Mechanisms governing neuronal degeneration and axonal regeneration in the mature retinofugal system

Observations concerning the consequences of injury to the central nervous system, the spinal cord and the retina of higher vertebrates can be traced back to the early decades of the century (Cajal, 1928; James, 1933; Eayrs, 1952). In accordance with these observations, which have been confirmed later, the course of retrograde adult retinal ganglion cell degeneration commences a few days after intraorbital transection of the optic nerve and progresses during the weeks and months following the axotomy, finally resulting in depletion of the retinal ganglion cell layer (GCL) (Richardson et al. 1982; Barron et al. 1986; Thanos, 1988; Villegas-Perez et al. 1988; Carmignoto et al. 1989). It is assumed that the failure of lesioned ganglion cells to regrow their axons within the distal portion of the optic nerve is caused by the presence of differentiated oligodendrocytes whose myelin exerts inhibiting influences both on embryonic (Schwab and Caroni, 1988) and on adult ganglion cell axons (Vanselow et al. 1990). In addition to the inhibiting environment, it has been assumed that insufficient growth-supporting agents within the optic nerve (Cajal, 1928) determine the fate of lesioned neurons, namely the progressive degeneration. External neurotrophic influences introduced by the apposition of peripheral nerve segments at the time of severing the optic nerve could rescue some ganglion cells, which then can regenerate into growth-permitting peripheral nerve transplants (Vidal-Sanz et al. 1987; Villegas-Perez et al. 1988). Factors released from peripheral nerves also support regrowth of axons in cultured retinal expiants (Thanos et al. 1989). The responsiveness of lesioned ganglion cells to external administration of nerve growth factor (NGF) during the first weeks after lesion (Carmignoto et al. 1989) is in line with all previous observations that epigenetic influences can regulate the quantities of neurons which survive axotomy.

Further exploration and intervention into the responses of ganglion cells and of their glial environment to external violence may result in persistent increase of the numbers of neurons which can be then recruited to regrow and reconnect their axons. Since ganglion cell lesion and subsequent death causes a local inflammation within the ganglion cell and inner plexiform layers, local or generalized mechanisms of recognition and removal of cell debris are probably essential in order to protect the surrounding tissue from further lytic destruction. The investigation of the entire local response within’ the retina is therefore of crucial importance for understanding the mechanisms of lesion-induced cell death, and perhaps for preventing it. In addition to neurons and the macroglia (astrocytes and the Müller’s glia), microglia are the third major population of cells within the retina of mammals (del Rio-Hortega, 1932; Cammermeyer, 1970). Microglia are also localized in various areas of the developing, adult and lesioned CNS (Cammermeyer, 1970; Streit et al. 1990; Perry and Gordon, 1988; Schnitzer and Scherer, 1990). The function of brain and retinal microglia in the repair process at the sites of injuries is not well defined, although several lines of evidence support the view that these cells are responsible for the immune response and phagocytosis (see Perry and Gordon, 1988, and Giulian, 1990 for reviews). Observations based on the temporal relation of the microglia to neuronal cell death in the lesioned adult rabbit retina (Schnitzer and Scherer, 1990) have assigned to the microglia a role in the dying cells.

The present work was based on the expectation that retinal proteases, probably produced by microglia, are directly involved in the cascade of regressive events initiated by optic nerve transection. It basically monitored whether blockade of the proteolytic activities can rescue the lesioned neurons from degeneration. If so, the model predicts that axotomized and functionally altered ganglion cells are recognized by immunocompetent microglial cells (Streit et al. 1990), whose response to the alteration is the onset of enzymatic neuron degradation and subsequent endocytosis. This possibility, called neuro-phagy, implies that microglia represent a local immune system devoted to the protection of functional and structural retinal integrity. Consequently, such a mechanism also presumes that microglial cells use proteases (Giulian, 1990) to degrade severed ganglion cells. Administration of protease inhibitors and anti-inflammatory treatment would therefore influence ganglion cell degradation. To confirm this, the optic nerves of adult rats were transected beyond the eye cup and the retinal ganglion cells whose axons form the optic nerve were retrogradely labelled with the fluorescent dye 4DI-10ASP which accompanies the membrane particles after cell degradation and phagocytosis (Thanos et al., unpublished observations). Protease inhibitors (Table 1) were injected into the vitreous body during the surgical optic nerve interruption. It is shown here that substantial numbers of vital ganglion cells can be rescued and express regenerative capacities in organ cultures and in vivo.

