Address correspondence and reprint requests to Dr Milos Pekny, Department of Medical Biochemistry, University of Gothenburg, Box 440, SE-405 30 Göteborg, Sweden. E-mail: Milos.Pekny@medkem.gu.se
Intermediate filaments (IFs) compose, together with actin filaments and microtubules, the cytoskeleton and they exhibit a remarkable but still enigmatic cell-type specificity. In a number of cell types, IFs seem to be instrumental in the maintenance of the mechanical integrity of cells and tissues. The function of IFs in astrocytes has so far remained elusive. We have recently reported that glial scar formation following brain or spinal cord injury is impaired in mice deficient in glial fibrillary acidic protein and vimentin. These mice lack IFs in reactive astrocytes that are normally pivotal in the wound repair process. Here we show that reactive astrocytes devoid of IFs exhibit clear morphological changes and profound defects in cell motility thereby revealing a novel function for IFs.
Migration of cells involves cell-substratum adhesion via specific receptors, transmitting forces and signals necessary for locomotion. Cell-substratum receptors of importance for cell motility include the integrins (Tooney et al. 1993). Both a low and a high affinity between cell and substratum inhibit cell motility, the maximal speed being obtained at an intermediate binding affinity (Palecek et al. 1997). The basic engine for movement of cells is the actin cytoskeleton that assists in the protrusion of lamellipodia, and provides the contractile forces necessary for cell displacement (Lee et al. 1993; Mitchison and Cramer 1996). Actin function is regulated by the small GTPases of the Rho family (Rho, Rac and Cdc42), which seem to be of major importance in the control of cell locomotion, inducing the formation of lamellipodia, filopodia and stress fibers by modulating actin polymerization and distribution (Tapon and Hall 1997). The MAP kinases (ERK1 and ERK2) influence the motile machinery by phosphorylating, and thereby increasing, myosin light-chain kinase activity leading to a subsequent phosphorylation of myosin light chains, which in turn promote the acto-myosin contraction (Klemke et al. 1997). Microtubules are of importance in maintaining cell polarity, movement of chromosomes during interphase and, in some cases, in cellular motility (Vasiliev 1991), although certain cells devoid of microtubules (e.g. keratinocytes or neutrophils) are capable of moving rapidly (Schliwa and Höner 1993). Recent data showing the role of plakins as integrators of the cytoskeleton (Yang et al. 1996, 1999) support the scenario of a co-operative action between individual cytoskeletal components, and strengthens the importance of assessing the role of non-actin cytoskeleton in activities such as cell locomotion.
The role of intermediate filaments (IFs) in the regulation of cell shape and motile behaviour has remained elusive. Inactivation of the vimentin gene by gene targeting revealed no obvious differences between wild-type and vimentin-deficient mouse tissues or organs (Colucci-Guyon et al. 1994). In vitro studies have given contradictory results. On one hand no obvious effects of tetracycline-regulated vimentin expression were observed in fibroblast lines derived from vimentin-null mice, as regards the distribution of microfilaments or microtubules, the shape of the nucleus, the directed migration as evaluated by a wound scrape assay, or resistance to mechanical trauma (Holwell et al. 1997). On the other hand, Eckes et al. (1998) reported that fibroblasts from vimentin–/– (vim–/–) mice in primary culture displayed an impaired mechanical stability and a reduced migration in both wound scrape and Boyden chamber assays as compared with fibroblasts derived from wild-type mice. Using primary cultures of astrocytes from glial fibrillary acidic protein (GFAP)–/– mice, it was found that even though astrocytes extended processes in response to neuronal contacts, the ability to introduce blood–brain barrier properties in endothelial cells seemed compromised (Pekny et al. 1998a,b).
Astrocytes in vitro, similarly to reactive astrocytes in vivo, produce three IF proteins, GFAP, vimentin and nestin; we have recently shown that in such astrocytes the absence of both GFAP and vimentin completely abolishes the IF formation, whereas the absence of GFAP or vimentin results in a partial IF deficit (Eliasson et al. 1999). Here we have used mice carrying mutations in the GFAP and vimentin genes separately (Colucci-Guyon et al. 1994; Pekny et al. 1995) and in combination (Eliasson et al. 1999; Pekny et al. 1999b) in order to investigate the effects on astrocytic morphology and motility in vitro. We show that astrocytes devoid of IFs exhibit changes in cell morphology, and that their motility is severely compromised. The impaired motility of astrocytes devoid of IFs may explain the impaired formation of glial scars following brain or spinal cord injury observed in GFAP–/–vim–/– mice (Pekny et al. 1999b).
