Formation of primary and secondary myotubes in rat lumbrical muscles

This is the first of two papers concerned with the determination of skeletal muscle fibre numbers during embryonic development. It has previously been shown (Harris, 1981) that primary myotubes develop autonomously, but generation of secondary myotubes requires both the presence and the activity of motor innervation. Now it is of interest to ask whether innervation is regulatory or permissive for the formation of secondary myotubes. In this paper, we give a quantitative description of some aspects of development of the IVth lumbrical muscle in the rat hindlimb and of its innervation; in the next paper (Ross, Duxson & Harris, 1987), we describe the effects of partial reduction of the muscle’s innervation during development.

The major features of skeletal muscle formation are well known. Following the seminal work of Kelly & Zacks (1969a), several authors have confirmed the biphasic nature of muscle development (Ashmore, Robinson, Rattray & Doerr, 1972; Ontell & Dunn, 1978; Harris, 1981; Stickland, 1982; Ontell & Kozeka, 1984a,b). Mononucleate myoblasts fuse to form primary myotubes which extend from tendon to tendon of the embryonic muscle (Dennis, Ziskind-Conhaim & Harris, 1981; Ontell & Kozeka, 1984a), forming a cellular framework for the later generation of secondary myotubes. Secondary myotubes form near the midpoint of primary myotubes under the basal lamina, resulting in clusters of muscle cells within a single basal lamina. The secondary myotubes then grow longitudinally, eventually connecting with the muscle tendon and separating from the primary myotube as they acquire their own basal lamina. Subsequently, secondary myotubes are difficult to distinguish ultrastructurally from primary myotubes. Here we give a quantitative description of the normal development of a particular muscle (the IVth lumbrical in the rat hindfoot) and of the timing of moto-neurone death in its motor pool, as a prelude to a study of the neural regulation of muscle development (Ross et al. 1987).

We chose the lumbrical muscle for this study because an entire cross section fits on a single electron microscope grid, allowing all the myogenic cells to be identified and counted. The motor innervation to the IVth lumbrical, previously described by Betz, Caldwell & Ribchester (1979), leaves the spinal cord in the L4, L5 and L6 ventral roots.

All experiments were performed on White Wistar rats. Pregnancies were dated by the appearance of a copulatory plug, with 9 a.m. on the day the plug was found being EO. Embryos from at least three different litters were taken at daily intervals from E15 onwards; the numbers of muscles and ventral roots from which photomontages were prepared are given in Table 1. PN28 tissues came from three litters of control animals employed in another study, currently in progress. The mother was killed with a blow to the head, and the embryos quickly removed and placed on ice for anaesthesia. They were then perfused through the heart with heparinized saline followed by fixative, both at room temperature. The fixative contained paraformaldehyde 1%, glutaraldehyde 1%, and CaCl₂ 0·4mM in 0·1 M-phosphate buffer, pH7·4. Glucose was added to the buffer to increase its tonicity, with total molar concentrations (glucose + phosphate) of 0·25 M for E15, E16 and E17, 0·2 M for E18 and E19, and 0·15 M for E20, E21 and E22 embryos. Following perfusion, the lumbrical muscles and lumbar spinal cords were immersed in fixative at 4°C for a further 2h. Final dissections were done in 0·1 M-phosphate buffer with glucose added as above to match the age of the embryo. The tissue was then postfixed in osmium tetroxide and uranyl acetate, dehydrated and embedded in TAAB epoxy resin for transverse sectioning. The plane of section and progress through the tissue was assessed by viewing 1 μm sections with light microscopy. Muscles were sectioned in the midbelly endplate-containing region. Ventral roots were cut 1-2 mm proximal to the dorsal root ganglion. In one embryo at each day from E16-E19, branches of the plantar nerve were visualized in a cross section of the distal hindlimb at the level of the malleolus. This was chosen as the most compact region of this nerve, allowing the best comparison of axon number between animals of different ages. Ultrathin sections were in all cases mounted on single-slot Formvar-coated copper grids, stained with uranyl acetate and lead citrate and then viewed with a Phillips 300 or 410 electron microscope. Photomontages of entire lumbrical muscles, ventral roots and all branches of the plantar nerve were produced at final magnifications in the range X6300-11400. Cells in the cross sections were classified and counted as described in the Results.

