De- and remyelination in spinal roots during normal perinatal development in the cat: a brief summary of structural observations and a conceptual hypothesis

Abstract

We have studied the perinatal development of large myelinated axons (adult D>10µm) in cat ventral and dorsal lumbo-sacral spinal roots using autoradiography and electron microscopy (serial section analysis). These axons acquire their first myelin sheaths 2–3 weeks before birth and show nearly mature functional properties first at a diameter of 4–5µm, i.e. 3–4 weeks after birth. The most conspicuous event during this development takes place around birth, when a transient primary myelin sheath degeneration strikes already well myelinated although short ‘aberrant’ Schwann cells. The aberrant Schwann cells become completely demyelinated, then measuring about 10 µm in length, and are subsequently eliminated from their parent axons. Morphometry indicates that on average 50% of the Schwann cells originally present along a prospective large spinal root axon suffer elimination. Here it should be noted that in cat lumbo-sacral spinal roots, the longitudinal growth of myelinated Schwann cells that belong to the group containing what will be the largest fibers is on average twice that of their parent axons. The elimination phenomenon is particularly striking in the dorsal roots close to the spinal cord where CNS tissue invades the root for several hundred micrometres. Our observations suggest that, once demyelinated and then eliminated, Schwann cells (i.e. aberrant Schwann cells) colonize neighbouring axons, future myelinated as well as future unmyelinated ones. In the former case the immigrant Schwann cells appear to start myelin production, possibly risking a second demyelination and elimination. We take our observations to indicate that Schwann cells in the cat, during normal development, may switch iteratively between a ‘myelin-producing’ and a ‘non-myelin-producing’ phenotype. From a functional point of view the transient presence along a myelinated axon of intercalated unmyelinated segments ∼10µm long, due to aberrant Schwann cells, would mean a slowing down of the action potential. The rapid disappearance of aberrant Schwann cells during the two first postnatal weeks could then explain the progressing normalization of the leg-length conduction time.

Introduction

The functional maturation of feline peripheral hind limb nerves and associated lumbo-sacral spinal roots has been studied extensively by Skoglund and co-workers (for references and reviews see Skoglund, 1966, 1969; Ekholm, 1967; Nilsson & Berthold, 1988). They found, in the newborn kitten, that the peak of the largest myelinated fibres measured 2–3 µm in diameter and were equipped with myelin sheaths of 20–30 compact lamellae. (In the adult cat the corresponding values are 15 µm and 120–150 lamellae.) At birth these fibres were unable to support sustained or high-frequency activity, and their absolutely refractory periods were about twice the adult value. To this is added a leg-length conduction time that was double the adult value.

The fibre group began to show nearly mature properties as it reached a diameter of 4–5 µm and a myelin sheath thickness of 40–50 compact lamellae (which in the case of lumbo-sacral – L 6,7, S1 – spinal roots corresponds to the third postnatal week). Thus a remarkable functional maturation had taken place, during which there was little change in such fundamental fibre variables as fibre diameter and myelin sheath thickness. During the same period, however, the organization and the distribution of internodes and of nodes of Ranvier showed striking changes, and the ‘future large fibres’ became miniatures of their 3–4 times thicker and fully mature counterparts (for review and references see Berthold, 1996).

The most conspicuous feature during this maturation process is a primary myelin sheath degeneration that strikes already well-myelinated Schwann cells which, after complete demyelination, become eliminated from their parent axons. Morphometric studies, based on serial sections and ultrastructural analyses, have indicated that on average 50% of the Schwann cells originally present along the spinal root part of prospective alpha motor axons become eliminated. Thus the longitudinal growth of myelinated Schwann cells in the cat is on average twice that of their parent alpha axons. A particularly extensive elimination of demyelinated Schwann cells has been noted close to the spinal cord in the dorsal spinal roots L7 and S1 (Carlstedt, 1981).

