Quantifying cerebrospinal fluid dynamics: A review of human neuroimaging contributions to CSF physiology and neurodegenerative disease

Abstract

Cerebrospinal fluid (CSF), predominantly produced in the ventricles and circulating throughout the brain and spinal cord, is a key protective mechanism of the central nervous system (CNS). Physical cushioning, nutrient delivery, metabolic waste, including protein clearance, are key functions of the CSF in humans. CSF volume and flow dynamics regulate intracranial pressure and are fundamental to diagnosing disorders including normal pressure hydrocephalus, intracranial hypotension, CSF leaks, and possibly Alzheimer’s disease (AD). The ability of CSF to clear normal and pathological proteins, such as amyloid-beta (Aβ), tau, alpha synuclein and others, implicates it production, circulation, and composition, in many neuropathologies. Several neuroimaging modalities have been developed to probe CSF fluid dynamics and better relate CSF volume and flow to anatomy and clinical conditions. Approaches include 2-photon microscopic techniques, MRI (tracer-based, gadolinium contrast, endogenous phase-contrast), and dynamic positron emission tomography (PET) using existing approved radiotracers. Here, we discuss CSF flow neuroimaging, from animal models to recent clinical-research advances, summarizing current endeavors to quantify and map CSF flow with implications towards pathophysiology, new biomarkers, and treatments of neurological diseases.

1. Background

The human brain, the organ responsible for centralizing and regulating cognitive, somatic, and key life functions, is, along with the spinal cord, protected through a variety of mechanisms. The macroscopic physical boundary against mechanical insults is the skull and spinal column. Pertinent to our discussion of protection on a chemical scale, the brain, spinal cord, and cerebrospinal fluid (CSF) are together separated from a free communication with the systemic blood circulation by the tight junctions of endothelial cells comprising the blood-brain barrier (BBB). The BBB at the capillary level, regulates ionic and molecular permeability to safeguard the central nervous system (CNS) (Daneman and Prat, 2015). The blood-CSF barrier (BCSFB) is found at the major site of CSF production, the choroid plexus (ChP) of the ventricular system. The BCSFB is characterized by epithelial cell fenestrations forming a type of tight junction, thus maintaining another fluid barrier that protects the CNS (Redzic, 2011; Liddelow, 2015; Cousins et al., 2021). The epithelial cells of the BCSFB are a major site for CSF production. The passage of molecules through the BBB is via passive diffusion and facilitated transport. However, the passage of molecules through the BCSFB is a function of CSF production and its clearance. Despite having the BBB and the BCSFB, the brain uses the CSF as the vehicle to eliminate metabolic waste compounds and clear the CSF to the venous circulation to avoid toxic exposures. Accordingly, the CSF circulation significantly modulates the removal of toxic metabolites and metabolic byproducts, in order to maintain physiological homeostasis of the brain (Rennels et al., 1985; Ghersi-Egea et al., 1996; Bakker et al., 2016; Benveniste et al., 2017)

The production, circulation, function, and clearance of CSF from the central nervous system holds crucial pathological implications. Functioning against the build up of toxic products, CSF is responsible for clearing amyloid-beta (Aβ) and tau peptides, whose toxic aggregations in the parenchyma form, respectively, the extracellular senile plaques and intracellular neurofibrillary tangles of Alzheimer’s disease (AD) (Selkoe and Hardy, 2016). Their reduced CSF clearance has been implicated in their toxic accumulation and proposed as a cause of AD (Silverberg et al., 2001; Tarasoff-Conway et al., 2015). In another example, biomarkers of Parkinson’s disease including α-synuclein species, lysosomal enzymes, and cellular receptors have also been studied in the CSF, although their relationship to CSF flow dynamics remains unclear (Parnetti et al., 2019; Wilson et al., 2020). In idiopathic normal pressure hydrocephalus (iNPH), irregular pressure waves and not physical blockage reflect altered CSF fluid dynamics (Bradley Jr., 2015; Stöcklein et al., 2022). Similarly, Chiari malformations (cerebellar tonsillar ectopia) and syringomyelia (spinal cord central canal ectasia or adjacent destructive channel formation) can occur in relation to a structural restriction of CSF flow at the foramen magnum (Holly and Batzdorf, 2019; Mantha et al., 2021). The relevance of CSF circulation and clearance to diverse neurological diseases has led to extensive efforts by numerous groups to visualize CSF circulatory patterns (de Leon et al., 2017; Ringstad et al., 2018; Benveniste et al., 2021; Li et al., 2022a). In this review, we present the diverse imaging modalities, including cisternography, myelography, phase-contrast MRI and PET, used to image cerebrospinal fluid clearance and pathways across various pathological conditions.

2. CSF production

Most of the CSF is produced in a highly vascularized portion of the brain’s ventricular linings defined as the choroid plexus (ChP) (Shapey et al., 2019). The ChP tissue is located along the lateral ventricles, the roof of the third ventricle, and the caudal roof of the fourth ventricle (Sakka et al., 2011; Nakada and Kwee, 2019). The ChP comprises a vascular stroma covered by a monolayer of epithelial cells containing a variety of transporters, membrane proteins, and immune cells that regulate this epithelial blood-CSF barrier (see Fig. 1) (Lun et al., 2015). The ChP of the lateral and third ventricles receives blood, the source of the CSF, from the anterior and posterior choroidal arteries, while the ChP of the fourth ventricle is supplied by the anterior inferior and posterior inferior cerebellar arteries (Damkier et al., 2013; Tumani et al., 2017). The ChP passively filters plasma through its capillary endothelial cells and stroma after which the ChP epithelium actively filters and secretes ions into a modified plasma that is pushed along an osmotic gradient to become the intraventricular CSF (Sakka et al., 2011; Marques et al., 2017; Zhao et al., 2020). As compared with plasma, CSF is composed of increased concentrations of Na⁺, Cl⁻, and ${\text{HCO}}_3^{-}$, and lower concentrations of cations (K⁺, Mg²⁺, Ca²⁺). It contains actively transported proteins (including folate, thiamine), and low amounts of Albumin (Spector et al., 2015). Relative to blood, CSF contains significantly lower protein levels (~2 g protein/100 mL vs 7 g protein/100 mL in blood) (Spector et al., 2015; Podkovik et al., 2020).

