In this study, the distribution of fluorescent tracer up to 60 min after injection into the spinal cord interstitium was limited to the adjacent two to three spinal cord levels. Tracer was distributed in a radial pattern after delivery into the grey matter, with dissemination into white matter. The absence of statistically significant differences between tracer fluorescence intensities of the grey and white matter after grey matter injections (Fig.
3a, b) is consistent with this observation. However, there was limited redistribution of tracer from white into grey matter after white matter injections. The statistically significant differences between grey and white matter tracer fluorescence intensities after white matter injections support this observation. There was some evidence these patterns were amplified over time. Greater spread of tracer along white matter tracts longitudinally was also observed. There was prominent labelling of all vascular structures by AFO-647. Tracer appeared to be conducted away from the injection site towards the pial surface by depositing around radially projecting arterioles and venules. Support for this inference was provided by the detection of tracer fluorescence around extramedullary vessels. This finding was unlikely to have been secondary to diffusion (or other means of contiguous solute transport such as bulk flow) of tracer because of the general absence of subpial fluorescence (Fig.
6e, i), particularly after grey matter injections. Further microscopic analysis revealed accumulation of ovalbumin both in the perivascular and paravascular spaces of arterial vessels, which will be discussed below.
Although it was not possible in this study to conclude whether diffusion or bulk flow governed interstitial tracer movement, our findings are in concordance with theoretical and animal models of spinal ISF movement from other groups. Confinement of tracer to white matter tracts is characteristic of anisotropic diffusion, well described in the literature on CNS diffusion tensor imaging [
23], and has been confirmed in the developing rat spinal cord in ex vivo experiments [
28,
29]. Here, fluid diffuses along, and is constrained by, myelinated white matter fibres that run parallel to its long axis. The unmyelinated grey matter, however, is the site of penetrating arteries and its extracellular space (ECS) is rich with somas and neurites that have no preferential orientation. Diffusion here is isotropic which may explain why in our experiments tracer in the grey matter was able to redistribute in all directions. Convection enhanced delivery (CED) studies in animal spinal cord have yielded similar findings of anisotropic movement of ligands through the white matter tracts [
30,
31]. Endo et al. [
21], employing Evan’s blue tracer, observed comparable results to ours but described almost no tracer penetration into grey matter after white matter injections. Moreover, Evan’s blue tracer was redistributed further rostrocaudally from the injection site in the white matter compared to the grey matter. These differences could be secondary to the larger delivered volume of tracer (2 µL) in their experiments, and the smaller molecular size of Evan’s blue compared to ovalbumin. Like other CED models and earlier ex vivo work on spinal cord ISF movement [
14,
28,
30‐
34], in Endo’s study a large durotomy was performed, resulting in substantial CSF leak and altered hydraulic integrity of the subarachnoid space and perivascular spaces, which may in turn alter fluid inflow dynamics. Computational simulation of the rat spinal cord by Sarntinoranont et al. [
35] yielded a lower hydraulic conductivity in the grey matter and thus increased tissue resistance. They showed that diffusion of macromolecules
through ECS is limited by tortuosity (higher in grey matter) and efficacy of diffusion diminishes as the square of distance. Diffusion
along ECS is unaffected by these factors [
1,
36]. This would account for the greater penetration of tracer from grey to white matter (compared to white to grey matter) in this study, and the higher fluorescence signal rostral to the injection site observed at 60 min in white matter injections (Fig.
2b). It may also explain the apparent increase in white matter fluorescence after white matter injections at 60 min compared to 20 min, and the absence of this in the grey matter after delivery of tracer into the grey matter (Fig.
7a, d).
Perivascular clearance
Various authors have used the terms “Virchow-Robin space”, “perivascular space” and “paravascular space” interchangeably, but also at times to refer to discrete anatomical compartments. A comprehensive review of the ultrastructure of the “perivascular” space is beyond the scope of this article, but readers are referred to excellent treatises by Bakker et al. [
37] and others [
3,
38‐
40]. For our purposes, we distinguish the “peri-arterial space”, which consists of multiple compartments within the pial sheath that accompanies the arteriole/artery as it enters the CNS parenchyma, from the “para-arterial space”. The latter is the space formed by the glia limitans and the pial sheath of the penetrating artery. The “para-venular space” is formed by the venular adventitia and the glia limitans. Collectively the “para-arterial” and para-venular” spaces form the paravascular compartment. Henceforth, “perivascular space” loosely refers to all the compartments between vessel and glia limitans. These descriptions have been derived from brain studies [
2,
3,
37,
38]. Ultrastructural studies of the rat spinal perivascular space suggest similar anatomy [
41].