A serial and semiserial EM study was also made of a single E21 IVth lumbrical muscle in order to clarify some aspects of the muscle organization. Sectioning commenced in the midbelly region of the muscle and continued into the tendon, a total length of about 400μm. If the position of the origin is designated as zero along the longitudinal axis of the muscle, regions from 0-10μm, 105-160 μm and 220-400μm were serially sectioned and all sections collected; in remaining regions ultrathin sections were collected at intervals of 2·5 μm.

Choline acetyltransferase (ChAT) (acetyl-CoA: choline O- acetyltransferase, EC 2.3.1.6.) activity was assayed in homogenates of whole hindfeet as an index of nerve terminal numbers in the intrinsic muscles of the foot. Embryos were taken as described above, and hindfeet removed by section through the ankle joint and placed in 20 or 50 μl of homogenization buffer depending on the age of the embryo, quick frozen in liquid nitrogen and stored at —80°C until they were assayed. Assays were made with the technique of Fonnum (1969), using [¹⁴C]acetyl CoA (Amersham), specific activity 58mCimM^-1, at a final concentration of 82μM. Total enzyme activity in the foot was calculated taking into account the different volumes of feet of different ages. Control experiments ensured that activities were independent of the volume of homogenization buffer and that the assay was linear. Specificity of enzyme activity was tested by running assays in the presence of 4x10⁻³M-4-(l-naphthylvinyl) pyridine (Calbiochem), said to be a specific inhibitor of ChAT (White & Cavallito, 1970).

All embryos from at least two litters were examined on each embryonic day from E15 to E20 inclusive; numbers of tissues sampled are given in Table 1. Motoneurone death was inhibited by placing slow-release capsules containing T1X into the amnion of individual embryos at E15 (Harris, 1981; Harris & McCaig, 1984). The criterion for successful paralysis was complete flaccidity of the embryo and a lack of reflex response to electrical stimulation of the snout.

Development of the IVth lumbrical muscle in the rat hindfoot is similar to that described by Kelly & Zacks (1969a) for intercostal muscles, except that each developmental stage is retarded by 2-3 days. The first signs of lumbrical muscle formation were seen on E15 when a condensation of mesenchymal cells appeared between two whorls of connective tissue running up into the IVth and Vth digits. Axons appeared, singly and in bundles, within this condensation and frequently made close contacts (≤20 nm) with the undifferentiated cells. At these points of contact, some axons showed a local enlargement lying in a surface depression on the mesenchymal cell while others had no such specializations.

One day later (E16), a few small (3-5μm) diameter myofilament-containing cells were present within the muscle mass. These were the earliest primary myotubes. They were irregular in cross section with central nuclei, modest numbers of myofilaments dispersed throughout the cell, and small deposits of lipid and glycogen. Small patches of amorphous material, presumably a rudimentary basal lamina, occurred on the plasmalemma. Large numbers of axons were seen within the muscle and these made prolific contacts with both myotubes and undifferentiated cells. During the period E16-E18, intramuscular axons could be defined as belonging to motor or sensory neurones only rarely, when motor nerve terminals contained synaptic vesicles.

By E17, the primary myotubes had enlarged and contained abundant myofilaments, lipid and glycogen. They occurred in groups of three to eight cells with an incomplete basal lamina irregularly surrounding the group. Gap junctions between myotubes were frequent. Large numbers of axons coursed through the muscle, while nerve-myotube junctions were still simple appositions between an unspecialized region of sarcolemma and a varicose nerve terminal containing an occasional dense-core vesicle. Two primary myotubes sometimes shared an individual nerve terminal. Fibroblasts, distinguished as elongated cells containing abundant rough endoplasmic reticulum, were also present within the muscle, along with irregularly shaped undifferentiated cells. The latter had a moderately dense cytoplasm and heterochromatic nuclei and are presumed to be a mixture of myoblasts and fibroblast precursors (Quinn, Namer- off & Holtzer, 1984).

