INTRODUCTION

Type 2 astrocytes are known to be differentiated in vitro from the bipotential oligodendrocyte-type 2 astrocyte (O-2A) progenitors, depending on the type of culture conditioning (Raff et al., 1983; Stallcup and Beasly, 1987; Asou et al., 1995; Linsky and Gilbert, 1995). However, not only the differentiation to but also the existence of type 2 astrocytes are still under debate as regards in vivo studies (Espinosa de los Monteros et al., 1993a, b; Groves et al., 1993; Amberger et al., 1997). Even if type 2 astrocytes are indeed located in the brain, immunocytochemical studies have rendered the distinction between astrocytes and oligodendrocytes more complicated. This is because type 2 astrocytes are immunoreactive for both anti-GFAP (glial fibrillary acidic protein) and anti-A2B5 antibodies (Raff et al., 1983; Linskey and Gilbert, 1995). Anti-GFAP antibody and anti-A2B5 antibody are known to be specific markers for astrocytes and immature oligodendrocytes, respectively (Raff et al., 1983). However, the prevalence of unanswered questions as regards these cells emphasizes the incompleteness of a morphological characterization for type 2 astrocytes. Minute observation of type 2 astrocytes is strongly required for such a characterization.

Recently, atomic force microscopy (AFM) has been applied for observation of fine surface structures of various cells and tissues (Ushiki et al., 1994a, 1996; Haga et al., 1998; Sasaki et al., 1998). AFM images are recorded by gently scanning throughout the cell surface with an AFM tip, a device much like the stylus of a record player. To prevent the tip from damaging the cells during the scanning procedure, a feedback system keeps constant the interatomic forces between the apex of the AFM tip and atoms in a cell surface. In the field of neurobiology, AFM has been shown capable of detecting very fine three-dimensional surface structures of cultured nerve and glial cells that otherwise could not have been observed by light microscopy or scanning electron microscopy (Henderson et al., 1992; Parpura et al., 1993; Nagayama et al., 1996, 1997; Tojima et al., 1998; Yamane et al., 1998).

Generally, eucaryotic cells can adopt a variety of shapes depending on a complex network of protein filaments called “cytoskeletons”, which are known to commonly shape the fine structures of cell surfaces (Poste and Nicolson, 1981; Borisy et al., 1984). Cytoskeletal filaments are composed of a distribution of proteins that can be detected by immunocytochemical analysis using anti-cytoskeletal filament antisera. Various AFM studies have also described the structures and the dynamic properties of cytoskeletons (Henderson et al., 1992; Tojima et al., 1998; Yamane et al., 1998). In our preliminary AFM observations of type 1 and type 2 astrocytes and microglia, we were able to detect rugged, wavy, or radial structures on the surfaces of the glia (Yamane et al., 1998). However, it was not always possible to determine if a correspondence between the AFM-detected structures and the cytoskeletons exists. To clarify the manner in which the fine structures of cell surfaces and the cytoskeletons relate to each other, immunocytochemical analysis of the cytoskeletal filaments, including microtubules, intermediate filaments, and microfilaments, aids in identifying the cytoskeleton that reflects these fine surface structures.

In the present study, we observed the three-dimensional fine structures of the surfaces of type 2 astrocytes by AFM. The heights of the structures in the AFM images were then estimated. One of the most important features of AFM is the capacity to observe the morphology of living cells (Nagayama et al., 1996). However, we observed strongly fixed cells, because AFM observation can sometimes reveal inner structures by pressing on living soft cell membranes (Henderson et al., 1992). Due to this reason, it is often difficult to compare AFM data with observations obtained by the conventional methods such as scanning electron microscopy (SEM) (Tojima et al., 1998). Even though one advantage of AFM was sacrificed in conducting the present study, other features remained helpful. In particular, in contrast to AFM observation, SEM generally requires a metal coating on cell surfaces, a process which can bury fine structures. In addition, with AFM, cells could be observed in water, which helps avoid drying and shrinkage. The distribution of cytoskeletons was examined immunocytochemically to confirm which type of cytoskeletal filament reflects the surface structures observed in AFM images of type 2 astrocytes. As a result, AFM was recognized as a powerful tool for the morphological characterization of vaguely defined cells such as the type 2 astrocytes.

