementary fragment: ␣ 1S (III-IV) (in some instances N-terminal tagged with YFP) or ␣ 1S (I-II), respectively. Previous studies have shown that two fragment constructs of ␣ 1S (without a BAD) are able to restore EC coupling in dysgenic myotubes (15,16). Indeed, co-expression of either YFP-␣ 1S (I-II)-BAD ϩ ␣ 1S (III-IV), or ␣ 1S (I-II) ϩ BAD-␣ 1S (III-IV)-YFP, in dysgenic myotubes resulted in restoration of EC coupling (see above) and the appearance of yellow punctate foci (Fig. 8, A and B), indicating that the BAD-tagged fragments were correctly targeted to junctions. After avidinrhodamine staining, red fluorescent puncta co-localized with As a test of the ability of a BAD to become biotinylated when placed in the center of an intact protein, and of the environment of the II-III loop in a one-piece ␣ 1S construct, we constructed ␣ 1S (I-II)-BAD-(III-IV)-YFP, where the BAD replaced residues 672-685 of ␣ 1S . The BAD introduced into the II-III loop was biotinylated and accessible to avidin-rhodamine (11/11 cells), as illustrated by the tightly co-localized puncta of YFP and rhodamine fluorescence in Fig. 9. Thus, the proximal portion of the ␣ 1S (II-III) cytoplasmic loop is accessible to 60-kDa avidin molecules both when ␣ 1S is expressed as a single protein or as two-protein fragments divided at the proximal II-III loop.
The results above indicate that avidin has access to several sites of ␣ 1S and ␤ 1a . In contrast, the C-terminal of ␣ 1S appears to be inaccessible. Fig. 10 illustrates dysgenic myotubes expressing GFP-␣ 1S long-BAD and GFP-␣ 1S short-BAD (BAD following residues 1860 and 1667, respectively). Both constructs were functional in EC coupling (see above) and produced distinct green fluorescent puncta. However, there were no puncta of avidin staining that co-localized with the green puncta, a result that was consistently observed. Specifically, there was a clear absence of any co-localized puncta in 29/29 cells expressing GFP-␣ 1S long-BAD and in 30/32 cells expressing GFP-␣ 1S short-BAD, with the other two cells showing some regions of aggregated red and green fluorescence that were difficult to interpret unambiguously as either being puncta or not. In principle, the lack of co-localized red and green puncta for GFP-␣ 1S long-BAD could have been a consequence of proteolytic cleavage of the C-terminal, which has been reported to cause truncation (between residues 1685 and 1699) of the vast majority of ␣ 1S in adult skeletal muscle (17). However, green fluorescence appears in junctional puncta after expression in myotubes of full-length ␣ 1S tagged on the C-terminal with GFP (18), which indicates either that relatively little truncation occurs in myotubes or that the distal segment remains anchored to the DHPR (19). In any case, red puncta coincident with green puncta were also not observed when the BAD was fused at a position upstream of the potential proteolytic site (i.e. GFP-␣ 1S short-BAD in Fig. 10).
A second possibility, a lack of biotinylation, could explain the absence of co-localizing red and green puncta for BAD fused to the ␣ 1S C-terminal. However, near the site of cDNA injection, there was both strong (but non-punctate) green fluorescence and red fluorescence with similar intensity and subcellular distribution. Farther from the injected nucleus, the rhodamineavidin staining had a granular appearance, which as already mentioned did not overlap with the green puncta. Similar granular binding of avidin was also observed in non-injected myotubes. The differential pattern of avidin staining at sites near and far from the injected nucleus suggests that early in the biosynthetic pathway, BAD on the C-terminal of ␣ 1S is biotinylated but becomes inaccessible once ␣ 1S is inserted into fully assembled junctions.
A third possibility, which could explain the lack of avidin binding to the C-terminal of junctionally targeted ␣ 1S constructs, is that RyR1 occludes access to the ␣ 1S C-terminal. This possibility was tested by expression of GFP-␣ 1S -BAD in RyR1-lacking (dyspedic) myotubes. Indeed, as shown in Fig. 11, punctate avidin staining that co-localized with green fluorescent foci was observed both when GFP-␣ 1S long-BAD (3/3 cells) and GFP-␣ 1S short-BAD (13/19 cells) were expressed in dyspedic myotubes. Thus, the presence of the ryanodine receptor appears to prevent avidin from accessing the ␣ 1S C-terminal. DISCUSSION To probe the topology of the proteins at the triad junction, we have used endogenous biotinylation to investigate the ability of avidin, a 60-kDa molecule, to access specific sites of the DHPR within junctions. Effective biotinylation in living myotubes occurred as a consequence of insertion of a modestly sized BAD sequence (70 or 97 residues) at the N-and C-terminals of both ␣ 1S and ␤ 1a as well as within the cytoplasmic II-III loop of ␣ 1S . The DHPR/BAD fusion proteins were correctly targeted as determined by punctate XFP fluorescence near the cell surface and by restoration of excitation-contraction coupling. To test whether avidin could access sites of the DHPR within fully assembled (RyR1-containing) junctions, permeabilized myotubes were exposed to rhodamine-avidin. As judged by the occurrence of red fluorescent foci that co-localized with the XFP foci, avidin had access to all but one of the sites tested for both ␤ 1a and ␣ 1S . The only exception was the C-terminal of ␣ 1S , which was inaccessible in junctions containing RyR1, but accessible in junctions lacking RyR1. These results are summarized in Fig. 12.
