Deep-etch structure of astrocytes at the superficial glia limitans, with special emphasis on the internal and external organization of their plasma membranes

Structure and functions of aquaporin-4-based orthogonal arrays of particles

Orthogonal arrays or assemblies of intramembranous particles (OAPs) are structures in the membrane of diverse cells which were initially discovered by means of the freeze-fracturing technique. This technique, developed in the 1960s, was important for the acceptance of the fluid mosaic model of the biological membrane. OAPs were first described in liver cells, and then in parietal cells of the stomach, and most importantly, in the astrocytes of the brain. Since the discovery of the structure of OAPs and the identification of OAPs as the morphological equivalent of the water channel protein aquaporin-4 (AQP4) in the 1990s, a plethora of morphological work on OAPs in different cells was published. Now, we feel a need to balance new and old data on OAPs and AQP4 to elucidate the interrelationship of both structures and molecules. In this review, the identity of OAPs as AQP4-based structures in a diversity of cells will be described. At the same time, arguments are offered that under pathological or experimental circumstances, AQP4 can also be expressed in a non-OAP form. Thus, we attempt to project classical work on OAPs onto the molecular biology of AQP4. In particular, astrocytes and glioma cells will play the major part in this review, not only due to our own work but also due to the fact that most studies on structure and function of AQP4 were done in the nervous system.

Ultrastructure of intermediate filaments of nestin- and vimentin-immunoreactive astrocytes in organotypic slice cultures of hippocampus

Glial cells in rat hippocampal slices cultured for 4 weeks were examined with immunocytochemical and cryoelectron microscopical methods. Astrocytes possessing long processes were similarly stained with antibodies against nestin, vimentin, and glial fibrillary acidic protein as seen by confocal microscopy. The three antibodies also labeled intermediate filaments in these astrocytes. In order to examine the fine structure of these intermediate filaments, slices were rapid-frozen for freeze-substitution and freeze-etching. By freeze-substitution the processes of the astrocytes were packed with large hundles of intermediate filaments. In rapid-freeze deep-etched slices, these filaments were often interconnected with filamentous cross-bridges. These cross-bridges were rather uniform in size and shape (mean 2.9 nm thick and 14.8 nm long). These results suggest that the filament network with these cross-linkers is important for shaping the long processes of nestin- and vimentin-immunoreactive astrocytes in slice cultures.

Basement-membrane stromal relationships: interactions between collagen fibrils and the lamina densa

Collagens, the most abundant molecules in the extracellular space, predominantly form either fibrillar or sheet-like structures-the two major supramolecular conformations that maintain tissue integrity. In connective tissues, other than cartilage, collagen fibrils are mainly composed of collagens I, III, and V at different molecular ratios, exhibiting a D-periodic banding pattern, with diameters ranging from 30 to 150 nm, that can form a coarse network in the extracellular matrix in comparison with a fine meshwork of lamina densa. The lamina densa represents a stable sheet-like meshwork composed of collagen IV, laminin, nidogen, and perlecan compartmentalizing tissue from one another. We hypothesize that the interactions between collagen fibrils and the lamina densa are crucial for maintaining tissue-tissue interactions. A detailed analysis of these interactions forms the basis of this review article. Here, we demonstrate that there is a direct connection between collagen fibrils and the lamina densa and propose that collagen V may play a crucial role in this connection. Collagen V might also be involved in regulation of collagen fibril diameter and anchoring of epithelia to underlying connective tissues.

The organization of neurofilaments accumulated in perikaryon following aluminum administration: relationship between structure and phosphorylation of neurofilaments

Neurofilaments accumulated in perikarya and dendrites of anterior horn cells and Purkinje cells of rabbit treated by aluminum chloride were analysed with a variety of techniques. Four different monoclonal antibodies against phosphorylated and nonphosphorylated epitopes on neurofilament H subunit were used to compare phosphorylation state of these accumulated neurofilaments with that of axonal neurofilaments. Although immunoblotting revealed no significant difference in phosphorylation between control and aluminum-treated brains, accumulated neurofilaments were immunocytochemically more phosphorylated than control perikaryal or dendritic neurofilaments. With detailed analysis of cryothin-section immunogold labeling, accumulated neurofilaments were, however, significantly less phosphorylated than axonal neurofilaments. With quick-freeze deep etching, core filaments of accumulated neurofilaments are as dense as axonal neurofilaments but much less regularly aligned. Cross-bridges of accumulated neurofilaments were less frequent and more branched than those of axonal neurofilaments, and when examined with combined immunocytochemistry and deep etching, were less phosphorylated. These results suggest that there is a relationship between the phosphorylation and the structural organization of neurofilaments. The phosphorylation of neurofilament H subunit may be necessary for formation of frequent and straight cross-bridges and resulting regular alignment of core filaments.