During the next day (E18), the primary myotubes (Fig. 1A) progressively separated from their clusters, but no new cells appeared. On E19, however, small myofilament-containing cells were seen beneath the basal lamina of a few of these separate primary myotubes (Fig. IB). These were the earliest secondary myotubes and their number progressively increased during the remainder of the embryonic period. At E19, motor nerve terminals could be identified lying outside the basal lamina of the muscle cells while sensory nerve terminals lay close to the plasmalemma. Muscle spindles were first obvious on E20. They were distinguished by the capsule surrounding the midregion of the intrafusal fibre and by the presence of sensory nerve terminals; the IVth lumbrical muscle typically contained two or three spindles. By E21, most primary myotubes had one to three associated secondary myotubes of varying maturity (Fig. 2A). Undifferentiated cells, presumably myoblasts, were present beneath the basal lamina of primary myotubes. In addition, medium-sized myotubes began to appear in isolation (Fig. 2A): these cells are thought to be secondary myotubes which have separated from their primary clusters (Ontell & Kozeka, 1984a). By E22, it became difficult to distinguish between primary and maturing secondary myotubes, except on the basis of the presence or absence of associated young myotubes.

The maturation of nerve-muscle junctions occurred in a way similar to that described by Teräväinen (1968) and Kelly & Zacks (1969b). We noted the presence of close contacts between nerve terminals and mononucleate cells (Fig. 2B) during the period when secondary myotubes were forming.

All myotubes in the endplate region of lumbrical muscles were counted in a series of muscles aged from E15-E22 inclusive (Fig. 3) (see Table 1 for the number of muscles sampled). Primary myotube generation began after E15 and reached a plateau of 102 ± 4 cells by E17. No further myotubes were generated for two days, while the existing myotubes matured. A few secondary myotubes (4 ± 0·8, n = 9) were seen for the first time on E19, presaging a period of rapid generation; 65 secondaries, on average, were added between E19 and E20; 100 between E20 and E21; and 114 between E21 and E22. Myotube generation was far from complete by birth: six muscles from PN28 rats contained 919 ± 15 muscle fibres; hence about 530 secondary myotubes were added postnatally. Betz et al. (1979), in a different strain of rat, counted an adult number of 938 ±79 lumbrical muscle cells.

The number of myoblasts or presumptive myoblasts in lumbrical muscles of different ages was estimated by counting all undifferentiated mononuclear cells in one midbelly cross section from each muscle (Fig. 4). There was a constant number (mean value 22 ± 2·3) during E16-E18, which then increased during E18-E21 to reach a mean value of 55 ± 4 on E21. The number of undifferentiated mononucleate cells present on each day during E19-21 was well correlated (r² = 0·97) with the rate of production of secondary myotubes on that day (Fig. 5).

The time course of motoneurone death in the lumbar spinal cord was estimated by counting L4 ventral root axon numbers on successive embryonic days (Fig. 6). The major period of ventral root axon loss was between E16 and E17 when axon numbers fell from 12000 to 4000 (67% of the total loss). Occasional signs of axon degeneration, in the form of myelin figures and small regions of cellular debris, were seen in the ventral roots over this same period. Axon numbers continued to slowly decline over the next two days before stabilizing at 2175 by E19; overall 81 % of axons present on E15 were lost from the ventral root by E21.

The lumbrical muscles develop later than most proximal muscles in the hindlimb, and their motoneurones make a very small contribution to axons in the L4 ventral root. To obtain a parameter better related to lumbrical motoneurone numbers, we counted axons in all branches of the plantar nerve at the level of the ankle, using one embryo at each day during E16-E19 (Table 2). Growth cones were present in the nerve branches at all these times, indicating that axons were still growing into the foot throughout this period. There was a peak number of axons on E18, which declined by 23 % by E19. In one embryo (E17), axons were counted above and below the ankle; there were 8 % more axons below the ankle, indicating that some axons had branched. Judging by the large number of axons in the plantar nerve branches, it is likely that the majority was sensory.

