Electron Tomographic Analysis of Cytoskeletal Cross-Bridges in the Paranodal Region of the Node of Ranvier in Peripheral Nerves

2.1 High Pressure Freezing

The procedures for high pressure freezing of chemically fixed tissue followed by freeze substitution follow that described in detail in Sosinsky et al., (2008). Rats were first anaesthetized by intraperitoneal injection of Nembutal (35 mg per kg body weight), and whole blood was flushed out with oxygenated rat Ringers solution at 37° C. They were then fixed by intracardial perfusion with an aldehyde mixture composed of 2% paraformaldehyde and 2.5% glutaraldehyde in 0.15 M cacodylate buffer (pH 7.4) at 37° C. The spinal root was dissected out, immersed in the same fixative and transferred to the high-pressure freezing (HPF) unit. These samples were trimmed and blotted (to remove excess water) before placing in the carrier. Samples were loaded into the carriers of the Bal-Tec HPM010 HPF unit (Bal-Tec, Liechtenstein) and operated according to manufacturers instructions. Carriers with a depth of 0.2 mm were used with thin bundles of spinal root to provide uniform ultrastructural preservation. The inert liquid, hexadecene, was added to the carrier to fill in the well, excluding air from surrounding the sample.

2.2 Freeze Substitution

After the samples were high-pressure frozen, they were freeze-substituted in dry acetone containing 0.1% tannic acid for 24 hrs at -90° C in a Leica EMAFS freeze substitution unit (FSU, Leica Microsystems, Bannockburn, IL), then in acetone containing 2% OsO4 and 0.1% of uranyl acetate for the remaining time of the procedure. Both tannic acid and uranyl acetate were added not only to help stain the material but also to aid in membrane preservation. During this procedure, the water in the sample is completely extracted and replaced by the fixative, i.e., freeze-substituted. The temperature for substitution was maintained at −90° C for 3 days to ensure that the biological structures were cryo-immobilized during this substitution. The temperature was raised gradually at a rate of 2° C per hour to −60° C and held there for 8 hours because OsO4 does not start to cross-link until around −70° C. The samples were then warmed to -20 °C at a rate of 5° C per hour and held at this temperature for 12 hours. The carriers were then taken from the FSU, the tissue removed from the carriers and quickly washed several times in dry acetone at room temperature (RT) to remove the OsO4 and uranyl acetate. Acetone above 70% can extract lipids at RT. Hence, the exposure time to it was minimized. The tissues were then incubated in a 50:50 mixture of acetone:Durcupan for 4 hour at RT followed by 100% Durcupan resin of 24 hrs infiltration with 3 changes. Resin polymerization was accomplished in a vacuum oven at 65-70 °C for 48 hours. The resin-embedded samples were removed from the oven and sections were cut using a Leica ultramicrotome. Sections were then stained 30 min in 2% aqueous uranyl acetate, followed by 15 min in lead salts. Fiducial cues consisting of 15 and 20 nm colloidal gold particles were deposited on opposite sides of the section.

2.3 Electron Tomography

Two tomograms from HPF/FS samples (both single-tilt) and two tomograms from conventionally prepared samples (1 single-tilt, 1 double-tilt) were analyzed. For each reconstruction, a series of images at regular tilt increments was collected with a JEOL 4000EX intermediate-voltage electron microscope (JEOL USA, Peabody, MA) operated at 400 kV. The specimens were irradiated before initiating a tilt series in order to limit anisotropic specimen thinning during image collection. Tilt series were recorded on film at 20,000 times magnification with an angular increment of 2° from -60° to +60° about an axis perpendicular to the optical axis of the microscope using a computer-controlled goniometer to increment accurately the angular steps. For the double-tilt series, the grid was subsequently rotated 90° and a second tilt series was collected with the same angular increment and range. The illumination was held to near parallel beam conditions and optical density maintained constant by varying the exposure time. The negatives were digitized with a Nikon CoolScan at 1800 dpi producing images of size 4033 × 6010 pixels. The final pixel size in our reconstructions was 1.41 nm. The IMOD package (Kremer et al., 1996) was used for rough alignment with the fine alignment and reconstruction performed using the TxBR package (Lawrence et al., 2006). One of the HPF/FS single-tilt series reconstructions (310 nm thickness) had a (global) resolution estimated by using the even/odd Fourier shell correlation (Cardone et al., 2005) and software provided by Alasdair Steven (NIH) to be 8.4 nm at a threshold of 0.3; the other two single-tilt reconstructions analyzed likely had similar resolution estimates because they had similar thicknesses and angular increment and range. The double-tilt series reconstruction (215 nm thickness) had a resolution estimated to be 4.6 nm by the same criterion. Volume segmentation was performed by manual tracing in the planes of highest resolution with the program Xvoxtrace (Perkins et al., 1997b). The reconstructions were visualized using Analyze (Mayo Foundation, Rochester, MN) or the surface-rendering graphics of Synu (Hessler et al., 1996). These programs allow one to step through slices of the reconstruction in any orientation and to track or model features of interest in three dimensions. Measurements of structural features were made using NIH ImageJ (Rasband, W.S., ImageJ, National Institutes of Health, Bethesda, Maryland, USA, http://rsb.info.nih.gov/ij/, 1997-2004.) Movies of surface-rendered volumes and slices through the reconstructions were made using Amira (Mercury/TGS, San Diego, CA). For the averages shown in Fig. 3, we had to first select for cross-bridges of each type that appeared straight, not kinked along their lengths. Subvolumes of similar dimension were cut out and aligned together in 3D using Analyze. After alignment, multiple subvolumes were averaged using EMAN (Schmid and Booth, 2008; Tang et al., 2007) to increase the signal-to-noise ratio, enabling visualization of greater structural details.

