In this study we examined four different strains of AxD mutant mice at three different ages to observe stages of RF formation. At all times and in all lines of the mice we found a high degree of variability in the sizes and shapes of RFs. The appearance of RFs ranged from large, dense structures to small, less dense structures that appeared to be deposited on intermediate filaments. We also found that DAPI was an excellent marker for RFs, allowing us to view the distributions of RFs.
How do RFs form?
We found many examples in which small RFs were connected to adjacent, larger RFs by filamentous and/or granular material, suggesting that RFs may form by the continued incorporation of small RFs into the larger inclusions. The smaller RFs were less electron-dense and in many we found structural heterogeneity. Intermediate filaments coursed through the small inclusions, suggesting that an early stage of RF formation is the deposition of electron dense material on filaments. We also observed in the small RFs a granular, electron-dense material of a relatively homogeneous size, about 15-30 nm. We found similar granular material attached to filaments outside the RFs, suggesting that small deposits on filaments represent an even earlier stage. These observations are consistent with a model in which the deposition of electron dense material on filaments is followed by more deposition and growth of the inclusions, followed by aggregation of small inclusions or incorporation into larger RFs. The deposited material likely attracts more material to the growing aggregate. Figure 9 depicts a possible scheme of RF formation.
Immuno-electron microscopy showed that the granular electron-dense material contains both alphaB-crystallin and GFAP. We did not look for other RF proteins in the smallest aggregates, however, and other proteins may be there as well and may be important in RF formation. Nevertheless, since alphaB-crystallin binds to GFAP [32], these two proteins could be sufficient to form aggregates. The small, granular material attached to filaments contains GFAP and alphaB-crystallin [8]. Thus, one possible mechanism underlying RF formation is the binding of alphaB-crystallin to wild type and/or mutant GFAP monomers or oligomers, which then can associate with filaments. Aggregation into RFs is likely to be a phenomenon dependent upon protein concentration and the stoichiometry of GFAP and alphaB-crystallin levels. We previously found that overexpressing alphaB-crystallin in cultured astrocytes with normal GFAP levels debundled intermediate filaments, but did not depolymerize them or lead to the formation of RFs [8].
However, genetically overexpressing alphaB-crystallin in astrocytes in AxD mice reduced the oxidative stress response, RFs, seizures, and restored astrocyte glutamate uptake ([6]). We do not know the stoichiometry of GFAP and alphaB-crystallin in individual astrocytes in either the AxD mice or in the AxD mice that overexpressed alphaB-crystallin. The alphaB-crystallin transgenics produce more alphaB-crystallin from the outset, a state somewhat different from AxD itself, and which might alter RF formation from the beginning.
Whether proteolytic degradation of GFAP plays a role in RF formation is not known. We previously performed Western blots of mouse AxD model brains and observed only a small band of GFAP- immunoreactive material migrating more rapidly than the main protein band, indicating a minimum level of GFAP proteolysis. Western blots, however, do show a series of less rapidly migrating bands, which are likely to represent ubiquitinated GFAP and GFAP oligomers [32,33,34]. However, Chen et al. [2] present evidence that GFAP is a substrate for caspase 6 and using an antibody to the cleaved peptide, found proteolyzed GFAP in cultured cells that expressed the R239H GFAP mutation and in the heterozygous KI mice. The small, 26 kDa fragment is largely insoluble, using a deoxycholate and Triton X-100 buffer, and thus it is possible that this fragment might be found in RFs and contribute to their formation.
Mouse RFs appear identical to human RFs
The idea that RFs are formed from an accumulation of smaller inclusions is consistent with reports of human biopsy and autopsy material. Thus, EM reports of human AxD note small RF-like material at the borders of larger RFs [1, 10, 19, 23]. Borrett and Becker (1985) reported a brain biopsy from a 14 week-old infant, which showed many small RFs in a bed of intermediate filaments, and concluded that these inclusions represented early RFs. We have observed the same small RFs in many of the mouse astrocytes. A biopsy of a 34 month-old child, described in the same report, showed a large RF and also several smaller ones within the large skein of filaments (see Fig. 6 in [1]). Variation in the size of RFs within a single astrocyte was also reported by Schochet et al. [23], who reported ultrastructural findings in an 8 month-old child showing a spectrum of sizes of RFs, including many small ones. RFs were connected to each other by filaments and granular material. Astrocytes in adult forms of AxD also show variation in the sizes of RFs with small inclusions situated among the larger ones [25, 29]. Neuropathologists have not previously commented on the small, granular profiles or the less dense matrix of small RFs, although Herndon et al. [10] commented on the “thickening” of filaments in the vicinity of a RF (their Fig. 13), and granular deposits on the filaments. Of course, nothing was known about the molecular composition of RFs at that time.