Fig. 1.

Anatomy of the Choroid Plexus. Schematic diagram reproduced with permission from (Lun et al., 2015) highlighting the membrane of epithelial cells composing the choroid plexus (ChP). Steady blood flow, ion and solute exchange, and immune regulation across the stroma allow for secretion of CSF across the epithelium into the ventricles. This figure was originally published in Nat Rev. Neuroscience (DOI: https://doi.org/10.1038/nrn3921).

Besides the ChP, there is experimental evidence that CSF may arise directly from the capillaries in the ventricular ependyma and even the brain parenchyma (Sakka et al., 2011; Buishas et al., 2014) Generated at an estimated adult averaged rate of 0.3–0.4 mL per minute, traditional estimates define a total CSF volume of around 150 mL in adult humans (Spector et al., 2015; Brinker et al., 2014). However, recent neuroimaging data suggest volumes closer to 250 mL (Chazen et al., 2017). Further, with normal aging and in cases of AD, there is a 50–100% increase in the CSF volume due to the passive filling of atrophy with compensatory ventricular and sulcal enlargement (Takeda and Matsuzawa, 1985; Shear et al., 1995; Symonds et al., 1999; de Leon et al., 2002; Ott et al., 2010).

3. CSF circulation

In 1828, Francois Magendie’s study of CSF circulatory pathways reported that the ventricular and subarachnoid fluid are connected through a midline fourth ventricule “communication”. A simple dogma formed that CSF circulates in a connected, enclosed, and regular path from the ventricular choroid plexi to the subarachnoid space surrounding brain and spinal cord and then to the superior convexity arachnoid granulations. Intraventricular CSF circulation is predominantly unidirectional (craniocaudal), from the lateral ventricles through the interventricular foramen of Monroe into the third ventricle (Roesch and Tadi, 2021), then through the cerebral aqueduct of Sylvius into the fourth ventricle (Orěskovíc and Klarica, 2010) and passing through the foramina of Magendie and Luschka into the subarachnoid space of the brain and spinal cord (Killer, 2013). Recent imaging work has revealed a cardiac cycle associated-pulsatile motion of the CSF with a predominant directionality (Mestre et al., 2018), as well respiratory influences on CSF flow (Lloyd et al., 2020).

4. CSF in perivascular spaces

Covering the brain are three meningeal layers comprising the dura mater, the arachnoid membrane, and the pia mater respectively. The subarachnoid space is bounded at the exterior by dura and the arachnoid membrane, with arachnoid trabeculae traversing the subarachnoid space (Nicholas and Weller, 1988; Zhang et al., 1990; Weller et al., 2018). Trabeculae divide the subarachnoid space into compartments and suspend arteries and veins within the subarachnoid space (Weller et al., 2018). CSF communication between the subarachnoid space and the topologically “separate” perivascular space occurs via micron-sized leptomeningeal pores, called stomata, that are webbed by fibers (see Fig. 2) (Abbott et al., 2018). These anatomically complex features are responsible for physiological (e.g. postural, diurnal) regulation, pathological circumstances, and allow a potent CSF connectivity between the subarachnoid space and perivascular space (Mastorakos and McGavern, 2019).

Fig. 2.

Visualization of the perivascular space (PVS). Schematic diagram from (Abbott et al., 2018) (CC by 4.0) highlighting key features surrounding a penetrating leptomeningeal artery in the subarachnoid space (SAS). The fluid filled perivascular compartment within the outer layers of blood vessels is defined by the astroglial and endothelial basement membranes. Distinct anatomical regions of potential fluid exchange mechanisms are specifically highlighted: 1. Micron-sized pores deemed stomata that allow for communication between the SAS and PVS 2. Fenestrations within pia mater cells providing fluid communication between the subpial space and the SAS.

Electron microscopy studies reveal that the pia mater lining the surface of the brain reflects on arteries that enter into the cortex from the subarachnoid space. This results in a perivascular compartment composed of the basement membrane of the pia mater and the basement membrane of the astrocyte end feet (glia limitans) known as the glial basement membranes (see Fig. 2). This perivascular compartment is also referred to as the “Virchow-Robin” space, representing the pathways through which CSF enters the brain as convective influx/glymphatic entry (Rennels et al., 1985; Zhang et al., 1990; Weller et al., 2018; Abbott et al., 2018; Benveniste et al., 2019; Nycz and Mandera, 2021). Despite the numerous accounts of perivascular or Virchow-Robin spaces between the glia limitans and the walls of arteries within the cerebral cortex, transmission electron micrographs show there are no true perivascular (Virchow-Robin) spaces in either the human or rodent brains (Zhang et al., 1990; Morris et al., 2016). The problem consists of the misinterpretation of artefactual separation of brain tissue from vessel walls in paraffin sections of post-mortem human brain specimens (Sapsford et al., 1983) and misinterpretation of artefactually swollen perivascular astrocyte processes mistaken for perivascular spaces. In the white matter and in the basal ganglia, there are two leptomeningeal sheets in the perivascular compartment, and they can expand to create true perivascular spaces visible with MRI (Pollock et al., 1997; Weller et al., 2015; MacGregor Sharp et al., 2019).

