CSF transport at the brain–meningeal border: effects on neurological health and disease

Summary

Groundbreaking studies uncovered specialized structures that allow a continuous exchange of cerebrospinal fluid between different anatomical compartments at the brain–meningeal border, challenging the conventional concepts of molecular transport within the brain. Experimental findings have shined light over the conduits and cellular structures governing the transport of cerebrospinal fluid and immune cells between the brain parenchyma (via the glymphatic system), the subarachnoid space—enclosed by the meningeal pia and arachnoid layers—and the outmost meningeal dural layer and calvaria, via the so-called arachnoid cuff exit points. Studies performed in both rodent models and human patients unraveled new mechanisms of brain glymphatic molecular transport, meningeal lymphatic vascular drainage, and immune surveillance at the brain-draining skull bone marrow and cervical lymph nodes. Pathological alterations at the brain–meningeal border have been implicated in disorders of diverse etiologies, ranging from acute sterile traumatic brain injury all the way to chronic-like neurodegenerative diseases like Alzheimer’s.

Introduction

The brain is “immersed” in the cerebrospinal fluid (CSF), which fills the brain ventricles and subarachnoid space (SAS). The molecular composition of the CSF is largely influenced by the highly selective (and controlled) transport of blood-derived water, ions, and molecules at the blood–CSF and the blood–brain barriers, as well as the de novo secretion of molecules by the ventricular choroid plexus epithelial and ependymal cells. These processes lead to the generation of “fresh” CSF that travels from the lateral ventricle into the third, and from there into the fourth ventricle, via the cerebral aqueduct, before reaching the SAS.¹ Amongst many crucial properties of the CSF, which have been discussed elsewhere by experts in the field,^2–4 it preserves the brain’s physical integrity and homeostasis, serving as a highway for molecular transport and exchange between different neural compartments. After being produced at the choroid plexuses, the CSF might suffer changes as it flows through the ventricles—and its molecular composition is shaped by different cells present at the brain parenchyma, perivascular spaces, and meninges—on its way to the SAS.^1,5–12 Yet, despite decades of research, the field is still trying to figure out how all these different anatomical compartments, structures, and cells act jointly to promote CSF production, transport, and clearance from the brain.

A recently published body of work has reshaped our understanding about the architecture of the brain’s meningeal layers and SAS, and unraveled previously unappreciated physical routes of communication between different anatomical compartments that have access to CSF molecular content. In this review, we will discuss new experimental evidence obtained from rodents and primates, including humans, that is changing our view regarding the transport of molecules from the CSF-filled SAS into the brain (and back) via the glymphatic system, or all the way into the outmost meningeal dural layer and skull bone marrow (SBM). As we will relay throughout different sections, these new findings regarding CSF transport and clearance via the glymphatic–lymphatic systems, and about the neuroimmune interactions taking place at the meninges and SBM, might fundamentally influence our understandings of neural physiology and pathology.

The brain–meningeal border

The meninges, long viewed as a passive barrier that protects the brain and spinal cord, is now recognized as multifunctional and complex structures that play active roles in immune regulation, CSF dynamics, maintenance of brain homeostasis, neuroinflammation, and the development of neuropathology. Classically, in both rodents and humans, the meninges have been shown to consist of the inner pia mater, intermediate arachnoid mater (the arachnoid and pia together form the leptomeninges), and dura mater, all of which are fibroblast-rich connective tissues.^13–15 The highly vascularized inner pia and the avascular tight arachnoid layer, with barrier-like properties, are connected by trabeculae and delimitate the SAS filled with flowing CSF (figure 1). Inside the cranium, the dura mater is further subdivided into the periosteal layer, which is in direct physically contact with the skull bone, and the inner meningeal layer (figure 1). Between the dural periosteal and meningeal layers one can find a panoply of cell types and structures including the venous sinuses, resident immune cells, and a network of functional lymphatic vessels (figure 1).^{7,8,10,13,14,16} However, recent studies involving rodent models have revealed novel cellular and structural features of the meninges that have reshaped our understanding about CSF transport across the different meningeal layers and uptake by cells and lymphatic vessels at the outmost dural layer (figure 1). In fact, recent advances in cell lineage-tracing mouse models, coupled with state-of-the-art whole tissue clearing and imaging, intravital imaging techniques, and single-cell transcriptomics, have identified specific cell populations and specialized structures within the pia, arachnoid, and dura mater. These different cellular players, namely fibroblasts and macrophages, have complementary roles in both CSF transport and immune surveillance.^{6,8–10,15,17–24}

Figure 1: Molecular transport and cell migration across the brain–meningeal border.

The brain meninges are composed of three layers—namely, the dura mater, arachnoid mater, and pia mater. The arachnoid mater and pia mater enclose the CSF-filled subarachnoid space. Studies have proposed the existence of a permeable layer, known as the subarachnoid lymphatic-like membrane (SLYM), which further divides the subarachnoid space into inner and outer compartments. Bridging veins with sparce smooth muscle cells arising from brain cross the pia mater and arachnoid mater to merge with the dural sinuses ensheathed by smoother muscle cells. At the ACE points (purple dashed circle) the endothelial cells of the bridging vein down-regulate the expression of the tight junction protein CLDN5 (green solid lines). (1) Brain parenchymal cells are a source of molecules (orange hexagons), such as growth factors, inflammatory mediators, or metabolic byproducts, which might cross the pia mater and reach the CSF in the subarachnoid space. These brain-derived molecules can then cross the arachnoid layer at the ACE points along the bridging veins, and either be phagocytosed by innate immune cells in the dura mater or be drained by dural lymphatic vessels (orange arrows). (2) Continuous and efficient drainage by dural lymphatic vessels is needed for optimal brain myelination and neural activity (green arrow and green dashed circles). (3) Adaptive and innate immune cells in the dura mater can cross the ACE points to reach the CSF in the subarachnoid space and the perivascular space (brown arrow). (4) Molecules secreted by dural and leptomeningeal cells (yellow triangles) can follow the same path and ultimately signal to cells in the perivascular space and brain parenchyma (yellow arrows). ACE=arachnoid cuff exit. CLDN5=claudin 5. CSF=cerebrospinal fluid. SAS=subarachnoid space. SLYM=subarachnoid lymphatic-like membrane. SMC=smooth muscle cell.