Materials and methods
Animals and cell cultures
Primary astrocyte cultures were prepared from brains from newborn (postnatal day 0.5–1.5) wild-type, GFAP deficient (GFAP–/–), vimentin deficient (vim–/–) or GFAP and vimentin deficient (GFAP–/–vim–/–) mice as described previously (Pekny et al. 1998a,b). Five animals were included in each group. Cells were grown in Dulbecco's modified Eagle's medium (DMEM) with 2 mm GlutaMAXTM (Gibco BRL, BRL Gaithersburg, MD, USA), supplemented with 20% (v/v) heat-inactivated fetal bovine serum (FBS; Gibco BRL), 100 U/mL penicillin, 100 µg/mL streptomycin and 2.5 µg/mL fungizone (Amphotericin B, Gibco BRL). Cells were dislodged with a 0.05% trypsin/0.02% EDTA solution (Gibco BRL), seeded in a 6-well tissue culture plate (Nunc, Roskilde, Denmark) at a density of 3000 cells/cm2 and grown in a humidified atmosphere of 5% CO2 at 37°C for 16–24 h before video-recording.
For the ultrastructural evaluation of IFs, 10-day-old primary astrocytic cultures were used. Medium was removed and the cultures in 24-well Falcon plates were rinsed three times with isotonic phosphate buffer. The wells were filled with fixative consisting of a mixture of 2% paraformaldehyde and 2% glutaraldehyde dissolved in an isotonic phosphate buffer. The cultures were kept in the fixative overnight at 6°C, osmicated for 2 h, treated en bloc for 2 h with 2% uranyl acetate dissolved in 50% ethanol in water, dehydrated in a graded series of ethanol and embedded, still in the wells, in Epon. After polymerization the specimens were freed from the wells and sectioned on an LKB Ultrotome, picked up on Formvar-coated one-hole copper grids, contrasted first with uranyl acetate and then with lead citrate and examined in a Philips EM 400.
Time-lapse video-recording was performed using a computer-assisted microscope work-station consisting of a Nikon Diaphot 300 inverted microscope (Nikon, Copenhagen, Denmark) equipped with phase-contrast optics and a thermostat-controlled plexiglas incubator. A computer-controlled movable stage was mounted on the microscope allowing simultaneous recordings from several microscopic fields. Video-recordings were performed using a black and white CCD video camera (Burle, Lancaster, PA, USA) attached to the microscope producing 512 × 512-pixel images. Before recording, lids on culture plates were sealed tightly with adhesive tape and plates were placed on the microscope stage for approximately 30 min for temperature equilibration. All recordings were performed at 37°C. Images from 20 microscopic fields per cell culture were recorded at intervals of 15 min for a period of 4 h using Prima software (Protein Laboratory, Copenhagen, Denmark).
Evaluation of individual cell morphology and motility was carried out using Prima software. Contours of live cells were recognized semi-automatically by means of thresholding and binary transformation of the recorded images. For motility determinations, manual marking of the centres of cell nuclei was performed. From each animal, 90–170 cells in a cell culture were outlined and tracked. In order to exclude the possible influence of cell–cell interactions on cell behaviour, only single cells were evaluated.
Evaluation of cell morphology
Cell contours were used for determination of mean cell area, form factor, process domain and process index. The form factor is defined as 4π × area/perimeter2. Thus, perfectly round cells have a form factor of 1, whereas more elongated or stellate cells have lower form factors (Soll et al. 1988). Process index has been demonstrated to correlate with the number of cellular processes and is defined as the number of areas contained outside of the cell contour but within the cellular convex hull (Kawa et al. 1998). Process domain has been shown to correlate with process length and is defined as the difference between the area of the cellular convex hull and the cellular area (Kawa et al. 1998).
Evaluation of cell motility
The track of an individual cell was defined as a sequence of positions of its nucleus at different times. The dispersion of a cell is the Euclidean distance between two points on a plane measured in µm. The dispersion of a single cell after a given time of observation was calculated as
where x(·) and y(·) are the x- and y-coordinates of the cell, respectively, and tobs is the time during which a given displacement dtobstakes place. The rate of diffusion, R was estimated by plotting the mean squared displacement, 〈d2〉, against time with subsequent curve fitting to the equation given below
where ti is the time interval of interest, in this case 15 min. The fitting of this curve simultaneously permits an estimate of P, the persistence time (Gail and Boone 1970; Dunn 1983).
The average speed of the individual cells (mean cell speed) was calculated as the mean displacement of each cell for several identical time intervals with different starting points. This was performed according to the equation
where k denotes a given cell, N is the size of the investigated sample of a population of cells, x(·) and y(·) are the x- and y-coordinates of the cell, ttot is the time elapsed from the first to the last image constituting a recording and τ is the time interval between discrete observations.
