Journal of Neurocytology 26, 799–809 (1997)

Postnatal development of Schwann cells at neuromuscular junctions, with special reference to synapse elimination K . H I R A T A * , C . Z H O U , K . N A K A M U R A and M . K A W A B U C H I Department of Anatomy, Faculty of Medicine, Kyushu University, Maidashi 3-1-1, Higashi-ku, Fukuoka 812, Japan Received 26 September 1996; revised 19 May and 8 August 1997; accepted 15 August 1997

Summary The neuromuscular junctions (NMJs) of postnatal rat soleus muscles were examined by immunohistochemical staining for S100, a marker of Schwann cells (SCs), and for protein gene product 9.5, a neuronal marker, to elucidate the involvement of SCs in synapse elimination. The morphological maturation of S100-immunoreactive terminal SCs at NMJs proceeded with the gradual increase in their number. The number of terminal SCs per NMJ was one or two at postnatal day (P) 7, reaching the adult number at P28, when it became three or four. Confocal laser scanning microscopic analysis of multi-innervated NMJs, whose number decreased between P7 and P14, revealed a change in the ratio between terminal SCs and axons with age. At P7, the ratio between axons and terminal SCs per NMJ was 52 : 1, which was exactly the reverse of that in adults, while at P14 this had changed to 2 : 2. A structural change appeared to occur at the same time at the preterminal region, this being prior to the establishment of a 1 : 1 relationship between axon and SC sheath which was detected at P14, with the 52 : 1 relationship seeming to occur at P7. Thus, synapse elimination seems to proceed, at least for one week, with the gradual loss of axons which are at different stages of maturation with respect to their spatial relationship with SCs. From our results it seems unlikely that SCs play an active role in selecting a single axon to survive.

Introduction It is known that the pruning of axonal branches and contemporaneous synapse elimination occur postnatally in the CNS and PNS and may be a critical event in neural development. The motor axons which initially co-innervate the same muscle fibres lose some of their connections during early postnatal life, leaving muscle fibres singly innervated (for review see Purves & Lichtman, 1980; Colman & Lichtman, 1993). Synapse elimination is considered to be competitive simply because the fate of one set of terminal motor axons known as ‘synaptic cartel’ is related to that of another on the same muscle (Balice-Gordon et al., 1993). This competition is thought to be due to direct interactions between axons which converge at the same neuromuscular junction (NMJ) or else due to indirect interactions involving other cells in the vicinity, i.e. muscle cells or Schwann cells (SCs). In the latter case, relatively little is known about the role of SCs in this event, whereas considerable evidence exists showing that muscle cells play a crucial role in synapse elimination (Colman et al., 1987; Colman & Lichtman, 1993). * To whom correspondence should be addressed.

0300–4864/97 ( 1997 Chapman and Hall

As for the involvement of SCs in synapse elimination, two theories have been proposed. One is a hypothesis introduced by Brown and colleagues (1976): SCs ‘select’ a single axon to survive by establishing a mature relationship with it. This theory is interesting in that SCs absolutely control the first step of synapse elimination, however, subsequent electron microscopic studies have provided little evidence to support this theory (Bixby, 1981). The second is that SCs ‘guide’ the axons to withdraw. Recent studies on the NMJs of frogs (Chen & Ko, 1994) and rats (Son & Thompson, 1995a,b) revealed, through the use of double fluorochrome labelling, that the extension of the processes of SCs preceded the axonal outgrowth in regenerating and remodelling motor nerves. Their results suggested that SCs play a major role in promoting nerve growth. The authors concluded that SCs might also play a similar role in the regression of nerve processes. However, these studies were performed on adult muscles. In the present study, neonate and young rat soleus muscles were used to elucidate the involvement of SCs in synapse elimination during development. To this

