Figure. In vivo (A) and in vitro (B) illustrations. A, The schematic drawing illustrates the anatomic relationship between the neuropil and the ventricular/subarachnoidal cerebrospinal fluid on the one hand, and the neuropil and the Virchow-Robin spaces on the other. B, Illustration of the contact between substances within the cerebrospinal fluid and neural cultures on microelectrode arrays.
Hartung H, Dihné M. Volume Transmission–Mediated EncephalopathiesA Possible New Concept?. Arch Neurol. 2012;69(3):315-321. doi:10.1001/archneurol.2011.933
Author Affiliations: Medical Faculty, Department of Neurology, Heinrich-Heine University Duesseldorf, Dusseldorf, Germany.
There is strong evidence that the composition of cerebrospinal fluid (CSF) influences brain development, neurogenesis, and behavior. The bidirectional exchange of CSF and interstitial fluid (ISF) across the ependymal and pia-glial membranes is required for these phenomena to occur. Because ISF surrounds the parenchymal compartment, neuroactive substances in the CSF and ISF can influence neuronal activity. Functionally important neuroactive substances are distributed to distant sites of the central nervous system by the convection and diffusion of CSF and ISF, a process known as volume transmission. It has recently been shown that pathologically altered CSF from patients with acute traumatic brain injury suppresses in vitro neuronal network activity (ivNNA) recorded by multielectrode arrays measuring synchronously bursting neural populations. Functionally relevant substances in pathologically altered CSF have been biochemically identified, and ivNNA has been partially recovered by pharmacologic intervention. It remains unclear whether the in vivo parenchymal compartment remains unaffected by pathologically altered CSF that significantly impairs ivNNA. We hypothesize that pathologic CSF alterations are not just passive indicators of brain diseases but that they actively and directly evoke functional disturbances in global brain activity through the distribution of neuroactive substances, for instance, secondary to focal neurologic disease. For this mechanism, we propose the new term volume transmission–mediated encephalopathies (VTE). Recording ivNNA in the presence of pure human CSF could help to identify and monitor functionally relevant CSF alterations that directly result in VTEs, and the collected data might point to therapeutic ways to antagonize these alterations.
Cerebrospinal fluid (CSF) acts not only as a mechanical protection system that cushions forces impinging on the brain from outside the skull, but it also provides neutral buoyancy to prevent brain trauma caused by its own mass during normal body movement. The currently underestimated anatomic feature of CSF exchange with the interstitial fluid (ISF) that constitutes the extracellular milieu of the brain provides a basis for interesting functional considerations.
The largest portion of CSF in the ventricles and the subarachnoidal spaces (SAS) is produced by the choroid plexus (CP) at a rate of approximately 0.4 mL/min. Its production follows a circadian rhythm,1 up to a total volume of 150 to 160 mL, and the fluid has a turnover rate of approximately 4 volumes per 24 hours.2,3 The main bulk flow of CSF follows the neuraxis from the lateral ventricles through the third ventricle and continues into the fourth ventricle and through the hindbrain foramina into the cisterna magna and the basal SAS. There is also evidence for limited CSF flow from the cisterna magna into the cortical subarachnoidal regions and arachnoid villi.4
Cerebrospinal fluid is reabsorbed via 2 functionally distinct routes. The first route involves a rapid drainage through the arachnoid villi into the venous sinuses.5 The second is a slow reabsorption route in which the fluid follows the olfactory and optic nerves along the cribriform plate and the nasal submucosa, finally draining into the cervical lymphatic system. These routes are valid for several species, including nonprimates and humans6- 14 and provide a first clue for a sink function of the CSF.
Many interactions occur between the CSF and the ISF of the brain parenchyma. Both fluid masses move by bulk flow, CSF through larger cavities (eg, the ventricles) and ISF through narrow pathways within the perivascular or Virchow-Robin spaces and along myelinated fiber tracts.3,15,16 Importantly, CSF and ISF compartments are penetrable and are in contact across the ependymal cell layer of the ventricles and the pia-glial surface of cortical brain regions, leading to a dynamic exchange of water and substances.15- 19 In the ventricles, this exchange is possible because ependymal cells lack tight junctions. On the cortical surface, arteries and arterioles within the SAS are coated by a sheath of pia mater–derived cells that accompanies those arteries on their way into the cortical parenchyma.20 Perivascular spaces of arteries within the SAS and cortical parenchyma are in direct continuity. As the pial membrane appears to be fully permeable at least for solutes and, additionally, becomes perforated and incomplete after branching of arteries in the cerebral cortex, also the subarachnoidal CSF is in contact and exchanges substances with Virchow-Robin spaces. In deeper parenchymal cortical regions, the microvasculature is invested by astrocytic end feet of the glia limitans. Here, the Virchow-Robin spaces are bordered by the endothelium and the glia limitans.
