Molecular anatomy and functions of the choroidal blood-cerebrospinal fluid barrier in health and disease

The choroid plexuses form an interface between two circulating fluids, the blood and the CSF. All choroid plexuses are constituted by a cuboidal epithelium with numerous villi that ensheathes a connective stroma. Blood vessels penetrate this stroma and form large fenestrated capillary loops juxtaposed to the tight epithelium that constitutes the blood-CSF barrier (BCSFB) per se. Different types of immune cells are associated with the choroid plexuses, both in the stroma and adhering to the CSF-facing membrane of the epithelial cells. Three factors favor the exchanges between the blood and the CSF across the choroid plexuses. First, the choroidal local blood flow is very high, around 4 ml g⁻¹ min⁻¹, actually the highest found in the brain [153]. This results from an extensive network of wide and anastomosed capillaries throughout the choroidal stroma (Fig. 1a). Second, the choroidal endothelium that forms the first cellular layer between the blood and CSF (Fig. 1b) is highly permissive to a large range of molecules. The extent to which the paracellular pathway at this endothelium is sealed remains unclear (see infra), but the diaphragmed fenestrations that are landmarks of the choroidal capillary loops (Fig. 1c) allow a facilitated diffusion into the stroma of molecules with molecular weight up to ~ 800 kDa (reviewed in [149]). The formation and maintenance of the fenestrations and their diaphragms involve the plasmalemma vesicle associated protein (Plvap) gene [143] and are under the control of the choroidal epithelium via a basolateral secretion of vascular endothelial growth factor (VEGF). Neutralizing choroidal VEGF production leads to strong morphological and functional alterations of the choroidal vasculature, with presumable consequences on CSF secretion and composition [86]. This mechanism may underlie the central side effect of anti-VEGF antibodies observed in some patients under antitumoral therapy. Third, the choroidal epithelial cell membranes display basolateral infoldings and apical microvilli (Fig. 1b, d) that greatly increase the surface area available for transfer between the stroma and the CSF [66]. Extrapolating rodent surface measurements to the adult human, the apical surface developed by the epithelium of the four choroid plexuses represents 10% of the surface area developed by the brain microvasculature forming the blood–brain barrier (BBB). The exchanges across the choroidal epithelium are, however, highly regulated. They can be facilitated for various nutrients, micronutrients, hormones, polypeptides, or restricted for many other compounds, by the combined action of tight junctions, transport processes and metabolic enzymes. This implies high-energy requirements as illustrated by the presence of numerous mitochondria in the choroidal epithelial cells. Thus, the BCSFB should be viewed as an active interface rather than a simple barrier. Such an array of transport regulatory mechanisms would be rather unexpected if CSF was only fulfilling the usually cited buoyancy role and sewer function for brain metabolites. It makes more sense in the light of the current understanding of CSF functions, which go far beyond a drainage function.

The CSF is contained within the boundaries formed by the barriers separating the CNS from the blood. This fluid is, therefore, an integral part of the CNS, and is renewed several times a day. It circulates through the ventricular system into the vela such as the velum interpositum, the internal cisternal spaces such as the ambient cistern, and into the external subarachnoid spaces surrounding the brain and spinal cord. These cisternal and subarachnoid spaces, illustrated in [48, 129], are characterized by a complex leptomeningeal membrane network in which blood vessels are running. These CSF spaces are connected to the perivascular spaces (PVS) surrounding the vessels penetrating the brain parenchyma. The anatomical organization of this CSF-PVS connection differs according to the vessel type (veins versus arteries) and the brain area. In particular, one or two cell layers of leptomeningeal origin extend along the penetrating arteries, and complexify CSF/perivascular fluid exchanges [171]. Exchanges of solutes between the CSF and the extracellular fluid occur also across the permissive ependyma, and across the less permissive glia limitans bordering the outer boundaries of the brain [48]. Extensive deep brain parenchyma diffusion of a substance is not expected following intraventricular or intrathecal bolus infusion. The pattern of brain penetration from the CSF can, however, be much broader when considering substances continuously secreted by the choroid plexuses into the CSF and associated perivascular fluid. An example is transthyretin, a carrier protein involved in the regulation of thyroid hormone concentration in the extracellular fluids. In the brain, transthyretin is uniquely secreted by choroid plexus epithelial cells into the CSF in a continuous manner and throughout life [117], allowing long-term diffusion within the neuropil. Periventricular cellular uptake from CSF followed by retrograde transport also occurs, as exemplified for manganese [4, 12]. Circumventricular organs (CVOs) which are either sensing or secreting organs interacting with the peripheral blood are juxtaposed to CSF spaces and a specific CSF-CVO dialogue takes place across the tanycytes bordering the CSF side of CVOs [65, 95].

The functions of the CSF are, therefore, multiple. They include the overall drainage of brain metabolites diffusing into the CSF, inorganic ion osmoregulation, and pH buffering for the brain extracellular fluid [108]. They also include a role in volume transmission of different biologically active compounds and micronutrients within the brain, a function that is particularly relevant to brain development. The proportion of CSF is especially large during brain development [46], stem cells are lining the ventricles, and the CSF bears various trophic and differentiation factors, guidance molecules, some of them secreted by the choroid plexus cells themselves (see infra). This volume transmission function remains significant in adulthood for the maintenance of brain homeostasis, and likely for brain repair. At the ventricular level, it benefits from an organized pattern of cilia modules whose beating controls substance distribution along the ependymal wall [43].