In thin section electron micrographs, much of the junctional gap between the SR and plasma membrane is filled by electron dense material contributed by the foot region of RyR1. Despite the presence of RyR1, however, it appears that these junctions can accommodate substantial additional mass (ϳ60 kDa), inasmuch as avidin had access to the N-and C-terminals of ␤ 1a , and to the N-terminal and II-III loop of ␣ 1S . It is useful to compare this accessibility with results of other studies aimed at identifying potential sites of interaction between the DHPR and RyR1. The ␤ 1a subunit is required for trafficking of ␣ 1S to the plasma membrane (20) and also has modulatory effects on kinetics and voltage dependence of the L-type calcium current (reviewed in Ref. 21). Moreover, functional analyses of ␤ 1 cDNA constructs expressed in ␤ 1 -null myotubes have revealed an important role for the ␤ 1a C-terminal in EC coupling (22). Indeed, deletion of the final 29 residues of ␤ 1a largely eliminates skeletal-type EC coupling (23). In addition, preliminary studies have shown that the ␤ 1a subunit binds to RyR1 (24). These results raise the possibility that the distal portion of the ␤ 1a C-terminal interacts with RyR1. However, if this interaction does occur, it is not sufficient to occlude access of avidin to a site nearby since BAD fused to the C-terminal of the ␤ 1a subunit did not grossly affect EC coupling and was accessible for avidin binding.
A number of studies have suggested that the II-III loop of ␣ 1S plays an important role in the orthograde and retrograde signaling between the DHPR and RyR1. For example, application of small peptides corresponding to the proximal portion of the II-III loop (␣ 1S residues 671-690; "peptide A") activates RyR1, as measured by ryanodine binding, single channel activity, and calcium release (25)(26)(27)(28). However, when the DHPR is expressed in dysgenic myotubes, scrambling the peptide A sequence, replacing it with non-related sequence, or deleting it entirely, does not appear to impair function of ␣ 1S as voltage sensor or Ca 2ϩ channel (15,16,29,30). Consistent with these results, insertion of BAD in place of ␣ 1S residues 671-686 (␣ 1S (I-II)-BAD-(III-IV)-YFP) does not interfere with the ability of ␣ 1S to mediate EC coupling. An additional argument that this general region of the II-III loop does not interact with RyR1 is that BAD introduced into this site is accessible to avidin binding. Moreover, the accessibility in three different constructs (YFP-␣ 1S (I-II)-BAD, BAD-␣ 1S (III-IV)-YFP, ␣ 1S (I-II)-BAD-(III-IV)-YFP) suggests that this accessibility is not an artifact of a particular construct. It is also important to consider the site of the BAD placement with respect to downstream residues that are important in EC coupling. In particular, ␣ 1S residues 720 -765 (31,32), or more minimally 734 -748 (33) in ␣ 1C chimeras, are able to restore full orthograde and retrograde DHPR/RyR1 coupling. Moreover, yeast two-hybrid analyses indicate a weak interaction between ␣ 1S loop residues 720 -765 and RyR1 residues 1837-2168 (34). In the constructs BAD-␣ 1S (III-IV)-YFP and ␣ 1S (I-II)-BAD-(III-IV)-YFP, the BAD was attached at ␣ 1S residue 686 that was 49 residues upstream of the minimal sequence identified by Kugler et al. (33). Of course, the three-dimensional structural relationship between BAD and this minimal sequence remains uncertain. However, the accessibility of avidin makes it unlikely that this minimal sequence binds to a deep pocket within RyR1, particularly because biotin is almost completely embedded within the binding pocket of avidin (35).
In regard to the ␣ 1S C-terminal, the distal portion (residues 1543-1647) has been shown to be important for targeting to junctions (29,36), which may indicate a binding interaction with other junctional proteins. However, this putative, targeting interaction seems unlikely to account for the differential ability of avidin to bind to BAD at the ␣ 1S C-terminal in dyspedic but not dysgenic myotubes because ␣ 1S targets to junctions in both cell types. Besides being important for targeting, several studies suggest that the ␣ 1S C-terminal may interact with RyR1. For example, a peptide corresponding to ␣ 1S residues 1487-1506 inhibits RyR1 in vitro (37). Moreover, ␣ 1S residues 1393-1527 bind calmodulin and, in the absence of calmodulin, bind directly to RyR1 residues 3609 -3643 (38). Thus, the inability of avidin to access the C-terminal of ␣ 1S within RyR1-containing junctions may be a reflection of this binding interaction.
