Fetal development of deep back muscles in the human thoracic region with a focus on transversospinalis muscles and the medial branch of the spinal nerve posterior ramus

Abstract

Fetal development of human deep back muscles has not yet been fully described, possibly because of the difficulty in identifying muscle bundle directions in horizontal sections. Here, we prepared near-frontal sections along the thoracic back skin (eight fetuses) as well as horizontal sections (six fetuses) from 14 mid-term fetuses at 9–15 weeks of gestation. In the deep side of the trapezius and rhomboideus muscles, the CD34-positive thoracolumbar fascia was evident even at 9 weeks. Desmin-reactivity was strong and homogeneous in the superficial muscle fibers in contrast to the spotty expression in the deep fibers. Thus, in back muscles, formation of the myotendinous junction may start from the superficial muscles and advance to the deep muscles. The fact that developing intramuscular tendons were desmin-negative suggested little possibility of a secondary change from the muscle fibers to tendons. We found no prospective spinalis muscle or its tendinous connections with other muscles. Instead, abundant CD68-positive macrophages along the spinous process at 15 weeks suggested a change in muscle attachment, an event that may result in a later formation of the spinalis muscle. S100-positive intramuscular nerves exhibited downward courses from the multifidus longus muscle in the original segment to the rotatores brevis muscles in the inferiorly adjacent level. The medial cutaneous nerve had already reached the thoracolumbar fascia at 9 weeks, but by 15 weeks the nerve could not penetrate the trapezius muscle. Finally, we propose a folded myotomal model of the primitive transversospinalis muscle that seems to explain a fact that the roofing tile-like configuration of nerve twigs in the semispinalis muscle is reversed in the multifidus and rotatores muscles.

Introduction

The transversospinalis muscles are medially located deep back muscles that extend between the transverse processes and the vertebral arches or spinous processes of vertebrae. They are usually classified as spinalis, semispinalis, multifidus and rotatores muscles, and are innervated by the medial branch of the spinal nerve posterior ramus (Newell, 2005). In muscles other than the spinalis, the more superficial muscle bundles span a greater number of vertebrae, whereas the deeper bundles extend between fewer vertebrae (semispinalis muscle: more than four vertebral arches; multifidus muscle, two to four vertebrae; rotatores muscles, one vertebra). Thus, simply put, longer muscle bundles override or cover the shorter bundles to create a roofing tile-like arrangement. In the thoracic semispinalis muscle, a nerve twig-supplying area of the posterior ramus is located superficial to another area innervated by twigs from the lower ramus. Thus, the intramuscular nerve distribution pattern is also similar to roofing tiles. Strangely, however, this configuration is reversed in the thoracic rotatores muscles and, possibly, in the multifidus muscle; that is, the upper nerve crosses into the deep side of the lower nerve (Fig. 1A; see below).

Fig. 1.

Schematic depiction of the configuration of the human thoracic transversospinalis muscles and their supplying nerves (modified from Sato, 1973). Panel (A) (posterior view) displays the roofing tile-like configuration of the thoracic rotatores muscles and their intramuscular nerve distribution (many short twigs are omitted). A supplying area of nerve twigs from the lower posterior ramus of the spinal nerve lies on other muscle bundles innervated by twigs from the upper posterior ramus. Communication may exist between segmental nerves (arrow in the center of panel A). Panel (B) shows the topographical relation between muscle bundle and nerve in the multifidus and rotatores muscles (M, R). A muscle bundle of the multifidus longus muscle appeared to be innervated by two segmental nerves. The topographical relation varies depending on the supero–inferior level. Panel (C) is a cross-section showing the nerve course in the transversospinalis muscles. The nerve medial branch passes between the multifidus and semispinalis muscles to reach the spinous process of the vertebrae. ICN, intercostal nerve; MC, medial cutaneous nerve of the back; SS, semispinalis muscle.

