Visualization of inositol 1,4,5-trisphosphate receptors on the nuclear envelope outer membrane by deep etching electron microscopy

Summary

The receptors for the second messenger InsP₃ comprise a family of closely related ion channels that release Ca²⁺ from intracellular stores, most prominently the endoplasmic reticulum and its extension into the nuclear envelope. The precise sub-cellular localization of InsP₃Rs and the spatial relationships among them are important for the initiation, spatial and temporal properties and propagation of local and global Ca²⁺ signals, but the spatial organization of InsP₃Rs in Ca²⁺ stores is poorly characterized. Using nuclei isolated from insect Sf9 cells and freeze-dry rotary shadowing, we have addressed this by directly visualizing the cytoplasmic domain of InsP₃R located on the cytoplasmic side of the nuclear envelope. Identification of ~15 nm structures as the cytoplasmic domain of InsP₃R was indirectly supported by a marked increase in their frequency after transient transfections with cDNAs for rat types 1 and 3 InsP₃R, and directly confirmed by gold labeling either with heparin or a specific anti-InsP₃R antibody. Over-expression of InsP₃R did not result in the formation of arrays or clusters with channels touching each other. Gold-labeling suggests that the channel amino terminus resides near the center of the cytoplasmic tetrameric quaternary structure. The combination of nuclear isolation with freeze-drying and rotary shadow techniques allows direct visualization of InsP₃Rs in native nuclear envelopes and can be used to determine their spatial distribution and density.

1. Introduction

Inositol 1,4,5 trisphosphate receptors (InsP₃Rs) are ligand-gated channels through which calcium (Ca²⁺) is released from intracellular stores in many eukaryotic cells [1]. The channels mediate a wide variety of biological processes, such as proliferation, programmed cell death and gene expression [2]. They are assembled from four large subunits, each of ~2700 residues with a single InsP₃ binding site. In mammals three different genes encode very similar subunits (types 1–3) that assemble as large homo- and hetero-tetrameric structures [3, 4]. The three types of InsP₃Rs share approximately two thirds of their amino acid residues; however they have different affinities for InsP₃ [5] and are differentially regulated by Ca²⁺ and ATP [6, 7]. The structure and domain organization of the full-length type 1 InsP₃R have been determined by single particle analysis of purified receptors, mainly derived from brain cerebellum microsomes. The channel exhibits a four-fold symmetry and, in the presumably closed state in the absence of Ca²⁺, its large hydrophilic domain presents a square profile when viewed from the cytoplasm [3, 4, 8].

The precise sub-cellular localization and distribution of InsP₃Rs are believed to be critical for the spatial and temporal properties of localized Ca²⁺ release and for the initiation and propagation of Ca²⁺ signals [3, 9, 10]. Accordingly, the physiological consequences of Ca²⁺ released are strongly influenced by the localization of InsP₃Rs on Ca²⁺ stores, including the Golgi complex, sarco/endoplasmic reticulum (ER) and nuclear envelope [8, 10]. Optical imaging has revealed the presence of discrete Ca²⁺ release sites within cells [11], but the relationship between such sites and the localization of InsP₃Rs is not well understood. For example, whereas spatially-segregated clusters of 10–100 InsP₃Rs are believed to be responsible for the generation of so-called Ca²⁺ puffs in different cell types [9, 12], immunolocalization studies or imaging of GFP-tagged InsP₃Rs suggests a more uniform distribution of the channels throughout the entire ER in many cell types [10, 13–17]. Furthermore, the number of channels and their spatial proximity determines the properties of localized Ca²⁺ release events such as puffs, but these features have been inferred and not directly examined.

Direct localization of InsP₃Rs has been described in neuronal Purkinje cells of the cerebellum, where InsP₃Rs, mainly type 1, are highly concentrated [18, 19]. In these cells, the ER forms stacks of cisternae decorated by semi-crystalline arrangement of InsP₃Rs [19–23]. For other cells and cellular locations, however, the arrangement and density of InsP₃Rs in membrane sites have been difficult to visualize and remain poorly characterized. Freeze-drying and rotary shadowing have been successfully used to determine the structure, position and arrangement of RyR Ca²⁺ release channels in isolated triads of skeletal muscle [22–24]. Here, we have used this technique, together with gold particle conjugation either to heparin or to specific InsP₃R antibodies, to reveal the spatial disposition and density of InsP₃Rs within the outer membrane of the nuclear envelope under native and over-expression conditions. The nuclear envelope is an InsP₃-dependent Ca²⁺ store [25], whose outer membrane is continuous with the ER, and contains functional InsP₃R channels [26–31]. The cytoplasmic domain of the nuclear envelope can be visualized in its entirety by freeze drying and rotary shadowing of isolated nuclei, facilitating observation of the spatial relationships among InsP₃R channels.

