The main arterial supply of the spinal cord is provided by posterior (dorsal), lateral, and anterior (ventral) vessels that span the entire cord length (Fig. 2.1c, see also [120]). The anterior spinal artery is located directly at the indentation of the anterior median fissure (Fig. 2.1c). In contrast, two posterior spinal arteries are located just ventral of the dorsal root entrance to the spinal cord, and the lateral spinal arteries can be observed somewhat halfway between the dorsal and ventral roots (Fig. 2.1c). Various transverse arteries extend from this arterial transverse anastomotic circle into the spinal cord and predominantly branch in the gray matter. The origin can be variable, but usually the anterior spinal artery is a derivate of two branches of the vertebral arteries at the level of the pyramidal decussation. The posterior spinal arteries either derive directly from the vertebral artery in the neck region or stem from the posterior inferior cerebellar artery. The posterior spinal arteries also receive blood from segmental arteries (ascending cervical artery, profound cervical artery, posterior intercostal artery, and lumbar arteries). The most prominent supporting artery is the artery of Adamkiewicz, which supplies the thoracolumbar and sacral spinal cord. In most cases it originates from the abdominal aorta approximately at T9, traverses the intervertebral foramen, and eventually anastomoses with the anterior spinal artery. Radicular arteries enter the spinal canal alongside the spinal nerves in the intervertebral foramina, then branch outside the dura mater, and follow the dorsal and ventral roots, respectively, to enter the spinal cord (Fig. 2.1c). Within the gray and white matter of the spinal cord, an intricate capillary network is formed and is significantly more elaborated in the gray matter. It was discovered in the mid-1940s that this imbalance of microvascularization mostly correlates with the number and density of synapses with high levels of mitochondria, of which there are many more in the gray than white matter [103]. Furthermore, some regions also seem to contain more capillaries than others, for example, the corticospinal tract contains about twice the amount of capillaries than the cuneate fasciculus [129].

Along the surface of the spinal cord, an elaborate venous plexus can be observed. Similar to the arterial layout, an anterior and posterior spinal vein spans the longitudinal axis of the spinal cord but can show significant caliber variations. Both the anterior and posterior veins anastomose via a network of smaller veins that circumvent the spinal cord. Especially at the cervical and lumbar enlargements, where the majority of spinal nerves innervating the extremities exit the cord, a larger venous transverse anastomotic circle can be found.

The first significant step during vertebrate nervous system development is the achievement of neural identity from dorsal ectodermal cells positioned along the midline of the gastrulating embryo. Subsequently, the neural plate will form and, along its two major axes (anterior-posterior and dorsal-ventral), it supports the generation of all neural cell types that will ultimately shape the mature CNS.

The CNS derives entirely from ectoderm. Its earliest derivatives include the neural plate and the neural crest. The neural plate is a somewhat oval-shaped region with elevated epithelial cells rostral to the primitive pit (Fig. 2.2a). Beginning with Carnegie stage 6 (day 18pc), the neural plate can be distinguished from the surrounding ectoderm in the anterior germinal disc and is wider there than at its caudal end. The neural crest and the embryo’s placodes develop in a band of cells directly adjacent to the neural plate. These cells are in direct contact with the surrounding epidermal ectoderm.

The neural plate is induced due to an inhibition of epidermis formation, which in turn is driven by signals that derive from the primitive node at the cranial end of the embryo (Fig. 2.2a), [101]. Within 24 h of the first appearance of the neural plate, folds appear at its respective edges, thereby creating the neural fold along the embryo’s midline (Fig. 2.2a, b).

Several factors contribute to the complex event of axial patterning and neurulation – the latter describing the event of embryonic folding not only along the midline but also in a dorsoventral orientation to ultimately close and form the neural tube. Cell intercalation into the embryonic midline results in a narrowing along the medial-lateral axis (convergence) and simultaneous rostrocaudal lengthening (extension) [50]. Specifically, the caudalizing event is driven by signaling molecules of the fibroblast growth factor (FGF) family, while the rostralizing event relies heavily on retinoic acid signals [56].