The dura and leptomeninges are home to resident immune cells and act as key hubs for immune surveillance, and sensing of inflammatory molecules, infectious agents, or injury to the brain.^8,12 Leptomeningeal and brain perivascular macrophages in particular have recently been shown to regulate arterial pulsation, and, by doing so, modulate CSF molecular transport via the glymphatic system (a topic that will be developed further ahead in a dedicated section).⁷ Depletion of these border-associated macrophages in healthy adult mice resulted in accumulation of extracellular matrix proteins, leading to arterial stiffness (a process observed in aging) and impairment of brain CSF influx. Conversely, enhancing colony-stimulating factor 1 signaling to brain/leptomeningeal innate immune cells in aged mice resulted in a recovery of glymphatic influx of CSF.⁷ A similar experimental approach reduced toxic amyloid beta (Aβ) accumulation in transgenic mice, directly linking deleterious immune cell activation at the brain–meningeal border to AD-like brain amyloid buildup.⁷

Recent compelling experimental data also challenged the classical view about the architecture of the meningeal layers. Studies performed in mice and humans have proposed the existence of a previously unrecognized meningeal layer, called subarachnoid lymphatic-like membrane (SLYM).^17,18 The SLYM has been characterized as a monolayer of cells expressing prospero-related homeobox-1 (PROX1) and podoplanin, interspersed with loosely organized collagen fibers, diving the SAS in inner and outer compartments (figure 1). Due to the lack of the tight junction protein claudin-11 expression, the SLYM has been proposed to represent a relatively permeable mesothelial layer that spans the SAS in the meninges above the human cerebral cortex.^17,18 Furthermore, functional magnetic resonance imaging (MRI) studies provided evidence for the existence of such a semipermeable membrane surrounding human pial arteries, facilitating CSF influx along the periarterial space while preventing unspecific mixing in the greater SAS.²⁵ While these studies suggest that the SLYM is a distinct fibroblast-rich layer, other studies have argued against it. Alternative studies claim that the SLYM is an artifact of human (or murine) tissue processing and imaging or result from transient alterations in CSF flow during in vivo experiments involving mouse models.^15,24,26,27 The actual existence of the SLYM and its involvement in the process of CSF transport at the SAS in humans are topics of ongoing investigation.^17,18,26,27

Newly discovered specialized structures found throughout the arachnoid layer, the arachnoid cuff exit (ACE) points, serve as routes for direct molecular and cellular exchange between the SAS and the dura.^6,8,21 The ACE points represent unsealed discontinuities in the barrier formed by arachnoid fibroblasts around the bridging veins that permit the passage of fluids, molecules, and cells (figure 1).¹⁰ At the ACE points, venous endothelial cells start expressing the tight-junction protein claudin-5 and also express high levels of laminin, which mediates the adhesion and trafficking of myeloid cells, especially in neuroinflammatory conditions.¹⁰ In fact, an experimental autoimmune encephalitis mouse model of multiple sclerosis (MS; a canonical white matter autoimmune disease) showed exacerbated numbers of T cells and neutrophils (and other myeloid cells) at ACE points and in the leptomeninges, which suggests that these might represent avenues of pathogenic immune cell migration from the dura into the SAS and brain in autoimmune diseases.¹⁰ The presence of a SAS-to-dura CSF pathway along the bridging veins was also observed in humans. Using 3 Tesla real time-inversion recovery MRI scans of gadolinium-based contrast agent enhancement coupled to sagittal anatomical scans in 10 healthy human participants (3 female and 7 male patients, aged between 31 and 71 years), the authors were able to detect gadolinium exchange between the SAS CSF and the bridging veins at the leptomeninges, and observed what appeared to be structures similar to the mice ACE points, which also permitted the transport of CSF into the dura.¹⁰ We envision that research on the roles of the ACE points at the human brain–meningeal border will continue, as the field will try to understand if they share similar anatomical, molecular, and functional characteristics to the ACE points seen in mice. It will be important to develop appropriate imaging tools aimed at evaluating to what extent these structures contribute to the clearance of CSF from the SAS into the dura, or the movement of cells and molecules from the dura into the SAS (and brain parenchyma) in humans. Altogether, emerging evidence suggests that the transport CSF at the brain–meningeal border is more organized than previously understood.

The next section will be focused on novel findings concerning the transport of brain-derived antigens into the meningeal dura and SBM, and its involvement in neuroimmune surveillance in health and disease.

The meningeal dura–skull bone marrow interface

Complementing recent advances in immune interactions between the meninges and the brain came the discovery of vascular channels directly connecting the bone marrow in the skull to the underlying dura mater.^{21,23,28–30} The SBM represents a hematopoietic reservoir at the brain’s borders. The channels present within the SBM and their proximity to the brain enables a specialized role in neuroimmune surveillance and response (figure 2).^23,30 First identified in mice, the presence of osseous channels was confirmed in humans through micro-computed tomography (micro-CT) imaging of 3 patient craniectomy samples following decompression surgery and 3D tissue imaging of 8 postmortem human skull and dura. These studies revealed channels typically ranging from 40–150 μm in diameter, present across frontal, parietal, and temporal bones.^23,30 Functional evidence in mice shows that these channels provide a bi-directional route for cellular and fluid exchange between the SBM and the brain.^6,21,23 Immune cells generated within the SBM traverse these channels to reach the dura mater under homeostasis (figure 2).^21,23,29 In parallel, CSF from the SAS flows into the SBM, delivering brain-derived cues that shape local immune responses (figure 2).^6,19 During pathology, cells of the SBM can infiltrate the leptomeninges and brain parenchyma via the ACE points.^10,21,23,29 This unique anatomical setup enables SBM to act as a sentinel, sampling antigens and inflammatory signals from the brain via CSF and responding dynamically with bespoke immune activity. Structurally, the channels in both mouse and human are composed of a perivascular space, with an inner vascular lining and an outer bone surface, acting as highways for immune cell trafficking and passage for CSF solutes. The channels begin in the sinusoidal vasculature of the bone marrow, and typically terminate in the dura mater.³⁰ Because of the ACE points identified in mice, cells exiting SBM channels in the dura mater can subsequently transit to underlying SAS CSF and meningeal layers, ultimately accessing the brain.¹⁰

Figure 2: Molecular and cellular transport to and from the SBM.