Hindfoot ChAT activity progressively increased with increasing embryonic age, except for a marked inflexion during E17-E19 (Fig. 7). This inflexion was not present in the curve of activity versus time for hindfeet from embryos paralysed with TTX, which is known to inhibit motoneurone death (Harris & McCaig, 1984). Accordingly, we interpret this inflexion as reflecting death of a substantial proportion of motor axons projecting to the intrinsic muscles of the hindfoot on E17, with consequent loss of their ChAT content, while those remaining continued to accumulate Ch AT at the normal rate.

A serial and semiserial EM study was made of a single E21 lumbrical muscle to examine the frequency and distribution of intercellular junctions between developing muscle fibres. The muscle was sectioned from the midbelly to the tendon, as described in the methods.

Gap junctions between myotubes were common throughout the entire length of the E21 lumbrical muscle. These junctions consisted of parallel electron-opaque membranes separated by a gap of 2-4 nm. The entire junctional complex was about 15 nm wide, similar to that reported by Schmalbruch (1982) in newborn rat lumbrical muscles. The distribution of gap junctions between cells varied according to the maturity of the myotubes. Young secondary myotubes within clusters were always extensively electrically coupled to the parent primary myotube along their entire length. More mature secondary myotubes, which had separated from the primary myotubes in their midzone, were still extensively coupled to the primary by gap junctions located near the ends of the secondary myotubes. Even the most mature secondary myotubes present in the E21 muscle still rejoined their primary cluster for some part of their length, and gap junctions between primary and secondary myotubes were commonly seen in these regions.

Gap junctions between neighbouring primary myotubes were also occasionally observed, close to the tendon. In two cases, junctions were seen between mononucleated cells and a filamented cell, and on a single occasion, between two mononucleated cells. Gap junctions involving mononucleated cells were less electron dense than those between myotubes and their apparent rarity may result from difficulties in identifying them.

Mechanical attachment of developing myotubes to tendons was also examined in the serial section study. At E21, primary myotubes had a classical insertion onto the collagen fibrillar network which needs no further description here. Secondary myotubes, how-ever, rarely inserted directly onto the tendon but at most stages of maturity were attached only to the primary myotube of their cluster. Along much of the length of the secondary myotube, numerous finger-like processes from the cell delved deeply into the primary myotube, forming an intricate attachment between the cells (Fig. 8). Desmosome-like structures were common between secondary and primary myotubes in these regions. These junctions had parallel electron-dense membranes separated by a regular gap of about 25 nm, but the sub-membranous dense plaques, which are a feature of classical desmo-somes, were never apparent.

Only the most mature secondary myotubes, those which had separated from the primary cluster for most of their length, formed direct attachments to the collagenous network of the muscle tendon. Even these myotubes, at this age, maintained some small regions of attachment to the primary myotube.

This study extends those of Kelly & Zacks (1969a); Ashmore et al. 1972; Ontell & Dunn (1978); Ontell & Kozeka (1984a,b) and others, by tracing and quantifying mammalian myotube production and differentiation from the earliest times, and is the prelude to a study on the mechanism of regulation of muscle fibre numbers.

Primary myotube generation in the lumbrical muscle begins after E15 and is complete by E17. No new myotubes are added for a further two days, until the first secondary myotubes appear on E19. A similar delay between primary and secondary myotube generation has been described in the mouse extensor digitorum longus (Ontell & Kozeka, 1984a). Ontell & Kozeka (1984b) suggest that primary myotubes require at least 2 days to mature sufficiently for them to act as scaffolding for secondary myotube formation. A contrasting view, supported by the experiments of White, Bonner, Nelson & Hauschka (1975), Rutz, Haney & Hauschka (1982) and Quinn et al. (1984), is that the myoblasts that fuse to form primary and secondary myotubes descend from different classes of myogenic stem cell, so that the time delay reflects the absence of activated secondary myotube precursor cells.