The other directional folding event along the dorsoventral axis is carried by bone morphogenic proteins (BMPs) for the dorsalizing orientation and sonic hedgehog (SHH) for the major ventralizing factor of this event [127]. Both factors significantly contribute to so-called hinge point (HP) formation [108]. Median and dorsolateral HPs are required for correct neural tube folding and closure, and lack of correct signaling results in the emergence of a number of neural tube closure defects along the embryo, which, depending on the position of the defect, can range from anencephaly to craniorachischisis, lumbosacral spina bifida, or spinal dysraphism and encephalocele [18, 37].

In humans, the next major time point in the development of the spinal cord is at around Carnegie stage 9 (20 days pc). Here, fusion of the opposing neural folds and subsequent formation of the neural tube occur at the hindbrain – spinal cord junction. This event requires several days to be completed and is coordinated simultaneously both at the rostral and caudal ends [82]. After closure of the neural folds, both ends temporarily remain open and are referred to as neuropores (Fig. 2.2b). When by 28 days pc both neuropores have been closed, the rostral neural tube initiates brain vesicle development, while the caudal pore is triggered to form the primitive spinal cord. This process of neural tube closure and extension of the spinal cord to S4/S5 is termed primary neurulation, with the secondary neurulation describing the event during which more caudal levels of the spinal cord are generated via connection and fusion of mesodermal cells. These cells then epithelize and subsequently merge with the rest of the tube [101].

Cells that emerge along the edge of the neural folds during the above-described process form what is termed the neural crest. They reside along the entire length of the neural tube, and while those that migrate along a dorsolateral pathway give rise to a number of neural cell populations later in life, those that transform into sensory ganglia are of distinct significance for spinal cord development: they are the origin of all dorsal root ganglia (DRG). Morphologically, the somata of these neurons remain in place, but the cells project a peripheral process to innervate somatic or visceral regions, while a central projection extends into the forming dorsal horn of the spinal cord. The extension of the peripheral projections heavily depends on appropriate trophic factor-mediated guidance and survival cues provided by the respective target sites. At the same time, a solid balance with repulsion events contributes to process pathfinding. The notochord, ventral spinal cord, and dermomyotome of the developing embryo have putative chemorepulsive effects on the extending peripheral DRG processes. Masuda and Shiga suggested members of the semaphorin family along with chondroitin sulfate proteoglycans be included among factors secreted by the notochord and dermomyotome, respectively, while the chemorepulsive carriers secreted by the ventral spine remain elusive [61].

A fundamental step during spinal cord development is the formation of the regions that will later form sensory and motor areas, respectively. A cross section through a 6-week-old human embryo is depicted in Fig. 2.3. Here, the roof and floor plates as well as a ventricular zone of the developing embryo are separated clearly. They are surrounded by neuroepithelial cells that produce a pseudostratified wall to the neural tube lumen. In this region, a number of cells with similar fates come to lie close to each other and ultimately perform cell-type-specific coupling, which depends on both chemical and bioelectrical signaling [1, 8], thus building the mantle layer between the ventricular zone and the marginal zone. The mantle layer, which resides around the primitive spinal cord, arises from multiple neuroepithelial cell divisions. These then lead to the subsequent emergence of primitive nerve cells and the accumulation of postmitotic neuroblasts beneath the external limiting membrane of the neural tube. The mantle layer will become the gray matter of the spinal cord. In turn, the marginal zone will ultimately become the white matter of the mature spinal cord.

Within the mantle layer, two distinct, paired regions can be observed: the dorsal thickening is referred to as the alar plate (future sensory areas of the spinal cord) and the ventral thickening is the basal plate (future motor areas of the spinal cord). Notably, the cells that reside outside of the roof and floor plates will ultimately serve as dorsal and ventral white commissures that facilitate the crossing of axons in the mature spinal cord.