(A) Bone marrow is present throughout the skull and ossified vascular channels connect this bone marrow with the underlying dura mater. (1) Under homeostasis, the skull bone marrow contains HSCs that generate innate and adaptive immune cells that can traffic to the dura mater using these ossified vascular channels (brown arrows). (2) These ossified vascular channels also permit continuous surveillance of brain-derived molecules, by way of ACE points that permit CSF access to the dura mater (orange arrows). (3) Blood-born molecules can reach the bone marrow and signal to HSCs (blue arrow). (B) During infections or sterile brain injuries such as stroke, pathogens and brain-derived molecules can access the skull bone marrow through ossified channels (orange arrows), resulting in HSC expansion and the generation of innate immune cells. These innate immune cells can use the ossified channels to traffic into the dura mater, underlying leptomeninges, and brain parenchyma to contribute to inflammatory responses (brown arrows). (C) In neurodegenerative diseases, inflammatory responses triggered by brain-derived misfolded proteins (orange arrows) are present in the skull bone marrow and innate and adaptive immune cells can be generated here and potentially traffic into underlying meningeal and brain tissues by ossified channels (brown arrows), contributing to disease responses. (D) Peripheral tumor cells can utilize ossified channels to metastasize from the blood into the leptomeninges tissues (brown arrows). Brain tumor derived molecules can also access the skull bone marrow by ossified channels (orange arrows) and generate innate and immune cells that can traffic into the tumors to contribute to anti-tumor immunity (brown arrows). ACE=arachnoid cuff exit. CSF=cerebrospinal fluid. HSCs= hematopoietic stem cells. SAS=subarachnoid space. SSS=superior sagittal sinus.

Recent studies underscore the involvement of the SBM in diverse neurological diseases. Unlike most bone marrow niches that attenuate during aging, both murine and human SBM display resistance to age-related decline. Murine skulls contain a greater proportion of haemopoietic stem cells with age, with micro-CT imaging of human aneurysm patients revealing an expansion in SBM area indicative of comparable changes in human SBM. These findings may have implications for age-related neurodegenerative decline as myeloid cells are deeply implicated in neurodegeneration and peripheral immune cells can be recruited into disease brains, meninges, and CSF.^{11,23,28,31,32} The skull may represent an active player in this trafficking, and mouse models of stroke and MS demonstrate preferential immune cell recruitment of myeloid cells from the skull compared to other bone marrow sites.^21,23 Positron emission tomography (PET) imaging for translocator protein (TSPO) ligands can act as a proxy for inflammatory activation in humans, as TSPO is upregulated during neuroinflammation. PET imaging of MS patients reveals distinct patterns of inflammation within the SBM, suggesting possible involvement of the SBM in MS disease processes.³⁰ Similar patterns of TSPO ligands are also present in AD patients, demonstrating inflammatory activation of the SBM and indicating a potential role for the SBM in AD.³⁰

The role of SBM in neuroinflammation extends beyond neurodegenerative diseases. In bacterial meningitis, pathogens within the CSF can enter the SBM via these channels, triggering hematopoietic responses that amplify local inflammation (figure 2).^6,19 Methods to encourage SBM trafficking could represent an attractive intervention to augment bactericidal activity. Additionally, mouse models of stroke demonstrate preferential recruitment of neutrophils from the skull compared to peripheral bone marrow.²³ The SBM in humans suffering a stroke also display distinct patterns of TSPO activation, demonstrating an immunological response in this niche that may underly the delivery of neutrophils.³⁰ Interestingly, SBM-derived myeloid cells display phenotypic differences from those present in other bone marrow sources that may have clinical impacts in neurological function.^21,30 For example, those originating from peripheral bone marrow display elevated proinflammatory phenotypes.^21,30 The majority of studies have focused on the role of innate immune cells, particularly monocytes and neutrophils, which use these channels to egress from the SBM to the dura mater. However, recent findings also demonstrate that B cell progenitors and mature B cells populate the dura and can be derived from the SBM via these channels.²⁹ The dura provides a distinct microenvironment for B cell progenitor residence and development during homeostasis and inflammation. Dural B cells contribute to immune tolerance by encountering brain-derived antigens, a process that helps maintain the brain’s immune privilege.^29,33 SBM also shows implications for brain tumors. Using patient-matched human tissue, immune cell clustering in SBM regions adjacent to glioblastoma multiforme (GBM) lesions in humans was observed, likely driven by tumor antigens accessing the calvaria through CSF.³⁴ These antigens elicit adaptive immune responses, with SBM-derived cells contributing to anti-tumoral activity (figure 2). Despite the challenges of studying the SBM in humans, T cell receptor sequencing—essentially an endogenous cellular barcode—revealed shared T cell clones between intracerebral tumors and adjacent SBM in 3 patients diagnosed with isocitrate dehydrogenase-wildtype glioblastoma, providing indirect evidence of trafficking through these channels.³⁴ Antigen presenting cell-like neutrophils and professional antigen presenting dendritic cells were also identified in 17 patients diagnosed with isocitrate dehydrogenase-wildtype glioblastoma, and mouse models of glioblastoma, demonstrating their preferential recruitment from local SBM to promote antitumor responses (figure 2).³⁵ Therapeutic strategies targeting SBM, such as the C–X–C chemokine receptor-4 inhibitor AMD3100, enables egress from SBM resulting in anti-tumoral immunity in mouse models, underscoring the potential for SBM-focused interventions in GBM and other cancers.^21,35 Conversely, SBM channels also provide a route for metastatic infiltration of breast cancer and acute lymphoblastic leukemia, which both have a propensity to metastasize to the leptomeninges, significantly worsening patient outcomes.^28,36 Using mouse models with human acute lymphoblastic leukemia, these cells fail to breach an intact blood–brain barrier, instead, using SBM channels to infiltrate the brain borders (figure 2).²⁸ Similarly, mouse models engrafted with human bone-trophic breast cancer cells demonstrate metastasis to the SBM, subsequently invading the leptomeninges through the vascular channels.³⁶ Clinical imaging of these sites may represent a target for early detection of leptomeningeal metastasis and a target for intervention.

