The choroid plexus: a missing link in our understanding of brain development and function

Abstract

Studies of the choroid plexus lag behind those of the more widely known blood-brain barrier, despite a much longer history. This review has two overall aims. The first is to outline long-standing areas of research where there are unanswered questions, such as control of cerebrospinal fluid (CSF) secretion and blood flow. The second aim is to review research over the past 10 years where the focus has shifted to the idea that there are choroid plexuses located in each of the brain’s ventricles that make specific contributions to brain development and function through molecules they generate for delivery via the CSF. These factors appear to be particularly important for aspects of normal brain growth. Most research carried out during the twentieth century dealt with the choroid plexus, a brain barrier interface making critical contributions to the composition and stability of the brain’s internal environment throughout life. More recent research in the twenty-first century has shown the importance of choroid plexus-generated CSF in neurogenesis, influence of sex and other hormones on choroid plexus function, and choroid plexus involvement in circadian rhythms and sleep. The advancement of technologies to facilitate delivery of brain-specific therapies via the CSF to treat neurological disorders is a rapidly growing area of research. Conversely, understanding the basic mechanisms and implications of how maternal drug exposure during pregnancy impacts the developing brain represents another key area of research.

CLINICAL HIGHLIGHTS.

The choroid plexuses have the potential to exert either favorable or deleterious effects on the brain via their secretome. For example, epithelial cells secrete many important health- and growth-promoting factors for the brain. However, the release of some factors, including release of cytokines and other damaging effector molecules previously thought to originate solely from immune cells, can now be attributed at least in part to the choroid plexus.

The choroid plexuses provide a novel route of entry to the brain and spinal cord for peripherally derived molecules that are required for normal development and healthy brain function. On the other hand, cellular barriers present in this tissue may hinder the delivery of therapeutic agents required for the treatment of infection and disease in the central nervous system (CNS).

Similarly, trafficking of peripherally derived immune cells and immune cell-secreted molecules through the choroid plexus represents a major strategic location for peripheral-central cell-cell interactions. An abundance of drug efflux transporters on the blood side of the plexus epithelial cells add to the physical barrier and protection provided by the choroid plexus. Many disease-associated mutations in both developmental and neurodegenerative pathologies decrease the ability of the plexus to produce such protective efflux.

Future therapies that reduce drug efflux, or allow specific transport of other brain-nonpenetrant drugs into the CSF and thence the brain, could provide a novel and effective drug delivery route to help ameliorate a wide range of diseases.

1. INTRODUCTION

What a privilege it is to provide a review of the choroid plexus in the centenary year of Physiological Reviews. A year after the foundation of Physiological Reviews in 1922, Lewis Weed contributed an article entitled “The cerebrospinal fluid” (1). It opens in the very first sentence: “It is obviously impossible within the limitations of this review to present a truly comprehensive account of a characteristic body-fluid such as the cerebrospinal liquid.” It is a daunting task to provide an up-to-date review of the choroid plexus a hundred years later. Of necessity, this should include its intimate biological relationships to the cerebrospinal fluid (CSF) that it secretes. Although the choroid plexuses and CSF have a long and venerable history, it is the blood-brain barrier proper that has received much more attention since it was recognized early in the twentieth century (2). The term “blood-brain barrier” was first coined in French by Stern, “barrière-hémato-encéphalique” (3), to describe the mechanisms that keep dyes and toxins from entering the brain from the circulation. However the term “blood-cerebrospinal fluid barrier” (blood-CSF barrier) was first used even earlier, also by Stern (4). Despite this, the contrast in research activity in the two barrier fields can be clearly illustrated by comparing PubMed entries for the comparable structural entities or their barrier functions (FIGURE 1).

FIGURE 1.

Historical publication data for choroid plexus and vascular endothelial cell barriers in the brain. Although minimal work was published from 1834 to the mid-1900s, an explosion in publications focusing on the blood-brain barrier and brain endothelial barrier properties began around 1945. Data collected from PubMed May 14, 2022 with the following search terms: “choroid plexus,” “cerebral blood vessels,” “blood-brain barrier,” “blood cerebrospinal fluid barrier.”

There are several excellent historical reviews of CSF and choroid plexus available (see, e.g., Refs. 5–7), so only a brief summary is provided here. The presence of a fluid within the ventricular system was probably first reported in writing in the Edwin Smith Surgical Papyrus (1700 BC). However, it was not until 1664 that Thomas Willis suggested the choroid plexus as the source of the fluid (cerebrospinal fluid) within the ventricles. A variety of terms to describe the tissues within the ventricular system, such as choroeidé sustremmata (Herophilos) and choroeidé plegmata (Galen) were used (see Refs. 5–7 for details and references). Most studies in the 300 years following Willis focused on the structure of the ventricular system and the choroid plexuses. There were also numerous arguments about whether CSF was a filtrate or secretion of the plexuses. Serious studies of the functions of the choroid plexus and properties of CSF did not begin in any substantial way until the middle of the twentieth century (5).

In this review we briefly consider the secretory properties of the choroid plexuses in producing the CSF within the ventricular system and its importance for the stability of the internal environment of the brain, both in development and in adulthood. This in turn is essential for normal brain function, particularly with respect to ion concentrations. This is followed by a review of available experimental evidence suggesting that some specific features of the CSF are present even before the appearance of the choroid plexuses. We then discuss how early in development the CSF begins to provide a local environment that is distinct from that of the rest of the embryo. Related to this is consideration of the features of this distinctive environment that may be important for specific features of brain development.

Understanding physiological function cannot be divorced from morphological structure. We therefore describe briefly the structure of the choroid plexuses during their development and in the adult. This includes consideration of differences between the plexuses in all four ventricles as well as some striking species differences.

Choroid plexuses provide an interface of exchange between the CSF and the circulating blood. There are comprehensive cellular transport mechanisms in the choroid plexuses that are important for inward movement of nutrients such as glucose and amino acids from blood into CSF. There are also outward transport mechanisms from CSF to blood that remove metabolites as well as prevent entry and/or remove drugs and toxins. There are numerous reviews that cover these aspects of choroid plexus function. Where appropriate, these are cited and our focus is on the more recent studies that extend understanding of the functions of the choroid plexuses.