Phosphorylation of neurofilament H subunit as related to arrangement of neurofilaments

To find out what causes differences in phosphorylation states in neurofilaments (NF), we selected two types of dendrite, one provided with very few NFs (Purkinje cell) and the other with relatively many (anterior horn cell). We examined these with four monoclonal antibodies selected by the Western blot analysis, two (NE14 and SMI31) recognizing only phosphorylated, SMI32 recognizing only nonphosphorylated, and N52 recognizing phosphorylation-independent epitopes of NF-H. The immunoperoxidase labeling of dendrites, and also of perikarya, in both neurons was detectable with all four antibodies. After the tissue was treated with Triton X-100, the labeling was still detectable with SMI32 or N52, but undetectable with NE14 and SMI31. The brain homogenate Triton-extracted supernatant after centrifugation at 100,000g for 1 hr showed the staining of NE14, SMI31, and N52 but not that of SMI32. In Purkinje cell dendrite and perikaryon, NFs always appeared singly. In the immunogold labeling, they were labeled only with SMI32 or N52. Labeling by NE14 or SMI31 was distributed throughout the cytoplasm and hardly associated with NFs. In the anterior horn cell dendrite and perikaryon, NFs appeared both singly and in bundles. They were predominantly labeled with SMI32 or N52 when they were single, and with NE14, SMI31, or N52 when they were bundled. Even in one NF, portions that appeared single were labeled mostly with SMI32 or N52, while the remainder, to which other NFs approached closely, were labeled mostly with NE14, SMI31, or N52. Thus, when NFs appear singly, NF-H in their projections or cross-bridges with other organelles is not phosphorylated, while when NFs are bundled, NF-H is phosphorylated in crossbridges between NF core filaments. These data may explain why the NF-H is heavily phosphorylated in axons, where NFs are abundant, and not in dendrites and perikarya, where NFs are sparse.

Reactive astrocytes in myelin-deficient rat optic nerve reveal an altered distribution of orthogonal arrays of particles (OAP)

Reactive astrocytes are a common feature of various pathological conditions within the CNS. Morphological changes of reactive astrocytes include an altered nucleus-cytoplasm relationship, nuclear indentations, an increased amount of intermediate filaments, and an immunologically immature phenotype. Additionally, the number of orthogonal arrays of particles (OAP) was found to be increased within parenchymal membranes of reactive astrocytes. This observation prompted us to investigate the distribution of astroglial OAP in the amyelinated CNS of the myelin-deficient (md) rat in which reactive astrocytes prevail. In the present freeze-fracture study, astroglial OAP were determined within endfoot and nonendfoot (parenchymal) membranes in the developing optic nerve of md rats and normal littermates at the age of 12, 23, 40, and 64 days postnatally (dpn). The endfoot OAP density in md astrocytes remained constant during the entire period of investigation. In myelinated littermates, OAP densities continuously increased up to adult values. In contrast, the parenchymal OAP density in md astrocytes increased during the entire period of investigation. In normal littermates, OAP densities remained constant during the first 40 dpn and thereafter increased rapidly. These observations suggest that the absence of myelinogenesis in the md mutant may be a stimulus for parenchymal membranes of reactive astrocytes to insert OAP or to assemble OAP subunits into complete arrays.

