The main findings of this study are that a CSF tracer administered at the lumbar level enriched the entire parenchyma of the human upper cervical spinal cord and the brain stem. Tracer enrichment in parenchyma is associated with tracer enrichment in adjacent SAS and the distance from the parenchymal surface. Furthermore, the tracer enrichment in both the parenchyma and the SAS is associated with CSF-to-blood clearance and is altered in iNPH compared with younger reference patients.

Gadobutrol as a CSF tracer substance. For this investigation, we used the MRI contrast agent gadobutrol as a CSF tracer substance injected into the SAS at the lumbar level. Gadobutrol is a low-weight (605 Da), hydrophilic molecule that does not pass the BBB. Lipophilic molecules are more rapidly removed from the intrathecal space (18) and do not diffuse as long into the spinal cord as hydrophilic molecules (19). In pigs, 4 different intrathecally administered opioids showed markedly different pharmacodynamic behavior, which in part could be related to differences in hydrophobicity (20). The distribution of tracers in both the brain and the spinal cord in animals is size dependent with slower passage to deeper parenchymal structures for larger molecules compared with smaller molecules (13, 21). The recently discovered subarachnoid lymphatic-like membrane may also restrict the movement of molecules > 3,000 Da within the intracranial SAS (22). Animal studies are essential to understanding physiology, but anatomical differences between species exist. For example, small sensory axons are located in the substantia gelatinosa in the dorsal horn of the spinal cord (23). Sensory impulses are suppressed by morphine injected into this site (24). In rats, this area is 10–20 μm from the spinal cord surface; in dogs, this distance is 200–300 μm; and in humans, it is 500 μm (18). After intrathecal delivery of morphine, the time to onset of effect is 2–3 minutes in mice (25), 30 minutes in cats (26), and more than 1 hour in humans (27). Table 1 summarizes physiochemical and human pharmacokinetic properties of CSF tracers and different intrathecal pharmaceuticals in clinical use. Human in vivo distribution data for the spinal cord for these agents are lacking, but many have approximately the same size as gadobutrol. Depending on other properties, like hydrophobicity, the distribution could be roughly estimated from our data, but intrathecal distribution patterns are complex. To illustrate this, nusinersen, an intrathecally administered antisense oligonucleutide for treatment of SMA, is detectable throughout the CNS in 3 human autopsy cases (15), despite higher molecular weight compared with many other substances (Table 1). Nusinersen is hydrophilic and has a long terminal elimination half-life at 135–177 days (28), which may explain the wide spread of the molecule despite its relatively higher weight.

Tracer distribution in the SAS. The fast distribution of the tracer from the lumbar to the cervical level could be explained by a potent dispersion effect in the spinal canal (29). The less tracer enrichment in the interpeduncular fossa and tracer retention at this site after 48 hours is probably a result of increased distance from the injection site (11) and less dispersion of CSF in a more remote space compared with the SAS in the spinal canal (30, 31).

Tracer distribution in the spinal cord and brainstem. Our results show that tracer in CSF enriches brainstem and spinal cord centripetally from the periphery to the center. Evidence for this includes: (a) there is a faster and stronger enrichment in the superficial regions compared with the deep regions (Figure 3); (b) for the largest structure, pons, the correlation between the tracer enrichment in the SAS and the superficial regions is slightly stronger than the correlation between the SAS and the deeper regions (Figure 4); and (c) the differences in parenchymal enrichment between superficial regions and deep regions are more prominent in larger structures — e.g., pons, compared with smaller structures, such as the spinal cord (Figures 3 and 4).

Based on previous animal studies, the tracer probably moves through perivascular spaces in the parenchyma in a centripetal direction, not through the central canal from the fourth ventricle (11, 13), but how the tracer moves from the SAS to the deep regions of the spinal cord is not possible to assess with the spatial resolution in this study (1 mm voxel size). Apart from perivascular bulk flow, diffusion through the parenchyma is probably also involved (9).

Clearance of tracer. Clearance (T1/2,abs) to the blood and the tracer enrichment in the SAS and in the spinal cord are highly associated (Table 3 and Figure 7); the pharmacokinetic model for CSF to blood clearance provides an estimate of how fast a substance is cleared to the blood, which in turn can estimate the remaining tracer levels in the SAS and parenchyma. We injected the tracer into the SAS and followed the tracer from the SAS into the parenchyma and to the blood. Whether the tracer, when in the spinal cord and brain stem, then returns to SAS through the parenchyma along perivascular spaces and to which extent it is cleared through other routes — e.g., directly to the blood — is not possible to assess with our methodology. The blood-spinal cord barrier is more permeable to tracers compared with the BBB (32) and a tracer injected into the SAS may pass to the vessel lumen (13). As proposed in refs. 13 and 31, substances can possibly pass over the vessel wall from CNS to blood even if passage the other way is restricted. Most gadobutrol injected into the SAS at the lumbar level exits in the SAS before it arrives outside the upper brain convexities, indicating that gadobutrol is mostly cleared at the spinal level, with maximum tracer levels in blood occurring much earlier than in the parasagittal dura mater (30, 33). In mice, tracers in the CSF pass to spinal lymphatic vessels outside of the dura mater (34). In humans, molecules in CSF may pass to lymphatic vessels in the dura mater at the convexity of the skull (35), but other CSF outflow routes also may exist, and a quantitative comparison of different CSF outflow routes in humans is lacking (36). CSF efflux to meninges and bone marrow enables for CNS-immune crosstalk (37–39). Our current findings suggest immunologically derived solutes, and perhaps cells, may also access the spinal cord and brain stem.