A contentious point that has been difficult to evaluate experimentally is the relative contribution of transport in and out of the brain by the cerebral blood vessels (blood-brain barrier) and the choroid plexuses (blood-CSF barrier). The latter mechanisms are probably much more important early in brain development, as plexus development occurs much earlier than the vascularization of the brain (8) and it has a notably high blood flow compared to the brain itself (9).

A related clinically important point is the extent to which the composition of the CSF reflects properties of the blood-brain barrier itself. It is often assumed that measurements of CSF reflect properties of the blood-brain barrier, but comparison with measurements of brain interstitial fluid shows that this is generally not the case (10, 11).

There is also increased understanding of possible roles of the choroid plexuses in a variety of pathological processes. These include inflammation and infection following trauma, hydrocephalus, and neurodevelopmental, psychiatric, and degenerative diseases such as Alzheimer’s disease. The choroid plexus is notably of interest as being a possible route for entry of drugs into the central nervous system (CNS), a problem that has largely eluded many researchers over several decades. These topics are considered below.

Overall, this review has two themes. The first theme deals with the contribution of the choroid plexuses to the properties of CSF that are essential for normal brain development and function as previously proposed (12–15). This approach replaces the nearly century-old notion that brain barrier mechanisms are “immature ”or “leaky” (as discussed in Ref. 2). Consideration of specific properties of the CSF is being increasingly recognized as appropriate given the developmental context (16–18). The second theme concerns properties of the choroid plexuses and their communication with other parts of the brain via the CSF. This extends the understanding of the CSF interactions with these tissues and their contributions to brain functions that go well beyond their importance in controlling the internal environment of the brain. These include aspects such as immune and chemical surveillance, inflammation/pathology, regulation of choroid plexus function by sex hormones, and other factors like the possible role of the choroid plexuses in circadian rhythms and sleep.

2. MORPHOLOGY OF THE DEVELOPING AND ADULT CHOROID PLEXUSES

2.1. Gross Morphology and Cross-Species Comparisons

For general orientation, FIGURE 2 illustrates the location of the four choroid plexuses (2 lateral, 3rd, and 4th) within the cerebral ventricles. Netsky and Shuangshoti (19) have provided a detailed description of the morphology of the developing and adult choroid plexuses, mainly in the human. Additional information on the developing human plexuses has been published (20–25). Animal species that have been reported include rabbit (26–28), sheep (29), mouse (30–32), rat, Monodelphis domestica (South American gray short-tailed opossum) (33), chick, frog (34, 35), and zebrafish (36, 37). A variety of fish species have also been reported on, including elasmobranchs (sharks, rays, skates, and sawfish) (38, 39). In contrast to the four plexuses in most vertebrate species, only two sites are present in teleosts such as zebrafish. They correspond to the 3rd and 4th ventricular plexuses (40). Some species, for example amphioxus (39) and hagfish (41, 42), do not possess any choroid plexuses. An extensive evolutionary perspective on the development of the ventricular system and choroid plexuses has been provided elsewhere (38, 40, 43, 44).

FIGURE 2.

Gross morphology of cerebral ventricles and choroid plexuses within them in the mammalian brain.

All four choroid plexuses share a similar gross structure (FIGURE 3), although they are very different in degree of compactness and size. The 3rd ventricular choroid plexus is the smallest, and the 4th is the most compact (45). The lateral plexus in mouse has anterior and posterior domains, with distinct lineages contributing to their development (46) and selective vulnerability to tumorigenesis (47, 48). In the adult, all choroid plexuses consist of a single layer of cuboidal epithelial cells joined by apical tight junctions, which are the morphological basis of the blood-CSF barrier. Each plexus has an inner core of well-vascularized mesenchymal tissue, folded into villi protruding into the ventricles. The surface area of the cells, in particular the apical surface, is considerably increased by the presence of microvilli superimposed on the villous macrostructure of the plexuses. The detailed molecular mechanisms underlying choroid plexus morphogenesis are beginning to be understood (49). For example, the particularly elaborate branching structure of the 4th ventricle choroid plexus is regulated in part by MEIS1-WNT5A signaling (50).

FIGURE 3.

Schematic cross section of 2 choroid plexus villi. The mammalian and nonmammalian vertebrate choroid plexuses consist of numerous fronds projecting into the cerebrospinal fluid (CSF), which are composed of several villous processes. An outer simple cuboidal epithelium lies on a basal lamina surrounding an inner stromal core of connective tissue that is highly vascularized. It is derived from the adjacent ependyma lining the ventricle walls. The plexus differs from the ependyma by the presence of apical tight junctions. The epithelial cells are polarized. The apical membrane facing the CSF has uneven borders of irregular microvilli and numerous groups of cilia. The lateral membranes of the epithelial cells display complex infoldings at their basal ends. Each villus contains a large fenestrated capillary consisting of very thin endothelial cells. The stromal connective tissue is composed of a loose network of collagen fibers, secreted by occasional elongated fibroblasts (yellow). Globular macrophages (green), rich in phagolysosomes, are also present in the stromal core; these are distinct from the star-shaped dendritic cells. A few of the dendritic cells lie between the basal lamina and the choroid epithelium, extending processes between epithelial cells. Kolmer cells (epiplexus cells, dark blue) lie on the ventricular surface of the epithelial cells, closely associated with the microvillous border. Adapted from Ref. 45, with permission from Journal of Neuropathology & Experimental Neurology.

