The effects of protease inhibitors on axon growth through astrocytes

Axons will not normally regenerate in the mature mammalian central nervous system (Cajal, 1928). This regenerative failure is due, at least in part, to the CNS environment being inhibitory to axonal growth: experiments in which CNS tissue is transplanted into the PNS and PNS tissue into the CNS have demonstrated that axons from most neurons will regenerate readily in PNS tissue, while CNS tissue blocks regenerative axonal growth (Weinberg and Spencer, 1979; David and Aguayo, 1981; Aguayo et al. 1978; Anderson and Turmaine, 1986; Hall and Kent, 1987). Some progress has been made in defining the features of the CNS environment that inhibit growth. The cell types surrounding damaged axons in the CNS are predominantly astrocytes and oligodendrocytes. Oligodendrocytes in vitro repel growing axons by causing growth cones which contact them to collapse and withdraw, and this has been shown to be a property of various molecules found in CNS myelin (Caroni and Schwab, 1988; Fawcett et al. 1989a; Savio and Schwab, 1989; Pesheva et al. 1989): this, and other in vivo evidence, is consistent with the view that oligodendrocytes are, at least in part, responsible for the failure of axons to regenerate in the CNS. However, the predominant cell type in regions of damaged CNS is the astrocyte, and abortively sprouting damaged axons are usually seen in contact with these cells (Liuzzi and Lasek, 1987; Wolburg et al. 1986; Reier and Houle, 1988). Axons in vitro grow well on monolayers of astrocytes (Noble et al. 1984; Tomaselli et al. 1988), so it has been assumed that astrocytes encourage axon regeneration. However, we have recently shown that, when neurons are put in contact with mature astrocytes grown in more life-like three-dimensional cultures, axons grow much less readily than on astrocyte monolayers (Fawcett et al. 1989b): in three-dimensional astrocyte cultures, axons from postnatal dorsal root ganglia (DRGs) or retina will only grow for distances of up to 100/mi. On the other hand, if embryonic retina or DRG neurons are put in contact with mature astrocytes, their axons can grow for a maximum of 1.5 mm within 2 days; thus embryonic axons can penetrate astrocytic tissues which block the growth of axons from postnatal neurons. This difference between growth from embryonic and postnatal neurons is much less apparent on astrocyte monolayers. Here axons from both embryonic and postnatal DRGs or retina grow readily for distances of up to 2 mm in 2 days.

We have been investigating some of the factors that might allow axons from embryonic neurons to penetrate three-dimensional astrocyte cultures so much better than those from postnatal neurons. A general feature of invasive cell types, whether tumour cells, trophoblast cells, or vascular endothelial cells is the secretion of proteolytic enzymes, or proteases (Mignatti et al. 1989; Moscatelli and Rifkin, 1988; Liotta and Stetler-Stevenson, 1989; Bazer and Roberts, 1983). In the case of tumour invasion, the invasiveness of a cell type correlates well with its degree of protease secretion, and protease inhibitors can diminish the ability of tumour cells to invade. Neurons and axonal processes have been shown to secrete a variety of proteases (Pittman and Patterson, 1987; Pittman and Buettner, 1989), while astrocytes secrete two protease inhibitors, nexin-1 and alpha 2 macroglobulin (Rosenblatt et al. 1987a; Gloor et al. 1986; Bauer et al. 1988). In this paper, we investigate the effects of a variety of inhibitors to the main types of protease on axonal growth from embryonic dorsal root ganglia on flat collagen surfaces, on astrocyte monolayer cultures, and in three-dimensional astrocyte cultures.

Dorsal root ganglia were obtained from embryonic CFHB rats, the plug date being counted as embryonic day 1 (El).

50 μl Vitrogen 100 was spread over a 35 mm culture dish, and allowed to air dry. After washing with Hanks, DRGs were placed on the dish, and covered with a thin layer of medium, which was topped up after 16h.

