Neuronal maintenance and neurite extension of adult mouse neurones in non-neuronal cell-reduced cultures is dependent on substratum coating

Adult mammalian neurones have generally proved more difficult to isolate and maintain in vitm (Scott, 1977; Fukuda, 1985; Unsicker et al. 1985; Smith & Mclnnes, 1986), than have embryonic or avian sensory neurones (e.g. see Bunge et al. 1967; Rogers et al. 1983; Bray et al. 1987; Gundersen, 1987). We recently demonstrated, however, that adult mouse neurones would survive in vitro, even when non-neuronal cell (NNC) proliferation was prevented by addition of cytosine arabinoside (AraC) to the medium, providing that the substratum was adequately coated : the glycoprotein fibronectin (FN) was particularly beneficial in this respect over a 14-day period, and a microexudate derived from mouse hepatocytes was also effective in the short term at least (Smith & Orr, 1987).

FN and laminin (LAM) are both present in the basal lamina and endoneurium of sensory ganglia (Bannerman et al. 1986), and in sciatic nerve segments, in which Schwann cells were shown to produce laminin (Cornbrooks et al. 1983). A putative cell surface receptor for FN and LAM has been identified on the perikarya, axons and growth cones of embryonic chick dorsal root ganglion (DRG) cells in vitm and implicated in adhesion (Bozyczko & Horwitz, 1986). LAM and FN have been found to increase neuronal adhesion and to promote neurite extension and guidance in many studies of cultured chick sensory neurones (Rogers et al. 1983; Roufa et al. 1983; Hammarback et al. 1985, 1988; Bray et al 1987; Gundersen, 1987), and cultured neonatal rat sympathetic neurones (Hawrot, 1980; Lander et al. 1985). With human foetal neurones, although neurite extension was enhanced by separate substrata of both FN and LAM, it was more pronounced on laminin (Baronvan Evercooren et al. 1985). Fridman et al. (1985) produced three-dimensional extracellular matrices from two sources (corneal endothelial cells and embryonic fibroblasts) and cultured ciliary ganglion neurones upon these. The cells responded differentially to the two matrices, suggesting that subtle changes in composition and organization could affect survival and differentiation of neurones.

In view of these previous studies for primarily embryonic neurones, the present work aimed to investigate in detail: (1) whether laminin in addition to fibronectin enhanced the maintenance of adult sensor neurones in NNC-reduced cultures over a 14-day culture period; (2) whether neurone survival and neurite extension were significantly improved if both glycoproteins were present simultaneously; and (3) whether a fibroblast-derived microexudate proved more effective than hepatocyte-derived material in the long-term survival of adult neurones. We consider it is important to extend the maintenance of neurones in culture to include cells taken from adult, rather than merely embryonic, sources, since we believe a better understanding of the requirements of adult neurones will lead to the establishment of a system pertinent to the study of nerve regeneration following trauma. A scanning electron-microscopic (SEM) study was undertaken in order that the numbers of neurites and the extent of their regrowth could be determined more accurately than is possible by our previous phase-contrast microscopic studies.

Preliminary reports of parts of this study have been presented (Orr, 1988; Smith & Orr, 1988).

Neurones were prepared as described (Smith & Orr, 1987). Briefly, 25–40 dorsal root ganglia were excised from adult (3–6 months) CB A male mice and incubated overnight with 0·125% collagenase, followed by gentle mechanical dispersion. The cells released by this procedure, together with any small fragments of ganglia remaining as aggregates, were then plated at a density of 15000–20000 neuronesml⁻¹. Cultures were maintained in standard Dulbecco’s medium supplemented with 10gI⁻¹ glucose, 10% foetal calf serum (depleted of fibronectin in some experiments), and 0·1% gentamycin, and incubated at 37 °C in a humidified atmosphere of 95% air/5% CO2. Where non-neuronal cell (NNC) proliferation was to be inhibited, 5×10⁻⁶ M-cytosine arabinoside (AraC) was added to the medium from the onset of culture. Media were changed at 2-day intervals and cultures monitored for 14 days in litro (d.i.v.).