2. Materials and methods

2.1. Cell culture and transfections

Spodoptera frugiperda (Sf9) cells (Invitrogen) were incubated at 27°C and maintained in serum-free SF900II medium (GIFCO). A baculovirus expression system was used to transiently express either rat types 1 or type 3 InsP₃R. The chicken B lymphocyte-derived DT40 cell line with all three InsP₃R isoforms genetically deleted (DT40-KO) [32] was maintained in suspension culture at 37°C (95/5% air/CO₂) in RPMI 1640 medium (GIBCO/BLR) supplemented with 10% (v/v) FBS, 1% chicken serum, 2 mM glutamine, 100 U ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin.

2.2. Western blotting

Cellular extracts were lysed in 100 µl ice-cold Cytobuster protein extraction reagent (Novagen) supplemented with a broad-spectrum protease inhibitor cocktail (Roche). Protein concentrations of the supernatants were determined with BSA as standard. Protein extracts were suspended in Laemmli buffer, separated in 5% SDS-polyacrylamide gels and transferred to PDVF membranes (Millipore). Blocking was at room temperature for 1 hr in 5% fat-free milk, and the membranes were incubated overnight with antibodies against either types 1 (recognizes the carboxyl-terminus; provided by Dr. S. Joseph, Thomas Jefferson Univ.) or 3 (recognizes the amino terminus; BD Transduction Laboratories) InsP₃R. Immunoreactive proteins were detected using ECL reagents (Pierce Biotechnology) according to the manufacturer's instructions.

2.3. Sf9 and DT40 cell nuclear isolation

Nuclei were extracted from cells by a combination of hypotonic shock and mechanical disruption in a Dounce homogenizer, as previously described [33]. Briefly, both Sf9 and DT40 cells were pelleted, washed three times in phosphate-buffered saline and then re-suspended in hypotonic buffer (10 mM Tris-HCl, pH 7.8, 10 mM β-mercaptoethanol, 0.2 mM PMSF and protease inhibitors). After 5 min on ice, the swollen cells were broken in a Dounce homogenizer. The nuclei were separated by slow centrifugation (400 g for 7 min at 4°C), re-suspended in wash solution (10 mM Tris-HCl, pH 7.2, 110 mM KCl, 2.2 mM MgCl₂ and protease inhibitors) and re-centrifuged (800 g for 5 min at 4°C). Finally, the isolated nuclei were re-suspended in 10 mM Hepes-Tris, pH 7.6, 110 mM KCl, 1 mM MgCl₂ plus protease inhibitors.

2.4. Isolation of individual Xenopus oocyte nuclei

Ovary extraction from Xenopus laevis and oocyte nuclei isolation were performed as was described [34]. Briefly, nuclei were manually teased out of freshly isolated oocytes with fine forceps. The nuclei were cleaned of cytoplasmic material by gently sucking them up and down in a pipette in basic oocyte nucleus solution (BONS, containing 140 mM KCl, 10 mm HEPES, 3 mM MgCl₂, 1 mM BAPTA, 0.543 mM CaCl₂, pH 7.3). The isolated nuclei were then transferred to a glass coverslip previously treated with 0.1% poly-L-lysine for freeze-drying and replication.

2.5. Immunostaining, and confocal microscopy

Isolated nuclei were fixed in methanol at −20°C for 12 min, blocked in 1% BSA and incubated with the primary antibodies against types 1 or 3 InsP₃R overnight at 4°C, with gentle rotation. The nuclei were washed with PBS/1%BSA, incubated with Alexa-488 secondary antibody (Molecular Probes) for 1 hr at room temperature, also rotating, mounted in Vectashield mounting medium (Vector Laboratories) and examined in a Zeiss Axiovert 510 LSM Pascal confocal microscope, using a high numerical aperture water immersion 63× objective.

