The evidence against somatotopic organization of function in the primate corticospinal tract

Abstract

We review the spatial organization of corticospinal outputs from different cortical areas and how this reflects the varied functions mediated by the corticospinal tract. A long-standing question is whether the primate corticospinal tract shows somatotopical organization. Although this has been clearly demonstrated for corticofugal outputs passing through the internal capsule and cerebral peduncle, there is accumulating evidence against somatotopy in the pyramidal tract in the lower brainstem and in the spinal course of the corticospinal tract. Answering the question on somatotopy has important consequences for understanding the effects of incomplete spinal cord injury.

Our recent study in the macaque monkey, using high-resolution dextran tracers, demonstrated a great deal of intermingling of fibres originating from primary motor cortex arm/hand, shoulder and leg areas. We quantified the distribution of fibres belonging to these different projections and found no significant difference in their distribution across different subsectors of the pyramidal tract or lateral corticospinal tract, arguing against somatotopy. We further demonstrated intermingling with corticospinal outputs derived from premotor and supplementary motor arm areas.

We present new evidence against somatotopy for corticospinal projections from rostral and caudal cingulate motor areas and from somatosensory areas of the parietal cortex. In the pyramidal tract and lateral corticospinal tract, fibres from the cingulate motor areas overlap with each other. Fibres from the primary somatosensory cortex arm area completely overlap those from the leg area. There is also substantial overlap of both these outputs with those from posterior parietal sensorimotor areas.

We argue that the extensive intermingling of corticospinal outputs from so many different cortical regions must represent an organizational principle, closely related to its mediation of many different functions and its large range of fibre diameters.

The motor sequelae of incomplete spinal injury, such as central cord syndrome and ‘cruciate paralysis’, include much greater deficits in upper than in lower limb movement. Current teaching and text book explanations of these symptoms are still based on a supposed corticospinal somatotopy or ‘lamination’, with greater vulnerability of arm and hand versus leg fibres. We suggest that such explanations should now be finally abandoned. Instead, the clinical and neurobiological implications of the complex organization of the corticospinal tract need now to be taken into consideration. This leads us to consider the evidence for a greater relative influence of the corticospinal tract on upper versus lower limb movements, the former best characterized by skilled hand and digit movements.

Lemon and Morecraft use neuroanatomical evidence to argue against somatotopy in the brainstem and spinal course of the primate corticospinal tract and to suggest instead that intermingling of fibres from different cortical regions reflects the tract's multiple functions. The findings have implications for understanding the effects of incomplete spinal injury.

Introduction

Early neurologists such as William Gowers stressed the importance, for nervous system disorders and damage, of the long descending pathways that connected the cerebrum with the spinal cord. Particular emphasis was placed on the corticospinal tract (CST).¹ Subsequent neuroanatomical, neuropathological and neurophysiological investigations have firmly established the importance of the CST for normal motor control.² The function of the CST is often compromised in spinal cord injury (SCI); for example, a significant proportion of patients with incomplete cervical SCI are diagnosed as having central cord syndrome (CCS),^3,4 characterized by much greater impairment of upper limb (particularly hand function) than lower limb movements.

Background: early studies of ‘lamination’ within the corticospinal tract

Early studies on the organization of fibre pathways in the human spinal cord searched for signs of a somatotopic pattern or lamination. The first papers of which we are aware are by Flatau⁵ and Foerster⁶ which were based on neurosurgical interventions. Flatau proposed a ‘Law of Eccentricity’ which stated that the shorter propriospinal fibres remained closest to the spinal grey, pushing the longer fibres (e.g. those supplying the leg) to the periphery. In 1936 Foerster⁶ (who does not cite Flatau) published a figure claiming extensive lamination of all the long pathways in the cord, including the CST, with fibres directed towards the leg lying lateral and those to the arm lying medial (Fig. 1A). This concept was central to the explanation advanced by Schneider et al.⁷ in their study of a number of cases of CCS. They suggested that spinal injury had damaged the central grey matter of the cord and that the damage extended to medially located arm/hand fibres within the lateral corticospinal tract (LCST), while sparing leg fibres located more laterally within the tract, thereby explaining the disproportionate effects on upper versus lower limb function.

Figure 1.

Textbook versions of the concept of somatotopy or ‘lamination’ in the CST. (A) The original diagram from Foerster (1936)⁶ published in the Handbook of Neurology. Note the orientation is with dorsal to the bottom of the figure and ventral to the top. ‘Lamination’ in all the major pathways, including the LCST, is shown. Redrawn from Foerster.⁶ (B) The diagram as it first appeared in the 32nd edition of Gray’s Anatomy (1958). Redrawn. (C) The diagram as it appeared in the 41st edition of Gray’s Anatomy (2016)¹⁰; reprinted from Figure 20.14 with permission from Elsevier. Dorsal is on the top of the figure in B and C.

