Abnormal Reaction to Central Nervous System Injury in Mice Lacking Glial Fibrillary Acidic Protein and Vimentin

Mice with null mutations in the GFAP (Pekny et al., 1995) or vimentin (Colucci-Guyon et al., 1994) genes were crossed to generate double mutants. These mice reached adulthood and did not display any overt anatomical or behavioral defects. Genotypes were determined as described (Colucci-Guyon et al., 1994; Pekny et al., 1995). All the mice used in these studies were on C57BL/129 mixed genetic background and were maintained in a conventional animal facility.

Mice subjected to spinal cord or brain injury were killed shortly after the injury (2 or 3 d) or were allowed to survive for 2 or 3 wk. In one group of animals the spinal cord dorsal funiculus was transected at the upper thoracic level, and the injury was extended 5 mm rostrally by a longitudinal incision (Frisén et al., 1992). In other animals, a 27-G needle was inserted through the skull and into the frontal cerebral cortex. After both types of injury, wild-type, GFAP−/−, or vimentin−/− animals did not display any recognizable symptoms after any of the lesions. The following number of animals were used for the spinal cord injury studies (shown as number of animals allowed to survive 2 d, 2 wk): wild-type (n = 10, 5), GFAP−/− (n = 8, 4), vimentin−/− (n = 2, 2), and GFAP−/−vim−/− (n = 3, 3). For the brain injury, these animals were used (shown as number of animals allowed to survive 3 d, 3 wk): wild-type (n = 3, 3), GFAP−/− (n = 4, 3), vimentin−/− (n = 2, 2), and GFAP−/−vim−/− (n = 5, 3).

Tissues were fixed by transcardial perfusion of anesthetized mice with Tyrode's solution followed by 4% formaldehyde and 0.4% picric acid in PBS or immersion fixation in Bouin's fixative (75 ml of saturated picric acid, 5 ml of glacial acetic acid, 25 ml of 40% formaldehyde). Specimens were cryo-sectioned or embedded in paraffin and sectioned using a microtome. Sections from all animals were stained with hematoxylin and erythrosin/eosin (HE) using standard protocols. Immunohistochemistry with rabbit antiserum to nestin (no. 130, diluted 1:2,000; Dahlstrand et al., 1992; Tohyama et al., 1992) was performed as described (Frisén et al., 1995; Pekny et al., 1998b). Binding of primary antibodies was visualized with Cy2-conjugated donkey anti–rabbit antibodies (Jackson ImmunoResearch) or HRP-conjugated goat antiserum against rabbit immunoglobulins (Dako A/S). Peroxidase was detected with DAB (Dako A/S). Photomicrographs were scanned and mounted using Adobe Photoshop.

Nestin mRNA expression in the injured spinal cord was detected by in situ hybridization using digoxigenin labeled cRNA probes as described by Schaeren-Wiemers and Gerfin-Moser (1993). The animals were allowed to survive 4 d after the incision in the dorsal funiculus. The antisense probe was prepared with T3 RNA polymerase from a NotI linearized plasmid containing a mouse nestin cDNA corresponding to nucleotides 1630–2972 (according to GenBank accession number M34384) and the sense probe was generated with T7 polymerase from the same plasmid linearized with XhoI. No specific hybridization was seen with the sense probe.

Two mice of each genotype were used for the bromodeoxyuridine (BrdU) incorporation experiments. 1 h after the brain injury, which was performed as described above, the mice were injected i.p. with BrdU (100 μg/g body weight). The injections were repeated 17 times for 50 h; injections were applied every second hour with a 10-h injection-free window each night. The mice were anesthetized and transcardially perfused with 0.1 M phosphate buffer followed by 4% paraformaldehyde. Vibratome or frozen sections were used for either simultaneous or sequential detection of BrdU and S-100. For BrdU detection sections were treated with 2 M HCl for 30 min followed by 0.25% pepsin in 0.1 M HCl for 2 min at 37°C before incubation with the primary antibody. Mouse monoclonal antibody (M0744) and goat anti–mouse FITC-conjugated antibody (F0479; both Dako A/S) were used. For S-100 immunodetection, rabbit anti–cow S-100 antibody (Z311) was used followed by swine anti-rabbit TRITC-conjugated antibody (R0156; both Dako A/S).