2.6. Freeze-drying and replication

A suspension of freshly isolated nuclei was placed on fragments of glass coverslips previously treated with 0.1% poly-L-lysine, and allowed to attach for 5 min. The coverslips were then rinsed with 100 mM ammonium acetate, treated with 2% (w/v) uranyl acetate for 30–60 sec, and rinsed extensively with 40% (v/v) methanol. The solution was dried to a very thin film using the sandwich technique, and frozen in liquid nitrogen [35]. After mounting on the cold stage of a Balzer’s freeze-fracture apparatus, the nuclei were freeze-dried at 10⁻⁶ mbar pressure at −90°C for at least 30 min, and then re-cooled to −110°C, rotary shadowed with platinum at a 25° angle and replicated with carbon. Finally, the glass coverslips were dissolved with hydrofluoric acid and the replicas cleaned with bleach (6%) for 10 min, washed with water and mounted on an EM grid. Replicas were viewed and photographed in an electron microscope (Philips EM 410; Philips Technology, Cheshire, CT).

2.7. Heparin-gold and immuno-gold labeling

Shadowed replicas of gold-labeled nuclei were obtained as described [36], with modifications, using nuclei isolated from Sf9 cells transfected with rat type 3 InsP₃R .

2.7.1. Heparin gold

Nuclei were incubated with 100 µg/ml heparin-biotin (Sigma) for 3 hr at 4°C while rotating, washed several times and then incubated with Alexa Fluor 488-streptavidin conjugated to 5 nm gold particles (Molecular Probes) for 1 hr. After the labeling, the isolated nuclei were fixed in 0.05% glutaraldehyde for 20 min and freeze-dried and replicated as above, except bleach was diluted to 0.5% and applied for 1 min, followed by three washes in water for 1 hr each. To avoid nonspecific interactions, all samples were treated for 1 hr with 0.05% avidin and 0.005% biotin before incubation with heparin-biotin. Streptavidin-gold incubation in the absence of heparin-biotin, as well as heparin-biotin incubation with DT40-KO cell nuclei, were used as negative controls.

2.7.2. Immunogold labeling

Nuclei were fixed with 4% formaldehyde and 1% glutaraldehyde for 20 min, blocked in 1% BSA for 1 hr, incubated overnight with anti-InsP₃R-3 antibody, washed with PBS/1%BSA and incubated overnight with a gold-conjugated secondary antibody (Molecular Probes). After labeling, the nuclei were freeze-dried and replicated as described above. Incubation with the secondary antibody in the absence of primary antibody, and immunolabeling of DT40-KO cells were used as negative controls.

2.8 Image processing and analysis

All image-processing routines describe below were developed in one of the author’s laboratories (www.scian.cl) on the basis of IDL 7.0 (Interactive Data Language, ITT, CO, USA), including interactive tools for the segmentation of InsP₃R-like structures in the freeze drying images into regions of interest (ROIs), calculation of domain area, and visualization (see Figure 6A). A region of interest in this case is a delimited area that differs in gray scale from the background. Segmentation of the images was achieved by applying gradient filters and selecting threshold values of light intensity in the gradient histograms, resulting in a homogeneous definition of the borders of the InsP₃R-like structures. Remaining holes inside the regions and small segmented signals outside the structures were corrected either by morphological filters or manually. The quality of the segmentation in terms of recognizing the particles that we define as InsP₃R-like structures was controlled interactively by overlaying the original images with the segmented regions (see Figure 6A top and bottom). Although the gray scale threshold values varied between individual images, other segmentation criteria were kept constant for a total of N = 445 InsP₃R-like structures from three different experiments (see Figure 6B). To determine if the surface area of the segmented regions corresponding to the InsP₃R-like structure were from a single population, a Lilliefors test for normality was performed with MATLAB. The Lilliefors test estimates the population mean and variance of the data, computes the empirical cumulative distribution function (ECDF), and the cumulative distribution function (CDF) of the normal distribution using the estimated mean and variance. Finally, it calculates the maximum discrepancy between the ECDF and the CDF, and determines if the discrepancy is statistically significant leading to a rejection of the null hypothesis that the data is normally distributed.

Figure 6. Segmentation and area distribution of InsP₃R-like structures as revealed by image analysis techniques.

(A) Electron micrographs of shadowed InsP₃R-like structures before (top) and after (bottom) segmentation into Regions of Interest (ROIs) outlined by white borders. The areas of 445 ROIs were determined and analyzed for normal distribution by the Lilliefors test. Note that the two particles indicated by a red arrow have a smaller footprint because they partially overlap. Such images account for the smaller size footprints at the left of the distribution. (B) Frequency plot of the area distribution (columns) in combination with a normal distribution (line).