3.1 Overall organization of the paranode

The paranodal region is rich in juxtapositions of membrane-bounded structures of the Schwann cell cytoplasm and neuronal axoplasm and includes linkages of cytoskeletons, organelles, with numerous examples of cross-bridges between them (Fig. 1). In the Schwann cell cytoplasm and the axoplasm there are extensive cytoskeletal networks and numerous organelles such as mitochondria, endoplasmic reticulum, autophagosomes, lysosomes, and tubular and cisternae that are linked to the cytoskeleton, sometimes forming anatomosing networks. The mitochondria in the axoplasm appear different from those normally seen in tissues because they are in the condensed state (Hackenbrock, 1966), a structural configuration that is now starting to become more recognized in the axoplasm far from somas and terminals (Perkins et al., unpublished results). Condensed mitochondria have a dark matrix that appears like a long twisted sausage in 3D and are thought to be indicative of greater ATP production. The crista junction (Frey and Mannella, 2000; Mannella, 2006; Perkins et al., 1997a) may be either a narrow tube or an elongated opening (Renken et al., 2002).

The use of electron tomography revealed a variety of cross-bridges (Fig. 2), including the presence of arrays of transcellular junctions, PJBs, traversing the paranodal glial-axonal juxtaposition (Fig. 2a, a′). This juxtaposition is constituted by the terminal cytoplasmic loops of the myelin forming Schwann cell, the extracellular matrix studded with septate junction-like structures that span the space between the glial and axonal membranes, and the axolemma. The PJBs not only span the cleft between cells, but also appear to insert into both glial and axonal membranes (Fig. 2a, h, a′, h′). Tubular cisternae (TC) consist of tubules of smooth endoplasmic reticulum of various diameters and are closely associated with juxtapositional membranes in the glial loops. Filamentous cross-bridges of diameter similar to F-actin linking the TC to the loop membrane were revealed in tomographic volumes (Fig. 2b, c, b′, c′). Focal attachments of cytoskeletal filaments to the paranodal membranes were visualized in both cytoplasms of the glial loops and axon (Fig. 2d, d′). A subset of these filaments attaches to the transmembrane PJB (Fig. 2a′).

Organelles are also connected to the cytoskeleton and occasionally to each other in the axoplasm. Mitochondria were seen connected by short tethers to neurofilaments (NF) (Fig. 2e, e′) or microtubules (MT) (Fig. 2f, f′). These tethers had a diameter similar to that of neurofilaments. These filaments may be comprised of desmin or keratin (Toivola et al., 2005), which attach to mitochondria. Mitochondrial-ER cross-bridges were also observed (Fig. 2g, g′). Multiple cross-bridges of the same diameter and similar lengths were found connecting the two organelles. Microfilaments in the glial cytoplasm running more-or-less perpendicular to the filaments and microtubules in the axoplasm were found in the sub-volume closest to the juxtapositional loop membrane (Fig. 2h, h′). These actin filaments did not appear bundled, nor did they appear to have cross-bridges with each other, other cytoskeleton, or membranes. Cross-bridges are also common between neurofilaments (Fig. 2i, i′) and between neurofilaments and microtubules (Fig. 2j, k, j′, k′). These cross-bridges have a diameter about the same as neurofilaments. We noticed that a small subset of neurofilaments appear to terminate directly at the axolemma, including at the PJBs (data not shown).

Averages of tomographic subvolumes of cross-bridges enhanced the contrast of the sometimes faint individual examples of such bridges (Fig. 3). These averages have an increased signal-to-noise ratio, enabling visualization of greater structural details. A general difference in cross-bridge diameter between the NF-NF, NF-MT, and mito-MT averages on one hand and the NF-Axl and TC-PNL averages on the other hand is noticeable. In addition, a variation in diameter along the length of each bridge is evident, which might reflect variability in stain deposition.