Almost 40 years ago, the first author described the intramuscular nerve distribution in the human transversospinalis muscles (Sato, 1973); a modification of this description is shown in Fig. 1. In short, this anatomical architecture is characterized by (i) multisegmental innervation (Fig. 1B); (ii) site-dependent changes in the topographical relationship between a muscle bundle and segmental nerves (i.e. one nerve crosses the muscle superficially, whereas the other crosses deeply; Fig. 1B); and (iii) a border between the semispinalis and multifidus muscles demarcated by the course of the medial branch of the posterior ramus (Fig. 1C). Later, according to whole-mount acetylcholinesterase histochemistry of mouse embryos, Nakao & Ishizawa (1994) considered the multisegmental innervation to be a result of the ‘desegmentation’ (fusion and interdigitation) of the myotomal arrangement. although their major interest was the nerve configuration. However, Cornwall et al. (2010, 2011) recently demonstrated that in the mouse and human, the transversospinalis muscle does not have ‘in-series’ architecture even for muscle bundles that extend over two or three segments. Their observations indicate that, in contrast to Fig. 1B, a single muscle bundle of the transversospinalis muscle is unlikely to be supplied by multisegmental nerves.

Our major interest was to determine how the roofing tile-like configuration of nerve twigs is formed in the transversospinalis muscles in human fetuses. According to Deries et al. (2010) and Paul et al. (2004), if a muscle bundle of the human transversospinalis muscle is innervated by a single segment, the muscle fibers, intramuscular tendons, downward nerve courses, and a series from the superficial long to the deep short muscle fibers would together develop as a ‘set’. This hypothesis may exclude the possibility of another simple way, a secondary fusion between segmentally differentiated muscles followed by sorting to a regular laminar arrangement. It seems to be especially difficult to explain the fact that the nerve configuration seen in the semispinalis muscle is most likely to be reversed in the multifidus and rotatores muscles (see above). Consequently, we examined fetal development of the transversospinalis muscles using mid-term fetuses with the aid of immunohistochemistry. To identify each muscle fiber direction, we prepared near-frontal sections along the back skin in addition to the traditional horizontal sections. In the present study, we use the term ‘branch’ specifically for the medial and lateral branches of the posterior ramus that are likely to become cutaneous branches, whereas the muscle nerve is called ‘twig’.

Materials and methods

This study was performed according to the provisions of the Declaration of Helsinki 1995 (as revised in Edinburgh 2000). The use of fetuses for research was approved by the Chonbuk National University (Korea) Ethics Committee. Fetuses obtained by induced abortions were donated to the Department of Anatomy, Chonbuk National University with the informed consent of the mother and the agreement of the family. The mother consulted orally with an obstetrician, and then they gave written consent. There was no encouragement to make the donation. The privacy of patients was assured by removing all patient-identifying information and assigning random numbers to specimens. Fourteen mid-term fetuses, comprising five fetuses at 9 weeks of gestation (Crown-Rump Length or CRL 40–55 mm), five fetuses at 12 weeks (CRL 90–95 mm), and four fetuses at 15 weeks (CRL 100–135 mm) were examined. All specimens were stored in 10% v/v formalin solution for more than 3 months prior to preparing them for paraffin-embedded histology.

After removal of the thoracoabdominal viscera, the extremities and the head with the upper neck, specimens from thorax and lumbar regions were decalcified by incubating at 4 °C in a 0.5 m EDTA solution (pH 7.5; Decalcifying solution B, Wako, Tokyo) for 1–3 days, depending on the size of the material. After preparing for paraffin-embedded histology using routine procedures, near-frontal sections at 20-micron intervals along the back skin were prepared from eight fetuses (three at 9 weeks, three at 12 weeks, and two at 15 weeks). A separate group of six fetuses (two per age group) was used to prepare horizontal sections at 20-micron (9 weeks) or 50-micron (12 and 15 weeks) intervals. All sections were 5-μm thick. For each fetus, 50–200 sections were obtained, depending on fetus size. Most sections were stained with hematoxylin and eosin (HE), but some sections (four sections for every 10 sections) underwent immunohistochemical staining for (i) peripheral nerves (S100 protein; Niikura et al. 2010; Abe et al. 2011), (ii) striated muscle fibers (desmin; Abe et al. 2010), (iii) vessels and stromal mesenchyme (CD34; Katori et al. 2011a), and (iv) macrophages (CD68; Kim et al. 2011). All studies cited for immunostaining applications were recently performed by our group using human fetuses. In particular, S100 is useful as a marker of peripheral nerves including nerve terminals in human fetuses after deep fixation and long storage, although it is usually used as a Schwann cell marker.