The perivascular “spaces” composed of pial-glial basement membranes are filled with penetrating CSF (Bedussi et al., 2018) and may taper off towards capillary level where the compartment is composed of the fused endothelial and astroglial basement membranes, with no leptomeningeal (pial) sheet. Along the penetrating arterioles, astrocytes carry aquaporin-4 responsive channels which assist in a unidirectional convective influx of CSF (Bacyinski et al., 2017). This physiological process reported as delivering the draining CSF and its contents to lymphatic structures has been termed the glymphatic system (Iliff et al., 2012). This process depends on an osmotic pressure gradient for convective influx of CSF into the parenchyma (Pirici et al., 2018). This is one of several purported CSF transport mechanisms that also includes vascular pulsations (Daouk et al., 2017), respiration (Dreha-Kulaczewski et al., 2018), diurnal variations (Xie et al., 2013) and posture (Muccio et al., 2021). Our understanding of the anatomy of this circulation is emerging, and there remains methodological hurdles to refining our knowledge of the volume and transport dynamics of CSF, the latter comprising the speeds and directionality of the flow. Estimating CSF volume turnover has been difficult and a target of recent imaging approaches. This review will examine current in-vivo imaging modalities used to study CSF circulatory and clearance pathways.

5. CSF clearance

During its circulation, CSF is cleared from the subarachnoid space into venous blood by a variety of mechanisms. Classically, the major resorption of CSF was thought to occur via arachnoid granulations, small projections of arachnoid tissue through the dura mater to the venous sinuses (see Fig. 3) (Proulx, 2021). There are differences between species, with arachnoid granulations present in humans but not in rodents. There are also human maturational changes with increased arachnoid granulations with aging (Radoš et al., 2021). Arachnoid granulations contribute to the drainage of CSF into the venous sinuses in humans, whereas in rodents a greater fraction of the CSF drains across the cribriform plate into the nasal mucosa. There is direct clearance of CSF into the lymphatic vessels of the dura mater (Louveau et al., 2015; Aspelund et al., 2015; Ma et al., 2017). As well as strong evidence in multiple species of CSF clearance via cranial nerves, particularly the olfactory (cn. I) and optic (cn. II) nerves and spinal nerves into lymphatic vessels (Proulx, 2021; Kida et al., 1995; Johnston et al., 2004; Johnston et al., 2005). As reviewed by (Mehta et al., 2022), these circulatory and clearance pathways are important for the physiological homeostasis of the brain and are implicated in several neurological diseases. As such, the development of modalities to image CSF circulation and flow parameters, including localization, and pressure dynamics, are of fundamental importance.

Fig. 3.

Summary of various proposed CSF clearance pathways. Schematic diagram highlighting strength of scientific evidence from animal and human studies supporting CSF clearance routes across arachnoid granulations, dural lymphatics, cranial nerves, and lumbar nerves from Proulx (2021) (CC by 4.0).

6. Aging and CSF flow

The physiology and composition of CSF changes with human aging and neurodegeneration. The aging of the human CNS can be described in metabolic terms as a combination of imbalanced functions including: ion status, nucleotide and protein biosynthesis, and chemical degradation compromising the metabolic homeostasis of the brain (Ivanisevic et al., 2016). These changes are believed to be reflected in an altered CSF composition (Hwangbo et al., 2021; Peters et al., 2021). At an anatomical level, aging is associated with ChP atrophy with reduced CSF production due to reduced ChP vascularity, flattened ChP epithelium, and shortened microvilli (Serot et al., 2003). Age-related extracellular fibrosis in the ChP stromal layer is amplified with senescence resulting in reduced production and circulation of CSF (Serot et al., 2000). In mice, these age-related processes underlie a decline in production of CSF over the life span of up to 33% (Liu et al., 2020). Human CT scan imaging demonstrates marked age-related enlargements (Alisch et al., 2021) and calcifications of the ChP (Modic et al., 1980).

Overall, the aging process directly influences the robustness of and mechanisms of CSF clearance. In AD, the reduced clearance of amyloid-beta and tau peptides via CSF egress potentially marks a key factor of disease pathogenesis (Tarasoff-Conway et al., 2015; Sharma and Singh, 2016). Similarly in mice, for example, in-vivo imaging of intracisternally infused fluorescently labeled albumin showed a significant decrease in CSF clearance with increasing age (Brady et al., 2020). This age-dependent reduction in the flow of CSF was also observed in humans using endogenous phase contrast (PCC or PC MRI) which allowed the observation of reduced CSF pulsations due to reduced CSF stroke volumes through the aqueduct and cervical cord in the elderly (Stoquart-ElSankari et al., 2007). Using bipolar gradients and phase data present in MR signals, PCC MRI quantifies fluid velocity in diverse clinical settings (Wymer et al., 2020). Here, decreased CSF clearance associated with aging has neurodegenerative implications. In another study using PCC MRI to quantify ventricular and spinal CSF, reduced spinal CSF flow in 92 geriatric patients was associated with memory loss, decreased visuo-constructive ability, and decreased verbal fluency (Attier-Zmudka et al., 2019). Further, reduced CSF clearance was hypothesized as a potential contributor to the cognitive decline experienced in patients with AD (Attier-Zmudka et al., 2019). Ultimately, the connection between neurodegenerative processes of aging and CSF dynamics underscores the need for physiologically robust imaging modalities that can localize, trace, and visualize CSF dynamics.