The following genotypes and number of mice were used for ultrastructural analysis: wild-type (n = 4), GFAP−/− (n = 3), vimentin−/− (n = 2), and GFAP−/−vim−/− (n = 3). 3 d after a cortical stab wound, the mice were anesthetized and transcardially perfused with Tyrode's solution followed by 500 ml of a solution containing 2% paraformaldehyde and 2% glutaraldehyde dissolved in PBS. After postfixation overnight, the tissues were cut on an Oxford Vibratome and osmicated for 4 h, dehydrated in acetone, and embedded in Vestopal W. Ultra-thin sections (80–100 nm thick) were cut from injured brain cortex of all animals using an LKB Ultrotome. These sections measured 1.5 × 0.4 mm and included at one side the wound, covered the reaction zone, and displayed compact neuropile at the opposite side. The sections were picked up on Formvar-coated one-slot copper grids, contrasted with lead citrate (Reynolds, 1963) and uranyl acetate, and examined in a Philips EM 400 electron microscope.

Blood vessel diameters were measured in cross sections of thoracic spinal cord segments in a microscope fitted with a scale in the eyepiece. The smallest distance across an individual blood vessels was assigned as the diameter of the blood vessel. The distance from the dorsal spinal cord surface to the bottom of the indentation in the dorsal funiculus was considered as the depth of the indentation below the superficial dorsal vessel (dorsal spinal vein). The genotypes of the animals were unknown to the person performing the measurements. Blood vessel diameters and dorsal funiculus indentations were measured in at least six animals of each genotype and at least 4 sections were analyzed from each animal. The average number of blood vessels wider than 15 μm per section and the average depth of the indentation in the dorsal spinal cord was calculated for each animal, and were used to calculate the mean and standard error of the mean for animals of each genotype. Statistical significance was tested by Student's two-tailed t test. The differences were considered significant if the P values were lower than 0.01.

We first analyzed the uninjured nervous system of the mutant mice. Many blood vessels appeared dilated in the brain and spinal cord of GFAP−/−vim−/− mice (Fig. 1). The lumina of these structures were lined with von Willebrand factor–immunoreactive endothelial cells indicating that they were true blood vessels rather than, for example, cavities formed by degeneration (data not shown). Measurements of blood vessel diameters in these mice revealed an almost threefold increase in the number of blood vessels with a diameter >15 μm compared to wild-type mice (P < 0.0005; Fig. 2 A). We frequently noticed that there was an unusually deep indentation in the tissue of the dorsal spinal cord in many of the mutant mice. In wild-type mice, there is often an invagination below the dorsal spinal vein, but measurements of the depth of the invagination revealed a statistically significant increase in the depth compared to wild-type mice in GFAP−/− (P < 0.005), vimentin−/− (P < 0.0001), and GFAP−/−vim−/− mice (P < 0.0001; Figs. 2 B and 6, A–D).

Endothelial and ependymal cells in the adult CNS express both nestin (Dahlstrand et al., 1992; Frisén et al., 1995; Fig. 3 A) and vimentin (Franke et al., 1979), but not GFAP in wild-type mice. Nestin-immunoreactivity (IR) in these cells appeared normal in GFAP−/− mice (data not shown). However, in vimentin−/− mice, nestin-IR was reduced to very low levels in both endothelial and ependymal cells (Fig. 3 B). Moreover, nestin-IR did not display a filamentous pattern but was spread diffusely throughout the cytoplasm in these cells (Fig. 3 B). The same nestin-IR pattern was seen in GFAP−/−vim−/− mice (data not shown).

We next analyzed scar formation in wild-type and mutant mice in response to an incision in the spinal cord dorsal funiculus. The mice were allowed to survive 2 d or 2 wk after the injury, and tissue sections from the site of the injury and from a spinal cord segment 10 mm rostral to the lesion were analyzed. There were no consistent differences in the histology of the scar tissue between wild-type, GFAP−/−, or vimentin−/− mice revealed by HE staining (Fig. 4). However, the scar tissue was much less dense in the GFAP−/−vim−/− mice, and the scar tissue was interrupted by numerous fissures most often running in a dorso-ventral orientation (Fig. 4, H and L). These fissures were filled with blood, tissue fluid or debris (Fig. 4 L). Moreover, in the GFAP−/−vim−/− animals there was a large number of red blood cells within the scar tissue, both 2 d and 2 wk after the injury, indicating more pronounced bleeding and/or defective clearance of blood from the injury site (Fig. 4, H and L).