800

H I R A T A , Z H O U , N A K A M U R A and K A WA B U C H I

Schwann cells in synapse elimination end, immunostaining for the calcium-binding protein S100, which is known to be a marker of SCs (Brockes et al., 1977; Stefansson et al., 1982; Reynolds & Woolf, 1992), was first performed to demonstrate the morphological details of SCs in developing NMJs, since no study had previously done so. S100 immunohistochemistry combined with cholinesterase staining or with DAPI nuclear staining was used to identify the site of the endplate or to clarify the change in the number of terminal SCs, respectively. On the other hand, immunostaining for the protein gene product 9.5 (PGP 9.5), an enzyme known as ubiquitin carboxylterminal hydrolase, which is understood to be a highly sensitive marker for neuronal elements (Thompson et al., 1983; Wilkinson et al., 1989) including motor axon terminals (Ann et al., 1994), was performed to demonstrate developmental changes in motor axons. Finally, confocal laser scanning microscopy was applied on the S100- and PGP 9.5-double immunofluorescencelabelled sections to analyse the relationship between SCs and motor axons during the process of synapse elimination. Materials and methods Wistar rats were housed in a 12-h light–dark cycled environment, with free access to water and standard food. The day of birth was designated as postnatal day 0 (P0). Soleus muscles were examined on P1, P7, P10, P14, P21, P28 and in adult rats. Fifteen to 18 pups at each postnatal day and five 2-month-old adult rats were used.

Immunohistochemistry Animals were killed by means of ether anaesthesia. The hind limbs were dissected and tied at the heel with strings in order to extend the soleus muscles. They were immersed in 0.1 M phosphate buffer (PB) (pH 7.4) containing 4% paraformaldehyde overnight, placed in 0.1 M PB containing 7% sucrose for 12 h and then placed in 0.1 M PB containing 15% sucrose for 12 h. The soleus muscles were removed, embedded in embedding matrix, and immediately frozen with dry-ice isopentane. Floating longitudinal

801 sections (50 lm) were cut on a cryostat and collected in 0.1 M PB. For labelling with anti-PGP 9.5 or anti-S100 antibody, the sections were incubated in 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h at room temperature and then stained with: (1) rabbit polyclonal anti-PGP 9.5 (Ultraclone), at a dilution of 1 : 2000, or rabbit polyclonal anti-S100 (Nichirei) in PBS for 3 days at 4° C; (2) biotinylated goat anti-rabbit IgG (1 : 200) (Vector) for 12 h at 4° C; and (3) fluorescein (FITC)-conjugated streptavidin (1 : 200) (Vector) for 12 h at 4° C. Neuronal elements or SCs were clearly stained with anti-PGP9.5 or anti-S-100 antibody, respectively. Control sections were processed identically and in parallel; however they were incubated with PBS instead of with the primary antibody. No stained structures were seen in these controls. Some of the immunostained sections were then processed for cholinesterase (ChE) histochemistry (Pestronk & Drachman, 1978) in order to identify the sites of the NMJs. To count the number of terminal SCs located in the NMJs, some of the sections for S100 immunostaining were counterstained with 5 lm 4@,6diamidino-2-phenylindole (DAPI, Polyscience); they were processed with a mixture of DAPI and FITC-conjugated streptavidin, instead of the procedure (3) described above. About 100 NMJs which comprised entire NMJs, as judged from a surface view, were randomly selected from the muscles of three animals at each age. The number of terminal SCs was counted under a fluorescence microscope (Zeiss Axiophot). The nuclear site of FITC-labelled terminal SCs was identified by switching from filters with maximum excitation at 485 nm to ultraviolet filters with maximum excitation at 365 nm. Specimens obtained from P7, P10, P14 and adult rats were double-labelled with two rabbit polyclonal antibodies, using the following quenching technique (Volpe et al., 1993). The sections were first immunostained with rabbit polyclonal anti-PGP 9.5 antibody and labelled with FITC as mentioned above. They were then saturated with an excess amount of goat anti-rabbit IgG. Secondly, they were incubated with: (1) rabbit polyclonal anti-S100 for 3 days at 4° C; and (2) Texas red-conjugated donkey anti-rabbit Ig (Amersham) (1 : 200) for 12 h at 4° C. No difference in morphology was noted on either the PGP 9.5- or S100-immunolabelled structures between the single and double fluorescence immunohistochemistry. The sections were examined by

Figs 1–4. Longitudinal sections of adult and developing soleus muscles immunostained with anti-PGP 9.5 antibody. Fig. 1. Adult. Two PGP 9.5-IR motor axons extend left and right to form each NMJ. On the left is the lateral view of an axon terminal riding on the lower muscle fibre, whereas on the right is the surface view of an axon terminal innervating the upper muscle fibre. The axon terminals are characterized by arborization and terminal varicosities (arows). A fine PGP 9.5-IR, presumably an autonomic nerve (arrowheads) running with the blood vessel, can be seen near the left NMJ. Fig. 2. At P7. Intramuscular nerve bundles consisting of fine PGP 9.5-IR axons and their branches can be seen. Two terminal regions (asterisks) are in focus. Fig. 3. At P14. An NMJ innervated by two axons is in focus (asterisk). Three retraction bulbs (arrowheads) can be seen close to the site of dispersion of the axon bundle. Fig. 4. At P28. An axon terminal with three terminal branches, each of which appears to connect, and terminal varicosities (arrow). Fig. 5. At P7. The section is double-stained for S100 (a) and ChE (b). The regions of S100-IR terminal SCs (arrows) are identical to the ChE-positive site (arrows). Terminal SCs form unperforated discs. Scale bar: 20 lm.