Differences in barrier properties due to these morphologic differences are not fully understood. A summary illustration of the anatomic details is provided in the Figure. These characteristics support the idea that the ISF and CSF have a sink function that operates to guarantee extracellular homeostasis and to discard metabolic waste. The clearance of unneeded and/or toxic substances from the brain parenchyma occurs either in reverse, across the blood-brain-barrier and into the cerebral capillaries, or by bulk flow and diffusion into the CSF compartment. From the CSF, clearance proceeds by active reabsorption at the choroid plexus epithelium and through diffusion along concentration gradients and bulk flow toward arachnoidal sites16 and the cervical lymphatic system.6- 14
Substances in the large cavities of the CSF are able to enter the brain parenchyma, as shown by experiments involving the intracerebroventricular injection of interleukin-1β, which is also produced in situ by the choroid plexus,19 and by experiments using carbon 14 (14C)inulin17 or and 14C sucrose.21 In the interleukin study,19 fluorescence-labeled interleukin-1β penetrated the periventricular tissue and spread along fiber bundles and blood vessels into the caudate-putamen, hypothalamus, and amygdala. Importantly, subarachnoidal-space fluid that originated from the choroid plexus–produced ventricular CSF entered the cerebral cortex through perivascular or Virchow-Robin spaces.17,22,23
In other studies, evidence of CSF-ISF penetration of a variety of growth factors has been reported.24 Radiolabeled nerve growth factor (NGF) has been shown to enter the brain parenchyma and accumulate in basal forebrain cholinergic neurons and in the brainstem after intraventricular administration.25 Similar results were obtained for leukemia inhibitory factor (LIF), insulinlike growth factor II (IGF-II)25 and neurotrophin 3 (NT-3).26 Glial cell line–derived neurotrophic factor (GDNF) was shown to accumulate in the cerebral cortex, septum, diagonal band, fimbria, striatum, hippocampus, hypothalamus, substantia nigra/ventral tegmental area, and cerebellum after intraventricular injection.27 In these CSF-ISF exchanges, CSF was competent in widely distributing information within the brain. Because neuronal survival and function critically depend on precise extracellular concentrations of ions and other substances such as neuropeptides, growth factors, neurohormones, nucleosides, and vitamins, the clearance function from the ISF into the CSF and the distribution of substances from the CSF into the ISF suggest that the composition of CSF can influence the development and function of the parenchymal compartment.3
Factors produced by the choroid plexus within CSF are known to influence neurogenesis during central nervous system (CNS) development.28- 31 Among them are factors important for neural precursor proliferation and differentiation such as fibroblast growth factor-2 (FGF-2),31 NGF,32 somatotropin,33 brain-derived neurotropic factor (BDNF),34 IGF binding proteins (IGFBPs)35 and transforming growth factor α (TGF-α).36 Owen-Lynch et al37 report an example of disturbed CNS development due to CSF abnormality in fetal-onset obstructive hydrocephalic Texas rats, which have impaired cortical development that is attributed to a CSF-mediated proliferation block of neural precursors. In that study, impaired development of cortical cells was shown by directly applying the CSF from hydrocephalic rats to cortical precursor cells in vitro; this led to cell cycle arrest within S-phase. These reports suggest that the cortical abnormalities in these animals are primarily caused by a pathologic CSF composition during development and that the pathologic condition may have secondarily evolved as a result of disturbed CSF turnover.