Surgical interventions involving the skull present another consideration for the SBM role in neurological health. Procedures such as decompressive craniectomy disrupt the calvaria and may sever the connections between SBM and the dura, impairing immune cell trafficking and compromising the SBM niche. Frozen autogenous skull cranioplasties can result in non-viable cell recovery for re-transplanted bone and disruptions to SBM could hinder acute recovery and long-term neurological outcomes.³⁷ For instance, in a GBM resection, the removal of overlying calvarial bone could alter the immunological environment surrounding the tumor, diminishing SBM contributions to immune surveillance and response and potentially worsening patient outcomes. Similarly, craniectomy in MS patients may alter the local immune environment, as an experimental autoimmune encephalitis mouse model of MS demonstrated that a proportion of both innate and adaptive immune populations in the CNS are SBM-derived.²¹ The consequences of this are unclear, but SBM-derived monocytes do possess an immunoregulatory phenotype compared to their peripheral counterparts.²¹ Therefore, careful consideration must be given prior to craniotomy, especially in immunologically challenged patients, to ensure best patient outcomes.

CSF drainage through lymphatic vasculature

Evidence collected from rodent models shows that SAS CSF can exit the skull through the porous cribriform plate, a process thought to be heightened in pathological conditions due to an apparently more permissive meningeal tissue arachnoid layer along the axons of the olfactory nerves.^38–42 However, albeit the reported transport of CSF-injected fluorescent molecules into the nasal mucosa in rodents, recent MRI data has shown little to no transport of CSF tracers (contrast agents) from the SAS to the nasal mucosa via the cribriform plate route in humans.^43,44 Contrarily, different independent imaging studies involving human subjects did observe transport of tracers from CSF-filled compartments into the parasagittal dura, further supporting the existence of direct anatomical pathways mediating the CSF molecular exchange between the SAS and the remotest meningeal layer.^44–48 This same phenomenon has been repeatedly reported in mice, suggesting that this anatomical route of rapid CSF transport into the dura via the aforementioned ACE points is conserved between mammalian species.^6,8,10,22,31 Despite the current uncertainty regarding the exact contributions of the nasal (across the cribriform plate) or ACE-to-dura routes to CSF molecular clearance, elegant experimental evidence, generated with mouse models and state-of-the-art imaging techniques, clearly show that molecules in the SAS CSF can reach the bona fide lymphatic vascular networks found at the extracranial nasopharyngeal mucosa or the intracranial meningeal dura mater (figure 3).^{9,16,22,31,41,42,47,49,50} In mice and humans, these lymphatic networks continuously drain CSF-derived molecular content directly into the peripheral collecting lymphatic vessels afferent to the neck cervical lymph nodes (CLNs; figure 3).^{9,16,31,42,51} At the CLNs, brain-derived molecules can be either ingested, processed, presented, and recognized by immune cells (for immune tolerance or activation), or continuously drained with the rest of the lymph, until they reach the venous circulatory system via the thoracic duct to be ultimately excreted from the body.^9,31,52–55

Figure 3: Anatomy of the CSF-draining lymphatic vessels.

Scheme showing a lateral view of the murine skull enclosing the meninges and brain (illustration at the center), and depicting the extracranial nasopharyngeal lymphatic vessels, the intracranial meningeal lymphatic vessels, and the superficial and deep CLNs. Lymphatic vessel networks (in green) surrounding blood vessels (arteries in red and veins in blue) in the dorsal and ventral meningeal dura are shown in illustrations to the left and right, respectively. CLNs=cervical lymph nodes. CS=cavernous sinus. ICA=internal carotid artery. IEV=interpterygoid emissary vein. IJV=internal jugular vein. IPS=inferior petrosal sinus. MMA=middle meningeal artery. PSS=petrosquamosal sinus. SS=sigmoid sinus. SSS=superior sagittal sinus. TS=transverse sinus.

The transport of CSF content into the dura, suggests that the recently re-discovered intracranial meningeal dural lymphatic vascular system is well-equipped to encounter and drain brain-derived antigens. An ever-growing body of evidence generated by numerous independent groups has confirmed the presence of functional meningeal lymphatic vessels composed of lymphatic endothelial cells (LECs) co-expressing all the classical lineage-specific markers, including PROX1 and vascular endothelial growth factor receptor-3 (VEGFR3), in rodents and humans alike.^16,48,50 In mice, a fully matured meningeal lymphatic vasculature is only observed at around postnatal day 28.⁴⁹ Both the assembly and maintenance of a functional meningeal lymphatic vasculature depend on continuous VEGFR3 signaling and expression of the mechanosensitive Piezo1 ion channel by PROX1-expressing cells, including LECs.^49,56 In the adult mouse dura, the network of initial lymphatic vessels converges into collecting-like lymphatics containing luminal valves, but devoid of ensheathing smooth muscle cells that can be found at the dorsolateral and ventral dura. At the base of skull, collecting lymphatics exit via the foramina to drain molecular content from the brain directly into the peripheral CLNs (figure 3).^22,47,49,50 Experiments involving mice have shown that the rate of CSF lymphatic drainage into the deep CLNs is modulated by the acute topical application, on the exposed deep CLNs, of either the alpha 1 agonist phenylephrine or the muscle relaxant sodium nitroprusside (in a dose-dependent manner).⁴² However, the mechanism(s) behind this effect remain unknown, and more work will be necessary to unveil the exact roles of adrenergic or nitric oxide signaling pathways in regulating CSF lymphatic drainage into the CLNs. Another interesting study involving mouse models has shown that CSF outflow into the deep or superficial CLNs occurs via distinct mechanisms. Lymphatic drainage of CSF into the deep CLNs happens spontaneously, independently of lymphatic vessel pumping, and it is coupled to fluctuations in intracranial pressure.⁵⁷ Instead, CSF drainage into the superficial CLNs is driven by the pumping of the collecting lymphatic vessels. Of note, impaired drainage into the deep CLNs led to elevated CSF outflow resistance and delayed CSF-to-blood efflux despite the recruitment of the nasopharyngeal lymphatic-to-superficial CLNs’ pathway, which only emerged to mitigate perturbances in CSF lymphatic drainage.⁵⁷