Understanding the immune privilege of the CNS requires appropriate consideration of CNS anatomy, especially regarding the localization and the different functions of the respective endothelial, epithelial and glial brain barriers, as these establish various CNS compartments that significantly differ in their communication with the peripheral immune system [40]. Taking brain barriers into account, one can appreciate that the CNS parenchyma proper is an immune privileged site according to its original definition, namely a site in which experimentally implanted tissue allografts are incapable of provoking an immune response leading to their rejection. The immune privileged status of the CNS parenchyma has been further substantiated by observations that viral antigens introduced into the CNS parenchyma evade immunological recognition. In fact, the CNS parenchyma also lacks the ability to mount a rapid innate immune response when exposed to bacterial products or experimentally induced cell death, which in peripheral organs triggers the rapid infiltration of neutrophils and macrophages (summarized in [40]). The afferent arm of the immune system, i.e. immune cell entry into the CNS parenchyma, is prohibited by the glia limitans surrounding the entire CNS parenchyma. The efferent arm of the CNS parenchymal immune system is limited as the resident tissue macrophages, the microglial cells, do not migrate to peripheral draining lymph nodes to present antigens, and thus only CNS parenchymal antigens reach the draining lymph nodes via interstitial fluid [37]. The CNS parenchyma proper is thus protected from the immune system. This is very different from the ventricular and subarachnoid CSF spaces, which do not exhibit the same immune privileged status as the CNS parenchyma proper. Allografts transplanted directly into cerebral ventricles are readily rejected (summarized in [40]). Furthermore, the CSF drains along lymphatic vessels that pass through the cribriform plate and the dura mater into deep cervical lymph nodes (summarized in [40]). This pathway also allows for the trafficking of immune cells from the CSF space to CNS draining lymph nodes, thereby establishing fully functional afferent connections of the immune system with the CNS.

The immune privilege of the CNS is thus established by its specific morphological architecture, which we have previously proposed to resemble the architecture of a medieval castle that is surrounded by two walls embracing a castle moat [38]. Using this comparison, the epithelial choroidal epithelium and the arachnoid membrane establish the outer wall of the castle, which in the case of peripheral danger followed by peripheral immune cell activation, can be breached by activated immune cells in search of their specific antigen in the CNS [129]. Within the CSF-drained space—the “castle moat”—the immune cells encounter antigen-presenting cells—“castle-moat guards”—, which are exposed to CNS antigens within the CSF-drained subarachoid space and the ventricles. If immune cells encounter their specific antigen on these “castle moat” antigen presenting cells, they will be reactivated allowing them to cross the second wall made of the perivascular and superficial glia limitans, and subsequent entry into the CNS parenchyma, the “castle” proper [1] to fight against dangers and infections.

In light of these well-recognized and newly documented functions of the CSF, the choroid plexuses located upstream of the CSF circulation become important actors of both molecular and cellular exchanges between the blood and the CNS, and a relevant source of biologically active molecules for the developing and adult brain. They appear as main CNS sensors for insults of both peripheral and central origin. Their role in protecting the CSF against homeostatic changes and toxic challenges also becomes crucial to prevent dysfunction of volume transmission and immune surveillance within the brain (Fig. 2). Our current understanding of these choroid plexus functions are described in the next sections.

Owing to the “classical” buoyancy and drainage functions attributed to the CSF, its secretion can be considered as the first of the neuroprotective properties attributed to the choroid plexus. The historical demonstration that the choroid plexus is the major CSF secretion site in the brain was based on (1) the increased ventricular volume after foramen occlusion that was prevented by plexectomy, (2) the similar composition of choroid plexus-secreted fluid collected under oil in situ as compared to CSF, and (3) the existence of an arterio-venous positive concentration gradient for hematocrit and for polar tracer in choroidal blood (reviewed in [32]). Since these pioneering findings, the molecular mechanisms underlying CSF secretion by the choroidal epithelium have been largely deciphered.

The current model of CSF secretion is described in [47, 108]. A unidirectional active transcellular blood-to-CSF flux of Na⁺, Cl⁻, and HCO₃⁻ creates an osmotic gradient followed by water movement. This active ionic flux results from the activity of carbonic anhydrase and Na⁺/K⁺-ATPase, the latter enzyme being located on the apical membrane, a unique feature among all epithelia in the body. A large number of channels and transporters for monovalent inorganic ions, and aquaporin-1 participate in the secretory process. A number of issues relative to the mechanisms of CSF secretion remain, however, unresolved, among them the potential implication of the pore-forming claudin-2 either in CSF-to-blood paracellular K⁺ transfer, or as a water channel ([108], and see below).

Dysfunction of choroidal CSF secretion could be part of the pathophysiological process leading to neurodegenerative diseases. Clinical observations showing that the choroid plexus-CSF system is altered in Alzheimer’s disease support this hypothesis [134]. An alteration of choroid plexus vascular and epithelial morphology, a decrease in choroidal synthesis and transport capacity, and a decrease in CSF secretion is observed in aging, and is exacerbated in Alzheimer’s disease [132, 134]. The CSF circulation plays a role in amyloid beta distribution and clearance from the brain [49]. An impaired drainage of the brain induced by a decreased CSF secretion may, therefore, favor amyloid deposition. In addition, dysfunction of amyloid beta choroidal clearance through one or several low density lipoprotein receptor-related proteins (LRPs) or a decreased secretion into the CSF of polypeptides such as transthyretin that bind and stabilize soluble amyloid beta [157] can sustain the pathology. The reason why the choroid plexuses are damaged and dysfunction in Alzheimer’s disease is unclear, but may involve amyloid beta accumulation within the choroidal tissue. The hypothesis that a global choroid plexus-CSF system dysfunction favors neurodegeneration is supported by experimental work using Alzheimer’s disease mouse models [53].