3. Results

3.1. Nuclear pore complexes in rotary shadowed images of isolated nuclei allow identification of nuclear envelope side

An initial survey of thin sections (not shown) indicated that isolated Sf9 cell nuclei had a surface that was devoid of cytoskeletal and/or cytoplasmic remains, thus enabling a clear view of the cytoplasmic surface of the nucleus in shadowed images.

The structures of the nuclear pore complex (NPC) on the nucleoplasm and cytoplasm sides of the nuclear envelope are quite different [37]. We used this well-established difference to identify nuclear surfaces in isolated nuclei. Rotary shadowed replicas of intact nuclei isolated from Sf9 cells showed NPCs characterized by a ring surrounding a central depression (Fig. 1A) indicating that the cytoplasmic side was shadowed. By comparison, fragments of manually isolated nuclei from Xenopus leavis oocytes, showed both cytoplasmic (Fig. 1B) and nucleoplasmic (Fig. 1C) sides of the envelope. The two NPC structures were easily differentiated: one side showed the ring structure, while the other showed a highly organized fibrous network forming the NPC basket [37]. Under the experimental conditions used in this work we only observed NPCs with the structural signature of the cytoplasmic side of the envelope, indicating that the nuclei were intact.

Figure 1. Identification of nuclear envelope surfaces in isolated nuclei by freeze-drying and rotary shadowing.

A. Cytoplasmic surface of a nucleus isolated from wild-type Sf9 cells. The nuclear pore complexes (NPC), arrows) are identifiable as annuli with a diameter of 100–110 nm, decorated by a ring of prominent particles. The structure is typical of the cytoplasmic side of the envelope [36] and intact nuclei always showed only this side The black square encloses a particle identified as a presumptive InsP₃R based on evidence shown later. Scale bar = 100 nm. B and C. Details of freeze-dried rotary-shadowed NPCs in the cytoplasmic (B) and interior face (C) of nuclear envelope fragments from Xenopus oocytes Note that in B the NPC is dominated by a particulate ring, while in C the dominant structure is a fibrillar basket.

3.2. Small structures in the shadowed nuclear envelope might represent InsP₃Rs

The membrane of the outer nuclear envelope in nuclei isolated from Sf9 cell is relatively featureless between individual NPCs, except for occasional small protrusions. Most of these were not identifiable, but a few were consistent with the appearance of the cytoplasmic domains of InsP₃R, as shown in situ [20] and in isolated molecules [38–40]. Although somewhat variable in appearance, the small presumptive profiles of shadowed InsP₃R fit into a square with an approximate size of 15×15 nm, some time had a central depression, and were clearly raised above the envelope surface (Fig. 1, black rectangle).

In order to establish the nature of these presumptive InsP₃R profiles, we imaged the protein after over-expression in Sf9 cells and further labeling with specific ligands.

3.3. Small presumptive InsP₃R- like structures are very frequent in nuclei of Sf9 cells after over-expression of rat types 1 and 3 channels

To provide a more definitive identification of the InsP₃Rs, we over-expressed either the rat type 1 or type 3 InsP₃Rs at very high levels in Sf9 cells using recombinant baculoviral infection, as evidenced by Western blot analysis of infected cell lysates (Fig. 2A). Because a pure population of nuclei could not be isolated for biochemical analysis, we verified that over-expressed recombinant InsP₃R was present in the nuclear envelope by confocal microscopy of isolated nuclei immunolabeled with specific anti-rat types 1 and 3 InsP₃R antibodies. In stacked serial optical sections of 0.5 µm thickness, both InsP₃R isoforms were highly expressed in the nuclear envelope and had similar discontinuous distributions characterized by intense hot spots (Fig. 2B). Single optical sections across the middle of the nuclei (Fig 2B right) show that the spots are located at the nuclear periphery. The nuclei of native cells were totally negative with these antibodies, since the antibodies used to detect the rat channel isoforms do not recognize the endogenous Sf9 InsP₃R (Fig. 2C). For that reason, the relative expression levels of the recombinant and endogenous channels in the transfected cells could not be directly quantified. However, baculovirus-mediated expression of recombinant InsP₃R enhances the probability of recording InsP₃R channels in electrophysiological experiments by over two orders of magnitude [43, 44] suggesting that the expression levels of the recombinant channels greatly exceed that of the endogenous InsP₃R.

Figure 2. Heterologous expression of rat types 1 and 3 InsP₃R in two separate sets of Sf9 cells.