Experiments involving the transfer of CSF from sleep-deprived goats into rats have further demonstrated the existence of functionally relevant CSF signaling.38 In these experiments, CSF-transplanted rats began to exhibit fatigue, indicating that CSF signals were transferred and that they influenced brain activity. Other experiments showed that cerebellar Purkinje neurons were able to extract small and large molecules from the CSF, which led to behavioral changes in the animals.39
A groundbreaking experiment was performed by transplantation of suprachiasmatic nuclei (SCN) tissue into the third ventricle of SCN-ablated hamsters.40 Ablation of SCN in hamsters leads to disturbed circadian activity rhythms. Following transplantation of SCN tissue, circadian rhythm activities were restored. As axonal outgrowth and synapse formation were prevented by encapsulating the transplanted SCN tissue within a semipermeable polymeric membrane, functional restoration was attributed to a diffusible signal.40
Cerebrospinal fluid composition and turnover thus represent crucial factors in CNS development and function. In physiologic senescence or neurodegenerative diseases like normal-pressure hydrocephalus and Alzheimer disease, CSF turnover rates have been shown to be markedly reduced, leading to a compromised CSF sink function and changes in CSF composition.2,41,42 The changes in CSF resulted in an accumulation of toxic substances such as amyloid-β peptide (Aβ) and might lead to pronounced entry of these substances into the ISF, which might contribute to neuronal dysfunction and degeneration. Concentrations of many extracellular parenchymal substances are regulated locally (eg, at synaptic sites by astrocytic absorption), but as a result of bidirectional CSF-ISF exchange, the CSF additionally influences the extracellular neuronal environment and function of the brain parenchyma.
The CSF also nonsynaptically influences the so-called CSF-contacting neurons that act as sensory cells of the chemoreceptor type, sending their dendritic processes into the brain ventricles.43 Because these neurons send their axons toward telencephalic, mesencephalic, and rhombencephalic brain nuclei, constituents of ventricular CSF not only can influence neuronal activity after penetrating the ISF but also can directly influence the activity of paraventricular brain nuclei that are known to regulate multiple aspects of the autonomic nervous system (ie, hormone release, circadian rhythms, hunger, and thirst). These findings have been substantiated by electrophysiologic experiments showing that paraventricular hypothalamic neurons are highly sensitive to the composition of ventricular CSF.43
Because the CSF is in continuity with Virchow-Robin spaces, substances within the CSF gain access to the CNS parenchyma. Also, immune cells can cross the endothelial wall and enter the Virchow-Robin spaces (Figure). If they, additionally, are able to cross the glia limitans, they gain access to the parenchymal neuropil. This route of invasion seems to play a pivotal role in multiple sclerosis44 and viral encephalitis45 and involves, for instance, CXCL12 protein and the CXCR7 and CXCR4 receptors on CD4+ T cells. There is also evidence for central memory CD4+ T-cell trafficking from the vasculature through the choroid plexus into the CSF. These CSF cells then function to provide CNS immune surveillance within the SAS and inspect the surfaces of antigen-presenting cells.46,47
Thus, the distribution of neuroactive substances and even immune cells to widespread sites of the CNS, including the ISF of the brain parenchyma, is an important assumed role of the CSF. It is called volume transmission (VT),5,48,49 in contrast to the term wiring transmission (WT), which covers classic chemical synapses and electrical synapses (gap junctions). Volume transmission includes endocrinelike and paracrinelike modes of signal transmission that use the ISF and CSF as carriers of information via diffusion and convection.5
Understanding how pathologically altered CSF may influence brain activity in different neurologic diseases or unclear encephalopathies is important for developing new therapeutic strategies to treat these diseases. It would be desirable to directly visualize and dissect the functional impact of normal and pathologic CSF and/or ISF and of isolated, disease-related substances on electrophysiologic activity of the parenchymal compartment under defined conditions. A variety of technologies have been used to investigate the morphologic characteristics and function of single cells or of the entire brain. On a single-cell level, the patch clamp method is widely used to directly measure the activity of distinct ion channels.50 Functional magnetic resonance imaging (fMRI) measures indirect parameters of neural activity such as changes in local cerebral blood flow, and the blood oxygen–level dependence (BOLD) signal measures blood oxygenation.51 These methods permit noninvasive determination of local brain activity under certain paradigms with high spatial resolution in the human brain. Neural activity of synchronized, cortical, and subcortical neurons can be directly visualized using magnetoencephalography (MEG), which detects the small magnetic fields that are produced during neural activity.52
One of the main advantages of these functional imaging technologies is the opportunity they provide to correlate clinical and functional conditions. However, for ethical reasons, the human parenchymal compartment cannot be simply electrophysiologically recorded under experimentally controlled conditions. Thus, a functional in vitro system that represents basic features of the parenchymal compartment that could be exposed to human CSF would supplement the currently available in vivo repertoire. In vitro, single-cell or complex multicellular systems are applicable. Brain disorders often involve pleiotropic mechanisms that originate from disturbed development, changes in signaling pathways, compensation of neuronal network function, or neuronal degeneration.53 Thus, the preferable in vitro system for measuring parenchymal function should go beyond a target-centric approach, which measures the activity of a single cell. A suitable in vitro system would represent an integration of the fundamental aspects of brain activity that are exhibited by functionally important collections of receptors, excitatory and inhibitory neurons of neurotransmitter-specific phenotypes, supporting astrocytes, and extracellular matrix molecules.