802

H I R A T A , Z H O U , N A K A M U R A and K A WA B U C H I

Schwann cells in synapse elimination a fluorescence microscope (Zeiss Axiophot) or a confocal laser scanning microscope.

Confocal laser scanning microscopy To detect the relationship between motor axons and SCs the sections double-labelled with FITC and Texas red were scanned using excitation at 488 nm (argon laser) for FITC and 568 nm (krypton laser) for Texas red by the confocal laser scanning imaging system (LSM-GB200, Olympus). Serial optical sections of each fluorescence (at consecutive focal levels of 0.5 lm) were separately taken to Channel 1 and Channel 2 in order to avoid any cross talk and then they were superimposed. Some images were taken using a ]20 or ]40 objective lens to capture the entire structure of terminal nerve branches leading to NMJs. Other images were taken at higher magnification through means of a ]60 oil-immersion objective lens in order to count the number of axons and terminal SCs per NMJ. In the latter case, more than 20 multi-innervated NMJs which possessed the complete structure of an NMJ, presented as a surface view, were randomly selected from the muscles of three animals at each of P7, P10 and P14. By summing the images from consecutive sections, threedimensional morphology of terminal nerve branches or NMJs was reconstructed into a single two-dimensional image. In some cases, z-series of images were examined side by side.

Results PGP 9.5 immunohistochemistry Within adult muscles, a single PGP 9.5-immunoreactive (-IR) axon of the motor terminal nerve branch was observed to extend to and end on a given muscle fibre. Sensory endings, which terminated at muscle spindles, and autonomic nerves surrounding blood vessels were also labelled with anti-PGP 9.5 antibody. PGP 9.5-IR motor axons formed an elaborate arborization with varicosities at the terminal region (Fig. 1), which was identified as an NMJ by ChE staining. In developing NMJs, PGP9.5-IR axon terminals were first detected at P7 (Fig. 2). At this stage the immunoreactivity was usually weak, however, the

803 terminal regions were expanded and showed an immature pattern of arborization. This pattern varied from terminal to terminal. The number of axons entering an NMJ could not be counted under a conventional fluorescent microscope because of their fineness, although the high resolution of a confocal laser scanning microscope revealed that more than two fine axons usually innervated an NMJ (cf. Figs 17 and 19). At P10, a number of NMJs (70 out of 100) showed multi-innervation. A reduction in multi-innervated NMJs occurred at P14 and only eight out of 100 NMJs were innervated by more than two PGP 9.5-IR axons (Fig. 3). From P10 to P14, thin PGP 9.5-IR axons ending with a round structure could be detected (Fig. 3). This round structure was regarded as a retraction bulb, since it was found at various levels, e.g. close to dispersion of the axon bundle or near the level of the NMJs of neighbouring axons. At P21, almost all the NMJs were single-innervated. At P28, PGP 9.5-IR axon terminals possessed clear arborization and varicosities, characteristic of those seen in adults (Fig. 4). S100 immunohistochemistry In the adult and developing muscles examined, the S100-IR SC sheath of terminal nerve branches extended to each muscle fibre where it was transformed into a disc of S100-IR terminal SCs. S100-IR terminal SCs (Fig. 5a) were located exactly at the site of NMJs as identified by ChE staining (Fig. 5b). In the adult muscles, the morphology of S100-IR terminal SCs differed markedly from that of SC sheaths. The perikarya of terminal SCs were recognized as round or elliptical areas, from which a few processes extended, sometimes ramifying, toward neighbouring terminal SCs in order to make contact with them. There was often a weakly stained frill around the processes. The complex figure of the processes seemed to be similar to that of the branches of PGP 9.5-IR axon terminals. Two to four terminal SCs were usually present at each NMJ. As a whole, terminal SCs formed a perforated disc which had several holes at the NMJ (Fig. 6).