But in 1955 Nathan and Smith,¹ in their comprehensive review of descending pathways, pointed out that Foerster had not offered any evidence to substantiate his claim of lamination within the LCST. Nevertheless, the concept continued to be adopted as the likely explanation for the motor symptoms arising from CCS and other forms of incomplete SCI. It was suggested that ‘cruciate paralysis’ also resulted from selective damage of arm versus leg corticospinal fibres, claimed to cross at different levels of the pyramidal decussation (PD).⁸ Many reviews have repeated the idea that corticospinal somatotopy exists and could explain the much greater disturbance of upper versus lower limb function after incomplete SCI.^9–12

Challenging the concept of ‘lamination’

However, the concept was challenged early on by the discovery that, after stroke in restricted parts of motor cortex, degenerating corticospinal fibres were always found throughout the pyramidal tract (PT) and not confined to one sector.^1,13 A very early experimental study in the chimpanzee¹⁴ and later ones in macaque monkeys^15,16 all found that discrete lesions within the primary motor cortex (M1) resulted in a widespread distribution of degenerating axons throughout the medullary PT and LCST (see also Kuypers¹⁷), which is unexpected if there were a precise, somatotopic organization of descending fibres.

Later, Pappas et al.¹⁸ used the anterograde wheat germ agglutinin-horseradish peroxidase (WGA-HRP) technique to label projections from macaque and squirrel monkey motor cortex. Irrespective of the injection site (forelimb or hindlimb motor cortex), the labelled fibres were found throughout the medullary PT. Fibres from both cortical representations crossed at the same level of the cranio-vertebral junction (CVJ), discounting an anatomical explanation for cruciate paralysis. Thus, although investigators confirmed the somatotopical arrangement of fibres in the more cranial course of the corticofugal system as it passes through the subcortical white matter, internal capsule and the initial parts of the cerebral peduncle, this arrangement was clearly lacking in the brainstem and spinal cord.

The concept of lamination lives on

Levi et al.³ summarized the evidence and argued strongly against somatotopy, and this view has been repeated in subsequent reviews.^19–22 Nevertheless, the idea of lamination persisted, particularly in clinical teaching and textbooks of anatomy and neurosurgery.^9–12,20,23 A Foerster-based figure appeared for the first time in the 32nd edition of Gray’s Anatomy (1958) (Fig. 1B) and subsequently in all editions until the 41st, published in 2016 (Fig. 1C),¹⁰ although some editions did refer to the issue of somatotopic localization in the medullary pyramids and spinal cord as a ‘matter of controversy’, and in the most recent (42nd) edition, the Foerster-based figure has been withdrawn.

How can one explain the longevity of the lamination concept? First, we should consider the attractive simplicity of a somatotopical scheme and its apparent resemblance to other somatotopically organized pathways, such as the dorsal columns (see Barnard and Woolsey¹⁵). Second, it could be argued that the earlier evidence against the concept is only partial: we have no definitive anatomical data from human studies, and MRI tractography and intraoperative stimulation cannot yet provide the resolution required, and has instead been more concentrated in the somatotopy of the corticofugal fibres over their more cranial course in the subcortical white matter, internal capsule and cerebral peduncle.^24–26 Finally, the evidence from animal studies is from older degeneration and HRP labelling methods, which fail to label the many fine fibres in the CST.²⁷ A detailed examination of the localization of fibres as they pass through the PT, CVJ and LCST is missing. Further, nothing is so far known about the organization within the lower brainstem and LCST of the very significant contribution of corticospinal fibres from secondary motor areas lying outside the primary motor cortex. Together these areas make up around 50% of the frontal lobe output into the CST.²⁸

Resolving the question of somatotopy in the corticospinal tract

What was needed to re-evaluate the question of somatotopy in the primate CST was a new, systematic study of the descending course of corticospinal fibres originating from the arm/hand, shoulder and leg regions of M1 through the CVJ and cervical spinal cord (C1–T1) in the macaque monkey,²⁹ which is the best available model for the human motor system.^2,17 This approach used modern neuroanatomical anterograde tracers that clearly label individual fibres, and which are transported by even the finest of CST fibres.³⁰ It also considered the significant contribution to the corticospinal projection (CSP) that arises from secondary motor areas of the frontal lobe, including the ventral and dorsal premotor areas and the supplementary motor area. Finally, quantitative methods were used to confirm the distribution of labelled axons within the PT and LCST.

Results from the study of the corticospinal projection from the frontal lobe

Morecraft et al.²⁹ analysed labelling in a total of 11 adult rhesus macaques that had previously been used to assess the pattern of CST termination in the cervical grey matter.^31–33 In each of these monkeys, one or more localized injections were made of different types of high-resolution anterograde tracers. These included biotinylated dextran amine (BDA), lucifer yellow dextran (LYD), fluorescein dextran (FD), phaseolus vulgaris leucoagglutinin (PHA-L) or dextran tracer (DA488). Injections were made into three subdivisions of M1 (arm/hand, shoulder or leg) and into the arm/shoulder representations of the ventral or dorsal premotor cortex (LPMCv and LPMCd) and supplementary motor area (SMA/M2). In each case, the injection site was confined to the somatotopic region revealed by movements evoked by low threshold intracortical microstimulation (ICMS). After a survival period ranging from 25 to 33 days, monkeys were sacrificed and prepared for the cutting of frozen sections from cortex, brainstem and spinal cord. These sections were then processed immunohistochemically to localize the fibres labelled with the different tracers used.³⁴