The following primary antibodies were used: mouse monoclonal anti-human S100 protein (dilution 1 : 100; Dako Cytomation, Kyoto, Japan); mouse monoclonal anti-human desmin (dilution 1 : 50; Dako, Glostrup, Denmark); mouse monoclonal anti-human CD34 (dilution 1 : 100; Dako); and mouse monoclonal anti-human CD68 (dilution 1 : 100; Dako). Pretreatment by autoclaving was not conducted because of the loose nature of the fetal tissues. For the same reason, detergents such as Triton-X were not used. Nonspecific binding was blocked by incubating specimens in a skim milk solution. Primary antibodies were detected with horseradish peroxidase (HRP)-conjugated secondary antibodies (Dako Chem Mate Envison Kit; Dako), and antigen–antibody reactions were visualized via HRP-catalyzed reaction with diaminobenzidine. Samples were counterstained with hematoxylin. Negative controls (i.e. without primary antibody) were performed for all immunohistochemical analyses, but only one photo of a negative control is shown in the present figures.

Results

In contrast to the morphology of cross-sections (Fig. 2), the unique frontal sections clearly demonstrated muscle bundle directions (Figs 3–8). However, because of a slight curvature of the vertebral column, the sectional plane was somewhat oblique to the vertebral arch: usually superficial in the superior site and deep in the inferior site. Thus, in a section, the semispinalis muscle appeared in upper thoracic segments, whereas the rotatores muscles appeared in the lower segment (Fig. 3). Likewise, the depth was different between left and right sides of the back depending on the specimen. Therefore, within a section, we were able to observe a continuous change in morphology of the transversospinalis muscles from the long semispinalis muscle to the short rotatores muscles. The semispinalis and multifidus longus muscles were seen in sections superficial to the transverse process, whereas multifidus brevis and rotatores muscles were identified in sections that included the transverse process and ribs. The semispinalis muscle contained abundant intramuscular tendinous or aponeurotic tissues, most of which connected to the transverse process (Fig. 3A); the amount of such tissue was, notably, much greater than that in iliocostalis and longissimus muscles (Fig. 3B). The serratus posterior muscle was identified as a sheet on the lateral side of the iliocostalis and longissimus muscles extending from the cervical region (Fig. 3).

Fig. 2.

Cross-sectional morphology of the upper thoracic body wall and deep back muscles. HE staining. In both panel (A) (15 weeks; CRL 135 mm) and panel (B) (9 weeks; 40 mm), the scapula is shown at the right-hand margin. The iliocostalis and longissimus muscles (IL) and the transversospinalis muscles (SS, M, R) appear to be independent masses in panel (A), but the demarcation is unclear in panel (B). Because of difference in magnifications (see scale bars in panels A and B), the mediolateral length of the deep back muscle area at 15 weeks is almost four times as large as that at 9 weeks. Asterisk in panel (A) indicates an artifact during the histological procedure. After a downward course, the medial cutaneous nerve branch (MC) of the first thoracic nerve reaches the spinous process (panel C; 0.5 mm from panel A). In panel (D) (3.1 mm from panel A), it penetrates the trapezius muscle at the level of the second thoracic nerve root. Star in panel (C) indicates a small muscle mass medial to the cutaneous nerve course. DRG, dorsal root ganglion; ICM, intercostal muscles; ICN, intercostal nerve; LC, levator costae muscle; RH, rhomboideus muscle; SP, serratus posterior muscle.

Fig. 3.

