Mechanisms of Glial Regeneration in an Insect Central Nervous System

A major problem in the study of the mechanisms of recovery from neural damage is the complexity of the ensuing cellular reactions and interactions. These involve a variety of cell types and it is frequently difficult to recognize their relative contributions. Recent research has focused not only on the regeneration of the nerve cells, but also on the role of glia and exogenous cells. Glial cells have, for example, been shown to play a critical role in directing axonal growth in central and peripheral neural grafts (Aguayo, David & Bray, 1981) possibly through their production of extracellular matrix materials and secretory products which appear to be an essential feature of axonal development and regeneration following nerve injury (Bunge & Bunge, 1983; Bunge, 1987). Several studies have also implicated exogenous cells in neural repair, notably the blood monocytes which invade the lesion zone and have the capability of transforming into brain macrophages (Adrian & Schelper, 1981; du Bois, Bowman & Goldstein, 1985). Furthermore, it is also known that blood cells, principally T-lymphocytes, can release glial growth factors (Fontana, Grieder, Arrenbrecht & Grob, 1980; Benveniste et al. 1985) and could, conceivably, play an important role in glial regeneration.

Much of the accelerating progress in this field has been derived from in vitro Studies using cultured neural cells. It is, however, still difficult to relate the findings from such studies to those obtained from in vivo investigations, often involving gross lesioning, in the complexity of the vertebrate brain.

In our recent studies, we have used an invertebrate preparation, the abdominal connectives of the cockroach (Penplaneta americana L.), to investigate some of the basic features of neural repair. Our preparation is relatively simple and well-characterized, consisting only of axons and two basic glial types with their associated extracellular matrices (Fig. 1A). The perineurial glia form a barrier round the central nervous system, excluding small water-soluble cations, such as potassium, from the axonal surfaces (Fig. 1B: Schofield, Swales & Treherne, 1984; Schofield & Treherne, 1984). The experimental situation is further simplified by the technique of selective glial disruption (Smith, Leech & Treherne, 1984). This results in swift and ordered repair, which contrasts with protracted and complex cellular responses observed following surgical lesioning of the connectives (Treherne, Harrison, Treherne & Lane, 1984). Encouragingly, the sequence of glial repair observed under these circumstances bears a number of similarities with the equivalent processes which occur in the vertebrate central nervous system, especially in the role of blood cells and their interactions with reactive endogenous cells.

The abdominal connectives used in our experiments contain only glial nuclei, the neuronal soma being confined to the ganglia. Thus, local application of a DNA-intercalating agent, ethidium bromide (Nelson & Tinoco, 1984), to a mid-portion of the connectives affects only glial nuclei. This compound, which has been previously used to demyelinate mammalian nerve cords (Yajima & Suzuki, 1979; Blakemore, 1982) is taken up from low external concentrations by the cells in cockroach nerve cords (Leech, 1984) and causes extensive glial disruption when applied locally to the connectives in vivo (Smith et al. 1984). This is seen within 24 h by drastic changes in the ultrastructure of the superficial, perineurial glia: notably, a clumping of nuclear chromatin and extensive cytoplasmic damage (Fig. 2A). During this stage there is a breakdown of the perineurial blood–brain barrier as shown by the rapid decline in amplitude of the axonal action potentials, following exposure to high-potassium saline (Fig. 3B), and the penetration of ionic lanthanum into the underlying extracellular system (Fig. 2C). There is also extensive damage to the sub-perineurial and adaxonal glia (Fig. 2B). The ultrastructural appearance of the axons is little modified by the ethidium treatment (Fig. 2B) although, at the light microscope level, axons can appear constricted in diameter and sometimes collapsed. Normal action potentials can, however, be recorded from them at all stages of disruption and repair (Smith et al. 1984). Protein synthesis by the glia, including contributions to the axons and acellular neural lamella, is greatly reduced post-lesion (Smith & Howes, 1984).