In the light of these observations we performed a series of autoradiographic and morphometric (EM serial section analysis) studies in order to investigate the proliferation and the distribution of Schwann cells in the L7 ventral root and the S1 dorsal root during perinatal development in the cat (Berthold & Nilsson, 1987; Nilsson & Berthold, 1988; Nilsson, 1988; Nilsson-Remahl et al. 1998). Based on our results we have formulated the hypothesis that in cat lumbo-sacral spinal roots, during normal perinatal development, Schwann cells, which have first been myelinated, then demyelinated and subsequently eliminated, should have several alternative options: (1) to colonize neighbouring axons and in the case of prospective myelinated ones to take part in their myelination, (2) to remain dormant in the endoneurial space and (3) to remove themselves by apoptosis. The first alternative includes the possibility of a second demyelination and elimination after which the three options again become valid.

Materials and methods

Fourteen cat fetuses (aged 25–63 days after mating, d.a.m.; full term = 63 d.a.m.), nine kittens (aged 1–6 weeks) and one adult cat (aged 1 years) were used. The fetuses were removed by Caesarean section. All animals were anaesthetized intraperitoneally and perfusion-fixed with phosphate-buffered 5% glutaraldehyde. The ventral (VR) and dorsal (DR) roots L6, L7 and S1 were dissected out and their length from the spinal cord to the dorsal root ganglion measured. After postfixation and osmication the specimens were embedded in Vestopal-W (for details regarding anaesthesia, fixation, preparative procedures and autoradiography see Berthold & Nilsson, 1987; Nilsson, 1988; Nilsson-Remahl et al. 1998).

Series of between 600 and 8000 consecutive ultrathin cross-sections, each section about 100 nm thick, were cut in the proximo-distal direction, beginning at the middle of the ventral roots and in dorsal roots at the root–spinal cord junction. In the adult cat and the older kittens sectioning was performed in batches of 2–5 ultrathin cross-sections followed by 100 semithin consecutive ones (about 0.5 µm thick). Between 15 and 20 such batches were produced.

In each of the serially sectioned roots (the L7 ventral and the S1 dorsal roots), a root fascicle of 100–400 and 200–600 axons, respectively, was selected for analysis. Its contents of axons and Schwann cells were reconstructed to scale for distances of between 120 and 500 µm in the fetuses, and between 500 and 1000 µm in the kittens and the adult cat. Calculations based on data obtained from measurements on the reconstructed fibre samples in combination with earlier obtained estimations of calibre spectra and internodal lengths in VR L7 (Nilsson & Berthold, 1988) and DR S1 (Carlstedt, 1980) enabled us to compare the Schwann cell content of a defined group of axons from cats of different ages.

Full definitions of terms, calculations of ‘standardized’ and ‘growth-compensated’ variables, together with original quantitative data, are given in Berthold & Nilsson (1987) and Nilsson-Remahl et al. (1998).

Calculation of the standardized Schwann cell internuclear distance (SSCD)

This variable describes the theoretical average axonal length available to each of the Schwann cells associated with a diameter class. In the case of Schwann cells associated with a single axon, the variable corresponds roughly to mean Schwann cell length. With regard to completely myelinated fibres, SSCD translates to internodal length. In the case of Schwann cells associated with more than one axon, the variable becomes purely theoretical:

$$\text{SSCD}={\text{T}}_{\text{d}}/\sum {\text{SCN}}_{\text{d}}$$

where T_d = total axon length of a diameter class = Ax_d^*L; where Ax_d = number of axons of a diameter class and L = the length of the reconstructed root fascicle; and Σ SCN_d = the sum of all Schwann cell nuclear values (SCN) of a particular diameter class.

All completely reconstructed Schwann cell nuclei, and all partly reconstructed ones found at the distal end of the series, were given a value depending on the number of axons associated with the Schwann cell at the nuclear region, SCN = 1/number of associated axons.