As in all epithelia, the structural barrier properties of the choroidal epithelium are determined by the tight junctions. They form at the most apical part of the lateral space between neighboring cells (Fig. 1b). They control paracellular movements and guarantee the cell polarity by impeding the diffusion of transmembrane proteins between the apical and the basolateral membrane domains. Intercellular tight junctions are formed by a combination of transmembrane proteins, namely claudins and occludin, which interact with the F-actin cytoskeleton through various scaffold cytoplasmic proteins [172]. Claudins, which are actually responsible for controlling the paracellular permeability, are of two types. Barrier forming claudins seal the paracellular pathway, and channel forming claudins allow the charge- and size selective passage of electrolytes and small solutes. The claudin repertoire expressed in a given barrier tissue and the way these proteins dimerize between adjacent cells, confer to this barrier its intrinsic properties in term of permselectivity and impermeability. A thorough analysis of claudin transcripts in rat choroid plexus highlighted three major claudins expressed at high levels, Cldn-1, -2 and -3 [74]. The distribution of these proteins is restricted to the apical tight junctions in the choroidal epithelial cells. Claudin-1 and -3 are typical of “tight” epithelia, and are impermeable to ions, water, and larger solutes [121]. Claudin-2 is typical of “leaky” epithelia with secretory functions like the kidney proximal tubule. It forms a paracellular channel selective for small cations (K⁺, Na⁺) and is also the only channel forming claudin that cotransports water [120]. It is sufficient to induce a low transepithelial resistance in the highly resistant MDCK C7 cell line, without affecting the paracellular diffusion of non-charged solutes. The expression of claudin-2 in the BCSFB brings some understanding to the fact that electrophysiologists classified the choroid plexus in the “leaky” epithelium category. The lack of claudin-2 expression in the BBB presumably explains the large difference in electrical resistances measured in the two interfaces. The paracellular movement of water through the channel forming claudin-2 in choroidal cells may also explain, at least partially, why knocking out aquaporin-1 in mouse decreased CSF secretion only by 35% [103]. The basis for having both a transcellular water pathway through aquaporin and a paracellular water pathway through claudin-2 could be to provide an energy saving mechanism, as shown in the renal proximal tubule [106].

Claudin expression is developmentally regulated, which may reflect physiological adaptation to microenviromental changes during brain development. Similar data concerning claudin-1, -2, and -3 ontogenic regulation were reported for rat, mouse, and human choroid plexus [74, 146]. Importantly, the barrier forming claudin-1 and -3 appear very rapidly after the choroid plexuses start to bulge in the ventricles and differentiate (Fig. 4a). A functional correlate has been established by permeability studies at early embryonic stages [80]. Claudin-2 expression appears to follow the developmental increase in CSF secretion capacity. Besides these three major claudins, a number of other barrier forming and channel forming claudins are expressed in the choroid plexus [74], and some of them, such as claudin-9 and -19 have been immunodetected in apical junctions of the rat choroidal epithelium. Claudin-5, the prototypical BBB claudin, distributes at the junctions in large penetrating stromal vessels in rat and mouse and possibly in some smaller vessels [74, 82, 146]. Its expression in the choroidal epithelium was not detected in rat and mouse, but was recently described in human neonates [164]. Claudin-11 was immunodetected at epithelial junctions in mouse choroid plexus [166]. These data need to be confirmed because the antibody used was later found to cross-react with claudin-10 [99], also expressed at the choroid plexus [74]. The peculiar parallel organization of tight junction strands in the choroidal epithelial barrier, initially attributed to claudin-11 [166], could alternatively be related to the choroidal expression of claudin-19, which also forms tight junctions with parallel strands in Schwann cells of the peripheral nervous system [94]. The permeability at tricellular junctions, which form at the meeting point of three neighboring cells, is regulated by tricellulin. This transmembrane protein mainly restricts the paracellular movement of large molecules. It has little effect on ion diffusion, given the very limited area of exchange provided by the tricellular junctions in comparison to the length of intercellular lateral tight junctions, at which ion control is effective. Tricellulin has been detected at tricellular junctions in the mouse choroidal epithelium [61]. A new partner critical for the apical formation of tight junctions in choroid plexus is the protein Alix, involved in endocytosis and endosome sorting processes, and expressed at high levels in the BCSFB. Alix has a critical role in promoting the formation of apical tight junction complexes. Absence of Alix leads to a total disorganization of the choroid plexus tight junctions and to hydrocephalus [15].

In peripheral organs, claudins as well as tricellulin are regulated by inflammation, oxidative stress, or ischemia/hypoxia. A loss of claudin-1, -2 or -3 was observed in the choroidal epithelium in animals with experimental autoimmune encephalomyelitis or in multiple sclerosis patients [71, 166]. The functional consequences of these changes on the BCSFB permeability to paracellular markers have not been explored. By contrast, a functional loss of the choroidal barrier properties, assessed by measuring inulin permeation was observed during ischemia/reperfusion in the rat [42]. It suggests that tight junctions are altered by the injury, but alterations in claudin expression and localization were not examined. Tricellulin may be a target in lead-induced cerebral toxicity. The metal accumulates in the choroid plexus in humans and in animal models, leading to a dysregulation of TTR production and thyroid hormone transport. It may also affect tricellulin expression and functionality by inducing the level of the regulating microRNA mir203, as suggested by in vitro studies [152].