3.1 Analysis of cytoskeletal elements in the paranodal region

Figure 4 is a 3D model that incorporates the structural information obtained in the present study. The associations of the major cytoskeletal components are shown: neurofilaments, glial filaments, microtubules, microfilaments, and cross-bridges of various sizes (Fig. 4a). The cross-bridges that connect neurofilaments to each other or to microtubules are few and far between. We found only one or two such bridges per connecting cytoskeletal components (Fig. 4b). This limited connectivity was also seen with the cross-bridging of neurofilaments to the axolemma. The PJBs are seen to be rod-like structures (Fig. 4c) of uneven lengths (mean±std dev=45±34 nm; range:16-128 nm) and arranged in nearly parallel arrays separated by remarkably uniform spacing (HPF/FS samples: mean±std=19±4 nm; conventional samples: mean±std=21±.3 nm). The focal attachment of axoplasmic neurofilaments to the PJBs appears to be complemented by a similar attachment of glial cytoskeletal filaments on the glial side. Occasionally, instead of being connected to the axolemma via cross-bridges, neurofilaments would connect by terminating at this membrane, often at the site of a PJB (Fig. 4d). This site-specific link is suggestive of intercellular association between axon and glia. Tubular cisternae, possibly involved in regulating intracellular calcium levels, are linked to the loop membrane also by only a small number of cross-bridges (Fig. 4d). Mitochondria and ER were found positioned within a dense “jungle” of cytoskeleton (Fig. 4a), yet the connection of these organelles to the cytoskeleton was limited to either a single neurofilament or microtubule (Fig. 4e) and moreover, a single cross-bridge. Intercellular cytoskeletal connections, not always apparent in thin-section micrographs or in slices through the tomographic reconstruction, were revealed after volume segmentation. An example is an axoplasmic cross-bridge connected to a glial cross-bridge through more than 40 nm of intervening PJB rod (Fig. 4f). Whereas neurofilaments and microtubules predominate the axoplasm, microfilaments are more pronounced in the loop cytoplasm (Fig. 4g, h). What role the microfilaments play is not implied from ultrastructure revealed here as we observed no cross-bridges to each other, to other cytoskeletal elements or any attachments to organelles or loop membranes (see supplemental movie).

Measurements of diameters uncovered two distinct classes of cross-bridges (Fig. 5). The range and number of measurements are found in Table 1. Those bridges connecting neurofilaments, microtubules, and organelles have different cross-bridging diameters to the second class--those connecting membranes: either filaments to the axolemma or between the TC and paranodal loop membrane. The two different classes are reflected in the probability of sampling from the same population values, p, derived from the Student's t-test. For example, p=0.0014, 0.000038, 0.00020, 1.2×10^-10, and 3.1×10^-6 for NF-NF vs. TC-PNL, NF-MT vs. TC-PNL, NF-NF vs. NF-Axl, NF-MT vs. NF-Axl, and mito-MT vs. NF-Axl, respectively. No other t-test between cross-bridge types had p<0.01, which we used as the cutoff for statistical significance, including comparisons between the measured diameters from HPF/FS and conventional data. The lengths show similarity between NF-NF, NF-MT, and mito-MT cross-bridges (Fig. 5b). A notable length dissimilarity exists between the TC-PNL cross-bridge and all the NF cross-bridges, i.e., NF-NF, NF-MT, and NF-Axl cross-bridges (p<0.01 for all 3 comparisons) and is remarkable because the NF-Axl and TC-PNL mean diameters are similar. It may be that the links position the TC close to the PNL for the purpose of limited-diffusion calcium release or sequestration. The PJB diameter of 15±2 nm (mean±std dev; range 12-19 nm) for HPF/FS samples or 16±4 nm (range 10-22 nm) for conventional samples is much greater than the fibrous cross-bridges (Fig. 5a) and emphasizes that PJBs are rods with an axis running lengthwise along the juxtapositional membranes, not between the membranes. By plotting cross-bridge diameter vs. length, we found some indication that there exists a linear correlation between four of them: NF-NF, NF-MT, NF-Axl, and TC-PNL (Fig. 5c, 5d) and more noticeable with the conventionally prepared samples. In other words, as the diameter of the cross-bridge increased, it's length also tended to increase linearly. The one exception was the mito-MT cross-bridge, consistent with our observations that these are generally short-linkers. Because of the size of mitochondria, a thicker cross-bridge might be required.

4.1 Protein composition of the PJB

An additional observation is that cross-bridges to cytoskeleton in the cytoplasm and occasional neurofilaments terminations were found connected to PJBs. These cytoskeletal components might also serve to separate this septate-like junction with a nearly uniform spacing we report here. We found that similar filamentous cross-bridges exist in the glial cytoplasm and the axoplasm originating from the PJB that might be part of an intercellular link of glial and axonal cytoskeletons. We propose that the elegant architecture of the transcellular cytoskeletal network provides a mechanical link that stabilizes the two-cell spacing particularly for efficient action potential conduction.