The earliest ultrastructural study of choroid plexus epithelial cells was that of Maxwell and Pease in the mid-1950s (51). This study, predominantly in adult rat, focused on surface pedicels and folding of the cell membrane, particularly of the basal cell surface, as well as intracellular organelles. Cilia are only mentioned as a “persistent embryological remnant.” Intercellular junctions are not mentioned, although their Figure 1 shows clearly what would later be described as a “tight junction.” This term was introduced by Farquhar and Palade in 1963 (52) in a study of various epithelia. Luse (53) provides a single image of adult mouse choroid plexus, which she describes as having morphology similar to ependymal cells with a more complex microvillus and only sparse cilia. Millen and Rogers (54) give a useful general description of the ultrastructural features of plexus epithelial cells. Tennyson and Pappas (55, 56) provide a much more detailed study with high-quality electron micrographs. Along with Brightman (57) and Brightman and Reese (58), they provide the first description of intercellular junctions between the epithelial cells that impede the movement of at least large molecules. Doolin and Birge (59, 60) published a study of their development in chick embryo choroid plexus from embryonic day (E)8 and in the adult. More recently, collagen expression in the epithelium was shown to be essential for maintenance of the epithelial basement membrane and junctional integrity. This observation came from experiments in which disrupted collagen (Col9a3) expression altered microtubule dynamics needed for correct polarization of cells, including tight junction assembly (61).

2.1.1. Cilia.

Many of the epithelial cells of the choroid plexus have multiple (8–12) cilia, but the possible physiological functions of these cilia remain unclear (62). Differentiating epithelial cells have distinct programs for spatial organization and ciliogenesis expression (63). Maturing epithelial cells in each choroid plexus transiently express ciliogenesis genes, including Deup1/Ccdc67, which drives centriole biogenesis, a key step of multiciliation (64). According to Sturrock (32), cilia are apparent in choroid plexus epithelial cells by embryonic day (E)12 in the mouse and their numbers increase during development. Early electron microscopical studies described the 9 + 2 axoneme in several species including human (65), newly weaned pig (66), and chick (60). However, the contemporaneous account of Peters and Swan (67) in the rat showed that they have a 9 + 0 configuration. These authors demonstrated how tissue displacement might give rise to the appearance of 9 + 2. The functional significance of the different configurations is that 9 + 2 cilia are motile whereas 9 + 0 are typically not motile, with some notable exceptions including single nodal cilia, which exhibit a rotational motion rather than a fully beating motion. The 9 + 0 configuration was confirmed in pig choroid plexus (68). However, Takeda and Narita (69) consider choroid plexus epithelial cells to be unusual because although they have the 9 + 0 configuration, each cell has many cilia. Narita and colleagues also identified the typical configuration in mouse choroid plexus, but in postnatal day (P)1 animals they also found 9 + 2 and some 9 + 1 on the same plexus cells in addition to those with 9 + 0 (70). These cilia showed evidence of motility. In a further study Nonami et al. (71) showed a peak in cells with motile cilia at P2, which declined to negligible numbers by P14. It is unclear what the functional significance is, as these studies discussed that this motility did not produce a directional flow of CSF (70). Their time-lapse studies showed that there were two types of movement: rotational and back and forth.

DNAHC11 (encoded by the Dnah11 gene) is an axonemal dynein required for the motility of some cilia. Dnah11 gene expression preceded the changes in ciliary motility. Expression of this gene was already detectable at E13 and was downregulated by P1, which is days before the number of motile cilia began to decline (71). These authors suggested that the embryonic motility implies that the motile cilia in embryonic and perinatal periods might act as mechanosensors regulating CSF production at a time of ventricular development. Another possibility is that they might be important for producing a stirred layer at the surface of the developing plexus epithelial cells at a time when their secretory activity is low (72, 73).

Cilia and CSF flow are also believed to provide key ingredients for a healthy functioning ventricular system but appear to vary across species (e.g., Refs. 74–76). For example, during development in mouse, progenitors lining the brain ventricles are monociliated and nonmotile until around birth, when multiciliated ependymal cells differentiate and mature (62). In contrast, during development in zebrafish, there are distinct ependymal cell populations along the ventricles that are either mono- or multiciliated (75). In addition, ablation of cilia at the zebrafish choroid plexus leads to ventriculomegaly (76).

2.2. Morphological and Functional Heterogeneity in Choroid Plexus Epithelial Cells

In standard histological preparations the epithelial cells appear to be a rather homogeneous population, but more detailed morphological techniques show that this is not the case.

2.2.1. Morphometric studies of the developing choroid plexuses.

A detailed morphometric study of the developing lateral ventricular choroid plexus in the rat has been published (77). It covers E16 (3 days after the first appearance of this plexus) to a young adult age of 30 days. The basolateral surface of the epithelial cells increases only slightly, but the apical surface increases by threefold, where the surfaces are measured as surface area per unit volume of cell, which allows for changes in cell volume. Keep and Jones (77) estimate the total apical surface area of all choroid plexuses as 75 cm² for young adult rats (30 days). They point out that this surface area is similar to estimates of surface area of the cerebral capillaries (155 cm²) at this age. This similarity underscores the importance of the microvilli of the choroid plexus epithelial cells, which are not present in cerebral capillary endothelial cells. Although epithelial cells are by far the dominant cell type in the choroid plexuses, there are other cell types present within the stromal connective tissue (macrophages, fibroblasts, dendritic cells). It can be estimated from cell counts that epithelial cells in the different choroid plexuses comprise 61–69% of the total plexus cell numbers (78).

Glycogen has been reported to be present in epithelial cells of the 3rd and 4th ventricular choroid plexuses in human fetuses, but for a briefer period of development (23). References to studies of choroid plexuses in other species are listed at the beginning of this section and in reviews by Brocklehurst (43), Dziegielewska et al. (79), Lun et al. (14), and Fame et al. (44). The general pattern of development appears to be similar. Tennyson (80) summarized the reports on the occurrence of glycogen in different species. There has been some discussion about the functional significance of the glycogen in stages II and III of plexus development. Several authors have suggested that it may have general nutritive function or be related to synthesis of specific structural components such as mucopolysaccharides of the basement membrane (23) or glycosaminoglycans (32). The choroid plexuses of developing marsupials (79) and chicks [commented on by Birge and Doolin in 1965 as referenced by Tennyson (80)] lack glycogen. Glycogen levels are also reduced and have a different distribution in choroid plexuses of hibernating animals (81) similar to that in adult frogs after adrenaline injection (82). When cultured in vitro the postnatal mouse choroid plexus epithelial cells were found to lose their glycogen (83). Possibly there is a relation between the presence of glycogen early in development and low oxygen levels in the fetus.