Figs 6–11. Longitudinal sections of the adult and developing soleus muscles immunostained with anti-S100 antibody. Fig. 6. Adult. Two perforated discs seen at the terminal regions. Each disc consists of three S100-IR terminal SCs. Terminal SCs are characterized by a round or elliptical perikarya (asterisks) and the cytoplasmic processes with ramification, showing an irregular, rough contour. Note some holes (h) between the processes. The figure comprises two photographs with different focal planes. SC sheaths leading to these terminal SCs are out of focus. Fig. 7. At P7. Several small unperforated discs can be seen at the terminal regions. Each disc consists of S100-IR terminal SCs with poor cytoplasmic processes. An S100-IR SC sheath (arrowhead) transiting to the disc is clearly shown at the centre. Fig. 8. At P14. Two discs with newly-formed holes (arrows) can be seen. Note that the S100-IR SC sheaths of terminal nerve branches arising from the bundle are irregular in width and longer than those seen at P7. Fig. 9. At P14. A disc with numerous finger-like processes arising from S100-IR terminal SCs can be seen. Fig. 10. At P28. The top and bottom perforated discs are clearly portrayed. The number of S100-IR terminal SCs at each disc is greater and the size of the discs is longer compared with those at P14. Arrows indicate holes in the discs. Fig. 11. At P28. Another example of a perforated disc seen at P28. The disc is formed by two terminal SCs with considerably rich cytoplasm. Scale bar: 20 lm.

804

H I R A T A , Z H O U , N A K A M U R A and K A WA B U C H I

During development, terminal SCs gradually matured in their morphology and gradually increased in number. At P7, one or two S100-IR terminal SCs were found at each NMJ (Figs 5a and 7). The terminal SCs had only small tapering processes (arrowheads in Fig. 7) or small finger-like processes (Fig. 5a) extending from the margin. They formed an unperforated disc, with no structure corresponding to PGP 9.5-IR terminal arborization being found at P7. Thus, the arborization of the axon terminals preceded that of the terminal SCs. At P14, in addition to the unperforated discs, perforated discs with a few small holes appeared (Fig. 8). They usually consisted of two or three S100-IR terminal SCs with more developed processes. Sometimes terminal SCs had numerous fingerlike processes (Fig. 9). At P28, the number of terminal SCs per NMJ and their morphology resembled those of adults. As a result, terminal SCs formed a perforated disc with a small number of holes (Figs 10 and 11). The gradual increase in the number of terminal SCs in an NMJ during development is shown in Fig. 12. Confocal laser scanning microscopy The images of FITC and Texas red were taken separately (Figs 13a, 13b, 14a, 14b, 15a and 15c) and then superimposed to analyse the relationship between axons and SCs (Figs 15b and 16). In adults, a one : one relationship of axon-to-SC sheath was detected at the preterminal region; a single FITC-labelled axon was overlaid by a single Texas red-labelled SC sheath. At the terminal region a single FITC-labelled axon with a complex branching pattern was completely overlaid by several Texas red-labelled terminal SCs with elaborate processes; the ratio between axons and terminal SCs was one : multiple (Figs 15a–c), respectively. The change in their relationship during synapse elimination was revealed by analysing the multiinnervated NMJs seen from P7 to P14. At P7, the multiple axons innervating a single NMJ were apparently associated with a single SC sheath (Fig. 17). The axons were usually extremely fine and were about 0.6 lm in maximum width. Some of the axons had a round expanded structure (2.0 lm in maximum width) in the preterminal region (Fig. 17, arrow). On the other hand, at P14, the multiple axons converging on a single NMJ were more well-developed both in width and length, with each one often showing an individual one : one relationship with the SC sheath (Fig. 18). The size of axons was about 2.0 lm in maximum width at the preterminal region of some axons which showed a relatively regular and smooth outline, as seen in Fig. 18 (arrow), whereas it varied from region to region in other axons, which often possessed a round expanded structure (3.0–4.0 lm in maximum width). The ratio between axons and terminal SCs also changed with age. At P7 in most of the multi-innervated

Fig. 12. Frequency histograms of the number of S100-IR terminal SCs per NMJ. The number of terminal SCs was verified on double-labelled preparations of S100-immunostaining and DAPI nuclear-staining, by counting the number of DAPI-labelled nuclei within the S100-IR terminal SCs. The number of NMJs examined at each age group is indicated within parentheses.

