Multiple ependymal structures have been observed in regenerating spinal cord of the teleost Apteronotus albifrons. Evidence is presented for two modes of formation of the secondary ependymas: budding off from the primary ependyma, and de novo origin of a tube-like ependymal structure within a group of undifferentiated cells. Serial sections of regenerated cord provide evidence that undifferentiated cells not in immediate contact with the main ependymal layer can organize and differentiate into an ependymal structure in the regenerating spinal cord. These findings suggest that a significant amount of morphological organization can take place independent of the normal developmental sequence and environment.

The ependyma, or layer of cells surrounding the central canal in the vertebrate central nervous system, is of interest to developmental biologists because of its role in development and in the adult individual. In the embryo this layer, termed the generative neuroepithelium (Jacobson, 1978), is responsible for producing all the cells (both neuronal and non-neuronal) of the growing spinal cord. In the adult vertebrate, the cells lining the central canal (the ependymal cells) have differentiated specialized junctional complexes and cilia at their luminal ends, and may function in regulating the content and circulation of the cerebrospinal fluid.

It has been hypothesized that the ependymal layer in postnatal vertebrates may still harbour undifferentiated ‘stem cells’ which retain a relatively unrestricted potential for mitosis and differentiation under the appropriate circumstances. In fact, production of new cells has been demonstrated in postnatal rat brain (Altman & Das, 1965; Kaplan & Hinds, 1977), in adult rodent olfactory bulb (Graziadei & Graziadei, 1979; Wilson & Raisman, 1980), and in certain brain nuclei in adult birds (Goldman & Nottebohm, 1983). The newly produced cells in these instances include neurones which are recruited into functional circuits (Easter, 1983; Paton & Nottebohm, 1984), glial cells, and ependymal cells which form the wall of the lengthening ependymal tube. The ultimate source of the new cells appears to be the ependymal layer, since studies examining [3H]thymidine incorporation after a series of intervals (e.g. Altman & Das, 1965; Goldman & Nottebohm, 1983) show labelled cells first in the ependyma or ventricular layer, then later in other brain layers. Thus there is evidence that, under certain circumstances, some of the cells in the ependymal layer of normal adult vertebrates retain the ability to divide and produce new neurones and non-neuronal cells.

The ependyma plays an important role in the regeneration of spinal cord in lower vertebrates (Hughes & New, 1959; Kiortsis & Droin, 1961; Butler & Ward, 1965; Simpson, 1964, 1968, 1983). In the regenerating lizard tail, the presence of ependyma and contact with it have been shown to be necessary in order for spinal cord regeneration to proceed (Kamrin & Singer, 1955). In teleosts and amphibians (Anderson & Waxman, 1981, 1983a; Egar & Singer, 1972; Nordlander & Singer, 1978), ependymal cell mitoses generate the new neurones, glia and ependymal cells of the spinal cord that regenerates after amputation of the tail. In the teleost Apteronotus albifrons* (Black Ghost Knife fish) (Fig. 1), regeneration of new spinal cord can occur from ependyma at levels up to halfway rostral along the rostrocaudal axis of the fish (Anderson, Waxman & Laufer, 1983). Such a dramatic case of regeneration is possible because in this gymnotiform teleost all of the gut and major organs are located in the rostral third of the fish; the caudal two-thirds of the fish’s length contains spinal column, electric organ and muscle. The regeneration of relatively complete spinal cord from rostral nongrowing ependyma in Apteronotus indicates that histologically mature, stable ependyma can be stimulated by the amputation (injury) to generate a new ependyma, which in turn generates new neurones, glia and ependymal cells (Anderson et al. 1983; Anderson, Waxman & Tadlock, 1984).

Fig. 1.

Apteronotus albifrons with 9-month regenerated tail. Apteronotus is a weakly electric gymnotiform teleost. The body cavity and organs are located in the first third of the fish, with the anus situated ventrally just behind the mouth. The tail and spinal column of this species readily regenerate in the wild as well as in the laboratory. The previous amputation site in this individual (arrow) is clearly visible as a notch in the top of the tail. Magnification reduced from life size by 30 %.

Fig. 1.

Apteronotus albifrons with 9-month regenerated tail. Apteronotus is a weakly electric gymnotiform teleost. The body cavity and organs are located in the first third of the fish, with the anus situated ventrally just behind the mouth. The tail and spinal column of this species readily regenerate in the wild as well as in the laboratory. The previous amputation site in this individual (arrow) is clearly visible as a notch in the top of the tail. Magnification reduced from life size by 30 %.