In terms of cellular derivatives, the alar and basal plates and their induction to produce the cell types of the spinal cord have been investigated mostly in animal models. However, it is reasonable to assume similar mechanisms for human spinal cord development, as many of the signaling molecules and guidance factors along with their gene regulation are highly conserved among vertebrate species. While the roof plate is most likely initiated by signaling from the overlying ectoderm [16, 59], the floor plate is induced by axial mesoderm [2, 72]. Once established, the floor plate has been shown to generate a SHH gradient, which in turn induces the production of distinct pools of progenitors in the neuroepithelial ventricular zone ([13, 133], Fig. 2.3 left panel). Likewise, the roof plate and its external ectoderm, by generating a BMP gradient that is also reciprocal to the floor plate’s SHH gradient, produces six progenitor domains, which in turn generate six distinct early-born and two late-born dorsal interneuron cell populations [133] (Fig. 2.3 right panel). Like in the developing brain, the majority of neuroblasts use migration paths along radial glia cells to leave the matrix zone and reach the mantle zone (Fig. 2.3). However, not all neuroblasts follow this radial migration pattern. A subset of neuroblasts has been observed to migrate tangentially. For example, axons of some early-differentiating alar plate-derived neurons extend toward the ventral spinal cord and intersect in the floor plate. These arcuate fibers build a pathway, the so-called circumferential pathway [42], which in various species serves as both a dorsal-to-ventral [54] and ventral-to-dorsal migration path for neuroblasts [87]. Thus, it provides important commissural axon corridors that are essential for correct target innervation.

The earliest time point at which discernible ascending (spinocerebellar) neurons appear in humans is 10 weeks pc [17]. In a similar time window, by 13 weeks pc, the lateral corticospinal tract in humans has reached the caudal medulla oblongata. Just 2 weeks later, the pyramidal decussation is completed (reviewed in [112]). Next, the cervical spinal cord is reached between 14 and 16 weeks pc, and then a brief stop in further progression can be observed. The corticospinal tract invades the more caudal regions of the spinal cord at a later stage of development: 17 weeks pc for the low thoracic spinal cord, 27 weeks pc for the lumbosacral spinal cord. The early contact between descending axons of the corticospinal tract and corresponding synaptic targets in the spinal cord represents an important event for proper structural and functional maturation of the spinal cord. The maturation of spinal motor centers requires neuronal activity provided by the aforementioned contact [31]. This activity-dependent pattern for correct tract formation is especially important, considering that the developmental hallmark of initial exuberant axonal growth with subsequent substantial axon loss depends largely on neuronal activity and growth factors (whose expression can also be activity dependent). Hence, descending inputs are essential for the induction of plasticity and appropriate circuit formation [110]. In fact, recent data suggest that different motor systems can interact during development to drive each other’s adult specification [123]. The authors suggest that the developing rubrospinal system is under activity-dependent regulation by the corticospinal system for establishing mature rubrospinal connections and a corresponding red nucleus motor map.

Contrary to many mammals where myelination is part of the postnatal developmental period, first myelination of axons in the human can be observed in the early fetal spinal cord (e.g., less than 16 weeks pc shown by [78]). However, the majority of myelination cannot be seen until the second trimester. A distinct temporal order of myelinating events has been described previously. The earliest structure to myelinate is the medial longitudinal fasciculus (20 weeks pc), with the corticospinal tract being the last, still undergoing myelination at birth [121]. The same authors also described a spatiotemporal gradient of myelination, in which the three major protein classes myelin basic protein (MBP), proteolipid protein (PLP), and myelin-associated glycoprotein (MAG) appear along an anteroposterior and rostrocaudal gradient, with MBP emerging first. On a molecular level, oligodendrocyte progenitors – after migration and arrival at the correct location – undergo terminal differentiation into so-called premyelinating oligodendrocytes. This event follows a rostrocaudal gradient in the spinal cord [11]. Their terminal differentiation then includes the actual process of myelination, which requires physical contact to the axon that is to be myelinated. Again all these steps underlie intricate molecular regulation that depends on a number of developmental genes and factors (for a detailed review, see [70]).

Based on a fundamental anatomical nomenclature for the cat spinal cord gray matter first published by Rexed [95], a similar laminar cytoarchitecture of the gray matter also applies to humans [105]. An overview is provided in Table 2.1. The dorsal horn includes laminae I–VI, while laminae V–X reside in the base of the dorsal horn and the central region of the ventral horn (Fig. 2.4). However, not all laminae can be clearly distinguished from each other. Laminae I–VIII span the entire length of the spinal cord, although their size can vary significantly from one segment to the next. Of note, lamina IX cannot be understood as one section in the classical sense, but rather as motoneuron pools interspersed within laminae VII and VIII (Fig. 2.4).

