Emerging evidence supports the concept that cerebrospinal fluid (CSF) acts as a quasi-lymphatic system in the central nervous system (
1). Cardiovascular pulsatility drives CSF inflow along periarterial spaces into deep brain regions (
2,
3), where CSF exchange with interstitial fluid, facilitated by glial aquaporin 4 (AQP4) water channels (
4), takes place. Fluid and solutes from the neuropil are cleared along multiple routes, including perivenous spaces and cranial nerves, for ultimate export to the venous circulation via meningeal and cervical lymphatic vessels (
5,
6). CSF reabsorption may also occur at the sinuses via the arachnoid granulations—although this has not been described in rodents (
7–
10). Despite the efforts dedicated to studying CSF flow along the glymphatic-lymphatic path, it remains to be determined how CSF is transported within the large cavity of the subarachnoid space (
11,
12). In this study, we explored how CSF and immune cell trafficking are organized within the subarachnoid space surrounding the brains of mice and humans.
The meningeal membranes were first analyzed by in vivo two-photon microscopy in the somatosensory cortex of Prox1-EGFP
+ reporter mice (Prox1, prospero homeobox protein 1; EGFP, enhanced green fluorescent protein). Prox1 is a transcription factor that determines lymphatic fate (
13,
14). Second harmonic generation was used to visualize unlabeled collagen fibers, while the vascular volume was labeled by a Cascade Blue conjugated dextran, and astrocytes were labeled by sulforhodamine 101 (SR101, intraperitoneally) (
15,
16). Below the parallel-oriented collagen bundles in dura, we noted a continuous monolayer of flattened Prox1-EGFP
+ cells intermixed with loosely organized collagen fibers. This subarachnoid lymphatic-like membrane (SLYM) divides the subarachnoid space into an outer superficial compartment and an inner deep compartment lining the brain (
Fig. 1A). Quantitative in vivo analysis of the somatosensory cortex revealed that the thickness of SLYM itself was 14.2 ± 0.5 μm, hence thinner than dura (21.8 ± 1.3 μm,
n = 6 mice). The dura vasculature is surrounded by collagen fibers, whereas SLYM covers the subarachnoid vessels. The organization and calibers of the two sets of vasculature also exhibit distinct differences (
Fig. 1, B and C).
A key question is whether SLYM constitutes an impermeable membrane that functionally compartmentalizes the subarachnoid space. To test this, Prox1-EGFP
+ mice were first injected with 1-μm microspheres conjugated to a red fluorophore into the subdural outer superficial compartment of the subarachnoid space along with an injection of 1-μm microspheres conjugated to a blue fluorophore distributed within the inner deep subarachnoid space compartment by cisterna magna injection (
Fig. 2A). In vivo two-photon microscopy showed that the red microspheres were confined to the outer superficial compartment, whereas the blue microspheres remained trapped in the inner deep subarachnoid space compartment. Quantitative analysis showed that the 1-μm microspheres did not cross SLYM from either side. Yet, many solutes in CSF, such as cytokines and growth factors, are considerably smaller than 1 μm in diameter (
17). Therefore, we sought to determine whether a small tracer could pass through SLYM. In these experiments, tetramethylrhodamine (TMR)–dextran (3 kDa) was administered into the deep inner subarachnoid space via the cisterna magna in Prox1-EGFP
+ mice. In six mice, the small tracer did not cross the EGFP-expressing SLYM (
Fig. 2B and fig. S1). Yet, in mice with dural damage and leakage of CSF, the tracer was observed on both sides of the EGFP
+ membrane (fig. S1). Thus, SLYM divides the subarachnoid space into an upper superficial and a lower deep compartment for solutes ≥3 kDa. SLYM is therefore a barrier that limits the exchange of most peptides and proteins, such as amyloid-β and tau, between the upper and lower subarachnoid space compartments.
Live brain imaging avoids fixation artifacts (
18) but cannot immunophenotypically characterize the meningeal membranes. To preserve the integrity of the meningeal membranes, sections were next obtained from whole heads of Prox1-EGFP
+ mice. Immunohistochemistry revealed that Prox1-EGFP
+ cells lined the ventral parts of the entire brain surface (
Fig. 3A). Immunolabeling showed that the Prox1-EGFP
+ SLYM cells were positive for another lymphatic marker, podoplanin (PDPN) (
19), but not for the lymphatic vessel endothelial receptor 1 (LYVE1) (
20) (
Fig. 3, A, lower right panels, and D). SLYM also labeled for the cellular retinoic acid–binding protein 2 (CRABP2) (
Fig. 3, A and D), which is restrictively expressed in dural and arachnoid cells during early development (
21). In contrast to SLYM, lymphatic vessels in dura were positive for all the classical lymphatic antigens, Prox1-EGFP
+, PDPN
+, LYVE1
+, and VEGFR3
+, but was CRABP2
− (fig. S2). Notably, analysis of adult human cerebral cortex depicted that above the pia mater, a CRABP2
+/PDPN
+ membrane was present in the entire subarachnoid space (
Fig. 3, B and C). Thus, SLYM also surrounds the human brain. We infer that the SLYM monolayer of Prox1-EGFP
+ cells organizes into a membrane rather than vessel structures and exhibits a distinctive set of lymphatic markers (
Fig. 3E). To distinguish SLYM from the structures forming the arachnoid mater, we used immunolabeling against claudin-11 (CLDN-11), a main constituent of the tight junctions that create the arachnoid barrier cell layer (ABCL) (
22). CLDN-11 was densely expressed in ABCL as well as in the stromal cells of the choroid plexus, but SLYM was CLDN-11
− (fig. S3, A and B). Additionally, ABCL was distinctively positive for E-cadherin (E-Cad) (
Fig. 3C), as previously reported (
23,
24). We also compared SLYM to the arachnoid trabeculae (
25), collagen-enriched structures that span the subarachnoid space, finding that cells surrounding the arachnoid trabeculae are Prox1-EGFP
−/LYVE1
− (fig. S3C). Pial cells covering the cortical surface also exhibited an immune-labeling profile that differed from that of SLYM (figs. S3 and S4). We conclude that SLYM constitutes a fourth meningeal layer surrounding the mouse and human brain displaying lymphatic-like features (Prox1-EGFP
+, PDPN
+, LYVE1
−, CRABP2
+, VEGFR3
−, CLDN-11
−, and E-Cad
−) and that SLYM is phenotypically distinct from dura, the arachnoid, and pia mater (
Fig. 3E). Interestingly, SLYM expressed PDPN, sharing a trait with the mesothelium lining the body cavities (
26). Accordingly, we observed PDPN
+ cells lining the kidney, as well as PDPN
+ podocytes in the kidneys of adult C57BL/6J mice (fig. S5A). In a human fetus, a PDPN
+ membrane corresponding to pericardium, pleura, and peritoneum encases the developing heart, lungs, and intestinal tract, respectively. PDPN
+ lymphatic vessels were also observed in the lungs and intestinal tract (fig. S5, B and C). Thus, SLYM may represent the brain mesothelium and, as such, covers blood vessels in the subarachnoid space (
Fig. 1) (
26). The mesothelium is present where tissues slide against each other and is believed to act as a boundary lubricant to ease movement (
27). Physiological pulsations induced by the cardiovascular system, respiration, and positional changes of the head are constantly shifting the brain within the cranial cavity. SLYM may, like other mesothelial membranes, reduce friction between the brain and skull during such movements.
Does SLYM have additional functions? The arachnoid villi and granulations are defined as protrusions of the arachnoid membrane into the lateral walls of the sinus veins and are believed to act as passive filters that drain CSF from the subarachnoid space into the venous sinus system (
7–
10). The arachnoid villi and granulations are present in the brains of humans, primates, and larger animals such as dogs, but not in the brains of rodents (
28,
29). We critically reexamined this issue to evaluate the distribution of SLYM in relation to the superior sagittal and transverse sinus. Sections obtained from decalcified heads of Prox1-EGFP
+ mice showed that Prox1-EGFP
+ SLYM cells often were in direct contact with the venous sinus endothelial cells (
Fig. 4A). Thus, the arachnoid barrier cell layer (CLDN-11
+/E-Cad
+), which normally separates dura from the subarachnoid space, was lacking in discrete areas allowing SLYM to directly contact the venous sinus wall (
Fig. 4B). Prox1-EGFP
+ SLYM cells were not positive for CLDN-11 or E-Cad, which distinguish the arachnoid barrier cell layer (fig. S6).
Are the close appositions of SLYM and the venous endothelial cells permeable, allowing the exchange of small molecules between blood and CSF? To test this, we used the principles of bioluminescence, wherein the convergence, in the same compartment, of an enzyme with its substrate is needed to trigger light emission. First, we delivered the luciferase enzyme from
Oplophorus gracilirostris (NanoLuc) fused to the fluorescence tag mNeongreen (GeNL, 44 kDa) (
30) into CSF via the cisterna magna of wild-type (C57bl/6) mice, and allowed it to circulate for 30 min to ensure thorough distribution by the glymphatic system. The distribution of GeNL was verified by mNeongreen fluorescence. Then, the blood-brain barrier (BBB)–impermeable substrate fluorofurimazine (FFz, 433 Da) (
31) was administered intravenously (fig. S7, A to C) (
32). After intravenous injection of FFz, a bright bioluminescence signal catalyzed by GeNL was detected specifically near the large venous sinus wall (fig. S7, A and B). The bioluminescence signal was particularly strong around the confluence of sinuses (fig. S7B). The distribution of the bioluminescence signal was quantified by plotting the mean signal intensity profiles perpendicular to the venous wall of the transverse sinus and superior sagittal sinus. The mean bioluminescence signal profiles intersected with the fluorescence signal profiles of the intravascular tracer (TMR-dextran, 70 kDa) or with shadow imaging of the inverted GeNL signal outlining the vascular wall (fig. S7C). Thus, the bioluminescence signal was restricted to the venous wall of the two major sinuses lacking a BBB (
33,
34), consistent with the notion that FFz is BBB-impermeable and requires the catalyzation enzyme NanoLuc to generate photons (fig. S7, A to C). In control experiments, FFz was delivered intravenously, while the GeNL injection into CSF was omitted. In these control experiments, no bioluminescence signal was detected from the exposed cortex, including from the sinus venous wall (fig. S7D). In another set of control experiments, GeNL was injected into the soft ear tissue, while FFz was delivered intravenously. Consistent with the notion that peripheral blood vessels are leaky (
11), light emission was clearly observed in the region of the ear injected with NanoLuc but not in surrounding noninjected regions of the same ear. No signal was observed in the venous compartment, likely reflecting that blood flow rapidly diluted the bioluminescence signal (fig. S7E). Together, this analysis shows that a small molecule, FFz, can enter the central nervous system (CNS) from the blood and activate an enzyme, NanoLuc, present in CSF, resulting in the generation of photons along the wall of the venous sinus. On the basis of the juxtaposition of SLYM and the venous endothelium in histological examination (
Fig. 4A), the selective generation of photons when luciferase was injected into CSF, and the fact that the substrate was present in the vascular compartment (fig. S7, A to C), we propose that the apposition of the venous endothelia and SLYM represents rodent arachnoid villus–like structures, comparable to those in human brain.