Frontal sections of the thoracic deep back muscles at 9 weeks (CRL 55 mm). HE staining. Panel (A) shows a level 0.5 mm superficial to panel (B). Because of a slight curvature of the vertebral column, the superficially located muscles, such as the rhomboideus muscle (RH), are seen in the upper part of the figure, in contrast to the deeply located rotatores muscles, which are evident in the lower part of the figure. Note a continuous change in the morphology of the transversospinalis muscles from the long semispinalis muscle (SS) to the short rotatores brevis muscles (Rb). Many tendons are developing in the transversospinalis muscles (arrows in panel A) in contrast to only a few tendons (arrowheads in panel B) in the iliocostalis and longissimus muscles (IL). RH, rhomboideus muscle; SP, serratus posterior muscle; TP, transverse process of the vertebra.

Fig. 8.

Depth-dependent differences in intramuscular nerve distributions in transversospinalis muscles. S100 immunohistochemistry. Panels (A), (B) and (C) are higher magnification views of a site lateral to the Th-5 vertebral arch and correspond to squares in Figs 6A and 7A,B, respectively. All these areas appear to be innervated by the Th-4 posterior ramus medial branch. Some nerve twigs (arrowheads in panels A and B) run along the muscle fibers in the multifidus muscle (ML, Mb) and rotatores longus muscles (RL). In the rotatores brevis muscles (Rb), a few nerves are seen (short arrows in panels B and C). In panel (A), two nerves, indicated by long arrows, appear to go deeply into muscles shown in panels (B) and (C). The developing vertebra (cartilage) is positive for S100 immunostaining (asterisk in panel C). Mb, multifidus brevis muscle.

Before detailed descriptions on frontal sections, we describe cross-section morphologies for comparison. The mediolateral distance between the spinous process of the thoracic vertebrae and the scapula increased almost fourfold from 9 to 15 weeks, but the increase in the thickness of the entire dorsal back muscle layer was smaller (Fig. 2). The iliocostalis, longissimus, and levator costae muscles were easily identified in horizontal sections in all four specimens at 15 (Fig. 2A) and 12 weeks. In agreement with Sato's concept, shown in Fig. 1C, the semispinalis and multifidus muscles following a course of the medial branch of the posterior ramus were also discriminated in the same horizontal section. However, at 9 weeks, each of the deep back muscles was difficult to identify in horizontal sections because of irregular fusions between muscle masses at any given thoracic level (Fig. 2B). At 12 and 15 weeks, the medial cutaneous branch of the posterior ramus showed a downward course along nearly half the length of the vertebral arch (Fig. 2C). Moreover, the medial cutaneous nerve penetrated the trapezius muscle at the level of the inferiorly adjacent spinal nerve root (Fig. 2D). However, at 9 weeks, the medial cutaneous branch did not penetrate the trapezius, but reached the prospective thoracolumbar fascia. A thin muscle layer was seen between the medial branch and spinous process in two specimens at 15 weeks (Fig. 2C).

Using frontal sections, first, we focus on identifying which muscle was located beneath the thoracic back skin. In all eight specimens for this section, the trapezius and rhomboideus muscles were readily detected in the superficial layer. Moreover, a prospective semispinalis muscle with oblique muscle fibers was found along the spinous process at all ages examined (Fig. 4A,D,E). In addition, in one of two specimens at 15 weeks and one of three specimens at 12 weeks, we found a few muscle fibers extending between the adjacent spinous processes (i.e. interspinalis muscle) at the lower thoracic level (not shown). In close proximity to the deep side of the trapezius and rhomboideus muscles, CD34-positive mesenchymal tissue was evident and corresponded to the developing thoracolumbar fascia (Fig. 4B). This fascia covered the semispinalis muscle running obliquely. At 9 weeks, the medial cutaneous branch of the posterior ramus reached the thoracolumbar fascia near or superior to the spinous process of the same numbered vertebra (e.g. a branch from the third thoracic nerve between the second and third thoracic spinous processes; Fig. 4C). However, this topographical relation changed depending on age, and the nerve appeared at a more inferior site at 12 and 15 weeks. In two specimens at 12 weeks and two specimens at 15 weeks, CD68-positive macrophages were concentrated along the thoracic spinous process and in the medial attachment of the semispinalis muscle (Fig. 4F). Such a concentration was also seen in the serratus anterior muscle in the same specimens (not shown). A prospective spinalis muscle was not found along the spinous process in any specimen examined.