The mean path length, 〈L〉, for a sample of a population of cells at a given time of observation, tobs, was calculated as
The locomotive index, LI, was calculated as the relationship LI = 〈d/L〉, and used as a measure of cellular persistence in a certain direction (i.e. the tendency of a cell to move in approximately the same direction for a certain amount of time).
Statistics and graphical presentations
Statistics and graphical presentations were performed using Prima (Protein Laboratory, Copenhagen, Denmark), Statistica v4.5 (StatSoft Inc., Tulsa, USA) and Origin v5.0 (Microcal, Northampton, MA, USA) programs.
Ultrastructural examination of IFs in GFAP–/–, vimentin–/– and GFAP–/–vimentin–/– astrocytes in primary cultures
Examination of wild-type, GFAP–/–, vim–/– and GFAP–/–vim–/– astrocytes in vitro showed reduced quantities of IFs in GFAP–/– astrocytes, unusually dense IF bundles in vim–/– astrocytes, and a complete absence of IFs in GFAP–/–vim–/– astrocytes (Fig. 1 and Eliasson et al. 1999). As we have recently shown, this reflects the manner in which GFAP, vimentin and nestin assemble to form IFs in reactive astrocytes in vitro and in vivo: IFs are formed of GFAP, vimentin and nestin in wild-type astrocytes, whereas in GFAP–/– and vim–/– astrocytes they are composed of vimentin and nestin, and of GFAP only, respectively (Eliasson et al. 1999). Nestin can neither self-assemble into IFs nor can it assemble with GFAP, and consequently, it does not form IFs in vim–/– or GFAP–/–vim–/– astrocytes (Eliasson et al. 1999). Thus, three model systems could be used to study the role of IFs on astrocyte morphology and motility representing (i) reduced quantities of IFs, (ii) abnormally bundling IFs and (iii) the complete absence of IFs, respectively.
Effects of IF deficiency on astrocyte morphology
The GFAP–/– or vim–/– astrocytes did not differ morphologically from wild-type cells (data not shown). However, GFAP–/–vim–/– astrocytes exhibited a profoundly altered morphology. The morphological appearance of cultured primary astrocytes from wild-ype and GFAP–/–vim–/– mice is shown in Fig. 2(a). The GFAP–/–vim–/– astrocytes were smaller and exhibited shorter processes than did the wild-type astrocytes.
In order to quantify the morphological differences between wild-type and GFAP–/–vim–/– astrocytes, 90–170 cells/animal were used to calculate the mean cell area, form factor, process domain and process index. Figure 2(b) shows that the mean cell area of GFAP–/–vim–/– astrocytes was smaller (∼ 35%) than that of the wild type. The shape of GFAP–/–vim–/– astrocytes also changed compared with the wild-type cells as reflected by an increase in form factor and a decrease in process index and process domain. As form factor, process index and process domain describe roundness, number of processes and length of processes, respectively, the observed changes indicate that astrocytes from GFAP–/–vim–/– mice were more rounded and had fewer and shorter processes than astrocytes from wild-type mice. Immunodetection of actin filaments revealed substantial heterogeneity among the cultures, but no difference between wild-type and GFAP–/–vim–/– astrocytes was detected (data not shown).
Effects of IF deficiency on astrocyte motility
In order to test the effect of the absence of GFAP and/or vimentin on astrocytic motility, images were recorded with 15-min intervals for a total recording time of 4 h. The positions of the centres of astrocytic nuclei from consecutive video frames were connected in order to generate cell tracks. In Fig. 3(a), the tracks of cells from representative animals from each of the four genotypes are shown as composite diagrams. Each diagram represents the tracks of 100–120 cells with the starting points of the tracks being superimposed. Each track demonstrates the motile behaviour of a single cell recorded for 4 h, and the radius of the circle in each diagram is the square root of the mean-squared displacement ( ) of all recorded cells from the investigated animal during the period of observation. These diagrams indicated a reduction in the motile behaviour of GFAP–/– or vim–/– astrocytes, and this was even more evident for the GFAP–/–vim–/– cells.
To examine the effect of the absence of GFAP and/or vimentin on motility of astrocytes in more detail, the data were expressed as the mean-squared displacement of the cells, <d2>, at various times of observation (Fig. 3b). The <d2> values at the final time point of 4 h were 1030.5 ± 132.26, 836.34 ± 82.83 and 854.38 ± 156.93 µm2 for wild-type, GFAP–/– and vim–/– astrocytes, respectively (no statistically significant difference, Fig. 2b). In contrast, the simultaneous absence of both GFAP and vimentin resulted in a reduction of cell motility, as compared with wild type astrocytes, <d2> being 506.32 ± 87.5 µm2; p < 0.02.