Lamina I (zona marginalis, substantia spongiosa) has a reticular appearance and is composed of predominantly fusiform neurons (marginal cells). It borders with the tract of Lissauer, the region, where the dorsal horn contacts the pial surface of the spinal cord. It is visible as a band of unmyelinated or sparely myelinated afferent fibers also known as the posterolateral tract and contains axons projecting toward central regions, carrying nociceptive and thermoreceptive information. Neurons in lamina I are important for nociception and thermoreception and are the principal site for termination of nociceptive afferents. They use mainly glutamate, substance P, and calcitonin gene-related peptide (CGRP) as well as other peptides as transmitters. Lamina I neurons can also respond to mechanical stimulation from Aβ-fibers [124]. Interneurons in lamina I modulate nociception and signal predominantly via opioids, γ-aminobutyric acid (GABA), and glycine.

Like lamina I, lamina II (substantia gelatinosa) is a major nociceptive transmission and computation center and principal site for termination of Aδ- and C-fibers. It is composed of a great number of smaller, densely packed neurons and many unmyelinated axons. Neurons in lamina II form mostly local connections, with terminal axon collaterals reaching from lamina III as deep as lamina V [55]. Lamina II neurons function mostly as interneurons, and an estimated third of these contain GABA and glycine [114]. Other neurotransmitters in lamina II interneurons include CGRP, substance P, and vasoactive intestinal polypeptide (VIP).

Neurons in laminae III and IV (nucleus proprius) are larger and less densely packed than in laminae I and II due to numerous myelinated axons that traverse this lamina. Again, a large portion of neurons here are interneurons, with dendrites that extend mostly rostrocaudally. Transmitter types include GABA, glycine, dopamine, and substance P, among others. Functionally, some neurons seem to respond to fine tactile information, while others respond to rather strong, coarse pressure application. In the upper cervical segments, the medial region of lamina IV builds the internal basilar nucleus, a caudal extension of the cuneate nucleus [7].

Lamina V is the site of convergence between somatic and visceral input. It can be further divided into a medial part (ascending projection neurons) and a lateral part (preganglionic neurons in the lateral horn). The latter contains large somata, intercalated with thick myelinated axon tracts. In the human, this lamina cannot be clearly separated from lamina VI [104]. Dendrites in lamina V extend mostly in a dorsoventral orientation and can reach into laminae II and VII, respectively. Neurons in lamina V receive afferent input from the skin, musculature, and visceral organs with mixed modalities (mechanosensory, nociceptive, chemoreceptive). The major neurotransmitters present are GABA, dopamine, glycine, and substance P, among others.

Lamina VI is the deepest layer of the dorsal horn. As mentioned before, this lamina is almost indistinguishable from laminae V and VII, only visible in the cervical and lumbar enlargements. Similar to lamina V, it is also divided into a medial and lateral zone. Afferent input comes from collaterals of Aα-fibers (from muscle spindles). Lamina VI also receives input from cutaneous and nociceptive fibers. The majority of lamina VI neurons are either propriospinal neurons or interneurons. The latter are more local in their projection, many of them innervating ventral horn motoneurons. Hence, the major function of lamina VI interneurons seems to be the control of reflex pathways.

Lamina VII comprises most of the intermediate zone. The cell population in lamina VII is evenly sized and spaced, and the majority of cells are premotor interneurons projecting to the large motoneurons of lamina IX. In fact, lamina VII is a motor computation hub in that descending motor pathways control spinal motoneurons by synaptic transmission via interneurons in lamina VII. It is assumed that lamina VII interneurons are typically prewired to execute all aspects of a voluntary movement [120]. The modular arrangement of these interneuron-to-motoneuron connections thus provides a very potent way for the brain to execute directed voluntary movements. Indeed, the topographical arrangement of motoneuron columns matches the distribution of interneurons in lamina VII, with those modulating proximal musculature being placed more medial, and those driving distal movements being located more laterally.