The development of the dural lymphatic vascular system and its capacity to continuously drain CSF is essential for the preservation of appropriate behavioral skills in mice, namely motor and cognitive functions.^{9,20,52,56,58} This might be attributed to the recently described effects of meningeal lymphatic vessel loss on altered brain immune surveillance and activation, and its downstream impact on mature oligodendrocyte survival and deficient axonal (re)myelination (figure 1).⁵⁴ Of note, patients presenting white matter hyperintensities—a study involving 27 males and 27 females, with mean ages of 57 and 59 years, respectively—have shown concomitant reductions in lymphatic-like flow at the parasagittal dura. This decrease, measured by 2D-T2-fluid-attenuated inversion recovery MRI, correlated with lower levels of intrathecally-injected gadodiamide at the deep CLNs, which was measured by neck T1-fat-suppression MRI. Both measurements point to reduced lymphatic drainage of CSF in patients with white matter pathology.⁵⁹ Patients diagnosed with MS (7 males and 26 females; mean age of 35 years) have shown decreased levels of the VEGFR3 lymphangiogenic ligand VEGF-C in the CSF when compared to healthy controls (17 males and 19 females; mean age of 29 years).⁵⁴ The drop in CSF VEGF-C was more prominent soon after relapses, which might be a sign of defective meningeal lymphatic function during active neuroinflammation.⁵⁴ However, more robust clinical studies will be needed to understand whether reduced CSF VEGF-C is in fact linked to poor meningeal lymphatic drainage into the CLNs in patients diagnosed with neuroinflammatory white matter diseases like MS. It will also be important to take advantage of animal models to understand the immune mechanisms through which the meningeal lymphatics promote brain myelination and white matter integrity and understand how this newly described role might ultimately impact neuronal function and cognition.

By mediating CSF outflow and regulating brain immunity, the meningeal lymphatic vascular system is also able to modulate fluid transport within the brain via the glymphatic system.^9,52,60,61 In the next sections, we will be focusing on brain fluid circulation through the glymphatic system and on novel experimental observations corroborating the importance of appropriate molecular transport across the glymphatic–meningeal lymphatic systems for brain health.

The brain’s glymphatic system

The glymphatic system is a brain-wide fluid transport network relying upon fast fluid flow along the perivascular spaces demarked by astrocytic endfeet (figure 4). This system plays a crucial role in maintaining neural homeostasis by facilitating the exchange of CSF and interstitial fluid within the brain parenchyma.^62–65 Complementary studies involving mouse models have contributed to a better understanding of the glymphatic fluid transport process, which can be described in 3 steps (figure 4): (1) CSF from the SAS enters periarterial spaces surrounding cerebral arteries, propelled by arterial pulsatility and slow vasomotion; (2) CSF exchanges with the brain’s interstitial fluid (ISF) via the perivascular (paravascular) spaces, a process facilitated by the water channel aquaporin 4 (AQP4), which is highly expressed at the astrocyte endfeet forming the glia limitans; (3) waste-laden ISF/CSF is cleared along perivenous pathways into cranial and spinal nerves, or into the dura via ACEs, to then be drained by the lymphatic vasculature into the CLNs before reaching the systemic circulation and excretion via the kidneys and liver.^{9,11,16,61,62,65–68}

Figure 4: Molecular transport within the brain’s glymphatic system.

(A) The brain’s glymphatic system consists of PVSs that exist between the basement membrane of arteries and veins and the glia limitans formed by the astrocytic endfeet. Increased periarterial influx of CSF molecules (illustrated by yellow triangles) from the SAS into the parenchyma is observed during sleep, in part due to decreased adrenergic signaling. Perivenous efflux of parenchyma-derived molecules (illustrated by orange hexagons) is also heightened during sleep. Glymphatic influx and efflux are dependent on AQP4 expression by astrocytes. (B) During awake state or fragmented sleep, the glymphatic influx and efflux are reduced, in part due to increased adrenergic signaling. (C) In the aged AD brain, increased deposition of amyloid plaques in the brain, CAA, reduced AQP4 expression at the astrocytic endfeet, and ensuing glial activation, have been linked to an enlargement of the PVS and reduced glymphatic influx/efflux. (D) During the acute phase of ischemic stroke, there is a deleterious increase in glymphatic influx of CSF and enlargement of the PVS. (E) Molecular influx and efflux via the brain’s glymphatics are both diminished after TBI, largely due to increased adrenergic neuronal activity. AQP4=aquaporin 4. BAM=border-associated macrophage. CAA=cerebral amyloid angiopathy. CSF=cerebrospinal fluid. PVS=perivascular space. SAS=subarachnoid space. TBI=traumatic brain injury.