More recently, the metabolic status and ATP production capacity of the choroid plexus were evaluated pre- and postnatally in mice (84). Mitochondria number and size increased from E16.5 to 2 mo of age, whereas cellular glycogen load decreased. Furthermore, the subcellular distribution of mitochondria shifted such that while embryonically mitochondria were preferentially distributed along the basal side of epithelial cells, by the end of the first postnatal week mitochondria attained a predominantly adult apical distribution (85). Because mitochondrial subcellular localization can respond to regional energy demands in other systems, the shift in epithelial cell mitochondrial distribution in mature cells supports a model in which there is a requirement for increasing ATP supply. This change is in response to higher demands and functions of the adult apical surface (e.g., secretory functions). In agreement with this model, Xu and colleagues (84) report increased ATP production in postnatal choroid plexus.

2.2.2. Dark cells.

Historically, an early indication of diversity in the epithelial cells of the choroid plexuses was the observation of “dark” cells. This has fueled a long-standing discussion concerning their nature among the overall population of plexus epithelial cells. Light and dark cells were already identified in histological preparation in the late nineteenth century but were dismissed by many investigators as artifacts of the preparation techniques used (86). They were described in one of the first electron microscopy (EM) studies of the choroid plexus by Wislocki and Ladman (87). In a careful electron microscopical study of choroid plexuses from anesthetized dogs and human biopsies, Dohrmann (86) was able to demonstrate dark and light epithelial cells in well-preserved tissues. The main difference between dark and light cells was the greater density of the cytoplasm, nuclei, and matrix of the microvilli in the dark cells. In the mouse embryo, dark cells are apparent from E14 (32); the proportion of epithelial cells is approximately constant at ∼11–12% up to P90. Johanson et al. (88) reported that dark cells were only infrequent in infant rat choroid plexus.

The functional significance of dark cells, if any, remains enigmatic. In rats, arginine vasopressin (AVP) increased in vitro the number of dark cells in adult but not in developing plexus (88). Fibroblast growth factor 2 (FGF-2) administered in vivo has been shown to increase the number of dark cells and reduce CSF secretion (89). This is possibly via a reduction in Cl⁻ efflux from choroid plexus epithelial cells. However, it remains an association rather than direct demonstration of an effect on these cells (88). A recent study of choroid plexus organoids reported that dark cells were richer in mitochondria (90).

Transcriptome analyses report that all choroid plexuses share some genes (e.g., Ttr, Htr2c, Kcnj13, Aqp1). However, there are also differences: for example, the 3rd ventricle plexus expresses higher levels of Ins2 and the 4th ventricle plexus expresses Shh and Wnt5a (63). Choroid plexus-secreted factors in the CSF can be delivered to many target cells in the developing and adult brain (91, 92). As the choroid plexus matures there are age-dependent increases in expression of transporters and channels involved in ion and water movement across the epithelium, e.g., Slc12a2 (NKCC1), Kcnj13 (Kir7.1), and Slc4a5 (NBC4) in mice (84) and rats (93, 94). These expression patterns strongly suggest that the key adult function of choroid plexus epithelial cells to secrete CSF water matures over the course of development (13, 84). Age-dependent downregulation of gene expression is observed particularly for genes encoding secreted proteins with roles in early development (e.g., Shh), whereas expression of some secreted signals increases with age, such as the “antiaging” gene Klotho (95, 96).

2.2.3. Plasma protein-containing epithelial cells.

Immunohistochemical staining for plasma proteins showed that there are subpopulations of epithelial cells containing a variety of these proteins. The numbers of these plasma protein-containing cells are generally higher in choroid plexuses of brains from larger animals. There is evidence that some of these cells contain specific proteins whereas others may contain more than one plasma protein (21, 22, 33, 97). Nevertheless, in all studies the proportion of plasma protein-containing cells is higher earlier in development than in the adult (21, 22, 33, 97). There also appears to be more specific individual protein transport in the developing choroid plexus compared with the adult, where several plasma proteins are transported by the same cell (93, 98), although the functional consequences of such a loss in specificity remain unknown. There does not appear to be an association between presence of plasma protein and type of plexus epithelial cell, at least in the case of albumin (21). It has been suggested that this presence of plasma proteins reflects transfer of these proteins from the blood to the CSF via the interstitial space of the choroid plexuses and then transport across the epithelial cells of the plexus. This transfer has been demonstrated in a variety of species (97, 99, 100). Combined with the low turnover of CSF, this suggested that transport results in a high concentration of individual plasma proteins and total protein concentration in CSF in the developing brain compared with the adult (8). The total protein concentration in CSF can be 10 times in fetal rats (101) and postnatal marsupials (97, 102) compared with the adult and up to 30 times in fetal sheep (99) and fetal pigs (103) compared with the adult.

2.2.4. Nonepithelial cells in the choroid plexuses.

The choroid plexus is a site of diverse cell types, subtypes, cell states, and expression programs at all ages and in all ventricles. Single-cell transcriptomic analyses have shown that in addition to epithelial cells there are five major cell classes present in the choroid plexus during mouse embryonic development. These are mesenchymal (mural and fibroblast), endothelial, immune, neuronal, and glialike cells (63). The mesenchymal population also includes pericytes and vascular smooth muscle cells. This is in agreement with cranial mesenchyme and neural crest lineage contributions to the choroid plexus stroma (104, 105). Similar to epithelial cells, the mesenchymal cells harbor ventricle-specific, regional transcriptional identities. During development, mesenchymal cells tend to express genes involved in morphogenesis and development. However, postnatally in adult and aged tissue mesenchymal cells upregulate expression of transporters, receptors, peptidases, and proteolytic activity genes (63). There is additional discussion of some of these diverse cell types in sect. 5.