The mesothelium surrounding peripheral organs acts as an immune barrier (
26). Does SLYM also impede the entry of exogenous particles into CSF? In vivo two-photon imaging of Prox1-EGFP
+ mice injected intravenously with rhodamine 6G (Rhod6G) to label leukocytes (
35) showed that a large number of Rhod6G
+ myeloid cells are embedded in SLYM (
Fig. 5A). The number of Rhod6G
+ leukocytes in dura and SLYM was directly comparable, suggesting a prominent role of SLYM in CNS immune responses, which supports the finding that leptomeninges are densely populated with immune cells (
36) (
Fig. 5A). How do systemic inflammation and aging affect the immune cell populations residing in SLYM? Ex vivo analysis of brain sections obtained from Prox1-EGFP
+ mice showed that, in the control group, CD45
+ cells were abundant, located mostly along pial vessels in the surface of the brain (
Fig. 5B). This observation, together with the significant increase in CD45
+ in inflammation-prone conditions [aging and lipopolysaccharide (LPS)–treated mice, 4 mg/kg of body weight, intraperitoneally (ip), 24 hours] (
Fig. 5C), suggests that SLYM can act as a CD45
+ recruiting and/or proliferating site in pathological conditions. Of note, the dose of LPS used (4 mg/kg) did not affect the BBB (fig. S8). Additional immune markers showed that LYVE1
+ (
Fig. 5D), CD206
+ (
Fig. 5E), and CD68
+ (fig. S9) macrophages can be found in SLYM, together with dendritic cells (CD11c
+) (fig. S9). Despite the absence of CD3
+ and CD19
+ lymphocytes (fig. S9), our results indicate that SLYM functions as a niche for immunological surveillance. Thus, in young, healthy mice, SLYM hosts CD45
+ cells, but the number and diversity of innate immune cells rapidly expands in LPS-induced inflammation and was also significantly altered in aged mice. We conclude that SLYM fulfills the characteristics of a mesothelium by acting as an immune barrier that prevents exchange of small solutes between the outer and inner subarachnoid space compartments and by covering blood vessels in the subarachnoid space.
Discussion
The critical roles of the meningeal membranes lining the brain have only recently been acknowledged (
5,
37). It is now known that CSF is drained by a network of lymphatic vessels in the meninges and that suppression of this drainage accelerates protein aggregation and cognitive decline in animal models of neurodegeneration (
38–
40). SLYM is Prox1
+/PDPN
+/LYVE1
−/CRABP2
+/VEGFR3
−/CLDN-11
−/E-Cad
− and thereby distinct from the traditional meningeal membranes, including dura, arachnoid, and pia, as well as the meningeal lymphatic vessels and the arachnoid trabecula. SLYM subdivides the subarachnoid space into two compartments, suggesting that CSF transport is more organized than currently acknowledged. For example, SLYM covering the vasculature in the inner subarachnoid space will guide CSF influx along the penetrating arterioles into the brain parenchyma without circulating solutes present in the outer subarachnoid space compartment. Yet the discovery of a fourth meningeal layer, SLYM, has several implications beyond fluid transport. The observation that SLYM is a barrier for CSF solutes that have a molecular weight larger than 3 kDa will require more detailed studies but indicates a need to redefine the concept of CNS barriers to include SLYM. The meningeal membranes are hosts to myeloid cells responsible for immune surveillance of the CNS (
5,
37), and SLYM, owing to its close association with the brain surfaces, is likely to play a prominent role in this surveillance. Herein, we showed a large increase in the number and diversity of immune cells residing in SLYM in response to acute inflammation and natural aging. Physical rupture of SLYM could, by altering CSF flow patterns, explain the prolonged suppression of glymphatic flow after traumatic brain injury as well as the heightened posttraumatic risk of developing Alzheimer’s disease (
41,
42). Rupture of SLYM will also permit the direct passage of immune cells from the skull bone marrow (
33,
43) into the inner subarachnoid space, with direct access to the brain surfaces, possibly explaining the prolonged neuroinflammation after traumatic brain injury (
44). SLYM may also be directly involved in CNS immunity, in addition to being host to many immune cells. Lymphatic-like tissues can transform quickly in the setting of inflammation, which in the brain may be of notable relevance for diseases such as multiple sclerosis (
45).
Acknowledgments
We acknowledge P. S. Froh and H. Nguyen (Department of Cellular and Molecular Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark) for their excellent technical assistance for the histology and immunohistochemistry of the decalcified samples. We also thank D. Xue for expert graphical support, B. Sigurdsson for analysis, and H. Hirase, N. Cankar, and N. C. Petersen for critical reading of the manuscript.
Funding: Funding was provided by Lundbeck Foundation grant R386-2021-165 (M.N.), Novo Nordisk Foundation grant NNF20OC0066419 (M.N.), the Vera & Carl Johan Michaelsen’s Legat Foundation (K.M.), National Institutes of Health grant R01AT011439 (M.N.), National Institutes of Health grant U19NS128613 (M.N.), US Army Research Office grant MURI W911NF1910280 (M.N.), Human Frontier Science Program grant RGP0036 (M.N.), the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (M.N.), and Simons Foundation grant 811237 (M.N.). The views and conclusions contained in this article are solely those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the National Institutes of Health, the Army Research Office, or the US Government. The US Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein. The funding agencies have taken no part on the design of the study, data collection, analysis, interpretation, or in writing of the manuscript.
Author contributions: K.M. and M.N. designed the study. F.R.M.B., P.K., L.M.M., C.D., V.P., M.K.R., R.S.G., N.L.H., T.E., and Y.M. performed the experiments, collected the data, and performed the analysis. K.M. and M.N. wrote the manuscript. All authors read and approved the final version of the manuscript.
Competing interests: The authors declare that they have no competing interests.
Data and materials availability: All data are available in the main text or the supplementary materials.
Is the central nervous system enclosed by a mesothel?
After detailed comparison of the here presented findings by Møllgård et al. (Møllgård et al., 2023) with our own optical coherence tomographies of the meninges (Hartmann et al., 2020; 2019a), multiple previous electron microscopic studies (Andres, 1967; Reina et al., 2002; 1998; Schachenmayr and Friede, 1978) as well as daily medical observations during neurosurgical practice (De Bonis et al., 2009; Zainel et al., 2021; Edlmann et al., 2017; Santarius et al., 2009) we consider an alternative but most relevant interpretation of their results.
We assume that the neurothelial cell layer that encloses the central nervous system (CNS) could physiologically function as a mesothel. Situated inferior to the dura mater and superior to the arachnoid mater, it could act as an immunological tissue and as a physiological shifting layer between the rigid dura mater and the pulsating CNS. Analogue to numerous other mesothelia of moving organs such as the pleura of the lung, the pericardia of the heart, or the peritoneum of the gut.
This concept would not only contribute significantly to the understanding of the recently discovered lymphatic system of the CNS (Louveau 2015, Apelund, 2015, Absinta 2017) but clarify numerous longstanding contradictories concerning the physiology and pathophysiology of the controversially discussed "subdural space" (Haines et al., 1993).
We also published this concept in more detail in a recent peer reviewed article (Hartmann et al., 2023).
In Fig. I. a sketch demonstrates the position of the neurothelial cell layer in relation to the meninges. [LEGEND FIGURE I: Neurothelial Cell Layer in Relation to the Meninges; Simplified sketch of the meningeal composition: (1) skull at the level of the os frontale, (2) dura mater, with its inner and outer membrane and intradural veins as well as parallel running intradural lymphatic vessels, (3) neurothelial cell layer, which we assume acts as a mesothelial fold of the CNS, (4) subarachnoid space (SAS) filled with cerebrospinal fluid (CSF) enclosed by the impermeable arachnoid barrier cell membrane (ABCM) and arachnoid trabeculae spanned across. Note here the intra arachnoid blood vessels. Arteries show a thicker tunica media due to a strong tunica muscularis and supply the brain cortex. Veins show an increased diameter with a thin tunica media and drain off venous blood from the brain parenchyma. Venous blood is then transported via bridging veins to the intradural venous system, (5) pia mater and grey matter of the brain cortex, (6) white matter, (7) sinus sagittalis superior and (8) parasinusoidal intradural lymphatic vessels.]
In regard to the study by Møllgård et al. we assume (A) that the two-photon imaging in Prox1-EGFP+ reporter mice did not stain an undescribed cell layer inside the subarachnoid space – like suggested by the authors - but rather the previously well described neurothelial cell layer, which is situated inferior to the inner membrane of the dura mater and superior to the arachnoid barrier cell membrane of the arachnoid (Andres, 1967) (Reina et al., 1998) (Reina et al., 2002). These cells show a nuclear thickness of 0.5 - 1 µm, length of 100 µm, forming up to 8 parallel rows, and are surrounded by amorphous material, leading to an approximate thickness of this cell layer of 5 - 7 µm. Their appearance under electron microscopy was well demonstrated e. g. by Reina et al. 2002, see Fig II. [LEGEND FIGURE II.: Electron Microscopic Imaging of Neurothelial Spinal Cells; A: Dura-arachnoid interface is seen below the dural lamina. The dura-arachnoid interface is filled with neurothelial cells and amorphous material. Here demonstrated with transmission electron microscopy [TEM], magnification ×5000, bar = 1 µm. Orange box indicates a probable scanning point for B: Subdural space seen by scanning electron microscopy, limited by the dura mater with a thickness of 300 µm and the arachnoid barrier cell membrane with a thickness of 40 µm. The internal surface of both layers is covered with neurothelial cells.]