NMJs analysed the ratio between axon and terminal SC was two or more : one (Fig. 19), whereas at P14, the ratio was mostly two : two (Fig. 14). Figure 20 shows the gradual change in the ratio of axon and terminal SC in multi-innervated NMJs from P7 to P14. Confocal laser microscopy also revealed the relationship between SCs and the axon ending with the retraction bulb, which was found at P10 or P14. The retraction bulb in the eight analysed axons was round in structure with a maximum width of 2.5–3.5 lm. It was always overlaid by a single round SC with no processes (Fig. 16). The retraction bulb was similar to the expanded structure which was detected at the part more proximal to the terminal region. Therefore, a careful analysis of the z-series of images prior to reconstruction was undertaken in order to determine whether they comprised retraction bulbs or expanded structures (Fig. 13). Discussion In the present study, the morphology of S100-IR terminal SCs in adult NMJs was extremely similar to that observed by scanning and transmission electron microscopy (Ogata & Yamasaki, 1984): the cell body swelled roundly, from which a few cytoplasmic processes then extended downward and divided several times. The lateral sides and ends of terminal SC processes were frilled with small projections. Thus, the entire outline of the cells could be expressed by S100 immunostaining. S100 immunohistochemistry also showed a gradual change in the morphology of the

Schwann cells in synapse elimination Fig. 13. Confocal images of FITC-labelled motor axons (a) and Texas red-labelled SCs (b) at P14. Four section images (12, 15, 18 and 21) from a 30-section series with 0.5 lm intervals are shown in each of (a) and (b), in order to visualize the portion of the retraction bulb (arrows). Scale bar: 10 lm. Fig. 14. Images of FITC-labelled axons (a) and Texas red-labelled SCs (b) in an NMJ at P14, comprising a 15-section series, Example of the two : two relationship between axon and terminal SC. In this NMJ, the relationship between axon and SC sheath at the preterminal region is clear: each of the axons shows a one : one relationship with the SC sheath. Scale bar: 10 lm.

805

806

H I R A T A , Z H O U , N A K A M U R A and K A WA B U C H I

Schwann cells in synapse elimination developing terminal SCs, including the gradual increase in their number within each NMJ. PGP9.5 immunohistochemistry also revealed the entire morphology of the adult motor axon terminals, as previously reported by Ann and colleagues (1994). In addition, it demonstrated the morphological details of the developing motor axon terminals, the occurrence of retraction bulbs and the time course of the synapse elimination. Thus, S100 and PGP 9.5 can both be regarded as suitable markers for studying the morphology of SCs and the axons of the NMJs, respectively, in the developing peripheral motor nerves. The combined technique of double immunofluorescence staining plus analysis by confocal laser scanning microscopy enabled us to show the full details of the morphological relationship between axons and SCs. The time course of the synapse elimination was demonstrated in relation to the number of multiinnervated NMJs. At P7, the majority of the NMJs were multi-innervated. Subsequent abrupt reduction occurred between P10 and P14. The peak pattern of multiinnervation at P7 has been reported in physiological studies (Brown et al., 1976; O’Brien et al., 1978). Brown and colleagues (1976) suggested that the loss of extra synapses may be well under way long before P10, since three or more e.p.p. components were seen in many muscle fibres immediately after birth. With PGP 9.5 immunohistochemistry, retraction bulbs, which are characteristic of axonal withdrawal (Riley, 1977, 1981; Balice-Gordon et al., 1993), could not be detected at P7. Instead, an expanded structure whose diameter was similar to that of the retraction bulbs was detectable. This structure presumably may be involved in the dynamic process of axonal withdrawal, since it was observed in the withdrawing axons ending with the