Newborn rat pups, aged P0 to P2, were killed by open ether anaesthesia. The brains were removed, demembranated, chopped and then incubated with 0.1% trypsin (Sigma no. T0134) in HBSS for 10min, then 20 μg DNAse (Sigma no. D5025) per ml was added, and the tube shaken briefly. The tissue was triturated in the following solution: (300 mg BSA-Sigma no. A3912, 1 mg DNAse-Sigma no. D5025, 50 mg Trypsin inhibitor-Sigma no. T9003 per 100ml HBSS), then allowed to settle, and the supernatant mixed with medium (DMEM+10% FCS) and plated onto poly-lysine coated glass cover slips in 6 well plates, approximately 1 brain per plate. The cells generally formed a monolayer after about 6 days in culture. Before use in an assay the cells of the oligodendrocyte lineage were removed by complement mediated lysis, as described below. Dorsal root ganglia were placed on these cultures, and held in place by surface tension from a thin film of medium, until the medium was topped up on the following day.

These were made by the same technique described above, approximately 1 brain being put into each flask. After 6 to 10 days, the majority of cells of the oligodendrocyte/type 2 astrocyte lineage were removed by shaking the culture overnight (McCarthy and de Vellis, 1980). Final purification of flask and coverslip cultures was achieved by complement mediated lysis. Cultures were incubated in tetanus toxin (Calbiochem 20 μgml ⁻¹), followed by anti-tetanus toxin (1:100, gift from Dr R. Thomson, Wellcome labs) and antigalactocerebroside (1:100, gift from Dr G. Giotta) with complement.

These were made by putting astrocytes cultured in flasks into cellulose ester tubes of 0.5 mm internal diameter and 0.2 [im pore size. These were obtained by cutting open Microgon minikap filter capsules (Northumbria biologicals); the tubing was boiled in water to remove any detergent. 1 –1.5 cm lengths of tube were glued with Araldite into the end of reamed pipette tips and the open ends blocked with the same glue. The tubes were filled by resuspending purified astrocytes trypsinised from primary cultures in approx. 0.25 ml of medium, and putting a short column of the suspension into each pipette tip-tube assembly. The column of cell suspension was pushed into the porous tube using air pressure, followed by centrifugation of the whole assembly at 100 g. The filled tubes were then cut from the pipette tips, and divided into 3 or 4 lengths. DRGs were introduced into one end of the tubes, and the cultures were maintained in DMEM with 10%FCS, NGF and the appropriate protease inhibitors on a gentle shaker.

Tubes were fixed in 4% paraformaldehyde, embedded in 12% gelatin and sectioned on a cryostat. Staining was with anti-GFAP (DAKO) 1:250 and 3A10 anti-68kD neurofilament monoclonal 1:20 (gift from Dr T. Jessell), followed by FTTC-labelled anti-rabbit (Caltag) and biotinylated antimouse (Caltag), then TRITC-conjugated streptavidin (Serotec). Cover slip cultures were fixed with paraformaldehyde and stained with the same antibodies.

The results were measured and quantified in the same way for both two- and three-dimensional cultures. Tube sections were examined under a fluorescence microscope, and the distance from the interface between DRG and astrocytes to the furthest limit of axonal growth was measured. 5 sections from each tube were measured. To quantify growth from DRGs on flat surfaces, the distance from the DRG to the furthest point of axonal outgrowth was measured for each individual DRG. For both tubes and DRGs, the controls for each individual run had their maximum outgrowth length measured, then these were averaged together, to give a control maximum outgrowth. The maximum outgrowth lengths of the individual experimental DRGs were then measured, and calculated as a proportion of the control outgrowth. The results from the various runs of each category of experiment were then pooled and averaged to give a figure for the ‘average maximum length’ of outgrowth for each category. Similarly, the number of axons leaving the DRGs cultured on astrocytes was counted; however, even under high power, it was not always possible to resolve how many axons were in the larger fascicles, so the accuracy of these counts is lower for DRGs with many axons than for those with fewer.

We have examined the effects of protease inhibitors on axonal growth from embryonic day 15 DRGs in three culture models; DRGs were plated directly onto air--dried collagen, onto two-dimensional astrocyte cultures, and were inserted into tube cultures filled with astrocytes that had been in culture for 3 weeks or more. The cultures were treated with inhibitors for the three main classes of protease. Serine protease inhibitors were ∈ amino caproic acid (EACA) 4mM combined with Soyabean trypsin inhibitor (SBTI) 100 μgml ⁻¹, and Trasylol 200 U ml ⁻¹. Collagenase-type metalloproteases were inhibited with tissue inhibitor of metalloproteases (TIMP) 15 μgml ⁻¹ (gift from Dr G. Murphy), and enteroprotease type metalloproteases by Cbz-Gly-Phen-amide, 1 or 2mM. o-phenanthroline, which is a chelating agent that inhibits metalloproteases, was also used at 10 μgml ⁻¹ or less.