These morphological characteristics and the neuroanatomic position is in clear conformity with the Prox1-EGFP+ stained cells described by Møllgård and colleagues.
For example, Fig. 1. A. and C. of Møllgård et al. depict that the cell layer is situated inferior to the dura mater and superior to the arachnoid mater. If the stained layer were inside the arachnoid mater – like suggested by the authors – the arachnoid blood vessels needed to be around them, see an excerpt of Fig. 1. A. and C. by Møllgård et al. in Fig. III which clearly depicts that the arachnoid green stained cell layer is situated superior to the arachnoid blood vessels. [LEGEND FIGURE III.: Excerpt from Fig. 1. A. and C. by Møllgård et al.; A: The lateral view - top row on the right and down row - enables a clear overview of the composition of layers. Note that the green cell layer - named by the authors: subarachnoid lymphatic-like membrane, SLYM - is situated inferior to the dura mater but superior to the arachnoid blood vessels and therefore superior to the arachnoid.]
Here it is to be noted that the arachnoid vessels, which are the "vasa privata" of the brain, solely exist inside the subarachnoid space, which is physiologically enclosed by the arachnoid barrier cell membrane, as clearly depicted by our first in vivo optical coherence tomography of the human subarachnoid space (Hartmann et al., 2019b), see Fig. IV. [LEGEND FIGURE IV: In Vivo Visualization of Human Subarachnoid Space and Cranial Meninges with OCT; A: Light microscopic image after right fronto-lateral craniotomy. Opened segment shows frontal brain cortex. Orange line indicates region of scan. B: OCT-scan of frontal arachnoid mater, pia mater and cerebral cortex. C: Enlarged excerpt demonstrating details of the subarachnoid space. Note that the arachnoid barrier cell membrane is covered by a slim inhomogeneous layer which might represent neurothelial cells. Further note that the arachnoid vessels are inside the subarachnoid space which is enclosed by the arachnoid barrier cell membrane.]
Fig. 1 B. and especially C. of Møllgård et al. also demonstrate the classic morphological character of neurothelial cells and again its position inferior to the dura mater and superior to the arachnoid blood vessels and therefore superior to the subarachnoid space.
In their second experiment, Møllgård et al. injected red fluorophore inferior to the dura mater and blue fluorophore at the position of the cisterna magna into the subarachnoid space. Since in vivo two-photon imaging then depicted red fluorescence superior and blue fluorescence inferior to the green stained cell layer, the authors concluded that the green stained cell layer states an impermeable membrane inside the subarachnoid space, see Fig. 2. A. Møllgård et al.
In contrast to the authors – we assume (B) that the subdural injection tore the neurothelial cell layer apart and separated the arachnoid barrier cell membrane from the dura mater, leading to the formation of a subdural space.
If the neurothelial cell layer separated the subarachnoid space – as postulated by the authors – the cell layer needs to be in close proximity of the subarachnoid vessels and not superior to them, see excerpt of Fig. 2. A. and Fig. S1. of Møllgård et al. in Fig V. [LEGEND FIGURE V: Excerpt from Fig. 2. A. and Fig. S1. by Møllgård et al.; Excerpt from Møllgård et al. Fig 2. A. and Sup. 1. The left row demonstrates a measurement after subdural injection of red fluorophore and subarachnoid injection of blue fluorophore. The right row shows subarachnoid injection of red fluorophore in Prox1-EGFP+ reporter mice. If the green stained cell layer would be an impermeable membrane of the subarachnoid space – like suggested by the authors – the cell layer would be in close proximity to the arachnoid blood vessels. Instead here it is clearly depicted, that the cell layer is superior to the shadow of the undyed subarachnoid vessels.]
The here described formation of a subdural space is regularly seen in various medical conditions and during microsurgical dissection of the dura mater:
After a portion of the skull is removed during craniotomies, only the superior layer of the dura mater, the osseous or endosteal layer, is incised so that the two layers of the dura mater can be slightly lifted. At this moment, the arachnoid barrier cell membrane is still adherent to the intact inner meningeal layer of the dura mater and lifts with it. The inner membrane of the dura mater is then very carefully incised without harming the arachnoid barrier cell membrane beneath. At this moment, the arachnoid barrier cell membrane independently begins to separate from the dura mater. Blunt instruments can then enter the newly formed subdural space easily, and the dura mater can be cut without harming the now separated arachnoid barrier cell membrane. The internal side of the dura mater and the outer surface of the arachnoid barrier cell membrane appear to be moist but dry quickly. No CSF is drained, and the "vasa privata" of the brain remain untouched in the intact subarachnoid space.
This formation of the subdural space in vivo was clearly depicted with optical coherence tomography by our group, see. Fig VI (Hartmann et al., 2020). [LEGEND FIGURE VI. Surgical Formation of a Subdural Space Visualized with OCT; D: OCT scan of dura mater. During craniotomy the inner membrane of the dura was punctured leading to the formation of a subdural space by detaching of the arachnoid barrier cell membrane from the dura mater. E: Enlarged excerpt demonstrating details of the dura mater, subdural space and arachnoid mater. F: Schematic drawing of microstructures: (1) + (2) dura mater, (1) outer endosteal layer, (2) inner meningeal layer, (3) subdural space, (4) subarachnoid space (4) arachnoid barrier cell membrane, (5) subarachnoid blood vessels, (6) subarachnoid space, (7) trabecular system, (8) brain cortex and (9) reflection artifacts.]
Our interpretation of the neuroanatomic position of the green died cell layer is further supported by the here published e-letters from various expert groups (1) Rieck et al., (2) Siegenthaler, (3) Pan et al. and (6,7) Bestholtz et al.
If our two interpretations (A) and (B) are correct, the logical conclusion would state that the CNS could be covered by a mesothelium.
Interestingly, this concept would be consistent with numerous physiological, pathophysiological, immunological, and neuroanatomical observations for which current explanations and therapeutic approaches tend to fail.
It would solve the physiological question, why the pulsating CNS is able to shift effortlessly in relation to the rigid dura mater - like the lung in relation to the thorax, the heart in relation to the pericard, or the intestines in relation to the fascia abdominalis.
It would solve the question what the subdural space is and why it is formed.
It might further solve various pathophysiological inadequacies concerning diseases of the subdural space and its therapies:
Finally, the here described concept, could clarify and rapidly progress the current discussion on the structure of the lymphatic system of the CNS. Serous fluids produced by the neurothelial cell layer would physiologically drain via the dural lymphatic vessels. An increased understanding of the CNS lymphatic system might have further impact on the therapy of numerous infectious but especially autoimmune neuroinflammatory diseases (Louveau et al., 2018).
The ontogenetic origin of such a mesothelic cavity is not farfetched, since it could have developed analogously to other mesothelia of numerous organs by enfolding of the neural tube in the mesodermic germ layer during embryogenesis.
As a first step, we would suggest a reevaluation of Møllgård et al.’s findings.
Whereas spatial resolution of the naked eye, classic neuroimaging techniques like 3 or 7 T MRI, neurosonography, intraoperative light microscopy, and conventional OCT are insufficient to display this cell layer, in vivo imaging methods with increased spatial resolution, e.g. psOCT or speckle modulated OCT might be able to redefine its anatomic position and tract its physiological movements (Hartmann et al., 2021).
In a second step, the internal side of the dura mater and the surface of the arachnoid barrier cell membrane need to be screened for cells of mesothelial origin before the arachnoid barrier cell membrane is opened.
References
Alfageme, I., Muñoz, F., Peña, N., Umbría, S., 1993. Empyema of the thorax in adults. Etiology, microbiologic findings, and management. Chest 103, 839–843. doi:10.1378/chest.103.3.839
Andres, K.H., 1967. [On the fine structure of the arachnoidea and dura mater of mammals]. Z Zellforsch Mikrosk Anat 79, 272–295.