807 retraction bulb at P14 (Fig. 16). This finding indicates that the synapse elimination may occur as early as P7. Thus, our result is consistent with the physiological data mentioned above. The frequent occurrence of the retraction bulbs also supports the idea that the mechanism of synapse elimination is mainly due to the retraction (Korneliussen & Jansen, 1976; Riley, 1977, 1988; Balice-Gordon et al., 1993) rather than the degeneration of redundant axons (Rosenthal & Taraskevich, 1977). Confocal laser scanning microscopy demonstrated that the relationship between axons and SCs in multiinnervated NMJs gradually matured with age. The sequence of the event can be illustrated as in Fig. 21, based on the change in morphology and ratio in the majority of NMJs examined, although NMJs were not always in the same state of maturation at any one time (Riley, 1981; Steinbach, 1981; Slater, 1982). At P7, axons demonstrated an immature relationship to SCs. At the terminal region, two or more axons were covered by a single terminal SC, whereas at the preterminal region these multiple axons seemed to be associated with a single SC sheath, as seen at a more proximal level of the developing peripheral nerve fibre (Webster et al., 1973; Ziskind-Conhaim, 1988; Jessen & Mirsky, 1991), although the latter finding at the preterminal region needs to be confirmed by electron microscopy. Even in this stage, axonal withdrawal may take place as discussed. At P14, in multi-innervated NMJs at the terminal region their ratio was two : two, although these axons had already established a mature one : one relationship with the SC sheath at the preterminal region. Finally, the monoaxonal innervation was completed after one of the two axons withdrew. Thus, even under the condition of multi-innervation, the relationship between axons and SCs seems to change,

Fig. 15. Pseudocolour images of FITC-labelled axons (a) and Texas red-labelled SCs (c) and their superimposed image (b) of a single optical section through an adult NMJ. The yellow in (b) indicates the neuronal element overlaid by the SC element. Note that a single axon is associated with a single SC sheath at the preterminal region but with three terminal SCs at the terminal region. Asterisks indicate the nucleus of terminal SCs. Scale bar: 10 lm. Figs 16–19. Superimposed images of developing motor terminal nerve branches and NMJs. Scale bars: 10 lm. Fig. 16. At P14. Image comprising a 30-section series, including four sections in Fig. 13, shows an axon with a retraction bulb (arrow) covered by a single SC. Note that this axon possessed the expanded portion (arrowhead) at a more proximal part. Fig. 17. Image of single optical section through two NMJs showing the immature features of multi-innervation at P7. Three and two axons run toward top and bottom NMJ, respectively. The former axons are located close to each other and are apparently associated with a single SC sheath. Note that each of the terminal nerve branches is extremely short in length at this age. Fig. 18. Image comprising a 20-section series showing more developed features of multi-innervation at P14. In multiinnervated NMJs (asterisks), two axons running toward the right NMJ wind separately, whereas two axons innervating the left NMJ appear to be pulled apart at the bifurcating site. Compare the length of the terminal nerve branches with those in Fig. 16. Fig. 19. Images of z-series visualizing the entire features of a multi-innervated NMJ at P7. Example of a two : one relationship between axon and terminal SC. Note that the outline of the nucleus of the terminal SC can be detected in the second or third section.

808

H I R A T A , Z H O U , N A K A M U R A and K A WA B U C H I

Fig. 20. Frequency histogram of terminal SCs in multiinnervated NMJs. The majority of multi-innervated NMJs examined were associated with one or two terminal SCs, while only a few NMJs were associated with three or four terminal SCs. The number of axons and terminal SCs per NMJ was counted by confocal laser microscopic analysis on double-labelled sections with anti-S100 and anti-PGP 9.5 antibodies. The number of NMJs examined in each age group is indicated within parentheses.

Fig. 21. A scheme showing the change in the relationship between axons and SCs during synapse elimination. Stage 1: Several axons ensheathed with a single SC sheath innervate a single NMJ with one terminal SC. Axons with an expanded structure appear. Stage 2: An axon with an expanded structure retracts within a single SC sheath. Stage 3: SCs proliferate both at the preterminal and the terminal regions, whereas axons increase in calibre at the preterminal region and expand the synaptic area. Subsequent segregation results in the convergence of two well-developed nerve fibres in an NMJ with two terminal SCs. Stage 4: A single innervation on a muscle fibre is accomplished by retraction of one of two axons, along with SCs.

possibly due to some interaction between them (Jackson & Suter, 1994). The present study revealed that the synapse elimination was protracted for at least one week by the gradual loss of axons (Korneliussen &