In control cultures in which E15 DRGs were put in contact with astrocytes, the average maximum outgrowth length was 867 μm over 2 days, a figure very similar to that which we saw in our previous experiments (Fig. 1A). Both of the serine protease inhibitors produced significant (P<0.001 by t-test) decreases in the length of axon outgrowth, to 40% of control in the case of EACA+SBTI, and to 50% of control in the case of Trasylol (see Fig. 2). Tissue inhibitor of metalloproteases (TIMP) had no significant effect on the length of outgrowth, but the inhibitor of enteroprotease-type metalloproteases, Cbz-gly-phen-amide, had a clear dose-related effect; growth in 1mM gly-phen was 81 % of control, and in 2 mM gly-phen was 17% of control. A control peptide, cbz-gly-gly, at the same concentration of 2mM had no effect on axon outgrowth, o-phenanthroline at 5 or 10 μgml ⁻¹ almost completely prevented any axon outgrowth in tube cultures. Both o- phenanthroline and cbz-gly-phen-amide had obvious effects on the astrocytes in the tube cultures. Whereas astrocytes in tubes normally form a solid mass of interweaving processes, tightly adherent to one another (Fig. 3), in gly-phen treated tubes the astrocytes were clearly much more rounded, and there seemed to be significant amounts of space between them. They still stained strongly for GFAP, however, o-phenanthroline also caused the astrocytes to round up, and often the volume of cells was greatly reduced, leaving the centre of the tubes empty. Most of the remaining cells did not stain for GFAP, and many appeared to be dead.

We stained some tubes with antibodies to laminin, to see how much extracellular matrix there was between the astrocytes. As shown in Fig. IB, there was no laminin staining in the cultures, except for a few brightly stained patches; these patches were probably groups of leptomeningeal cells, which are always a contaminant of astrocyte cultures, and which stain strongly for a variety of extracellular matrix components including laminin.

Axonal growth from E15 DRGs on astrocyte monolayers was prolific, extending for an average distance of 1188 μm in two days (Fig. 4A). The pattern of inhibition of axon growth by protease inhibitors in two-dimensional cultures was mostly similar to that seen in three--dimensional cultures (Fig. 5). That is, cbz-gly-phen-amide and o-phenanthroline inhibited axonal growth in a dose-related manner, and to about the same extent as in three-dimensional cultures, although o-phenanthro-line almost completely prevented growth in the three-dimensional cultures, but only partially prevented in on two-dimensional cultures (Fig. 4C). As in three-dimen-sional cultures, TIMP had no effect on axon outgrowth in two-dimensional cultures. The main difference between two- and three-dimensional culture systems was that neither Trasylol nor EACA+SBTI had any significant effect on the length or the numbers of axons growing in monolayer cultures, whereas both reduced growth significantly in three-dimensional cultures, o-phenanthroline and gly-phen both affected the astrocytes in the cultures: o-phenanthroline at 10 μg ml ⁻¹ appeared to be toxic over the two days of the experiment, many of the astrocytes floating off into the medium, and many of the remaining ones losing their GFAP staining, and becoming much less flattened onto the culture surface (Fig. 4D) lower doses of o-phenanthroline produced the same appearance, but less so. Gly-phen also affected astrocytic morphology; the cells remained flat, adherent to the surface, and retained their normal cytoskeletal morphology as revealed by GFAP staining, but the cells were less closely packed than normal, and not as obviously adherent to one another as in normal cultures (Fig. 6). The lower cell density suggests that some cells had floated free of the culture surface, but this was not as marked as with o-phenanthroline.

