Structure of the Kidney Slit Diaphragm Adapter Protein CD2-Associated Protein as Determined with Electron Microscopy

Abstract

CD2-associated protein (CD2AP) is a multidomain scaffolding protein that has a critical role in renal function. CD2AP is expressed in glomerular podocytes at the slit diaphragm, a modified adherens junction that comprises the protein filtration barrier of the kidney, and interacts with a number of protein ligands involved in cytoskeletal remodeling, membrane trafficking, cell motility, and cell survival. The structure of CD2AP is unknown. We used electron microscopy and single particle image analysis to determine the three-dimensional structure of recombinant full-length CD2AP and found that the protein is a tetramer in solution. Image reconstruction of negatively stained protein particles generated a structure at 21 Å resolution. The protein assumed a roughly spherical, very loosely packed structure. Analysis of the electron density map revealed that CD2AP consists of a central coiled-coil domain, which forms the tetramer interface, surrounded by four symmetry-related motifs, each containing three globular domains corresponding to the three SH3 domains. The spatial organization exposes the binding sites of all 12 SH3 domains in the tetramer, allowing simultaneous binding to multiple targets. Determination of the structure of CD2AP provides novel insights into the biology of this slit diaphragm protein and lays the groundwork for characterizing the interactions between key molecules of the slit diaphragm that control glomerular filtration.

CD2-associated protein (CD2AP) was originally identified as an interaction partner of the mouse T-cell membrane protein CD2, where it is involved in CD2 clustering and T-cell polarity.¹ Mice deficient in CD2AP (the mouse homolog) die secondary to renal failure,² whereas heterozygous mice are more likely to develop glomerular lesions similar to those lesions described in patients with FSGS.³ In the kidney, CD2AP is primarily expressed in glomerular podocytes, and its absence results in improper formation of the slit diaphragm² and the protein filtration barrier of the kidney.⁴ In humans, homozygous mutations in the CD2AP gene resulted in FSGS.^3,5–7

CD2AP is a 70 kDa multidomain protein containing several identified protein–protein interaction domains (Figure 1A). The protein is a homo-oligomer, an association driven by a coiled-coil (leucine-zipper) domain at the extreme C terminus.⁸ The amino terminus consists of three Src homology 3 (SH3) domains (SH3A, -B, and -C) separated by intervening sequences of undetermined structure. SH3 domains are found in various proteins, where they exert their typical function: molecular recognition and subsequent binding.⁹ Although they form a highly conserved family of domains, the amino acid composition of SH3 domains varies at a few key sites, allowing for a wide range of molecular targets. The three SH3 domains of CD2AP are structurally similar to known SH3 domains^10,11,12 and also recognize a target protein sequence consisting of an extended polyproline 2 helix.¹³ The recognition sequence of the CD2AP SH3 domains is unusual in that each domain has an exact requirement for the sequence PXXXPR, with a preference for PXPXPR sequences,¹⁴ where the final arginine is critical for binding.¹⁰ The SH3 domains are necessary for CD2AP interactions with a number of signaling molecules, regulators of the cytoskeleton, and modulators of apoptosis, including CD2,^1,10 SETA binding protein 1,¹⁵ the tyrosine kinase negative regulator c-Cbl,^10,16–18 ubiquitin ligase Cbl-b,¹⁹ ADP-ribosylation factor GTPase-activating protein ASAP1,²⁰ apoptosis-linked gene 2-interacting protein X/apoptosis-linked gene 2-interacting protein 1,^21,22 and ubiquitin,²³ each of which contains the target recognition sequence within their cystosolic domains. Simultaneous interactions with multiple CD2AP SH3 domains have been postulated as necessary for receptor clustering and subsequent internalization. CD2AP also interacts through its SH3 domains with the transcription factor dendrin²⁴ and the actin-associated protein synaptopodin²⁵ proline-rich possibly natively unfolded proteins, which also contain the target sequence. Loss of CD2AP drives dendrin translocation from the slit diaphragm to the nucleus, where it promotes expression of cytosolic cathepsin L,²⁶ which cleaves dynamin,²⁷ synaptopodin,²⁸ and CD2AP itself (between SH3B and SH3C domains) and leads to podocyte effacement and progressive disease. After the SH3C domain, CD2AP contains a stretch of approximately 100 amino acids that contains three proline-rich stretches probably organized as polyproline 2 helices and thus, serving, in turn, as recognition sequences for external SH3 domains, because none of the proline-rich sequences contain the PXXXPR sequence necessary for internal recognition. However, the related protein CIN85, which is 35% identical to CD2AP, does contain the target sequence in its proline-rich domains. Evidence suggests that the first SH3 domain in CIN85 forms an intramolecular association, which inhibits binding by the other SH3 domains.^29,30