Jansen, 1976; Riely, 1977; Balice-Gordon et al., 1993), which showed a different state of maturation regarding their spatial relationship with SCs. It seems to go against the hypothesis if SCs determine a single axon by completing their mature relationship, i.e. myelination, with the axon (Brown et al., 1976), since axons appeared to withdraw even in an immature state with the SC sheath, and it seems most unlikely that only the first axon could establish such a mature relationship; two axons innervating an NMJ were noted to have already established a one : one relationship at P14. In adult muscles, multi-innervation to NMJs transiently occurs during the process of regeneration following denervation. On subsequent withdrawal, SCs are always associated with axons (Reynolds & Woolf, 1992; Woolf et al., 1992; Son & Thompson, 1995a). Some experiments for inducing axonal sprouting, such as partial denervation, paralysis by botulinum toxin and nerve implantation (Chen & Ko, 1994; Son & Thompson, 1995b) have suggested that SCs may not be merely associated with retracting axons, but probably play an active role on them, since SCs lead and guide extending axons. During development, the withdrawing axons were also always associated with SCs. If the monoaxonal innervation to muscle fibres is undertaken in essentially the same way between development and regeneration, as is thought in the case of axonal outgrowth (Brown, 1984; Keynes, 1987 and Martini, 1994 for review), it could be assumed that SCs also play an active role in axonal withdrawal during development. However, a recent study (Trachtenberg & Thompson, 1996) has revealed that SCs during early postnatal days are quite different from those in adults with regard to their dependence upon axons; following denervation, SCs are not able to survive but do induce apoptosis. From our observations, SCs seem to perhaps interact with retracting axons, as suggested by their close association between the axon with a retraction bulb and SCs, although it remains unclear whether or not they play an active role in axonal withdrawal. The present study provides fundamental data on a change in the morphology of SCs and their relationship with axons during the period of synapse elimination. This basic phenomenon will contribute to our further understanding of the mechanisms involved in synapse elimination. Acknowledgements We thank Dr B. Tandler for his suggestions and critical reading of the manuscript, Dr M. Hirata for his encouragement and Mr T. Kanemaru for photographic assistance. The English used in this manuscript was revised by Miss K. Miller (Royal English Language Centre, Fukuoka, Japan).

Schwann cells in synapse elimination

809

References AN N , E . S., MIZ OG UC H I, A ., O K A J IMA , S. & IDE , C .

(1994) Motor axon terminal regeneration as studied by protein gene product 9.5 immunohistochemistry in the rat. Archives of Histology and Cytology 57, 317–30. BAL I C E - G OR DON , R . , C HU A, C . K . , N EL S ON , C . C. & L IC H TM AN, J. W. (1993) Gradual loss of synaptic

cartels precedes axon withdrawal at developing neuromuscular junctions. Neuron 11, 801–15. BIXB Y, J . L . (1981) Ultrastructral observations on synapse elimination in neonatal rabbit skeletal muscle. Journal of Neurocytology 10, 81–100. BROCKES, J. P., FIELDS, K. L. & RAFF, M. C. (1977) A surface antigenic marker for rat Schwann cells. Nature 266, 364–6. B ROWN , M . C ., JAN SEN , J. K . S. & V AN E SS E N , D. (1976) Polyneuronal innervation of skeletal muscle in newborn rats and its elimination during maturation. Journal of Physiology 261, 387–422. BROWN, M . C. (1984) Sprouting of motor nerves in adult muscles: a recapitulation of ontogeny. Trends in Neurosciences 7, 10–14. CH E N, L. & KO , C . P. (1994) Extension of synaptic extracellular matrix during nerve terminal sprouting in living frog neuromuscular junctions. Journal of Neuroscience 14, 796–808. COL M AN , H . & L IC H T MA N , J. W . (1993) Interactions between nerve and muscle: Synapse elimination at the developing neuromuscular junction. Developmental Biology 156, 1–10. COL M AN , H . , N A BE KU RA , J. & LI C HT M AN , J . W .

(1997) Alterations in synaptic strength preceding axon withdrawal. Science 275, 356–61. GE O RG IO U , J . , R OB I TA IL L E , R. , T RI M B L E , W . S. & C H AR L TON , M . P. (1994) Synaptic regulation of glial

protein expression in vivo. Neuron 12, 443–55. JACK SON , S. G . & SUTE R , U. (1994) Signaling pathways

mediating axon-Schwann cell interactions. Trends in Neurosciences 17, 399–401. JE S SEN , K . R. & MIRSK Y, R. (1991) Schwann cell precursors and their development. Glia 4, 185–94. KE YN E S, R . J . (1987) Schwann cells during neural development and regeneration: leaders or followers? Trends in Neurosciences 10, 137–9. KO RN E L I U SS E N , H . & J A N SE N, J . K. S. (1976) Morphological aspects of the elimination of polyneuronal innervation of skeletal muscle fibers in newborn rats. Journal of Neurocytology 5, 591–604. MA RTI NI , R . (1994) Expression and functional roles of neural cell surface molecules and extracellular matrix components during development and regeneration of peripheral nerves. Journal of Neurocytology 23, 1–28. O’ B R I EN , R. A . D. , OST B E RG, A. J . C . & VRB OVA , G .

