Changes in intermediate fiiament content and extracellular matrix material showed that the injury response of ependymal cells in lesioned axolotl spinal cord involves an epithelial-to-mesenchymal transformation, and that fibrous astrocytes are excluded from the remodeling lesion site. Antibody localization was used to visualize cytokeratin-, vimentin-, and glial fibrillary acidic protein-(GFAP-) containing intermediate filaments, as well as the adhesive glycoprotein fibronectin. In normal axolotl spinal cord cytokeratins were found near the apical surface of the ependymal cells. Transmission electron microscopic examination suggested that these cytokeratins were in tonofilaments. Cytokeratin expression was lost and vimentin production was initiated in ependymal cells 23 weeks following spinal cord injury. There was a period of approximately 1-2 weeks when cytokeratins and vimentin were co-expressed in vivo. This co-expression was maintained in vitro by culture on a fibronectin-coated substratum. As the central canal reformed, vimentin expression was lost. Ependymal cells lacked GFAP intermediate filaments, but GFAP was present in fibrous astrocytes of the neuropil and white matter. Following injury, GFAP localization showed that fibrous astrocytes disappeared from the remodeling lesion site and reappeared only after the ependymal epithelium reformed and newly myelinated axons were found. Fibronectin expression closely followed the expression of vimentin during mesenchymal ependymal cell outgrowth. These results suggest that the ependymal cell outgrowth requires changes in cell shape followed by changes in production of extracellular matrix.

In urodele amphibian spinal cord regeneration, the ependymal cells lining the central canal remodel the lesion site to favor axonal regrowth. We profiled the production of matrix metalloproteinases by injury-reactive mesenchymal ependymal cells in vivo and in vitro and found that matrix metalloproteinases are involved in this remodeling process in the axolotl (Ambystoma mexicanum). The production of cell-associated matrix metalloproteinases in vivo was shown to be identical to that in our cultured ependymal cell model system. Activated and zymogen forms of matrix metalloproteinases were identified using zymography, chemical inhibitors of matrix metalloproteinases, and cleavage of propeptides by organomercurials. The principal cellular proteinases consisted of matrix metalloproteinase-2 (gelatinase A) and matrix metalloproteinase-1 (type I collagenase), which display characteristic shifts in molecular weight following proenzyme processing by organomercurials. In addition, ependymal cell conditioned medium contained secreted forms of the enzyme undetectable in situ. Matrix metalloproteinase-9 (gelatinase B) as well as matrix metalloproteinase-2 and matrix metalloproteinase-1 were secreted and casein substrate zymography showed the presence of a small amount of a very high molecular weight matrix metalloproteinase-3 (prostromelysin) secreted into the culture medium. Matrix metalloproteinases were still present at 4 weeks post-lesioning when the ependymal cells have just re-epithelialized, but decreased near the completion of regeneration (8 weeks post-lesioning). Zymography showed no detectable matrix metalloproteinases in unlesioned cord but the presence of tissue inhibitor of metalloproteinase-1 in intact cord was seen by Western blotting. This study shows that matrix metalloproteinases are associated with urodele spinal cord regeneration and validates the use of our ependymal cell tissue culture model system to evaluate ependymal cell behavior during spinal cord regeneration.

No abstract