Figure 1.

Purified CD2AP forms a tetramer in solution. (A) Schematic diagram of the CD2AP primary structure. Regions in the primary structure corresponding to the identified sequence motifs discussed in the text are indicated. Shaded areas indicate regions with unidentified structure. The three shorter sequences containing polyproline regions and the CARMIL peptide, which binds the actin capping protein CP as an extended peptide,^31,32 are indicated above the diagram. (B) Electron micrograph of negatively stained CD2AP. Scale bar, 1000 Å. (C) Gel filtration profile for purified CD2AP. Displayed on the same graph are peak locations for the standards thyroglobulin (ThyG; Stokes' radius=85 Å), ferritin (Fer; 61 Å), IgG (52 Å), ovalbumin (Ova; 29 Å), and cytochrome C (CytC; 16.5 Å). The curve shows the fit for the Stoke’s radius to the elution volume for the standards. (D) Gel electrophoresis showing oligomerization of CD2AP. Left panel shows Coomassie-stained native gel run as indicated in Concise Methods. Positions of the native standards are indicated on the left. Right panel shows silver-stained SDS-PAGE with the results from DTSSP crosslinking. The sample in each lane is indicated at the top, and the positions of the standards are again indicated on the left. (E) Symmetry evaluation in selected particles from electron micrographs. Left panel shows a reference-free average calculated from 22 particles selected as possessing C4 symmetry. Right panel shows an autocorrelation plot showing the correlation coefficient calculated between rotated and unrotated particle averages.

CD2AP is also directly involved in control of the actin cytoskeleton. The C-terminal segment of CD2AP contains an approximately 20 residue sequence (termed the CARMIL peptide^31,32) that binds the actin capping protein (CP).^29,33 CP binds to the barbed ends of actin filaments and thus, caps them, inhibiting the addition and loss of actin monomers at that end. CP is ubiquitous in eukaryotes and important for actin assembly and actin-based processes of morphogenesis and differentiation.³⁴ In vitro, CP promotes dendritic branching of actin networks through the Arp2/3 complex by inhibiting actin polymerization at the barbed ends.^35,36 By sequestering CP, CD2AP inhibits branching and promotes development of thin bundles of actin filaments.³⁴ CD2AP can also interact directly with actin filaments and possesses actin binding sites at the extreme C terminus.^37,38 Interestingly, the proline-rich region of CD2AP interacts with the actin binding protein cortactin in a ligand-dependent manner, recruiting it to the endocytic CD2AP–Cbl–epidermal growth factor receptor complex.³⁹ Recently, Zhao et al.⁴⁰ found that this interaction recruits CD2AP to the cell periphery and mediates formation of lamellipodia through a feedback loop generated between cortactin, CD2AP, and CP. In podocytes, CD2AP localizes to lipid rafts at the slit diaphragm with nephrin and podocin, and it contains a binding site at its C terminus for the membrane proteins nephrin² and podocin.⁴¹ In addition, CD2AP links through its SH3 domain Rac1 to cortactin and CP, another actin binding protein.⁴² CD2AP, thus, seems to provide a structural link connecting the podocyte membrane with the actin cytoskeleton.

