Segmental Spinal Dysgenesis: Neuroradiologic Findings with Clinical and Embryologic Correlation

Multiple Vertebral Segmentation Disorders

In 1938, Jarcho and Levin (13) described two siblings with multiple segmentation defects of the vertebral column and fusion of several ribs. Several types of MVSD, characterized by short-trunk dwarfism and variably associated with costal abnormalities, have since been gathered under the collective term Jarcho-Levin syndrome; however, there is still considerable semantic and nosologic confusion in the literature. Mortier et al (14) categorized MVSD into three distinct types: Jarcho-Levin syndrome proper, spondylothoracic dysostosis, and spondylocostal dysostosis. Jarcho-Levin syndrome has autosomal recessive inheritance, and affected infants usually die before age 2 years from respiratory infections; it is radiologically characterized by a symmetric crablike thorax (15) and vertebral segmentation defects, usually involving the entire spine. Spondylothoracic dysostosis has autosomal recessive inheritance, and early infantile death from lethal respiratory infections is common. Radiologically, vertebral defects usually involve the entire spine, with predominance in the thoracic spine, and anomalous ribs may be present (16). Spondylocostal dysostosis is inherited in an autosomal dominant mode, and affected patients have a normal life span; radiologically, it is characterized by severe kyphoscoliosis, milder vertebral segmentation anomalies, and the invariable presence of abnormal ribs. Associated neural tube defects have been reported in several patients with MVSD (16, 17); however, little, if any, attention has been paid so far to the morphology of the spinal cord, probably because neurologic impairment is not a principal feature with these disorders.

We were able to identify several clinical and neuroradiologic features that appear to be typical of SSD and distinctly uncommon with MVSD: a) the disorder is sporadic; b) affected patients have a normal life span (2), and life-threatening infantile respiratory diseases are uncommon; c) neurologic impairment is the rule, although some newborns may be neurologically intact at presentation; d) the vertebral malformation is typically localized in a determined spinal segment, usually in the thoracolumbar or upper lumbar spine, although minor vertebral anomalies may involve other spinal levels; e) not surprisingly, abnormal ribs may be associated when the malformation involves the thoracic spine, because the ribs develop from ventrolateral migration of cells from the thoracic sclerotomes (18); and f) most important, the spinal cord is segmentally abnormal.

We conclude that SSD is a separate entity from MVSD, with the main differential feature being the typically segmental malformation of the spine and spinal cord.

Congenital Vertebral Displacement

CVD is defined as dislocation at a single vertebral level that results in an abrupt offset of the spinal canal; two patterns of deformity may be identified according to the presence of single-plane or multiple-plane offset (19). Neurologic deficits can be amazingly mild; when present, they result from marked spinal instability combined with compression of the spinal cord. Closed spinal dysraphisms are typically not associated. CVD is a single-level spinal abnormality, and the vertebral pedicles are consistently present in CVD and absent in SSD (Dias MS, personal communication). The spinal cord is stretched or compressed in CVD, but is otherwise normally developed (see Fig 2A in [19]), whereas it is segmentally abnormal in SSD. Therefore, we conclude that CVD is a disorder different from SSD because it is a single-level deformity, the vertebral abnormality does not involve the posterior neural arch, and the spinal cord is stretched or compressed, but not segmentally malformed.

Medial Spinal Aplasia

In 1992, Carstens et al (20) described six infants with an isolated absence of spinal segments, mainly involving the lower thoracic and lumbar regions. All patients had congenital kyphoscoliosis and complete paraplegia, usually at a level that corresponded to the bony aplasia. MR imaging was performed in one patient, and showed absent spinal cord at the level of the aplasia with intraspinal neural structures distally. These authors believed they were dealing with a new syndrome, and named it “medial spinal aplasia” (20). However, their description of clinical and radiologic findings coincides with severe SSD; we therefore suggest that this entity probably represents the most severe end of the SSD spectrum, and discourage the use of the term medial spinal aplasia.