In Apteronotus, a complete spinal structure, including new nerve cell bodies, neurosecretory cells, axons, and glial components, is regenerated after amputation of the tail (Anderson & Waxman, 1981, 1983a). As in regenerated spinal cord of amphibians and reptiles (Nordlander & Singer, 1978; Egar, Simpson & Singer, 1970), new cells in regenerating Apteronotus spinal cord are produced from the ependymal layer in a time-dependent fashion. That is, there is a rostrocaudal gradient of cell production and of differentiation within the regenerating cord (Anderson et al. 1983). New cells are generated from ependyma at the caudal extremity of the cord; differentiation proceeds in a rostral-to-caudal direction, so that the rostral portion of regenerated cord contains the ‘oldest’ and most differentiated tissue. In Apteronotus, the processes of differentiation and morphogenesis within the regenerating tissue create a complete spinal cord: Apteronotus cord that has regenerated for relatively long time periods (12–29 months) is indistinguishable in morphology from normal cord (Anderson et al. 1984; Anderson & Waxman, 1985b).

These studies on regeneration demonstrate the capacity of the intact, adult ependyma of lower vertebrates to generate new spinal neurones and glia in vivo and to direct the organization of a new (regenerating) ependymal and spinal axis. However, several questions arise concerning the developmental potential of these ependymal cells and the relationship between the intact structure of the ependyma and the subsequent morphogenesis of the spinal cord produced during regeneration. One important question concerns whether development of regenerated ependymal structures depends on contact with the pre-existing ependymal tube, which could serve as a template or induction element, or can occur independently as a result of inherent properties of the stem cell population. Recent observations on regenerating spinal cord of Apteronotus albifrons have allowed us to address this issue and suggest that the cells generated by the ependyma in this species exhibit an inherent capability for selforganization into differentiated ependymal tubes in the absence of direct contact with a pre-existing ependyma.

Adult Apteronotus albifrons, more than 12 cm in length, were used in this study. Spinal cord regeneration was initiated by amputation of the tail. Fish were first anaesthetized in a 1:15 000 solution of tricaine methane sulphonate (MS222; ‘Finquel’, Ayerst Labs). After amputation of the tail with a razor blade, the fish was returned to its tank and maintained at 26–28 °C. After the regeneration period (generally 3–12 months), the fish was reanaesthetized, the regenerated tail was cut off several mm rostral to the previous amputation site, and the spinal cord was dissected out in fixative.

The previous amputation site is clearly visible as a notch in the dorsal aspect of the fish’s tail (Fig. 1), accompanied by a 2–6mm decrease in size of the tail along the dorsal-ventral axis. In addition, scales covering the regenerated portion of tail are half the size (or less) of the unregenerated scales. Internally at the amputation site there is a tough fibrous scar and the bone of the spinal column forms a very tough, unsegmented mass.

After the amputation the large blood vessels beneath the spinal column were held with a haemostat until bleeding stopped (30–120 s), then the fish was returned to its tank. Triple sulfa (Dyna-Pet, Inc.; 100mg per 5 gallon tank) was added to the tank water as a preventive; in only one case out of over 100 operations has an infection developed after tail amputation. The wound was not sutured because sutures usually pull out within hours, due to the constant undulating motion of the fish (and because it was not necessary). In this fashion, fish can repeatedly regenerate a new tail and spinal cord after repeated amputations. Some of the fish in our laboratory have completed four or five such amputation/regeneration cycles.

The results presented in this paper are based on detailed examination of three fish that had regenerated multiple ependymas. The cord presented in Figs 2,3,4,6 and 7 had regenerated for 10 months and represented the third regeneration cycle for that fish (the tail had been cut and regenerated three times previously). The cord presented in Figs 5 and 8 was from a fish that had regenerated for 20 months and was in its fourth regeneration cycle. The third fish had regenerated for 12 months and was in its fourth regeneration cycle.

Fig. 2.

Diagram of a 10-month regenerated cord and its subsidiary ependymas. This regenerated cord has subsidiary ependymas both rostral and caudal to a narrow middle section. In the rostral portion, four of the subsidiary ependymas (shown in detail in Figs 3, 4) are located close to the peripheral border of the cord and have no direct connection with the major ependyma. A fifth subsidiary ependyma is present, in continuity with the major ependyma. In the caudal portion of this regenerated cord, several pieces of spinal tissue split off from the major part of the cord. The subsidiary ependymas (13) which are observed in these offshoots of spinal tissue are shown in detail in Figs 6, 7.

Fig. 2.

Diagram of a 10-month regenerated cord and its subsidiary ependymas. This regenerated cord has subsidiary ependymas both rostral and caudal to a narrow middle section. In the rostral portion, four of the subsidiary ependymas (shown in detail in Figs 3, 4) are located close to the peripheral border of the cord and have no direct connection with the major ependyma. A fifth subsidiary ependyma is present, in continuity with the major ependyma. In the caudal portion of this regenerated cord, several pieces of spinal tissue split off from the major part of the cord. The subsidiary ependymas (13) which are observed in these offshoots of spinal tissue are shown in detail in Figs 6, 7.

Fig. 3.