7. CSF imaging modalities in animals

In understanding CSF production, circulation, and egress, a number of imaging modalities have been applied to various animal species. In rodents, intrathecal (IT) administration of gadolinium (Gd) or other paramagnetic molecules (i.e. application of contrast agents to the spinal subarachnoid space), allows direct communication with circulating CSF. This approach when examined by dynamic contrast-enhanced MRI, is used to study the time course of CSF clearance. In one pivotal study, Gadopentetic acid (Gd-DTPA), a low molecular weight tracer (0.938 kDa), as well as polymeric Gd-chelate (GadoSpin), a larger molecular weight (200 kDa), was administered IT in 14 rats (Iliff et al., 2013). The transport of agents was monitored by T1-weighted MRI and a dynamic time series traced the time-dependent pathways of CSF flow through brain parenchyma (Iliff et al., 2013). This study provided a proof of concept in using MRI to visualize CSF as it entered brain tissue along the peri-arterial space and exchanged with interstitial fluid (Iliff et al., 2013) In another study, intrathecally administered low weight (Texas Red–dextran-3, MW: 3 kDa) and high weight (FITC-d2000, MW: 2000 kDa) molecular tracers were used in mice to clarify CSF flow pathways (Iliff et al., 2012). Following tracer administration, two-photon laser scanning microscopy mapped dextran tracer fluorescence along CSF pathways (Iliff et al., 2012). Here, subarachnoid CSF was found to readily cycle through the brain interstitial space. In both studies, the use of both low and high-weight tracers were used to clarify the characteristics of waste products cleared and helped expand CSF neuroimaging in rodents. A similar attempt to image CSF in mice in another study relied on conventional fluorescence microscopy of injected tracer (Texas Reddextran-3, MW 3 kD; FITC- dextran-500, MW 500 kD) at both the cisterna magna and lumbar spine (Yang et al., 2013). This study provided evidence of the feasibility of using lumbar intrathecal contrast with dynamic contrast enhanced MRI to examine glymphatic drainage. Here, conventional fluorescence microscopy and contrast was used to compare the impact of tracer location (cisterna magna vs lumbar spine). Evidence for CSF movement in both regions was observed (Yang et al., 2013). While fluorescent microscopy (Yang et al., 2013) and two-photon laser microscopy (Iliff et al., 2012) are useful at providing anatomical maps of CSF flow at a depth of 250 μm in shaved skull in mice, they are not suitable for human studies. However, the authors concluded that lumbar intrathecal delivery of contrast tracer can be coupled with the dynamic contrast enhanced MRI as a clinically relevant imaging modality for CSF movement and may be informative in humans.

While the application of fluorescence microscopy, laser-microscopy, and contrast-enhanced neuroimaging in rodents provides a preliminary understanding of how various imaging modalities can be used to map CSF flow, its application to CSF dynamics must be validly translated to human subjects. There are large differences in CSF turnover between species (Lu et al., 2013), in part due to physiological differences in cardiovascular function. For example, the mouse heart beats at approximately 600 beats per min and the human at approximately 60 beats per min (Janssen et al., 2016), and there is a nearly et al. 1000 fold greater size and complexity in the human nervous system (Hodge et al., 2019). Consequently, a push to further develop the imaging modalities applied to CSF dynamics has begun.

In a recent innovation of MR imaging on murine CSF flow pathways, one group examined fluid flow around the middle cerebral artery of mice using diffusion weighted MRI sequences (Harrison et al., 2018). Ultra-long echo time, low b-value, six-direction diffusion-weighted MRI was used to identify the direction of CSF fluid flow in the small predefined anatomical region of interest (Harrison et al., 2018). Glymphatic inflow of CSF was imaged, solidifying a proof of concept for the use of MRI in CSF mapping. This non-invasive and contrast-free MRI method measures random Brownian motion of molecules in mm²/s and uses different gradient probing directions to identify the direction of greatest D* related to fluid flow. Most recently, the expansion of non-invasive MR imaging in the murine model has turned to phase-contrast MRI in which a novel and generalized encoding-based multi-band scheme was incorporated to increase signal-to-noise ratio (Li et al., 2022b). This methodology was subsequently used to identify the reduction in CSF flow that occurs with anesthesia in living C57BL/6 mice (Li et al., 2022b). Overall, these developments in rodent neuroimaging modalities have a potential application to human CSF flow analysis.

Alongside the literature using various imaging modalities to understand CSF circulation in rodents, a growing attempt to trace CSF pathways in dogs has relied on PCC MRI. One study on Cavalier King Charles Spaniels used PCC MRI to examine CSF flow at the foramen magnum and C2–3 intervertebral disc level to clarify how obstruction of CSF flow patterns are related to syringomyelia, the destructive development of fluid-filled cysts in the spinal cord (Cerda-Gonzalez et al., 2009; Hechler and Moore, 2018). Similarly, a recent prospective study used cardiac-gated PCC MRI at 3 anatomical locations in Beagles (mesencephalic aqueduct, foramen magnum, and cervical spine) to examine CSF flow and velocity (Christen et al., 2021). Using PCC MRI, an influx and efflux of CSF consistent with the “to and fro” motion was observed at all locations in the beagles (Christen et al., 2021). In short, a growing literature ranging from microscopic techniques to phase contrast MRI has created a baseline understanding of how neuroimaging can be used to map the circulation of CSF in animals (Brinker et al., 2014). These various neuroimaging studies in animal systems have complemented MRI studies, and more recently PET imaging, to map CSF flow in the human brain and spinal cord.