As shown in Fig. 4(a), the rate of diffusion of astrocytes from GFAP–/– or vim–/– mice did not differ from that of wild-type astrocytes, median values being 5.969, 4.719 and 5.631 µm2/min for wild-type, GFAP–/– and vim–/–, respectively. In contrast, GFAP–/–vim–/– astrocytes exhibited a decrease in the rate of diffusion (median value 2.385 µm2/min) as compared with wild-type astrocytes (p < 0.05).
A reduction in both the mean-squared displacement <d2> and the rate of diffusion R in GFAP–/–vim–/– astrocytes may reflect a change in either cell speed or cellular persistence in direction. The latter was estimated by determination of the locomotive index of the astrocytes from each group as shown in Fig. 4(b). No differences between any of the four groups were found, implying that the reduced rate of diffusion observed for the double knock-outs was not caused by a reduction in persistence time.
Subsequently, the mean cell speed Sti of the astrocytes of all experimental groups was analysed. Individual values of the mean cell speed of 540–560 cells of each genotype were determined. For each set of data, histograms with increments of 0.02 µm/min were created. In Fig. 5(a), it can be seen that wild-type astrocytes judged by the Sti value constituted the most heterogeneous cell population compared with GFAP- and/or vimentin-deficient astrocytes. The removal of the IF proteins clearly led to an altered, more narrow distribution of individual mean-cell speed in both single and double mutants.
Subsequently, medians of the Sti distributions (Sti 50) were calculated for each animal (90–170 cells) and grouped by genotypes, see Fig. 5(b). The differences in Sti 50 values between wild-type and GFAP–/– or vim–/– astrocytes were statistically significant (p < 0.01 and p < 0.05, respectively). The GFAP–/–vim–/– cells exhibited a further reduction in the Sti 50 value as compared with either wild-type, GFAP–/– or vim–/– astrocytes (p < 0.01 in all cases).
Thus, the absence of GFAP and vimentin and consequently of IFs from astrocytes in culture, profoundly affected astrocyte morphology and motility. Astrocytes deficient in both GFAP and vimentin were smaller, more rounded in shape and had fewer and shorter processes. The GFAP–/–vim–/– astrocytes also displayed a significantly lower rate of diffusion and mean cell speed than did single mutants and wild-type astrocytes.
In reactive astrocytes in vitro or in vivo, the absence of GFAP or vimentin leads to partial deficit in the IFs, whereas the absence of both of these proteins prevents the formation of any IFs (Eliasson et al. 1999). Using primary cultures of reactive astrocytes from mutant mice lacking GFAP and/or vimentin we have addressed the effect of the absence of IFs on astrocyte morphology and motility.
Morphology of GFAP–/–vimentin–/– astrocytes
Lack of expression of either GFAP or vimentin had no effect on astrocytic morphology, whereas the absence of both IF proteins influenced cell shape profoundly. Astrocytes from GFAP–/–vim–/– mice were smaller, more rounded and extended fewer and shorter processes as compared with wild-type astrocytes, indicating that IFs are directly involved in the regulation of cellular process extention, probably in co-operation with other cytoskeletal components (Schliwa and Höner 1993). As shown before, IF-free astrocytes still form cell processes both in vivo and in vitro (Pekny et al. 1995, 1998a). However, this quantitative study shows a clear effect of the IF absence on astrocyte morphology, and this may potentially explain some of the defects exhibited by the IF-deficient astrocytes in vivo and in vitro.
In the CNS, astrocytes play many important roles, such as the induction of the blood–brain barrier, recycling of neurotransmitters, or the repair of the injured brain or spinal cord. We have previously reported that in all these situations, the astrocytes deficient in IFs are functionally affected (for a review see Pekny 2001). Thus, GFAP–/– astrocytes are not capable of inducing blood–brain barrier properties in the dynamic blood–brain barrier model system in vitro (Pekny et al. 1998b). Similarly, the absence of GFAP in astrocytes leads to an increase in intracellular levels of glutamine, the conversion product of glutamate, a major neurotransmitter in the CNS (Pekny et al. 1999a). Finally, glial scar formation following brain or spinal cord injury is impaired in GFAP–/–vim–/– mice (Pekny et al. 1999b). Although these functional differences may reflect more specific roles of IFs in the respective processes, they can also be a direct consequence of a decrease in cell size and in number or length of astrocytic processes. It is easy to imagine that this would limit the cell surface area available for cell–cell interactions, as well as the average volume of the space controlled by an individual astrocyte. This would result, for example, in the suppression of astrocyte–endothelial cell interactions required for the induction and maintenance of the blood–brain barrier or for its reconstitution after CNS injury. The medical relevance of GFAP has recently been further strengthened by identification of mutations in the GFAP gene as the cause of Alexander disease, a rare and fatal leucoencephalopathy (Brenner et al. 2001).