De Bonis, P., Anile, C., Pompucci, A., Labonia, M., Lucantoni, C., Mangiola, A., 2009. Cranial and spinal subdural empyema. Br J Neurosurg 23, 335–340. doi:10.1080/02688690902939902
Dennis, B.M., Bellister, S.A., Guillamondegui, O.D., 2017. Thoracic Trauma. Surg Clin North Am 97, 1047–1064. doi:10.1016/j.suc.2017.06.009
Edlmann, E., Giorgi-Coll, S., Whitfield, P.C., Carpenter, K.L.H., Hutchinson, P.J., 2017. Pathophysiology of chronic subdural haematoma: inflammation, angiogenesis and implications for pharmacotherapy. J Neuroinflammation 14, 108–13. doi:10.1186/s12974-017-0881-y
Haines, D.E., Harkey, H.L., al-Mefty, O., 1993. The “subdural” space: a new look at an outdated concept. Neurosurgery 32, 111–120. doi:10.1227/00006123-199301000-00017
Hartmann, K., Neyazi, B., Stein, K.-P., Haghikia, A., Sandalcioglu, I.E., 2023. Is the central nervous system enclosed by a mesothel? Therapeutic Advances in Neurological Disorders. doi:10.1177/17562864231180335
Hartmann, K., Stein, K.-P., Neyazi, B., Sandalcioglu, I.E., 2021. Theranostic applications of optical coherence tomography in neurosurgery? Neurosurg Rev 1–7. doi:10.1007/s10143-021-01599-x
Hartmann, K., Stein, K.-P., Neyazi, B., Sandalcioglu, I.E., 2020. Optical coherence tomography of cranial dura mater: Microstructural visualization in vivo. Clin Neurol Neurosurg 200, 106370. doi:10.1016/j.clineuro.2020.106370
Hartmann, K., Stein, K.-P., Neyazi, B., Sandalcioglu, I.E., 2019a. First in vivo visualization of the human subarachnoid space and brain cortex via optical coherence tomography:. Therapeutic Advances in Neurological Disorders 12, 175628641984304. doi:10.1177/1756286419843040
Hartmann, K., Stein, K.-P., Neyazi, B., Sandalcioglu, I.E., 2019b. Aneurysm Architecture: First in vivo Imaging of Human Cerebral Aneurysms with Extravascular Optical Coherence Tomography. Cerebrovasc. Dis. 48, 1–6. doi:10.1159/000502450
Louveau, A., Herz, J., Alme, M.N., Salvador, A.F., Dong, M.Q., Viar, K.E., Herod, S.G., Knopp, J., Setliff, J.C., Lupi, A.L., Da Mesquita, S., Frost, E.L., Gaultier, A., Harris, T.H., Cao, R., Hu, S., Lukens, J.R., Smirnov, I., Overall, C.C., Oliver, G., Kipnis, J., 2018. CNS lymphatic drainage and neuroinflammation are regulated by meningeal lymphatic vasculature. Nat. Neurosci. 21, 1380–1391. doi:10.1038/s41593-018-0227-9
Møllgård, K., Beinlich, F.R.M., Kusk, P., Miyakoshi, L.M., Delle, C., Plá, V., Hauglund, N.L., Esmail, T., Rasmussen, M.K., Gomolka, R.S., Mori, Y., Nedergaard, M., 2023. A mesothelium divides the subarachnoid space into functional compartments. Science 379, 84–88. doi:10.1126/science.adc8810
Reina, M.A., De Leon Casasola, O., López, A., De Andrés, J.A., Mora, M., Fernández, A., 2002. The origin of the spinal subdural space: ultrastructure findings. Anesth Analg 94, 991–5– table of contents. doi:10.1097/00000539-200204000-00040
Reina, M.A., López García, A., de Andrés, J.A., Villanueva, M.C., Cortés, L., 1998. [Does the subdural space exist?]. Rev Esp Anestesiol Reanim 45, 367–376.
Santarius, T., Kirkpatrick, P.J., Ganesan, D., Chia, H.L., Jalloh, I., Smielewski, P., Richards, H.K., Marcus, H., Parker, R.A., Price, S.J., Kirollos, R.W., Pickard, J.D., Hutchinson, P.J., 2009. Use of drains versus no drains after burr-hole evacuation of chronic subdural haematoma: a randomised controlled trial. Lancet 374, 1067–1073. doi:10.1016/S0140-6736(09)61115-6
Satoh, Y., 2016. Management of chest drainage tubes after lung surgery. Gen Thorac Cardiovasc Surg 64, 305–308. doi:10.1007/s11748-016-0646-z
Schachenmayr, W., Friede, R.L., 1978. The origin of subdural neomembranes. I. Fine structure of the dura-arachnoid interface in man. Am J Pathol 92, 53–68.
Vakamudi, S., Ho, N., Cremer, P.C., 2017. Pericardial Effusions: Causes, Diagnosis, and Management. Prog Cardiovasc Dis 59, 380–388. doi:10.1016/j.pcad.2016.12.009
van Baal, J.O.A.M., Van de Vijver, K.K., Nieuwland, R., van Noorden, C.J.F., van Driel, W.J., Sturk, A., Kenter, G.G., Rikkert, L.G., Lok, C.A.R., 2017. The histophysiology and pathophysiology of the peritoneum. Tissue Cell 49, 95–105. doi:10.1016/j.tice.2016.11.004
Zainel, A., Mitchell, H., Sadarangani, M., 2021. Bacterial Meningitis in Children: Neurological Complications, Associated Risk Factors, and Prevention. Microorganisms 9, 535. doi:10.3390/microorganisms9030535
Møllgård et al. Science, 2023: Concerns of rigor and objectivity
In our initial Science eLetter of March 1, 2023 (1), we sought to bring to the authors’ and readers’ attention our concerns over scientific problems identified in data and claims in Møllgård et al. (2). However, the authors’ eLetter response (3) to our eLetter, as well as their responses to the eLetters by others (4-6), raise even more serious concerns. The authors’ defense of their claims and denial of the validity of all criticisms and alternative interpretations weigh against conventional scientific rigor, objectivity, and skepticism of their findings.
In their paper (2) and eLetter-responses to criticisms (3, 7, 8), the authors also attempt to discredit the investigators who submitted critical eLetters by claiming that they lack the experience necessary to comment and that their comments are scornful. How else can readers interpret the authors’ concluding remarks in their eLetter:
“…we would like to note the limitations of an unregulated online forum, in which unvetted and somewhat unhinged remarks, whose factual bases have been neither objectively validated nor independently sustained, can masquerade as critical review - in contrast to the careful, well-informed and unbiased peer review to which their object of scorn have already been subject.” (3).
As described in Ragnar Levi’s article, Reasons for Scientific Rigor and Skepticism Revisited:
“A hallmark of sound scientific thinking is to be self-critical rather than self-serving. {….} core principles that should guide all scientific efforts include disinterestedness – acting for the benefit of a common scientific enterprise, rather than for personal gain – and organized skepticism – exposing claims in scientific papers to critical scrutiny before submitting or accepting them for publication (9).”
Stated another way by Igor Grossman and Justin Brienza: “Wise reasoning characteristics include intellectual humility, recognition of uncertainty, consideration of diverse viewpoints, and an attempt to integrate these viewpoints (10).”
Seven examples of misinterpretations, inadequate supportive evidence, and projection of the problems onto others that raise questions of scientific rigor and objectivity are described at the end of this eLetter. Six of them are directly linked to the central claims of a 4th meningeal layer that subdivides and forms a barrier between “lower” and “upper” compartments of the subarachnoid space. The seventh questions why the authors attempted to discredit their critics’ expertise (“shoot the messenger”) instead of providing balanced responses to criticisms. Why not welcome comments from readers with all levels of experience and points of view?
In this context, we are puzzled by the authors’ emphasis of a single interpretation of their data and rejection of alternative explanations of their data that do not fit their hypothesis. Objectivity requires balanced consideration of interpretations that compare the investigator’s views against other interpretations, and the scientific method requires a concerted effort to disprove the investigator’s hypothesis (11, 12). If extraordinary claims are made, as in Møllgård et al. (2), extraordinary evidence in support of the claims is required for acceptance. If their data regarding the existence of a 4th meningeal layer were solid, and their interpretations were compelling, beyond reproach, and completely consistent with the findings, there would be no need to argue that their “critics have little experience” and that comments in an eLetter are “unhinged” (3) (“unhinged” means “deranged”, “crazy”, or “mentally unbalanced”). Instead, it would seem more constructive and scientifically rigorous to welcome comments and alternative interpretations that could lead to new experiments, reevaluation of existing data, and progress in understanding the meninges.
Also puzzling is the authors’ criticism of the eLetter process by calling the submission of Science eLetters “an unregulated online forum” in which readers’ comments and alternative interpretations “can masquerade as critical review.” The authors do not explain why they consider signed eLetters by established investigators (1, 4-6) less meaningful than unsigned reviewer comments they describe as “careful, well-informed and unbiased peer review” (3). Even papers that are retracted after publication have presumably undergone careful, well-informed, and unbiased peer review. And mistakes can happen during the review process.
Further discussion of these issues would benefit from other investigators weighing in with their views and the original reviewers of Møllgård et al. (2) identifying themselves, as we have, and providing their updated assessment(s) in light of the concerns raised in this and other eLetters (1, 4-6).
Examples that raise questions of scientific rigor and objectivity of data, figures, and interpretations in Møllgård et al. (2):
References
Response to eLetter by Christer Betsholtz et al.
We thank our colleagues for the repeated posting of their comments regarding our study describing a 4th meningeal layer, which we have named the SLYM layer (1). Our colleagues raise 3 sets of critique: They state that SLYM is identical to the arachnoid membrane, that the data does not show that SLYM has a barrier function and that the subarachnoid space therefore not is compartmentalized. They also state that no convincing evidence is presented to justify the description of "SLYM” as “lymphatic-like” or as a “mesothelium."
We believe that it is important to note that most of the critique is based on old histological studies of postmortem tissue. The original studies of meningeal membranes surrounding the adult brain date from the 1920’s through the mid ‘80s and are based on ordinary histology and EM analysis. The fixation techniques used in these classical histological methods yield significant changes in tissue morphology, and in particular affecting fluid filled spaces; fixatives are hyperosmotic, resulting in deformation and shrinkage of the thin meningeal membranes, and their artifactual adhesion to adjacent layers. This typically leads to the collapse and often disappearance of smaller fluid-filled spaces, a limitation of classical histology that has been well-recognized for decades. The more recent phenotypic characterization of the meningeal layers has primarily been done in the developing brain, which may not be representative of adult meningeal phenotype.
No convincing evidence is presented for the existence of a fourth membrane of the meninges separating two compartments of the subarachnoid space.
Our study is based on a combination of ex vivo immunophenotyping and in vivo functional studies. To our knowledge, the barrier functions of the meningeal layers had never before been studied using real-time 2-photon imaging, which is free of fixation artifacts. Our report defines a subdivision of the subarachnoid space into spatially and functionally distinct compartments, separated by a Prox1-defined lymphatic mesothelium. We should note that our critics have little experience with, and neglect the importance, of optical imaging in live mice. This was a critical part of our study, and the means by which we identified the subcompartmentalization of the subarachnoid space.