2.2.5. Blood supply to the choroid plexuses.

The anterior choroidal and posterior cerebral arteries largely feed the lateral ventricle choroid plexus. The arteries give rise to several branches that extend across the tissue. Ultimately, veins converge toward the free margin of the tissues and eventually join the internal cerebral vein to drain the choroid plexus. This general structure is best studied in the lateral ventricle and is largely conserved across species ranging from mouse to human (63, 106, 107). The blood vessels within the plexus stroma are fenestrated and allow the passage of large protein molecules from the blood into the stromal tissue (58). In the adult rat, choroid plexus blood flow is the highest of any region in the brain at ∼4 mL/g/min in the 4th ventricular plexus. It is ∼3.25 mL/g/min in the lateral ventricular choroid plexus. At the earliest age measured (2 wk postnatal), the blood flow was lower, at 2.7 mL/g/min (4th ventricular plexus) and 2.5 mL/g/min (lateral ventricular plexus). In contrast, the cerebral cortex blood flow at 2 wk was only ∼0.31 mL/g/min, increasing in the adult to ∼0.6 mL/g/min (9). The relatively high flow rate in the developing choroid plexus compared with the brain emphasizes the importance of the plexuses for exchange across this interface early in development (8).

2.3. Development of the Choroid Plexuses

After the neural tube has closed, each choroid plexus emerges independently from the roof plate, from which capillaries and mesenchymal and neural crest cells invaginate the neuroepithelium (46, 105, 108) (FIGURE 4). There is a period following closure of the neural tube (109) when the tube expands before the development of any choroid plexuses. This process is considered in sect. 4. Most papers on the choroid plexuses do not specify which choroid plexus has been studied, and when they do, it is usually the lateral ventricular plexuses. This important experimental detail was pointed out by Strong (27) and is only beginning to be consistently documented in publications.

FIGURE 4.

Changing fluid sources and targets during neurulation and early neurogenesis. Developing brain schema derived from mouse development. Major changes occur throughout neural tube closure (NTC) and early brain development. A: the neuroectoderm of the open neural tube folds to generate neuroepithelium, and the fluid environment changes from amniotic fluid (AF) to cerebrospinal fluid (CSF). Ages are shown for mouse development at embryonic days (E)8.5, E10.5, and E12.5. Scale bars, 1 mm. B: before NTC, the fluid environment of the neuroectodermal progenitors is AF that contains evidence of high levels of translation and glycolysis that occurs in the progenitors during this time. Early CSF before choroid plexus formation contains unique markers of brain parenchyma development, as well as some components that have been shown to signal to neural progenitors. Later in development, CSF components can arise from the neural progenitors or from the choroid plexus. CSF also mixes and flows to move components to different regions of the ventricular system. AF composition includes organelle components (e.g., ribosomal or mitochondrial proteins). CSF contents include free molecules [e.g., 5-hydroxytryptamine (5-HT)], molecules bound to carrier proteins (e.g., RA+ RBP4); organelle components (e.g., ribosomal or mitochondrial proteins), free proteins or peptides (e.g., WNT3A or cytokines), or membrane-bound particles (e.g., exosomes) that act as carriers for additional proteins (e.g., SHH) or nucleic acid (e.g., miRNA loops). These different components have unique signaling modalities including receptor binding to cells or fusion with cells in direct contact with CSF, including neural progenitors or choroid plexus cells (epithelial cells, mesenchymal cells, endothelial cells, immune cells, neurons).

The timing of choroid plexus development varies across species. It does not seem to be related to the stage of gestation or time of birth, which are very different across species. Thus, in humans the 4th ventricular choroid plexus appears around 6 wk of gestation, the lateral ventricles at 7 wk, and 3rd ventricular plexus at 8 wk, corresponding to 16-, 19-, and 23-mm crown-rump length, respectively (110). These authors point out that estimates of gestation age in the literature are quite variable. We suggest that this is probably due to differences in how gestational length is defined and the practical difficulty of establishing when fertilization occurs. For that reason, measurements of crown-rump length are probably more useful for developmental studies.

Sheep are characterized by a relatively long gestation period (150 days) compared with similar-sized nonprimate vertebrates. The choroid plexuses in sheep appear at E21 (4th), E24 (lateral), and <E60 (3rd) (29, 111). Conversely, marsupials are characterized by their extremely short gestational periods. For example, in Monodelphis domestica the 4th plexus appears just before birth (21 days of gestation). The lateral plexuses are present at birth as buds protruding into the ventricles, and the 3rd does not appear until P5 (112). In Sminthopsis crassicaudata (fat-tailed dunnart), which has an even shorter gestational period (16 days), the 4th ventricular choroid plexus is only just present at birth as a rudimentary structure, with the other choroid plexuses developing postnatally (79).