Observations and comments

Developmental features common to the L6, L7 and S1 ventral and dorsal spinal roots

The large myelinated axons in the L6, L7 and S1 spinal roots of the newborn kitten (age = 63 d.a.m) display three types of internodes. About 85% of the internodes are ‘long’ ones, c. 300 µm in length. They are generally equipped with a ‘complex’ paranodal region at one end and a ‘simple’ one at the other (Figs 1 and 2). Some 5% are ‘short’ internodes, 50–150 µm long, and usually equipped with ‘complex’ paranodes at both ends. About 10% are ‘very short’ internodes, ∼10–50 µm in length. The ‘very short’ internodes lie intercalated between ‘long’ and ‘short’ internodes in an apparently random fashion. They contain Marchi-positive (degenerating myelin and fat droplets) debris and acid phosphatase-positive material. The myelin sheaths of the ‘very short’ internodes appear, when not obscured by the debris, highly irregular and ‘accordion-like’. The occurrence of ‘short’ and ‘very short’ internodes declines rapidly during the first two postnatal weeks and becomes virtually nil after about 1 month, the ‘short’ ones ahead of the ‘very short’ ones (Berthold, 1973). Most likely the appearance of ‘complex’ paranodes, and of ‘short’ and ‘very short’ internodes is due to Schwann cell crowding along overpopulated axons (see below) in combination with some sort of selective contact inhibition.

Fig. 1.

Electron micrograph. Longitudinal section through ventral root L7 of a newborn kitten. The main part of the nerve fibres are alpha axons (A) equipped with 30–40 myelin lamellae. Myelinated gamma axons (G) are scarce. Six nodes of Ranvier are marked out (N1–N6). Node N1 is bordered by ‘simple’ paranodes (asterisks), i.e. the internodal end-segments have fairly smooth myelin sheaths surrounded by Schwann cell cytoplasm free of disintegrating lamellar bodies. The nodes N2 and N3 are both bordered by one ‘simple’ and one ‘complex’ paranode (C; cf. Fig. 2). Node N4 is delimited by ‘simple’ paranodes but also associated with a juxtaposed aberrant Schwann cell (S1; cf. Fig. 5). The nodes N5 and N6 delimit a ‘very short’ myelinated internode and its aberrant Schwann cell (S2; cf. Fig. 3). Scale bar = 10 μm. (Reproduced from Berthold & Skoglund, 1968; by permission of Almquist and Wiksell.)

Fig. 2.

Electron micrograph. Ventral root L7 of a newborn kitten. Longitudinal section through a node of Ranvier (arrow) and its associated complex paranode (heavy vertical bar). The main paranodal axon is outside the sectioning plane. The latter passes through the outer Schwann cell compartment, which is packed with lamellar bodies, many in various stages of disintegration (white asterisks). Serial section analysis showed that those marked by black asterisks connected to the main myelin sheath. SN, Schwann cell nucleus of adjacent nerve fibre; A, axon. Scale bar = 1 μm. (Reproduced from Berthold & Skoglund, 1968; with the permission of Almquist and Wiksell.)

Electron microscopy has revealed two types of ‘very short’ internodes: myelinated and unmyelinated ones (Figs 3 and 4). The former type (Figs 1, 3, 10 and 11) is characterized by a highly irregular myelin sheath that includes a correspondingly irregularly shaped axon segment. The Schwann cell cytoplasm contains fat droplets and numerous acid phosphatase positive and disintegrating lamellar bodies. Unmyelinated ‘very short’ internodes lack a myelin sheath but contain varying amounts of disintegrating lamellar debris and fat droplets (Fig. 4). The axon segments of unmyelinated ‘very short’ internodes show few irregularities and are 2–15 µm long. In some cases the debris-containing Schwann cell of an apparent unmyelinated ‘very short’ internode is in fact juxtaposed to the fibre and connected to the nodal axon in a restricted fashion (Fig. 5). We refer to the Schwann cells of the different types of ‘very short’ internodes as ‘aberrant Schwann cells’, and distinguish between myelinated, unmyelinated and juxtaposed ones.

Fig. 3.