Intermingling of M1 corticospinal outputs from different regions of M1

One of the main results of the study is shown in Fig. 2. The labelling shown in the upper panel came from a case in which the arm/hand M1 representation was injected with LYD. In this case, the labelled fibres were distributed throughout the ipsilateral PT (PT in Fig. 2). As the labelled fibres approached the rostral part of the PD, most fibres began to arch dorsally and toward the midline in a widely dispersed manner. In the spinal cord, labelled fibres were densely and widely distributed in the contralateral LCST throughout both its medial and lateral parts. These fibres terminated throughout the cervical grey matter, and particularly in the C5-T1 segments, with heavy labelling in lamina VII and also in lamina IX, where the motor nuclei supplying hand and finger muscles are located (for details, see Morecraft et al.³³).

Figure 2.

Distribution of corticospinal fibres from M1 arm/hand area and M1 leg area in the macaque monkey. Line drawings depicting selected transverse sections (tissue was cut perpendicular to the long axis of the spinal cord and brainstem) showing the distribution of labelled axons in the PT, CVJ and cervical spinal cord in Case SDM54 (top) which received an injection of LYD in the M1 arm/hand area, and Case SDM84 (bottom), which received an injection of DA488 in the M1 leg area. For each case the transverse sections are shown from rostral (top left, lower medullary pyramid) to caudal (bottom right, T1). CN = cuneate nucleus; DC = dorsal column; ECN = external cuneate nucleus; GN = gracile nucleus; ICP = inferior cerebellar peduncle; IO = inferior olive; LCST = lateral corticospinal tract; ML = medial lemniscus; NA = nucleus ambiguus; NS = nucleus solitarius; PD = pyramidal decussation; PT = pyramidal tract; RtN = reticular nucleus; SSN = supraspinal nucleus; VC = trigeminal complex; VCST = ventral corticospinal tract; X = dorsal motor nucleus of the vagus. Reproduced from Morecraft et al.^29©The authors, published with permission. CC BY-NC-ND 4.0.

Importantly, the lower panel in Fig. 2 shows that the fibres labelled by a DA488 injection in the M1 leg area of a different animal travelled through exactly the same territory of the PT, decussation and LCST, showing an almost complete overlap with arm/hand fibres, with no sign of any somatotopy. This was also true of fibres found in another two cases with injections in the M1 shoulder region (Fig. 3, top). Thus fibres from all three regions were heavily intermingled in the lower course of the CST.

Figure 3.

Distribution of corticospinal fibres from M1 shoulder area and from the SMA/M2 in the macaque monkey. Line drawings depicting selected transverse sections showing labelled axons in the PT, CVJ and cervical spinal cord in Case SDM90 (top) which received an injection of BDA in the M1 shoulder area, and Case SDM54 which received an injection of FD the SMA/M2 arm area. The transition region between the caudal-most part of the medullary PT and the VCST is indicated by the asterisk in SDM54 (top row, section before C2). For layout of sections, see Fig. 2. Reproduced from Morecraft^29©The authors, published with permission. CC BY-NC-ND 4.0. CN = cuneate nucleus; DC = dorsal column; ECN = external cuneate nucleus; GN = gracile nucleus; ICP = inferior cerebellar peduncle; IO = inferior olive; LCST = lateral corticospinal tract; ML = medial lemniscus; NA = nucleus ambiguus; NS = nucleus solitarius; PD = pyramidal decussation; PT = pyramidal tract; RtN = reticular nucleus; SSN = supraspinal nucleus; VC = trigeminal complex; VCST = ventral corticospinal tract; X = dorsal motor nucleus of the vagus.

Within the PD, heavily labelled fibres from arm/hand, shoulder and leg areas continued throughout the rostral, middle and caudal parts of the PD with no obvious somatotopy.²⁹ Thus, axons from the M1 arm/hand area do not selectively lie medially at rostral levels of the PD, nor do M1 leg axons occupy an exclusively lateral position at rostral levels of the PD, and they do not selectively cross the midline at caudal levels of the PD, as proposed by Bell.⁸

At the CVJ, a very small contingent of fibres remained in the ventromedial PD to form the uncrossed ventral CST. Most fibres were located ipsilaterally and there was no topography present here either, as labelled arm fibres were scattered throughout the dorsal, central and ventral regions of the VCST and overlapped with the location of uncrossed shoulder and leg fibres.

Quantifying the degree of intermingling

We confirmed our findings by quantifying the number of labelled fibres, cut in cross-section, in different subsectors of the PT and LCST. We adapted a stereological approach that allowed us to make unbiased estimates of the distribution of labelled fibres as they passed through the PT just above the PD and through the LCST at the mid-C5 and the mid-C8 level.²⁹

The results from four of the eight cases investigated in this way are shown in Fig. 4. Figure 4B and C shows how the PT and LCST were divided into three subsectors or regions of interest (ROIs) (lateral, central and medial: PTl, PTc and PTm) and also indicates how the counting frame was moved in a series of steps to allow unbiased samples to be taken across the whole subsector. Figure 4A shows a high-power photomicrograph of five labelled axons within a counting frame. Figure 4D–K shows the results from four M1 injections. In each panel, the number of labelled axons at each site sampled is shown by the grey circles, and the grey bars show the mean number (and standard error, SEM) of fibres per sample, for each ROI.