Multiple ependymas in rostral regenerated spinal cord. (A) A transverse section, located approximately 0·9 mm caudal to the transition zone (amputation site) showing three subsidiary ependymal structures (labelled 1, 3 and 4) and an area of undifferentiated cells (labelled 2) which develops an ependymal structure in a nearby section (see Fig. 4). The major ependyma (e) is located in the ventral portion of this section, as sometimes occurs in regenerated cord, and the subsidiary ependymas are located dorsal to it. Large diameter myelinated axons (ax) are also prominent in the dorsal part of the cord. Groups of axons growing in connective tissue near the cord, as at left here, are common in regenerating Apteronotus cord. Magnification x440. Bar, 50 µm. (B) Higher power of the subsidiary ependyma no. 4, showing lumen, cilia, basal bodies, and junctional specializations. Magnification ×6500. Bar, 5 μm.

Fig. 3.

Multiple ependymas in rostral regenerated spinal cord. (A) A transverse section, located approximately 0·9 mm caudal to the transition zone (amputation site) showing three subsidiary ependymal structures (labelled 1, 3 and 4) and an area of undifferentiated cells (labelled 2) which develops an ependymal structure in a nearby section (see Fig. 4). The major ependyma (e) is located in the ventral portion of this section, as sometimes occurs in regenerated cord, and the subsidiary ependymas are located dorsal to it. Large diameter myelinated axons (ax) are also prominent in the dorsal part of the cord. Groups of axons growing in connective tissue near the cord, as at left here, are common in regenerating Apteronotus cord. Magnification x440. Bar, 50 µm. (B) Higher power of the subsidiary ependyma no. 4, showing lumen, cilia, basal bodies, and junctional specializations. Magnification ×6500. Bar, 5 μm.

Fig. 4.

Serial sections of subsidiary ependymas from Fig. 3. The sections A to H progress in a rostral-to-caudal direction and cover 65 μm along the rostrocaudal axis of the cord. The total length of cord covered by the subsidiary ependymal tubes is 350–400 μm. The section in G corresponds to Fig. 3A. Boxes mark the significant areas (nos 14 from Fig. 3). The length of cord from A to H is 65 μm. Areas adjacent to the closing-off or opening-up of an ependymal structure (e.g. no. 2 and no. 3 in C or no. 4 in H) exhibit undifferentiated cells with large, clear nuclei. Magnification X650. Bar, 50 μm.

Fig. 4.

Serial sections of subsidiary ependymas from Fig. 3. The sections A to H progress in a rostral-to-caudal direction and cover 65 μm along the rostrocaudal axis of the cord. The total length of cord covered by the subsidiary ependymal tubes is 350–400 μm. The section in G corresponds to Fig. 3A. Boxes mark the significant areas (nos 14 from Fig. 3). The length of cord from A to H is 65 μm. Areas adjacent to the closing-off or opening-up of an ependymal structure (e.g. no. 2 and no. 3 in C or no. 4 in H) exhibit undifferentiated cells with large, clear nuclei. Magnification X650. Bar, 50 μm.

Fig. 5.

Migration and reorganization of cells from the ependymal layer in a 20-month regenerated cord. This fish was in its fourth regeneration cycle. The series of thick sections (A to C) progresses in a rostral-to-caudal direction and covers a length of 270μm along the longitudinal axis of the cord. Cells of the ependymal layer of this regenerated cord stained darker (with toluidine blue) than the other cells. In A, darkly staining cells (arrowheads) can be observed apparently migrating away from the ependymal layer. Edge of cord (arrow) is visible at upper left, in the same orientation as in B and C. In B, a section 230 μm caudal to A, where the cord is noticeably smaller in diameter than in A, two groups of darkly staining cells (arrows) are present near the dorsal border. In C, a section 40µm caudal to B, a lumen (arrow) and ependymal structure are clearly discernible within each of the two groups of dark cells. Magnification ×295. Bar, 50μm.

Fig. 5.

Migration and reorganization of cells from the ependymal layer in a 20-month regenerated cord. This fish was in its fourth regeneration cycle. The series of thick sections (A to C) progresses in a rostral-to-caudal direction and covers a length of 270μm along the longitudinal axis of the cord. Cells of the ependymal layer of this regenerated cord stained darker (with toluidine blue) than the other cells. In A, darkly staining cells (arrowheads) can be observed apparently migrating away from the ependymal layer. Edge of cord (arrow) is visible at upper left, in the same orientation as in B and C. In B, a section 230 μm caudal to A, where the cord is noticeably smaller in diameter than in A, two groups of darkly staining cells (arrows) are present near the dorsal border. In C, a section 40µm caudal to B, a lumen (arrow) and ependymal structure are clearly discernible within each of the two groups of dark cells. Magnification ×295. Bar, 50μm.

Fig. 6.