Caudal Regression Syndrome

Faciszewski et al (2) raised the problem of the relationship between SSD and CRS, and made clear that they excluded from their definition of SSD all cases in which a segmental defect in the thoracolumbar or lumbar region was associated with caudal spinal anomalies. However, they provided no embryologic reasons for their doing so and, moreover, accepted that two separate segmental levels of dysgenesis could be found in the same patient. We see little, if any, difference between this concept and the association of SSD with caudal spinal anomalies, such as coccygeal or sacrococcygeal agenesis. We saw three cases of SSD associated with partial sacrococcygeal agenesis, whereas renal abnormalities, such as crossed ectopia and horseshoe kidneys, were present in four cases. The coexistence of some features of CRS in typical cases of SSD probably reflects a form of multisegmental early embryonic chorda-mesodermal derangement (see “Embryologic Correlation” below). We speculate that if such derangement involves an intermediate segment along the longitudinal embryonic axis, SSD results; if it involves the most caudal segment, CRS results. If two segments are involved, double levels of SSD, or the association between SSD and sacrococcygeal agenesis, may be explained. To cite but one reported case, the association between anencephaly and sirenomelia, described by Rodriguez et al in 1991 (21), is another confirmation of the theory that multiple midline, segmental chorda-mesodermal derangements may be found in association, and that the clinical and neuroradiologic picture essentially depends on what segments of the developing embryo are involved. We conclude that SSD and CRS probably represent two faces of a single spectrum of segmental malformations of the spine and spinal cord, only differing, from an embryologic point of view, in the segmental location of the derangement along the longitudinal axis of the embryo. However, although these entities are probably embryologically related, major differences exist both in terms of clinical expression and neuroradiologic features; moreover, one of the most compelling reasons to separate SSD from CRS or other forms of dysraphism is that SSD is unlikely to benefit significantly from untethering procedures, since the neurologic disturbance seems to be related to the congenital hypoplasia or absence of roots or an entire segment of the spinal cord, rather than to cord tethering (1).

Dysgenetic Segment

Programmed cell death, or apoptosis, is a process of cell elimination that occurs in the course of normal development and represents a crucial phenomenon during various steps of embryogenesis (24). Maden et al (25) demonstrated that quail embryos developing in the absence of vitamin A have a completely missing hindbrain as a result of ectopic apoptosis occurring at the four to six and six to eight somite stages, whereas these apoptotic events are prevented by retinol injection before gastrulation. These authors concluded that since these apoptotic events occur at the time of specification of axial identity and segmentation in the mesenchyme and neuroepithelium, wrongly specified cells in terms of their rostrocaudal position are eliminated (positional apoptosis) (25). We surmise that similar events occurring during human gastrulation may explain the segmental malformations typical of SSD. In fact, it is known that prospective chorda-mesodermal cells migrating through the primitive pit and into the ectoderm-endoderm interface have a genetically determined destination along the longitudinal embryonic axis; an error in this positional imprinting could activate apoptotic events so that wrongly specified cells are eliminated and, eventually, fewer cells or even no cells at all form the chorda-mesoderm at a given abnormal segmental level. The consequences of such segmental chorda-mesodermal paucity are manifold, and affect the development of the spinal column, spinal cord, and nerve roots as follows:

The induction and development of the somites will be altered. Normally, 42 to 44 pairs of somites are formed (18); each somite differentiates in a dermatome (future skin), a myotome (future muscles), and a sclerotome (future cartilage, bone, and ligament of the spinal column). Histologic studies on avian embryos have proved that the removal of a notochordal segment causes massive death of sclerotomal cells, whereas the myotome and dermatome are essentially unaffected (26). A wide array of segmental vertebral malformations, including segmentation defects, indeterminate or block vertebrae, or even absence of several vertebrae, results from segmental chorda-mesodermal derangement; abnormal ribs will also be found if the thoracic spine is involved, owing to their common embryologic origin (18). Moreover, because the dermatome survives regardless of the state of sclerotome differentiation (26), the consistent absence of open spinal dysraphism in cases of SSD may also be explained.

Because the chorda-mesoderm, probably together with the underlying endoderm (22), acts as a neural inducer, at the beginning of primary neurulation there already exists an established anomaly in which fewer prospective neuroectodermal cells, or even no cells at all, are induced to form the neural plate in the pathologic segment. In milder cases, neurulation of a depauperated neural plate results in a hypoplastic cord segment; conversely, no neurulation occurs in the most severe cases, so that no cord is formed. In a study on chick embryos, van Straaten et al (27) proved that the extirpation of a notochordal fragment at 1.5 days of incubation resulted in a segmental abnormality characterized by smaller than normal or even unrecognizable somites and severe reduction in spinal cord size; caudal to this site, a normal morphology of the spine and spinal cord gradually reappeared. In other words, these authors were able to obtain an experimental model of SSD in chick embryos as a result of segmental notochordal damage. Because of the metameric relationship between each vertebra and the corresponding spinal cord level, the hypoplastic or absent cord segment begins cephalad to the bony dysgenesis.