Figure 4.

Quantitative assessment of the distribution of labelled CST axons within the PT and LCST in the macaque monkey. (A) High power photomicrograph showing an example of a field of BDA-labelled axons in the LCST with an unbiased counting frame randomly placed by the computer software over a sampling area. Labelled axons that were counted in the quantitative analysis included axons located completely within counting frame (green arrowheads), or axons that were in contact with the green ‘inclusion’ line (none in this example). Labelled axons that contacted the red ‘exclusion’ lines (red arrowhead) were not counted in the analysis. (B) An example of a systematic random placement of the counting grid (250 µm × 250 µm) and corresponding counting frames placed over the central sector of the medullary pyramidal tract (PTc), just above the pyramidal decussation. A single enlarged counting frame is shown on the top left with the counting frame dimensions (100 µm × 70 µm) along with the green ‘inclusion line’ and red ‘exclusion line’. (C) An example of a systematic random placement of the counting grid (250 µm × 250 µm) and respective counting frames (100 µm × 70 µm) placed over the central sector of the LCST (LCSTc) at C5. (D–K) Results from quantitative assessment of labelled fibres in the three subsectors (l, m and c) in representative C5 (D, F, H and J)

and PT (E, G, I and K) sections for four different cases with tracer injection in M1 (D–G: arm/hand; H–I: shoulder; J–K: leg area). The circles represent the number of fibres within each of the counting frames sampled at random over the entire subsector (see B and C). A small random jitter along the x-axis was introduced for visualization purposes only. The grey bars represent the mean (±SEM) number of fibres/counting frame, and the dashed line the regression calculated across the three subsectors. None of the regressions shown were significant [P = 0.17 (D), 0.16 (E), 0.80 (F), 0.26 (G), 0.24 (H), 0.08 (I), 0.60 (J) and 0.22 (K)]. Cent = central; Lat = lateral; Med = medial; LCSTm, c, l = respectively, medial, central and lateral subsectors of the LCST; PTl, c, m = respectively, lateral, central and medial subsectors of the PT. Reproduced from Morecraft et al.^29©The authors, published with permission. CC BY-NC-ND 4.0.

The main finding is that in every case (M1 arm, Fig. 4D–G; M1 shoulder, Fig. 4H–I; and M1 leg, Fig. 4J–K), labelled fibres were present in rather similar numbers in all three subsectors. This was true at all three levels analysed: PT, C5 (Fig. 4D–K), and C8 (data not shown). The distribution of labelled fibres was broadly similar for the M1 arm/hand and leg projections. A one-way ANOVA was performed in each case to determine if the ROI was a significant factor in explaining the variance in fibre distribution across the three ROIs, but in none of the M1 cases did this test return a significant result. None of the gradients shown in Fig. 4 (dashed lines) were statistically significant (ns). This quantitative analysis confirmed that there was no evidence for somatotopy or lamination in the fibre distribution from different M1 representations.

Intermingling of corticospinal tract fibres from secondary motor areas

We also examined the CSP from three of these areas: ventral and dorsal premotor cortex, and M2/SMA. We again found that each of these projections, although less dense than those from M1, had identical distributions, occupying the full extent of the PT and contralateral LCST. The results from an injection in the SMA are shown in Fig. 3 (bottom). Projections from secondary motor areas showed a very different pattern of termination within the cervical cord,^29,31,32 but their course through the PT and LCST completely overlapped that from M1. Again, the quantitative analysis revealed that the CSP from the SMA/M2, LPMCv and LPMCd arm areas showed similar distributions across three ROIs within the PT/LCST.

New findings on the corticospinal projections from cingulate and parietal areas

We have now extended our investigation into the frontal CSP by examining the projections revealed by dextran tracer injections in key regions of the cingulate and parietal cortex. These new investigations were carried out using the same methods,²⁹ where details of all the animals and procedures are provided. The results reported here were studied in a total of six purpose-bred adult rhesus monkeys, listed in Table 1, which also gives details of the gender, weight, area injected and volume of neuroanatomical tracer used, together with the post-injection survival time. All of the injected sites were restricted to the regions indicated. In Case SDM75, the injection sites were characterized by observing motor responses in different body parts evoked by ICMS. The injection site in all other cases listed in Table 1 were localized using anatomical landmarks, cortical cytoarchitectonic maps and cortical stimulation maps.

Table 1.