Electron micrograph. Ventral root L7 of a newborn kitten. Longitudinal section through a ‘very short’ myelinated internode, i.e. through a myelinated aberrant Schwann cell. Bordering nodes of Ranvier are at arrows. The tortuous myelin sheath forms complex outgrowths. There are many isolated lamellar bodies in the Schwann cell cytoplasm. Sn, Schwann cell nucleus; Ax, axon. Compare with Figs 10 and 11. (Reproduced from Berthold, 1996; by permission of Wiley-Liss.)

Fig. 4.

Electron micrograph. Ventral root L7 of a 1-week-old kitten. Longitudinal section through a ‘very short’ unmyelinated internode, i.e. through an unmyelinated aberrant Schwann cell. Bordering hemi-nodes of Ranvier are at the arrows. The Schwann cell cytoplasm contains a number of lamellar bodies (asterisks) in different stages of disintegration. The arrowhead points at a cluster of more or less circular clear spaces which are taken to indicate the position of fat droplets dissolved during dehydration of the specimen. Sn, Schwann cell nucleus; Ax, axon. (Reproduced from Berthold & Skoglund, 1968; by permission of Almquist and Wiksell.)

Fig. 10.

Electron micrograph. Ventral root L7 of a cat fetus; age 55 d.a.m. Transverse section through a myelinated aberrant Schwann cell. This is electron microscopic image no. 47 of the series used for the 3D-reconstruction in Fig. 11. The position of the image in the reconstruction is indicated by the thin white line in Fig. 11. Thin arrows point at positions of presumed fat droplets. Arrowheads indicate myelin and axonal outgrowths. Asterisk indicates a lamellar body. SN, Schwann cell nucleus. Scale bar = 1 μm.

Fig. 11.

Computer assisted three-dimensional reconstruction of the myelinated aberrant Schwann cell of Fig. 10. The reconstruction is based on a series of 250 about 90-nm-thick transverse sections documented electron microscopically. In all the five panels (a−e) the circumscribing light line indicates the outer contour of the aberrant Schwann cell as viewed from its nuclear side with a transparent cell membrane (see Fig. 10). A, normally appearing internodal axon segments proximal and distal to the aberrant Schwann cell. Ax, distorted axon segment surrounded by the aberrant Schwann cell; My, myelin sheath of the aberrant Schwann cell; SN, nucleus of the aberrant Schwann cell. White arrows point at the two nodal axon segments that demarcate the aberrant Schwann cell and its ‘very short’ internode. Light grey rounded bodies of panels a, b, and e show solitary disintegrating lamellar bodies, two of which are marked out by black asterisks. Tiny white spheres show lipid droplets. Vertical scale bar in e = 10 μm. (a) All reconstructed components are illustrated. White asterisk indicates myelin sheath outbulging. (b) The Schwann cell nucleus has been ‘removed’ by the computer and the irregular myelin sheath becomes partly visible. (c) Disintegrating lamellar bodies and lipid droplets have been ‘removed’ and the irregular myelin sheath becomes fully visible. (d) All components except the axon have been ‘removed’. Note the irregular shape of the axon as found inside the aberrant Schwann cell. (e) The Schwann cell nucleus and the myelin sheath have been ‘removed’ leaving a naked axon surrounded by disintegrating lamellar bodies and lipid droplets which tend to aggregate close to the nodes. The reconstruction was performed in co-operation with Drs Ragnar Pascher and Marie Kreus. The facilities of the MEDNET laboratory at the Medical Faculty in Goteborg were used.

Fig. 5.

Electron micrograph. Ventral root L7 of a 1-week-old kitten. Longitudinal section through a juxtaposed aberrant Schwann cell. The aberrant unit lies alongside the paranode to the right of the node of Ranvier (arrow). It is connected with the nodal axon segment by an irregular cytoplasmic tongue. The cytoplasm of the aberrant Schwann cell contains disintegrated lamellar bodies (*). Sn, nucleus of the juxtaposed aberrant Schwann cell; Ax, axon. (Reproduced from Berthold & Skoglund, 1968; by permission of Almquist and Wiksell.)