Adult female rats (24) weighing 200 to 230 g from the Sprague-Dawley strain were used for the present study. Under intraperitoneal chloralhydrate anesthesia (42 mg kg⁻¹ body weight), the left optic nerve was intraorbitally exposed and after longitudinal incision of its meningeal sheath the nerve was completely transected. Care was taken to avoid damage of the retinal blood supply. In 20 rats, solid crystals (0.2 to 0.4 mm in diameter) of the fluorescent styryl dye (D291, 2V-4-(4-didecylaminostyryD-AT-methylpyridinium iodide, (4DI-10 ASP), Molecular Probes, Oregon) were deposited immediately after transection at the ocular stump of the optic nerve as reported for Dil (Thanos, 1988), in order to label the retinal ganglion cells retrogradely before they undergo degeneration. Solubilization of the dye at the site of deposition was made by directly pipetting 10 to 20 μl of complete Freund’s adjuvant (Sigma) on the crystal. The adjuvant appeared to be the ideal solvent for the dye and to result in labelling of all retinal ganglion cells. In 10 of these rats, 5 μl of a freshly prepared protease inhibitor and dexamethasone solution (Table 1) were injected into the vitreous body of the axotomized eye with a 10 μl Hamilton syringe. The remaining 10 control rats which received optic nerve transection and 4DI-10ASP, but not inhibitors, were either used to determine the normal course of retrograde ganglion cell degeneration and microglial labelling (8 rats), or they received intravitreal injection of 5 #x03BC;l phosphate buffer (2 rats). In 6 further control rats (3 with optic nerve cut and protease inhibitor injection, and 3 with nerve cut but without injection), the dye was deposited at the optic nerve stump during a second surgical intervention, 2 days prior to the animal’s death, in order to label only the population of ganglion cells which survived the primary axotomy.

Following survival times of 2, 8,14, 30 and 60 days, the rats were deeply anesthetized with 7% chloralhydrate. After intracardial perfusion with phosphate-buffered saline, the animals were killed and fixed with 200 ml aqueous 4% paraformaldehyde and their retinae were dissected, incised into four quadrants and flat-mounted on filters with the nerve fiber layer upwards.

Under chloralhydrate anesthesia, the left optic nerve of 20 rats was surgically exposed and crushed within its intraorbital segment with a jeweller’s forceps, in order to produce a conditioning lesion, which has beneficial effects on the regenerative response of ganglion cells (Thanos et al. 1989). In 8 rats, 5 μl of the protease inhibitor/dexamethasone solution was injected into the vitreous body. The remaining 12 rats were used as controls without any injection during the crushing procedure. One week after severing the optic nerve, the retinae of animals with protease inhibitor/dexamethasone injection and those of the controls were dissected under sterile conditions and were used to produce organ cultures in a chemically defined medium devoid of serum and growth factors, according to the technique of Thanos et al. (1989). Each retina was divided into 8 optic nerve head centered pieces, which were then explanted on petriperm dishes (Heraeus) coated with polylysine (375 000 to 410 000 × 10^sM_r, Boehringer; 200ugmU¹, overnight at 37 °C) and with laminin (BRL, 20μgml⁻¹, lh at 37°C) with the ganglion cell layer facing the substrate. For examining the effects of protease inhibitors/dexamethasone injected in situ, the 64 expiants obtained from pretreated retinae were not supplemented with neurotrophic factors. Control expiants were substituted with each of the following neurotrophic factors: (1) 20 expiants with basic fibroblast growth factor (bFGF, 5μgml⁻¹); (2) 30 expiants received purified brain-derived neurotrophic factor (Barde et al. 1982) at dilutions identical to these described by Thanos et al. (1989); (3) in an additional 32 expiants derived from protease inhibitor-treated retinae in situ, brain-derived neurotrophic factor (BDNF) was added to the organ cultures, in order to examine additive effects on axonal growth; (4) further controls consisted of cultures of 32 retinal expiants with in situ conditioned sciatic nerve exudate collected into implanted teflon tubes during nerve regeneration (Thanos et al. 1989) and dissociated cells from sciatic nerves which were precrushed 1 week prior to explantation.