Description of the experimental parameters in each case

Case	Gender	Weight (kg)	Area injected	Tracer/injections	Total volume (µl)	Post-injection survival (days)
Cingulate motor cortex cases
SDM23	Male	8.5	M3 Arm	BDA/2	0.4	29
SDM23	Male	8.5	M4 Arm	LYD/2	0.4	29
SDM68	Female	7.9	M3 Arm	FD/3	1.2	33
SDM68	Female	7.9	M4 Arm	LYD/3	1.2	33
Parietal cortex cases
SDM75	Male	9.5	S1 Arm (area 3/1)	BDA/3	1.2	33
SDM75	Male	9.5	S1 Arm (area 1/2)	LYD/3	1.2	33
SDM86	Male	15.1	S1 Leg (area 1/2)	BDA/3	1.2	33
SDM73	Female	4.9	Area PE	LYD/3	1.2	33
SDM75	Male	9.5	Area PE	FD/3	1.2	33
SDM86	Male	15.1	Area PEc (lateral)	FD/3	1.2	33
SDM30	Male	6.1	Area PEc (medial)	LYD/1	0.3	28

Cingulate motor areas: M3 and M4

Figure 5 (top panel) shows the results of a FD injection in the arm area of the rostral cingulate motor cortex (CMAr; M3).³⁵ In the PT, labelled fibres were chiefly concentrated in the medio-dorsal region of the PT, but with further scattering throughout the tract. Labelled fibres were seen to cross the midline at all levels of the PD and were then found scattered throughout the LCST within the contralateral cervical spinal cord. Far fewer fibres were found in the ipsilateral LCST. In the same animal, an injection of LYD was also made in the caudal cingulate motor cortex (CMAc; M4). The fibres labelled in the PT (Fig. 5, middle panel) from this injection overlapped completely with those from CMAr, and the same was true for those in the PD and LCST. These results in Case SDM68 were confirmed in a second animal (Case SDM23).

Figure 5.

Corticospinal projection from cingulate motor cortex and area PEc. Line drawings of cortical injection sites on the medial wall of the cerebral hemisphere (far left) and selected transverse sections showing the distribution of labelled axons in the PT, CVJ and cervical spinal cord. Top: Medial view of Case SDM68 showing the FD injection in the arm area of the rostral cingulate motor cortex. The injection site primarily involved cortex lining the ventral bank and fundus of the cingulate sulcus, corresponding to cytoarchitectonic areas 24c and 24d, respectively. Middle: Medial view of Case SDM68 showing the LYD injection in the arm area of the caudal cingulate motor cortex. The injection site primarily involved cortex lining the ventral bank and fundus of the cingulate sulcus, corresponding to cytoarchitectonic areas 23c and 23d, respectively. Bottom: Medial view of Case SDM30 showing the LYD injection in the posterior medial region of the parietal cortex, placed immediately behind to the posterior extent of the cingulate sulcus. The injection site involved parietal cortex corresponding to area PEc. However, it is clear in the sections shown through the CVJ that scattered fibres cross the midline at rostral as well as caudal levels of the pyramidal decussation. Note also the relatively dense distribution of fibres located laterally in the PT of the medulla, compared to the lighter and scattered distribution of labelled fibres located medially. This general pattern of PT fibre distribution was also found following injections placed in the lateral parietal cortices (Fig. 6). cgs = cingulate sulcus; ros = rostral sulcus; rs = rhinal sulcus.

Somatosensory cortex

In Case SDM75, we examined the projection from the arm region of primary somatosensory cortex (S1; areas 3 and 1). Fibres labelled by an injection of BDA into the caudal bank of the central sulcus were clustered in the most ventro-lateral region of the PT (Fig. 6, top row) and these fibres again crossed at all levels of the PD and then descended in the LCST, again scattered throughout the tract. A broadly similar pattern of labelling was found when a different tracer (LYD) was injected into the convexity of the postcentral gyrus (arm region of area 1/2; Fig. 6, second row).

Figure 6.

Corticospinal projection from somatosensory and posterior parietal cortex. Line drawings of cortical injection sites located in somatosensory areas of the parietal cortex/posterior parietal cortex and selected transverse sections showing the distribution of labelled axons in the PT, CVJ and cervical spinal cord. From top down the cases shown are: S1 arm, lateral view of Case SDM75 showing the BDA injection in the arm/hand region of S1 localized with low threshold ICMS. The injection site was confined to cortex lining the posterior bank of the central sulcus, involving cytoarchitectonic areas 3b and 1. Area 1/2, lateral view of Case SDM75 showing the LYD injection in the posterior arm/hand region of S1 localized with low threshold ICMS. The injection site involved cytoarchitectonic area 1 on the dorsal convexity of the central sulcus and area 2 on the gyral surface. S1 leg, lateral view of Case SDM86 showing the BDA injection in the leg region of S1 localized with low threshold ICMS. The injection site was confined to cortex lining the posterior bank of the central sulcus, involving cytoarchitectonic areas 3b and 1. Note in all cases the relatively heavier distribution of fibres laterally in the PT compared to the medial region. However, there is no somatotopy as complete overlap occurs amongst the dense and lightly labelled fibre fields. Area PE, lateral view of Case SDM75 showing the FD injection in the superior parietal lobule, involving cytoarchitectonic area PE on the gyral surface and opercular portion of the intraparietal sulcus. Area PEc, lateral view of Case SDM86 showing the FD injection the superior parietal lobule, involving cytoarchitectonic area PEc on the gyral surface. Note the widespread or scattered distribution of labelled fibres at all levels of the CVJ and within the cLCST, but far less intense in fibre number when compared to the frontomotor CSP. as = arcuate sulcus; cs = central sulcus; ios = inferior occipital sulcus; ips = intraparietal sulcus; lf = lateral fissure; ls = lunate sulcus; ps = principle sulcus; sts = superior temporal sulcus.