The BCSFB harbors a versatile machinery that protects the brain and CSF against a wide spectrum of xeno- and endobiotics. This machinery uses efflux systems, which belong to the ATP-binding cassette (ABC) transporter family and distinctive solute carrier (SLC) subfamilies, and are selectively distributed between the apical and basolateral membrane domains of the choroidal epithelial cells. These transporters either restrict the penetration of compounds into the CNS or favor the clearance of potentially deleterious metabolites from the CSF. The neuroprotective machinery also comprises a panel of metabolizing enzymes, which generally inactivate their substrates and work in concert with efflux transporters to prevent the intracerebral accumulation of noxious molecules (Fig. 4b).

Specifically, the choroidal energy-dependent ABC transporters ABCC1 and ABCC4, located at the blood facing membrane (basolateral side) can extrude a wide range of structurally unrelated lipid soluble compounds following their passive diffusion within the epithelial cells. These compounds include various environmental pollutants and drugs from diverse therapeutical classes (anti-tumoral, anti-epileptic, anti-infectious, anti-depressant, immunosuppressant (reviewed in [149]). The transporter ABCC1 may also restrict the penetration of plasma unconjugated bilirubin into CSF [45], and thus contribute to protect the brain in case of moderate transient hyperbilirubinemia, such as occurring in neonatal jaundice. The expression in the choroidal epithelium of ABCB1 and ABCG2, two hallmark transporters of the BBB, is debated [44, 91, 119]. Given their supposedly apical/subapical localization, these proteins would accelerate the transfer of deleterious compounds into the CSF, and further studies are awaited to clarify these inconsistencies. Multispecific SLC transporters are located at the apical membrane, with the exception of Slco1a4 present in the basolateral domain. Slco1a5 (in rodent) and SLC22A8 at the CSF-facing membrane function as inwardly directed transporters and mediate the cellular uptake of CSF-borne molecules. They transport various drugs such as antiviral, antibiotic, and non-steroidal anti-inflammatory agents. They also transport endogenously produced hormones (estrone sulfate, dehydroepiandrosterone sulfate) and inflammatory modulators (prostaglandins, leukotrienes) (reviewed in [149]). More recently, other carriers of the SLC family, relevant to efflux processes, have been identified in the BCSFB, such as the multidrug and toxin extruder SLC47A1, or else the organic cation transporter SLC29A4 [159, 160]. This last efflux system transports various monoamine neurotransmitters, serotonin, dopamine, and histamine, as well as 1-methyl-4-phenylpyridinium-like neurotoxins. It is located at the apical membrane of the choroidal epithelial cells. By removing monoamines from CSF, SLC29A4 contributes to terminate central signaling of neurotransmitters, and in the case of histamine, the transporter could play a key role in sleep–wake regulation [160]. Di-and tripeptides are specifically eliminated from the CNS across the BCSFB by SLC15A2 located at the apical membrane [139]. Species-dependent variations in the identity of certain efflux transporters have been highlighted by a recent proteomic study on rodent and human choroid plexus. This study suggests that there may be differences in the abundance of some transporters, SLC15A2 or SLC47A1 for example, between rat and human choroid plexus [159]. The human data, based on the analysis of a unique sample collected from an aged individual need to be validated, and the functional consequences of species differences remain to be delineated as the efflux systems have usually large and partially overlapping substrate specificities.

To complement this specific set of efflux transporters, the choroid plexuses express a large panel of metabolizing enzymes, which transform endogenous molecules, drugs, or toxicants into water-soluble and usually inactive metabolites. The BCSFB likely represents the major site of xenobiotic metabolism in the brain [149]. High specific activities have been measured for epoxide hydrolase and various conjugating enzymes, which catalyze the addition of hydrophilic moieties such as glucuronic acid, glutathione or sulfate. Growing evidence shows that these conjugating enzymes coupled to the ABC or SLC efflux transporters located at the basolateral membrane of choroidal epithelial cells form an efficient metabolic barrier, which limits the penetration of xenobiotic substrates into the CSF [50, 148]. High glutathione-S-transferase and sulfotransferase activities, as well as high ABCC expression levels were measured in human choroid plexuses [44, 50, 116], suggesting that similar barrier mechanisms occur in human. Such an enzymatic barrier process also exists for endogenous molecules, like monoamine neurotransmitters, which intracerebral concentration needs to be strictly controlled. The choroid plexus has a high monoamine oxidase B activity, which prevents the passage of circulating dopamine and may potentially assist in the breakdown of amines cleared from the CSF [81]. It is tempting to extend this mechanism to another enzyme, the 15-hydroxyprostaglandin dehydrogenase specifically expressed in the choroid plexus, especially in early postnatal life, and mediating prostaglandin E2 catabolism [2]. Finally, several antioxidant enzymes, catalase, glutathione peroxidases, and superoxide dismutase are highly active in the choroid plexus, providing a protective mechanism against reactive oxygen species and organic peroxides [158].