The presence, along developing peripheral nerves, of very short internodes corresponding to our ‘aberrant Schwann cells’ was reported by Vignal (1883a, 1883b) more than a century ago. In the light of the then current concepts regarding myelination, Vignal interpreted the very short internodes as myelinating cell units added to those already present along an immature myelinated fibre. Modern concepts, the ultrastructure of the lamellar debris, its Marchi positivity and its acid phosphatase activity, suggest just the opposite situation: myelin-producing Schwann cells, having got rid of their myelin sheaths, are leaving immature myelinated axons (Berthold, 1973). Obviously the final setting of the number of nodes of Ranvier along the large axons takes place several weeks after the axons acquired their first myelin.

Ventral spinal root L7

The ventral spinal root L7 of the adult cat is 35 mm long. It contains about 70% myelinated and 30% unmyelinated (C) axons (Coggeshall et al. 1974; see also Nilsson & Berthold, 1987). About two-thirds of the myelinated axons are large ones with a mean diameter of 9 µm and a mean internodal length of about 1400 µm. They represent the group of alpha axons (including a few beta axons). The remaining myelinated axons are small ones with a mean diameter of 3 µm and a mean internodal length of about 650 µm. They represent the group of gamma axons.

The alpha axons acquire their first myelin during the period from 45 d.a.m. to 55 d.a.m. when the root increases in length from about 6 mm to 8 mm. The gamma axons start their myelination about a week later at full term, when the root measures 10 mm in length. All gamma axons are myelinated about 3 weeks later at a root length of 16 mm (Nilsson & Berthold, 1987). The unmyelinated (C-) axons begin to enter the distal end of the root around 45 d.a.m. (Risling et al. 1981; Nilsson & Berthold, 1988).

At the very beginning of their myelination, axons show cuffs of compact myelin that alternate with intercalated unmyelinated segments (Figs 7 and 9). As a rule, at this early stage, one and the same Schwann cell displays both myelinated and unmyelinated stretches. In a few cases, and then usually along myelinating gamma axons, the whole extent of a Schwann cell is unmyelinated. Such intercalated non-myelinating Schwann cells often contained smaller unmyelinated axons in addition to the larger, otherwise myelinated gamma axon (Fig. 9).

Fig. 7.

Serial section-based reconstructions of axons from immature cat L7 ventral roots. The drawings illustrate the distribution of myelinated and unmyelinated segments along axons of different sizes. The mean axon diameter for each fibre is given to the left. A nuclear value less than 1, for instance ¼, indicates that the depicted axon is one out of four axons ensheathed by the same Schwann cell. (Reproduced from Berthold & Nilsson, 1987; by permission of the editor and publisher of the Journal of Neurocytology.)

Fig. 9.

Serial section-based reconstructions of a group of axons from the L7 ventral root of a 1-week-old kitten. The reconstruction illustrates the mixture, during early myelination, of myelin- and non-myelin producing Schwann cells along the same axon. Mean axon diameter to the right. Same symbols as in Fig. 7. (Reproduced from Berthold & Nilsson, 1987; by permission of the editor and publisher of the Journal of Neurocytology.)

Our calculations of ‘the mean standardized Schwann cell internuclear distance’, a variable that includes the number of aberrant Schwann cells and translates to ‘mean standardized Schwann cell length’, i.e. to internodal length in fully myelinated axons (see Nilsson & Berthold, 1987), show that the mean standardized Schwann cell length along the alpha axons during their initial myelination between the 45th and 55th d.a.m. averages about 140 µm. During the same period the mean standardized Schwann cell length along the still unmyelinated gamma axons can be calculated to about 280 µm. Evidently the Schwann cells of the alpha axons, as noted from the beginning of myelination to adulthood, increases in length from about 140 µm to about 1400 µm, i.e. with a factor of 10. During the same period the root increases in length from about 7 mm to about 35 mm, i.e. with a factor of about 5. So there seem to be 50% too many Schwann cells present along the alpha axons at the beginning of their myelination. Using the data given for the gamma axons it can be calculated that these axons, during the period between the 45th and 60th d.a.m., lack about 50% of the Schwann cells necessary to satisfy the adult values. At full term, when the gamma axons have just begun their myelination, the Schwann cell surplus along the alpha axons drops from about 50% to about 10%, and the deficit along the gamma axons changes to a surplus of 5%. During the first postnatal month the calculated Schwann cell surplus along the alpha axons varies between 10% and 25% and that along the gamma axons increases from 5% to 25% during the first postnatal week and remains high throughout the first month (Berthold & Nilsson, 1987).