Tracer distribution in patients with iNPH. The delayed enrichment of tracer in the parenchyma and slower CSF clearance in the iNPH group compared with the reference group could be related to both disease and age. In patients with iNPH, the clearance from the brain was reduced (8), but patients in the iNPH group in our study were significantly older, and age is correlated to higher CSF tracer levels and decreased CSF clearance (33). The delayed clearance from the parenchyma in the iNPH group could be explained by reduced clearance from the SAS. With more tracer available in the SAS for a longer period, it is reasonable to expect more tracer in the nearby parenchyma. Clearance from the parenchyma to the SAS is not possible to address with the temporal resolution of this study. To assess the clearance from the parenchyma, injection of tracer directly into the spinal cord parenchyma would be preferred, but that is not feasible in humans in vivo. To this end, since CSF clearance differs between patient groups and at an individual level (40), dosage of intrathecal drugs should be individually tailored by assessing CSF clearance capacity before treatment (33).

Limitations. Patients with tentative CSF disorders were investigated in this study with off-label intrathecal use of gadobutrol. Patients without any diagnosis of CSF disorder after assessment at the neurosurgery department were assigned reference patients. While a healthy age-matched control did not exist, we consistently observed tracer enrichment in the spinal cord and brain stem for both groups. For practical reasons, the upper cervical spine and not the entire spinal cord was investigated. The abrupt signal change at the interface between the SAS and the parenchyma may result in Gibbs artifacts in the spinal cord and biased measurements (41). To minimize this, obvious artifacts and measurements in the most superficial parts of the parenchyma were avoided. Also, when artifacts were suspected, the regions of interest (ROIs) were placed in the inflection point of the wave-formed artifacts for the most accurate measurements (41).

Effects of sleep and arousal state for clearance from the parenchyma and CSF were not assessed in this study. In a previous study by our group, with the same method as in the present study, we showed delayed clearance of tracer from the brain, but not from the CSF, after 1 night of total sleep deprivation (42). In a more recent modeling study, which also added the volumes of different brain regions, there was no significant difference between the 2 groups after 24 hours, but groups were significantly different after 48 hours (43). Animal studies have so far shown more prominent effects of sleep for clearance from the brain and CSF (44, 45). All patients stayed at the hospital for 2 nights in a similar setting, and we see it very unlikely that any minor differences in sleep between the groups could have an effect on the present results.

Implications. Altogether, the main implication of our study concerns the potential of distributing drugs in CSF to bypass the barrier constituted by the blood vessel walls and reach the extravascular compartment via CSF from the spinal cord and brainstem surface. The pharmacokinetics of gadobutrol could be used as a rough estimate of the pharmacokinetics for an agent with similar properties. In elderly patients with iNPH, delayed clearance from the SAS and higher amounts of tracer were detected in the parenchyma, which should be considered in intrathecal therapy regimens. The wide parenchymal distribution of this CSF tracer could also apply to antigens and possible immune cells, suggesting CSF as an important mediator between the spinal cord and the immune system in the meninges and the bone marrow.

Further work. In animal studies, the delivery of intrathecal drugs to the CNS via the glymphatic system can be enhanced by systemic drugs (46, 47) and ultrasound (48). The physiochemical complexity, the possibility to modulate the distribution of agents in the thecal space, and differences between species emphasize the importance of further human CSF tracer studies. This could include studies with tracers of varied sizes, addition of systemic drug therapy, and imaging of the entire spinal cord. MR sequences able to quantify the amount of gadolinium are evolving (49) and could be useful for pharmacokinetic assessment of tracer concentrations in the spinal cord.

Conclusion. Intrathecally administered gadobutrol, a small molecular weight CSF tracer, was shown to enrich all ROIs in the upper spinal cord and brain stem from the surface. Drugs with features like gadobutrol can, thus, also be expected to access the extravascular compartment of the spinal cord and brain stem from CSF; however, with different clearance profiles between patients and patient groups. To this end, CSF clearance capacity can be measured by assessing tracer pharmacokinetics from CSF to blood, which may guide dosage of intrathecal treatment regimes.