Fig. 4.

Superficial muscles of the back and the thoracolumbar fascia. Frontal sections. Panels (A) (HE staining), (B) (CD34 immunohistochemistry), and (C) (S100 immunohistochemistry) display near-frontal sections of a 9-week fetus (CRL 55 mm; the same specimen as that shown in Fig. 3). Because of artifactual folding of the skin, the trapezius and rhomboideus (RH) muscles take on a wavy configuration (panel A). CD34-positive mesenchymal tissue (arrows in panel B) is evident along the developing thoracolumbar fascia. The medial cutaneous nerve (MC) of the posterior ramus reaches the thoracolumbar fascia near to or on the superior side of the spinous process of the same numbered vertebra (e.g. MC-3 from the third thoracic nerve is seen between Th-2 and 3). Panels (D) and (E) (HE staining) and panel (F) (CD68 immunohistochemistry) show a 15-week fetus (CRL 125 mm). Panel (D) is located 0.3 mm superficial to panel (E). The semispinalis muscle is located in close proximity to the deep side of the trapezius muscle. Panel (F) displays macrophages concentrated along the semispinalis muscle (SS) attachments to the spinous process. In these specimens, the spinalis muscle is not found. Panels (A–E) are prepared at the same magnification (scale bar in panel D).

Secondly, we describe results of desmin immunoreactivity in frontal sections from seven of eight specimens. The positive expression extended along the entire length of muscle fibers of the iliocostalis and longissimus muscles and superficial parts of the semispinalis muscle. However, it was restricted to muscle fiber ends (origin or insertion sites) in most parts of the transversospinalis muscles. Thus, spotty expression characterized the multifidus and rotatores muscles (Fig. 5). However, the desmin-positive spots did not show any segmental configuration suggesting intercalation or fusion between segmental muscle fibers. All tendons of the semispinalis and rotatores muscles were desmin-negative (Fig. 5, inset, and 5C). At three or four sites in the specimens, we found spotty desmin expression in the multifidus longus muscle bundle running along the vertebral arch (Fig. 5B); the muscle bundle appeared to be pushed onto the developing vertebra. In addition, tendons and aponeuroses of the lateral deep muscles started developing at 9 weeks.

Fig. 5.

Desmin immunohistochemistry showing developing muscle attachments to the tendons and vertebrae. Frontal section (same specimen as that shown in Fig. 3, a section slightly superficial to that in Fig. 3B). The inset at the top of panel (A), as well as panels (B) and (C), are higher magnification views of squares in panel (A); for panel (C), the square is rotated at almost a right angle. Another inset at the bottom of panel (A), which corresponds to a muscle in the left-hand side of the Th-7 vertebra, displays a negative control without primary antibody. Desmin expression is seen along entire fibers of the iliocostalis and longissimus muscles (IL), but is restricted to the muscle fiber ends of transversospinalis muscles (SS, ML, Rb). Desmin expression is absent in the tendon itself (stars in the inset of panel A, and panels B and C). SS, semispinalis thoracis muscle; ML, multifidus longus muscle; Rb, rotatores brevis muscles; IL, iliocostalis and longissimus muscles; MB, medial branch of the posterior ramus; TP, transverse process of the vertebra.

Thirdly, we show intramuscular nerve distribution in the transversospinalis muscle from the superficial to deep aspect in Figs 6–8. S100-positive nerve terminals displayed a linear pattern in the semispinalis and multifidus longus muscles (Figs 6 and 7), whereas mostly spotty expression was observed in the multifidus brevis and rotatores muscles (Fig. 8). All these figures are for a 9-week fetus. But beginning at 15 weeks, the linear terminal in the semispinalis muscle developed into a thicker structure, recently termed the ‘endplate band’ by Mu & Sanders (2010) and Abe et al. (2011) (not shown). We attempted to trace the segmental nerve in semiserial sections based on S100 immunostaining, but we were not able to identify a single multifidus muscle bundle innervated by two segmental nerves. However, we confirmed the rule described by Sato that a nerve twig to the semispinalis muscle crossed on the superficial side of the corresponding lower nerve. Thus, a segmental supplying area overrode another area supplied by the lower segmental nerve.