At first sight it might appear that there was some localization of these somatosensory fibres within the PT, with the fibres being more concentrated in its ventro-lateral region. However, it is clear that there is no somatotopy from comparison with a case with an S1 leg area injection of BDA in a different animal (Case SDM68). Thus, Fig. 6 (third row) shows that S1 leg fibres are particularly concentrated in exactly the same region of the PT as the arm fibres and are again, like the arm fibres, distributed throughout the LCST in the cervical cord.

Posterior parietal cortex

We also examined the CSP through the PT, PD and LCST from posterior parietal areas PE (Fig. 6, fourth row) and its caudal subdivision PEc, located on the lateral surface of the cerebral hemisphere (Fig. 6, fifth row). In all cases, labelled fibres occupied the full extent of the PT, PD and LCST, without the notable ventro-lateral clustering of fibres in the PT found in the somatosensory cortex cases. However, fibres labelled by an injection placed in the medial region of the posterior parietal cortex corresponding to cytoarchitectonic area PEc (Fig. 5, bottom) were found to be clustered in the most ventro-lateral region of the PT as well as being scattered medially. But once within the PD, these fibres again crossed at all levels of the PD in dispersed fashion. A processed series of spinal cord tissue sections were not available in this case for analysis, but labelled fibres were broadly dispersed in the most rostral level of the LCST.

Conclusion: lack of somatotopy in the LCST

These new studies considerably extend previous evidence against somatotopy in the PT and LCST, since they include data on the finest fibres in the CST, and include the projections from other frontal motor areas, together with those from cingulate motor areas and parietal somatosensory areas. The lack of somatotopy is further substantiated by a rigorous quantitative assessment of the pattern of distribution of M1 fibres across the medio-lateral extent of the tract. These observations should bring to an end any continued ideas of applying the concept of ‘lamination’ to the spinal course of the CST.

Nevertheless, it is worth considering why a detailed somatotopic organization within the LCST, similar to that seen in ascending sensory pathways (Fig. 1C), might be lacking. Is there any evidence that such an organization would support the different functions mediated by the CST? In this context, we must first reiterate that the CST is much more than a descending motor pathway, because there is strong evidence for its involvement in a multiplicity of other functions. As well as its role in the generation and control of movement,² it may also act as part of the ‘mirror neuron system’ during observation of the movement of others.^36,37 The CST also mediates a role in somatosensory control,¹⁷ including effects on tactile and nociceptive perception,^38,39 and the prediction of sensory events that arise from movement.^40,41 It also has a widespread influence on autonomic function.⁴² Finally, the CST can exert long-term actions during development^43,44 and during learning.^45,46

The multiple functions mediated by the CST are unsurprising, given the large number of cortical motor, premotor, sensory and limbic areas that contribute to it, and its wide range of fibre sizes.^{2,27,28,47,48} All of these characteristics make it unlikely that there is a ‘functional’ rather than a simple anatomical somatotopy within the LCST.

But even if we restrict our attention to the motor functions of the CST, it is hard to come up with a good argument in favour of strict somatotopy. There is now wealth of evidence in both humans and non-human primates that M1, for example, is not organized along the lines of a detailed ‘homunculus’^49–51 or ‘simunculus’,^52–55 respectively. In M1, the evidence instead suggests that particular muscles and movements have multiple representations which overlap heavily with those of other muscles and movements. Current thinking is that such an organization allows motor cortex to combine and recombine closely adjacent outputs which gives rise to an enormously varied repertoire for actions such as reach and grasp.^2,56 So, given this type of organization at the cortical level, a strict somatotopy (for example, an orderly location of outputs to the five digits as suggested by the classic ‘motor maps’ of Penfield and others) within the CST seems very unlikely. Instead, the organization of fibres reflects the mosaic complexity of the cortex from which they originate, and the particular pattern of target muscles that characterize individual output neurons in the primate M1.^57,58 In Fig. 7, we present a revised scheme for the intermingling of fibres from the arm/hand, shoulder and leg regions of M1.

Figure 7.

Schematic diagram of the intermingling of fibres within the CST. Summary diagram illustrating the heavily intermingled pattern of CSP fibres from the M1 arm/hand, M1 shoulder and M1 leg regions in the PT of the lower medulla (A), rostral pyramidal decussation (B), caudal pyramidal decussation (C) and throughout all cervical levels of the contralateral LCST as shown at the C5 level (D). This summary is based on the data reported by Morecraft and colleagues.²⁹ In no case was there any evidence of somatotopy in the pyramidal decussation as suggested by Bell⁸ or within the cLCST as suggested by Foerster⁶ in 1936. This same striking pattern of fibre intermingling throughout the CVJ and cervical spinal cord was found following tracer injections in the arm/trunk regions of the premotor cortex and supplementary motor cortex.