The drastic change at full term from a Schwann cell deficit to a surplus along the gamma axons concomitant with a marked drop in the Schwann cell surplus along the alpha axons could perhaps be explained hypothetically by a wave of programmed Schwann cell death along the alpha axons in combination with a burst of Schwann cell proliferation along the gamma axons. None of the cell units encountered during our reconstructions of the selected root fascicles has shown the structural characteristics of apoptosis (Wyllie, 1980; Kerr et al. 1995); yet a recent re-examination of our material has demonstrated the very rare occurrence of pyknotic cell units in the endoneurial space outside the analysed fascicles (Fig. 6). The unlikely possibility thus exists that at least a part of the drop in the Schwann cell surplus along the alpha axons may depend on apoptosis. Also unlikely is a burst of Schwann cell proliferation along the gamma axons. As judged from the situation in the L6 ventral root, Schwann cell proliferation has its maximum between the 30th and the 45th d.a.m., whereafter it declines rapidly, becomes close to zero at 55 d.a.m and ceases completely after birth (Nilsson, 1988). An alternative explanation is that the Schwann cell surplus along the alpha axons is exported to the gamma axons and covers up their Schwann cell deficit as they begin to myelinate. The obvious main actor in this transfer process should be the aberrant Schwann cells.

Fig. 6.

Electron micrograph. Ventral root L7 of a newborn kitten. Longitudinal section through a nodal region (arrow) and its bordering paranodes. Associated to the region is a pyknotic cell unit (asterisk). This is a most unusual observation. The unit is tentatively interpreted to represent an eliminated apoptotic aberrant Schwann cell. Fi = fibroblast. Scale bar common for Figs 3–6 = 4 μm. (Reproduced from Berthold, 1996; by permission of Wiley-Liss.)

Aberrant Schwann cells are first seen on myelinating alpha axons at 47 d.a.m. They make up 10–15% of the Schwann cells associated with alpha axons during the period 50–60 d.a.m. (Berthold, 1973; Berthold & Nilsson, 1987). Their occurrence then drops to 7% at full term and becomes nil during the third postnatal week. When searched for systematically at the age of 55 d.a.m., five out of 10 juxtaposed aberrant Schwann cells showed contacts with unmyelinated axons (Fig. 8), an arrangement in line with the suggested transfer. Concomitant with the change along the gamma axons from a Schwann cell deficit to a surplus the first aberrant Schwann cells appeared in this axon group. This suggests that supernumerary Schwann cells have to be eliminated also from the gamma axons in order to satisfy the inconsistency in longitudinal growth between the Schwann cells and the axons. The fate of the Schwann cells eliminated from both the gamma axons and the alpha axons during the first postnatal month is obscure. Several possibilities exist. Some of the eliminated cells may associate with the ingrowing C-axons which during this period show a moderate Schwann cell deficit (Berthold & Nilsson, 1987). Others dedifferentiate and become free-lying inconspicuous members – Schwannoid cells – of the endoneurial space, demarcated by a more or less fragmented basal lamina (Fig. 12). Such cells are found scattered in the spinal roots and in comparatively high numbers close to the spinal cord (Nilsson-Remahl, unpublished observations). The rest may suffer apoptosis. An elimination of supernumerary Schwann cells similar to the one observed here during normal development is well known to occur during regeneration after nerve injuries (for references and review see Hildebrand et al. 1994).

Fig. 8.