Fig. 6.

Nerves and vessels in the transversospinalis muscles. Frontal sections (same specimen as that shown in Figs 3–5, a section slightly superficial to Fig. 3A). Panel (A) (S100 for nerves) and panel (B) (CD34 for vessels) display adjacent sections. Panel (A) exhibits five clusters of intramuscular nerves (denoted by an oval, a square, and three circles) in the multifidus muscle (ML). The oval area appears to be innervated by the Th-3 posterior ramus medial branch, the square by a Th-4 nerve, a circle lateral to Th-6 by a Th-5 nerve, and so on. Panel (B) shows diffusely distributed vessels in multifidus muscle.

Fig. 7.

Nerves in the multifidus and rotatores muscles. Frontal sections (same specimen as that shown in Figs 3–6). S100 immunohistochemistry. Panel (A) is located in the 0.4-mm-deep side of Fig. 6A, whereas panel (B) is 0.6 mm deeper than panel (A). Panel (A) exhibits several clusters of intramuscular nerves (circles and a square) in the rotatores longus muscles (RL). The square lateral to the Th-5 vertebral arch appears to be innervated by the Th-4 posterior ramus medial branch (see also Figs 6 and 8). However, in panel (B), nerves are considerably decreased in number in the rotatores brevis muscles (Rb; see Fig. 8 for details). Arrows in both panels indicate the lateral branch of the posterior ramus. Other abbreviations are the same as in Fig. 5.

Nerve terminals in multifidus and rotatores muscles from a segmental nerve were distributed in a limited area of a section, thus forming several territories or clusters. Because of the downward course of the nerves, a Th-X nerve cluster was located on the inferior side of the mother Th-X nerve (i.e. the medial branch of the posterior ramus). This site corresponded approximately to the lateral side of the inferiorly adjacent (Th-X + 1) vertebral arch (Figs 6 and 7). Thus, for instance, peripheral twigs to the multifidus brevis and rotatores muscles from the fourth thoracic nerve were seen on the lateral side of the fifth vertebral arch (Fig. 8). Therefore, as shown in Fig. 1A, the corresponding twigs from the fifth thoracic nerve in the lateral side of the fifth vertebral arch were located in the superficial side of or covering the fourth nerve-supplying area. In contrast to the patterned distribution of nerves, CD34-positve vessels and fasciae appeared to be randomly distributed in muscles, although they tended to be concentrated around the nerve and transverse process (Fig. 6B).

Discussion

The frontal sections employed here clearly demonstrated obliquely running muscle bundles of each part in the transversospinalis muscle group, even at 9 weeks, discriminating between muscle types that are difficult to distinguish in horizontal sections. Immunostaining for desmin revealed a difference in expression between the superficial and deep muscles: the trapezius muscle as well as the iliocostalis and longissimus muscles expressed the reactivity along the entire length of the muscle fiber bundles. In the transversospinalis muscle, the semispinalis muscle carried desmin immunoreactivity that was stronger than that of the multifidus and rotatores muscles, although in the present fetuses, the semispinalis muscle was located in close proximity to the deep side of the trapezius muscle. According to Abe et al. (2010), a spotty expression of desmin, such as typically seen in the transversospinalis muscle at sites of prospective attachments to the ligament and bone, suggests formation of the first anchoring or enthesis. Because of the restricted staining in muscle fibers, there seemed to be little possibility of nonspecific staining of the superficial muscles due to dehydration during long storage of the present specimens. However, desmin is one of the earliest muscle-specific proteins and the diffuse expression along a muscle fiber is likely linked to functional demands rather than maturation (Li et al. 1993). In back muscles, formation of the myotendinous junction may start from the superficial muscles and advance to the deep muscles.