In the context of this discussion, it is also interesting that even the CSP from the somatosensory cortex, which, unlike the primary motor area, has a well-defined somatotopy,⁵⁹ still obeys the same principle of intermingling of fibres from different cortical representations. Thus, the fibres originating from the S1 arm area (Fig. 6, top row) travel through the PT, decussation and LCST in exactly the same location as those from the S1 leg area (Fig. 6, third row).

Which factors guide corticospinal projections to their spinal targets?

While it is now clear that somatotopy cannot provide the spatial clues for LCST fibres to find their spinal targets, such a guidance system must exist. For example, despite the intermingling of fibres from different M1 regions in the LCST, there were no terminals in the cervical grey matter from fibres labelled from the M1 leg area,²⁹ meaning that these axons pass through the entire cervical cord without giving off any collaterals, before continuing to the lumbo-sacral cord, where they make extensive terminations in the grey matter.⁶⁰

We also know that each of the cortical areas contributing to the CST terminate within the spinal grey matter in a highly characteristic way, with differences in the degree of bilateral termination [e.g. the M1 hand CSP terminates almost exclusively contralaterally (98%) versus LPMCd with only 79%]. There are also key differences in the level of the cord to which projections reach; for example LPMCv projections are mainly focused on the upper cervical segments,³¹ while the opposite is true of M1 hand fibres.³³ Finally, there are also important differences in the pattern of termination within the spinal laminae. Probably the most striking example of this is the relative absence of terminations in the dorsal horn from the M1 hand CSP, whereas this is the main target of the CSP from the postcentral S1 hand representation.⁶¹

Another key difference between the various CSPs is the extent to which they establish cortico-motoneuronal (CM) connections. Corticospinal neurons in the caudal subdivision of M1 (‘new’ M1) establish the strongest and fastest connections with upper limb motor neurons,^{54,55,62–64} although there are some CM projections from the more rostral subdivision (‘old’ M1).^55,64

Of course, in addition to the coding of each of these different connections, one also has to consider that each CST neuron has an extensive ‘connectome’, which includes all the other cortical and subcortical sites that each CST neuron projects to via its collaterals. These collaterals innervate structures such as the red nucleus, pontine nuclei, reticular formation, etc.^17,65 These patterns of collateralization seem to be species specific.

Current research has shown that each axon carries a number of genetically controlled molecular markers that help the developing growth cone of the axon to interact with its immediate environment and guide it to its targets,^66–68 a process that is further reinforced by early spiking activity in the projection neurons.⁶⁹ A recent report⁷⁰ identified different subpopulations of CST neurons in the mouse motor cortex that extended their axons during development either to the grey matter of the brainstem/cervical cord or to the thoraco-lumbar cord. They showed that the molecular identity of these two subpopulations could be differentiated in the extending axons early in development and long before the axons reach their targets. Further, the molecular specificity of an axon not only defines how far it extends along the cord, but also defines how it collateralizes within the appropriate region of the spinal grey matter: ‘segmentally distinct corticospinal neuron subpopulations are also molecularly distinct’.⁷⁰ Further, developing axons from both brainstem/cervical and thoraco-lumbar subpopulations arrive at the caudal cervical cord at the same time. Hence, there does not appear to be a separation of these axons into either somatotopical space or developmental time, but they clearly have their own distinctive molecular signature. In another paper,⁷¹ the same group identified two of the genes that control the two subpopulations of corticospinal neurons. There are undoubtedly other molecular markers yet to be discovered.

Of direct relevance to this review, this group also showed that in the mouse cervical cord the two groups of axons are heavily intermingled within the dorsal columns, through which the CST passes in rodent species, with no evidence of somatotopy.

How can we explain CCS and cruciate paralysis? Corticospinal tract influence over upper versus lower limb

Early on in the debate about the causes of CCS, another explanation for these syndromes emerged: that incomplete SCI is characterized by diffuse injury to both the central grey and the surrounding white matter.^72–74 Given the intermingling of fibres, a diffuse injury to the CST would result in a greater upper limb (UL) than lower limb (LL) deficit if the CST had a more essential role in the execution and control of hand and arm movements than for leg and foot movements.²⁹ Here we review the salient findings supporting this explanation.

First, there is evidence that there are more corticospinal axons directed to the spinal segments innervating the arm and hand than to those innervating the leg and foot. Weil and Lassek⁷⁵ estimated that 50% of the CST projected to the cervical segments versus 30% to the lumbo-sacral. Their results were based on measurements of the total area of the LCST at cervical versus lumbosacral levels in post-mortem tissue from 10 patients with lesions in the precentral gyrus or internal capsule. They found evidence of many more fibres leaving the tract at cervical levels, presumably terminating in the cervical grey matter, than in the lumbo-sacral cord. This question would be well worth re-investigating with modern methods.