There is compelling evidence from our study supporting the importance of the vascular basement membrane in fluid outflow in the spinal cord. Controversy still surrounds the relationship of CSF, ISF and perivascular flow in the brain. There are two prominent contemporary theories of brain perivascular flow—the “glymphatic” system [
10,
42] and a vascular basement membrane model [
11]. The former was borne out of experiments on transgenic mice where intraventricular, intracerebral, and intracisternal injections of CSF tracers established CSF inflow into brain via a “para-arterial” route, bulk interstitial flow, and “para-venous” outflow. Other groups later raised concerns regarding the methodology and interpretation of observations [
2‐
4,
8,
43‐
46]. The Carare–Weller group has long promulgated that fluid influx is via the para-arterial space and clearance of solutes and ISF occurs via the “peri-arterial” vascular basement membrane found within the tunica media. Their model has been backed by experiments from their own laboratory [
11,
38,
47,
48] and from other groups employing intravital multiphoton microscopy [
49]. They also assert that at the level of the capillaries, there is adjacent bidirectional flow of fluid, with inflow occurring adjacent to the glia limitans, and outflow of ISF occurring next to the endothelium. Notably, there is no venular involvement in fluid transport. The major points of dissent are: (1) the types of vascular structure(s) that are involved in ISF and solute egress; and (2) the precise relationship of the outflowing fluid to the various compartments surrounding these vessels. In this study, confocal microscopy demonstrated the presence of tracer not only in the “para-arterial” and “para-venular” spaces, but also in the “peri-arterial” compartment. It appeared that arterioles, venules and even capillaries are implicated in fluid outflow, and hence elements of both dominant theories (that are based on brain studies) have relevance in the spinal cord. Moreover, the vascular basement membrane (as proposed by Carare–Weller) has been shown here to play an important role in solute clearance in the spinal cord, which in turn suggests ISF outflow occurs both within and outside the wall of the arteriole. While injection pump pressure could confound our interpretation of interstitial perivascular tracer deposition, it is unlikely to explain tracer accumulation around only some extramedullary vessels and “remote” arterioles that are far removed from the bulk of the tracer. Moreover, our infusion rate of 0.12 µL/min is lower than that employed by other groups [
3,
50], and thus is unlikely to alter the physiological drainage pathways. Uniform perivascular distribution of tracer around the spinal pial surface, which has been previously observed following cisterna magna injections [
20,
41], was not detected in our experiments. Therefore, accidental delivery into, or recirculation of tracer from the cord back into the subarachnoid space are highly improbable.
Our findings raise the possibility of a model of spinal perivascular fluid dynamics characterised by rapid bidirectional movement. Some authors have suggested that there is little or no directed net fluid displacement in the perivascular space, a concept that aligns with earlier experimental data [
2,
51]. A recent mathematical modelling study proposed that although there might be fast water movement to-and-fro in the perivascular space, solute transfer is facilitated by advection or dispersion [
43]. Dispersion is in turn driven by arterial pulsations, which authors of disparate theories can all agree underlie the mechanism of perivascular flow [
11,
16,
18,
52]. Additionally, solutes may freely communicate between the “para-arterial space” and the “peri-arterial” space through porous barriers that have been confirmed in ultrastructural studies of the spinal cord [
2,
39,
41]. If this “convection” [
3] theory of bidirectional fluid displacement also applies to the “para-venous” space, then tracer molecules injected into the cord parenchyma would disperse along both arteriolar and venular pathways by way of the capillaries [
11]. Initially, there is “peri-arterial” drainage of solutes via the vascular basement membrane, but tracer is then able to infiltrate the “para-arterial” space. As pulsations are much stronger in arteries, tracer is propelled further along arterioles (towards the extramedullary arteries) compared to venules. This is reflected in the preponderance of tracer around extramedullary arteries and “remote” arterioles. However, we would also expect greater “para-venular” tracer deposition at 60 min compared to 20 min. This was, however, not observed, challenging this conjecture on perivascular fluid outflow. Future studies would mandate longer time points to investigate para-venular tracer distribution.