Our study was designed to understand how fluid flow is organized in the large subarachnoid compartment. To achieve this end, we needed to collect in vivo data, since it is known that fluid-filled compartments are deformed and essentially disappear during tissue fixation. (Please see our prior publication: https://www.nature.com/articles/s41467-018-07318-3. The movie included in this publication shows that the CSF-filled perivascular spaces do not survive fixation. See: https://static-content.springer.com/esm/art%3A10.1038%2Fs41467-018-07318-3/MediaObjects/41467_2018_7318_MOESM4_ESM.mov). The same phenomenon, cavity disappearance revealed only by their surviving boundary membranes and cells in ex vivo histology, was noted with the recent designation of a new organ, the interstitium (2). See: https://www.scientificamerican.com/article/meet-your-interstitium-a-newfound-organ/. This study revealed that the collagen surrounding peripheral organs does not form the dense wall observed in histological sections, but rather that the connective tissue surrounding organs is in fact an “open, fluid-filled highway”, whose internal structure and porosity are lost with fixation. This discovery shares many similarities with our report on SLYM. In essence, the collapse of the subarachnoid space in histological sections is the reason that SLYM appeared to merge with other meningeal layers, such that its functional significance in separating the subarachnoid space into discrete compartments was not previously described, and in fact could not have been done before the advent of the live in vivo analysis that we conducted.
We would like to refer our colleagues to the Krisch, Leonhardt and Oksche (1983, 1984) (3, 4), in which the meninges surrounding the mouse cortex were studied by tracer injection. Fig. 1 (see: https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511) shows a schematic drawing from these authors illustrating the meningeal layers. Using their original nomenclature and moving down from the skull to the brain, the meningeal layers were defined here as: (1) dura, (2) the inner dura layer, (3) the neurothelium and (4) the outer arachnoid layer. Below the outer arachnoid layer is the (5) cerebrospinal fluid (CSF)-filled arachnoid space, followed by (6) the inner arachnoid layer and (7) the outer pial layer. The inner arachnoid layer and the outer pial layer are often fused, and the resultant dual layer is called the intermediate lamella. Below the intermediate lamella is (8) the pial space, the floor of which is created (9) by the inner pial layer. (10) A subpial space is also described. Injection of a tracer, horseradish peroxidase (HRP) in either the arachnoid or pial spaces demonstrated that the intermediate lamella is indeed a barrier that separates the CSF-filled arachnoid and pial compartments.
The barrier function of this intermediate lamella was defined by injecting horseradish peroxidase (HRP) into either the arachnoid or the pial spaces (3, 4). Fig. 2 (see https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511) displays the original data by which the authors demonstrated the existence of two separate CSF-filled compartments: the arachnoid space (A), and the pial space (P). The barrier between the two compartments is created by the inner arachnoid layer and the outer pial layer that together create the intermediate lamella that effectively separates the subarachnoid space into two distinct compartments.
Krisch et al., 1983 concludes: “Due to the development of the pia mater and the arachnoidea from a common matrix primitiva and due to the cytologic characteristics common to the pia mater and the arachnoidea, one should avoid the terms "pia" and "arachnoidea". Both terms, having their origin in gross anatomy, should be replaced by the terms of "inner, intermediate", and "outer leptomeninges" encompassing the inner and outer leptomeningeal spaces (3). These observations are strikingly similar to what we reported in Møllgård et al., 2023 (1).
No convincing evidence is presented for the Prox1-positive layer of cells acting as a barrier between two compartments.
Our critics state that we injected in the subdural space, an artificial space created when dura and the outer arachnoid layer (EA) are separated by cleavage of the inner dura layer (Fig. 1). It is important here to note that the outer arachnoid layer (EA) is a barrier expressing the tight junction claudin-11 (5), and that the inner dural cells and the neurothelium (also called the dural barrier cell layer) do not express tight junctions, as such, they do not form a barrier layer facing the collagen-rich dura mater.
We, therefore, studied the classical literature and found that bleeding into the subdural space penetrates directly into the dura mater and distributes along the dural collagen fibers. See Fig. 3, a diaminobenzidine (DAB)-reacted Nissl-stained cryosection (s) of a pig with an experimental subdural hemorrhage, established via subdural injection of blood containing Horseradish Peroxidase (HRP). The underlying cortex is also shown (6). The HRP reaction (black) is clearly restricted by the arachnoid barrier cell layer (ABCL); no blood enters the subarachnoid space (SAS). The key observation here is that HRP enters dura outlining the collagen fibers demonstrating that no barrier exists between the arachnoid barrier cell layer (ABCL) and the dura (6). Similarly, Nabeshima and Reese (1972) found that HRP injected into the dura reaches the arachnoid barrier cell layer (7). All these studies simply reflect the overwhelmingly typical clinical observation that neither acute nor subacute subdural hemorrhages traverse into the subarachnoid space in the absence of traumatic penetrating rupture of the latter.
Based on these findings – that no barrier exists between the subdural space and dura mater, we re-examined the experiments in which microspheres were injected in the upper subarachnoid space - or on top of SLYM (Møllgård et al., 2023, Figs. 2A-B). As described, the dura mater was gently punctured at the edge of the craniotomy and lifted using fine forceps, inserting a 35G needle mounted to a stereotaxic arm at an 85° angle into the subarachnoid space. Because the dural opening was larger than the needle, only a fraction of the injected solution containing microsphere solution remained in the upper subarachnoid space. The purpose of this slow infusion was to gently allow a fraction of microspheres to enter the CSF space before mounting a coverslip. Did we inject in the subdural space? We do not believe so. We are in Fig. 4A (see https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511) displaying a representative set of images from an experiment repeated successfully in seven mice. Second-harmonic 2-photon imaging was used to visualize the collagen fibers along with the 1 µm microspheres (red). As noted above, we would expect that microspheres injected into the subdural space on top of the outer arachnoid layer (also called the arachnoid barrier cell layer, ABCL) have free access to dura, since the inner dura layer is not a barrier. It is in panel A clear that the microspheres distributed independently of the collagen fibers. In Fig. 4B, we believe that the microspheres accidentally were injected on top of ABCL because the microspheres distributed in a pattern closely resembling that of dural collagen fibres, suggesting that the microspheres travelled in between the fibres. The similarity in distribution of collagen and microspheres were noted in two mice, that were consequently excluded from the analysis showed in Fig2A of Møllgård et al., 2023.
The second line of data was generated from Z-stack reconstructed from 2-photon imaging of Prox1-EGFP reporter mice in which a very small tracer, 3 kDa tracer was infused into the inner subarachnoid space by cisterna magna injection. Thus, the semi-invasive procedure was > 2 cm from the field of view. In these experiments, nothing was injected into the upper subarachnoid compartment and dura was left intact. Second harmonic and regular 2-photon excitation was used to collect optical Z-stacks and cross-sections that are displayed below (Fig. 5, see https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511). The images clearly demonstrate that the dura collagen fibers and SLYM are separated in some regions, but that the two layers are in close contact to each other in other areas. The remarkable observation is that the 3 kDa tracer did not leak into the upper subarachnoid space, i.e., the space between the collagen fibers and SLYM within the observation period (1-2 hrs). These data show that SLYM indeed acts as a barrier.
We described in the submitted manuscript the procedures in detail because the methodology was developed to interrogate the functional properties of SLYM in live brains free of fixation artifacts. Both Figs 4 and 5 demonstrate that SLYM creates the pial space or an inner subarachnoid space CSF-filled compartment.
The major point of these two sets of in vivo studies in live mice was to show that the subarachnoid compartment is subdivided into two functional compartments. As such, our data demonstrate that subarachnoid CSF transport is organized separately within at least two subcompartments and is not simply distributed uniformly within a single large compartment, a concept that might seem simple, but substantially extends our conception beyond that provided by the current literature.
No convincing evidence is presented to justify the description of "SLYM” as “lymphatic-like” or “mesothelium." Or to justify the proposed functions of “SLYM”
Our phenotypic characterization of the Prox1-EGFP+ SLYM showed that its cells can be defined as PDPN+, LYVE1 , CRABP2+, VEGFR3 , CLDN-11 , and E-Cad . This pattern of immunostaining differs from all other layers of the meningeal layers, as documented in Figs. 3, 4, 5 and Supplementary Figs. 2-6 of Møllgård et al., 2023. We show in higher magnification some of the data displayed in Møllgård et al., 2023 (Fig. 6, see https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511). A double layered membrane consisting of 2 phenotypically distinct layers is clearly identified. Critique has been raised about SLYM being identical to the arachnoid barrier cell layer (ABCL). We do not think it is fruitful to go into a discussion about what is reported with regard to immunolabelling of SLYM compared to the previous published data. Our colleagues are citing developmental papers or referring to studies of peripheral tissues. It is well-known that lymphatic and mesodermal markers exhibit variations in expression pattern across species and change dramatically during development. Single cell transcriptome data will be needed to obtain a better understanding of the meningeal layers and their evolutionary and developmental similarities and differences.
Critique has also been raised about what justifies the name SLYM, i.e., Subarachnoid lymphatic-like membrane. The SLYM expresses two markers of lymphatic tissues, Prox1 and podoplanin. To be clear, the SLYM is not lymphatic tissue; it does not express VEGFR3 or LYVE1. In general, multiple lymphatic markers are needed to characterize traditional lymphatic tissue. We named the layer SLYM because its cells express the lymphatic master transcription factor, Prox1. Finally, critique is raised for the concept that SLYM may represent a brain mesothelium, analogous to that which typically surrounds other organs and body cavities. Mesothelial cells express glycoproteins that bind water to reduce friction when organs are moving. Typically, mesothelial membranes support and wrap around vessels and nerves, and are hosts for immune cells. Our conception that the SLYM comprises part of the brain’s mesothelium was based on (1) that SLYM surrounds the brain similar to the mesothelium around organs and body cavities, (2) the shared immunogenicity of SLYM with the mesothelium surrounding peripheral organs, (3) that SLYM has the classical appearance of a mesothelium consisting of a simple single-layered, squamous cell layer (Fig. 7, https://www.alzforum.org/news/research-news/and-then-there-were-four-new-meningeal-membrane-discovered#comment-48511), 4) the presence of a large number of immune cells residing in SLYM similar to mesothelium surrounding other organs, (5) that SLYM clearly surrounds the pial vasculature. Future single-cell transcriptomic data will elucidate how many similarities SLYM shares with traditional mesothelium membranes. There will be, without doubt, differences, since the mesothelium surrounding different organs differ, but the five sets of evidence listed above justify that SLYM can be described as a mesothelium – in fact, our paper was not the first to do so. Many classical histology texts refer to a layer of the meninges as mesothelia, and as far back as 1938. Weed stated that “…the villi are essentially continuations of the arachnoid mesh into the lateral walls of the great dural sinuses, so that the arachnoid mesothelium comes to lie directly beneath the vascular endothelium” (8). Also, Andres et al., (1967), Illen & Woollam (1962) and Kirsch et al., (1983) used the term mesothelium (9) (10) (3). We suspect that our colleagues were unaware of the literature, which in fact previously demonstrated a compartmentalization of the subarachnoid space and discuss the existence of a mesothelium in the subarachnoid space.