8. CSF imaging modalities in humans

In parallel with the development and implementation of advanced imaging modalities in animals, a similar endeavor to image CSF circulation has been applied in a variety of clinical contexts (Whedon and Glassey, 2009). MR cisternography and MR myelography are often used for the diagnosis and neurosurgical treatment of anatomical CSF disorders (Mokri, 2014). Rhinorrhea, defined as the leakage of CSF from the subarachnoid space into the nasal fossa, is a common reason for MR cisternography (Abboud et al., 2020). Extending beyond leaks, CSF clearance pathways may have direct relevance to Alzheimer’s disease and possibly to other neurodegenerative diseases (Tarasoff-Conway et al., 2015; Mehta et al., 2022; Kylkilahti et al., 2021). The important function of CSF in clearing beta-amyloid and tau protein, whose toxic aggregation defines AD, has been widely documented (Kress et al., 2014; Ueno et al., 2014; Xin et al., 2018; Carare et al., 2020). As such the circulation and clearance of CSF from the brain has been an important focus of neuroradiological development, starting with MRI studies with and without the use of Gd contrast, and more recently, dynamic PET studies (see Table 1). Here we examine these neuroimaging modalities, discussing methodology, application to disease processes, and potential future advancements.

Table 1.

Summary of CSF neuroimaging modalities.

Imaging modality	Information given	Primary pathological use	Invasive?	Sample studies
MR Cisternography	CSF localization	CSF leakage out of the intracranial compartment	Can be done with or without intrathecal injection of Gd contrast	Aydin et al. (2008), Tedeschi et al. (2009), Dogan et al. (2020)
MR Myelography	CSF localization and flow parameters	CSF leak in the spine; CSF flow tracking from the spine to the head	Can be done with or without intrathecal injection of Gd contrast	Ringstad et al. (2018), Chazen et al. (2014), Hattingen et al. (2009), Kim et al. (2020), Ringstad et al. (2017), Eide and Ringstad (2019), Dyke et al. (2020)
Phase-Contrast MRI	CSF Flow Parameters	CSF flow dynamics in structural disorders (iNPH, Chiari malformation) and recent AD work	No	Attier-Zmudka et al. (2019), Kartalci et al. (2021), Ahmad et al. (2021)
Dynamic PET	CSF Flow Parameters	CSF clearance in neurodegenerative diseases	No	de Leon et al. (2017), Li et al., 2022a, Schubert et al. (2019)

Comparison of various neuroimaging techniques of CSF localization, flow and clearance based on diagnostic value, invasiveness, and imaging data obtained.

9. MR cisternography

MR Cisternography is a proven imaging technique often utilized before surgery to diagnose the site of CSF leakage out of the intracranial compartment (Algin and Turkbey, 2013). In one of the first examinations of the efficacy of MR cisternography, a cohort of 30 patients with post-traumatic CSF fistulae were examined with two T2-weighted MRI sequences and a 3D constructive interference steady-state (CISS) sequence in the supine position (Eberhardt et al., 1997). Identification of CSF leakage in the skull was confirmed through intraoperative inspection of the skull during neurosurgical correction of suspected fistulae (Eberhardt et al., 1997). Here, MR cisternography provided higher detection sensitivity for CSF leakage than CT cisternography and was confirmed to be of direct diagnostic value (Eberhardt et al., 1997).

In a similar attempt at the pre-surgical localization of CSF leakage in 51 patients with CSF rhinorrhea, another study used Gd-enhanced MR cisternography (Aydin et al., 2008). Here, lumbar IT injection of Gd tracer was coupled with T1-weighted MR images through the head to identify the site of CSF leakage (Aydin et al., 2008). Neurosurgical intervention again confirmed the site of the leak. It was determined that MR cisternography can accurately localize CSF leakage in the setting of rhinorrhea with 84% yield (Aydin et al., 2008). With surgical validation and promising neurosurgical outcomes, MR cisternography for this purpose has become an important addition to the neuroradiology of CSF leakage (Tedeschi et al., 2009; Satoh et al., 2005; Dogan et al., 2020; Duman et al., 2021). While MR cisternography is a robust, radiation free method to identify CSF leakage, it often uses an MR contrast agent and false negative do occur. Further, MR does not offer detailed osseous anatomy like CT (Vemuri et al., 2017).

10. MR myelography

While MR cisternography targets the head to visualize CSF fistulae and skull base leaks, MR myelography is used to identify leaks in the spine and surrounding tissues (Stone, 2003). Often coupled with intrathecally administered Gd contrast in the lumbar spine, MR myelography uses T1 and T2 weighted MR images to visualize the thecal sac and spinal cord, assisting in the localization of spinal cord CSF leakage, venous fistulae, and other structural abnormalities (Hergan et al., 1996). In one recent study, 97 patients were examined with MR myelography to detect and localize suspected CSF leakage (Madhavan et al., 2021). Upon neurosurgical intervention, it was determined that the diagnostic yield for identifying CSF leaks via MR myelography was 14% (Madhavan et al., 2021). Here, authors concluded MR myelography with intrathecal Gd had a limited capability in diagnosing and localizing CSF leaks, CSF-venous fistulae and distal nerve root sleeve tears (Madhavan et al., 2021). Another study comparing traditional CT myelography with MR myelography with intrathecal Gd in 24 patients with suspected CSF leaks found a 38% sensitivity (Chazen et al., 2014). A growing literature examining the diagnostic potential for MR myelography has been established over the years, and the technique has been deemed useful in its practicality and relative safety (Yoo et al., 2008; Hattingen et al., 2009; Farb et al., 2019; Kim et al., 2020; Kranz et al., 2021). Despite this, the low-dose application of intrathecal Gd often used in MR myelography consists of off-label use and as associated risk of seizures (Patel et al., 2020; Pellis et al., 2020). The invasiveness of contrast application, potential for neurotoxicity, limited procedural training in radiologists and a limited diagnostic value have largely limited the widespread application of MR myelography.