Axons grew from E15 DRGs less prolifically than on astrocyte monolayers, and they also grew less far; the average length of growth in our experiments was 455 μm in 2 days (Fig. 7). We tested growth on collagen to see whether any of the protease inhibitors that had inhibited growth in the axon-astrocyte co-cultures had direct effects on axonal growth, independent of any effect on axon-glial interactions. EACA+SBTI did not have any effect on the length of axon outgrowth in these cultures, but Gly-Phen-amide did reduce growth, the extent of inhibition being very similar to that seen in two- and three-dimensional astrocyte co-cultures. Using time-lapse videomicroscopy, we observed that Gly-phen amide, when applied to axons growing on collagen, caused the growth cone to collapse within 15 min, and the distal part of the axons to withdraw; the effect is similar to that seen with tubulin depolymerizing agents, o-phenanthroline also almost completely prevented axon growth, although it did not cause the same rapid growth cone collapse as gly-phen amide.

Summarized, our results show that cbz-gly-phenamide and o-phenanthroline reduced axon growth in all three culture models by about the same amount, and therefore have a direct effect on axonal growth, o-phenanthroline is probably toxic to cells over the 2 days of our experiments, but gly-phen does not seem to be as toxic to astrocytes, and its effects on axonal growth remain reversible for at least 24 h. TIMP had no effect on axon growth in any of our three culture models. The serine protease inhibitors, EACA+SBTI and Trasylol, decreased axonal growth in three-dimen-sional cultures, but not on astrocyte monolayers or on collagen. These two serine protease inhibitors must therefore be inhibiting a protease-dependent process which is necessary for axons to penetrate three-dimensional cultures, but which is not required for successful axonal growth on flat cultures.

We have investigated the effect of various protease inhibitors on the growth of .axons from embryonic dorsal root ganglia. We have examined growth on air--dried collagen, on astrocyte monolayers and in three--dimensional astrocyte cultures. We have used a variety of inhibitors to the main classes of proteases implicated in invasive events, namely serine proteases, collagenase type metalloproteases, and enteroproteases/ neuropeptidases (Moscatelli and Rifkin, 1988; Paganetti et al. 1988; Mignatti et al. 1989; Ossowski, 19886; Mignatti et al. 1986). The serine protease inhibitors EACA and trasylol reduced the maximum extent of axonal growth in three-dimensional astrocyte cultures to 40% and 50% of its control value, respectively, but had little effect on growth on collagen, or on astrocyte monolayers. The peptide cbz-gly-Phen-NH₂, which inhibits metalloproteases of the enteroprotease type (Baxter et al. 1983; Couch and Strittmatter, 1983; Lelkes and Pollard, 1987), inhibited axonal growth in all three culture models, causing growth cone collapse and distal axon withdrawal, but TIMP, which inhibits metalloproteases of the collagenase family (Murphy et al. 1989; Docherty et al. 1985), had little effect in any of the culture situations, o-phenanthroline, a chelating agent, inhibits both families of metalloprotease, and inhibited axon growth in all our culture models; indeed, it appears to be toxic to both neurons and non-neuronal cells.

Since our three-dimensional cultures are tissues, the issue of penetration of molecules into the cultures is important. We know that large molecules penetrate the cultures little, since attempts to block stain live cultures with antibodies (M_r of IgG ≈150 000) have resulted in the staining of only the surface of the cultures. The molecules that inhibited axon growth in three-dimen-sional cultures are small; much the largest is trasylol, which is a polypeptide of M_r 6500. Soyabean trypsin inhibitor, which we mixed with EACA is relatively large (M/_r 20100) and may therefore have added little to its effect. It is possible that TIMP, which has a M_r of 29000, did not penetrate the cultures in sufficient amounts to reach an effective concentration between the cells. However, our three-dimensional cultures contained no stainable extracellular matrix except around meningeal cells, as is the case in the mature brain, so it is unlikely that growing axons need to break down matrix materials in order to grow through astrocytes. Our experiments were conducted in serumcontaining medium, and serum contains protease inhibitors, particularly alpha 2 macroglobulin and some serine protease inhibitors. These, particularly alpha 2 macroglobulin (M_r 720000), are large molecules, which are unlikely to have penetrated far into our three--dimensional cultures.