A consistent feature of the early stages of repair following selective glial disruption is the appearance of cells containing characteristic electron-dense granules within their cytoplasm (Fig. 3A: Smith et al. 1984). After 2 days, the granule-containing cells begin to form a mosaic of cells and organized layers beneath the neural lamella. By 4 days, the cells in this region have developed extensive, interdigitating cell processes with septate and tight junctional complexes between the membranes of adjacent processes, but cells carrying the electron-dense primary lysosomes have largely disappeared (Smith et al. 1984). At this stage, there is a variable degree of access of potassium ions to the axon surfaces (Fig. 4), although extraneously applied lanthanum is now excluded from the underlying extracellular spaces. Granulecontaining cells are also present at deeper levels among the sub-perineurial tissues. Some of these appear to be engaged in the phagocytosis of cellular debris within the reorganizing perineurium and at deeper levels in the lesion (Smith et al. 1984).

At 6 days, the superficial repairing cells begin to acquire the flattened appearance, and narrow processes extending downwards into the underlying tissues, characteristic of normal perineurial cells. By 11 days, a blood-brain barrier has been reestablished; the glia are similar to those in control connectives (Fig. 5) and exposure of the repairing lesion to high-potassium saline no longer produces a rapid decline in action potential amplitude. However, the electrical properties of the blood–brain interface are still abnormal, in that the potassium-induced extraneuronal potentials, corresponding to the depolarization of the outwardly directed perineurial membranes, are much smaller than in normal connectives (Fig. 6).

After 28 days the ultrastructural appearance of the perineurial and sub-perineurial glia is indistinguishable from that of untreated connectives. There has also been an increase in the extent of the potassium-induced extraneuronal potentials although these are still smaller than those recorded in control connectives. Thus, it seems that, despite the swift re-establishment of the perineurial blood-brain barrier, there is a relatively slow restoration of the original electrophysiological characteristics of the blood–brain interface.

The first question which arises is the origin of the granule-containing cells that are so prominently involved in the initial stages of glial repair. They are unlikely to be endogenous, for no cells containing such primary lysosomes are seen in normal nervous connectives. The granule-containing cells are, however, strikingly similar to the circulating blood granulocytes (Figs 3, 7: Smith et al. 1984), that are only very occasionally seen adhering to the surface of undamaged connectives (Treherne et al. 1984), but which swiftly accumulate on, and penetrate, the neural lamella following chemical lesioning of the underlying glia (Smith et al. 1984; Treherne, Smith & Edwards, 1987b). These and other observations suggest that the granule-containing cells may be of exogenous origin, being derived from circulating haemocytes (Smith, Howes, Leech & Treherne, 1986), a supposition which is confirmed by recent experiments using antibodies raised against haemocytes (E. A. Howes, B. M. Chain, P. J. S. Smith & J. E. Treherne, unpublished observations). In these experiments, no cell within normal abdominal connectives binds monoclonal antibodies raised against cockroach haemocytes. Monoclonally labelled cells do, however, appear in the disrupted perineurial and sub-perineurial tissues within 24 h of selective glial disruption (Fig. 8). As with the granule-containing cells they increase in number over the first 1–3 days post-lesion and then decline.

Clearly, therefore, this category of reactive cell is of exogenous origin, invading the lesion zone in the early stages of repair to play an important role, not only in phagocytosis, but also in structural replacement of disrupted glia.

The exogenous, haemocytotic cells appear to work in concert with, and perhaps orchestrate, endogenous reactive cells. The involvement of such endogenous cells is indicated by several lines of evidence. For example, we have shown that glial cells are capable of division and repair following selective disruption in cultured nerve cords, in which there can be no haemocyte involvement and, consequently, in which granule-containing cells are never seen in the lesioned tissues (Howes, Smith & Treherne, 1987a,b). Glial repair in such in vitro preparations is slower and less organized than in vivo. Nevertheless, restructuring does occur. Initially, patches of viable cells appear among the disrupted glia eventually extending to form a continuous covering over the repairing lesion; aberrant repair of the sub-perineurial glia also occurs. Essentially similar sequences of repair occur when the appearance of granule-containing cells in the lesion is reduced by the DNA-scission drug bleomycin (Treherne, Howes, Leech & Smith, 1986) or by injection into the blood of physiologically inert particles (fluorescent microspheres), which are taken up by a relatively small proportion of the haemocytes (approx. 10%) (Fig. 8C) and thus prevented from entering the lesion area following selective glial disruption (Smith et al. 1986).