Katori et al. (2011b) demonstrated that, in the intermediate tendon of the fetal human digastricus muscle, involved muscle fibers of the anterior belly maintain desmin expression for several weeks. Conversely, desmin-negative intramuscular tendons of the semispinalis and multifidus muscles suggested little or no possibility of secondary fusion between segmentally differentiated muscles. Thus, it is likely that a set of intramuscular tendons, downward nerve courses, and a series from the superficial long to the deep short muscle fibers develops in situ. Actually, Deries et al. (2010) demonstrated that, using confocal microscopy and 3D reconstruction, the myotomal muscle fibers largely retain their cranial attachments, but re-orient and extend caudally and laterally as they transform into the transversospinalis muscles in the rat and mouse. According to Deries et al. (2010), moreover, the tendon progenitors are tightly associated with the sides and ends of the myocytes as they re-orient and elongate. However, we observed several examples of a multifidus longus muscle bundle in which the middle part exhibited spotty desmin expression and was pushed to the expanding vertebral arch (Fig. 5B). Thus, we cannot exclude the possibility that, with secondary degeneration, the multifidus longus muscle could change into the multifidus brevis muscle.

Strangely, we found no prospective spinalis muscle or its tendinous connections with other muscles, despite paying special attention to this possibility. Instead, the interspinous muscle was seen in the lower thoracic segment. Deries et al. (2010) showed no possibility of apoptosis in the early development of the transversospinalis muscle in mouse and rat fetuses. However, in the present study, abundant CD68-positive macrophages were evident along the spinous process at the later stages examined, suggesting a secondary change in muscle attachment. The large number of macrophages was unusual compared to other muscles, and the distribution pattern was quite different from that along the developing bone. Thus, some inserted muscle fibers seem to have detached and degenerated, whereas others may be newly connected. If true, this event may result in formation of the spinalis from the interspinalis muscle. The semispinalis muscle fibers may also be involved in this re-arrangement along the spinous process. Shindo (1995) suggested a postnatal change of the muscle fiber configuration in the lumbar multifidus muscle, which is very thick in adults because of a difference in the topographical anatomy between adults and late-stage fetuses (5 and 10 months of gestation). In relation to the missing spinalis, we found a limited number of tendons and aponeurotic tissues in and on the iliocostalis and longissimus muscles – much fewer than in adults. Likewise, we did not clearly identify the medially located aponeurosis of the serratus posterior: the fetal muscle belly appeared to be located more laterally than in adults. Aponeuroses and tendons in and on the deep back muscle, as well as bony tubercles on the transverse process for attachment (Sato & Nakazawa, 1982), are likely to become evident after birth due to tensile stress from the muscles.

The present S100 immunohistochemistry demonstrated a band-like arrangement of motor endplates similar to that in the mouse lumbar transversospinalis muscle shown by Cornwall et al. (2010). S100-positive intramuscular nerve twigs exhibited downward courses, not only in the semispinalis muscle but also (because they are independent of nerve twigs to the semispinalis) through the multifidus and rotatores muscles. Terminal twigs for the latter short muscles reached the rotatores brevis muscles in the lateral side of the inferiorly adjacent vertebral arch. Nakao & Ishizawa (1994), using mouse fetuses, clearly demonstrated developing primary rami of the spinal nerve but, as we interpret their findings, failed to identify each of the back muscles. They proposed a diagram of the nerve division to each of the transversospinalis muscles: branching first to the short rotatores muscles and second or later to the long multifidus muscle. However, the present observations support a reversed order, as shown by Sato (1973): branching to the multifidus longus muscle first and finally to the rotatores brevis muscles. Thus, the nerve twigs to the rotatores are most likely to be the terminal twig of the posterior ramus medial branch. On the basis of our current observations and those of Sato (1973), we hypothesize a model of segmental differentiation of the transversospinalis muscle in stages immediately prior to 9 weeks (Fig. 9). According to this model, a folded, two-layered arrangement of myotomes suitably describes (i) passage of the posterior medial cutaneous branch along a space between two layers; (ii) the downward courses of nerve twigs to the semispinalis muscle as well as to other short muscles, and (iii) long-to-short (superficial-to-deep) continuous laminations of the transversospinalis muscles. This model also provides an answer to our initial question (see Introduction), that is, why the roofing tile-like arrangement of nerve twigs in the thoracic semispinalis muscle is reversed in the rotatores and multifidus muscles (Fig. 1A).