From their studies of human embryos, Netsky and Shuangshoti (24, 110) divided the development of the lateral ventricular choroid plexuses into four stages. These were mainly determined by histological criteria and the presence or absence of glycogen in the epithelial cells. In stage I these epithelial cells are in the form of a pseudostratified epithelium that is continuous with the neuroepithelial cells that line the ventricles. The cells lack glycogen and have centrally positioned ovoid nuclei. There are occasional mitotic figures toward the luminal (ventricular) surface of the cells. Early signs of vascularization are already apparent at this stage of choroid plexus development. Stage II runs from the 9th week of gestation (30-mm crown-rump length) to the 16th week (110-mm crown-rump length). At 9 wk the choroid plexus occupies about one-third of the ventricular space. As it grows, it comes to occupy most of the volume of the ventricle by 11 wk. After this, there is less growth of the choroid plexus and the increase in volume of the ventricles outpaces that of the plexuses, due mainly to the onset of secretion of CSF by the epithelial cells. The choroid plexus becomes more lobulated as it develops, and the pseudostratified epithelium becomes a characteristically low columnar type. Some pseudostratified epithelium persists at the base of the plexus, which is the site of limited continued cell division; growth of the plexus occurs from its base but only from the dorsal side (33, 113) (FIGURE 5). The columnar epithelial cells are filled with glycogen, and the nuclei are displaced toward the apex (ventricular) surface of the cells. In some histological preparations, the glycogen is dissolved by the preparative procedure and the cells appear to be empty apart from the apical nucleus. The proliferative pseudostratified epithelial cells at the base of the plexus lack glycogen. In stage III the cells are cuboidal, glycogen deposits decline, and the nuclei have central or apical positions. Stage IV consists of cuboidal or squamous epithelial cells with central or basal nuclei and no glycogen. The only other species in which these stages in development of the lateral ventricular choroid plexuses have been confirmed and described in detail is the sheep (29). Although the four stages appear in the same order in the sheep fetus, their timing and duration are different. In a series of studies of both human and sheep choroid plexuses Møllgård and colleagues (21) identified six different types of epithelial cells; these represent some refinement of the criteria of cell shape and nuclear position as described by Netsky and Shuangshoti (110), with addition of the pre- and postmitotic cells at the base of the plexus as a separate class of cells. Information about the 3rd and 4th ventricular choroid plexuses is very limited compared with that for the lateral ventricle plexuses. Jacobsen et al. (21) describe the same four stages in the development of the 3rd ventricular plexus. They comment that there were fewer than six cell types present. In the 4th ventricular plexus they identified all six cell types but found that there was no clear demarcation between stages. In fetal sheep, Jacobsen et al. (29) identified the four stages of development and six cell types in the lateral ventricular plexus. In the 4th ventricular choroid plexus the four stages were more clearly delineated than in the human.

FIGURE 5.

Development of plexus epithelial cells and transcription factors involved in development of the lateral ventricular choroid plexus. The choroid plexus epithelial cells (CPECs) develop from modified neuroepithelium and are added to the structure from the dorsal (upper) surface. Experiments using Monodelphis domestica (South American gray short-tailed opossum) after a single injection of 5-bromo-2-deoxyuridine (BrdU) have been used to identify plexus growth in vivo. Positive nuclei initially seen in CPECs at the root of the plexus (top), with additionally added new cells pushing the plexus structure out from the neuroependymal wall, causing the complex invaginated structure to form. In this way, new epithelial cells are added along the stalk of the plexus, away from the root, displaced by newly dividing cells. No cells are described as being added from the ventral surface. Specific transcription factors involved in the promotion (green) or inhibition (red) of plexus epithelial development and growth are also shown. CSF, cerebrospinal fluid. Adapted from Ref. 33, with permission from Cerebrospinal Fluid Research, and Ref. 113, with permission from Frontiers in Neuroscience.

3. MECHANISMS CONTRIBUTING TO BRAIN WATER AND IONIC HOMEOSTASIS

One of the most distinctive features of the CSF, and one that is largely determined by the choroid plexuses, is the unique ionic composition and stability of ion concentrations. This in turn contributes to the internal environment (extracellular fluid) of the brain. These exchange mechanisms across the choroid plexus epithelial cells in the rat embryo and adult are summarized in FIGURE 6. There is also a contribution from the ion exchange mechanisms of the cerebral endothelial cells constituting the blood-brain barrier (130). Because of methodological constraints it is difficult to determine the relative importance of these two essential contributions to this ionic stability. In this review we focus on the contribution of the choroid plexuses to CSF ionic composition. This in turn is thought to be a main contributor to the brain extracellular fluid composition, at least in the adult brain. There is essentially free diffusional movement across the ependymal lining of the ventricular system; the ependymal cells are linked by gap junctions that do not impose a limit on exchange between CSF and brain extracellular fluid even for large molecules (58, 131). In the developing brain the situation may be different, as considered below. There are numerous other exchange mechanisms between blood and brain that the choroid plexuses contribute to such as organic ions and metabolically important molecules such as glucose. The focus of this section of the review is on water and ions because of the immediacy of their function in controlling normal neuronal function.

FIGURE 6.

Localization of proteins for ion transporters, channels, and associated enzymes and identification of their corresponding genes in adult and embryonic (E15) rat choroid plexus. Data for localization of the proteins are from Refs. 114–117. Cerebrospinal fluid (CSF) secretion results from coordinated transport of ions and water from basolateral membrane to cytoplasm, then sequentially across apical membrane into ventricles [for review see Davson and Segal (118)]. On the plasma-facing membrane there is parallel Cl⁻/$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ exchange (AE2, Slc4a2) and Na⁺-$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ cotransport (NBC1, Slc4a4), with net effect of bringing Cl⁻ into cells in exchange for $\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ (119). Na⁺-dependent Cl⁻/$\mathrm{HC}{\mathrm{O}}_3^{\mathrm{-}}$ exchange (NCBE, Slc4a10) at the basolateral membrane modulates pH and perhaps CSF formation (114,115, 120). Apical Na⁺ influx by NHE5 (Slc9a5) and ATB1 (Atb1b1, Na⁺-K⁺-ATPase, asterisk) maintains a low cell Na⁺ concentration ([Na⁺]) that sets up a basolateral gradient to drive Na⁺ uptake (121). Na⁺ is extruded into CSF mainly via the Na⁺-K⁺-ATPase pump (ATB1, Atb1b1) and, under some conditions, the Na⁺-K⁺-Cl⁻ cotransporter NKCC1 (Slc12a2; see Ref. 122 for review). Overall cell volume is maintained by the K⁺-Cl⁻ cotransporters NCCT (Slc12a3) and KCC1 (Slc12a4). Aquaporin (AQP1/3/4) channels on CSF-facing membrane (123) mediate water flux into ventricles (124). AQP1 is also localized to the basolateral membrane (125). Polarized distribution of carbonic anhydrase (CAR) and Na⁺-K⁺-ATPase, and aquaporins, enable net ion and water translocation to CSF (see Refs. 122, 126 for review). The gene Slc4a7 (NBCN1) was not detected by RNA sequencing (RNA-seq), although it has been reported in both rat and mouse choroid plexus (114). The genes for Clir (chloride inwardly rectifying) channels has not been previously identified but are probably Clica and Clicb. The gene for VRAC (volume-regulated anion channel) is not known (127). The carbonic anhydrases CAR2 and CAR8 have an intracellular distribution; CAR8 has been shown to lack the characteristic enzyme activity of these proteins (128). It is not known whether it is functional in the embryo. The CLIC chloride channels are also intracellular (129), but their location is unknown. Insets show the fold differences (FDs) for genes expressed at a higher level in the embryonic (red) or the adult (blue) choroid plexus. Adapted from Ref. 93, with permission from PLoS One.