These choroidal efflux transporters and metabolizing enzymes are for a large majority expressed at already substantial levels during the perinatal period, while the BBB is less mature with respect to neuroprotection. This is in line with an early maturation of the choroid plexuses during development and suggests that the BCSFB plays a predominant role in brain protection in the neonate and young infant [73, 150].

The transport and metabolic barrier mechanisms can be deregulated in different CNS diseases such as ischemia, inflammation, and infection, with potential consequences on the resolution of these pathologies. The BCSFB is involved in the continuous removal of prostaglandin E2 from the CSF, keeping low the intracerebral levels of this proinflammatory mediator [156]. A downregulation of prostaglandin E2 efflux transporters occurs in the models of bacterial meningitis and viral infection, and may exacerbate the effects of neuroinflammation [57, 67]. Chronic moderate hyperbilirubinemia depresses ABCC1 levels, which may actually potentiate bilirubin neurotoxicity [45]. Of importance, cytoprotective transporters and enzymes are under the transcriptional regulation of nuclear factors, which are expressed in the choroid plexuses [73]. The nuclear factor erythroid 2-related factor 2 pathway, in particular, represents an attractive target to protect the BCSFB and induce its antioxidant capacities in neurological disorders associated with oxidative stress [168].

A central role of the choroid plexuses is the production of bulk CSF. Importantly the choroid plexuses also contribute to the CSF composition by serving as an entry route for blood-derived factors into the brain across the BCSFB, and by actively secreting a wide variety of factors.

Molecules are transported in a highly regulated manner across the choroid plexuses, via facilitated or active transport or by transcytosis of ligands bound to receptors. These include inorganic ions, for maintaining ionic balance under homeostasis, small organic compounds, essential metabolic nutrients (glucose and amino acids), hormones and micronutrients (vitamin C), polypeptides and proteins [149].

The choroid plexus itself is an important source of diverse molecules including active signaling, trophic and guidance molecules. Bioactive polypeptides, cytokines, extracellular matrix, carriers, growth factors, and hormones are synthesized and released from the choroid plexuses into the CSF. These factors can reach distal areas of the brain via bulk flow of the CSF and can act in both an autocrine and paracrine manner [63, 137]. In addition to diffusible factors, exosomes containing signaling factors and non-coding (nc) RNAs, such as long ncRNAs and miRNAs, are present in the CSF [11] and can be choroid plexus-derived [6, 55]. Finally, the choroid plexuses also contain a large array of enzymes and act as an enzymatic or metabolic barrier, by modulating the activity and availability of neurotransmitters, peptides, growth factors, hormones, drugs and xenobiotics. Interestingly, the repertoire of factors secreted by choroid plexuses is different in each brain ventricle [85], suggesting that each choroid plexus may act as a local center for the secretion and processing of distinct pools of active factors.

In addition to its key role in brain homeostasis, recent work has shown that the CSF and choroid plexus-derived factors are important, stage-specific regulators of brain developmental processes and adult neural stem niches. Importantly, the CSF composition and protein concentration change throughout development and in the adult, as well as in aging, and can, therefore, exert distinct effects at different stages during development and in the adult (Fig. 5a, b).

With its position at the interface between the blood and the CSF, the choroid plexuses are uniquely placed to sense physiological states and integrate local and long-range signals from the circulation, nervous system, and immune system [9, 88] (Fig. 5c). Indeed, the choroid plexus transcriptome, proteome, and secretome are dynamic and rapidly change in response to external cues, such as inflammation [89, 155]. Choroid plexus epithelial cells have many villi, providing a large surface area for detection of signals in the CSF. They also have multiple primary cilia, which can sense and transduce signals from the brain environment. Epithelial cells also express a large repertoire of receptors for diverse molecules, including neurotransmitters, hormones, cytokines, growth factors, and bacterial toxins. As such, they can sense signals in both the CSF and blood, both of which can differentially affect the transcriptome and secretome of choroid plexus epithelial cells [9]. How other choroid plexus cell types respond to external inputs from the circulatory, autonomic, and immune systems and in turn change their secretome is still unknown.

Many physiological processes in mammals are influenced by both circadian and seasonal rhythms. The CSF composition exhibits circadian and seasonal fluctuations, which are at least in part due to dynamic changes in the choroid plexus [112, 138]. As outlined above, the choroid plexuses can modulate physiologiocal states, such as aging [9, 137]. The choroid plexuses are also targets for sex hormones, and can, therefore, elicit gender-specific effects [124] with important implications for physiology. The choroid plexus, via its transport function and secretory activity modulates neuroendocrine processes [138], providing a path to deliver physiological signals, such as leptin and thyroid hormone (T4), from peripheral blood to different brain areas including the hypothalamus and brain parenchyma via the CSF. It also directly synthesizes hormones and neuropeptides (including growth hormone, vasopressin, and insulin-like growth factor 1), as well as peptide processing enzymes (including for proopiomelanocortin and insulin) and modifying enzymes (such as prostaglandin D2 synthase), which can affect adult V-SVZ neural stem cells [105, 137], and impact global physiological states.

In sum, the choroid plexuses are emerging as a key sensor and modulator of physiological states by acting as a hub that translates physiological changes into molecular signals directly affecting adult neural stem cell behaviour, neurogenesis and global physiology (Fig. 5c). As such, changes in choroid plexus function can potentially lead to a variety of pathological conditions.