RFs can form anywhere in the astrocyte and continue to form over time
The presence of small RFs in all parts of the astrocyte, including cell body, processes, and endfeet, suggests that RFs can be independently formed throughout the cell, although we cannot rule out the possibility that some form in the cell body and are then transported to endfeet.
RFs continue to be generated over time in parallel with the accumulation of GFAP. We observed the small, electron dense aggregates in 1 month-old as well as in 1 year-old AxD mice. Although at every timepoint there was a large degree of variability in RF size, the numbers and sizes of RFs were greater in the older mice. Thus, RF formation is an ongoing process in AxD, rather than just occurring at early stages of the disorder.
However, even in the areas with high levels of astrogliosis not every astrocyte contained RFs. We were surprised that some astrocytes filled with GFAP did not contain RFs, whereas nearby astrocytes contained many RFs. Different levels of alphaB-crystallin and/or some other heat shock proteins that participate in RF formation might be responsible for the variable numbers of RFs in astrocytes. We could not exclude the possibility, however, that the very small RFs (200–300 nm in size) seen with the electron microscope are below the resolution of the confocal microscope.
Do RFs influence astrocyte phenotype?
Do RFs influence astrocyte phenotype or are they innocuous structures that have no effects on cell morphologies or functions? Given that RFs are always located within large skeins of intermediate filaments, it is difficult to dissect clearly the effects of RFs per se from those of the filament accumulations. Indeed, some changes of astrocytes, such as the new expression of CD44, do not depend on the presence of RFs, since high levels of CD44 were observed in cells with and without RFs. A similar case is true for other markers of astrocytes that we used (GLT-1, GLAST, ferritin, Kir 4.1, vimentin, nestin, not shown).
The two prominent changes that were consistently associated with large numbers of RFs were 1) a thickening and shortening of main processes and a loss of miniature leaf-like processes, and 2) abnormalities in nuclear morphologies. Astrocytes retract many of their processes during arrested mitoses and then extend them to their normal size after slippage from mitotic arrest (manuscript in preparation). Astrocytes filled with RFs may not be able to restore the normal shapes and sizes of their processes, presumably due to a disruption of proper cytoskeletal orientation by RFs. A loss of small, distal processes may have special significance because distal processes that isolate synapses are responsible for extraneuronal ion and transmitter homeostasis. Their absence might severely alter neuronal excitability [12, 18]. Note that the double mutant AxD mice develop seizures at 4-5 weeks of age. Transgenic and KI AxD mice are more susceptible to kainic acid-induced seizures compared to wild type mice [7, 31]. Those astrocytes that displayed such changes also showed enlargement and irregular forms of nuclei. We consider that the nuclear abnormalities originate due to arrested mitosis when RFs interfere with chromosome congression into the metaphase plate and subsequent segregation into two daughter groups, not allowing cells to fulfill cytokinesis. Finding RFs in mitotic astrocytes in mouse and in human indicates that at least some RFs do not depolymerize during mitosis. RFs between chromosomes would likely interfere with normal chromosome segregation and spindle formation. In addition, the formation of the nuclear envelope in telophase may also be influenced by RFs. Filaments are bound by the cytoskeleton linker protein plectin to nesprin-3, located in outer leaflet of the nuclear envelope [27]. Plectin accumulates in RFs [35], and thus could link RFs to the nuclear envelope and interfere with the formation or dissolution of the nuclear envelope.
We also observed that RFs with associated filament bundles often excluded membranous organelles, segregating those organelles either to a paranuclear position or to the periphery of the cell. In cultures of AxD astrocytes, we found separation of Golgi and ER complexes and fragmentation of the Golgi apparatus by large bundles of filaments and RFs (Guilfoyle and Sosunov, unpublished). These observations suggest that membrane trafficking may be disrupted in AxD astrocytes. RFs may increase the mechanical stability of filament bundles/aggregates and thus create mechanical barriers for intracellular trafficking as well for chromosome congression and segregation during mitosis.
DAPI as a new method for visualizing RFs
We found that DAPI can be used as a reliable and reproducible method of visualizing RFs. An advantage of this method is the ability to combine it with routine immunohistochemical procedures. Why RFs are stained with DAPI is not clear. One possible explanation may be that DAPI, like FJB, has an affinity for highly acidic structures. Fluorescent Nissl Stain (Neuro Trace, Molecular Probes) also gives positive staining of RFs (unpublished results) but in comparison with DAPI is much less reproducible and does not stain the small puncta-like RFs.