We will like to note that neurosurgeons, who professionally dissect meningeal membranes in live brain have for centuries been aware that the subarachnoid space is divided into multiple compartments. See for example: (11, 12).
Again, we would like to thank the respondents for their comments; we understand that vehemence can sometimes accompany passionate interest. As a broader point though, we would like to note the limitations of an unregulated online forum, in which unvetted and somewhat unhinged remarks, whose factual bases have been neither objectively validated nor independently sustained, can masquerade as critical review - in contrast to the careful, well-informed and unbiased peer review to which their object of scorn have already been subject.
References:
Comments on Møllgård et al. Science January 6, 2023: Is the subarachnoid space divided by a newly discovered 4th layer of meninges?
This recent paper in Science (1) claims the existence of a hitherto unrecognized 4th meningeal layer called “SLYM” with novel molecular properties, barrier functions, and immunological features.
As soon as this paper was published, we individually recognized that the claims were not convincingly supported by the data and that they contradicted findings in the same paper and many well-documented features of the meninges. We also realized that our diverse scientific backgrounds, expertise, and familiarity with the meninges and methods used in the paper put us in an effective position and gave us an obligation to critically assess the authors’ data and claims. Our collective firsthand knowledge and experience includes (i) intravital microscopy (near-infrared fluorescence, epifluorescence and 2-photon in vivo imaging) of the brain and spinal cord, including meninges, immune cells, and cerebrospinal fluid circulation and clearance, (ii) single-cell analysis of meningeal cell gene expression, (iii) structural and molecular analysis of meningeal cell adherens junctions and tight junctions, (iv) characterization of lymphatic structure and function, including lymphatics in the dura mater, (v) use of tracers to test cellular barriers, (vi) weighing the strengths and limitations of immunohistochemical staining, 2-photon imaging, confocal microscopy, and electron microscopy, and (vii) building on the history of the cellular basis of the blood-brain and meningeal barriers identified in the Laboratory of Neuropathology and Neuroanatomical Sciences at the NIH and subsequently confirmed by many others.
With this background, our collective assessment of the evidence presented in Møllgård et al. revealed serious flaws in the authors’ data and interpretations. The authors ignored conspicuous inconsistencies and contradictions. They also omitted essential controls in their own data and ignored centuries of work from prior investigators. The following are among the problems that undermine the main claims made by the authors. All figure citations refer to figures in Møllgård et al.
These concerns logically lead to the question of whether the data in Møllgård et al. can be reconciled with these problems and contradictions and still preserve the validity of the authors’ claims that a 4th meningeal layer divides the subarachnoid space into two compartments, serves as a barrier between the two subdivisions, and has properties of both lymphatics and mesothelial cells.
To address this question, we considered multiple counterarguments. The argument that the use of 2-photon in vivo imaging revealed layers and spaces not previously identified fails because spaces hundreds of micrometers in width can be identified without specialized imaging. The argument that the “outer subarachnoid space” was artifactually eliminated by tissue shrinkage in all previous studies using light microscopy, confocal microscopy, or transmission electron microscopy fails because the authors show the “outer subarachnoid space” in fixed, dehydrated, paraffin embedded, and sectioned mouse and human specimens (Figs. 3 and S3). Furthermore, the subarachnoid space of the same dimensions is visible regardless of the methods used for tissue processing and imaging. Spaces of this size do not just disappear. The argument that immunohistochemical staining proved that the Prox1-EGFP cells were not adjacent to the E-cadherin-positive arachnoid barrier layer (Fig. 3C, 3E) fails for the reasons described in points 2-4 above. The argument that the Prox1-EGFP-positive cells have a barrier function fails because Møllgård et al. did not test the barrier function of these cells apart from the arachnoid barrier layer, which is not visible in their 2-photon images. The argument that Prox1-positive cells in the meninges have properties of lymphatic endothelial cells and mesothelial cells fails for the reasons described in point 4. Even as a hypothesis, the relationship of a subset of meningeal cells to lymphatics and mesothelial cells would have to be objectively tested with consideration of the evidence in favor of or against the concept before it could be taken seriously.
After considering all of these issues, we conclude that Prox1 is expressed by a subset of arachnoid cells located close to the arachnoid barrier layer, but the authors’ claims of the discovery of a “4th meningeal layer” that divides the subarachnoid space and forms a barrier between the subdivisions are misrepresentations and misinterpretations of their data. Findings that underlie these claims can be explained on the basis of what is already known about meningeal cellular structure and function.
Response to Dr. Seigenthaler's eLetter
We appreciate Dr. Seigenthaler’s interest in our work. We have previously provided a lengthy explanation outlining the significant fixation artifacts that change not only the morphology, but also the apparent locations, of fragile membranes within fluid-filled spaces – including in this case the membranes within subarachnoid compartments. In addition, we have described how tracers are mislocated upon death, as first reported by Dr. Theise’s group 1. In her critique of our present study, Dr. Seigenthaler refers to histological studies, as well as to the phenotypic characterization of tissues harvested from other species, other developmental stages, and other CNS regions. Yet such imprecise comparisons can be as misleading as they are uninformative. Brain fluid transport and its vectorially-organized flow dynamics are highly specialized, and as such, so are the membranes covering the CNS. It is thus critical to avoid generalization based on fixation artifact, and to instead capitalize upon the availability of contemporary live imaging technology to accurately describe brain fluid dynamics. Importantly, these new technologies can validate as well as expand the scope of traditional histology. Indeed, our recent publication demonstrated that imaging of collagen around penetrating arteries using second harmonic generation replicated traditional immunohistochemistry 2, in contrast to Dr. Seigenthaler’s concerns in that regard. Going forward, we hope that Dr. Seigenthaler will participate in the acquisition of new and relevant in vivo data, which might better shed new light on the exciting world of glymphatic transport and its structural restrictions.
Comments on the letter from Rieck & Veh, Institut für Zell- und Neurobiologie, Centrum 2, Charité - Universitätsmedizin Berlin:There are no separate functional compartments within the subarachnoid space
Since 1956 it has been well established that the subarachnoid space is compartmentalized by initial findings based on encephalography (Liliequist, 1956), evidence that has been used among neurosurgeons all over the world since then.
Our colleagues from Berlin cite common textbook knowledge about meningeal anatomy including dural border cells and their involvement in chronic subdural hematomas (CSDH’s), though without mentioning the review in Nature from 2014 (Kolias et al., 2014). We have checked a number of CSDH-cases and found no evidence of PROX1-positive membranes associated with the hematomas, i.e. ‘the “SLYM” is NOT just a novel staining of the classical subdural neurothelium.
Rieck & Veh then claim that ‘All descriptions of the meningeal layers at the electron microscopic level identify the subarachnoid space as a single compartment’ obviously neglecting that their highly esteemed colleagues from Giessen and Hamburg (A Oksche, H Leonhardt and B Krisch) published two important papers in 1983 and -84 describing a convincing compartmentalization. For further discussion and references, see Alzforum “And then there were four”, pages 10-19, 2023.
Finally It is not correct that our ‘resolution of the immunohistochemical images is much too low to differentiate the meningeal layers in detail’. Quite the opposite, the single-layered SLYM and the ABC-layer are clearly shown in Figures 4, C and D. See also the high magnification of LYVE1-positive macrophages associated with SLYM in Fig. S4 and the SLYM covering the basilar artery in Alzforum, page 18.
Hopefully this answer the statements by Rieck & Veh concerning SLYM and the meningeal layers of the central nervous system.
Comment on Møllgård et al., 2023 “A mesothelium divides the subarachnoid space into functional compartments.”
Møllgård et al. define a meningeal structure designated as the subarachnoid lymphatic-like membrane (SLYM) that is located between the arachnoid and pia maters(1). On the basis of this definition, the authors claim the SLYM divides the subarachnoid space (SAS) into two functional compartments, the outer SAS and the inner SAS. We discuss the authors’ conclusions as they relate to the data presented and to previously published work characterizing the meninges.
In Figure 2 of Møllgård et al. the authors state the SLYM subdivides the SAS into two compartments, the outer SAS and the inner SAS. In the experiment upon which this conclusion is based, the authors injected red microspheres by lifting the dura and injecting into the subdural space. Blue microspheres were then injected via the cisterna magna into the SAS. In the figure legend for Figure 2A, the authors alternatively define the “outer SAS” as the “subdural space.” However, a subdural space is defined as a space below the dura but above the arachnoid(2), which is distinct from the SAS which lies below the arachnoid and above the pia. In addition, the subdural space is a well-known space that contains CSF, and additional CSF may accumulate within the subdural space in the setting of pathology or trauma(3–6). As the dura and arachnoid are not fused membranes, the subdural space can also be observed ex vivo due to the collapse of the arachnoid membrane secondary to the lack of CSF in the SAS(2, 6). Therefore because the subdural and subarachnoid spaces are distinct CSF-filled spaces and the arachnoid is not labeled in the projection provided in Figure 2A, it remains unclear into which CSF space the red microspheres were injected. We believe further clarification is needed before the authors can conclude that the SAS is divided into two functional compartments as it appears that Figure 2 depicts the subdural and subarachnoid spaces, and not subdivisions of the SAS.