Cytoplasmic architecture of the axon terminal: filamentous strands specifically associated with synaptic vesicles

Cytoplasmic architecture of axon terminals in rat central nervous tissue was examined by quick-freeze deep-etch method to determine how synaptic vesicles and their associated cytoplasmic environment are organized in the terminal and to know how these structures participate in the mechanism for neurotransmitter release. The axoplasm is divisible into two domains: one occupied by mitochondria in the middle of the terminal, called the mitochondrial domain, the other situated in the periphery and exclusively filled with spherical synaptic vesicles, 50-60 nm in diameter, the synaptic vesicle domain. The most characteristic feature of the mitochondrial domain was the appearance of many microtubules connected with mitochondria by filamentous strands. Large vesicles, 80-100 nm in diameter, were preferentially associated with the mitochondrial domain, and linked with microtubules wherever they appeared. The cytoplasmic matrix of the synaptic vesicle domain showed a more fibrillar texture than that of the mitochondrial domain because of the distribution of filamentous strands associated with synaptic vesicles. These strands were significantly thicker and longer (mean 11.7 nm thick and 42.7 nm long) than those linking membrane-bound organelles to microtubules (mean 8.3 nm thick and 23.0 nm long), and connected vesicles to one another or to the plasma membrane, making a complicated network around the vesicles. Further, both strands were significantly different in dimension from actin filaments (mean 9.9 nm thick and 73.5 nm long) showing 5-nm axial periodicity. These strands, especially synaptic vesicle-associated ones including their network, were readily broken down in the most part by detergent treatment or chemical fixation, indicating that they are very delicate in nature. Granular materials, which are spherical and vary in size (6-20 nm in diameter), are also more conspicuous in the synaptic vesicle domain than in the mitochondrial domain. More fibrillar and granular cytoplasmic structure of the synaptic vesicle domain may be crucial for synaptic vesicles to perform an essential role in releasing the transmitter.

Intermingling of central and peripheral nervous tissues in rat dorsolateral vagal rootlet transitional zones

The morphology of the CNS-PNS transitional zone of adult rat dorsolateral vagus nerve rootlets is uniquely complex. A typical rootlet contains a transitional zone over 300 microns long, consisting of a central tissue projection extending distally into each rootlet and a peripheral tissue insertion extending for a longer distance deep into the brainstem. The peripheral tissue insertion is continuous with the peripheral tissue of the free rootlet through channels traversing or running parallel to the central tissue projection. Accordingly, the vagal CNS-PNS interface is topologically much more complex than that found elsewhere. In some rootlets the peripheral tissue in the brainstem constitutes an isolated island deep within the neuraxis. In others, peripheral continuity is established only through a cross connection with the peripheral tissue insertion of a neighbouring rootlet. About one fifth of all vagal myelinated axons alternate between the CNS and PNS tissue compartments. This distinguishes the vagus from all other nerves studied to date. These axons are myelinated by Schwann cells distal to the transitional zone, by oligodendrocytes in the central tissue projection and by one or more short intercalated Schwann internodes further centrally, mostly in the peripheral tissue insertion, where their perikarya commonly form closely apposed aggregates. More than four fifths of all unmyelinated axon bundles alternate between central and peripheral tissue compartments, commonly more than once. In the peripheral tissue insertion axons are enveloped by series of non-myelinating Schwann cells. Schwann processes commonly extend for over 50 microns into the central compartment at each central-peripheral transition. Around one fifth of peripherally unmyelinated axons have an oligodendrocytic sheath in the central compartment. Of these axons possessing more than one intercalated Schwann internode, over one quarter display alternation of myelinated and unmyelinated segments in the peripheral tissue insertion. Astrocytes in the transitional zone segregate PNS tissue, a role played by sheath cells further peripherally in the vagal rootlets. Astrocytes form the surface limiting membranes of the central tissue projection and the barrier between the peripheral tissue insertion and the surrounding brainstem. The barrier consists only of an attenuated layer of processes. This is deficient in places, where oligodendrocytic myelin sheaths are directly exposed to the endoneurial space of the peripheral tissue insertion and in some instances are apposed to myelinating or non-myelinating Schwann cells. Such communication between the central and peripheral compartments is unique to the vagal transitional zone. The findings are consistent with a range of possible events during development.(ABSTRACT TRUNCATED AT 400 WORDS)

Orthogonal arrays of particles in astroglial cells: quantitative analysis of their density, size, and correlation with intramembranous particles

Astroglial cells were investigated by means of freeze-fracture in normal rat and mouse brain, cell culture and human gliomas. Membranes of these cells were quantitatively analyzed for their intramembranous particles (IMPs) and orthogonal arrays of particles (OAPs). Measurement of the size of OAPs and IMPs has permitted the search for a correlation between the 7-nm IMPs, which are distributed randomly in the membrane, and the subunits of OAPs (OAP-Su, also 7 nm in diameter). Using cultured astroglial cells treated with basic fibroblast growth factor (bFGF), arginine vasopressin, or sorbitol, good evidence for a relationship between the density of 7-nm IMPs and the size of OAPs can be demonstrated. These findings led to a hypothetical model of OAP modulation. A preliminary report has been published elsewhere (Neuhaus and Wolburg, 1989).