Experiments performed with rodent models have shown that glymphatic flow changes according to body posture and is not constant throughout the day. Glymphatic influx of CSF is most active during deep sleep, when there is reduced noradrenergic signaling and enhanced cerebrovascular pulsatility (figure 3). However, it still remains functional, albeit at lower levels, during wakefulness.^{62–64,69,70} The glymphatic system is regulated by the circadian clock and works efficiently during delta-rich non-rapid eye movement (NREM) sleep, enhancing CSF circulation and clearance of metabolic byproducts (figure 3).^61,62,71 Spontaneous infraslow oscillations of norepinephrine during NREM sleep have been recently identified as the primary driver of glymphatic flow. These oscillations, occurring approximately every ~50 seconds, induce cycles of vasoconstriction and dilation, which propel CSF along the periarterial spaces and facilitate glymphatic clearance.^70,72 On the other hand, chronic stress and sleep fragmentation were linked to a reduction in glymphatic function and cognitive decline in young mice.^72–75

Interestingly, experiments performed in mice have shown that nociceptive agents – including calcitonin gene-related peptide, which is released from the brain cortex during migraine aura (cortical spreading depression) – can utilize CSF glymphatic transport routes to reach trigeminal neurons, where they induce hyperactivity of nociceptive pathways.⁷⁶ This underlined a new role for the glymphatic system as a communication route between the central and peripheral nervous systems.

Experimental evidence obtained using mouse models shows that the glymphatic system plays an important role in the removal of toxic aggregation-prone peptides, thus potentially lowering the risk of cerebral protein aggregate deposition and angiopathy – pathological hallmarks of neurodegenerative diseases including Alzheimer’s disease (AD), Parkinson’s disease (PD) and amyotrophic lateral sclerosis.^{64,68,77–80} In an AD transgenic mouse model, sleep deprivation led to a reduction of glymphatic clearance of brain waste byproducts, which has been linked to accelerated parenchymal accumulation of insoluble Aβ and plaque deposition (figure 4).^9,74 Likewise, impaired glymphatic efflux has been linked to deleterious edema formation after traumatic brain injury (TBI), whereas an abnormal acceleration of periarterial CSF influx in a model of ischemic stroke contributed to unwanted tissue swelling (figure 4).^60,81 Interestingly, alongside impaired glymphatic function after TBI, mice also presented disrupted lymph outflow at the collecting lymphatic vessels afferent to the CLNs, an overall deficit in brain-derived fluid clearance that was attributed, at least in part, to excessive and long-lasting increases in central and peripheral norepinephrine (figure 4).⁶⁰ Pan-adrenergic inhibition improved both glymphatic and lymphatic efflux, eliminated the acute cerebral swelling and enhanced functional recovery, while limiting the histologic signs of neural injury and reactive gliosis.⁶⁰ These data corroborated previous experimental observations highlighting the importance of the constant functional crosstalk between the glymphatic and lymphatic systems and their involvement in the regulation of brain fluid homeostasis.^9,52,61 Insights into the glymphatic system using rodent models of disease have enhanced our core understanding of the anatomy of brain molecular waste removal, fluid flow and created new possibilities for clinical investigations, including early detection, diagnostic strategies, prognosis, and development of therapies for acute and chronic neurological disorders.^68,77 However, despite initially described in rodents, recent MRI studies provide strong evidence for a similar CSF/ISF glymphatic transport system in the human brain.^69,82,83 The next section will be dedicated to studies on glymphatic function and malfunction in humans.

Glymphatic (dys)function in the human brain

The impact of sleep and circadian dysregulation on glymphatic flow has also been studied in humans. Sleep disturbances are prevalent in adults older than 60 years of age, and multiple lines of work document that difficulty in initiating and maintaining a normal sleep pattern contributes to cognitive decline.⁸⁴ Prolonged sleep disturbances and its underlying circadian rhythm dysregulation are often the earliest symptoms in neurodegenerative disorders.^69,82 MRI studies involving injections of gadobutrol into the CSF revealed elevated levels of the contrast agent in the brain parenchyma of sleep deprived patients even 48 hours after injection.^82,85,86 In contrary, patients under NREM sleep showed decreased spectral entropy and strengthened brain pulsations in posterior brain regions, aligning with the hypothesis that sleep improves brain fluid circulation via glymphatics.⁸⁷ It is thus speculated that sleep disturbances increase the risk for neurodegenerative diseases by disrupting glymphatic clearance and accelerating the accumulation of metabolic molecular waste in the brain.^69,82

The “diffusion tensor image analysis along the perivascular space” (DTI-ALPS) MRI technique is proposed as an indirect measure of glymphatic function. Multiple studies using DTI-ALPS have documented impaired glymphatic function in neurodegenerative diseases, sleep disturbances, and other conditions. However, while promising, recent critical reviews have questioned the specificity of DTI-ALPS in representing glymphatic perivenous outflow and suggested that it might instead reflect fluid flow through white matter tracts.^88,89 This underscores the need for further research to validate and refine this MRI approach.

While sleep disturbances have emerged as a critical factor influencing glymphatic function, recent research has also begun to uncover the significant role of the glymphatic system in TBI, highlighting its potential in understanding and mitigating secondary brain damage in this highly prevalent neurological disorder.⁹⁰ Accumulating evidence from human studies using MRI shows that patients with history of repetitive head impacts or TBI have significantly larger perivascular spaces compared to healthy controls indicating glymphatic pathway stagnation.^91,92 However, so far, the preclinical finding that noradrenergic spikes after TBI inhibit lymphatic vascular outflow and contributes to post-traumatic edema have so far not been explored in the clinic. However, several indirect observations suggest that post-TBI noradrenergic surge or sympathetic overactivation can persist for weeks (sometimes even months) in patients leading to an array of neurocardiological symptoms.^93,94 A comprehensive meta-analysis of 15 studies, encompassing 12,721 patients, demonstrated that β-blocker administration after TBI significantly reduces in-hospital mortality, improves functional recovery, and results in few adverse effects.⁹⁵ Likewise, prophylactic treatment with prazosin, an α-1 adrenergic receptor antagonist provides a clinically meaningful efficacy in post-traumatic headache patients.⁹⁶ Moreover, treatment with blood–brain barrier permeable β-adrenergic blockers reduced the risk for dementia, including AD, in a retrospective study of a population of 69,081 individuals with hypertension.⁹⁷

Finally, and as aforementioned, the expression of the AQP4 water channels by brain astrocytes is essential for proper glymphatic function. Glymphatic transport is reduced by 20–70% in mice either constitutively lacking two copies of the murine Aqp4 gene or following pharmacological inhibition of AQP4.^62,65 In humans, genetic polymorphisms in the AQP4 gene may impact water channel function, and links have been established to both sleep disturbances and AD in humans.^98,99 Yet, additional clinical studies are needed to understand if these outcomes can in fact be attributed to altered glymphatic function.