MR myelography performed to localize CSF leaks usually involves imaging the patient soon after Gd injection and occasionally bringing the patient back for delayed imaging after 1–2 h. However, CSF flow has been observed over time when the intrathecal Gd is followed days and even weeks following injection. Extending the work of Gd contrast MRI in rodents, one prospective study examined glymphatic function using intrathecal lumbar injection of the contrast agent gadobutrol (Ringstad et al., 2017). Fifteen patients diagnosed with iNPH were imaged in the supine position with T1-weighted MRI with the same imaging parameters during and 24 h after lumbar IT contrast agent administration (Ringstad et al., 2017). Patients remained in the supine position throughout imaging. Following analysis of sequentially obtained MR images, gadobutrol clearance was traced from the subarachnoid space into the brain, establishing a clear diagnostic potential for MRI to track CSF flow dynamics (Ringstad et al., 2017). Following this work, gadobutrol usage was extended to another study obtaining T1-weighted MRI following a single dose of intrathecal gadobutrol at 24 h, 48 h, and 4 weeks in a cohort of normal control, dementia, and iNPH patients (Ringstad et al., 2018). Brain-wide enhancement of gadobutrol tracer following injection, provided evidence for a communication between CSF in the subarachnoid space and multiple brain compartments including the cerebral cortex, basal ganglia, thalamus, limbic system, cerebellar cortex, and cerebellar white matter (Ringstad et al., 2018). Here, tracer enhancement helps visualize CSF circulatory pathways and provides concrete evidence for the potential of MRI with intrathecal injection in this setting (Ringstad et al., 2018).

Following the introduction of T1-weighted MRI with intrathecal lumbar administered gadobutrol as a potential means of mapping CSF pathways, multiple additional studies attempted to expand on the use of this imaging modality to understand CSF fluid dynamics. One such study examined the clearance of gadobutrol from the entorhinal cortex, a region of the brain closely related to the hippocampus and a site of early AD tau pathology, in 30 patients with iNPH and potential CSF leakage (Eide and Ringstad, 2019). In this work, intrathecally administered gadobutrol was imaged with T1-weighted MRI, and CSF flow was analyzed as changes in normalized (calculated as the ratio of measured T1 signal unit from CSF or brain parenchyma with the value from any reference ROI) relative hyperintensity units (Eide and Ringstad, 2019). In this study, delayed clearance of CSF tracer in the entorhinal cortex and adjacent white matter was associated with iNPH prognosis (Eide and Ringstad, 2019). More recently, intrathecal injection of Gd contrast (0.5 cc Magnevist, Gd-DTPA) in 8 healthy volunteers was coupled with six sequentially obtained T1-weighted MR sequences to assess CSF transport kinetics (Dyke et al., 2020). Here, Gd was observed to peak within 1–3 h, and sequential enhancement was be used to observe differential CSF flow in the brain parenchyma (Dyke et al., 2020).

Overall, these studies characterize the novel use of T1-weighted MR imaging of Gd-based CSF tracer in analyzing CSF clearance pathways both in and outside the central nervous system. Despite these important findings, this technique is invasive, relying on the lumbar injection of a Gd-based tracer, and are clinically cumbersome to patients, requiring multiple follow-up imaging sessions, hours to weeks after injection. As such, a more widely accessible MRI methodology to explore CSF flow imaging is warranted.

11. Phase-contrast cine MRI (PCC MRI)

Among the most widely used imaging modalities of CSF flow is PCC MRI, a non-invasive technique without contrast administration or catheterization. PCC MRI generates signal contrast between flowing and stationary water nuclei, allowing for detection of CSF flow within the ventricular system as well as the subarachnoid space (Barkhof et al., 1994). Over the last 20 years, a great emphasis has been placed on using PCC MRI to understand CSF circulation in neurodegenerative pathology. In one study, CSF flow dynamics in the cerebral aqueduct were compared between 50 schizophrenia patients and 50 normal controls using PCC MRI (Kartalcı et al., 2021). Two-dimensional flow PCC MR images coupled with T1-weighted MRI in three planes were used to analyze CSF flow parameters in and around the cerebral aqueduct (Kartalcı et al., 2021). Here, the authors concluded that CSF flow between the third and fourth ventricle has a lower peak velocity, lower net forward volume, and lower average flow in schizophrenia patients (Kartalcı et al., 2021). A similar endeavor to contextualize CSF flow disruption to mild cognitive impairment used PCC MRI in a cohort of 92 geriatric patients (Attier-Zmudka et al., 2019). Flow images were obtained with a peripherally-gated velocity-encoded phase-contrast pulse sequence with peripheral gating (Attier-Zmudka et al., 2019). Findings confirmed the association of decreased CSF flow in the spinal column and ventricular system with cognitive deficits (Attier-Zmudka et al., 2019). In these cases, the use of PCC MRI was restricted to the ventricular anatomy where volumetric, velocity, and directional flow analysis provided information supporting a flow and clearance mechanism in neurodegenerative diseases.