In addition to the uncoated plastic of either Nunclon wells or Thermanox coverslips, coated substrata were prepared as described by Smith & Orr (1987) with minor modifications. Surfaces were covered with solutions of 10^gml⁻¹ fibronectin (FN) or 5-lOjzgml⁻¹ polylysine (both from Sigma Chemical Co, Poole, UK), air dried and ultraviolet (u.v.)-irradiated by a laminar flow cabinet light source for a period of 30 min, reduced to minimize possible changes in the coating (although Hammar-back et al. (1985) found no lowering of adhesivity of neurones on u.v.-irradiated FN coatings). As inactivation of LAM by u.v. irradiation has however been reported (Hammarback et al. 1985), this step was omitted; and instead these coatings were prepared by a 45 min contact with 10 μgml⁻¹ LAM solutions (Collaborative Research Inc., Bedford, USA) prior to the addition of culture medium. FN/LAM coverslips were coated first with FN and then with LAM.

Microexudates were prepared from monolayers of mouse hepatocytes obtained from perfused adult livers (Smith, 1984); the substratum-attached material was air dried after removal of the cells with 5 mM-EDTA for 30 min (Collins, 1980); this gave more consistent surfaces than the mechanical scraping method described previously (Smith & Orr, 1987). Microexudate coatings were also produced by similar procedures from confluent cultures of 3T6 cells (Flow Labs).

Exudate-substrata were screened by SEM to identify and monitor the attached material. Polyacrylamide gel electrophoresis was performed on fibroblast exudates to analyse their protein content using the Pharmacia PhastSystem (kindly loaned by Pharmacia Ltd, Milton Keynes, England). Solubilized exudate coating was sampled together with commercial controls of known molecular weights (M_r, ranging from albumin at 67×10³ to thyroglobulin at 670× 10³) and samples of the FN and LAM used to prepare substrata in this study. Gels were stained with Coomassie Blue and using the silver-staining kit supplied by Pharmacia.

Cultures were observed by phase-contrast microscopy for 14d.i.v. Isolated neurones and NNCs, free from any nondissociated ganglion fragments, were counted in three random reticle fields for between 16 and 65 separate cultures of each substratum coating at 7 and 14 d.i.v. Means and standard errors for each experimental group were expressed as percentages of the numbers of neurones maintained in the control culture regime of an uncoated substratum and medium that did not contain cytosine arabinoside for comparative purposes. Statistical analysis by ANOVA and Student’s f-test was carried out.

Neurone diameter, and neurite number and length data were collected from coverslips prepared for SEM, fixed for 30 min with 3% glutaraldehyde as reported (Smith & Mclnnes, 1986). Living cells were also measured by phase-contrast microscopy to ensure that no shrinkage resulted from the SEM preparatory steps.

Isolated cells and small, non-dissociated ganglion fragments attached to the substratum within 12–24 h of plating. Where cells were maintained in standard medium, NNC proliferation formed a confluent monolayer by 4–5 d.i.v. Large, spherical phase-bright neurones were evident on this cellular network, as were many regenerated processes, which either extended over the monolayer surface or descended into it. In AraC-in-hibited cultures NNC outgrowth was prevented at 4-5 d.i.v., so that the majority of surviving phase-bright neurones attached directly to the plastic surface. Cells maintained on the different coatings retained identical morphologies to those on uncoated dishes; neurones and NNCs were readily distinguishable (Smith & Orr, 1987).

The number of NNCs reached a maximum by 7 d.i.v. in standard medium and remained at this level throughout culture. In the presence of AraC a 25-fold reduction in NNC numbers was observed, irrespective of substratum, at both 7 and 14d.i.v., compared to cultures maintained in standard non-inhibited medium (Smith & Orr, 1987).

Numbers of neurones in cultures where NNC outgrowth was inhibited, however, were dependent on the choice of coating used, and in some cases on the duration of culture (Table 1). Data collected for cultures on each of the surfaces maintained in medium containing AraC were compared with those from the control uncoated, uninhibited cultures, in which the confluent bed of NNCs provided a natural substratum. In absolute numbers there was a 10% decrease in the initial plating density over the 14d.i.v. period for such controls. When neurones were maintained on the uncoated plastic in the presence of AraC, neuronal counts were significantly reduced by approximately 40% at both 7 and 14d.i.v., compared with non-inhibited cultures (Fig. 1).

AraC-inhibited cultures maintained on substrata of FN and LAM, either singly or in combination, did not show this reduction in neuronal numbers, but rather retained levels similar to those of the non-inhibited cultures where NNCs were present. Compared with uncoated plastic substrata, FN, LAM and FN/LAM coatings produced significantly greater numbers of neurones in the presence of AraC (Table 1).