Nestin expression is rapidly induced in astrocytes after CNS injury and serves as a sensitive marker for reactive astrocytes (Clarke et al., 1994; Frisén et al., 1995; Lin et al., 1995). Nestin-IR was induced in response to spinal cord injury in mice of all genotypes (Fig. 5). However, nestin-IR was more restricted in vimentin−/− or GFAP−/−vim−/− mice compared to wild-type or GFAP−/− mice (Fig. 5). Whereas 2 d after the injury nestin-IR cells were seen at the site of the lesion and in the surrounding gray matter in wild-type (Frisén et al., 1995) or GFAP−/− mice, nestin-IR was less pronounced in the gray matter in vimentin−/− or GFAP−/−vim−/− mice (Fig. 5). Furthermore, 2 wk after the lesion the levels of nestin-IR in the scar were lower in vimentin−/− or GFAP−/−vim−/− mice compared to wild-type or GFAP−/− mice (Fig. 5). Whereas the scar tissue appeared dense after 2 wk in wild-type, GFAP−/− or vimentin−/− mice, the labeled cells were more sparse in GFAP−/−vim−/− mice and nestin-IR was diffusely spread throughout the cytoplasm of the cells (Fig. 5). Nestin-IR was induced within 2 d after the injury in segments rostral to the lesion, in the degenerating axonal tract, to a similar extent in mice of all genotypes (Fig. 6). However, 2 wk after the injury the level of nestin-IR in the degenerating tract had decreased substantially in the vimentin−/− or GFAP−/−vim−/− mice, whereas it remained strong in wild-type or GFAP−/− mice (Fig. 6).

To determine whether the reduced nestin-IR in the injured spinal cord was caused by reduced transcription of the nestin gene in vimentin−/− and GFAP−/−vim−/− mice, we performed in situ hybridization with nestin antisense riboprobes in spinal cord sections from animals that underwent a dorsal funiculus incision 4 d before. Nestin mRNA was detected in mice of all genotypes, and no reduction in nestin mRNA levels could be found in vimentin−/− and GFAP−/−vim−/− mice (Fig. 7). Scattered cells in the dorsal funiculus expressed nestin mRNA and their distribution was highly reminiscent of the nestin-IR pattern. No specific hybridization was seen with the sense probe (Fig. 7, E and F).

Glial scar formation was also analyzed in the brain. A cortical stab wound was done with a fine needle, resulting in a much more restricted injury than the spinal cord lesion. Interestingly, 3 out of 11 GFAP−/−vim−/− mice died shortly after the injury and the necropsy showed extensive intracranial bleeding that was interpreted as the cause of death (data not shown). In the group of GFAP−/−vim−/− mice that were killed 3 d after the injury, 3 out of 5 mice showed extensive bleeding at the site of injury (Fig. 8 D). Bleeding of comparable magnitude was not seen in mice of other genotypes. In all wild-type, GFAP−/− and vimentin−/− mice, as well as in the GFAP−/−vim−/− mice that survived the operation, the discrete injury caused by the needle did not cause any apparent clinical symptoms and was sealed within 3 wk (Fig. 8). Nestin-IR was detected around the cortical lesion 3 d after the injury in mice of all genotypes (Fig. 8). However, whereas nestin-IR was comparably strong and revealed distinct reactive astrocytes in both wild-type and GFAP−/− mice, it was weaker and diffuse in vimentin−/− and GFAP−/−vim−/− mice (Fig. 8, G and H). Nestin-IR was reduced to undetectable levels after 3 wk in mice of all genotypes (data not shown).

To further characterize scar formation in the mutant mice, we analyzed the zone immediately adjacent to the central necrotic area of cortical injury by electron microscopy. This area was similar in wild-type, GFAP−/−, or vimentin−/− mice, but differed in GFAP−/−vim−/− mice. The difference was most apparent in the border zone between the severely disarranged tissue close to the wound and the more distant, compact and normal looking brain tissue. Here, the GFAP−/−vim−/− mice exhibited fragmented tissue with a high accumulation of extracellular debris. This debris was diffuse, finely granular or filamentous of moderate electron density (Fig. 9 d). This was in contrast to mice of the other genotypes (Fig. 9, a–c) in which the brain tissue in the corresponding area showed easily identifiable components (e.g., myelinated and unmyelinated axons, synaptosome-like profiles), narrow extracellular spaces and virtually no extracellular debris.

To evaluate cell proliferation in the area affected by the cortical injury we have compared BrdU incorporation within a 50-h time window after the injury in wild-type, GFAP−/−, vimentin−/−, and GFAP−/−vim−/− mice. Using confocal microscopy we have counted BrdU-labeled cells within the volume of 5 × 10⁶ μm³ that included the injury area in the middle and was reconstructed from superimposed confocal images. The number of BrdU-labeled cells within this volume ranged from 64 to 87 and no statistically significant differences were found between wild-type and mutant mice. To determine the proportion of astrocytes among the BrdU-labeled cells, S-100 positive cells were counted within the same volume. The number of S-100 positive cells ranged from 65 to 83 and no statistically significant differences were found between wild-type and mutant mice. Fig. 10 provides a comparison between individual (Fig. 10, a–d) and combined (Fig. 10, e and f) BrdU and S-100 immunostaining of the cortical injury area in wild-type and GFAP−/−vim−/− mice. Both contain comparable numbers of BrdU positive, S-100 positive and double positive cells (some of these are indicated by arrows). Thus, the cortical injury in wild-type and mice deficient for GFAP and/or vimentin triggers a comparable induction of cell division.