There has been a long-standing discussion, based on measurements of the ionic composition of CSF, of whether CSF is a secretion from the choroid plexuses or an ultrafiltrate of plasma. This was finally settled by Davson (132). He pointed out that because of the large difference in protein content of plasma and CSF in the adult brain it was essential to make measurements of ions based on units expressed as milliequivalents per kilogram of H₂O, not volume of plasma or CSF (132). Presumably because of the tedium of determining the water content of plasma and CSF samples, which requires drying them (133), it has become usual to use units of millimoles or milliequivalents per liter. This is potentially less accurate, as weighing is more precise than the measurement of volume. On this basis, the ionic composition of CSF cannot be accounted for by ultrafiltration, though some authors still suggest that CSF is an ultrafiltrate (e.g., Refs. 134, 135). Additional evidence for secretion of CSF by the choroid plexuses comes from the inhibitory effects of drugs on CSF secretion: acetazolamide, an inhibitor of carbonic anhydrase, is the primary driver of CSF production by choroid plexus epithelial cells (136, 137). Inhibitors of Na⁺-K⁺-ATPase that have been shown to reduce CSF secretion are ouabain (136, 138) and digoxin (139). It is generally held that the choroid plexuses produce the bulk of CSF in the adult brain. This model is based on experiments from many decades ago [e.g., Refs. 140, 141; reviewed by Davson and Segal (118) and Curl and Pollay (140)]. However, this notion has been questioned, and extrachoroidal sources of CSF certainly exist (142, 143). The matter is unlikely to be fully resolved until there are better methods of separating the contribution of the choroid plexuses and brain tissue via the ependyma. This aspect of choroid plexus function is not considered further here. A detailed review of the functioning of these CSF secretory mechanisms in the adult is provided elsewhere (144).

Proposed mechanisms for water transport across choroid plexus epithelial cells have been reviewed in detail elsewhere (145). We consider some of these mechanisms below but conclude that the evidence is insufficient to decide which is the most likely to explain this phenomenon. There has been a long-standing view that water transport across epithelial cells in a variety of tissues depends on generation of an osmotic gradient by ion exchange mechanisms [e.g., rat ileum (145); common roach (Rutilus rutilus) gallbladder (146); frog gallbladder (147)]. In these studies, the movement of water was attributed to active transport of Na⁺, with water following the osmotic gradient established by this transport. The mechanism was further clarified by the discovery of the Na⁺-K⁺ ATPase membrane pump, which transfers ions across the epithelial cell membranes, 3 Na⁺ in one direction and 2 K⁺ in the opposite direction accompanied by water transfer (148, 149). However, there is no consistent relation between the osmolality of the fluids in compartments on either side of the interface across which Na⁺ and K⁺ transport is occurring. For the choroid plexus, CSF and plasma have similar or at most only slightly different osmolalities (118). These observations led to the suggestion that any osmotic gradient would be highly localized within the epithelial tissue and, although small, would be sufficient to drive the passive movement of water (144, 150, 151). Damkier, Brown, and Praetorius provide a detailed discussion of this hypothesis, which includes consideration of other transported ions such as Cl⁻ that may be involved (144). Because the osmotic gradients identified are only small and in some cases water moves against a substantial osmotic gradient (152), an alternative hypothesis was proposed early in the discussion of water transfer across epithelia. This hypothesis posits that there is active transport of water across epithelial cell membranes (153). MacAulay et al. (145) have reviewed the evidence for this model as well as discussing the limitations of the osmotic gradient hypothesis. However, the problem of resolving whether or not osmotic gradients contribute depends on development of methods to measure local solute concentrations at a very local level where such a mechanism might operate. This was pointed out by Diamond as long ago as 1979 (154), but such methods are still not available.

Recent studies have been more focused on discovering possible mechanisms for active transport of water. Steffensen et al. (155) have used an ex vivo mouse choroid plexus preparation. They suggest that the bidirectional Na⁺-K⁺-2Cl⁻ cotransporter (NKCC1) that is expressed in the luminal membrane of choroid plexus contributes approximately half of the CSF production, via coupling of water transport to ion translocation and with overall outward transport. Gregoriades et al. (156) studied the function of this transporter in isolated choroid plexus cells and report a net inward flux that maintained intracellular Cl⁻ concentration and cell water volume, a conclusion that is the opposite of Steffensen et al. (155). Through the medium of a Journal of Physiology “CrossTalk,” Alvarez-Leefmans challenged the results and conclusions from the group of MacAulay (155). This provoked a response from MacAulay and Rose (157). The “CrossTalk” includes comments from a number of researchers with personal experience of studying transporters in choroid plexus (see supplemental material from Ref. 157 for extensive discussion and rebuttal). The consensus seems to be that this is a difficult problem to study with available preparations and methods, which all have their own strengths and limitations. For example, the ex vivo preparations may better match the native physiological setting of the choroid plexus compared with isolated epithelial cells. However, these isolated cells have lost their normal cellular contacts and the asymmetric environment between the interstitial fluid within the choroid plexus tissue and the CSF. It seems unlikely that the question can be fully answered until it is possible to make the relevant measurements in vivo (Keep and Xiang response in Ref. 157). The cryo-EM structure of NKCC1 was recently published (158). A conclusion from this study was that water transport by NKCC1 would be ∼5 times less than that proposed for the “water pump” hypothesis (159). However, the preparations studied were very different (Ref. 145: human transfected sf9 insect cells; Ref. 125: Xenopus oocytes). Given the plethora of transporters and channels localizing to choroid plexus epithelial cell membranes (FIGURE 6), it is likely that CSF water secretion is ultimately the collective result of many transporters and channels.

