Decline of intrinsic cerebrospinal fluid outflow in healthy humans with age detected by non-contrast spin-labeling MRI

Article activity feed

1. 

   eLife

   Nov 23, 2022

   eLife assessment

   Malis et al present a novel sequence attempting to non-invasively measure the outflow of cerebrospinal fluid, which is potentially an important contribution given the growing interest in the glymphatic system. Their reported findings are generally consistent with previous literature and prevailing theories, however, no robust validation of the sequence is supplied rendering the evidence base incomplete.

   Read the original source
2. 

   eLife

   Nov 23, 2022

   Joint Public Review:

   In this work Malis et al introduce a novel spin-labeling MRI sequence to measure cerebrospinal fluid (CSF) outflow. The glymphatic system is of growing interest in a range of diseases, but few studies have been conducted in humans due to the requirement for and invasiveness of contrast injections. By labeling one hemisphere of the brain the authors attempt to assess outflow through the superior sagittal sinus (SSS), one of the major drainage pathways for CSF, signal changes across time were assessed to extract commonly used metrics. Additionally, correlations with age are explored in their cohort of healthy volunteers. The authors report the movement of labeled CSF from the subarachnoid space to the dura mater, parasagittal dura, and ultimately SSS, evidence of leakage from the subarachnoid space to the SSS, and …

   Joint Public Review:

   In this work Malis et al introduce a novel spin-labeling MRI sequence to measure cerebrospinal fluid (CSF) outflow. The glymphatic system is of growing interest in a range of diseases, but few studies have been conducted in humans due to the requirement for and invasiveness of contrast injections. By labeling one hemisphere of the brain the authors attempt to assess outflow through the superior sagittal sinus (SSS), one of the major drainage pathways for CSF, signal changes across time were assessed to extract commonly used metrics. Additionally, correlations with age are explored in their cohort of healthy volunteers. The authors report the movement of labeled CSF from the subarachnoid space to the dura mater, parasagittal dura, and ultimately SSS, evidence of leakage from the subarachnoid space to the SSS, and decreases in CSF outflow metrics with older age.

   1. I don't think that the description of Parasagittal dura in figure 1 is correct. There is no anatomical structure at the top of SSS that is known as PSD. The location of the lymphatic structures is also incorrect. Please review "Anatomic details of intradural channels in the parasagittal dura: a possible pathway for flow of cerebrospinal fluid" Neurosurgery 1996 Fox at al. There is usually no obvious tissue between the upper wall of the SSS and the calvarium, which can also be seen in the authors' fig 2A and 2B. All of the tissues located lateral to the SSS are known as PSD. Also, the SSS wall is not as thick as the authors stated and is known as PSD in this region. For this reason, the authors need to revise Fig 1 and it should be changed to PSD in the areas referred to as the SSS wall in the article.

   2. The authors described tagged CSF in two pathways: from dura mater to PSD and SAS into the SSS and directly from SAS to SSS. Flow from dura mater to PSD and SAS in the main and supplement cannot be seen. Only a flow from PSD to SSS can be seen. Also, regular dura cannot carry flow-collagen-rich fibrous tissue, except parasagittal dura. There is no flow from dura to the CSF in the figures.

   3. The authors have conducted many tests to prevent venous contamination. However, measurements were made based on SSS flow rates in all tests. Small parenchymal venous structures, and small cortical-SAS veins might be tagged due to different flow patterns and T2- Relaxation times.

   4. The rate of CSF formation in humans is 0.3 - 0.4 ml min-1. ( Brinker et al 2014. Fluids Barriers CNS). We can assume that the absorption rate is also similar to the CSF formation for the entire system brain and Spine. Therefore, the absorption rate of this very small amount of CSF by SSS is very low in seconds. It is hard to detect by MR and especially CSF flow from the PSD to SSS. The authors concluded that using this technique the rate averaged less than a couple of seconds, rather than on the order of hours or days as previously reported with the use of intrathecal administration of GBCA (Ringstad et al., 2020).

   5. Overall, I think that the CSF flow from the PSD to the CNS described by the authors - the CSF flow, might be the venous flow that drains into the SSS slowly, predominantly in the rich venous channels, venous lacunae, and previously described channels in the PSD. Additional explanations are needed.