The role of astrocytic IF proteins has been the subject of numerous studies, but has been difficult to establish. In this study we have utilized mice carrying null mutations in the GFAP and/or vimentin genes. We found that GFAP and vimentin have partly overlapping functions in the astrocytic response to injury, and contribute to scar formation. Furthermore, blood vessel dilation was seen in the brain and spinal cord of mice lacking both vimentin and GFAP.

In the nervous system of adult wild-type mice, nestin expression is largely restricted to endothelial and ependymal cells (Dahlstrand et al., 1992; Frisén et al., 1995). Both of these cell types also express vimentin (Franke et al., 1979; Schnitzer et al., 1981). The nestin expression pattern was distinctly altered in vimentin−/− or GFAP−/−vim−/− mice. In wild-type or GFAP−/− mice, nestin-IR in endothelial or ependymal cells exhibited a distinct pattern in contrast to the weak and diffuse nestin-IR seen in vimentin−/− or GFAP−/−vim−/− mice (Fig. 3 and data not shown). Analysis of reactive astrocytes in vivo and astrocytes in culture from GFAP−/−vim−/− mice revealed the lack of IFs in these cells in spite of nestin expression, demonstrating that nestin fails to form IFs on its own in astrocytes (Eliasson et al., unpublished observations). The in situ hybridization data did not reveal any reduction in nestin mRNA expression in the spinal cord of vimentin−/− or GFAP−/−vim−/− mice compared to wild-type. This implies that the differences detected at the level of nestin-IR are posttranslational, probably caused by the changed equilibrium between unpolymerized and polymerized nestin in a situation when nestin polymerization cannot occur. This is further supported by the results from the Northern blot analysis of the RNA from astrocytes in vitro that showed rather an increase in nestin transcription in vimentin−/− and GFAP−/−vim−/− astrocytes (Eliasson et al., unpublished observations). Two-hybrid analysis has further indicated that nestin cannot bind to itself and nestin also fails to form IFs in SW13 cells devoid of other IF proteins, but incorporates into IFs once vimentin or desmin is transfected into these cells (Sjöberg, G., and T. Sejersen, personal communication; Marvin et al., 1998). These data suggest that the altered nestin-IR pattern observed in endothelial and ependymal cells in vimentin−/− or GFAP−/−vim−/− mice reflects the failure of nestin to form IFs in the absence of other IF proteins also in these cells.

Many blood vessels were dilated in the brain and spinal cord in GFAP−/−vim−/− mice. Vimentin, but not GFAP, is normally expressed both in endothelial cells and in perivascular cells such as pericytes and smooth muscle cells (Franke et al., 1979; Gabbiani et al., 1981). A reduction or lack of IFs in endothelial and perivascular cells may decrease the mechanical strength of blood vessels and consequently their capacity to withstand dilation. The observed increase in blood vessel diameter in the CNS of GFAP−/−vim−/− mice may be caused by altered properties of endothelial or perivascular cells (as a consequence of vimentin absence in these cells) and a reduced capacity of the surrounding tissue to withstand blood vessel dilation (the consequence of absence of GFAP and vimentin in astroglial cells). Such a concept is supported by the finding of a progressively deeper indentation beneath the dorsal spinal vein in GFAP−/−, vimentin−/−, and in GFAP−/−vim−/− mice, respectively.

Scar formation after a spinal cord or brain injury appeared comparable in wild-type, GFAP−/−, or vimentin−/− mice, but was defective in mice lacking both GFAP and vimentin, indicating that these proteins may have partly overlapping functions in this process. In such mice, the glial scar was less compact, contained more debris compared to the mice of the other genotypes, and the scar tissue was disrupted by fissures. Using BrdU incorporation, we did not detect any differences in cell proliferation in the injury area. Also, the amount of astrocytes, detected as S-100 positive cells, in the injury area was comparable between the mice of different genotypes. These results imply that even in GFAP−/−vim−/− mice, the injury results in a normal response at the level of cell proliferation.