The authors differentiate the SLYM, in part, from arachnoid and dura based on CRABP2 positivity, as they state CRABP2 is “restrictively expressed in dural and arachnoid cells during early development.” The authors do not present novel data to support this claim, and the paper cited does not substantiate this claim, as it characterizes CRABP2 expression in the postnatal and adult rat olfactory epithelium, respiratory epithelium, ovary, and retina, but not the meninges(7). Existing transcriptomic and immunofluorescence data show that both arachnoid and dural fibroblasts are enriched in CRABP2 in fetal development and postnatally(8–12). Therefore, it is difficult to differentiate the SLYM from the arachnoid with CRABP2 staining without further definitive experiments showing CRABP2 is not expressed in the arachnoid fibroblasts of adult mice.
As described by electron microscopy studies of the meninges conducted in the latter half of the previous century, the dura, arachnoid, and pia each contain multiple distinct layers of cells. The cranial dura is generally accepted to be two layers, the outer periosteal layer and the inner meningeal layer, each of which is composed of one or more cellular layers(13). The arachnoid is accepted to be composed of 2-3 layers of cells with distinct patterns of tissue organization, which includes the arachnoid barrier layer in addition to the underlying arachnoid trabeculae(13). Therefore, an alternative conclusion may be that the SLYM fits into a cell layer within one of the three canonically defined meningeal layers. There are no figures or schematics in the manuscript in which all four meningeal layers (dura, arachnoid, SLYM, and pia) are identified and simultaneously labeled, which makes it difficult to appreciate the distinction of the SLYM as a 4th layer. We believe that the identification of a new meningeal layer would be strengthened by additional histological and ultrastructural characterization of the SLYM and its anatomic relation to the dura, arachnoid, and pia, as well as the overlying skull and underlying brain parenchyma.
Finally, how the SLYM may fit into the context of common pathologies of the arachnoid and SAS such as arachnoid cysts and subarachnoid hemorrhage was not discussed, nor are there any known pathologies related to the SLYM. Neurosurgical access to the brain requires careful dissection of the dura and the arachnoid. There have been no neurosurgical reports of an additional layer separating the SAS into two compartments to date. We do agree that the arachnoid is a trabecular structure such that within the SAS there are known webbed membranes (in addition to the arachnoid barrier cell layer), however we believe these do not represent a distinct layer but are rather part of the arachnoid, as the authors also indicate in Figure 3E. Indeed, while it appears that the authors show the close approximation of the arachnoid barrier cell layer and the SLYM ex vivo (Figures 4C, 4D), for the SLYM to divide the SAS into truly functional compartments the arachnoid must be distinct from the SLYM in vivo with CSF separating the two layers, which would also allow for neurosurgical dissection. It may be argued that the SLYM is unable to be visualized by neurosurgeons due to its thin nature (14.2 µm in mice(1)) compared to the dura (21.8 µm in mice(1)) and arachnoid; however, we believe this is unlikely because the human dura, which can be seen by naked eye, is approximately 12 times the thickness of the mouse dura reported in this study(14, 15), which is a scale that would likely put the human SLYM in range for both surgical loupes (minimum 2.5x magnification) and microsurgical dissection with operating microscopes.
In summary, Møllgård et al. provide an interesting characterization of a fourth meningeal layer, SLYM, that draws attention to the perhaps underappreciated complexity of the meninges. However, there are a few key claims about the unique immunohistochemical, anatomical, and structural identity of the SLYM that are not optimally supported by the data presented, and further investigations are needed before we fully embrace the existence of a fourth meningeal layer that subdivides the SAS.
References
Post-publication comments on Møllgård et al., 2023 “A mesothelium divides the subarachnoid space into functional compartments.”
I read with great interest the recent paper by Møllgård et al. describing a subset of cells in the meninges co-expressing Podoplanin (Pdpn) and Prox1. I have spent over 15 years studying meninges function in brain and vascular development and, more recently, molecular heterogeneity of the meningeal cell layers in the fetal and adult brain. I believe the meningeal layers are functionally specialized to support brain health and function, this based on our work and literature on developing and mature meninges that has been steadily building over the last 20+ years.
In my view, the meninges are too often viewed as a single structure vs its individual parts. Decades of ultrastructural studies, now augmented with single cell profiling studies in developing and adult meninges show that there are likely 3 molecularly distinct dural fibroblast layers1, at least 2 arachnoid layers [our work in Developmental Cell2, arachnoid barrier cell, lower arachnoid fibroblasts that express Raldh2/Crabp2 and potentially cells associated with trabeculae that may be molecularly similar] in addition to pia and perivascular fibroblast layers that are molecularly similar but have different locations. Our validation in human fetal brain2 and a recent paper3 and preprint4 in human adult leptomeninges single cell profiling support layer-specific molecular profiles are present in human meninges. We are at the point that it is not accurate to say, ‘This is a function of the pia. This is a function of the arachnoid. This is a function of the dura’. Adopting additional sublayer specific descriptors, for example a molecular marker(s) and/or functional aspect like ‘E-cadherin+ arachnoid barrier cell’ or ‘Lama1+ pial cell’ or ‘Raldh2+ arachnoid cell’ is helpful to advance the meninges biology field. I think that the authors in this paper are bringing to the forefront the concept that studying meningeal sublayers and their potential functions is important to study.
That said, I have significant concerns about the disconnect between the anatomical location within the meninges of the Pdpn+/Prox1+ layer (which the authors provide data for from in vivo imaging Prox1-GFP mice and mouse and human histology in sections) and the authors’ claim of functional segregation and barrier within the subarachnoid space by the Pdpn/Prox1 layer. In addition to reading the paper, I have read their rebuttal posted on the ALZFORUM, written in response to critiques from knowledgeable scientists that study CNS barriers, CSF production and flow and the meninges. I support the content of those original critiques by my colleagues. Here I provide my critiques of the paper, which are similar to my colleagues, but I am taking a different tact. Also, I outline my concerns about the data and cited literature in their rebuttal document that I do not believe lend support to their original claims.
From reading their work, I have identified five characteristics that define the location of the Prox1/Pdpn+ layer in mouse and human meninges. These are:
1) Pdpn/Prox1+ forms a layer of cells that appears to be continuous by transmitted light or fluorescent microscopy imaging (in vivo Prox1-GFP: (Fig. 1A-C); Fig. 2B – histology (DAB or immunofluorescence for Pdpn or Prox1-GFP or antibody): Fig. 3A-D, Fig. 4C, E, Fig. 5B-E.
2) The Pdpn/Prox1+ layer is above all leptomeningeal vasculature (Fig. 1A, 1C; Fig. 3; Fig. 5). This vasculature is barrier blood vessels that sit within the subarachnoid space (SAS) or immediately on top of the pial cell layer (notably in mice that have very small SAS) and occasionally below the pial layer, most notably in species with thick meninges like human5 and pig6.
3) The Prox1+ positive cells (by GFP or antibody labeling) are above the arachnoid trabecula, as shown in Fig. 3A, C, Supp. Figure 3C, Supp. Figure 4. Pdpn+ cells appear to also be above the trabecula in human meninges, however Pdpn has broader expression than Prox1 by antibody labeling.
4) Prox1/Pdpn+ cells are below the E-cadherin+/Claudin11+ arachnoid barrier cells. There does not appear to be any intervening cell layer based on DAPI nuclei (Fig. 4E), nuclear counterstain (Fig. 4C, D).
5) The Prox1-GFP co-localizes with CRABP2 (Fig. 3D) and Pdpn+ labeling is in very similar cell layers as CRABP2, (Fig. 3A, B)
Based on these 5 characteristics and consideration of published work on meninges cellular organization and molecular markers, the Pdpn/Prox1 is a sublayer of the arachnoid layer, most likely the arachnoid cell layer immediately under the arachnoid barrier cell layer. This layer is referred to by a few different names in electron microscopy (EM) studies (Fig. 1), ‘arachnoid’7, ‘inner layer arachnoid membrane’8, ‘arachnoid reticular cells’6 and ‘external arachnoid’9 but shares all 5 characteristics with the Pdpn+/Prox1+: 1) it is below the E-cadherin+/Cldn11+ arachnoid barrier layer (all citations listed above), 2) it is a continuous layer of cells, of note connected by gap junctions and desmosomes7,8,10, 3) it is above the arachnoid trabecula (all citations listed above), 4) it is above all leptomeningeal vasculature (all citations listed above), 5) this arachnoid cell layer expresses CRABP2, based on our own work on CRABP2 and E-cadherin2 and likely expresses Raldh2, expressed by a subset of arachnoid cells based on our own work using scRNAseq and validation of expression in mouse fetal, postnatal and adult meninges. Of note, the authors refer to the Pdpn/Prox1+ layer as ‘subarachnoid lymphatic-like membrane’ or SLYM. Based on my evaluation, this layer is not ‘subarachnoid’ but is part of the arachnoid.
In reviewing EM images of the meninges from multiple species, there is no major physical space between the arachnoid barrier layer and this adjacent, inner arachnoid layer that shares all known features of the Pdpn/Prox1+ layer (Fig. 1). Also, there is evidence of connections between this layer and adjacent arachnoid barrier cell layer by hemi-desmosomes7,8,10, supporting that these layers are in close contact and argues against tissue fixation artifact causing collapse of the layers. If the Pdpn/Prox1 layer is this inner arachnoid layer, and all points above support this, it is difficult to conceive how injected microdots could accumulate above this layer to the extent reported in Figure 2 of their paper. Nor does it seem possible that the authors could selectively inject between this layer and the arachnoid barrier layer with a 35 gauge needle (135 microns outer diameter). A more logical interpretation is that the injections are on top of the arachnoid barrier layer (sub dura). This barrier layer is well known to keep out tracers of this size and smaller.