Substructure in the assemblies of intramembrane particles in astrocytic membranes

Assemblies are a specialization of intramembrane structure in astrocytes which is concentrated in astrocytic processes at the interface with vascular and cerebrospinal fluid compartments. When astrocytic processes are rapidly frozen and then freeze-fractured at very low stage temperature, the constituent particles of assemblies fracture in at least two planes in the lipid bilayer. The true outer surface of the astrocytic membrane can be exposed by etching rapidly frozen tissue and in such preparations the assemblies are seen to extend through the extracellular half of the membrane to be exposed on the surface to the extracellular fluid. The dimensions of the particles, their tendency to fracture at several levels and their exposure to the extracellular space all indicate that they are not composed solely of lipid. We conclude that assemblies represent a protein which partially or entirely bridges the membrane and which may serve a transport function.

Developmental alterations in membrane organization of rat subpial astrocytes

Subpial astrocytic processes were examined in developing rats, mainly with complementary replicas, to see how orthogonal arrays of particles (OAs) are formed and become numerous in membranes covered by basal lamina. Only a few (4.2%) endfeet in the membranes contacting the basal lamina (subpial membranes) had acquired OAs by the 19-day foetal stage. The number of endfeet provided with OAs increased drastically in the prenatal period, continued to increase at birth (P0), and somewhat more slowly in the early postnatal period (P0-P3), reaching 100% at P10. There were neuronal processes as well abutting on the basal lamina at the pial surface but they were easy to distinguish from astrocytic endfeet because of their larger intramembrane particles (IMPs), which are sparsely distributed and in patch-like aggregations. The distribution density of OAs in differentiated astrocytic endfeet also increased very gradually with age until P0, a little faster in the early postnatal period, and drastically from P10 to adult. Ordinary globular IMPs increased in number with age and continued to increase in the lateral membrane where OAs were still very few, though less rapidly in the subpial membrane as OAs became numerous. With maturation, larger IMPs became conspicuous in the lateral membrane but not in the subpial, suggesting that larger IMPs were predominantly used to form OAs. We have proposed the idea that relatively large IMPs line up to form single linear arrays (SLs), appearing as grooves on the E face, and that occasionally some SLs line up in multiple rows [multiple linear arrays (MLs)] and that SLs or MLs fuse with one another to become rod-like strands, then divide into squares to become OAs. SLs and MLs appeared ontogenetically earlier than OAs, and continued to appear in membranes provided with OAs. In areas where membranes were bent, transition of these three structures was observable and the proportion of OAs increased with age. Further, in such areas, alignment of OAs was different according to membrane curvature: concentric in and around protrusions, perpendicular to the edge of invaginations. This unique association of OA alignment with membrane curvature suggests that OAs contribute to some membrane stability in the area covered by the basal lamina and provide the membrane with special resistance to bending.

Substructure of cisternal organelles of neuronal perikarya in immature rat brains revealed by quick-freeze and deep-etch techniques

Membrane-bounded organelles possessing cisternae, i.e., rough endoplasmic reticulum and Golgi apparatus, in immature rat central neurons were examined by quick-freeze and deep-etch techniques to see how their intracisternal structures are organized and how ribosomes are associated with the membrane of the endoplasmic reticulum. Cisternae of endoplasmic reticulum, 60-100 nm wide, were bridged with randomly-distributed strands (trabecular strands, 12.5 nm in mean diameter). Luminal surfaces of cisternae of the endoplasmic reticulum were decorated with various-sized globular particles, some as small as intramembrane particles, and others as large as granules formed by soluble proteins seen in the cytoplasm. A closer examination revealed much thinner strands (3.3 nm in mean diameter). Such thin strands were short, usually winding toward the luminal surface, and sometimes touching the luminal surface with one end. Ribosomes appeared to be embedded into the entire thickness of cross-fractured membranes of endoplasmic reticulum, that is, their internal portions appeared to be situated at almost the same level as the cisternal luminal surface. From the internal portion of ribosomes, single thin strands occasionally protruded into the lumen, suggesting that these thin strands were newly synthesized polypeptides. A horizontal separation within ribosomes appeared to occur at the same level as the hydrophobic middle of the membrane of the endoplasmic reticulum. Interiors of the Golgi apparatus cisternae, which were much narrower than cisternae of endoplasmic reticulum, were similarly bridged with trabecular strands, but the Golgi trabecular strands were thinner and more frequent. Their cisternal lumina were also dotted with globular particles. No identifiable profiles corresponding to the thin strands in the endoplasmic reticulum were observed. Golgi cisternae showed a heterogeneous distribution of membrane granularity; the membrane in narrow cisternal space was granule-rich, while that in expanded space was granule-poor, suggesting a functional compartmentalization of the Golgi cisternae.