A. Western blots of whole cell homogenates probed with specific anti-type 1 or -type 3 antibodies, revealing intense bands at ~240 kDa corresponding to the InsP₃R specifically expressed in the infected cells. B. Images of isolated nuclei from either type 1 (IP3R1) or type 3 (IP3R3) InsP₃R expressing cells, labeled with their respective antibodies. The first three figures in each row were produced by summing stacked confocal images of single optical sections ~ 0.5 µm thick. Both InsP₃Rs (green) present a patchy distribution over the nucleus that was counterstained with TOTO-3 (red). Individual optical sections (at right) show that antibody labeling is at the periphery of the nucleus. C. Merged images, as in (B), but showing nuclei from untransfected cells. Scale bar, 10 µM.

Images from freeze-dried and rotary-shadowed nuclei from the baculovirus-transduced cells revealed small structures in the nuclear envelope similar in appearance and approximate size to those observed in the non-transfected cells but present at a much a higher density in patches of nuclear envelope devoid of nuclear pores (Fig. 3A and B). Despite the high density of the protein structures, we did not detect a semi-crystalline arrangement such as shown for InsP₃R in cerebellar Purkinje cells [20] and for RyRs in skeletal muscle [45].

Figure 3. Detection of InsP₃R in rotary shadowed images of isolated nuclei.

Nuclei isolated from Sf9 cells expressing either rat type 1 IP₃R (A), or type 3 IP₃R (B), or Wolframin, a non-related ER membrane protein, (C) were freeze-dried and rotary shadowed. Nuclear membranes from type 1 and type 3 InsP₃R-expressing cells showed large membrane patches occupied by a high density of small structures ~15 nm on the side similar to the rarer particles observed in the uninfected cell nuclear membranes. Nuclear pores were absent from these patches but visible elsewhere in the same nuclei. The small particles were not visible in cells transfected with cDNA for Wolframin. Scale bar, 100 nm.

Although the above data revealed a direct correlation between over-expression of InsP₃R and the density in the nuclear envelope of InsP₃R-like structures, nonspecific effects of the transfection, for example infection-induced over-expression of an unknown viral protein, had to be ruled out. To address this question, we infected Sf9 cells with either Wolframin, an unrelated ER membrane protein that assembles into high molecular weight complexes [46] or with the empty baculovirus vector. In both cases there was no increase in the appearance of the InsP₃R-like structures (Fig. 3C for Wolframin; data not shown for the viral vector).

The high frequency of presumed InsP₃R particles in the overexpressing nuclei strongly confirms their tentative structural identification.

3.4 Identifiable features of presumptive InsP₃R particles in overexpressing cells

The majority of profiles in the large patches of InsP₃R-overexpressing nuclear envelopes had several common structural features. Like the less frequent profiles on native nuclei, the profiles fit into a square of fairly uniform sixe (15×15 ± 2 nm, mean ± SD, from 300 particles) and some had a central depression (Fig. 4 A and B, black squares). The overall shape was approximately square, but the square had variably blunted corners. Most of the shape variations are to be expected by the superimposition of 3–4 nm size particles from the platinum shadow on the small molecule outline. The height of the particles was fairly uniform, as indicated by the intensity of the platinum shadow surrounding the particles in a given image. The intensity of the platinum shadow surrounding the particles is due to the fact that, since samples were rotary shadowed, each particle is surrounded by a ring of deposited platinum, and the taller the particle the larger the amount of platinum in the ring. In each cluster of particles, structures other than presumptive InsP₃R particles were present at low frequency and were clearly distinguishable due to their larger and more variable size and height (Fig. 4B, white and black squares and arrowheads). High magnification views of the most clearly defined presumptive InsP₃R particles revealed four equal components symmetrically disposed around a central depression, suggesting a tetrameric structure that measured ~15 nm on each side (Fig. 4C). These characteristics are consistent with previous descriptions of purified InsP₃R [41, 47], reinforcing the idea that these structures truly correspond to the InsP₃R.

Figure 4. Counting InsP₃R-like structures.

A and B show electron micrographs of the same area of a nuclear envelope from a cell over-expressing type 3 InsP₃R. In B, small squares with inside dimensions of 15×15 nm are superimposed on some of the presumed InsP₃R profiles. The profiles thus selected, even though not exactly “square”, fit well within the square outlines and some have a central depression. Additionally they all have similar height as indicated by the equal intensity of the platinum shadow at their edges. Numerous other particles with similar characteristics are seen in the same image. Arrowheads point to profiles that are clearly different: they are mostly larger, with a more variable shape and a darker shadow outline, indicating taller structures. The 15×15 nm black and white squares superimposed on these structures cover them only partially. Scale bar= 50 nm. C) High magnification views of InsP₃R-like structures from nuclei of cells over-expressing type 1 and 3 InsP₃Rs. The four particles shown here have equal sizes and an approximate square outline, but this is emphasized by the shadow only in the two particles at left. The hint of four subunits and a small central depression are typical details of the cytoplasmic domain in InsP₃Rs images obtained by single particle analysis [38,39,40]. Scale bar, 10 nm.