   6. The study is generally well described and to the best of my knowledge an innovative approach. The findings are broadly consistent with what might be expected from the literature and the authors make a good argument in support of their findings. However, the lack of validation is a major limitation of the presented work. In introducing a novel technique a comparison with an existing approach, such as Gd enhanced contrast techniques, or phase contrast would have been expected. Several considerations could have been mentioned/addressed in more detail e.g. what effect labeling efficiency, tortuosity of vessels, lack of gating, the effectiveness of the intensity thresholding to remove the signal from blood, etc may have on the quantification, etc. Without a more thorough validation, it is difficult to evaluate the findings. While scans were conducted on two volunteers to assess reproducibility this is a very small sample and it is notable that scans were conducted consecutively, which might be expected to reduce variance relative to scans further apart e.g. on different dates, scanned by a different operator and no information is provided on how the two scans were positioned (i.e. separately vs copied from the first to the second scan), some metrics showed large percentage differences, which were more pronounced in one subject than the other. Without further data, it is difficult to interpret the reproducibility results. No assessment of the effect of physiological parameters e.g. breathing, cardiac pulsations, or factors affecting glymphatic clearance e.g. amount of sleep the evening before was given.

   7. Given these limitations it is hard to adequately assess the likely impact or utility. In recent years several groups have published work e.g. doi.org/10.1038/s41467-020-16002-4 , doi.org/10.1016/j.neuroimage.2021.118755 assessing the blood-CSF barrier. However, previous work has generally focused on larger structures, and by labeling in the oblique-sagittal plane it is unclear how drainage and blood flow rates may affect the presented values here.

   8. Some validation data would greatly increase the value of the reported work. I would therefore encourage the authors to consider acquiring some additional datasets to compare measures of CSF draining against another method e.g. 2-D or 4-D phase contrast, or Gd-based contrast-enhanced techniques. Some additional points to consider are noted below.

   8. Abstract

   CSF outflow may also be imaged with phase contrast MRI (albeit in a limited way).
   Demographics would fit better in Results, breakdown could be given for the young and old groups i.e. n, ages, sex.
   Conclusion - unless further validation can be provided I think some of the claims should be toned down.

   9. Introduction

   The authors emphasise the role of Nedergaard, however, there was some relevant earlier work (e.g. Rennels et al, PMID: 2396537).

   10. Methods

   It would be more conventional to summarise the volunteer characteristics in the Results.
   Given the age difference between the two groups, and the fact that for conventional ASL we know of differences in labelling efficiency and the need for a different post-labelling duration in more elderly patients how did the authors account for this?
   More broadly what would the effect of differences in labeling efficiency be, given the labeling plane is unlikely to be perpendicular to the draining vessels?
   While the authors mention circadian effects there is no mention of controlling for other factors before the scan e.g. caffeine consumption, smoking, etc.
   Various mechanisms have been hypothesised to drive glymphatic pulsations. Assessing how physiological signals correlated with the flow may have been a useful proof of concept. Why was it not considered necessary to use a gated acquisition? Did the authors consider the potential impact of respiratory and cardiac pulsations on their measurements?
   ROI segmentation - manually selected by two raters, was this done independently and blinded? How were consensus ROIs agreed?
   Intensity values outwith MEAN +/- 2 SD were excluded from further analyses. This is justified as removing pulsatile blood. However, was this done independently for tag-on and tag-off? Does this mean slight differences were present in the number of voxels between the two?
   The starting points and parameter ranges are given in Eq'n 3, how were the ranges defined? Was there a reason for constraining the fit to positive values only, is there a risk of bias from this?
   While the main results appear to have a reasonable sample size n=2 for the reproducibility analysis is very limited. Additional datasets would be useful in properly interpreting the results.

   11. Results
   While the authors have taken some measures to reduce potential contamination from blood I would be concerned about the risk of surface vessels affecting the signal, and there does not seem to be an evaluation of how effective their measures are.
   The labeling pulse is applied in the oblique sagittal orientation, but in tandem with differing rates of blood flow and CSF drainage from the labeling plane does that not risk circulating flow from other slices potentially affecting the values?
   Figure 4. The authors focus on the parasagittal dura, but in both the subtraction image and panel C showing different slices at TI=1250 ms some movement appears visible in the opposing hemisphere. Similarly in S2 If the signal does represent CSF movement then this seems counterintuitive and should be explained.
   In Figures 4 and 5 the angulation of the TIME-SLIP tag pulse seems quite different. What procedure was used to standardise this, and what effect may this have on the results?

   12. Discussion
   Phrasing error 'which will be assessed in future studies'.
   I would suggest that some of the claims of novelty be moderated e.g. 'may facilitate establishment of normative values for CSF outflow' seems a stretch given multiple pathways exist and this is only considered one.
   More consideration should be given to some of the points mentioned in the results. The lack of validation should be properly discussed.

   Read the original source
3. 

   Version published to 10.1101/2022.07.14.500033v1 on bioRxiv

   Jul 18, 2022