(1978) Observations of the elimination of polyneuronal innervation in developing mammalian skeletal muscle. Journal of Physiology 282, 571–82. OG AT A , T . & YA MA SA KI , Y . (1984) Scanning electron microscope studies on the Schwann cells in rat motor endplates with special reference to their finger-like projections. Archirum Histologicum Japonicum 47, 533–9. PE ST RO N K, A . & DRA C H MA N , D. B . (1978) A new stain for quantitative measurement of sprouting at neuromuscular junctions. Muscle and Nerve 1, 70–4.

PU RVE S, D. & L I C H TMA N , J . W . (1980) Elimination of

synapses in the developing nervous system. Science 210, 153–7. RE YNO L DS, M. L . & WOO L F , C . J . (1992) Terminal Schwann cells elaborate extensive processes following denervation of the motor endplate. Journal of Neurocytology 21, 50–66. RI L E Y, D. A. (1977) Spontaneous elimination of nerve terminals from the endplates of developing skeletal myofibres. Brain Research 134, 279–85. RI L E Y, D. A . (1981) Ultrastructural evidence for axon retraction during the spontaneous elimination of polyneuronal innervation of the rat soleus muscle. Journal of Neurocytology 10, 425–40. RO SE NT H AL , J. L. & T AR ASK E VICH , P. S. (1977) Reduction of multiaxonal innervation at the neuromuscular junction of the rat during development. Journal of Physiology 270, 299–310. SLAT ER , C . R. (1982) Postnatal maturation of nerve-muscle junctions in hindlimb muscles of the mouse. Developmental Biology 94, 11–22. SON, Y. J. & TH O MPS ON , W. J. (1995a) Schwann cell processes guide regeneration of peripheral axons. Neuron 14, 125–132. SON, Y. J. & T H OMP S ON , W. J . (1995b) Nerve sprouting in muscle is induced and guided by processes extended by Schwann cells. Neuron 14, 133–41. STE FAN SSO N , K . , WO L L MA N N , R . L . & MO OR E , B . W . (1982) Distribution of S100 protein outside the

central nervous system. Brain Research 234, 309–17. STE IN B AC H , J . H . (1981) Developmental changes in acetyl-

choline receptor aggregates at rat skeletal neuromuscular junctions. Developmental Biology 84, 267–76. TH OM PSO N , R . J., DORA N, J. F . , JA CK SO N , P ., DH IL L ON , A . P . & RO D E , J . (1983) PGP 9.5 -a new marker

for vertebrate neurons and neuroendocrine cells. Brain Research 278, 224–8. TRA C H T EN BE RG, J. T . & TH O MPSO N , W. J. (1996) Schwann cell apoptosis at developing neuromuscular junctions is regulated by glial growth factor. Nature 379, 174–77. VO L PE , P., B IRA L , D. , PIZZ O, P., S AL V IA TI, G. & MA RGRE TH , A . (1993) Ontogenesis of chick iris intrinsic

muscles: Evidence for a smooth-to-striated muscle transition. Developmental Biology 159, 441–9. WE B STE R , H .

DE F . ,

M AR TI N , J. R. & O’ C ON NE L L , M. F .

(1973) The relationships between interphase Schwann cells and axons before myelination: A quantitative electron microscopic study. Developmental Biology 32, 401–16. WIL K IN SON , K . D. , L E E , K. , DE SH PA N DE , S. , DU E RKSE N - H U GH E S, P. , B O SS, J . M . & PO H L , J . (1989) The

neuron-specific protein PGP 9.5 is a ubiquitin carboxylterminal hydrolase. Science 246, 670–3. WOO L F , C . J . , RE YN OL DS , M. L . , C HO N G, M . S. , E MSON, P., I RW IN , N . & BE N OW IT Z, L . I . (1992)

Denervation of the motor endplate results in the rapid expression by terminal Schwann cells of the Growth-associated Protein GAP-43. Journal of Neuroscience 12, 3999–4010. ZI SKI N D- C ON H A I M, L . (1988) Physiological and morphological changes in developing peripheral nerves of rat embryos. Developmental Brain Research 42, 15–28.