The CNS parenchyma of healthy individuals harbors yolk sac-derived long-lived microglial cells [109] but is devoid of immune cells from the adaptive immune system such as T cells or B cells. T cell mediated CNS immune surveillance is rather strictly limited to the CSF-filled ventricles and subarachnoid spaces. This is supported by the observation that in healthy individuals 150,000–750,000 immune cells are present in the CSF [68] and they are equally distributed between the ventricular and lumbar CSF [110]. Other sterile bodily fluids, e.g. synovial or pleural fluid contain 10–100-fold more immune cells than CSF, but the majority is composed of innate immune cells rather than T cells or B cells. In contrast, the few immune cells circulating in the CSF are mostly T cells with a ratio of CD4⁺ T cells to CD8⁺ T cells of about 3.5 to 1 (summarized in [113]). The population of T cells found in the CSF of the ventricles and subarachnoid space contains mostly memory T cells that have characteristics of central memory T cells (T_CM cells) and effector memory T cells (T_EM cells) [33, 123]. Animal models have provided additional evidence for the presence of yet another subset of memory T cells in the CNS that do not recirculate but rather persist in the CSF-drained spaces after previous infections [165]. These T cells are referred to as CD103⁺ tissue resident memory T cells (T_RM cells) and provide an organ-autonomous defence system for the CNS for future infections [145].

Considering the restricted localization of these memory T cell subsets in the CSF compartment within the CNS it appears highly likely that the cells forming interfaces between the blood and the CSF play an important role in regulating T cell entry into the CNS. While there is no direct evidence for immune cell entry from the dura mater across the arachnoid membrane into the subarachnoid space, accumulating evidence points to immune cell entry into the CNS ventricles via the choroid plexuses [36, 101, 114, 129]. A key function of both the arachnoid and the choroid plexuses in regulating CNS immunity is, however, suggested by the high amount of MHC class II expressing myeloid cells found strategically positioned on either side of the barriers. On the blood facing side of the arachnoid membrane, the dura mater harbors a high number of macrophages and dendritic cells [24]. On the CSF side of the arachnoid the subarachnoid space is home to numerous tissue resident MHC class II expressing macrophages, some of them even in tight contact with the arachnoid barrier and perforating the arachnoid lining at the CSF side with their processes [13]. Subarachnoid macrophages exert scavenger functions and present CNS antigens to T cells entering this CSF space as shown by elegant in vivo live cell imaging studies [8, 84, 107]. Combined parabiosis and fate-mapping studies in transgenic mice provided recent evidence that subarachnoid macrophages arise from hematopoietic precursors during embryonic development and establish a stable long-lived population of innate immune cells in this compartment [52]. These myeloid cells share this origin with the Kolmer’s epiplexus cells that reside on the apical—i.e. CSF—side of the choroid plexus epithelium, and have first been described by Kolmer [70]. As the choroid plexuses are located in strategically important places in each ventricle, namely the narrow foraminae through which the CSF passes from the lateral ventricles to the third ventricle and from the fourth ventricle to the cisterna magna and to basal cisterns of the brain, it is tempting to speculate that Kolmer’s epiplexus cells serve similar functions as the subarachnoid macrophages as they are in a prime position to scavenge the CSF for antigens (Fig. 6). The unique characteristics and especially the previously unrecognized longevity of subarachnoid macrophages and Kolmer’s epiplexus cells may contribute to shaping CNS immune surveillance in a unique fashion as these antigen-presenting cells do not readily leave the CSF space towards the peripheral draining lymph nodes.

An important immunological function of the choroid plexus in its entity has long been proposed [36, 101, 144] and has been underscored by numerous descriptions of the high numbers of MHC class II expressing bone marrow-derived macrophages and dendritic cells [101, 133] residing in the highly vascularized choroid plexus stroma that is exposed to blood component due to the lack of an endothelial barrier (Fig. 6). As discussed above these myeloid cells are distinct from the myeloid subsets in the CSF spaces. Interestingly, choroid plexus stroma dendritic cells can also reside between choroid plexus epithelial cells and extend their dendrites into the CSF filled ventricles [133], where they might take up CSF antigens for presentation to T cells in the choroid plexus stroma [101].

This is supported by the observation that T cells found in the choroid plexus stroma are enriched for CNS-specific effector/memory CD4⁺ T cells [10] (Fig. 6). In the absence of neuroinflammation local activation of CNS-specific Th1 and Th2 cells induces expression of the Th1 and Th2 signature cytokines, IFN-γ and IL-4, which may maintain expression of trafficking molecules at the choroid plexus allowing for limited T cell trafficking into the ventricles [10, 75]. The choroidal endothelium, fenestrated and thus permissive to large molecules, does not display a BBB phenotype. Yet T cells need to adhere and get through this endothelium to reach the choroidal stroma, and then the CSF across the epithelium. The mechanisms allowing this two-step migration process remains to be fully elucidated (Fig. 6).

At the same time choroid plexus dendritic cells can promote immunological T cell silencing [19] via secretion of IL-10 [133]. The immunoregulatory function of the choroid plexuses has further been supported by local expression of complement proteins and their regulators (summarized in [41]). Overall, in absence of neuroinflammation the choroid plexuses thus seem to play an underestimated role in maintaining CNS immunity and may be critical structures ensuring local activation of CNS-specific T cells and their controlled subsequent migration across the choroidal epithelium allowing them to enter into the CSF niches of the CNS.