Central canal
Central canal labelling by tracer was detected in 50% of grey matter injection experiments, with a predilection for rostral migration. This corresponds to the earlier observation by Milhorat [
22] of cephalad flow within the central canal, which gave rise to the theory that it acts as a “sink” for excess solutes and fluid from the cord interstitium. Previous work by our group and others has indicated that the central canal is a route of clearance in normal and injured rat and ovine cords [
18‐
20,
24‐
26,
53,
54]. Bedussi et al. have suggested that in the brain, ISF drains preferentially towards the ependymal surface. However, this was only true in close proximity to the ependymal lining and the effect decreased away from the ventricles [
15]. In the spinal cord the distance between the ependymal and pial surfaces is much smaller, so the relevance of this hypothesis is unclear. A more likely explanation for the disparate central canal labelling between grey and white matter injected animals is that diffusivity differences (which in turn are dependent on factors such as tortuosity and distance) at the grey/white matter junction result in the central canal playing a critical role in fluid outflow from the grey matter. It was not possible to clarify, based on confocal micrographs, whether tracer migration into the central canal was transcellular, paracellular or both. Further ultrastructural studies may address this.
Clinical relevance
Findings from these experiments may contribute to a clearer understanding of various spinal cord pathologies. Although outflow pathways have been shown here to involve all vascular structures, there may be regional variations. In the spinal grey matter, there is prominent drainage of solutes and ISF via the numerous ramifying arteries from the ventral median sulcus, as well as into the central canal. In the white matter, outflow efficiency may differ as there are fewer arterioles, and greater reliance on the smaller arterial vasocorona. Extrapolating further, extra-canalicular syringomyelia (a consequence of spinal cord injuries) may be partly precipitated by pathological processes disproportionately compromising ISF drainage via white matter perivascular spaces. Outflow is unable to keep up with fluid influx, ultimately leading to fluid accumulation. Similarly, this may partially explain why spinal cord oedema preferentially follows white matter tracts, as in the grey matter there may be more robust drainage pathways.
Limitations
As some authors have emphasized [
10] that for maintenance of perivascular bulk flow the hydraulic parameters of the subarachnoid and perivascular spaces cannot be compromised. Although CSF losses were not observed during injections in this study, small leaks cannot be ruled out and may account for the relatively limited longitudinal displacement of ovalbumin, and the absence of statistical significance in tracer redistribution between 20 and 60 min. The apparent lack of tracer displacement between the two time points could also be explained by the possibility that these experiments were, at least partially, an investigation of spinal convection enhanced delivery. Other groups have noted that in CED studies, spinal cord injury results in migration of tracer across the grey–white border [
31]. We used the smallest calibre needle possible, but there was still some evidence of local parenchymal trauma due to the cyclical movements of respiration. Unlike in Endo’s study where Evan’s blue did not cross the grey/white junction after white matter injections, tracer in this study was not completely contained within white matter at the level of injection. As with other tracer studies, labelling of the “pial glial” layer and the smooth muscle basement membrane may be explained by selective binding of tracer or by a sieving effect [
8]. Fluid passage within the dorsal white columns was not directly investigated. As this area is isolated from the rest of the white matter the pattern of fluid outflow could theoretically be different. In future investigations of spinal cord fluid outflow, longer experimental time points are recommended. This would validate some of the observed differences between grey and white matter tracer distribution patterns. It may also provide insight into whether spinal CED was actually investigated in these experiments, as well as the role arterial pulsations might play in driving tracer outflow—para-venular tracer deposition may increase with time (see above). It was not possible to quantify the amount of tracer outflow via the various pathways due to the semi-quantitative nature of our results. Finally, these findings were obtained in anaesthetised prone small animals and extrapolation of these findings to upright large mammals should proceed cautiously as volatile anaesthetics are known to alter cardiovascular parameters and CSF production, which in turn affects CSF hydrodynamics [
8].