Alongside its application to neurodegenerative disease, PCC MRI has been utilized to study the impact of CSF flow on many neurologic disorders including iNPH, Chiari malformation, intracranial hyper and hypotension, and obstructive hydrocephalus. In one wide-ranging study, a group of 39 patients stratified along five subgroups (iNPH, hydrocephalus, idiopathic intracranial hypertension, brain atrophy, and Chiari I malformation type) were examined using cardiac-gated sagittal PCC MR image (Ahmad et al., 2021). In the iNPH group, PCC MRI at the level of the cerebral aqueduct found hyperdynamic CSF flow (Ahmad et al., 2021). Additionally, authors were able to use the PCC MRI data to successfully diagnose the exact cause of obstructive hydrocephalus (Ahmad et al., 2021). Across all pathological subgroups, PCC MRI provided novel and useful information on CSF flow parameters, confirming the technique’s effectiveness in the diagnosis and prognosis of a diverse cohort of neurological problems.

A similar examination of CSF flow dynamics in idiopathic intracranial hypertension was conducted using PCC MRI with a group of 19 patients and 11 healthy controls (Akay et al., 2015). Here, PCC MRI images taken through a cardiac cycle were used to calculate CSF flow parameters (cardiac gating): mean aqueduct area, mean peak rate, mean average rate, mean flow (Akay et al., 2015). The findings indicated significantly altered CSF flow in patients with idiopathic intracranial hypertension relative to normal controls, and once again validated the use of cardiac-gated PCC MRI as an imaging modality used to visualize CSF flow pathways and mechanisms (Akay et al., 2015).

The use of cardiac gated PCC MRI was extended to a cohort of 40 patients with cystic malformations of the posterior fossa (Yildiz et al., 2006). Cystic malformations of the posterior fossa cover a wide variety of findings, including the cisterna magna, arachnoid cysts, Dandy-Walker malformations, and include lesions that often intrude the subarachnoid space, interfering with CSF flow (Kollias et al., 1993). In this study, patients were subdivided into differing pathologies: communicating arachnoid cyst (n = 12), non-communicating arachnoid cyst (n = 7), Dandy-Walker malformation (n = 6), Dandy-Walker variant (n = 2), Blake’s pouch cyst (n = 3), and the normal variant configuration of the cisterna magna called “mega cisterna magna” (n = 10). In all subgroups, cardiac gated PCC MRI yielded distinctly different flow parameters of CSF within the foramen magnum based on individual cystic fluid patterns and pathological differences (Yildiz et al., 2006). PCC MRI examination of CSF fluid flow thus provided clear diagnostic criteria for varying malformations of the posterior fossa, defining another potential use of the technique.

While the development of PCC MRI has yielded a non-invasive, contrast-free, and practical capacity to image CSF flow and its respective properties, there are growing concerns about its clinical utility (Hladky and Barrand, 2014). Given slow CSF velocities relative to blood flow, and extremely small regions of imaging interest, PCC MRI may not provide accurate or complete visualization of CSF flow pathways and mechanisms (Battal et al., 2011; Hladky and Barrand, 2014; Korbecki et al., 2019). Moreover, quantitative thresholds for definition of pathology may vary across scanners and institution-specific technique. As a result, it becomes important to examine alternate imaging modalities that maintain the non-invasive nature of PCC MRI, while expanding its physiological reach and diagnostic potential; One such solution arrived in the form of dynamic PET imaging.

12. Dynamic PET imaging

Dynamic PET imaging has recently been applied to the visualization of CSF flow and clearance. In one study (de Leon et al., 2017), the radiotracer ¹⁸F-THK5117 was administered via intravenous bolus injection to examine 8 AD patients and 7 normal control subjects with a dynamic PET acquisition. Following decay correction and SUV normalization, PET data was coregistered with corresponding T1-weighted MRI. There were three observations made: 1. AD vs control subjects showed a 33% reduction in the ventricular CSF clearance rate; 2. The time-activity curves for extracranial voxels were correlated with the ventricular CSF to statistically define “CSF-positive voxels (r’s >.95),” and image the time-dependent anatomical egress of CSF. The results demonstrated in all subjects multiple potential CSF egress pathways. Most prominent was an egress signal under the cribriform plate at the superior nasal turbinate, which demonstrated a significant CSF egress reduction in AD; and 3. In the AD subjects that underwent additional PET scanning with ¹¹C-Pittsburgh compound B (¹¹C-PiB, Klunk et al., 2004) for amyloid-beta plaques, this study provided the first evidence of an inverse relationship between ventricular CSF clearance and amyloid lesions (de Leon et al., 2017).

Following this first attempt to use dynamic PET with coregistered T1-weighted MRI to identify and map brain CSF flow, another study used an intravenous bolus injection of ¹¹C-PiB to map CSF clearance in 11 AD, 12 mild cognitive impairment, and 20 multiple sclerosis subjects (Schubert et al., 2019). Here, dynamic PET was again coregistered with T1-weighted MRI to trace radioactive contrast and clarify CSF dynamics. The use of dynamic ¹¹C-PiB and quantitative modeling across a cohort of neurodegenerative diseases confirmed the potential of PET as a reliable, minimally invasive, and informative means to assess CSF pathways and dynamics. Results suggested significantly decreased clearance of tracer in Alzheimer’s patients relative to controls, confirming the successful imaging of CSF using PET. Most recently, Li et al. (2022a) used ¹⁸FTHK5351 and ¹¹C-PiB as PET tracers to assess CSF clearance in 15 AD and 9 control participants. Here, reduced ventricular CSF clearance was again confirmed in AD with a third PET tracer, and significantly associated with the PiB determined amyloid-beta peptide load (Li et al., 2022a). This study further establishes a role for PET imaging in evaluating CSF clearance pathology and pathways to AD.