An important caveat for the present studies is that most of the experiments were done on fixed and permeabilized myotubes. Thus, it is uncertain which functional states of the DHPR and RyR1 are responsible for the observed avidin binding. Because of these limitations, questions still remain: does avidin binding depend on the functional state, and does avidin binding alter function? However, the results from the BAD experiments involving fixed and permeabilized myotubes are supported by complementary studies of living myotubes (8) expressing DHPR constructs in which a CFP-YFP tandem was inserted at the same sites described in this paper for BAD insertion. In all cases, function of the DHPR as channel and voltage sensor for EC coupling did not appear to be disrupted. Moreover, for all but two of these sites, the fluorescent resonance energy transfer efficiency of the inserted CFP-YFP tandem was unaffected by whether or not RyR1 was present, indicating that it is unlikely that RyR1 is in close proximity to any of these sites. This conclusion is strengthened by the present results showing that avidin has access to the BAD at all of these sites, taking into account that avidin (8 ϫ 8 ϫ 8.5 nm; Protein Data Bank code 1AVD, Ref. 35) is larger than either CFP or YFP (5 ϫ 6 ϫ 7 nm; Protein Data Bank code 1EMA, Ref. 39). Thus, an important conclusion from both the CFP-YFP experiments and the BAD experiments is that functional junctions can accommodate substantial additional mass at a number of sites.
For two sites (N-terminal of ␤ 1a , shortened C-terminal of ␣ 1S ), the fluorescent resonance energy transfer efficiency of the inserted CFP-YFP did depend strongly on the presence of RyR1 as evidenced by a substantial increase in efficiency in dyspedic myotubes compared with dysgenic myotubes (8). Thus, both might represent sites at which the DHPR approaches closely to RyR1. It is interesting, however, that avidin was able to access biotinylated BAD at the ␤ 1a N-terminal, which raises the possibility that this site opens freely into the myoplasm. However, it is possible to reconcile both the CFP-YFP and BAD results if one were to assume that BAD on the ␤ 1a N-terminal lies in a fairly compact space but is oriented such that biotin extends sufficiently out of this space to allow avidin binding. In the case of the C-terminal of ␣ 1S , the CFP-YFP and BAD results are both consistent with localization within a fairly compact space, particularly because avidin lacked access to either the ␣ 1S short or long C-terminal. This postulated occlusion could occur either because the ␣ 1S C-terminal lies within a pocket of RyR1 or because RyR1 causes the DHPR to assume a configuration such that another protein (perhaps ␣ 1S itself) occludes the C-termi-FIG. 10. The C-terminal of ␣ 1S is biotinylated but inaccessible to avidin within fully assembled junctions. Left, dysgenic myotubes expressing GFP-␣ 1S long-BAD. Right, dysgenic myotube expressing GFP-␣ 1S short-BAD. Although puncta of green fluorescence were present for both constructs, there were no co-localized foci of avidin-rhodamine fluorescence. A complete lack of biotinylation did not appear to be responsible for the absence of punctuate avidin-rhodamine staining. In particular, there was a similar pattern of diffuse fluorescence for GFP and avidin-rhodamine in regions of highly expressed protein not yet in junctions (arrows). Scale bars ϭ 5 m.
FIG. 11. The presence of the ryanodine receptor is responsible for the inaccessibility of the ␣ 1S C-terminal to avidin. In contrast to expression in dysgenic myotubes, expression of GFP-␣ 1S short-BAD in a dyspedic (RyR1-null) myotube resulted in co-localized fluorescent puncta of both GFP and avidin-rhodamine. Scale bar ϭ 5 m. nal. It is important to state the obvious point that both the CFP-YFP tandem and BAD can at best be inserted near sites of interaction between junctional proteins because interrupting these sites would abolish function.
One can expect that future studies will provide improved information about the three-dimensional structure of both the DHPR and RyR1 and about the localization within these structures of specific sites (e.g. ␣ 1S (II-III) loop). Obviously this information will be of considerable value for the interpretation of both the CFP-YFP and BAD studies. Conversely, because the structural studies will likely be limited to the individual proteins (either DHPR or RyR), the CFP-YFP and BAD methods will be important for understanding the disposition of the DHPR and RyR with respect to one another within functioning cells. It will also be important to determine whether the access of avidin to biotinylated BAD depends on the functional status of the EC coupling apparatus (resting, activated, and inactivated) and whether the fluorescent resonance energy transfer efficiency of the CFP-YFP tandem displays changes that are correlated with such functional changes. However, even without awaiting these further refinements, the ability to introduce specific biotinylation at both terminal and internal sites of cellular proteins should provide a useful tool for studying the architecture of macromolecular assemblies in diverse cell types.