Counts of neurones cultured in AraC-inhibited media maintained on surfaces coated with the microexudates derived from hepatocytes or 3T6 cells (fibroblasts) differed depending on the length of time in vitro. Neurones maintained on fibroblast-derived substrata retained levels approaching those of glycoprotein coatings over the full 14d.i.v. period. However, although hepatocyte-derived material proved effective for 7d.i.v. when compared statistically with neuronal numbers on uncoated plastic in AraC-containing medium, by 14d.i.v. numbers had decreased and were no longer higher than on uncoated plastic surfaces. This was similar to our previous findings for maintenance on substrata of synthetic coating agents of polylysine and polyornithine (Smith & Orr, 1987).

The different cell types were readily distinguishable by SEM. This method of monitoring morphology was preferable to phase-contrast microscopy since neurone perikarya, the number of neurites and process length could all be observed with greater accuracy. In cultures maintained in standard non-inhibited medium, the confluent bed of NNCs consisted of a range of cell types, including flattened and fusiform cells, macrophages and fibroblast-like cells (Fig. 2). Large rounded neurones, often with roughened surfaces, rested upon this NNC network. Neuronal perikarya varied in diameter from 10—15 pm to large cells of over 35 μm. From neurones of all sizes processes extended and could be traced over the surface of the monolayer. Where AraC was included in the medium, the numbers of NNCs were reduced, so neurite extension was more easily seen (Fig. 3). Neurones appeared identical in morphology to those in non-inhibited cultures, with no evidence of damage to the perikarya as reported for adult rat DRG neurones when higher concentrations of AraC were used (Grothe & Unsicker, 1987). In AraC-inhibited cultures many neurones adhered directly to the various substrata, e.g. fibroblast exudate (Fig. 4), the numbers being dependent on the culture regime employed. Fewer neurones with a mean diameter of less than 15 gm were observed on coated surfaces, with a corresponding increase in those in the size range 15–25 μm and over. With some substratum coatings, neurites in excess of 500 μm were frequently extended. Such neurites were often branched and distal growth cones were evident (Fig. 5).

The numbers of neurites extended by neurones on each substratum were counted (Table 2). In the presence of AraC, on uncoated plastic surfaces, more than 50% of the neurones failed to extend any processes. In contrast, on FN and LAM over 60% of neurones produced neurites, while with a combination of FN/LAM almost 70% regenerated processes. Comparable numbers of neurones with neurites were evident when cultured on fibroblast-derived substrata. Process extension on hepatocyte exudates, however, was no better than on uncoated plastic. Polylysine was least efficient in stimulating neurite growth, with over 60% of cells failing to extend processes.

The microexudates produced by 3T6 cells and primary’ cultured mouse hepatocytes were monitored by SEM following cell removal by treatment with EDTA. Substratum-attached material was observed in each case (Fig. 6A,B). The fibroblast-derived exudate, together with standards that included fibronectin and laminin, was further analysed by electrophoresis using native PAGE gels. A protein, equivalent to the polymeric form of the fibronectin standard (.’V/_r 660×10³), was evident on silver-stained gels; no bands were discernible, however, that corresponded to the laminin standard.

We previously proposed that non-neuronal cells (NNCs) have an important role in furnishing a suitable environment for adult sensory neurones (Smith & Orr, 1987). We found that neuronal numbers were significantly reduced when cultured on plastic surfaces with a medium that eliminated non-neuronal cell proliferation. However, fibronectin substrata enhanced neuronal maintenance in NNC-reduced cultures, permitting the initiation of refined neurone cultures, in addition to co-cultures, for experimental use (Smith et al. 1987). The current work extends those studies by demonstrating that substrata that include LAM alone, or combined with FN, or microexudates prepared from fibroblasts, all gave consistent maintenance of adult mouse neurones in vitm. Recent findings of Grothe & Unsicker (1987) with adult rat sensory’ neurones are consistent with our present observation for LAM: they maintained their cultures for 7 d.i.v. only, however, since they were unable to prevent non-neuronal cells forming a dense network in their initially neurone-enriched cultures.

Cornbrooks et al. (1983) proposed that Schwann cells produced LAM, whilst fibroblasts secreted FN in DRG cultures. Both these glycoproteins could be expected, therefore, to be in close proximity to neurones in cocultures with NNCs. A putative receptor for FN and LAM has been identified on all areas of DRG neurones, indicating that potentially direct and specific interactions with these glycoproteins occur (Bozyczko & Horwitz, 1985). FN and LAM have been localized around the neurone-satellite cell complex in the DRG in vivo, with LAM confined to the basal lamina and FN distributed more widely throughout the endoneurium (Bannerman et al. 1986). The current study showed that adult mouse neurones were maintained in vitro when NNC numbers were reduced, providing they were supplemented with an exogenous source of LAM and FN, consistent with the in vivo data. It would be of interest to ascertain if different subsets of neuropeptide-reactive neurones are preferentially maintained by these coatings. Similar reports exist for adult rat neurones where cells containing somatostatin and cholecytokinin-8 are supported by LAM, but substance-? reactive cells declined in numbers after 7 days; although, since these cultures were initiated by centrifugation with resultant high neuronal loss during preparation, pre-selection for certain neuronal subsets cannot be ruled out (Grothe & Unsicker, 1987).