Electron micrographs of cross-sections from different levels (a−d) of a myelinated alpha-fibre (1) equipped with a juxtaposed aberrant Schwann cell (55-day-old cat fetus). Axon 2 is a gamma axon. Axon 3 is a C-axon. The inset electron micrograph shows the boxed area in d. Arrowhead points at a cytoplasmic contact between the juxtaposed aberrant Schwann cell and the two unmyelinated axons (2 and 3). ASCN, aberrant Schwann cell nucleus. Scale bar (in b) = 1 μm; inset = × 50 000. Reconstructions of the fibres are given below the electron micrographs. Arrows indicate levels shown in the electron micrographs. Dotted line indicates extent of the aberrant Schwann cell cytoplasm at the nuclear level. Mean axon diameter (μm) is given to the right of each reconstruction. (Reproduced from Berthold & Nilsson, 1987; by permission of the editor and publisher of the Journal of Neurocytology.)

Fig. 12.

Electron micrograph. Adult cat. Dorsal root S1. X, Schwannoid cell situated in the endoneurial space just outside the astrocytic part of the PNS—CNS transitional region (TZ). Serial section analysis showed that this cell consisted of just a thin cytoplasmic rim and a nucleus. Its cell membrane possessed several hemidesmosomes outside which there were traces of basement membrane-like material. Scale bar = 1 μm.

Therefore there is a possibility that a pro-myelin Schwann cell, originally present on a prospective alpha axon, first turns into a myelinating unit, then stops myelin production, adopts the traits of an aberrant Schwann cell, changes into a degenerative mode, destroys its myelin sheath, is eliminated, associates with a prospective gamma axon, iterates its previous behaviour, and becomes finally a non-myelinating Schwann cell holstering a bundle of C-axons.

Dorsal spinal root S1

Our study of the Schwann cell distribution in the most proximal part of the dorsal root S1 (Nilsson-Remahl et al. 1998), i.e. at the transition between the PNS and the CNS, was undertaken in order mainly to get support for the idea that Schwann cells eliminated from already myelinated axons during development may colonize neighbouring prospective C-axons. As noted earlier (Berthold & Carlstedt, 1977; Carlstedt, 1981), there is a particularly high occurrence of ‘complex’ paranodal regions, of ‘short’ internodes and of aberrant Schwann cells during the first postnatal weeks in the dorsal root adjacent to the spinal cord. To this are added numerous endoneurial Schwannoid cells.

The period during which axons are recruited from the group of unmyelinated axons into that of myelinated ones extended for about 1 month from 53 d.a.m. into the third postnatal week. The ‘mean standardized Schwann cell length’ along the earliest myelinated axons, i.e. the prospective largest myelinated axons can be calculated as about 100 µm, the root then measuring about 8 mm in length. A comparison with the values of the corresponding variables in the adult gives a situation similar to that in the ventral root: the ‘mean standardized Schwann cell length’ has increased to about 850 µm and the root to about 36 mm, i.e. the Schwann cell length increases 8.5 times and the root 4.5 times. The prospective large myelinated axons are evidently overpopulated with Schwann cells to an extent comparable with that in the ventral root. The consequent problem of crowding is further exaggerated here adjacent to the spinal cord by the successive outgrowing of CNS-tissue from the cord into the root (Carlstedt, 1981; cf. Fraher & Kaar, 1986; Fraher et al. 1988). The growth rate of the CNS extension seems to be particularly high from the 60th d.a.m. to the second postnatal week, a period during which the occurrence of aberrant Schwann cells and endoneurial Schwannoid both show a marked increase close to the CNS–PNS borderline (Table 1 in Nilsson-Remahl et al. 1998). Evidently the prospective large myelinated dorsal root axons have to get rid of their Schwann cell surplus just like the alpha axons. The virtual lack of apoptotic cell units (two of more than 3000 examined perikarya) speaks against programmed cell death as the fate of the supernumerary Schwann cells. Our calculations suggest instead a transfer to axons with a Schwann cell deficit, which in particular seems to be the case with prospective C-axons close to the CNS–PNS borderline.