Selforganization of subsidiary ependymas. The transverse sections shown in A to F proceed in a rostral-to-caudal direction along the caudal portion of a 10-month regenerated cord. (A) Two areas of tissue, 1 and 2, beginning to split off (arrows) from the major part of the cord (at right). Some large myelinated axons (ax) are present, as are undifferentiated cells (arrowheads) streaming out from the major ependyma of the cord. (B) Areas 1 and 2 have completely separated from the major part of the cord. (C) A distinct subsidiary ependyma is present in area 2 (arrow). Electron microscopy (see Fig. 7) reveals an ependyma forming in area 1 also. The major part of the cord is not shown in D-F. (D) The subsidiary ependymas in areas 1 and 2 are clearly visible. A third area of extraspinal tissue contains cells with large nuclei. Typical neuro-secretory tissue (ns) extends from areas 1 and 2. (E) The ependymal tubes in areas 1 and 2 continue. An ependymal structure has also formed in area 3. A bridge of tissue has merged areas 1 and 3 in this section. (F) The three subsidiary ependymas continue, parallel to the longitudinal axis of the major part of the cord (not shown). (A,B) Magnification ×420. Bar, 50μm. (C) Magnification ×520. Bar, 50μm. (D-F) Magnification ×310. Bar, 50μm.

Fig. 6.

Selforganization of subsidiary ependymas. The transverse sections shown in A to F proceed in a rostral-to-caudal direction along the caudal portion of a 10-month regenerated cord. (A) Two areas of tissue, 1 and 2, beginning to split off (arrows) from the major part of the cord (at right). Some large myelinated axons (ax) are present, as are undifferentiated cells (arrowheads) streaming out from the major ependyma of the cord. (B) Areas 1 and 2 have completely separated from the major part of the cord. (C) A distinct subsidiary ependyma is present in area 2 (arrow). Electron microscopy (see Fig. 7) reveals an ependyma forming in area 1 also. The major part of the cord is not shown in D-F. (D) The subsidiary ependymas in areas 1 and 2 are clearly visible. A third area of extraspinal tissue contains cells with large nuclei. Typical neuro-secretory tissue (ns) extends from areas 1 and 2. (E) The ependymal tubes in areas 1 and 2 continue. An ependymal structure has also formed in area 3. A bridge of tissue has merged areas 1 and 3 in this section. (F) The three subsidiary ependymas continue, parallel to the longitudinal axis of the major part of the cord (not shown). (A,B) Magnification ×420. Bar, 50μm. (C) Magnification ×520. Bar, 50μm. (D-F) Magnification ×310. Bar, 50μm.

Fig. 7.

Electron microscopy of ependymal formation in area 1 of Fig. 6. (A) A focus of junctional specializations (arrows), desmosomes, and cilia (c) are present in this thin section of area 1 taken at the level of Fig. 6C. Magnification × 12000. Bar, 1μm.

(B) Ependymal lumen is present in this section approximately 40μm caudal to A. Some of the cells surrounding the lumen have a more electron-dense cytoplasm than others (lower left), which are quite electron lucent. Magnification ×6200. Bar, 50μm.

The ependyma in area I in a thin section taken at the level of Fig. 6F. A complete ependymal structure is present. Magnification ×3900. Bar, lgm. (D) A higher power of C showing the lumen, cilia, and junctional complexes joining ependymal cells. Magnification × 15 000. Bar, 1μm.

Fig. 7.

Electron microscopy of ependymal formation in area 1 of Fig. 6. (A) A focus of junctional specializations (arrows), desmosomes, and cilia (c) are present in this thin section of area 1 taken at the level of Fig. 6C. Magnification × 12000. Bar, 1μm.

(B) Ependymal lumen is present in this section approximately 40μm caudal to A. Some of the cells surrounding the lumen have a more electron-dense cytoplasm than others (lower left), which are quite electron lucent. Magnification ×6200. Bar, 50μm.

The ependyma in area I in a thin section taken at the level of Fig. 6F. A complete ependymal structure is present. Magnification ×3900. Bar, lgm. (D) A higher power of C showing the lumen, cilia, and junctional complexes joining ependymal cells. Magnification × 15 000. Bar, 1μm.

Fig. 8.

A section from the 20-month regenerated cord shown in Fig. 5, approximately 500 μ m caudal to Fig. 5C. A group of lightly staining cells (arrows) is present in one area of the major ependyma. Other cells appear to be streaming from this to a subsidiary ependyma (sub) located at the peripheral border of this cord. Other undifferentiated cells with large nuclei (arrowheads) are also observed around the abluminal border of the ependyma. One cell with condensed chromatin (bold arrow), most probably in the process of mitosis, has bulged into the central canal. Magnification ×1050. Bar, 10 μm.

Fig. 8.

A section from the 20-month regenerated cord shown in Fig. 5, approximately 500 μ m caudal to Fig. 5C. A group of lightly staining cells (arrows) is present in one area of the major ependyma. Other cells appear to be streaming from this to a subsidiary ependyma (sub) located at the peripheral border of this cord. Other undifferentiated cells with large nuclei (arrowheads) are also observed around the abluminal border of the ependyma. One cell with condensed chromatin (bold arrow), most probably in the process of mitosis, has bulged into the central canal. Magnification ×1050. Bar, 10 μm.