The observed variability of the shape of the presumed InsP₃R on native nuclear envelopes is not an unexpected phenomenon, since exposed proteins tend to collapse to some degree once the supporting water is removed during the freeze-drying process. In addition the size of the platinum shadow grains is less than an order of magnitude smaller than the size of the molecule, thus affecting its appearance

It is important to note that both in the native nuclei and in those from overexpressing cells, the receptors are never associated with each other in a regular semi-crystalline disposition, even when they are in fairly close proximity. Occasional groups of two and short rows of 3–4 receptors are present in the overexpressing nuclei, but they represent less than 10% of the particles and do not show any precise arrangement.

3.5 Specific gold labeling identifies InsP₃R in shadowed images

To further confirm the identity of the observed structures as InsP₃Rs, we labeled the receptors with either heparin, a competitive InsP₃ antagonist of the InsP₃R, or a specific antibody, both followed by gold-conjugation and freeze-drying rotary shadowing. Both techniques require that the label be retained in the replica after limited cleaning in dilute bleach (0.5%) solution followed by repeated washings in water, and they exploit the fact that gold particles provide a strong contrast allowing their visualization after shadowing, even though the images are of lower quality due to remaining contamination from the poorly digested biological material [36, 48]. As recombinant type 1 and type 3 InsP₃R did not appear to behave differently in terms of expression, or distribution, we performed the labelling experiments using nuclei isolated from cells over-expressing the recombinant type 3 isoform, for which the amino terminal-directed antibody has a high affinity.

With either label, the large majority of the gold particles were associated with small profiles of the appropriate size (Fig. 5A and B) and only a very small fraction adhered to apparently non-specific sites (e.g., see the outlined dot indicated by an arrow in Fig. 5A). Counts from 70 pictures in 6 different preparations (three labeled with heparin and three with antibody) showed 398 gold particles (92%) associated with InsP₃R-like structures and 30 (8%) positioned over either unspecified structures or the background. This indicates a high specificity in the labeling. Omission of the heparin incubation step resulted in a complete absence of labeling by the streptavidin-gold conjugate (not shown). Importantly, neither incubation with heparin nor the antibody, followed by gold-conjugation, showed any labeling of nuclei isolated from the InsP₃R-deficient DT40-KO cells (n = 3 preparations, data not shown).

Figure 5. Shadowed images of nuclear envelopes from Sf9 cells expressing rat type 3 InsP₃R.

The nuclei were treated either with gold-labeled heparin (A) or with anti-type 3 InsP₃R followed by gold-labeled secondary antibody (B) followed by freeze-drying and rotary shadowing (see Methods). In both cases, the 5 nm gold particles are clearly associated with the InsP₃R-like structures, with the exception of 1 out of 11 gold particles in (A) (arrow). High magnification (insets) shows the gold particles mainly located in the central region of the structures. Note that the InsP₃R-like structures labeled by the gold particles have similar sizes but variable shapes and overall resemble the presumptive InsP₃R particles described above. The images are not as sharp as those in the previous figures because the technique used leaves some cellular debris associated with the shadowed replicas.

Both gold labels unequivocally identify InsP₃R as particles that have the structural characteristics describe above,

3.6 Image analysis identifies the presumptive InsP₃Rs as belonging to a single population

The question remains whether particles selected as InsP₃R-like profiles by visual criteria as above represent a single population, despite their minor structural variability. This question can be explored by determining whether their “footprints” occupy an area of the image that is fairly well-defined. To do so, we performed image analysis in combination with Lilliefors test to determine whether the areas of the presumed InsP₃Rs present a Gaussian distribution as would be expected for a single population of particles. Figure 6A (top) shows a sample shadowed image and Figure 6A (bottom) the outlines of the Regions of Interest (ROIs) that were segmented from the structures identified by us as presumptive InsP₃R in the same image (see Materials and Methods). Figure 6B plots the area histogram for 445 segmented InsP₃R–like structures with a mean size and standard deviation of 132 ± 31 nm² from three images. Lilliefors test for each sample determined that at a 5% significance, footprints had a normal distribution.