Major similarities exist between the epithelial response to injury and reactive gliosis in the CNS. In both cases, production of IF proteins accompanies cell activation and the process results in the closure of the wound. In case of stratified epithelia, such as the epidermis of the skin, the wound closure happens in an orchestrated action of keratinocyte migration from the surrounding healthy tissue and a contraction of the connective wound bed mediated by migrating fibroblasts (for review see McGowan and Coulombe, 1998). Keratinocyte activation and migration parallel induction of the wound repair-associated IF proteins in these cells, namely keratins 6, 16, and 17, which happens within 6–12 h after injury (Paladini et al., 1996). In the injured CNS, proliferation of stem cells is induced by injury, and the progeny of the stem cells migrate towards the injury where they differentiate to astrocytes (Johansson et al., 1999). Thus, cell migration and wound contraction may be phenomena playing a major role in wound closure in general. Our preliminary experiments suggest that in the scrape wounding assay in vitro, cultures of GFAP−/−vim−/− astrocytes require longer time to close the defect (unpublished data). Therefore, it is possible that reactive astrocytes without IFs exhibit a cell migration deficit. To address the contractile properties of astrocytes from wild-type and GFAP and/or vimentin deficient mice, we investigated whether primary cultured astrocytes from these mice can contract collagen gels. All type of astrocytes were able to contract the gels and no quantitative differences between wild-type, GFAP−/−, vimentin−/−, and GFAP−/−vim−/− astrocytes were detected (data not shown).

The defective scar formation in GFAP−/−vim−/− mice may not only be a result of astrocytic dysfunction but may also relate to the abnormal blood vessels seen in these mice. It is unlikely that the blood vessel abnormality is the major reason for defective scar formation in these mice, since scar formation appeared normal in vimentin−/− mice and only vimentin, and not GFAP, is expressed in cells of the blood vessel. The bleeding seen in GFAP−/− vim−/− mice therefore probably results from the combination of abnormal blood vessels and defective closure of the wound.

The amount of debris was clearly elevated in glial scar tissue of GFAP−/−vim−/− mice compared to mice of the other genotypes. This can either be a result of defective clearance of tissue debris or its increased production. Astrocytes are capable of phagocytosis and the abundance of debris in GFAP−/−vim−/− mice may result from impairment of this function. Alternatively, the bleeding observed in GFAP−/−vim−/− mice may result in the formation of more debris in the forming glial scar and its immediate vicinity.

Multiple fissures traversed through the scar in the injured spinal cord of GFAP−/−vim−/− mice. This may be a consequence of reduced astrocytic process formation with deficient bridging and sealing of the wound. Absence of IFs in astrocytes is likely to reduce the tensile strength of individual astrocytes, and as a consequence, of the tissue. A low resistance to mechanical forces is likely to inhibit scar formation, especially in the spinal cord that has to adapt to movements of the vertebral column, a situation distinct from that in the skull-enclosed brain. Irrespective of its cause, defective scar formation may lead to repeated tearing of the tissue and recurring bleeding, potentially explaining the abundance of red blood cells in the scar tissue of GFAP−/−vim−/− mice. The defective glial scar formation in GFAP−/−vim−/− mice described here reveals the importance of GFAP and vimentin for this process as well as a certain degree of functional overlap between these two proteins.

We wish to thank Dr. Kristofer Rubin for performing the collagen gel contraction assays, Dr. Charles Babinet for providing the vimentin null mice, Dr. Martha Marvin for the gift of nestin cDNA, Dr. Charles Babinet and Dr. Alain Privat for valuable discussions, Ms. Ulrika Wilhemsson for genotyping some of the animals, and Ms. Marianne Eriksson and Ms. Rita Grandér for technical help with processing the tissue for electron microscopy.

This study was supported by grants from the Swedish Medical Research Council (project nos. 11548, 09041, 12183, and 03157), the Swedish Cancer Foundation (project no. 3622), the King Gustav V Foundation, the Swedish Society for Medicine, the Swedish Society for Medical Research, A⁺ Science Invest, Stiftelsen Sigurd och Elsa Goljes minne, Stiftelsen Lars Hiertas Minne, Björklunds fond, Jeanssons stiftelse, Ostermans stiftelse, Magnus Bergvalls stiftelse, Marcus Borgströms stiftelse, Tore Nilssons stiftelse, the Wenner-Gren foundations, Åke Wibergs stiftelse, Kapten Arthur Erikssons fond, Inga-Britt and Arne Lundberg foundation, and Gustafsson foundation. M. Pekny was supported by a fellowship from the Swedish Society for Medical Research.

BrdU

bromodeoxyuridine

CNS

central nervous system

GFAP

glial fibrillary acidic protein

HE

hematoxylin and erythrosin/eosin

IF

intermediate filament

IR

immunoreactivity