Second, anatomical differences are paralleled by a number of neurophysiological indicators. It is known that the motor cortical representation of the hand and arm, as defined by electrical stimulation of the cortical surface, is considerably larger than that of the leg and foot. A recent re-analysis⁵⁰ of the classic work of Penfield and Boldrey⁴⁹ suggested that comparison of the number of points on the human precentral cortex that produced movement of the arm, hand or digits was around seven times greater than the number of points from which leg, foot or toe movements were evoked. Other measures, including precentral cortical surface area or length along the central sulcus occupied by these representations, confirmed the preponderance of upper over lower limb. Of course, this conclusion is predicated upon the fact that electrical stimulation of the cortex is equally effective in activating corticospinal outputs from the arm/hand area as from the leg area. In this respect, it is worth noting that, in humans, the M1 (area BA4) tissue representing the arm and hand is mostly buried in the rostral bank of the central sulcus, while the leg representation extends onto the surface of the precentral gyrus.⁷⁶ If the greater distance between surface stimulation and M1 CST neurons for the hand/arm versus those of the leg/foot representation has a significant effect on stimulus efficacy, the upper limb/lower limb difference might be even greater than shown by Catani’s analysis.

Third, studies using non-invasive stimulation of human motor cortex report short-latency motor evoked potentials in both upper and lower limb muscles, and these responses are generally larger in upper than lower limb. For example, there are larger responses in the extrinsic muscles acting on the fingers compared with those acting on the toes.^77,78 These findings are consistent with reports in non-human primates of cortical stimulation evoking larger CM effects in hand⁷⁹ than in foot motor neurons.⁸⁰

Fourth, it is possible that pathways additional to the CST are more important for lower limb than upper limb movements, and that these pathways are less vulnerable to damage by incomplete SCI. For example, the CSP to the lower limb⁶⁰ appears to be more bilaterally organized than that to the upper limb.³³ It is also important to recall Nathan et al.,⁸¹ who stated ‘What has never been emphasized is how few degenerating fibres there are below a complete transverse cord lesion caudal to the cervical enlargement. This is evidence that, apart from the lateral corticospinal tract below the cervical enlargement, most of the descending fibres are propriospinal’. We know from extensive work in animals that these propriospinal neurons receive extensive convergence from multiple descending pathways, not just from the CST.⁸² As a result, there might be some redundancy in these different descending inputs to propriospinal projections to the lumbosacral motor neurons, allowing greater recovery after damage. Having said that, we should also take note that intraoperative monitoring during human spinal surgery has revealed that the integrity of CST volleys directed to the lumbosacral cord is a good indicator of functional postoperative ambulatory function.⁸³

Finally, there are a number of interrelated findings that have shown that the dominant influence of the CST is in control of skilled hand and digit movements. For example, lesion studies involving complete bilateral pyramidotomy have a devastating and permanent effect on the capacity of macaque monkeys to perform the relatively independent finger movements required for skilled grasp and manipulation.^84–86 Locomotion was relatively unaffected.⁸⁵ A wide-ranging analysis by Heffner and Masterton⁸⁷ showed that cross-sectional area of the PT was highly correlated with the level of dexterity across a large sample of different primate species. Other evidence has implicated the involvement of fast-conducting CM neurons in dexterous behaviour.^88,89 These neurons have ‘thin’ spikes and fire bursts of high-frequency impulses.⁹⁰ It is interesting that the large axons of these and other neurons are particularly susceptible to spinal injury.⁷²

Finale: a new principle of corticospinal tract organization

Once they reach the lower brainstem (PT) and spinal cord, fibres from different motor cortical areas and projecting to different spinal targets are richly intermingled, and this appears to be a characteristic feature of the CST. As far as we are aware, this concept was first raised by Pappas et al.¹⁸ and was based on findings using anterograde HRP-WGA to label projections from M1 leg and arm areas to the brainstem in the monkey.

We have now explored whether this ‘principle’ applies to a much wider set of observations, made in the macaque monkey, including the CSP from three separate regions of M1 (arm/hand, shoulder and leg) and the CSP derived from the secondary motor areas in the lateral premotor cortex (LPMCd and LPMCv), SMA/M2, rostral cingulate (CMAr; M3) and caudal cingulate (CMAc; M4) motor areas. We have also presented here similar evidence for fibres derived from the somatosensory cortex and related posterior parietal areas (PE, PEc). We made a systematic examination of these projections at multiple levels of the CST, including the PT, CVJ and LCST. Our approach using anterograde dextran markers also provided a comprehensive account of these different CSPs, labelling axons of all diameters, including the numerous, small caliber fibres.

We can conclude that the principle does indeed apply to all these different CSPs. We have already made the point that the multiplicity of function within the CST would be difficult to reconcile with simple somatotopy. Our findings prompt questions about the neurobiological mechanisms responsible for the guidance of these intermingled, different projections to their spinal targets: recent work on the molecular identity of different subpopulations of rodent CST neurons is providing some of the answers. Marked differences in the organization of the CST between rodents and primates, especially in terms of CM projections, will mean extending some of this work to the primate model.

Finally, our findings have further prompted questions about the neurological basis of upper limb deficits after SCIs such as those resulting in CCS and cruciate paralysis. It seems there is ample evidence to support the idea that the especial importance of the CST for skilled upper limb movements means that diffuse injury to the CST that occurs in such syndromes reveals the vulnerable dependence of the upper limbs on these projections. These are important conclusions that should further our understanding of cortical control of movement and its vulnerability to injury and disease.

Funding

Supported by National Institutes of Health grant NS 33003, NS 046367, NS 097450.

Competing interests

The authors report no competing interests.