Orthogonal arrays of particles in plasma membranes of Müller cells in the guinea pig retina

Plasma membranes of guinea pig Müller cells were examined with a freeze-fracture technique to see how orthogonal arrays are distributed in the avascular retina. Examination of the portion approximately intermediate between the optic disc and equator of the eyeball showed that all end-feet of Müller cells were provided with arrays. Orthogonal arrays were concentrated on vitreal end-foot membranes, i.e., membranes that were covered by the basal lamina and contacted the vitreous body, called vitreal membranes here. The arrays were rarely observed in the portions of end-feet that did not contact the vitreous body, called lateral membranes. The distribution density of arrays in the vitreal membranes was 122.5 +/- 45.3/microns2, which was over 10 times higher than that (9.6 +/- 9.6/microns2) in the lateral membranes. The arrays became numerous and extended in shape at the periphery of the vitreal membrane, characteristically aligned in rows at the border where vitreal met lateral membrane, but never intruded into the domain of lateral membrane. Some arrays were composed of loosely attached particles and/or rod-like profiles. Sometimes rod-like profiles, 9-13 nm wide and 20-50 nm long, called linear structures here, were isolated, and sometimes they appeared in rows. Ordinary intramembrane particles (IMPs) were significantly smaller and less numerous in vitreal than in lateral membranes. IMPs larger than 9 nm in diameter were significantly fewer in the vitreal membranes, which suggests that they have been consumed to form the arrays. Although the distribution of orthogonal arrays is similar to that of K+ channels (Newman: J. Neurosci., 7:2423-2432, 1987), we consider the array an unlikely candidate for the ion channel, because its subunit particles do not protrude onto either the inner or outer surface of the membrane (Gotow and Hashimoto: J. Neurocytol., 17:399-413, 1988). Judging from their unique alignment in rows where the membrane is bent and vitreal and lateral membranes meet, the arrays may contribute to some membrane stability, resisting the physical tension at the interface with mesenchymal tissue.

Müller (glial) cells but not astrocytes in the retina of the goldfish possess orthogonal arrays of particles

Müller cells as the main glial component of the retina were investigated in the goldfish by means of ultrathin sections and freeze-fracture replicas. In the optic nerve head, they were directly compared with astrocytes. Whereas astrocytic endfeet bordering the vitreous body can easily be identified by their dense bundles of intermediate filaments, scarce membranous organelles, paravitreous caveolae, and lateral desmosomes, Müller cell endfeet reveal a looser arrangement of intermediate filaments, a characteristic pattern of triangularly shaped endoplasmic reticulum, large and pale mitochondria, and, if at all, very few desmosome-like junctions. The paravitreous membranes at the cytoplasmic face are covered by a fuzzy coat, which is less marked in astrocytic endfeet. Caveolae are lacking. Considering the freeze-fracture architecture of the membranes of both glial cell types, the Müller cells reveal orthogonal arrays of particles (OAP), which were predominantly located opposite to the inner limiting membrane; their density (109 +/- 33 OAP/microns 2) decreases abruptly with the loss of the contact between membrane and vitreous body. In contrast, astrocytes of the optic nerve head in the retina do not show any OAP in their membranes at all and are interconnected by tight junctions and desmosomes. The hypothesis suggesting that OAP might be correlated with K+ channels involved in the spatial buffering of the extracellular space is reconsidered with comparative reference to recent electrophysiological data. Further, the heterogeneity of Müller cell and astrocyte membrane equipment with OAP in the goldfish is briefly discussed.