MATERIALS AND METHODS

RESULTS AND DISCUSSION

Type 2 astrocytes prepared for AFM were first viewed by phase-contrast light microscopy (Fig. 1A), and then observed by AFM (Fig. 1B–D). An irregular meshwork on the cell surface, which was undetectable by light microscopy (Fig. 1A), was clearly observed by AFM (Fig. 1B–D). This meshwork structure consisted of ridges distributed throughout the cell surfaces. Some ridges ran along the cell processes; others were distributed radially toward the peripheral regions of flat membranes (Fig. 1C). In general, these ridges usually ran irregularly (Fig. 1D). Such surface structures were expected to reflect the structures of cytoskeletons (Poste and Nicolson, 1981; Borisy et al., 1984; Henderson et al., 1992; Tojima et al., 1998; Yamane et al., 1998). Typical noises in the recording process were seen along the horizontal scanning axis on the glass in Fig. 1B. We have already confirmed that these noises in the AFM observation in water have height of under 10 nm and that they do not exert a bad influence upon the analyses for our aims (Nagayama et al., 1996).

Fig. 1

Cell-surface structures of rat cerebellar type 2 astrocytes. (A) Phase-contrast photograph of type 2 astrocytes. (B) AFM image of the squared area in (A). Information of the height of the object is indicated as brightness of color; the brighter is color, the higher the object. (C) Magnified AFM image of the squared area in (B). (D) Magnified AFM image of the squared area in (C). (E) Cross-section image on the surface shown along the white line in (D). The heights of the ridges, measured from the bottom between the ridges, were estimated as 124 nm for H and 47 nm for L. (F) Frequency of the ridge-like structures emerged in (B). The data were collected at each 15 nm height. The solid line combined by 2 dashed curves is a fitted curve obtained by a logarithmic normal distribution function with 2 peaks.

Next, the heights of the ridges forming the meshworks observed in the AFM images were measured (Fig. 1E). A cross-section image of the surface indicated by a white line in Fig. 1D is shown as Fig. 1E. The height of the ridge was measured from the bottom between the ridges (see H and L in Fig. 1E). We thus obtained values of 124 nm for H and 47 nm for L. AFM is widely recognized for its capacity to measure the exact heights of observed samples (Nagayama et al., 1996). Here, “the exact heights” refer to the heights from the culture glass where the samples stick. Taking into account the difficulty of subtracting the roundness of the cell from the height between the culture glass and the top of the ridge, we must consider another method to estimate the heights of the ridges throughout the cell surfaces. To do so, we measured the angle (α) of elevation from the flat cell surface, or the bottom of the ridge, to the top of the ridge and the horizontal length (l) between the elevating point and the top of the ridge, and then calculated the height (l tan α). The histogram in Fig. 1F shows the emerging frequency, or the number, of the ridges, collected at each 15 nm height (n = 95). These data were fitted with a logarithmic normal distribution function by a nonlinear squares method. The curve with 2 peaks fit much better than that with 1 peak, that is the χ² value was significantly reduced in the case of 2 peaks (χ² = 3.1 for 1 peak; χ² = 1.4 for 2 peaks). Functions with more than 3 peaks were not able to improve the fits in a statistical sense (data not shown). The 2 peaks estimated here emerged at the heights of 31 nm and 82 nm. These analytical data suggest the existence of two possible structures under the cell membrane. It is possible that these two heights may reflect two structures consisting of two differently sized bundles of cytoskeletal filaments. However, it is also possible that two structures beneath the cell membrane are located at two different depths from the membrane.

The distribution of cytoskeletons was examined by immunocytochemical studies. The cytoskeletons are composed of three major filamentous components: microtubules which are produced by tubulins, intermediate filaments (or gliofilaments) that are mainly produced by glial fibrillary acidic proteins (GFAPs), and microfilaments (or actin filaments) (Borisy et al., 1984; Rutka et al., 1997). Double-stained immunofluorescence images for type 2 astrocytes are shown in Fig. 2A–D. Figures 2A and C represent GFAP; Figs. 2B and D represent tubulin. In these images (Fig. 2A, C and Fig. 2B, D), these two kinds of cytoskeletal filaments seemed to be similarly distributed. The white arrow in Fig. 2D points to a knotlike structure. This was observed in the anti-tubulin fluorescence image (Fig. 2D) but not in the anti-GFAP image (see white arrow, Fig. 2C). The knots are reliably confirmed by a phase-contrast photograph (see white arrow, Fig. 2E); such knots include only microtubules.