The laboratory of Michal Schwartz has coined the term “protective autoimmunity” [96] to develop this concept even further by including their experimental observations supporting the notion that presence of CNS-specific T cells in the CSF contributes to life-long protection, maintenance and repair of the CNS. This concept suggests that suboptimal trafficking of immune cells into the CNS via the choroid plexuses might be an underlying mechanism shared in the pathophysiology of neurodegenerative conditions [130]. The proposed underlying mechanism is a balance between type I and type II interferon signaling which may regulate immune cell entry into the CNS with important consequences for CNS physiology or pathology [34].

In neurological disorders as different as brain trauma [155], perinatal inflammation triggered by bacterial components [151], and experimental autoimmune encephalomyelitis (EAE) [41], the balanced cytokine and chemokine milieu within the choroid plexuses changes rapidly and may thus directly contribute to altered immune cell activation and their subsequent trafficking across the inflamed BCSFB into the CNS. Increased numbers of T cells and B cells are found in the CSF of patients with CNS infections or multiple sclerosis [3, 16].

A large body of our current knowledge on immune cell trafficking to the CNS is derived from studies in EAE, an animal model for multiple sclerosis. In EAE, activated autoaggressive CD4⁺ T cells enter the CNS where they cause neuroinflammation, demyelination and BBB breakdown, which resembles the pathological signs observed in multiple sclerosis. EAE is accompanied by significant morphological changes of the choroid plexuses, which are characterized by loss of apical microvilli and mitochondrial alterations in the choroid plexus epithelium and appearance of a higher number of electron dense, i.e. “dark cells” in the epithelium [41].

In EAE, inflammation in the choroid plexuses and CSF precedes the formation of brain and spinal cord inflammatory infiltrates and the development of demyelinating white matter lesions [14, 129]. Granulocytes [41] and myeloid cells taking up circulating very small superparamagnetic iron oxide particles (VSOP) accumulate in the stroma of the choroid plexuses prior to the onset of EAE [41, 93]. Interestingly, VSOP accumulation is undetectable in the non-inflamed CNS suggesting that the resident myeloid cells in the choroid plexus stroma require an inflammatory stimulus to actively phagocytose VSOPs. Increased numbers of MHC class II expressing myeloid cells have also been observed in the choroid plexuses from patients with multiple sclerosis or viral encephalitis but not in patients with amyotrophic lateral sclerosis [162]. This underscores a neuroimmune implication for the choroid plexus in immune system-mediated neuroinflammatory disorders rather than in CNS disorders in which primary events are initiated in the CNS. This is further supported by the observation that the choroid plexuses of patients with multiple sclerosis and viral encephalitis but not with amyotrophic lateral sclerosis show increased expression of VCAM-1 on the choroid plexus vasculature and also harbor increased numbers of T cells and CD138⁺ plasma cells in their stroma [162]. Accumulation of innate and adaptive immune cells in the choroid plexus stroma in multiple sclerosis patients is associated with the appearance of increased numbers of memory/effector T cells (Fig. 6) [51, 68], B cells, and antibody-producing cells in the CSF of these patients [26], suggesting that immune cells that invade the ventricles have to eventually pass the choroid plexus epithelium. Indeed, mimicking the increase of CNS danger signals by intracerebroventricular injection of TNF-α, besides eliciting an inflammatory response throughout the brain, leads to T cell accumulation in the choroid plexus stroma [169]. This further supports the important role of the choroid plexuses as key interphases for a bidirectional neuroimmune communication that may critically influence effector functions of those T cells locally activated within the choroid plexus stroma. The neuroimmune function of the choroid plexus is further underscored by the findings that in neuromyelitis optica, the BCSFB in the choroid plexus is significantly impaired [56].

Both ICAM-1 and VCAM-1 are constitutively expressed by choroid plexus epithelial cells and are significantly upregulated during EAE [144] and intriguingly also in CNS pathology remote to the choroid plexuses such as spinal cord injury [135]. In experimental spinal cord injury epithelial ICAM-1 and VCAM-1 on choroid plexus epithelium have been proposed to mediate immune cell trafficking into the CNS in vivo [135]. ICAM-1 and VCAM-1 have previously been identified to mediate the multi-step T cell migration across the inflamed BBB endothelium during EAE [39]. In the choroid plexus, however, although both adhesion molecules are functionally expressed [144] they are exclusively localized to the apical surface of the choroid plexus epithelial cells and thus not readily accessible for immune cells within the choroid plexus stroma ([41], and Fig. 7). Apical ICAM-1 and VCAM-1 may thus mediate the egress of immune cells from the choroidal epithelial cells allowing them to be released into the CSF or alternatively serve as anchors for immune cells after having crossed the epithelium allowing them to resist the shear forces exerted by the CSF flow. In addition, apical ICAM-1 and VCAM-1 may be the source for soluble ICAM-1 and VCAM-1 found in increased concentrations in the CSF of patients with multiple sclerosis, viral encephalitis or bacterial meningitis [118, 125].