In this paper, we used single particle image analysis to generate a three-dimensional structure of CD2AP at 21 Å resolution. CD2AP assumes a cuboid tetrameric structure with four of the faces related by rotational symmetry with clearly identifiable domains. CD2AP is organized into a central core surrounded by four symmetry-related motifs that each have three globular domains, corresponding to the three SH3 domains. Cathepsin L cleavage sites are located at the SH3 arms of CD2AP. The spatial organization exposes the binding sites of all 12 SH3 domains to allow simultaneous binding to multiple targets.

Results

CD2AP Is a Compact Tetramer in Solution

CD2AP examined by electron microscopy in negative stain presents compact, almost circular particles (Figure 1B). The uniformly compact particles indicate that the molecule does not possess an extended linear organization with the N and C termini spatially well separated from one another. The particles are large, with a characteristic diameter of approximately 130 Å, which is larger than expected for a monomer (if organized as a compact sphere, a 70 kDa protein would have a diameter of approximately 55 Å). The large particles are confirmed by gel filtration chromatography (Figure 1C), where the CD2AP elution peak approximately coincides with the largest standard (thyroglobulin; Stokes’ radius=85 Å). The oligomeric status of the protein was determined from chemical crosslinking and native gel electrophoresis (Figure 1D). Both methods provided consistent results, indicating that the purified protein forms a complex of approximately 300 kDa, compatible with a tetrameric organization. For a tetramer, there are three possible symmetry organizations: (1) point-group symmetry C4 with 4-fold rotational symmetry, with each subunit related by a 90° rotation about an axis, (2) point-group symmetry D2, where a pair of subunits is related to the other pair of subunits by a 180° rotation, and (3) a dimer of dimers, where a pair of subunits forms a dimer with another pair with a different interacting surface such that the four subunits are not related by a common symmetry axis. The C-terminal putative coiled-coil domain is believed to be responsible for CD2AP oligomerization,^8,37 which would suggest a single tetramerization interface with either C4 (parallel helices) or D2 (antiparallel helices) symmetry. To determine the symmetry of CD2AP in solution, we searched the entire dataset of selected particles for particles with internal rotational symmetry. Figure 1E shows the average from 22 raw particles selected for C4 symmetry. The average was generated with reference-free alignment and averaging; no symmetry was used during averaging, and no symmetric model was used for alignment. As can be seen, the average displays a clear four-lobed density, which is quantified in the adjacent rotational autocorrelation plot (Figure 1E). Autocorrelation peaks are spaced at 90° intervals, which was expected for C4 symmetry.

Electron Microscopy Structure of CD2AP

The dataset of 11,006 particles was processed with an iterative, projection-matching algorithm to generate a three-dimensional structure. Four-fold rotational symmetry was imposed on the map during refinement (Figure 2A). A Fourier shell correlation computed between maps generated from a split dataset indicates that the map has a resolution of approximately 21 Å using the 50% correlation criterion (Figure 2, B and C). The resulting map reveals a roughly cubic molecule with four of the faces related by the rotational symmetry (Figure 3). The map extends approximately 130 Å along the symmetry axis and has a widest dimension of approximately 140 Å on an orthogonal axis. Negative stain has penetrated throughout the structure to reveal clearly identifiable domains. The overall organization consists of a central core, comprising two segments (Figure 3B) along the symmetry axis. It is surrounded by four symmetry-related motifs (arm domains), each containing three globular subdomains (Figure 3A). Four symmetry-related segments, comprising a potential unknown domain (unk in Figure 3B), are at the base of the structure.

Figure 2.

Results from the final round of single particle image analysis of CD2AP and evaluation of the final map. (A) Histogram displaying the classification of the individual particles from the final round of refinement. Each spot represents a particular orientation of the two Euler angles Φ and Θ and is assigned a grayscale value based on the number of particles assigned to that orientation. White represents 1 particle, and black represents the maximum of 147 particles. (B) Fourier shell correlation plot for the final round of refinement. The map is calculated from a randomly split dataset using the Euler angle assignments for the final round. The resolution is calculated as the highest resolution displaying 50% correlation. (C) Gallery showing a comparison of class averages calculated at the final round of refinement and reference-free class averages. Reference-free averages were calculated as described in the text and classified against the complete set of class averages calculated during the final round of refinement. Each reference-free average is shown next to the refined average that it best matches. The assigned Euler angles for the refined average are shown in the right panel.