Fig. 9.

A model showing the hypothetical early stage development of the transversospinalis muscle. Lateroposterior view. A long surface faces to the vertebral arch and spinous process of the vertebrae. The deeply located extension of the ‘folded’ myotome-like mass would develop into a long-to-short (superficial-to-deep) laminar configuration of the multifidus and rotatores muscles (M & R), and the superficial short part into the semispinalis muscle (SS). The spinalis muscle is not included in this model because of the possible delayed development. The medial cutaneous branch of the spinal nerve posterior ramus passes through a narrow space between two layers those are folded and almost attached.

The medial cutaneous nerve of the posterior ramus had already reached the thoracolumbar fascia at 9 weeks at the level of the corresponding numbered spinous process, and at 15 weeks, the nerve penetrated the trapezius at the inferiorly adjacent level. Using six fetuses (dissection and histology), Pearson et al. (1966) described the cutaneous nerve course, but may have failed to take note of the topographical relation between muscle and nerve. Rather than considering that the lateral branch passes through the inferior border (i.e. postmyotomal branch), Nakao & Ishizawa (1994) suggested that a posterior ramus medial branch reaches the superiorly adjacent intermyotomal border (i.e. the premyotomal branch). Likewise, studies of adult cadavers indicate that the medial cutaneous nerve of the thoracic posterior ramus reaches the thoracolumbar aponeurosis alongside or near the spinous process of the superiorly adjacent vertebra, in contrast to its much lower exit to subcutaneous tissue (Aizawa & Kumaki, 1996). Notably, this inferiorly curved course in adults seems to follow fetal development during the 9–15-week period, as demonstrated in the present study. The developing trapezius muscle is evidently a strong obstacle for cutaneous nerve growth.

Mediolateral patterning of the somite is regulated by sonic hedgehog, and the deep back muscles originate from the medial half (Denetclaw et al. 1997; Olivera-Martinez et al. 2000; Venters & Ordahl, 2002). Sonic hedgehog also determines back muscle differentiation through activation of myogenic factor 5 (Myf5) (Borycki et al. 1999). ‘Resegmentation’, in which both the inferior half of one somite and the superior half of the next somite are postulated to develop into a single vertebra, has been a recent topic of somite studies (Aoyama & Asamoto, 2000; Huang et al. 2000). According to these studies, only the inferior half of a somite differentiates into a transverse process, in contrast to other parts from the two adjacent somites involved. In the present study, we found that transversospinalis muscles attached to the sixth transverse process were most likely supplied by the fourth thoracic nerve, owing to the downward migration of the segmental anlagen. Thus, the original segments appear to differ between a transverse process and the attached muscle. In a process similar to resegmentation of vertebral segments, two adjacent myotomal cell clusters may join to rearrange the supero–inferior topographical relation of segmental myoblasts. Actually, a figure in Huang et al. (2000) incidentally depicted only a limited part of the chick semispinalis originating from the grafted ‘one and half’ somites of donor quail. However, Deries et al. (2010) showed that, in rat embryos, a segmental block of myotomal muscle fibers largely retains its original attachment to the spinous process, but that the muscle fibers extend posteriorly and laterally beyond the segment of origin, as they mature into the transversospinalis muscles. Thus, the discrepancy in original levels between the transverse process and the innervating nerve is likely to occur without any mechanisms relating to re-segmentation.

Consequently, by way of re-arrangement of the segmental anlagen, the roofing tile-like configuration of the muscle fibers and nerve twigs is apparently established by 9 weeks in the human transversospinalis muscle. However, several delayed modifications, such as formation of the connecting tendon and aponeurosis as well as addition of the spinalis muscle, seem to occur later, depending on the functional demands for vertebral column movement. Our future research interests include the fetal iliocostalis and longissimus muscles as well as the possible ‘delicate position’ of the intermediate ramus of the spinal nerve (Sato, 1974).