In 6 rats weighing 200-230 g, the peroneus communis was used to replace the transected left optic nerve as described in former reports (Vidal-Sanz et al. 1987; Thanos and Vanselow, 1990). In addition to the grafting, 5 μl of the protease inhibitor/dexamethasone solution (Table 1) were injected into the vitreous body of the operated eye. Labelling of the ganglion cells whose axons regenerated into the grafts was performed 4 weeks later, when growing axons were considered to have reached the distal portion of the graft. For this, the grafted nerve was opened at its distal end, transected, and 4DI-10ASP was deposited into it. Three days later, the animals were killed and their retinae were dissected and flat-mounted to examine the ganglion cells which had regenerated their axons and were therefore retrogradely labelled with the fluorescent dye. Control animals were grafted but either did not receive injection of inhibitors (4 rats), or (4 rats) received injection of 5 μl phosphate buffer.

The retinae obtained after optic nerve transection and subsequent degeneration in situ, and those dissected from transplanted animals, were viewed as whole-mounts. Fluorescent ganglion cells and microglia were observed within the fluorescein filter, since 4DI-10ASP fluoresces green-yellowish.

For quantification of the ganglion cells and microglia, each retina was divided into three concentric areas with radii of about 1mm (central), 2 mm (middle) and more than 2 mm (peripheral) from the center of the optic nerve head, as viewed in the whole-mounted retina. Ganglion cell and microglia densities were determined in each concentric field by measuring in each quadrant 30 to 40 randomly distributed microscope fields with the 20 × lens. The data were averaged for each field and then used to obtain densities of ganglion cells and microglia across the retinal surface. Since microglia show a staggered, bilaminated distribution at late stages of degeneration (Thanos et al., unpublished observations), it was essential to measure the microglia within both layers, namely within the ganglion cell layer and within the deeper inner plexiform layer. Statistical analysis of the data obtained for each interval after axotomy was performed with the two-tailed Student’s t-test.

Retinal pieces cultured under the different conditions were scored for axonal growth 2 days after explantation by means of an inverted phase contrast microscope. Numbers of fibers were measured at a distance of 200 pm from the edge of each explant. Averaged numbers of axons from each group of expiants were compared by means of the t-test. Calculation of the ability of the entire retinae to regenerate axons in vitro was made by multiplying the average numbers of fibers with 8, that is with the number of expiants obtained from each retina.

Deposition and local solubilization of the fluorescent dye 4DI-10ASP at the stump of the transected optic nerve of control rats which had not received injections of protease inhibitors, and in control animals which had received injections of 6 μl phosphate buffer, resulted after 2 days in retrograde labelling of 1776±200 cells nun⁻² (12 rats), representing almost the total population of ganglion cells in the rat retina (Perry, 1979). When the animals were killed at later stages after axotomy and labelling, the ganglion cells were intensely labelled, indicating a longterm persistence of the dye within the surviving neurons (Fig. 1). The maximum observed persistence of dye within ganglion cells was 6 months (data not shown). Morphologically, cells which survived axotomy always resembled those shown in Fig. 1. Since axotomy causes a protracted degeneration of ganglion cells (Fig. la,c,e), as expected the densities of these cells continuously declined with time elapsed after axotomy to approximately 55 ±24 labelled cells mm⁻² at the end of the first, and to 20±12 labelled cells mm⁻² at the end of the second month after lesion (Figs 1 and 2).

In conjunction with the course of ganglion cell disappearance, non-ganglionic cells (Fig. lc,e,f), identified with different techniques as microglial cells (Thanos et al., unpublished data), first appeared in the optic fiber layer and ganglion cell layer on the 8th day after lesion, and their density increased in the GCL with time elapsed from lesion, peaking at 14 days (820±64 cells mm⁻²). This microglial density remained stable throughout the time of investigation, covering two months after lesion. Morphologically, microglial cells were characterized by their small perikarya and by the irregularly branching dendrites (Fig. lc,d,e,f). By the second month after optic nerve transection, microglia displayed a strong territorial arrangement within the ganglion cell and inner plexiform layers, and staggered, bilaminated distribution within the two layers (Thanos et al., unpublished data).