Fig. 2

Cytoskeletal structures of rat cerebellar type 2 astrocytes. (A) Immunofluorescence (fluorescein) image of a type 2 astrocyte visualizing the distribution of gliofilaments (GFAP). (B) Immunofluorescence (rhodamine) image for microtubules (tubulin). The cell in both (A) and (B) was identical and was doubly stained by anti-GFAP and anti-tubulin antibodies. (C) Magnified image of the squared area in (A). (D) Magnified image of the squared area in (B). (E) Phase-contrast photograph for the cell in (A) and (B). The white arrows in (D) and (E) point to knot-like structures, but such a structure cannot be observed in (C). Scale bars = 20 μm.

In contrast, no staining was observed for actin in our immunocytochemical study (data not shown). This result was confirmed by studies for oligodendrocytes, which are derived from the same precursors (O-2A progenitors). For example, even if immature oligodendrocytes are rich in actin filaments in their fine processes, mature oligodendrocytes do not have any actin filaments, either in the cell bodies or in the processes (Wilson and Brophy, 1989). Findings from the present study suggest that the ridges forming the meshwork revealed by AFM were structures reflecting very fine bundles of tubulin and GFAP. The exact correspondence between the heights of the ridges estimated from the AFM images and the thickness of the filamentous proteins (tubulin and GFAP) is still unclear. It should be noted that such fine ridges (tens of nanometers), revealed by AFM, cannot be observed with the resolution power of a fluorescence microscope. The threedimensional distribution of microtubules (tubulin) and gliofilaments (GFAP) in type 2 astrocytes were examined; optical sections of the stained cells were observed by confocal scanning fluorescence microscopy. The microtubules were located just underneath the cell surfaces and the gliofilaments were placed among or under the microtubules (data not shown). Rutka et al. (1997) also proposed in their model that gliofilaments link other filaments in astrocytes in the adult human brain. A summary of this indirect evidence leads to the speculation that the ridges revealed by AFM may be composed mainly of GFAP.

The heights of the observed ridges are varied; this is caused by cytoskeletal filamentous bundles. Possible conditions of these bundles are summarized in Fig. 3. First, the bundles of filaments possess at least two kinds of thickness. Hence, the two kinds of bundles of filaments with different thickness beneath the cell membrane may shape the meshwork (see (1) in Fig. 3). Second, the bundles of filaments are located in at least two different depths from the cell membrane (see (2) in Fig. 3). In such cases, it does not matter whether the thickness of the bundles is varied or unique. Third, the immunoreactive filaments shown in Fig. 2, with diameters exceeding hundreds of nanometers, change cell shape to a great extent. It is extremely hard that very fine observations by AFM can discriminate large inner structures from the curvature of the cell (see (3) in Fig. 3; see also Tojima et al., 1998 for the details).

Fig. 3

Possible configuration of ridges revealed by AFM and immunoreactive filaments. Closed and hatched circles denote the cross-sections of the very fine bundles of cytoskeletal filaments, which were reflected in AFM images, and the cross-sections of the thick bundles of cytoskeletal filaments, which were detected by immunofluorescence techniques, respectively. See text for (1)–(3).

Finally, we should discuss another protein, spectrin, which is known to form a lining structure of the cell membrane. Some isoforms of this protein were reported in neurons and glia in rats and mice (Opas et al., 1986; Goodman et al., 1989), as well as in erythrocytes (Terada et al., 1996). The spectrin network can extend or contract. The lengths between the intersections change according to mechanical strength. Even when this network was artificially expanded to form a meshwork and observed by transmission electron microscopy, the side of one mesh was estimated as being under 200 nm (Shotton et al., 1979; Byers and Branton, 1985; Shen et al., 1986). Therefore, we can neglect the possibility that the spectrin meshwork reflects the observed ridges in AFM images (Fig. 1B–D), because the maximum mesh formed by spectrin (0.1 μm²) is much smaller than the observed mesh (0.61 ± 0.02 μm², mean ± SEM, n = 130, Fig. 1B).

In summary, the present AFM study revealed that type 2 astrocytes present an irregular meshwork on the cell surface; this network is expected to reflect structures consisting of two cytoskeletons, microtubules, and gliofilaments. The morphological characterization of this meshwork is reported here for the first time for type 2 astrocytes. The present characterization is expected to aid further study of type 2 astrocytes.