The participation of endogenous reactive cells is also indicated by the substantial increase in cell numbers which occurs in undamaged tissues adjacent to the lesion zone following selective glial disruption (Treherne et al. 1987b). This increase is unlikely to have resulted from invading haemocytes, for granule-containing cells are rarely seen adjacent to the lesion zone, either in electron microscope sections or in monoclonal antibody-binding studies (E. A. Howes, B. M. Chain, P. J. S. Smith & J. E. Treherne, unpublished observations).

An increase in cell numbers within repairing connectives is first seen in the perineurial region 2 days after selective glial disruption, and is confined to the lesion zone (Fig. 9). This stage corresponds to the phase of haemocyte invasion into the lesioned tissues.

The increase in perineurial cell numbers in the undamaged tissues starts 3 days after selective lesioning (Fig. 9). It is then seen as a separate peak, anterior to the lesion and close to the fourth abdominal ganglia. This secondary peak is still evident after 5 days, with the appearance of having moved towards the lesion zone. By 7 days, the peaks appear to have fused and are now confined to the site of original lesion (Fig. 9). This increase in cell numbers is maintained in the lesion zone over the next 1–2 months, with further multiplication in the adjacent areas of the connectives (Treherne et al. 1987b).

An increase in cell numbers is also a feature of sub-perineurial repair following selective disruption, although the anterior polarity seen in the perineurium is less evident. As with perineurial repair, the increased numbers are maintained 1 month after selective lesioning (Treherne, Smith & Edwards, 1987b).

The above results are consistent with the notion that glial repair in penultimate cockroach connectives involves exogenous and endogenous elements: the haemocytes, which invade the lesioned tissues, and the reactive glia, that originate anterior to the damaged area, perhaps in the vicinity of the fourth ganglion, and migrate into the lesion.

We also have evidence that haemocytes play a critical role in the recruitment of the reactive glia. Thus, if haemocytes are excluded from the lesion (by prior injection and uptake of fluorescent microspheres, Fig. 8C) then there is no activation of the reactive glia equivalent to that seen in control lesion experiments (Fig. 10A,B). Similarly, in experiments with cultured cords (Howes et al. 1987a) and in the presence of bleomycin in vivo (Treherne et al. 1987b), there is no equivalent activation of endogenous reactive cells (Fig. 11A,B). Significantly, in the experiments with bleomycin (a drug which perturbs cell function and division: Kuo, 1981) there is a clear reduction of cells in the lesion area following selective glial disruption. It is unlikely that the activation of the reactive glia results from direct contact stimulus, as when, for example, growing neurites initiate division of cultured Schwann cells (Bunge & Bunge, 1983), for our studies show that the monoclonally labelled haemocytes appear to be confined to the damaged tissues. A remaining possibility is that the activation of the endogenous reactive cells is triggered by the release of diffusible growth factors from the invading haemocytes within the repairing connective.

We are, at present, still ignorant of the identity of the reactive glia that, in our preparation, appear to originate within, or in the vicinity of, the anterior ganglion and which impose an anterior polarity on the initial cellular responses to selective glial disruption. They could be uncommitted progenitor cells, such as have been (hown to exist in the vertebrate CNS (Raff, Miller & Noble, 1983) or, alternatively, the derived from existing glia, perhaps after dedifferentiation as has been postulated for the regenerating amphibian limb (Brockes, 1984).

The invading haemocytes, which so swiftly enter the lesioned tissues following selective glial disruption, clearly serve an important role in the early stages of repair, not only in initiating the recruitment of endogenous reactive cells, but also in the initial structural replacement of damaged glia. This is most clearly seen in their role in restructuring the superficial glia which constitute the perineurial blood-brain barrier. However, this process involves marked changes in the morphology of the repairing cells. There is also a progressive loss of cells containing characteristic cytoplasmic granules 2–4 days after selective glial disruption (Fig. 12). Furthermore, these changes are accompanied by an onset of cell division in the repairing perineurium, as indicated by the increasing number of nuclei labelled with tritiated thymidine (Fig. 12: Smith & Howes, 1987).

There are two possible explanations for these events: the granule-containing cells could be transformed or they could be replaced by cells of glial origin (Howes et al. 1987b; Treherne et al. 1987a). Both possibilities would account for the increase in nuclear thymidine labelling, for transformation could involve the division of granulecontaining cells, while replacement by reactive glia is also likely to be achieved by division.