The analysis indicated that despite some differences in shape, all structures tentatively identified as InsP₃R are part of a single population in respect to their surface area.

It is noticeable that the ROI areas presented in the histogram of Fig. 6B are all smaller than the area subtended by a 15×15 nm square, This is for two main reasons, One is the fact that the particle outline does not entirely fill the square outline, the second is that the segmentation process follows the outline of the dense platinum grains surrounding the protein and these leave small short crevices in the outline that subtract from the overall area. In addition, the areas for 10–15% of footprints at the left of the histogram are quite small. These footprints represent profiles of particles that are not fully included in the image because they are very closely associated with other particles (see arrow in Fig 6 A).

3.6 Can shadowed images be used to estimate InsP₃R density on the nuclear envelope?

The over-expression and gold labeling experiments clearly show that InsP₃R can be identified on the shadowed surface of isolated nuclei. However, the appearance of the shadowed receptors has a small inherent variability, and in addition, other unidentified structures occupy the same nuclear membrane patches. We were interested in defining how accurately we could estimate receptor densities from shadowed images. To that effect we used images from over-expressing nuclei and in them identified presumptive InsP₃R profiles using their fit into a 15×15 nm square in addition to an intensity of shadow indicating a fairly uniform height (Fig. 4 A and B). Most profiles in overexpressing nuclei clearly fit within the square outline, whereas a smaller number were questionable (question marks) and even fewer were simply undefined structures that did not fall into the InsP₃R-like category (Fig. 4B, arrowheads). In the example shown in Figure 4B, for example, 40 profiles are quite clearly presumptive InsP₃Rs, 17 profiles are equivocal and 10 are undefined structures. The disagreement in this assignation between two independent observers was less than 10%, providing an estimate of the possible error involved in such identifications. Counts of InsP₃R-like structures in randomly collected images showing extended patches of particles in cells overexpressing InsP₃R revealed an average density of 90.5 ± 7.4 µm⁻² of nuclear envelope (exclusive of nuclear pores) for type 1 and 92.1 ± 6.2 µm⁻² for type 3 (mean ± SD from 3 areas each in 6 different preparations). By contrast, the average density of similar structures was much lower in nuclei from native Sf9 cells: 3.2 ± 1.2 µm⁻² (mean ± SD, from 3 areas each in 6 different preparations), that is ~30-fold lower than in the overexpressing cells. Despite the uncertainty level defined above, the structural counts are consistent with functional studies from Sf9 cell nuclear patch clamp recordings, where ~2 channels per ~ µm² nuclear patch were detected on average in the same cells and a much higher level in overexpressing cells. [42, 43, 44].

4. Discussion

Using a direct shadowing technique and relying on the known structural signature of the InsP₃R cytoplasmic domain, we have been able to directly visualize the channels in the outer membrane of the nuclear envelope of Sf9 cells. The InsP₃R channels in this membrane are functional as shown by nuclear patch clamp electrophysiology [42]. The identity of the structures as InsP₃R channels is strongly suggested by four observations. First, by the structural similarity of presumptive InsP₃Rs in the images to native InsP₃Rs in cerebellar Purkinje cells visualized by shadowing techniques [20] and to heterologously-expressed channels in Sf9 cells imaged by atomic force microscopy [49]; second, by their increased density upon InsP₃R over-expression; thirdly, by population analysis confirming that the surface area of the InsP₃R-like segmented structures is distributed normally; and finally, but most importantly by their specific labeling with both heparin and a specific anti-InsP₃R antibody.

Gold-labeling before shadowing was first introduced by Pinto da Silva in 1984 [48] for freeze fracture samples to successfully study cytoskeletal networks [50, 51] and nuclear pores [36]. Labeling after fracture and shadowing has been effectively used for intrinsic membrane proteins [52, 53]. In both techniques it is essential that the digestion steps used to clean the tissue away from the shadowed replica allow retention of the label. We were successful using a brief exposure to a dilute bleach solution and obtained a high degree of specificity. The limited extent of labeling, which is clear in all above studies, is mostly related to the affinity of the specific primary and secondary probes for their ligands, rather than to a loss of the label during preparations.