Lamina VII contains three distinct nuclei: (1) the nucleus dorsalis (Stilling-Clarke, (C8) T1–L2/L3; see Sect. 2.4), with afferent innervation from proprioceptive fibers from muscles and joints; (2) the central cervical nucleus (C1–C4), with afferent input from the head and neck musculature, the vertebral joints, and the vestibular nuclei, which all coordinate head and neck movement; and (3) the intermediomedial nucleus (T1–L3), a part of the autonomic cell columns in lamina VII (cholinergic neurons). The latter are interconnected with sympathetic cells from the intermediolateral nucleus (lateral in C8–L3). These cells are preganglionic, visceroefferent neurons of the sympathetic and parasympathetic systems with input from supraspinal centers in the hypothalamus, pons, and medulla oblongata. Importantly, these nuclei also receive afferent visceral input. Both the intermediolateral and medial nuclei are connected with each other and with the contralateral side. Also located in lamina VII (lateral part) are the parasympathetic sacral nuclei (S2–S4), whose neurons innervate the large intestine, the urethral and anal sphincter, the reproductive organs, and the urinary bladder (see Sect. 2.4).

Neurons in lamina VIII are heterogeneous in size and shape at different levels along the spinal cord and their dendrites project mostly in the dorsoventral orientation. Dorsal dendrites, while not crossing the midline, are extended toward lamina VII and the ventral gray commissure. The majority of cells are GABAergic with input from propriospinal afferents and from supraspinal, vestibular, and reticulospinal afferents. Neurons of lamina VIII have a central role in the coordination of motor activity. For example, the left-right coordination of movement seems to be controlled by lamina VIII neurons [80]. Also, the majority of neurons with projections in the long propriospinal pathways that connect the lumbar and cervical enlargements are located in lamina VII, and they play a fundamental role for the coordination of front and hind limb movement [120]. Further details are discussed in Sect. 2.5.

Lamina IX is not a true lamina, but a set of columns in laminae VII and VIII and along the edge of X. It is the major site for large α-motoneurons (innervate striate musculature) and smaller γ-motoneurons (innervate muscle spindles) in the spinal cord. They are surrounded by a specific class of interneurons, so-called Renshaw cells, which provide recurrent inhibition for motoneurons and thus build an essential negative feedback mechanism for motoneuron activity. Motoneurons in lamina IX are organized in columns, which differ in size and number depending on the segmental level. Generally, four distinct columns are separated:

Lamina X (central gray of the spinal cord) extends along the entire spinal cord and encompasses the region around the central canal. Neurons in this lamina are smaller than those from the surrounding lamina VII and are more densely packed. Neurons of the central autonomic area are located here. Cells in lamina X receive input from C-fibers and Aδ-fibers of both somatic and visceral origin. Likewise, lamina X functions in both nociception and visceroception (including visceral pain and mechanoreception). Axons from lamina X project into laminae V, VI, and VII, among other targets in central regions of the brain.

As outlined in Sect. 2.1, the horns of the gray matter divide the surrounding white matter into three columns: a dorsal (posterior), a lateral, and a ventral (anterior). Yet boundaries between these parts are not necessarily distinct and clear. Within the white matter, ascending and descending pathways are distinguished from each other and will be discussed in the following sections (summarized in Table 2.2).

Overall, the ascending and descending tracts can be subdivided further into functional groups (Figs. 2.5 and 2.6). In the ascending orientation, we find (1) the posterior column-medial lemniscal system, which relays sensory information on vibration, proprioception, and fine touch (via the cuneate and gracile fasciculus), (2) the anterolateral system, which transmits nociceptive, thermoreceptive, and crude touch information (via the anterior and lateral spinothalamic tracts, spinoreticular tract, etc.), and (3) the cerebellar input system, which is responsible for proprioceptive sensibility of the upper and lower limbs (composed of the dorsal spinocerebellar tract, cuneocerebellar tract, and smaller tracts such as ventral and rostral spinocerebellar tracts).