Despite the apparently swift initial repair of selectively lesioned connectives, the cellular reorganization is, in fact, a protracted business. This is seen in the relatively slow restoration of the normal potassium permeability of the blood-brain interface. Thus, even after 28 days the potassium-induced extraneuronal potentials do not attain their original amplitude, though the blood-brain barrier is re-established within 14 days (Smith et al. 1984).

The early increase in cell numbers is also continued in the apparently fully repaired lesion. This is most clearly seen in the sub-perineurial glia. Here, the cell numbers at 28 days are more than twice those at 7 days. Furthermore, the increase in cell numbers now tends to spread beyond the limits of the original lesion (J. E. Treherne, P. J. S. Smith & E. A. Howes, in preparation). A similar long-term increase in glial cell numbers has been observed in cultured leech central nervous connectives after lesioning by crushing (Morgese, Elliott & Muller, 1983). A substantial increase in cell numbers also occurred, after 1 week, in regions adjacent to the lesion in the leech preparation.

Unlike the situation following surgical lesioning, the glial proliferation in the chemically lesioned cockroach connectives occurs in an existing compartment defined by the extracellular matrices and axons, so that substantially more cells are now deployed into what is effectively the same, limited, domain. That they proliferate to this extent implies the existence of powerful and sustained mitogenic signals within the repairing system which are capable of activating cell division in existing G-phase cell populations. If the mitogenic signals are powerful enough then they could override the contact-inhibition of cell division which, we presume, must arise as the available space is filled. Under these circumstances, proliferation in a restricted environment could result in an increased population of smaller cells. Alternatively, the increase in the number of glial cells in the repairing lesion could be merely a consequence of the recruitment of a population of progenitor cells which are smaller than those of the original neuroglia.

Our results show that glial repair in the insect CNS can be divided into three separate phases. The first involves an initial invasion of the lesion by circulating haemocytes. These accumulate on the surface of the nervous system within 24 h of selective glial disruption, penetrate the neural lamella and then transform into a novel cell class, the granule-containing cells, which are never seen in undamaged connectives. The granule-containing cells have three major roles: a limited involvement in phagocytosis, structural replacement of damaged glia and activation of endogenous reactive cells. The second phase involves a reduction in the number of granule-containing cells, proliferation of reactive glia from the vicinity of the anterior abdominal ganglion and restoration of the perineurial blood-brain barrier. The final phase is characterized by a progressive, and then maintained, increase in the numbers of perineurial and sub-perineurial glia within the lesion site and, subsequently, in the adjacent undamaged tissues.

This sequence bears a number of striking similarities to the equivalent events which occur in the repair of the vertebrate central nervous system. A notable similarity is the initial invasion of the lesion by haemocytes which parallels the entry of blood monocytes into vertebrate brain tissues following surgical damage (Adrian & Schelper, 1981). The timing of these events can be closely similar. As with the insect haemocytes, the vertebrate blood monocytes arrive within 2 days of lesioning and by 5 days transform to macrophage cells (du Bois et al. 1985): a timing which corresponds to the transformation/replacement of the insect granule-containing cells. Cell division of the astrocytes commences at 2 days and continues for 6–7 days, which coincides with the onset of thymidine labelling of the perineurial glia in the insect preparation.

The role of the haemocytes in triggering glial cell recruitment again parallels events in vertebrate systems, where it is known that blood cells, principally T-lymphocytes, can release glial growth factors (Fontana et al. 1980; Benveniste et al. 1985). The presence of glial mitogenic and morphogenic factors in the vertebrate central nervous system (Giulian & Baker, 1985; Giulian & Young, 1986; Giulian, Tomozawa, Hindman & Allen, 1985; Giulian, Allen, Baker & Tomozawa, 1986) also indicates the possibility of endogenous activation of glial proliferation. Equivalent activators in the insect CNS could provide the stimulus for the prolonged glial proliferation and might themselves be initially triggered by the arrival of the exogenous reactive cells, the haemocytes. This would be analogous to the situation in vertebrates. Mammalian microglia, which are probably derived during development from monocytic stem cells (see Rio-Hortega, 1932; Perry, Hume & Gordon, 1985), also produce growth factors that act selectively on populations of macroglia (Giulian & Baker, 1985).

This work was partly supported by the Leverhulme Trust.