Figure 3.

The three-dimensional density map (surface-shaded representation) of CD2AP determined by electron microscopy and single particle image analysis. The isosurface has been chosen to enclose 100% of the expected protein density. (A) View of the map with the 4-fold axis oriented in the y direction. Annotations on the map indicate putative positions of the SH3 domains within the arm domain. The asterisk indicates the approximate position of an identified cathepsin L cleavage site between SH3B and SH3C. (B) The same view as in A but rotated on the y axis 45°. The annotation C-C on the map indicates the putative position of the C-terminal coiled-coil domain. (C) The same view as in A but indicating a putative monomer in yellow. (D) The same view as in B but with the putative monomer colored as in C.

Segmentation of the Three-Dimensional Map and Domain Assignments

Identification of the individual subunits within the map was done with a watershed segmentation algorithm⁴³ to separate the map into 18 segments using the internal variations within the map density to determine the break positions (Figures 3 and 4). The two segments of the central core (red and blue in Figure 4, A and B) are presumably composed of the interacting regions forming the homotetramer and thus, shared between all four molecules. The 43-amino acid sequence at the C terminus³⁷ is strongly predicted to form a canonical left-handed coiled coil. A tetrameric coiled-coil structure⁴⁴ (red in Figure 4) provides a good fit for segment 1 of the central core (Figure 4, C and D). The second nonsymmtery-related segment 2 of the central core (Figures 3A and 4, A and B, blue) is not readily assignable to an identifiable structural element.

Figure 4.

Segmentation of the three-dimensional density map of CD2AP allows domain assignments. (A) Individual map segments generated by the program COAN.⁴³ Nonsymmetry-related segments are shown in different colors, and the remaining symmetry-related segments are displayed in white. Two of the segments, shown in red and blue, are along the 4-fold axis and contain densities from each of the subunits. (B) The same view as in A but rotated by 45° around the y axis. (C) Tentative assignments of CD2AP domains showing the structures of SH3A¹⁰ (green), SH3B (yellow), and SH3C¹¹ (cyan) within the density from the assigned segment. Also shown is a tetrameric mutant of the coiled-coil domain of GCN4⁴⁴ (red). The map segments are also shown with transparent isosurfaces. The orientation is the same as in A. The Protein Data Bank files were fit into segment densities with the program Chimera. (D) The same view as in C but rotated 45° to match the view in B.

The 12 segments (3 segments in each monomer) in the arm domains surrounding the central core as symmetry-related volumes are presumably formed from separate domains of individual CD2AP monomers. They (green, yellow, and cyan in Figure 4, A and B) likely represent the N-terminal SH3 domains and intervening sequences (Figure 4, C and D). The three SH3 domains have been tentatively assigned to their respective domains (A, B, or C) based on the connectivity of the segments (Figures 3A and 4A). The map segment assigned to SH3C in the monomer provides a good fit for this isolated domain, suggesting that it does not contain any intervening sequences, such as sequences from the proline-rich regions immediately after this domain. This spatial organization exposes the binding sites of all 12 SH3 domains in the tetramer to allow simultaneous binding to multiple targets.

The final map contains four additional segments related by symmetry (unk segments in Figure 3B and a monomer colored purple in Figure 4) that are not identified. However, by a process of elimination, it likely comprises the approximately 170-amino acid region (inclusive of the CARMIL peptide) between the proline-rich and the coiled-coil domains. This unassigned region itself is too small to enclose the whole approximately 20 kDa sequence and probably composed of at least one unidentified globular domain.