Additional observations in the context of EAE have supported a role for the BCSFB as CNS entry port for encephalitogenic CCR6⁺ Th17 cells during initiation of the disease [114]. The only known ligand for the chemokine receptor CCR6, the chemokine CCL20, is constitutively expressed at the BCSFB rather than the BBB implying that CCR6⁺ Th17 cells access the CNS via the choroid plexuses. In this context it is interesting to note that recognition of myelin oligodendrocyte glycoprotein (MOG) is limited to CCR6⁺ circulating T cells in multiple sclerosis patients [122] and that CCR6⁺ T cells are enriched in the CSF of multiple sclerosis patients [115]. On the other hand, CCR6 expression on CSF T cells does not strictly correlate with a Th17 phenotype [115]. Thus, the CCR6-CCL20 axis may be relevant for the CNS entry of additional CCR6⁺ T cell subsets, which is supported by the observation that CNS recruitment of CCR6⁺ FoxP3⁺ regulatory T cells (T_regs) has also been shown to require CCR6 [163]. The observation that in EAE the CCR6-CCL20 axis plays a dual role mediating autoaggressive Th17 cell versus T reg cell entry into the CNS supports the notion of the BCSFB as a critical interphase integrating neuroimmune signals from the periphery and CNS, leading to immune cell activation or tolerance [25].

The choroid plexuses may play an additional role during neuroinflammation by specifically promoting the recruitment of regulatory or “healing” immune cells. Evidence for this concept has been provided by the observation that in spinal cord injury pro-inflammatory monocyte-derived (Ly6c^hi CX3CR1^lo) macrophages reached the injury site in a CCL2 dependent manner from the spinal cord leptomeninges, while the inflammatory-resolving (Ly6c^lo CX3CR1^hi) monocyte-derived macrophages rather seemed to arrive to the injury site from the choroid plexuses in a α4-integrin/VCAM-1 and CD73 dependent manner [135]. The site of injury may, however, be relevant for the function of the choroid plexuses in promoting anti-inflammatory or inflammatory cues. In traumatic brain injury a rapid increase in synthesis and release of diverse CXCL chemokines and CCL2 from the choroid plexus epithelium has been observed [154, 155] and was accompanied by the invasion of monocytes and neutrophils in between choroid plexus epithelial cells [154]. Pathogen-derived TLR ligands also induce an accumulation of innate immune cells in the CSF and choroid plexus in the developing brain [98]. Whether in the latter two situations, the recruitment of immune cells at the choroid plexuses is beneficial or deleterious to the brain remains to be established.

Accumulating evidence has supported the role of the choroid plexus as key invasion route for immune cells into the CNS. The mechanisms of endothelial and epithelial transmigration, however, are not fully elucidated. Few studies were able to provide direct evidence for the migration of immune cells from the choroid plexus stroma across the BCSFB into the ventricle in vivo [154, 155]. Especially studies in EAE have failed to provide direct in vivo evidence for the localization of T cells between choroid plexus epithelial cells providing proof of their passage across the epithelium [41, 114]. One EAE study even failed completely to detect any T cells in the choroid plexuses [128]. Another study found only a marginal increase in T cells in the stroma [129], that can be interpreted as the choroid plexuses being gateway to the CSF during EAE rather than sites of accumulation per se. On the other hand in vitro studies employing inverse configurations of the BCSFB have provided solid evidence for the migration of activated T cells from the basolateral to apical site of choroid plexus epithelial cells in response to CSF derived factors such as CXCL12 [77, 147]. These observations raise the question if there is an alternative route for T-cell entry into the CNS from the choroid plexus stroma.

Investigating T-cell invasion into the peri-infarct cortex after ischemic stroke by employing in vivo cell tracking of photactivated T cells has indeed provided evidence for a directed intracerebral migration of T cells from the ipsilateral choroid plexus stroma to the peri-infarct cortex [83]. A CCR2 ligand gradient between the choroid plexus stroma and the ischemic cortex was identified as the potential driving mechanism for T cell invasion. At the same time blocking CSF flow with matrigel failed to alter T cell invasion at the peri-ischemic site suggesting that T cell migration via ventricular CSF does not play a significant role in cerebral T cell invasion after stroke [83]. A route of immune cell entry from the choroid plexus stroma into the CNS other than passing through the epithelium, e.g. from the base of the choroid plexus thus should be investigated. It may, similarly to the postulated T-cell migration across the epithelium, contribute to the preferential localization of focal demyelinated plaques in periventricular areas in patients with multiple sclerosis [76].

Furthermore, there is emerging evidence that the choroid plexuses are relevant entry sites for pathogens, such as bacteria, viruses, fungi and parasites into the CNS [31]. A number of pathogens, including Streptococcus suis, Neisseria meningitidis, Haemophilus influenzae, Listeria monocytogenes, and Trypanosoma brucei, to just name a few, have been shown to use the choroidal epithelium as a CNS entry site (summarized in [131]). These pathogens may cross the epithelium directly via paracellular or transcellular routes or use Trojan horse approaches to breach the barrier [31]. The choroid plexus responds to these pathogens by eliciting a stereotype programmed innate inflammatory host response leading to an increased expression of cytokines and chemokines that trigger BCSFB breakdown and eventually to an increased CNS invasion of pathogenic inflammatory cells into the CNS. In this context, the direct response of the choroid plexuses to the infectious agents will in addition to CNS invasion of the pathogens contribute to the pathogenesis of these neuroinflammatory diseases [131].

Finally, in light of the significant contribution of the meninges in brain metastasis, a role of the choroid plexuses when considering the pathway of tumor cell invasion into the brain may be underestimated. Recent in vitro evidence supporting a role for the migration of neuroblastoma cells across the choroidal barrier [161] warrants further studies into the role of this barrier in tumor metastasis to the CNS.