Spinal cord was dissected in a solution of 2·5% glutaraldehyde in 0·lM-sodium phosphate buffer, with 0·08M-sucrose (pH7-3; 300mOs). After dissection, the cord remained in fixative (at room temperature) for 2·3 h. The tissue was then washed in four changes of 0·1 M-phosphate buffer (with 0-08 M-sucrose) over a period of 1 h, and postfixed in 2 % OsO4 in phosphate buffer without sucrose (1 h, room temperature). Tissue was washed for 2 h in 0·2M-acetate-acetic acid buffer (pH 5) and stained en bloc with 2% uranyl acetate. Tissue was embedded in Epon-Araldite and observed with light microscopy and in a JEOL 100 CX electron microscope. In some cases, serial sections were taken, and serial thin sections were examined in the electron microscope.

Multiple ependymas

Multiple ependymal tubes are sometimes observed within regenerated spinal cord of Apteronotus. These extra ependymal tubes have been studied in three individual fish, each of which was regenerating after three or more amputation/ regeneration cycles (described in detail in Materials and Methods). Along the rostrocaudal axis in these regenerated cords, subsidiary ependymas were most frequently observed near the transition zone between regenerated and unregenerated cord (i.e. in rostral regenerated cord) and near the caudal tip. Among the three different cords, 25 separate subsidiary ependymas were observed. These could be subdivided into three groups: those which were clearly formed by budding from the major ependyma (five cases, observed in three different fish); those which appeared to be formed by selforganization (ten cases, observed in two different fish), and those which were of indefinite origin (ten cases, observed in three fish).

The diagram in Fig. 2 schematically depicts one regenerated cord and its subsidiary ependymas. This 10-month regenerated cord, from which Figs 3, 4, 6 and 7 were taken, extended for 4·5 mm. It represents the third cycle of amputation/ regeneration for this fish, i.e. the tail had been amputated three times previously. Several subsidiary ependymal tubes were observed in the rostral portion of the regenerated cord, and several others occurred near the caudal tip. Of the extra ependymal tubes in the rostral regenerated cord, all occurred within the major axis of the regenerated cord. This was invariably the case with all the cords we examined: the subsidiary ependymas in rostral cord were located within the major part of the cord. In the caudal region of regenerated cord, however, subsidiary ependymas frequently occurred in pieces of spinal tissue that had separated from the major portion of the cord (as in Fig. 2), as well as within the major portion of the cord. The diagram in Fig. 2 shows the areas of cord from which subsequent figures were taken. It also shows a fifth subsidiary ependyma in rostral regenerated cord (not shown in the subsequent figures) which was formed by budding or direct splitting from the major ependyma.

Fig. 3A shows a transverse section from the rostral portion of the 10-month regenerated cord depicted in Fig. 2. Three subsidiary ependymas can be observed close to the peripheral border of the cord. One additional area (number 2 in Figs 3, 4) shows a subsidiary ependyma at a point approximately 55 μm rostral to this section. These four subsidiary ependymas are tube-shaped cavities which lie parallel to the longitudinal axis of the cord. At each end, these subsidiary ependymal tubes terminate in a group of undifferentiated cells. At the time examined, each of these rostral subsidiary ependymas is totally independent of connection with the main ependyma of the cord.

Examination by electron microscope (Fig. 3B) confirms that the subsidiary structures are true ependymas, having a lumen, typical junctional complexes, and cilia and microvilli extending into the luminal cavity. The ultrastructure of the cilia appears identical to that of normal cilia. The cilia in secondary ependymas emanate from typical basal bodies, and striated fibres can sometimes be observed projecting from the basal body further into the cytoplasm, as in normal ependyma. However, in contrast to normal cord where cilia are restricted to the ependymal cells, cilia in regenerating spinal cord are also observed at distances of up to 65 µm from the ependymal lumen. In regenerating cord, cilia are observed not only on the ependymal cells, but also on cells which contain dense-cored vesicles.

In Fig. 4, serial sections through the rostral region depicted in Figs 2, 3 show that the four peripherally located subsidiary ependymas are the tubular structures, running parallel to the longitudinal axis of the cord. The termination of each ependymal tube in a group of relatively undifferentiated cells with large nuclei can be observed (Fig. 4A-D,H). In a cross-sectional view of the cord, the peripherally located subsidiary ependymas in Fig. 4 are 140– 180 μm from the major ependyma.

Selforganization of ependyma

Fig. 5 shows the migration and reorganization of ependymal cells in a series of thick sections through a 20-month regenerated cord that represented the fourth regeneration cycle of that fish. In this regenerated cord, the ependymal layer, and presumably cells derived from it, stained more darkly with toluidine blue than the rest of the cells in the cord. In Fig. 5A, darkly stained cells can be observed migrating away from the ependymal layer. In a more caudal section (Fig. 5B), the darkly stained cells are condensed into two discrete groups near the periphery of the cord. Subsequent (more caudal) sections (Fig. 5C) show a central canal and typical ependymal structure within each group of darkly stained cells.

In a second series (Fig. 6), serial thin sections from the caudal region of the regenerated cord illustrated in Fig. 2 document in detail the existence of three subsidiary ependymas within tissue that has separated from the major spinal cord. These subsidiary ependymas have no physical contact with the major ependyma at the time when this cord was fixed. Moreover, serial sections rostral to the series in Fig. 6 (not shown) depict a migration of undifferentiated cells from the major ependyma to the areas of tissue that split off from the major part of the cord (and in which the subsidiary ependymas are located), very similar to the migration of cells depicted in Fig. 5.