The “1 target–1 drug–1 effect” strategy is an oversimplistic model that could benefit from a more complex, but still reductionist, functional measurement of brain activity based on interactive parameters such as intercellular communication and neuronal network activity. This new strategy could supplement the important patch clamp method, which is more specific and can be used to clarify subcellular mechanisms.54 An in vitro system based on a human genetic background could improve the strategy's reliability and predictability with respect to the human in vivo situation. This has been corroborated by investigations on species-specific gene expression patterns in which differences between the human brain and that of other species were found to be pronounced.55 The use of allogeneic, syngeneic, or autogeneic backgrounds would even constitute a step toward individualized medicine. Furthermore, the pharmacokinetic and pharmacodynamic factors that affect animal trials owing to peripheral organ function or the blood-brain barrier and prevent direct measurement of parenchymal function under controlled conditions would be eliminated.
One possible means of implementing such a functional model is provided by the combination of human-induced pluripotent stem cell (iPS)56,57 and microelectrode array (MEA)58 technology. The combination of these methods provides the ability to measure the electrophysiologic network activity of human neural populations comprising astrocytes and a variety of different neuronal subtypes.59 The MEA technology allows the extracellular detection of action potentials simultaneously at multiple spatially separated sites of a neural population in vitro.60 The method goes beyond measuring single-cell activity: it detects the degree and quality of synaptic interactions of neurons expressed by their synchronous and oscillatory bursting activity. The action potentials detected are organized into bursts and are generally comparable to multiunit recordings that can be detected in vivo.
As an important first step toward the development of iPS-derived functional neuronal networks, it was recently shown that murine61,62 and human63 embryonic stem cell (ESC)–derived neural populations are able to generate functional neuronal network activities in successive developmental steps. Embryonic stem cell–derived neural populations spontaneously exhibited synchronous and oscillatory bursting activity that was highly responsive to changes in the extracellular milieu. This was owing to the presence of important ionotropic receptors such as N-methyl- D-aspartate (NMDA), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate and gamma-aminobutyric acid (GABA) receptors in these cell populations. The high sensitivity to extracellular changes was also due to the presence of excitatory glutamatergic and inhibitory GABAergic neurons, astrocytes, and extracellular matrix molecules. Thus, the network activity observed in these cell populations results from a complex interaction of a multitude of factors that converge into an integrated functional entity.
Because these qualities recapitulate basic parenchymal features of cellular diversity and connectivity, we propose that this in vitro system models, up to a certain point, parenchymal function. Furthermore, as ESC-derived populations can be differentiated into different neurotransmitter-specific cell types and site-specific neural phenotypes under controlled conditions,64,65 their electrophysiologic activities represent a reductionist functional image of distinct brain areas that use variable sets of receptors and neuronal subtypes. To date, most data on ESC-MEA hybrids have been collected on a murine background.61,62 However, we (unpublished observations) and others63 have successfully transferred this technology to the native human ESC background, which will further be adapted using human iPS cells.
In the murine ESC-MEA hybrid system, morphologically interconnected neural populations that possess similarities to the parenchymal compartment have exhibited synchronous burst behaviors under the influence of pure human CSF.66 Compared with artificial CSF, CSF specimens derived from healthy human controls significantly increased global activity parameters such as spike and burst rates and the synchrony of oscillating neuronal networks. These experiments showed for the first time that the composition of human CSF supports complex neuronal network function.
Additional experiments were performed66 to clarify whether CSF collected from neurologic patients negatively influenced neuronal network activity. In this study, CSF specimens were collected from patients after acute traumatic brain injury (TBI). Compared with control CSF specimens, TBI CSF significantly suppressed in vitro neuronal network activity (ivNNA). By biochemical separation of various components of the CSF, we demonstrated that a CSF fraction comprising small amino acids including glutamate, glycine, alanine, and serine was responsible for this effect.