4.2 Cytoskeletal cross-bridges serve to maintain the structural integrity of the paranode

We found three types of neurofilament cross-bridging. One type involved connecting to the axolemma via thin fibrous extensions, usually no longer than 100 nm. A second type of association was direct attachment of one end of the neurofilament to the axolemma. The third type of bridge was a cross-linking filament, about the same diameter as neurofilaments themselves, between neurofilaments or neurofilaments and microtubules (Fig. 5). The cross-bridges connecting to the axolemma had a diameter consistent with actin. Actin filaments are outnumbered by neurofilaments by about an order of magnitude in our tomographic volumes, consistent with the report that intermediate filaments can account for up to 85% of the total protein complement in fully differentiated cells (Fuchs and Weber, 1994). The cross-bridging of neurofilaments provides flexible resilience to the axoplasm in response to external stresses by reversibly cross-linking these filaments to the other cytoskeletal filaments and attachment sites at the axolemma (Fuchs and Cleveland, 1998). Part of the cross-bridging of neurofilaments to other neurofilaments and microtubules is likely accomplished by the long carboxyl tails of the neurofilament proteins NF-M and NF-H that protrude up to 30 nm from the filament backbone. Cross-bridging might also be through plectin, BPAG1n, and related proteins (Fuchs and Cleveland, 1998).

The cross-bridges between membranes and cytoskeleton reported here have both similar and dissimilar linkers to those reported previously by Ichimura and Ellisman (1991) who used three different methods for tissue preparation and three different electron microscopic techniques. Two cross-bridging types connecting neurofilaments with the glial-axonal junction were reported in the previous work. One is via 5-8 nm filaments equivalent to those we found that have a mean diameter of about 7 nm and a range of 6-9 nm (Table I). The slight difference in range of diameter could be due to the different contrasting mechanism—negative stain in the work of Ichimura and Ellisman and positive stain in our work. The other neurofilament cross-bridges previously reported were 3-5 nm short linkers. However, we did not observe these linkers and it may be that these fine filaments are just below the resolution or detection limit in our tomographic volumes. Another cross-bridge reported in the two studies was found connecting the tubular cisternae with the paranodal loop membrane (TC-PNL). However, in the previous work, 3-5 nm fine linking filaments were observed whereas we found that the measured diameter was ∼5-9 nm. Could it be that the positive stain is adding thickness to these filaments not present in the earlier work because of the negative contrast? Support for this interpretation is seen in fig. 3 where the stain accumulation is not uniform along the fiber lengths, giving the appearance of thicker and thinner regions. Further, the grains of uranyl acetate are about 1.5 nm across and so ∼3 nm to the thickness of our positively stained cross-bridges may be attributable to stain deposition.

Mutations that weaken or disrupt the cytoskeletal network are implicated in a number of human disorders. For example, missense and splice-site mutations of tau, a protein that stabilizes microtubules, cause aberrant filamentous assembly that is involved with a variety of human cognitive diseases, including Alzheimer's disease, inherited frontotemporal dementia, and Parkinsonism (Garcia and Cleveland, 2001). Disruption of cross-bridging and abnormal accumulation of neurofilaments are pathologic for a series of motor neuron diseases, among which are infantile spinal muscular atrophy, familial and sporadic amyotrophic lateral sclerosis (ALS), and hereditary sensory motor neuropathy (Lee and Cleveland, 1996; Xu et al., 1993). The mechanism of ALS disease is thought to stem from disorganization of neurofilament connectivity that would disrupt the transportation of cargo along the axon (Collard et al., 1995; Fuchs and Cleveland, 1998).

The cross-bridges that we have analyzed are central to the cyto-architecture and structural integrity of the paranode. The continued work to identify the proteins that constitute these bridges, how they link with each other to form filaments or rods, and how they attach to membranes and cytoskeleton is anticipated to enhance discovery of functional interactions. The combination of high-resolution 3D structures of the paranodal region with the localization of specific macromolecules that populate this region is essential for subsequent modeling studies (Sosinsky et al., 2005) and for a better understanding of the structural and molecular bases for diseases affecting the node and paranode (Bergoffen et al., 1993; Bhat et al., 2001; Boyle et al., 2001; Griffin et al., 1996; Martini, 2001; Rasband et al., 2003; Sherman et al., 2005; Sima, 1993; Weber et al., 1999). The cytoskeletal filaments with their various cross-bridges, including those that are transcellular, are assumed to play an important role in maintaining the electrophysiological properties required for functioning of the node of Ranvier by establishing a mechanical link between the glial and neuronal cytoskeletons, thereby stabilizing the relationship between these cells.