In Fig. 6A, two areas of tissue are starting to separate from the major part of the cord. This tissue contains axons, undifferentiated cells and neurosecretory cells and processes. Neurosecretory tissue is normally found in caudal spinal cord of Apteronotus (Anderson & Waxman, 1983b), and other teleosts (Bern & Takasugi, 1962), as well as in regenerating Apteronotus spinal cord (Anderson et al. 1983). The splitting off of spinal tissue, especially neurosecretory tissue, is common in the caudal area of regenerating spinal cord of Apteronotus. Following the serial sections in the rostral-to-caudal direction (Fig. 6A-F), subsidiary ependymas are formed in the two pieces of spinal tissue marked 1 and 2 (Fig. 6C). More caudally (Fig. 6E), a third subsidiary ependyma forms in a third area of extraspinal tissue close to areas 1 and 2. These three subsidiary ependymas continue as well-defined ependymal tubes (Fig. 6F), parallel to the major cord, and extend to the end of the tail.

The organization of the subsidiary ependymal structures in Fig. 6 was also followed by electron microscopy (Fig. 7). In rostral sections (at the level of Fig. 6B), the first indication of ependymal structure is the appearance of cilia between the cells. Next (Fig. 7A), a radial array of junctional specializations and cilia is observed. Sections caudal to this (Fig. 7B-D) show that a lumen and the typical ependymal structure are present. Due to the distance involved between the location of these subsidiary ependymas (Fig. 6, 7) and the major ependyma, their location in spinal tissue that has separated from the major portion of the cord, and the streaming of undifferentiated cells to these areas (more rostrally), we believe that these results strongly suggest that these subsidiary ependymas formed in the absence of contact with the major ependyma.

Budding from the major ependyma

The formation of subsidiary ependymas by budding or splitting from the major ependyma has also been observed in regenerating Apteronotus spinal cord. In these cases, a subsidiary ependymal tube forms directly from the major ependyma. It may remain in continuity with the major ependyma (as in Fig. 2) or may pinch off creating a separate ependymal tube. Five clear instances of the generation of subsidiary ependymas by budding have been observed in three individual fish. These instances occurred in both caudal and rostral regenerated cord.

General observations on ependymal structure

In both the major ependyma and in the secondary ones, differences in staining properties and electron density have been observed between ependymal cells. Some of the ependymal cells stain darker than average, while others stain much more lightly. Correspondingly, at the electron-microscope level, some of the ependymal cells have electron-dense cytoplasm, while others have cytoplasm which is very electron lucent. Differences in the density of regenerating ependymal cells have also been noted by Egar & Singer (1972) in the rostral part of regenerating amphibian spinal cord. In several instances in regenerating Apteronotus spinal cord, we have observed a group of lightly staining ependymal cells in continuity with other lightly staining undifferentiated cells that were streaming away from the major ependyma (Fig. 8). The differential staining properties of these ependymal cells may relate to the cellular processes that underlie the production of new cells, or may relate in some other way to differences in developmental potential or fate of these cells. Fig. 8 also shows a cell in the ependymal layer with condensed chromatin, which is most probably in the process of mitosis. The body of this apparently mitotic cell has bulged into the cavity of the ependymal lumen. This cell is located close to the focus of lucent ependymal cells at an apparent source of migration from the ependymal tube.

The results presented here show that multiple ependymas occur in spinal cord of Apteronotus that is regenerating after a series of previous amputation/regeneration cycles. Multiple ependymas form both in sections of tissue that have split off from the major spinal cord, and within the major spinal circumference itself. In addition, two modes of production of the secondary ependymas are suggested by the results presented in this paper: (1) budding or direct splitting from the major ependymal structure, and (2) selforganization within a group of undifferentiated cells. If confirmed, this capacity of ependymally derived cells to selforganize into a typical ependymal structure would demonstrate that these cells possess not only a wide range of potentialities with respect to differentiation (e.g. into various spinal cord neurones and glia), but also have extensive potential for morphogenetic organization.

Duplication of the spinal cord can occur spontaneously during embryonic development, and has also been observed in the regenerating tail of the lizard (Simpson, 1964). Duplication of the embryonic spinal cord can also be caused by various chemical or mechanical treatments (Jacobson, 1978). For example, duplication of the notochord and spinal cord in the zebrafish embryo can be induced by exposure to ethanol (Laale, 1971). In all of these cases, the spinal axis splits, producing two whole cords. In this respect, these processes may be qualitatively different from the process in regenerating Apteronotus spinal cord, in which numerous subsidiary ependymas can form within a single spinal cord and a whole secondary cord may not result.