Retrograde labelling of the axotomized ganglion cells and simultaneous injection of protease inhibitors into the vitreous body resulted in a different course of ganglion cell depletion and microglial labelling in the retina, observed 2 to 60 days after surgery (Figs 1 and 2). At 8 days, a normal pattern of ganglion cell distribution and of morphologies indicated that there was virtually no degeneration in the GCL (Figs lb and 2). At 14 days, the numbers of ganglion cells with intact perikaryal and dendritic morphologies were about 4-fold higher than in animals without injection of inhibitors/dexamethasone (Figs lc,d and 2). At the end of the first month after lesion and injection, 590 ± 100 cells mm⁻² indicated the efficacy of the inhibitors/dexamethasone treatment in rescuing ganglion cells (Figs If and 2), whereas ganglion cell degradation was not significantly different from controls at 60 days after axotomy and treatment (Fig. 2). Parallel to the reduction of ganglion cells, the densities of labelled microglial cells were significantly lower than in controls during the first month after lesion.

The appearance of more fluorescent ganglion cells within the axotomized and treated retinae does not necessarily document that they were alive, as they could be dead, but non-phagocytosed cells; therefore to confirm their viability it was necessary to demonstrate vital functions of the neurons. The ability of ganglion cells to regrow their axons was considered a major vital function of neurons. To assess whether growth properties of neurons are influenced by the presence of protease inhibitors and dexamethasone, retinal expiants were prepared 1 week after optic nerve transection and intravitreal injection, and cultured in a chemically defined medium not containing neurotrophic agents (Thanos et al. 1989). Massive regrowth of axons 2 days after explantation (Fig. 3a) was significantly higher when the retinae were taken from eyes injected with protease inhibitors (Fig. 3b). Substitution of the cultures with different potent neurotrophic factors like brain-derived neurotrophic factor (BDNF), fibroblast growth factor (Barde et al. 1982) or sciatic nerve-derived exudate (Thanos et al. 1989), was less efficient than pretreatment with protease inhibitors/dexamethasone. Sixty-four expiants obtained from 8 retinae treated with protease inhibitors/dexamethasone gave rise to an average of 650±120 axons/explant on polylysine/laminin. This efficacy of growth per explant corresponds to an incidence of regeneration that approaches 5 % of the total population of ganglion cells, when extrapolated to the entire retina. This incidence is higher than that estimated in vivo, when retinal neurites grew into transplanted peripheral nerve segments (Vidal-Sanz et al. 1987). The results indicate therefore that retardation of neuronal degeneration results in considerable acceleration of outgrowth in organ cultures and proves that non-phagocytosed ganglion cells in the treated retinae retained their ability to regenerate their transected axons.

All animals which received peripheral nerve grafts could be examined for regrowth of axons into the transplants. The numbers of retrogradely labelled ganglion cells varied between the animals of each group. The average number determined over the entire retinal surface was 3.335±980 cells mm⁻² in control animals which did not receive protease inhibitors (n=4, mean±s.D.) and 3.270± 800 mm ² in animals which received only phosphate buffer injections (n=4, mean±s.D.). A significantly increased average number of regenerated ganglion cells (8.250+1002 mm⁻², n=4, mean±s.D.) was determined in animals treated with protease inhibitors and dexamethasone during axotomy and transplantation. Fluorescent ganglion cells were consistently and uniformly distributed throughout the retinal surface. Although large ganglion cell bodies were predominantly represented in the population of regenerated cells (Fig. 4a), smaller neurons (Fig. 4a) were also observed in all animals analysed. Morphologically, regenerated ganglion cells displayed images reminiscent of normal large retinal ganglion cells (Fig. 4a,b). The patterns of dendritic arborizations were also reminiscent of normal ganglion cells. Neighbouring ganglion cells extended dendrites which overlapped each other over large areas (Fig. 4b). This dendritic overlapping was sufficient to cover uniformly the retinal inner plexiform layer, as one of the prerequisites for complete retinal presentation within the grafted piece of peripheral nerve. However, it remains to be analyzed whether this population of regenerating ganglion cells is sufficient to form a functional retinotopic connection when the growing axons are allowed to enter central targets.