More than half the lumbrical muscle secondary myotubes are added after birth. Postnatal formation of myotubes has been suggested to be common in rat development (Chiakulas & Pauli, 1965; Rayne & Crawford, 1975), but electron microscopy (Ontell & Dunn, 1978; Ontell, 1979) shows that in most muscles the mature number of myotubes is present at birth. The muscle fibres are packed in clusters and the apparent proliferation seen with light microscopy is the result of separation of myotubes from these clusters. The hindfoot lumbrical muscles are an exception to this rule and may be among the last muscles in the body to develop.

Primary myotubes run from tendon to tendon from the earliest times of their development (Dennis et al. 1981; Ontell & Kozeka, 1984a). It is therefore valid to assume that all primary myotubes will be seen in a single section across the midbelly endplate-containing region of a developing muscle, regardless of its age. In the lumbrical muscle, these will typically include three myotubes destined to be part of the muscle spindles (these were not counted in the older muscles where they could be identified). In rat extensor digitorum longus muscles, newly formed secondary myotubes were short and at first randomly distributed over the middle two thirds of the muscle so that only 60-70 % spanned its point of widest girth (Ontell & Kozeka, 1984a). Lumbrical muscles are fusiform so that muscle fibre-tendon relations are simpler than in the pennate EDL and, in the serial section study of an E21 lumbrical muscle, all secondary myotubes were centred on the endplate zone. Thus, our counts would include all secondary myotubes more than a few hours old.

Muscle cell death has a role in muscle morphogenesis, but we did not see degenerating myotubes in lumbrical muscles during embryogenesis. By analogy with other muscles, it would not be expected until most secondary myotubes had formed; our preliminary observations indicate that it occurs after PN10 in the lumbricals and so does not interfere with the accuracy of our counts.

The number of undifferentiated mononucleate cells in a single muscle cross section was used as an estimate of the number of myotube precursor cells in the muscle. The cells counted include myoblasts and myoblast and fibroblast precursor cells; at present, it is not possible to distinguish between these cell types.

The increase in number of these cells during the last four days in utero, concomitant with the appearance of secondary myotubes, we take to reflect an increased proliferation of myoblasts.

Myogenic cells have at least three possible fates: to give rise to primary myotubes, secondary myotubes or muscle satellite cells (Armand, Boutineau, Mauger, Pautou & Kieny, 1983). Whether these classes constitute a single lineage or reflect the presence of two or several stem cell lines is not known (Quinn et al. 1984), although it is clear that all are of somitic mesodermal origin (Armand et al. 1983).

Gap junctions between primary myotubes were prominent from their time of first appearance on E16. They have been observed from the earliest times of myotube development in rat intercostal muscles (Kelly & Zacks, 1969a;Dennis et al. 1981), rat hindlimb muscles (Rash & Staehelin, 1974; Schmalbruch, 1982) and in amphibian myotomes (Blackshaw & Warner, 1976), as well as in myoblast and muscle cell tissue cultures (Rash & Fambrough, 1973). Coupling is so extensive that excitation spreading laterally between myotubes gives rise to waves of excitation that propagate across the entire muscle (Dennis et al. 1981). Thus, all the coupled myotubes are subject to the same pattern of activity. Gap junctions may also be involved in the transfer of regulatory molecules between cells (Gilula, Reeves & Steinbach, 1972).

Secondary myotubes in the lumbrical muscle were all joined to primary myotubes via gap junctions at E21, and so can be assumed to be electrically coupled. Also they were mechanically coupled to the primary myotubes, via interdigitating insertions which pushed deeply into the primary myotube and by desmosome-like junctions. The structure of these desmosome-like attachment zones between primary and secondary myotubes closely resembled that of the fascia adherens seen in the intercalated discs of cardiac muscle.

In amphibia, birds and mammals, axons enter the premuscle mass before myotube formation begins (Filogamo & Gabella, 1967; Teräväinen, 1968; Bennett & Pettigrew, 1974; Kelly & Zacks, 1969b; Dennis et al. 1981; Cameron & McCredie, 1982). These observations are confirmed here. The failure of Ontell & Kozeka (1984a) to see axons in the EDL premuscle mass might reflect difficulties in fixation.