The proliferation of ependymal cells and the appearance of ependyma-like rosettes within the spinal cord have been observed in rat spinal cord after transection (Matthews, St. Onge & Faciane, 1979), and in several malignant conditions, including malignancies of the ependyma and subependymal layers (Rubinstein, 1972; Fu, Chen, Kay & Young, 1974; Azzarelli, Rekate & Roessmann, 1977). It should be emphasized that the important difference in the results from Apteronotus presented here is that the ependymally derived cells in Apteronotus are nonmalignant and that they form into complete ependymal tubes which exist in the correct orientation within the spinal cord.

The multiple ependymas in Apteronotus spinal cord have been observed only in cases of repeated amputation and regeneration cycles. The occurrence of multiple ependymas in Apteronotus only after multiple injuries implies that the previous amputations have had some effect on the uninjured part of the spinal cord (which is the source of new regeneration). This conclusion is consistent with earlier findings, which suggest that injury ‘primes’ the Apteronotus spinal cord for neurogenesis (Anderson & Waxman, 1985a). So far, the mechanisms underlying the effects of injury are not understood. It is known that in other neuronal systems (McQuarrie & Grafstein, 1981) the process of creating a lesion (injury) can ‘prime’ the system so that a subsequent lesion will elicit faster regrowth. Thus the process of an initial injury affects not only the regenerating part of the system, but also the adjacent uninjured nerve.

It may also be noteworthy that the multiple ependymas are observed primarily in two locations along the rostrocaudal axis: in rostral regenerated cord near the transition zone (between regenerated and unregenerated cord), and near the caudalmost tip (where new tissue is being generated). It seems reasonable that extra ependymas might be generated at the transition zone, which is the initial site of active proliferation of cells, and which exhibits a morphology characterized by the absence of normal landmarks. Likewise, the caudal tip of regenerating cord remains an active site of cell generation throughout the regeneration process. Hence it, too, is a prime candidate for the generation of secondary ependymas. It is interesting that the newly formed ependymal structures can be located at abnormal sites, i.e. close to the periphery of the cord. The expression of normal ependymal morphogenesis appears to involve processes, e.g. the development of cell polarity, differential cell adhesion, etc., that do not strictly depend on the characteristics of the surrounding milieu.

Our findings emphasize the wide degree of capabilities of the cells produced by the adult teleost ependyma in response to amputation of the cord. They strongly suggest that ependymally derived cells in regenerating Apteronotus spinal cord can selforganize into complete ependymal tubes in the absence of direct contact with the (template of the) major ependyma. These results, we believe, point to the possibility that mammalian ependyma may also contain cells with a wide range of developmental potential if the appropriate modes of stimulation, or conditions to encourage their division and differentiation, can be found.

This work was supported in part by the Medical Research Service, U.S. Veterans Administration, and by grant NS-15320 from the National Institutes of Health. We are also grateful to the Folger Foundation for support.