Interestingly, the suppression of network activity by CSF from injured patients was partially counteracted by glutamate and glycine receptor antagonists, showing that this system may help identify therapeutic approaches to acute encephalopathies following TBI. With these experiments, we showed for the first time that physiologic CSF supports neuronal network function. In contrast, pathologic CSF specimens negatively influenced neuronal network activity, and this negative influence was partially corrected by pharmacologic intervention. This study further showed that the extent of neuronal network suppression correlated, to some degree, with values of the corresponding Glasgow Coma Scales, demonstrating a correlation of in vitro results and the patients' clinical condition. However, these data need to be confirmed in larger cohorts.
The following 4 conclusions emerge: (1) CSF and ISF exchange across the ependymal cell layer and the pia-glial membrane; (2) the composition of the CSF affects CNS development and neurogenesis; (3) the composition of the CSF influences behavior via volume transmission; and (4) pathologically altered CSF from patients with acute TBI affects ivNNA differently than does human control CSF.
It is reasonable to ask whether the human parenchymal compartment of patients with TBI remains functionally unaffected in the presence of CSF that exerts significant suppressive influences on ivNNA. We hypothesize that, secondary to acute brain disease, pathologically altered CSF can itself directly influence brain activity and behavior. This hypothesis might also be valid for chronic neurologic diseases. The observation that CSF specimens collected approximately 1 week after TBI suppressed ivNNA points to the possibility that TBI may be accompanied by a CSF-mediated acute encephalopathy. According to this idea, a focal TBI-related CNS lesion might initially cause CSF alterations via a local breakdown of the blood-brain barrier. Substances that are normally not present in the CSF or that are normally present only at low concentrations would then be carried across endothelial and pia-glial membranes and into the parenchymal compartment via volume transmission to widespread sites of the CNS, resulting in functional disturbances. Brain activity might also be disturbed via CSF-contacting neurons that receive pathologic CSF signals. The existence of a CSF-mediated acute encephalopathy is also supported by much lower Glasgow Coma Scales in patients with TBI; these scores are not always thoroughly explained by the observed structural damage.
The relevance of pathologically altered CSF would change if it were shown that TBI is accompanied by CSF-mediated acute encephalopathy. It would go beyond passively indicating brain diseases. By distributing neuroactive substances, CSF could actively evoke acute or chronic functional disturbances of global brain activity. This mechanism could be called volume transmission-mediated encephalopathy (VTE). It would then make sense to search for possible VTEs at different times after onset of the primary insult, especially if the primary lesion or disease cannot fully explain the clinical condition. Also, because inflammatory signals and immune cells are spread by CSF, subarachnoidal spaces, and Virchow-Robin spaces, immunologic diseases like multiple sclerosis or viral infections, which commonly attack extended parts of the brain, similarly fit into the concept of VTEs.
If the effects of CSF on parenchymal function can be recorded and dissected by an integrative, reliable, and predictive in vitro model based on a human genetic background, the development of new therapeutic options that specifically counteract malfunction-causing substances within the CSF could fundamentally be supported. Unpublished results of our group demonstrate that this concept can also be applied to CSF alterations in the course of NMDA receptor encephalitis in which anti-NMDA receptor autoantibodies cause a defined clinical syndrome and suppress ivNNA. Since also ammonia, which is assumed to play a pivotal role in hepatic encephalopathy, influences ivNNA (unpublished results), the original concept of functional TBI-related CSF alterations can be broadened to include other neurologic diseases. Thus, the functional impact on ivNNA of antibodies or cytokines released into the CSF by immune cells during immune surveillance or inflammation can also be directly measured. Functional CSF parameters recorded by MEAs would then not only represent passive indicators of disease but also direct and functionally relevant measures of VTEs that can serve to clarify and monitor the influence of CSF on brain function.
Correspondence: Marcel Dihné, MD, Medical Faculty, Department of Neurology, University of Duesseldorf, Moorenstrasse 5, Duesseldorf, 40225, Germany (firstname.lastname@example.org).
Accepted for Publication: June 17, 2011.
Author Contributions:Study concept and design: Hartung and Dihné. Drafting of the manuscript: Dihné. Critical revision of the manuscript for important intellectual content: Hartung and Dihné.
Financial Disclosure: None reported.
Funding/Support: The ideas illustrated in this work have been supported by grant 0315641B from the Bundesministerium für Bildung und Forschung, Euro-Trans-Bio 4.