The interruption of ganglion cell axons within the optic nerve of adult rats leads to a cascade of neuronal and environmental responses culminating in the progressive and protracted depopulation of the retinal ganglion cell layer. Degenerating ganglion cells are phagocytosed by retinal microglial cells which become activated soon after injury and remove cell debris in a strong chrono-topological sequence that parallels the course of the neuronal degradation (Thanos et al., unpublished data). Fig. 5 illustrates the experimental paradigm used to assess the axotomy-induced responses of the retinal cells, and in particular the new technique of transcellular labelling used to resolve the role of microglia during neuronal degeneration. The principal new finding of the present study was that severed retinal neurons become enzymatically degraded by common proteases, whose activity can be specifically blocked, resulting in retardation of the degenerative events initiated by the axotomy. Consequently, more neurons can be recruited to regrow their axons under certain circumstances.

The marked reduction of ganglion cells following transection of the optic nerve in mammals is one of the major impediments to rescuing and reconnecting these neurons with central target cells in sufficient quantity and in a topographic fashion to ensure functional significance. Consequently, increasing the number of neurons which survive the axotomy and those which contribute to the regrowth of axons, are interdependent prerequisites for restoring function in the retinofugal system. Cell survival

after injury essentially co-regulates the efficacy of subsequent axonal regeneration. Thus, understanding the mechanisms which contribute to the destruction of lesioned neurons, and the development of strategies for manipulating these mechanisms in order to rescue cells from degradation, will help to achieve both goals, that of preventing cell death, and that of promoting axonal regeneration.

As has been consistently documented in experiments using peripheral nerve grafts to allow axons to regenerate within the growth-supporting Schwann’s cell environment and reach central targets, ganglion cells making use of the opportunity to grow (Tello, 1907; Politis and Spencer, 1986; Vidal-Sanz et al. 1987) can be rescued from cell death (Villegas-Perez et al. 1988). This intrinsic ability of lesioned ganglion cells to regrow their axons in favourable environments can also be used in organ cultures devoid of inhibiting central glial influences (Thanos et al. 1989). In addition, the intrinsic ability to regrow in vitro can be externally supported by using the peripheral nerve-derived growth supporting factors which significantly accelerate axonal elongation (Thanos et al. 1989). The major implication of these findings is that the axotomy-initiated cascade of degeneration can be, in principle, manipulated, indicating that insufficient influence of neurotrophic agents in situ is one of the factors which codetermine the fate of lesioned neurons. Among the molecules which can prevent cell death and promote axonal growth, brain-derived neurotrophic factor (BDNF, Barde et al. 1982; Thanos et al. 1989) and basic fibroblast growth factor (bFGF, Sievers et al. 1987) have so far been shown to influence ganglion cell survival. Similar effects have also been reported after repetitive intraocular injections of large amounts of nerve growth factor (NGF, Carmignoto et al. 1989), even though the numbers of rescued ganglion cells were lower than those obtained with bFGF (Sievers et al. 1987).