Calculations of SSCD along unmyelinated axons adjoining the CNS–PNS borderline and along those more distally in the reconstructed segments at the age of 3–4 weeks, when all prospective myelinated axons should have acquired their first myelin, give values around 480 µm and 715 µm, respectively. If the number of Schwann cells then associated with the unmyelinated axons were to remain constant, SSCD would increase by about × 2.4 (root length growth factor) to 1150 µm and 1715 µm in the adult, respectively. Calculations based on actually measured values in the adult give SSCDs of 522 µm and 1750 µm, respectively. Obviously the unmyelinated axons close to the cord more than doubled their number of associated Schwann cells (Table 2 in Nilsson-Remahl et al. 1998). One explanation for the increase in the number of Schwann cells along unmyelinated axons during development is, of course, the proliferation of already present ones. This explanation does not fit with our autoradiographic data that show that the mitotic activity ceases during the first postnatal week (Nilsson-Remahl et al. 1998).

Concluding remarks

Schwann cells originate from the neural crest in a sequence of differentiating steps where neural crest cells generate ‘Schwann cell precursors’, which in turn give rise to ‘immature Schwann cells’. The latter develop into either of the two ‘mature Schwann cell types’: the ‘myelinating’ and the ‘non-myelinating Schwann cell’. The two mature Schwann cell types are known to regress to the immature phenotype if their parent axons degenerate and, under such circumstances, to survive due to sustaining autocrine loops (for review and references see Jessen & Mirsky, 1999; also Mirsky et al. 2001). Schwann cell precursors and to some extent also immature Schwann cells depend for their survival on the growth factor beta-neuregulin as supplied by axons before the myelinating stage has been reached. Without axonal support (i.e. without access to neuregulin), Schwann cell precursors and premyelin immature Schwann cells are prone to die by apoptosis (Grinspan et al. 1996; Syroid et al. 1996).

When these facts about the development of the Schwann cell lineage are applied on our observations in cat lumbo-sacral spinal roots, it seems clear that our data refer mainly to Schwann cells that have passed the transition point from the immature to the mature cell stage and hence are able to survive due to autocrine loops (Meier et al. 1999). This should in particular apply to the aberrant Schwann cells.

Consequently we refute the idea that the Schwann cell surplus along the just myelinated prospective large axons is corrected by apoptosis and suggest that Schwann cells normally in cat lumbosacral spinal roots are eliminated from already myelinated axons and transferred to unmyelinated axons during the period of early myelination (cf. Berthold & Nilsson, 1987). Such a high degree of plasticity is in line with observations that Schwann cells are surprisingly tolerant under circumstances in which apoptosis could be expected, and able to swap between a myelinating and a non-myelinating phenotype, as seen for instance during and after wallerian degeneration (for discussion see Stewart et al. 1996).

We have used the concepts ‘crowding’ or ‘overpopulation’ as the cause behind the elimination phenomenon: though this is a crude simplification and says nothing about the molecular mechanisms behind the formation of ‘complex’ paranodes and aberrant Schwann cells. Here two recent works are of interest. First, Casella et al. (2000) have demonstrated the presence of a contact-mediated growth regulating factor, contactinhibin, in the cell membrane of human Schwann cells. Second, Zanazzi et al. (2001) showed, in established myelinated cultures, that neuregulin treatment induced deformation of paranodal myelin and demyelination of complete internodes. In line with this it is possible that the development of aberrant Schwann cells may be due to some imbalance between axonal neuregulin and its Schwann cell receptors ErbB2 and ErbB3.

Our observations indicate that the current idea, that already myelinated Schwann cells only are able to demyelinate and regress to an immature state in connection with axonal degeneration and then according to circumstances turn into the myelinating or the non-myelinating phenotype, needs revision and should include the early myelinating stage. It should also be noted that aberrant Schwann cells are rare in the rat and the mouse (Berthold, 1974), the nowadays most commonly used experimental animals in studies of the developing nervous system.