The size of the exposed cytoplasmic domains of InsP₃R has been reported to vary from a ~12×12 nm square in deep etched samples [20], to ~15×15 nm in our freeze dried preparations, to ~ 22×22 nm in single particle analysis [41], to ~30 nm in the atomic force images [49]. Obviously, the cytoplasmic domain of the InsP₃R is highly deformable, which is perhaps not surprising given its highly hydrated structure. The larger size of the structures observed with AFM seems to be intrinsic to the technique, because AFM images of type 1 RyR show a size of 40–50 nm [54], almost double the size detected by rotary shadowing [55]. In the most recent reconstruction of the molecule [41], the space-filling outline of the cytoplasmic domain shows a compact structure with a square outline and barely distinguishable subunits, which would appear quite similar to our images if superimposed with randomly arranged 3–4 nm dots of platinum shadow.

Our studies confirm that in native nuclei InsP₃R are located in the outer nuclear envelope, with the cytoplasmic domain projecting into the cytoplasm, supporting the concept of nuclei as InsP₃-dependent Ca²⁺ stores [25]. However, the density in native nuclei of Sf9 cells is very low, and preliminary, unpublished data also show a very low density in nuclei from cultured DT40 and from native mouse cardiac and liver cells. The low density agrees with functional data from patch clamping studies [42]. Thus, this approach offers the first direct means of detecting receptor distribution and frequency in the nucleus, and may be useful in assessing InsP₃R nuclear localization, including ligand-independent and -dependent clustering [14, 16, 17, 56–58]. Unfortunately, we could not visualize the inner nuclear envelope by this technique, and thus we cannot state whether InsP₃Rs are located facing the nucleoplasm.

Two important observations were made in our studies. First, we note that recombinant InsP₃R channels highly over-expressed in the nuclear envelope of Sf9 cells do not appear to organize into ordered semi-crystalline arrays, even when they are present at high levels and within small regions. By contrast, high density of InsP₃Rs in the ER membrane natively expressed in cerebellar Purkinje cells [20], or over-expressed in COS cells [19], is associated with formation of flat cisternae that are decorated by semi-crystalline array of the channels. To understand this discrepancy, it is informative to compare observed InsP₃R arrays with those formed by RyRs, Ca²⁺ release channels of ER/SR that share sequence similarity, large cytoplasmic domains and a tetrameric structure with InsP₃Rs. RyRs form extensive arrays when they are either natively expressed in muscle cells [59], or expressed in a heterologous system [60], or assembled with lipids under in vitro conditions [61]. Within RyR arrays, the cytoplasmic domains of adjacent molecules are in intimate, identical contact at their four corners, so that it is logical to assume that direct intermolecular interactions are responsible for array formation, particularly when the purified protein is considered [62]. By contrast, InsP₃Rs present in arrays in Purkinje cells are located within “touching” distance in one direction, but at a larger, even though perfectly regular, distance in another direction [20]. Thus, it seems likely that some component of cell, rather than a direct intermolecular contact between the channels, may be at least in part responsible for InsP₃R array formation observed in some cells. This presumed organizing element must be absent from the nuclear envelope, so that InsP₃Rs do not form arrays when over-expressed in the nuclear envelope of Sf9 cells (this work) and of Xenopus oocytes (unpublished results) even when the receptors are sufficiently crowded within the membrane to allow intermolecular interactions.

A second observation concerns the localization of the InsP₃ binding sites within the channels’ tetrameric quaternary structure. It has been speculated that they reside either about half-way along the sides [40, 63] or near the center [38] of the large cytoplasmic region of the receptor. Heparin is a competitive inhibitor of InsP₃R [64] that binds to the amino-terminal InsP₃-binding region [65]. Colloidal gold-conjugated heparin has been previously used to detect InsP₃R by electron microscopy of negatively stained preparations [47]. Peripheral binding of heparin-gold to immunopurified type 1 InsP₃R suggested that the InsP₃-binding domain is located far on the corners of the molecule [47]. However, in our studies with the channel in native ER membrane, heparin as well as the amino terminus antibody both marked sites very close to the center of the channel’s cytoplasmic domains, supporting a more central location. The channels in our study were not InsP₃-liganded and were likely in a “closed” configuration. It is possible that InsP₃ binding sites exist in different locations in different channel conformations, for example those associated with channel gating or binding to various ligands such as Ca²⁺. Only one gold particle was associated with each InsP₃R, whereas each channel contains four subunits and thus four amino terminals. This is probably accounted for by steric hindrance by the large size of the gold particles that prohibited multiple-site binding.

5. Conclusions

In summary, we have described a system and procedures for visualizing the organization and spatial relationships of individual InsP₃R channels in native nuclear envelope membranes. In the future, this technique may be valuable for quantification of channel distribution parameters that are critical parameters in models of InsP₃R-mediated cellular Ca²⁺ signals.