The authors argue in in their rebuttal posted on ALZFORUM that lack of overlap between red microdots and the second harmonic generation (SHG) imaging signal that corresponds to thick collagen fibrils in the dura suggests that they are injected into a supposed ‘upper subarachnoid space’. They further show images of likely subdural injections, leading to co-mingling of the microdots signal and the collagen fibril picked up by SHG imaging. These data were not in their original paper and these images do not have the Prox1-GFP signal or the blue microdots injected into the cisterna magna so full interpretation is challenging. Further, collagen fibrils detected by SHG are not cells and signal intensity depends on how concentrated that collagen fibrils are in a region, the dura by far has the most concentrated collagen fibrils in the meninges. Is that collagen distributed evenly across the dura? Very likely not. For example, in human, pig and rodent the dural border cells (bottom layer of dura) and adjacent arachnoid barrier cell layer have comparatively little collagen fibrils6–8,10. Therefore, I strongly suspect the SHG imaging is picking up dense collagen fibrils of the periosteal dura layer that is normally adjacent to the calvarial bone (removed in the craniotomy). This would put the dura SHG signal at least two cell layers (e.g., collagen poor dural border cell and arachnoid barrier cells) from the Prox1+ layer signal. This is supported by that fact that collagen fibrils are also found in the subarachnoid space as part of the trabecula and in the pia layer, but that signal (expected to be weaker as it is much less collagen fibrils vs dura) is not detected in their SHG imaging. This may also explain the varying size gaps that they point out in their rebuttal, it is not an empty space but instead is two cell layers that are not labeled in their in vivo imaging. The very large space may be caused by physical separation of the dura (which will be attached to the bone on adjacent sides of the craniotomy) from the underlying leptomeninges, but that is just speculation on my part and requires further investigation.
With these limitations in mind, I am skeptical of how useful SHG imaging is on its own for detection and resolution of spaces within the mouse meninges. At best it is an approximate landmark, but I would not rely on it to make conclusions as the authors have done to a fair extent in Fig 1 and 2 of their paper and in the rebuttal. At this point, I remain of the opinion that their barrier data in Figure 2 is simply showing the functionality of the arachnoid barrier layer. A potentially better, higher resolution approach would be to do these injections of microdots with HRP, perform in vivo imaging, collect the mice after for HRP staining and perform electron microscopy so that the location of the injected HRP can be correlated with the various meningeal layers at high cellular resolution. Alternatively, they could cross Prox1-GFP mice with Cdh1-mCFP (https://www.jax.org/strain/016933) mice for simultaneous in vivo imaging of the Prox1/Pdpn+ layer and the E-cadherin+ arachnoid barrier layer.
What other previously described layers of the meninges could be the Pdpn/Prox1+ layer? The authors in their analysis already ruled out that it is not part of the dura layers nor is it the arachnoid barrier layer, instead I conclude that the Pdpn/Prox1+ layer is just below the arachnoid barrier layer. The remaining potential cell layers are arachnoid cells that are in the subarachnoid space arranged around collagen fibrils to form the trabecula or the pia, a layer of fibroblasts with processes that are aligned with the brain surface. Pdpn/Prox1+ cells meet some characteristics of these layers but not all.
1) arachnoid cells in SAS – are CRABP2+ (in humans, our work2 and their data in Fig. 3) and are below the arachnoid barrier cell layer. Arachnoid cells in the SAS do not form a traditional, flat continuous layer and the data in paper puts Pdpn/Prox1+ above the trabeculae. The arachnoid cells and trabecula surround or are adjacent to blood vasculature but do not lie over the vasculature as a layer, a charteristic of the Pdpn/Prox1 cells. These cells fit only 2 of the 5 characteristics of Pdpn/Prox1 layer.
2) pial cells at brain surface – Pial cells form a layer that is generally thought of as continuous (though some papers report gaps in adult mouse EM8,11) and there is evidence from human5,12,13 and rat9 studies that it forms a subpial space that contains immune cells and may be restrictive to flow from perivascular spaces into the SAS, though the molecular size restriction is not well established. Pial cells are below the arachnoid barrier layer. However pial cells are CRABP2- in human and mouse (our work2), are below most blood vessels and below the trabeculae. Studies in human5 and rat show that the bottom part of the trabeculae attach to the pial layer, in human this is via desmosomes. These cells fit 2 of the 5 characteristics of Pdpn/Prox1+ layer.
The authors in their rebuttal point to two HRP tracer studies in rat from Krisch et al that detail spaces can be detected after a 5 min post-HRP injection into the meninges 2mm lateral to the sagittal suture or between the olfactory bulbs as well as HRP injection into ventricles. These are interesting studies and take advantage of the relatively larger meninges in rat to probe spaces that are not normally studied. However, the relatively short time frame of circulation (5 min) in Kirsch et al. vs this paper (30min+) limits comparisons. Kirsch et al. 1983 discuss in detail that the compartmentalization of HRP in their studies is a function of the short time frame of collection post-injection. The authors point to portions of the Krisch et al., 1983 paper that shows HRP accumulating in a space between what Krisch and colleagues refer to as the ‘pial space’, bordered by the ‘inner arachnoid layer’ and the ‘outer pial layer’. In their rebuttal the authors suggest that this space may be analogous to the ‘upper and lower SAS’ that they report in their paper, possibly this ‘lower arachnoid’ is the Pdpn/Prox1+ layer. However, that is not supported by the authors own data on the location of the Pdpn/Prox1+ which I detail above. Krisch et al put this lower arachnoid layer at the bottom of the trabecula and in some of Kirsch et al images (Fig. 1) it’s below most of the blood vessels, though not all. These studies from Krisch et al are certainly interesting and along with other literature support multiple pathways of fluid movement in the meninges however I don’t believe these papers provide support for the Pdpn/Prox1+ layer segregating the SAS into two separate compartments or is support for any barrier properties of the Pdpn/Prox1+ layer.
The final point I’d like to make is that transmission EM studies of rat and dog meninges show that by far the biggest ‘empty space’ in the leptomeninges is the SAS and there is no apparent bifurcation or separation. The CSF tracer studies in Fig. 2 and Fig. S1 (3kDa dextran injected into the cisterna magna) fill a big space that looks like the SAS by size and location of the vasculature. However, based on their data in Fig. 2A they suggest that the outer and inner SAS are essentially equal size. The physical bifurcation of the SAS into two similarly sized subspaces has not been reported, despite several transmission electron microscopy imaging studies, like those done on the meninges in rat14 and dog15 or more recent in vivo imaging modalities in human like optical coherence tomography16 (Fig. 2). I recognize that new imaging modalities and new markers are helpful to reveal structures than previously were not known or well recognized. However, in this case, I do not believe the Prox1/Pdpn+ or any layer of the meninges bifurcates the SAS as claimed in the paper by Møllgård and colleagues for all the reasons stated above. I also am, at this point, very skeptical that the Prox1/Pdpn+ layer has barrier properties on its own. Clearly more experiments (some of which I described above) are needed to determine which layers (if any) have size restrictive barrier properties besides the arachnoid barrier layer or the extent to which meningeal sublayers of the pia, arachnoid or dura regulate immune cell movement or accumulation, either via a physical barrier or via cell signaling. Loss of function or ablation studies targeting specific meningeal fibroblast populations in adult mice will be important to address this gap in knowledge.
Regarding the claim by Møllgård and colleagues that the Pdpn/Prox1+ layer constitutes a 4th layer of the meninges, I personally don’t believe that any distinct meningeal cell layer (of which there is ~7) needs a specialized name like the one proposed by the authors for the Pdpn/Prox1+ [likely] arachnoid layer. Adopting terms to denote specialized sublayers of the main three, in my opinion, is better for the meninges biology field and important for those outside the field who are not experts but reading and trying to understand new discoveries. Using sublayers provides location/anatomical reference, relates it to developmental origins and allows us to understand proximity to other meningeal sublayers and non-fibroblast cell types (immune cells) and structures (blood and lymphatic vasculature) with the meninges. Also, it helps us connect modern work and advanced techniques to the long standing and rich literature of observational studies of the meninges using light and electron microscopy. At the very least, it would be good to have accepted nomenclature for the various layers that can be applied going forward as new discoveries are made. For example, using one or more molecular markers that distinguish that sublayer from others combined with location (pia, arachnoid, dura). Given the numerous different names for all the sublayers that have accumulated over the years (Fig. 1), this is a long-standing problem that needs to be resolved.
References
There are no Separate Functional Compartments within the Subarachnoid Space
Quite recently a report in Science suggested that there might be functional compartments within the subarachnoid space separated by a mesothelium (1). Unfortunately, this hypothesis solely was based on immunocytochemical evidences at the light microscopic level.
In fact the layering of the meninges in mammals including man is well known in much detail (2, 3, 4, 5, 6, 7, 8, 9). The outermost layer, the dura mater is separated by the subdural neurothelium from the underlying arachnoid. Below that, the subarachnoidal space, a liquid space filled with cerebrospinal fluid, separates the arachnoid from the pia mater. In contrast to earlier assumptions, there is no subdural space in healthy individuals (2, 5, 8, 9). However, such space is easily generated in artificial situations, leaving a number of neurothelial cells attached to the arachnoid. The subdural neurothelium is described best in (2) and was termed border cells or arachnoid barrier cells by other authors. All descriptions of the meningeal layers at the electron microscopic level identify the subarachnoid space as a single compartment.
In contrast, Møllgård et al claim that there were different functional compartments within the subarachnoid space, separated by a mesothelium. This misleading conclusion apparently is based on two distinct errors: (i) The resolution of the immunohistochemical images is much too low to differentiate the meningeal layers in detail. (ii) The authors apparently are not precisely aware of the composition of the meninges, as supported by the missing references in the report.
Due to the low resolution of their images the authors could not precisely differentiate the individual meningeal layers. So, they misinterpreted the well known subdural neurothelium as a novel “membrane” and coined the term “SLYM” for it. Most likely, however, the “SLYM” is just a novel staining of the classical subdural neurothelium. The low resolution also precludes the realization that their “SLYM” is located outside the arachnoidea. This fact is evident after closer inspection of Figures 3B and 3C, where the staining is most intense at the outer side of the arachnoideal membrane.
Hopefully this letter avoids confusion in the scientific community, concerning the meningeal layers of the central nervous system.
Julian Rieck1, Rüdiger W. Veh1*
1Institut für Zell- und Neurobiologie, Centrum 2, Charité - Universitätsmedizin Berlin, Charitéplatz 1, D-10117 Berlin, Germany.
*Corresponding author.
Email: [email protected]
REFERENCES AND NOTES