The major findings of the present work were that, in addition to the described strategies of rescuing cells with trophic agents and promoting growth with neurotrophic molecules, pharmacological blockade of the enzymes which degrade lesioned neurons, and of axotomy-induced inflammation, can substantially contribute to the retardation of neuronal degeneration. Combined specific transcellular tracing of the neurophagic cells and manipulation of proteolytic activities in the retrogradely degenerating retina document highly specialized mechanisms which are initiated by axotomy and are devoted to the destruction of neurons (=neurophagy). In concert, the interactions between microglial cells and other retinal glial cells like astrocytes and Muller’s cells (Bignami and Dahl, 1979) are probably part of this cascade of responses. Such microglial-macroglial interactions may be mediated by astrocyte-activating molecules, like the interleukins, which stimulate the astrocytes to proliferate and to form the so-called gliosis (Giulian et al. 1989; Giulian, 1987). However, localization of the proteolytic enzymes remains to be demonstrated. The present data permit speculation about the function of the microglial cells, and may indicate that microglial proteases (for a review, see Giulian, 1990) are induced to degrade ganglion cells. Also the observations of Stoll et al. (1989) on the lesioned optic nerve demonstrated that optic nerve macrophages, which are correlated to retinal microglia, use proteases to perform myelinolysis and axonal degradation within the lesioned optic nerve. This would imply that microglial cells either become activated to secrete proteases, or directly attack the lesioned ganglion cells causing them to degrade, and then phagocytose the debris produced. The present data do not distinguish between secretable and cell-associated proteases. Alternatively, the findings are also consistent with the possibility that intraganglionic lysosomal activity can be blocked, thus resulting in retardation of cell death. Less likely, but not impossible, is that protease inhibitors are potent neurotrophic substances directly acting on the ganglion cells. Since, however, protease inhibitors are enzymes that catalyze chemical reactions dependent on specific substrates (Rich, 1986; Powers and Harper, 1986; Schnebli, 1975), the possibility that they also act as neurotrophic agents is less likely. In addition to the described features of protease inhibitors (Schnebli, 1975; Rich, 1986; Powers and Harper, 1986), the only experimental evidence at present opposing this possibility is that administration of protease inhibitors to organ cultures during the explantation procedure does not enhance outgrowth of axons, in contrast to the injection of inhibitors during the procedure of optic nerve transection, which does. An alternative explanation for the site of action of inhibitors is that they block the intracellular, lysosomal proteases which digest already degraded and endocytosed neuronal material. Although possible, such an explanation was not confirmed by the present study, since it documents a blockade of degradation at an earlier level, prior to the endocytosis. The spectrum of proteolytic activities which have been blocked with the inhibitors covers the major proteases used to degrade living cells (see Table 1). The spectrum can certainly be extended to more stable proteases, since pepstatin and leupeptin are less stable (Schnebli, 1975; Rich, 1986; Powers and Harper, 1986). In addition, the dose of protease inhibitors can be optimized, for example by repeated injections of small amounts of protease inhibitors (unpublished data). The combined inhibition of proteases and the promotion of axonal growth with neurotrophic agents and with antiinflammatory substances like corticosteroids in vivo is of considerable importance for the process of regeneration. The mechanism of dexamethasone treatment also remains to be elucidated, although it is widely accepted that dexamethasone treatment is beneficial for recovery of visual function in humans with traumatic optic nerve lesions (Spoor et al. 1990). Dexamethasone may just act as an anti-inflammatory drug, therefore decreasing the proliferative and neurophagic abilities of microglial cells.

Apposition of peripheral nerve pieces at the ocular stump of the optic nerve in order to provide suitable guides for transected axons to grow and reach their targets (Politis and Spencer, 1986; Vidal-Sanz et al. 1987; Villegas-Perez et al. 1988; Thanos and Vanselow, 1990) has helped in documenting the intrinsic ability of lesioned ganglion cells to regenerate their axons and to form synaptic contacts (Keirstead et al. 1989). However, the low numbers of neurons available for recruiting axons to regenerate is one of the circumstances that has hitherto limited restoration of the retinocollicular pathway in adulthood. The critical role of the proteases, and their inhibition combined with suppression of inflammation (Perry and Gordon, 1988; Lampson, 1987) and (micro-) glial activation seem to depend on their being effective soon after nerve lesion. It will be of considerable importance to maintain optimal inhibition of proteases over the critical period of post-traumatic axonal growth, which covers a few weeks (Vidal-Sanz et al. 1987; Thanos and Vanselow, 1990). Exploration of the sequence of interdependent events which are initiated by axotomy and retrograde transport of axotomy-induced signals (Singer et al. 1982) is of crucial importance in determining their reversibility, which might be beneficial for the lesioned neurons. The massive, protracted macroglial response (Bignami and Dahl, 1979; Perry and Gordon, 1988; Streit et al. 1990) and the hypertrophy of microglia (Schnitzer and Scherer, 1990), as well as recruitment of macrophages to carry out myelinolysis and removal of axonal debris in the optic nerve (Stoll et al. 1989), are evidence that these groups of cells are responsible for most of the post-traumatic activities in the retina.

The author thanks Dr Y. A. Barde for providing the BDNF and Dr W. Risau for providing the bFGF used in the cultures. The technical assistance of M. Wild is gratefully acknowledged. T. Rice proof-read the manuscript and the photographic team from our department helped in preparing the photographs. The work was supported by the Deutsche Forschungsgemeinschaft (grant Th 386 2-1).