Despite these recent demonstrations using dynamic PET to assess CSF flow mechanisms, the biological validation of the PET radiotracers to function as a surrogate or biomarker for CSF remains unknown. The combined use of dynamic PET with MRI with an emphasis on CSF mapping may, however, represent a promising new era for neuroradiology as we continue to expand and improve our study of CSF flow and clearance.

13. Novel and hypothetical approaches to understanding fluid egress pathways

The precise anatomy for all possible CSF clearance pathways remains vague. Important clues regarding the anatomy of fluid movement in the brain may actually come from peripheral sources. Recent studies using in vivo confocal laser endomicroscopy have shown that many different forms of fibroconnective tissue in the human body contain previously unappreciated interstitial spaces (Benias et al., 2018). The interstitium is considered the source of lymph and is therefore defined as the prelymphatic space. These newly identified structures may provide the missing links for CSF clearance.

The interstitium has previously and principally been thought to be comprised of minute channels between some cells (e.g. squamocytes, lipocytes) measuring under 1.5 μm (Tamaki et al., 2002; Calabrese et al., 2003; Aslan et al., 2006), and larger, peri-vascular channels – around capillaries and, in some tissues, venules and arterioles – measuring 5–10 μm, across. The peri-vascular interstitium is the principal site in which there is exchange of nutrients and waste products between the vascular system and the surrounding tissues (Swartz and Fleury, 2007; Wiig and Swartz, 2012; Stewart, 2020). The newly recognized interstitium is larger still, with spaces measuring sometimes >100 μm, even macroscopically evident in some connective tissues (Benias et al., 2018). This may be referred to as the reticular interstitium (Theise et al., 2022), in keeping with, for example, the terminology regarding the skin’s reticular dermis and the body-wide reticular framework of mammalian bodies as described over a century ago by Franklin Mall (Sabin, 1934) and confirmed by recent production of organ matrix scaffolds by decellularization (Kajbafzadeh et al., 2015).

Including the CNS all interstitial spaces, regardless of size and location, contain water and its solutes, collagens, elastin, proteoglycans and glycosaminoglycans, particularly hyaluronic acid (HA) (Swartz and Fleury, 2007; Wiig and Swartz, 2012; Stewart, 2020). In the intercellular interstitium, collagen and elastin are generally not seen. In the perivascular interstitium collagen and elastin fibrils are present and may begin to aggregate into larger bundles. Within the reticular interstitium, the collagen will aggregate into larger bundles ranging to 10–20 μm.

In tissue excised for histologic examination with dehydration and fixation, it is these reticular collagen bundles that collapse upon each other giving the appearance of dense collagenous connective tissue; however, the in vivo microscopy revealed that this appearance is arti-factual. Remnants of these spaces are, nonetheless, recognizable in standard routine sections (Benias et al., 2018) and can be easily confirmed by staining for hyaluronic acid in tissue sections utilizing peroxidase-labeled hyaluronic acid binding protein. In doing so, these spaces have also been shown to be continuous across tissue layers within organs and between organs (Cenaj et al., 2021). Such continuity has now been demonstrated from epidermis through the skin into the subcutaneous fascia (Cenaj et al., 2021), from lamina propria of the colon across all its layers into the mesenteric fascia (Cenaj et al., 2021), throughout the structures of gynecologic tract (Wang et al., 2022), from the bronchi of the lungs throughout the lung parenchyma and the pleura (Ordner et al., 2022), and in and around through the heart and lungs (unpublished data). As such, further research to examine a similar hypothetical continuity of the interstitial space in the CNS may provide additional clarity into CSF circulatory pathways, clearance mechanisms, and CSF-ISF communications.

Most importantly, the connective tissue surrounding all arteries and veins (whether loosely aggregated in smaller vessels or in distinct adventitia in larger ones) and nerves (perineurium) encloses these same hyaluronic acid filled spaces. These interstitial spaces are also continuous with all the tissues through which they pass and, presumably, along the entire vascular tree and peripheral nervous system (Cenaj et al., 2021). Thus, the interstitium appears to comprise a body-wide communication network of interconnected spaces for movement of solutes, signaling molecules, hormones, and cells, both in health and in disease. These findings, of course, suggest possible missing links between the hypothesized routes of CSF clearance and extracranial tissues along all vessels and nerves. These novel anatomic findings may therefore be important to explaining one or more of the hypothesized routes of CSF clearance and their possible inter-relationships.

14. Conclusion

Overall, efforts to image and map CSF fluid dynamics ranges broadly from imaging NPH and CSF leakage using MR myelography and cisternography to the newly introduced non-contrast MRI and dynamic PET imaging. The variable efficacy of high-molecular and low-molecular weight tracers have created a diverse range of neuroimaging techniques with diagnostic capabilities that are able to implicated altered CSF flow to diseases such as AD in the elderly to Chiari malformation in pediatric age groups. Despite this, these techniques differ in terms of accuracy, invasiveness, spatial resolution, and burden to patients, making standardization of CSF physiology and flow imaging difficult. Further research must be conducted to understand the efficacy of different CSF tracers and imaging modalities with an emphasis on biological validation and clinical practicality. Recent neuroradiological advancements in integrated PET/MR imaging, incorporating multiple techniques, may provide the next breakthrough in visualizing CSF fluid dynamics (Fink et al., 2015; Dupont, 2021). Ultimately, in expanding our insight into how different neuroimaging modalities can be used to image CSF extent, flow, and clearance, we will continue to improve diagnostic potential, clinical treatment effectiveness, and quantification of fluid dynamics in the central nervous system.