Altman
,
J.
&
Das
,
G. D.
(
1965
).
Postnatal origin of microneurones
.
Nature, Lond
.
207
,
953
956
.
Anderson
,
M. J.
&
Waxman
,
S. G.
(
1981
).
Morphology of regenerated spinal cord in Sternarchus albifrons
.
Cell Tiss. Res
.
219
,
1
8
.
Anderson
,
M. J.
&
Waxman
,
S. G.
(
1983a
).
Regeneration of spinal neurons in inframammalian vertebrates: Morphological and developmental .aspects
.
J. Hirnforsch
.
24
,
371
398
.
Anderson
,
M. J.
&
Waxman
,
S. G.
(
1983b
).
Caudal spinal cord of the teleost Sternarchus albifrons resembles regenerating cord
.
Anat. Rec
.
205
,
85
92
.
Anderson
,
M. J.
&
Waxman
,
S. G.
(
1985a
).
Neurogenesis in tissue cultures of adult teleost spinal cord
.
Devi. Brain Res
.
20
,
203
212
.
Anderson
,
M. J.
&
Waxman
,
S. G.
(
1985b
).
Neurogenesis in adult vertebrate spinal cord in situ and in vitro: A new model system. In Hope for a New Neurology
(ed.
F.
Nottebohm
).
Ann. N.Y. Acad. Sci
.
457
,
213
233
.
Anderson
,
M. J.
,
Waxman
,
S. G.
&
Laufer
,
M.
(
1983
).
Fine structure of regenerated ependyma and spinal cord in Sternarchus albifrons
.
Anat. Rec
.
205
,
73
83
.
Anderson
,
M. J.
,
Waxman
,
S. G.
&
Tadlock
,
C. H.
(
1984
).
Cell death of asynaptic neurons in regenerating spinal cord
.
Devi Biol
.
103
,
443
455
.
Azzarelli
,
B.
,
Rekate
,
H. L.
&
Roessmann
,
U.
(
1977
).
Subependymoma: A case report with ultrastructural study
.
Acta Neuropathol. (Berl.)
40
,
279
282
.
Bern
,
H. A.
&
Takasugi
,
N.
(
1962
).
The caudal neurosecretory system of fishes
.
Gen. comp. Endocr
.
2
,
96
110
.
Butler
,
E. G.
&
Ward
,
M. B.
(
1965
).
Reconstitution of the spinal cord following ablation in urodele larvae
.
J. exp. Zool
.
160
,
47
66
.
Easter
,
S. S.
, Jr
(
1983
).
Postnatal neurogenesis and changing connections
.
Trends Neurosci
.
6
,
53
56
.
Egar
,
M.
&
Singer
,
M.
(
1972
).
The role of ependyma in spinal cord regeneration in the Urodele, Triturus
.
Expl Neurol
.
37
,
422
430
.
Egar
,
M.
,
Simpson
,
S. B.
&
Singer
,
M.
(
1970
).
The growth and differentiation of the regen-erating spinal cord of the lizard Anolis carolinensis
.
J. Morph
.
131
,
131
152
.
Fu
,
Y.
,
Chen
,
A. T. L.
,
Kay
,
S.
&
Young
,
H. F.
(
1974
).
Is subependymoma (subependymal glomerate astrocytoma) an astrocytoma or ependymoma?
Cancer
34
,
1992
2008
.
Goldman
,
S. A.
&
Nottebohm
,
F.
(
1983
).
Neuronal production, migration and differentiation in a vocal control nucleus of the adult female canary brain
.
Proc. natn. Acad. Sci. U.S.A
.
80
,
2390
2394
.
Graziadei
,
P. P. C.
&
Graziadei
,
G. A.
(
1979
).
Neurogenesis and neuron regeneration in the olfactory system of mammals. I. Morphological aspects of differentiation and structural organization of the olfactory sensory neurons
.
J. Neurocytol
.
8
,
1
18
.
Hughes
,
A.
&
New
,
D.
(
1959
).
Tail regeneration in the Geckonid lizard Sphaerodactylus
.
J. Embryol. exp. Morph
.
7
,
281
302
.
Jacobson
,
M.
(
1978
).
Developmental Neurobiology
.
New York
:
Plenum Press
.
Kamrin
,
R. P.
&
Singer
,
M.
(
1955
).
The influence of the spinal cord in regeneration of the tail of the lizard, Anolis carolinensis
.
J. exp. Zool
.
128
,
611
627
.
Kaplan
,
M. S.
&
Hinds
,
J. W.
(
1977
).
Neurogenesis in the adult rat: electron microscopic analysis of light autoradiographs
.
Science
197
,
1092
1094
.
Kiortsis
,
V.
&
Droin
,
A.
(
1961
).
La régénération caudale des Urodeles (Induction et réactivité du territoire)
.
J. Embryol. exp. Morph
.
9
,
77
96
.
Laale
,
H. W.
(
1971
).
Ethanol induced notochord and spinal cord duplications in the embryo of the zebrafish, Brachydanio rerio
.
J. exp. Zool
.
117
,
51
64
.
Matthews
,
M. A.
,
St. Onge
,
M. F.
&
Faciane
,
C. L.
(
1979
).
An electron microscopic analysis of abnormal ependymal cell proliferation and envelopment of sprouting axons following spinal cord transection in the rat
.
Acta Neuropathol. (Berl.)
45
,
27
36
.
McQuarrie
,
I. G.
&
Grafstein
,
B.
(
1981
).
Effect of a conditioning lesion on optic nerve regeneration in goldfish
.
Brain Res
.
216
,
253
264
.
Nordlander
,
R. H.
&
Singer
,
M.
(
1978
).
The role of ependyma in regeneration of the spinal cord in the urodele amphibian tail
.
J. comp. Neurol
.
180
,
349
374
.
Paton
,
J. A.
&
Nottebohm
,
F. N.
(
1984
).
Neurons generated in the adult brain are recruited into functional circuits
.
Science
225
,
1046
1048
.
Rubinstein
,
L. J.
(
1972
).
Tumors of the central nervous system
.
In Atlas of Tumor Pathology
, 2nd ser., fase.
6
.
Washington, DC
:
Armed Forces Institute of Pathology
.
Simpson
,
S. B.
, Jr
(
1964
).
Analysis of tail regeneration in the lizard Lygosoma latérale. I. Initiation of regeneration and cartilage differentiation: the role of ependyma
.
J. Morph
.
114
,
425
436
.
Simpson
,
S. B.
, Jr
(
1968
).
Morphology of the regenerated spinal cord in the lizard, Anolis carolinensis
.
J. comp. Neurol
.
134
,
193
210
.
Simpson
,
S. B.
, Jr
(
1983
).
Fasciculation and guidance of regenerating central axons by the ependyma
.
In Spinal Cord Reconstruction
(ed.
C. C.
Kao
,
R. P.
Bunge
&
P. J.
Reier
), pp.
151
162
.
New York
:
Raven Press
.
Wilson
,
K. C. P.
&
Raisman
,
G.
(
1980
).
Age-related changes in the neurosensory epithelium of the mouse vomeronasal organ: extended period of postnatal growth in size and evidence for rapid cell turnover in the adult
.
Brain Res
.
185
,
103
113
.
*

Formerly classified: Sternarchus albifrons.