Discussion

The adaptor protein CD2AP acts as an anchoring protein that is involved in various signaling pathways,^{17,26,45–48} actin assembly,^{20,29,37,38,42,49} slit diaphragm complex in podocytes,^2,41,50,51 and endocytosis of the epidermal growth factor receptor.^33,39,52,53 The three N-terminal SH3 domains are known to play an important role in the internalization of a number of receptor proteins, such as CD2, SETA binding protein 1, c-Cbl, Cbl-b, ASAP1, apoptosis-linked gene 2–interacting protein X/apoptosis-linked gene 2-interacting protein 1, and ubiquitin, which bind CD2AP through a specific recognition sequence in their cytosolic domains.

Segmentation of the CD2AP structure shows that a monomer comprises a minimum of six domains, three of which correspond to SH3 domains (one domain at the C terminus corresponds to a coiled-coil domain and two map segments, lying between the third SH3 and coiled-coil domain in the primary structure, likely represent previously unidentified globular domains). Our structure analysis assigns the N-terminal SH3 domains to the arm domain (Figures 3 and 4). The C-terminal region to the arm domain is assigned to the remaining domains, including the central core (Figure 3B) and unassigned globular domains (unk in Figure 3B). The central core contains two domains identified by segmentation (red and blue in Figure 4) mediating the tetramer interaction. The remaining unassigned globular domains (purple in Figure 4) are composed of C-terminal regions between SH3C and the coiled-coil domain, but they are too small to contain the whole region. These sequences include the three polyproline peptides as well as the CARMIL peptide. Sequence analysis using the GlobPlot program⁵⁴ suggests that both of these sequences are disordered, which would facilitate binding to SH3 domains for the polyproline peptides and CP by the CARMIL peptide. We do not believe that these regions are themselves present in the map, because extended peptides would not be visualized in negative stain and the density for highly disordered domains averaged out during the electron microscopy refinement. It is possible that the unassigned globular domains might flank the polyproline regions and the CARMIL peptide. These C-terminal globular domains (purple in Figure 4) are on a separate face of the tetramer, allowing the approach of large proteins, even if the N-terminal SH3 domains have engaged ligands.

CD2AP forms a tetramer in solution, and the three-dimensional structure suggests that it is mediated by one or both components of the central core comprising the putative coiled-coil domain (red in Figure 4) and an unassigned core domain (blue in Figure 4). A direct connection between these domains is not observed in the map at an isosurface level enclosing the expected protein density. This finding suggests that the coiled coil is attached to the remainder of the structure by a more flexible strand, which would not be visualized on the map. Experimental evidence has shown that the coiled-coil domain mediates oligomerization.⁸ An alternative function for this domain may be binding to the podocyte slit diaphragm component nephrin.⁵⁵ The cytosolic tail of nephrin binds the region of CD2AP C terminus to the proline-rich regions and requires an intact coiled-coil domain.⁵⁶ The need for an intact coiled-coil domain suggests that the cytosolic tail of nephrin binds as a helix directly to the coiled-coil bundle. In the CD2AP structure, an individual α-helix would be able to access the coiled coil by entering the structure between the arm domains and the coiled coil. This type of interaction would still allow binding by the SH3 domains and other C-terminal motifs on the other faces of the structure. An alternative mode of interaction would have nephrin binding at the interface defined by the tetramer across two or more CD2AP molecules. In this case, the location of the unassigned core domain adjacent to the coiled-coil domain suggests that it may contribute to the nephrin binding site.

The structure of the CD2AP shows that 12 SH3 domains surrounding the central core are within a fairly small compass. The tetramer could, thus, potentially bind two or three receptor molecules on each face, facilitating clustering. This compact organization might facilitate the autoinhibition seen in the related molecule CIN85. In CIN85, the SH3A domain binds to the second internal polyproline region, which induces a conformational change that inhibits binding by the other SH3 domains.¹⁴ In CD2AP (which does not contain the internal recognition sequence), the first SH3 domain (SH3A) is positioned approximately 25 Å from the junction between the third SH3 domain (SH3C) and the proline-rich regions of a neighboring molecule. Movement of SH3A down to engage these regions would require a small rotation of the arm containing the three SH3 domains and might be the origin of the conformational change postulated to inhibit CIN85.