Ciliary ganglion neurones extended neurites more frequently on an extracellular matrix containing LAM, which had been produced by corneal endothelial cells, than on a substratum derived from fibroblasts, which did not contain LAM (Fridman et al. 1985). In the present study neuronal counts were significantly greater throughout the entire 14-day culture period for cells cultured on fibroblast-derived exudates than for neurones maintained on uncoated dishes. In contrast, neurones maintained on hepatocyte-derived exudate survived in statistically greater numbers for 7 d.i.v. compared with uncoated surfaces, but by 14d.i.v. no benefit was evident in NNC-reduced cultures. Such differences in neuronal response are indicative of variations in the matrices produced by the two cell types. In the present study we obtained evidence of fibronectin by electrophoretic analysis in the fibroblast exudate although no proteins with molecular weights relating to laminin standards were observed. Fibroblasts have previously been shown to produce cellular FN in vitro (Yamada & Olden, 1978), whilst there is evidence that hepatocytes synthesize plasma FN (Tamkum & Hynes, 1983), which may not have been retained in the exudate matrix prepared in this present investigation.

Greater numbers of neurones extended neurites when cultured on glycoprotein substrata, presumably due to specific interactions between the FN and LAM and the regenerating neurites. This is in agreement with the findings of Bozyczko & Horwitz (1986), which demonstrated putative receptors on growth cones and neurites in addition to those on DRG perikarya. There are also reports of the effects of FN and LAM on neurite extension in embryonic systems. Both glycoproteins enhanced neurite production in human foetal sensory neurones (Baron-van Evercooren et al. 1982), and in chick central nervous system (Manthorpe et al. 1983) and peripheral nervous system neurones (Rogers et al. 1983).

Recent work on foetal chick sensory neurones has shown that fibronectin or laminin do increase survival, but at certain embryonic ages this is dependent on cooperation with nerve growth factor as well, highlighting the complexities of systems relying upon foetal neurones (Millar-uelo et al. 1988). The interactions involved would appear to be highly specific, since neurite extension and guidance was prevented on areas of grids inactivated by lengthy u.v. irradiation for LAM coatings, although not on u.v.-irradiated FN, illustrating that minor molecular changes in LAM influenced the cellular response (Ham-marback et al. 1985). Our present findings show for the first time that neurite regeneration by adult sensory neurones may be governed by similar factors, with LAM and FN inducing neurite production compared to uncoated or polylysine-coated substrata.

Hawrot (1980) cultured neonatal sympathetic neurones on exudates from cardiac myocytes, which were thought to secrete FN, and which increased neurite extension. In the current study the fibroblast-derived exudate was more effective and therefore may contain greater numbers of components that stimulate neurite production than the exudate produced by cultured hepatocytes.

The fact that adult neurones extend processes on LAM or FN substrata could contribute to a better understanding of nerve regeneration in vivo. Carbonetto et al. (1987) reported that frozen tissue slices from different nervous system sources supported nerve fibre growth in cultured embryonic DRGs, provided that the tissue, i.e. sciatic nerve, spinal cord or optic nerve, was one that induced regeneration in vivo. Each slice type that stimulated neurite extension was shown to contain FN or LAM. In regeneration of adult nervous system in vivo fibres are often found in close contact with the basal lamina or Schwann tubes (Ide et al. 1983). The present work shows that LAM and FN in combination are probably important components for maintaining adult mouse sensory neurones in NNC-reduced cultures and for influencing nerve fibre extension. The role of the extracellular matrix and its components in regulating adult neurone survival and process regeneration are likely to retain a pivotal position in future in vivo and in vitro studies.

The critical reading of the manuscript by Professor R. J. Scothorne and Mr I. B. Mclnnes has been most helpful. The authors thank Miss C. A. Morris for preparation of artwork, and Miss M. Hughes for help with photography. Equipment purchased with assistance from Glasgow University Medical Research Funds is acknowledged.