During the development of the CNS, migrating glial cells contribute to the creation of the very organized layered structure of the brain and of the spinal cord. Similarly, post-traumatic CNS regeneration involves migration of astrocytes to the area of the lesion, with subsequent formation of a dense glial scar. It has been shown that following spinal cord injury, even in the adult, astrocytes are born from the ependymal cells of the central canal and migrate over considerable distances to the injured area (Johansson et al. 1999). The actin cytoskeleton seems to play a pivotal role in the coordinated events leading to cell relocation, although the participation of other cytoskeletal components is less well understood. Our data indicate that lack of expression of either GFAP or vimentin inhibits the mean cell speed of astrocytes, whereas the persistence in direction remains unchanged. Removal of both proteins, leading to a complete absence of IFs in astrocytes, induces a further decrease of the mean cell speed, still without any effect on persistence time. Thus, astrocytes with compromised or absent IFs change their motile behaviour, indicating that IFs are an integral part of the motile machinery of the cell. Interestingly, the directionality of movement is not affected even in the complete absence of astrocytic IFs, which indicates that the cellular events determining the ‘choice’ of direction of movement are probably independent of IFs. Interestingly enough, the CNS development in the GFAP–/–vim–/– mice does not exhibit any apparent abnormalities that might be expected as a consequence of deficient migration of glia and their progenitors. This may reflect a limited role of IFs on migration of immature glia during development, or successful compensation by other systems. A temporary delay in the CNS morphogenesis, undetectable in the adult GFAP–/–vim–/– mice, cannot be excluded and more detailed developmental studies remain to be performed.
The finding that IFs are required for normal cell motility has important implications. First, it provides a potential mechanistic explanation of the glial scar formation defect observed in GFAP–/–vim–/– mice following brain or spinal cord injury (Pekny et al. 1999b), and for impaired wound healing seen in vim–/– mice (Eckes et al. 2000). In GFAP–/–vim–/– mice, the astrocytes of which are devoid of IFs even in their reactive state, the glial scar development is slowed down, with the scar region containing a number of fissures and tissue debris (Pekny et al. 1999b). Impaired long- and short-distance migration of astrocytes may be at least partially responsible for this outcome. If so, modulating the production of astrocytic IFs can ultimately become a worthy approach in situations in which the interference with astrocyte motility is desired, such as the CNS injury. Moreover, the role of IFs in cell motility may apply also to other IF members beside the astrocytic IFs. Reactive gliosis in the CNS shares many similarities with the response of epithelial cells or fibroblasts to injury. In the epidermis of the skin, keratinocytes are known to migrate into the affected area from the surrounding normal tissue (McGowan and Coulombe 1998). This process involves interaction with fibroblasts in the underlying dermis and is accompanied by the expression of unique, wound repair associated IF proteins, namely keratins 6, 16 and 17 (Paladini et al. 1996). It is thus possible that a unique set of activation status-specific IF proteins is contributing to the ability of different cell types to migrate either during development or in the repair process. Such a scenario, though at the moment hypothetical, would provide a very elegant and fine control of the degree of the cell motility depending on the developmental phase and the cell activation status. It could also offer a new explanation – other than different mechanical demands imposed on individual cell types – for the puzzling diversity of IF proteins.
In conclusion, the reduction of the number and length of the cellular processes and the down-regulation of the cellular speed in astrocytes with absent IFs indicate that IFs (probably, in cooperation with an other part of the cytoskeleton) play a role in both process protrusion and cell locomotion.
This study was supported by grants from the Swedish Cancer Foundation (project no. 3622), the Swedish Medical Research Council (project nos 11548 and 03157), the Swedish Society for Medicine, the Swedish Society for Medical Research, the King Gustaf V Foundation, the Volvo Assar Gabrielsson Fond, the Swedish Stroke Foundation, the Sigurd and Elsa Goljes Foundation, the EU Biotechnology Programme (contract BIO4-CT98-0077), the Lundbeck Foundation and Danish Medical Research Council.
We wish to thank Ms Ulrika Wilhelmsson for genotyping some of the animals and Ms Rita Grandér for technical help with processing the cells for electron microscopy.