The unassigned domain (purple in Figure 4) has a similar size to the three SH3 domains, suggesting an alternate assignment of this segment to an N-terminal SH3 domain. The resolution of the map does not allow us to exclude this possibility; however, this seems unlikely based on the connectivity, because this domain attaches directly to the C-terminal central core. In contrast, our assignment places the three SH3s adjacent to one another and SH3A and SH3B in similar locations. Data from autoinhibition of CIN85 indicate that, in the absence of SH3A, intramolecular binding may be provided by SH3B.³⁰ It would be more consistent with our current assignment, because a similar motion would bring either domain adjacent to the proline-rich region.

CD2AP is a substrate for cytosolic cathepsin L, a protease that is induced in early podocyte damage.²⁶ Cathepsin L cleaves CD2AP at two sites: between SH3B and SH3C and between the first and second proline-rich regions that follow SH3C.²⁶ The latter cleavage generates a C-terminal fragment of CD2AP (P32) that remains bound to the podocyte cytoskeleton but releases its binding partner, dendrin,⁵⁷ that can now travel to the podocyte nucleus to promote apoptosis.²⁶

The organization of the CD2AP tetramer leaves both cleavage sites exposed, flanking SH3C at the outer radius of the tetramer (Figure 3). However, many of the other cleavage sites, which are revealed by computer modeling,^26,58 are inaccessible because of tetramerization of CD2AP.

The determination of three-dimensional conformation of CD2AP gives novel insights into the structural organization of this modular protein and provides a basis for probing the role of the unassigned domains and tetramerization in formation and stability of the slit diaphragm.

Concise Methods

Cell Culture and Transient Transfection

Human embryonic kidney 293 cells were maintained at 37°C with 5% CO₂ in DMEM (Invitrogen, Carlsbad, CA) supplemented with 10% heat-inactivated FBS (Invitrogen) in tissue culture-treated T75 flasks (Corning, Lowell, MA). After reaching 80%–90% confluency, the adherent monolayer cells were transfected with FLAG-CD2AP construct using Lipofectamine 2000 reagent (Invitrogen).

Purification of CD2AP Protein

Human embryonic kidney 293 cells expressing FLAG-CD2AP were harvested with ice cold PBS (Boston Bioproducts, Ashland, MA) containing 50 mM EDTA (Sigma-Aldrich, St. Louis, MO) and collected after pelleting at 1800 rpm for 5 minutes at 4°C. Washed cells were lysed for 30 minutes at 0°C in lysis RIPA buffer of 50 mM Tris⋅HCl, 200 mM NaCl, 1 mM EDTA, 1 mM EGTA, and 0.25% sodium deoxycholate (pH 7.5; Boston Bioproducts) containing protease inhibitors (Roche Diagnostics, Indianapolis, IN). The cell debris was removed by centrifugation at 13,000 rpm for 15 minutes, and cell lysate was incubated with anti-FLAG M2 affinity agarose gel beads (Sigma-Aldrich) prewashed in RIPA buffer on an overhead rotator in the cold room overnight. The next day, beads were washed with the same lysis buffer five times in the cold room. After the last wash, beads were washed one time with ice cold PBS, and bound CD2AP was eluted with 0.5 μg/μl FLAG-peptide (Sigma-Aldrich) after a 1-hour incubation on ice followed by centrifugation at 3000 rpm for 1 minute.