Effects of alterated meningeal lymphatic drainage

In the last decade or so, studies using mostly rodent models have uncovered the pleotropic roles of the meningeal lymphatic vascular system in health and disease. In mice, the meningeal lymphatics suffer drastic morphological changes with aging, which culminate in decreased drainage of CSF-derived molecules into the CLNs (figure 5), a feature that is shared between mice and humans alike.^9,22,100,101 Likewise, different types of insults to the murine brain and its meninges, including trauma, hemorrhage, viral infection, hydrocephalus, GBM, or genetically-induced familial AD-like amyloidosis, craniosynostosis, or Down syndrome resulted in significant impairments in either lymphatic morphology, CSF lymphatic drainage to the CLNs, or both (figure 5).^{20,56,100,102–106} Studies investigating the effects of lymphatic vessel loss-of-function—achieved via long-term loss of VEGF-C/D signaling—have failed to see differences in neuroinflammation or tissue pathology severity in mouse models of experimental autoimmune encephalitis or brain amyloidosis, respectively.^107–109 However, it is important to consider that the employed models are not cell lineage-specific and will deprive all murine brain and meningeal cells of VEGF-C/D signaling. This might result in unexplored off-target and confounding effects that go beyond meningeal lymphatic vessel ablation and might directly perturb brain cells expressing high levels of VEGFR2 and 3, which is the case of brain blood endothelial cells and oligodendrocyte precursor cells, just to name a few.⁵⁴ On the contrary, studies designed to assess the consequences of lymphatic vessel regression at the murine dura, or reduced immune cell egress via lymphatics (mimicking what is observed in old mice) have shown a loss of mature oligodendrocytes and myelin, and a worsening of cognitive function, gliosis, and Aβ plaque pathology.^9,20,52,54 Compromised lymphatic drainage and presentation of brain myelin components with immunosuppressive properties, like the murine regulatory self-peptide myelin basic protein 160–175, have been linked to reduced immune tolerance in the brain and increased neuroinflammation.¹¹⁰ Whether decreased tolerance to Aβ peptides in the CLNs due to defective meningeal lymphatic drainage is contributing to exacerbated brain amyloid pathology is a topic that warrants investigation. It will also be interesting to further explore the link between brain demyelination and exacerbated amyloidosis using mouse models of meningeal lymphatic vessel ablation, as brain myelination status has been recently connected to the severity of Aβ deposition.^54,111

Figure 5: Meningeal lymphatic drainage in neurological disorders.

(A) There is a marked loss and atrophy of initial lymphatic vessels within the dura in models of aging and Alzheimer’s disease. Transport of brain-derived Aβ aggregates across the meninges and into the dura changes in the activation status of adaptive and innate immune cells at the brain–meningeal border (orange arrows). This is accompanied by decreased lymphatic drainage of CSF-derived molecules, including Aβ (dashed orange arrow). Aging is linked with higher recruitment of adaptive immune cells into the dura (brown arrow) and progressive Aβ deposition at the ACE points and around the dural blood and lymphatic vasculature. (B) Multiple sclerosis is characterized by marked demyelination, heightened gliosis and recruitment of encephalitogenic leukocytes into the brain parenchyma and meninges. Myelin sheath content can reach the dura and enter the lymphatic vasculature to be drained into the cervical lymph nodes (orange arrows). Mouse models of multiple sclerosis present increased trafficking of leukocytes into the meningeal dura mater, and from there into lymphatics or the subarachnoid space via ACE points (brown arrows). (C) The formation of tumors in the brain has been linked to abnormal lymphangiogenesis in the meningeal dura, lymphatic drainage of tumor-derived antigens, metastasis of tumor cells into the meninges and draining cervical lymph nodes, and inflammation both at the level of the brain parenchyma and brain–meningeal border. Aβ=amyloid beta. ACE=arachnoid cuff exit. CSF=cerebrospinal fluid.

Contrarily, strategies aimed at rescuing or enhancing meningeal lymphatic drainage, namely by increasing the levels of VEGF-C or the Piezo1 agonist Yoda1, led to increased drainage rate of CSF molecular content into the CLNs, and concomitant improvements in behavioral performance and/or neuropathology (e.g., lower intracranial pressure and edema) in mouse models of aging, stroke, hydrocephalus, craniosynostosis, and Down syndrome.^9,56,100,112 In mouse models of GBM there is an aberrant remodeling of meningeal lymphatic vessels that culminates in reduced drainage capacity by this lymphatic system (figure 5).¹⁰⁵ Fine-tuning meningeal lymphatic function by increasing the levels of VEGF-C in mice with GBM improved the clearance of the brain tumor cells. This was achieved via enhanced drainage of brain tumor-derived antigens, and acquisition of long-lasting anti-tumor immune memory via enhanced priming of T cells in the CLNs.^105,106 Likewise, an approach combining VEGF-C delivery and passive anti-Aβ immunotherapy was more efficient in clearing Aβ plaques from the mouse brain.²⁰

In contrast to the above-mentioned conditions (mostly congenital and chronic-like), lymphatic drainage to the CLNs seems to have a deleterious role during active inflammation and autoimmunity by perpetuating encephalitogenic immune responses (figure 5). This was implied by the better outcomes observed upon surgical resection of the CLNs at time points of exacerbated neuroinflammation and brain leukocyte infiltration in mouse models of focal cerebral ischemia, TBI, or MS.^31,113,114 Altogether, these data highlight the potential therapeutic benefits of modulating meningeal lymphatic function in different neurological disorders for which there are very limited effective treatments.