Proteases have been implicated in the invasive process in tumour cells and other cell types (Mignatti et al. 1989; Moscatelli and Rifkin, 1988; Liotta and StetlerStevenson, 1989; Bazer and Roberts, 1983), the proteases often being bound to cell surface receptors after secretion. There is a close correlation between the production of proteases and the invasiveness of tumour cell lines, and tumour metastasis can be prevented in vivo by protease inhibitors (Ossowski, 1988a). Tumour cell invasion has been shown to rely on three main families of protease, serine proteases, collagenase-type metalloproteases and neuropeptidase-type metalloproteases. Individual cell lines secrete these proteases in varying proportions, but inhibitors active against all three families have been shown to slow or prevent tumour invasion in a variety of in vitro assays (Ossowski, 1988b; Paganetti et al. 1988; Mignatti et al. 1986), the effective protease inhibitors presumably being determined by the spectrum of proteases secreted by the cell line in question. Our experiments suggest that axonal growth cones are in some respects similar to other invasive cell types in that they also appear to rely on proteases for a part of their invasive properties. Axonal growth cones have been shown to secrete proteases of the serine protease type and also collagenase-type metalloproteases (Pittman and Patterson, 1987; Pittman and Williams, 1989; Machida etal. 1989). On the other hand, astrocytes secrete two potent serine protease inhibitors, nexin-1 and alpha 2 macroglobulin (Gloor et al. 1986; Rosenblatt et al. 1987b; Bauer et al. 1988).

In our assays, axon growth was inhibited by both serine protease inhibitors and inhibitors of a class of metalloproteases. The sites of action of these molecules are probably rather different. The serine protease inhibitors only inhibited growth in three-dimensional astrocyte cultures. In this culture model alone, the axons have to penetrate a three-dimensional matrix of closely apposed astrocytes, presumably needing to break cellular junctions in the process, and local proteolysis might aid this process. Growth of axons on astrocyte monolayers does not require axons to penetrate between cells, and it is therefore perhaps not surprising that the serine protease inhibitors had little effect on axon growth in this situation. The effect of serine protease inhibitors on axon outgrowth in our assay is particularly interesting, because astrocytes themselves secrete serine protease inhibitors, and these might have the effect of restricting axonal growth into astrocytic tissues. The peptide cbz-gly-phen-amide inhibited axon growth in all of our three culture models. Since growth of axons in the absence of glia was affected, the peptide must have a direct effect on neurons, and we have observed it to have a very rapid effect in causing growth cones to collapse and withdraw. Proteases of the type inhibited by this peptide have been reported to be involved in various cellular processes, such as axonal transport, transmitter release and cell fusion, and may be involved in calcium homeostasis (Hammerschlag et al. 1989; Lelkes and Pollard, 1987; Baxter et al. 1983; Couch and Strittmatter, 1983). The concentration of inhibitor required to affect these processes is very similar to that which was effective in our experiments. Enteroproteases are generally recovered from the membrane fraction of cells, and have been implicated in invasive events and destruction of extracellular matrix (Chen and Chen, 1987; Paganetti et al. 1988). It is possible that axonal growth cone progression also uses this family of enzymes to affect the extracellular environment. However, in our experiments the direct effect of the inhibitor on neurons makes it impossible for us to determine whether this is so.

Adhesion molecules, particularly N-cadherin, are critical for the growth of axons over astrocyte monolayers (Tomaselli et al. 1986; Neugebauer et al. 1988). Their role in controlling axonal growth into three--dimensional astrocyte tissues is likely to be more complex. Indeed, in tumour cells, increased invasiveness may be associated with a decrease in cell surface adhesion molecules (Behrens et al. 1989; Roman et al. 1989). N-cadherin is a homophilic adhesion molecule, and therefore presumably causes neighbouring astrocytes to adhere to one another as well as promoting axon-astrocyte adhesion. Localized proteolysis presumably affects the balance of the interactions between cells and axons.

In our previous study using the three-dimensional astrocyte culture model (Fawcett et al. 1989) we found that axons from postnatal DRGs or retina penetrated the astrocytes hardly at all, while axons from embryonic retina and DRGs grew for considerable distances. It is unlikely that this difference between embryonic and postnatal neurons can be entirely ascribed to differences in protease production, since inhibition of serine proteases did not reduce axon growth from embryonic DRGs to the same level as we found in postnatal DRGs.

Since proteases seem to play an important role in axon growth and invasiveness, it will be important to define the factors controlling their secretion by axons, and their roles inside neurons. It will also be useful to detail the proteases and protease inhibitors secreted by non-neuronal cells, and to define the resultant microenvironment that surrounds the growth cone as it interacts with the tissues that it encounters.

This work was supported by grants from the International Spinal Research Trust, and the Medical Research Council.