Nerve-myotube contacts were observed from the outset of appearance of myotubes. Dennis et al. (1981) observed that nerve stimulation evoked muscle contraction from the earliest time of appearance of primary myotubes and so we assume that the axon terminals seen on E16 in rat lumbrical were functional; at this time, they were growth cones containing occasional vesicles and the myotube membrane was unspecialized at the point of apposition. A study of the initiation of nerve contacts on secondary myotubes is presented separately (Duxson, Ross & Harris, 1986). Maturation of nerve-muscle junctions in lumbrical muscles was similar to that in rat inter-costal muscle (Teräväinen, 1968; Kelly & Zacks, 1969b), apart from the difference in timing, and needs no further description here.

Nerve-undifferentiated cell contacts were abundant in E16 lumbricals; these may have been maintained during fusion of myoblasts to form myotubes. Close contacts between nerve terminals and nonfilamented cells also were present during the period of secondary myotube formation, E19-22 (Fig. 2B). Serial section analysis confirmed that these cells were mono-nucleate and contained no myofilaments (M. Duxson, unpublished observations). Contacts between axons and undifferentiated cells have been seen in other tissues and species including chick myotomes (Sisto Daneo & Filogamo, 1973); chick ALD and PLD muscles (Oppenheim & Chu-Wang, 1983) and rat intercostal muscle (Kelly & Zacks, 1969b). The destinies of these undifferentiated cells (muscle or connective tissue) and the identity of the axons (motor or sensory) are generally unknown.

Motoneurone death is described in the chick embryo by Hamburger (1948), and has recently been quantified both in mice (Lance-Jones, 1982) and rats (Harris & McCaig, 1984). If, as suggested in the following paper (Ross et al. 1987), innervation is involved in regulating the number of myotubes generated in a muscle, it is important to understand how moto-neurone death is regulated so as to determine how many axons will supply the muscle during the period of generation of secondary myotubes.

Death of lumbrical motoneurones was monitored by assaying total ChAT activity in the hindfoot over time. Cairns, McCaig & Harris (1986) found a linear relation between the number of axons in the phrenic nerve and E21 rat diaphragm muscle ChAT activity, and we assume the same relation between muscle nerve axon number and muscle ChAT activity to hold during development of the hindfoot muscles. An inflexion in the ChAT activity curve indicates loss of a significant proportion of the axons innervating the hindfoot muscles over E17-E19. Motoneurone death in the lumbar spinal cord was assayed by counting L4 ventral root axons. The L3, L4 and L5 adult ventral roots have similar numbers of axons (Coggeshall, Emery, Haruhide & Maryward (1977) thus minimizing any risks of inaccuracy due to misidentification. Most motoneurone death in the lumbar spinal cord occurred during E16-E17 and was substantially complete by E19.

The lack of synchrony between removal of nerve terminals in muscles of the foot and the general timecourse of motoneurone death in the lumbar cord is evidence that death in an individual motor pool is regulated by feedback from the target muscle. The lumbrical is a late-developing muscle, and developmental death of lumbrical motoneurones is correspondingly later than most other lumbar motoneurones. At the time of motoneurone death, the lumbrical muscles had developed all their primary myotubes, but secondary myotube formation had not yet begun. All motoneurones that subsequently would innervate secondary myotubes must have survived the cell death period by contacting primary myotubes. If survival of motoneurones is determined by competition (Easter, Purves, Rakic & Spitzer, 1985) within a ‘target’, this target consists solely of primary myotubes, representing a small fraction of the number of adult muscle fibres (10 % in the case of lumbricals) or of adult synaptic sites. The role of motoneurone death, we suggest, is not to adapt the number of motoneurones to the number of synaptic sites, but to regulate the number of motoneurones in order to determine the ratio of secondary to primary myotubes in individual muscles.

This work is part of the PhD dissertation submitted to the University of Otago by Jenny Ross, who was the recipient of a postgraduate fellowship from the Southland Frozen Meat Company. The work was supported by the New Zealand Medical Research Council.