Gel Electrophoresis and Immunoblotting

The eluate was resolved on Invitrogen’s NuPAGE Bis⋅Tris gel (4%–12%) using morpholineethanesulfonic acid running buffer (Invitrogen) according to the manufacturer's instructions. The gel was stained with GelCode Blue Stain (Thermo Scientific, Rockford, IL), and the concentration of the eluate was measured by densitometric analysis of the protein bands using standard BSA (Bio-Rad Laboratories) samples (0.125–2.0 μg/well) and ImageJ software (National Institutes of Health, Bethesda, MD). The eluate was visualized using a silver staining kit (Bio-Rad Laboratories) to confirm purity. Gel-fractionated protein was also transferred to polyvinylidene difluoride membrane (Millipore, Billerica, MA) and probed with the rabbit anti-CD2AP¹ and anti-FLAG M2 mouse monoclonal (Sigma-Aldrich) primary antibodies for immunochemical characterization.

Chemical Crosslinking, Native PAGE, and Gel Filtration Chromatography

Chemical crosslinking was performed according to standard protocols with Pierce’s DTSSP crosslinking reagent (Thermo Scientific). Native PAGE was performed with the NativePAGE system (Invitrogen) according to the manufacturer’s instructions. Gel filtration chromatography was performed on a Superdex 200 10/30 column (GE Healthcare Life Sciences, Pittsburgh, PA) in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM sodium phosphate [pH 7.3]) with a flow rate of 0.4 ml/min. Gel filtration standards thyroglobulin, ferritin, IgG, ovalbumin, and cytochrome c are all from GE Healthcare Life Sciences.

Electron Microscopy

Aliquots (approximately 5 μl of 50 μg/ml protein) were allowed to adhere for 2–5 minutes to carbon-coated copper grids and then stained with 2% uranyl acetate (Ted Pella, Redding, CA). Images were recorded under minimum electron dose conditions using a CM10 electron microscope (Philips Electron Optics, Hillsboro, OR). Images were recorded on Kodak 4489 film at a nominal magnification of ×52,000 using 100 kV electrons. Micrographs were digitized with a CoolScan 9000 scanner (Nikon Instruments, Melville, NY) at 8 bits/pixel and 6.35 μm/pixel, and they were subsequently averaged to 12.7 μm/pixel. The OD for each negative was adjusted to give a mean value of approximately 127 over the total range of 0–255.

Image Reconstruction

Image processing was performed with the EMAN suite.⁵⁹ In total, 11,006 particles were selected from 25 micrographs. The contrast transfer function for each micrograph was manually determined with the EMAN program ctfit, and phase corrections were applied to the selected particles. Initial models were generated using the EMAN routine startcsym, which conducts a symmetry search of the particles for 4-fold and mirror symmetry (representing top and side views, respectively). These orthogonal projections are subsequently aligned with a common lines algorithm and back-projected to generate a three-dimensional structure. Starting models were subjected to refinement with the EMAN projection-matching routine refine. The final model was generated with C4 symmetry imposed and an angular increment of 5° for the projections. Progress of the refinement was evaluated by Fourier shell correlation between succeeding models and ceased when subsequent models generated by the refinement failed to display improvement in resolution. Resolution of the final model was determined by a Fourier shell correlation calculated between models generated by randomly splitting the particles in each projection class and recalculating two models by averaging the split classes and back projection of the resulting averages. Reference-free class averages to evaluate final model projections were generated from the raw dataset using the EMAN script refine2d.py. For segmentation analysis, the map was subjected to a low-pass Fourier filter at 21 Å⁻¹ before segmentation with the program COAN.⁴³

Visualization

Chimera molecular visualization system⁶⁰ was used for visual analysis of the three-dimensional structures and to produce all of the surface representations of the density maps. The isosurface for the final model was determined from the molecular weight of the tetramer (300,000), which encloses a volume of 360,000 ų using a protein partial-specific volume of 0.74 cm³/g. Atomic coordinates for the SH3 domains (Protein Data Bank IDs 2J6K,¹⁰ 3U23, and 2JTE¹¹) and the tetrameric coiled-coil domain (Protein Data Bank ID 1GCL⁴⁴) were visually fitted within the assigned map segments and refined with an internal Chimera routine.

Disclosures

B.D.A. and J.R. are inventors on pending patents regarding the structure of CD2AP. They stand to gain royalties from their future commercialization.