Nerve roots are missing in the abnormal segment, possibly as a result of apoptotic elimination of somatic motoneurons (28). This hypothesis was confirmed by prior studies in vitamin A–deficient quail embryos, in which the neural tube failed to extend neurites into the periphery and the neural crest cells throughout the embryo died by apoptosis (29).

Lower Spine and Spinal Cord

The localized, segmental cord abnormality does not prevent neurulation from occurring in the segments of the neural plate caudad to the malformation, resulting in the presence of a lower cord segment in the vast majority of cases. Van Allen et al (30) suggested that the neural tube, rather than closing in a zipperlike fashion, as is commonly believed, undergoes neurulation starting from five independent closure sites. If eventually proved true, this hypothesis might add substance to the embryogenetic understanding of SSD. In fact, these authors state that closure 1 of the neural tube forms the spinal cord down to L2, whereas closure 5 is initiated at S2 and proceeds rostrally to meet the caudal closure 1 at L2; therefore, the interface between closures 1 and 5 would be a sort of watershed area, explaining the comparatively higher prevalence of neural tube defects in this region than in other spinal segments (30). We surmise that the presence of multiple sites of neural tube closure might reflect regional differences in chorda-mesodermal induction of the neural plate. The high prevalence of SSD in the thoracolumbar and upper lumbar spine, namely, right at the interface between closures 1 and 5, could be explained by an intrinsic weakness, either genetically imprinted or teratogen-induced, of this area as compared with other notochordal segments. However, this speculation remains to be proved embryologically.

Other considerations regarding the presence of lower cord segments in most cases of SSD deserve mention. First, because this lower segment is invariably low-lying, postneurulation ischemic damage (4, 31) seems unlikely; rather, a malformative cause is strongly suggested. In fact, it should be stressed that spinal MR images in the term newborn show a 9-month-old spinal cord, whose metameric relationship with the spinal column is determined by the different growth rate between the cord and the column (32); the consistently low position of the lower cord segment in SSD probably reflects the original location of the caudal neuroectoderm as a result of impaired, or even absent, relative ascent of the spinal cord through the dysgenesis. Second, the lower cord segment is invariably bulky and thickened. As a possible explanation, it should first be considered that the elongation of the notochord occurs primarily by the addition of cells to its caudal end rather than by cranial extension of the notochordal process (22, 23). If, during gastrulation, a derangement occurs at a given segmental level, subsequent migrating waves of prospective chorda-mesodermal cells will be dammed below the malformed segment. However, as these cells are not wrongly specified in terms of their eventual location, they will not die by positional apoptosis; rather, they will induce an increased amount of cells to form the lower lumbar neuroectoderm, so that a thicker lower cord segment eventually results.

Finally, the neuroradiologic picture appears to depend not only on the degree of depletion of the chorda-mesoderm and, consequently, on the degree of prospective neuroectoderm in the dysgenetic segment but also on the level of the segmental damage along the longitudinal axis of the embryo. In fact, in order to have a lower cord segment it is necessary that a segment of normal neural plate exist below the segmental abnormality. This occurs in all cases of thoracolumbar or upper lumbar SSD (Figs 2–4⇑⇑), in which the future interface between closures 1 and 5 of the neural tube (30) is involved. However, the lowermost end of the neural plate lies at the level of S2 (32, 33); below that, caudal nervous structures (the tip of the conus medullaris and the filum terminale) are formed by the process of secondary neurulation and retrogressive differentiation. Therefore, it only seems logical to expect that if the malformation involves the upper sacrum, corresponding to the interface between closure 5 and the caudal cell mass, no lower cord segment can possibly exist caudally; rather, the hypoplastic cord segment will blend with the filum terminale (Fig 5). When the segmental level of the abnormality is low and associated caudal spinal abnormalities such as partial sacrococcygeal agenesis occur, derangements of secondary neurulation are expected to accompany SSD; consequently, apparently bizarre phenotypes, such as the presence of aberrant remnants of the conus tip continuous with the hypoplastic cord, may be explained.

To summarize, we believe there is sufficient evidence to surmise that the causal insult in SSD results from chorda-mesodermal derangement during gastrulation, possibly related to positional apoptosis of cells with wrong axial specification, whereas other theories dating the damage to subsequent developmental stages seem less tenable. The eventual malformative picture appears to correlate with two main factors: the severity of the abnormality and the segmental level at which it occurs.