The dura mater of non-human primates and humans also harbors lymphatic vessels, which, as in rodents, are also found in the vicinity of the dural sinuses and other large caliber blood vessels.^{16,39,115,116} Yet, despite the anatomical similarities between different mammalian species, nothing is known about the timing, ontogeny, and development process of meningeal lymphatics in humans. Recently-developed MRI modalities, involving the injection of contrast agents and different T1 and T2 acquisition sequences, allow the distinction between blood and lymphatic flow at the dura, and have provided the first evidence of perisinusal molecular transport which may reflect lymphatic drainage of CSF into the draining extracranial CLNs in humans.^{39,47,117,118} Of note, employment of such imaging modalities has revealed significant alterations in meningeal lymphatic flow in older subjects, or in patients diagnosed with idiopathic PD or Gorham-Stout disease.^47,117,118 Interestingly, in a case of Gorham-Stout disease, which is an idiopathic condition known as the “vanishing bone disease” and characterized by abnormal lymphangiogenesis at regions of bone loss, there was a prominent volumetric increase in meningeal lymphatic flow signal where the skull bone was nonexistent.⁴⁷ A systematic implementation of such imaging modalities at the clinical setting will be critical to detect putative alterations in meningeal lymphatic vessel morphology and drainage into the CLNs in neurological disorders of diverse etiologies. Likewise, examining the molecular content of the CLNs might inform about changes in brain CSF composition or even the presence/progression of neuropathology. Different Aβ and tau species, as well as neuronal damage and glial activation markers, have been detected in the interstitium of the human CLNs after minimally-invasive ultrasound-guided fine needle aspirations.⁵³ Surprisingly, the concentration of the biomarkers was higher in the CLNs than in the plasma, and, in the case of phosphorylated tau protein at serine 181, the concentrations at the CLNs were negatively correlated with age, whereas the inverse was observed in the plasma.⁵³ This study implies that measuring neurodegenerative disease biomarkers in the plasma and in the CSF-draining CLNs might yield different and perhaps complementary information. Yet, it remains to be seen whether measuring disease biomarkers in the CLNs has any actual diagnostic and/or prognostic value.

Conclusions and future directions

The latest evidence highlighted in this review challenges the idea of an “impenetrable” brain–meningeal border that prevents the exchange of fluid, molecules and cells between the SAS and the dura. In both rodents and humans, the presence of specific anatomical structures at the level of the meningeal arachnoid layer, coined as ACE points, facilitates the rapid transport of CSF molecules with putative immunogenic nature from the SAS directly into meningeal dura. Because the meningeal dura mater is home to varied types of innate and adaptive immune cells recruited from the bloodstream or produced at the SBM, these findings further emphasize how important the molecular crosstalk between these different compartments is when it comes to diseases with strong immune and/or inflammatory components.

Targeting the meningeal lymphatic system and SBM with specific therapies and drug delivery strategies leveraging the direct connection between the CSF, meningeal dura, and SBM could be exploited to recruit reparative immune cells or suppress pathological inflammation in different diseases, without affecting peripheral immunity. For clinicians, appreciating the interconnectedness between the brain, meninges and the SBM may prove crucial for optimizing patient outcomes, particularly in cases involving neurosurgical procedures.

The glymphatic and meningeal lymphatic systems play crucial roles in clearing brain molecular waste, with significant implications for neurological diseases. Optimizing glymphatic–lymphatic function through sleep promotion, circadian regulation, and pharmacological interventions, such as controlled delivery of VEGF-C or adrenergic blockers, might represent relatively non-invasive approaches to enhance brain cleansing, prevent the formation of deleterious immune responses and, ultimately, diminish neuropathology in the face of acute- or chronic-like insults. Integrating strategies to evaluate the functional status of the brain’s glymphatic and lymphatic systems, and of immune cells at the brain–meningeal border, SBM, and draining CLNs will be of outmost importance. In this sense, the development of personalized approaches based on genetic predispositions, offers a promising avenue for improving neurological care.

Key challenges remain when it comes to leveraging CSF transport at the brain–meningeal border for therapeutic purposes. On one hand, putative structural and cellular differences between mammalian species might limit clinical translation. On the other hand, the precise mechanisms driving CSF-to-dura transport at ACE points or immune cell trafficking between the SBM, dura, and CSF are still poorly understood, especially in humans. It will be demanding to develop MRI, or other imaging modalities, with enough resolution to accurately evaluate glymphatic–lymphatic CSF flow and meningeal/SBM immune cell dynamics in real-time and longitudinally in patients. Yet, the wide and systematic implementation of such advanced neuroimaging techniques focused on the brain–meningeal border could pave the way for more effective diagnosis, prevention, and treatment of neurological diseases of different etiologies.

Search strategy and selection criteria

References for this Review were identified by searches of PubMed and Google Scholar, through personal databases of relevant articles in the field, by cross-referencing, and from expert recommendations. Search terms included, but were not limited to, the following: “meninges”, “dura”, “subarachnoid space”, “arachnoid cuff exit”, “calvaria”, “skull bone marrow”, “glymphatic”, “adrenergic”, “meningeal lymphatic”, “cerebrospinal fluid”, “drainage”, “vascular endothelial growth factor C”, “Piezo1”, “aging”, “Alzheimer’s disease”, “brain tumor”, “traumatic brain injury”, “hydrocephalus”, and “craniosynostosis”. Additional references were recommended by the reviewers. From the 118 cited references, 102 were published between 2018 and 2024. There were no language restrictions. The final reference list was generated from peer-reviewed publications only, on the basis of the novelty and quality of each publication, and according to the topics covered in this Review.