In contrast to the detailed work that has been carried out in adult choroid plexus, relatively little is known about ion and water transport mechanisms in development. As described above, ion gradients between blood and CSF develop early and Aquaporin 1 is present in the epithelial cells when they first differentiate (125). Detailed transcriptomic analyses of developing choroid plexus epithelium have been published (93, 94, 160, 161). The results for ion channels and transporters in developing choroid plexus are summarized in FIGURE 6. They have not yet been followed up by the sorts of detailed electrophysiological studies that have been carried out in adult choroid plexuses as mentioned above. There does not appear to be any particular limitation in carrying out such studies since the cells are large enough for successful patch-clamping (77, 162). Nonetheless, based on the transcriptomic findings, Xu and colleagues (84) have recently provided in vivo evidence for the involvement of NKCC1 in efflux of K⁺ and water from CSF by choroid plexus epithelial cells of early postnatal mice. In these studies, endogenous NKCC1 expression by choroid plexus epithelial cells was found to be lower in developing mice than in mature mice. Unexpectedly, CSF from developing mice was found to have higher CSF K⁺ levels than CSF from more mature mice. To test whether the low NKCC1 expression levels could be rate limiting for removing K⁺ from the young CSF, Xu and colleagues overexpressed NKCC1 in choroid plexus epithelial cells during early mouse development in vivo (84). NKCC1 overexpression resulted in decreased CSF K⁺ levels and mice with small ventricles (84, 163). These findings are consistent with the model that choroid plexus NKCC1 functions as a bidirectional ion cotransporter in vivo, and its movement of ions and water have consequences for ventricle size.

4. EARLY CSF AND EXPANSION OF THE VENTRICULAR SYSTEM BEFORE CHOROID PLEXUS FORMATION

Hydrodynamic fluid properties and the composition of CSF components are thought to contribute to normal CNS development. Before appearance of the choroid plexuses, there is already expansion of the ventricular system following the closure of the neural tube. The fluid-filled neural tube is a hallmark of the chordate CNS (15, 164), and a closed neural tube is essential for normal development (15, 165). The cellular lining of the neural tube is a specialized epithelium, the “neuroepithelium,” which comprises the pool of stem cells that will generate the vast majority of CNS cells (166, 167). The neural tube develops into the brain ventricular system, a series of connected cavities filled with CSF.

During human brain development, the cerebral ventricles expand with CSF ∼3–4 wk before choroid plexus tissue appears (15, 168). This raises the question of the early sources of CSF. Ex vivo analyses in chick, rabbit, and cat demonstrated CSF production by isolated neuroepithelium. This is consistent with the hypothesis that the neuroepithelium produces the early CSF (141, 169, 170). In species where the neural tube forms by rolling or folding the neural plate (i.e., neurulation), the early cavity of the tube formed may initially be filled with amniotic fluid captured in the tube (92).

It is technically challenging to trace the origin of fluids in the CNS during development. Current evidence suggests that after neurulation neuroepithelial cells produce CSF in response to colloid osmotic gradients provided by proteoglycans (171, 172) and proteins mainly arising from transfer across the choroid plexuses (173) during the early stages of brain development (172, 174). Comparison of the adjacent neural progenitors both before and after neural tube closure indicates that these progenitors possess the mechanisms that are likely to drive fluid composition changes (92, 175, 176). In early embryonic zebrafish, a rod of cells opens to form the neural tube. It then develops into the ventricular system, filled with fluid that in this pre-choroid plexus stage appears to be made by the embryonic neuroepithelium (177). This fluid secretion is dependent on Na⁺-K⁺-ATPase expression on the neuroepithelial surface (178).

As indicated above, most of the nascent CSF including water and its ionic and protein composition (see below) appears to be derived from the neuroepithelium. However, multiple systemic factors may also contribute to the composition of early CSF (e.g., developing liver and placental circulation). There has been a long-standing view that the blood-brain and blood-CSF barriers are immature, leaky, or even absent in the mammalian fetus. However, this view is not supported by most published evidence (reviewed in Ref. 2). The brain barriers are by definition immature in the fetus as they are in a developing brain. An important physiological question is the extent to which they are functional and how they contribute to specific features of brain development (13) (e.g., Refs. 179, 180). Much of the work summarized in this section is beginning to provide answers. Nevertheless, the belief in a lack of an effective blood-brain barrier early in brain development persists, often without any indication of evidence (e.g., Refs. 179, 180).

The protein compositions of amniotic fluid and CSF diverge after neurulation. This includes differences in the cardinal signaling molecules Shh, bone morphogenetic proteins (BMPs), and retinoic acid (175). Amniotic fluid and CSF provide molecularly distinct fluid environments for the progenitor cells that they contact and support (92). Thus SOX2-positive progenitors of the developing olfactory placode epithelium thrive in amniotic fluid. In contrast Sox2-expressing neuroepithelial progenitors develop in CSF of a distinctive composition at the same age (92).

This early pre-choroid plexus CSF bulk fluid movement has directionality and thus the ability to transport and distribute CSF contents to CNS targets that are distant from the plexuses (181–183). In zebrafish, mutation of the sodium-potassium pump prevents CNS lumen expansion (178), indicating a requirement for ion pumping by neuroepithelial cells during early lumen expansion. Histological data in humans and some functional tracer labeling in chicks have suggested that the early spinal canal additionally undergoes a process of occlusion. This is suggested to contribute to ventricular expansion by reducing the available space for CSF and thus increasing CSF pressure (184, 185). In rat intraventricular tracer injections, measurements of CSF pressure and CSF-to-plasma ratios for radiolabeled sucrose reveal that the ventricular system connects to the spinal canal but is closed to the subarachnoid space until about E16. After this, it opens to the subarachnoid space through pores in the roof of the 4th ventricle and at birth with the formation of the foramina of Magendie and Lushka (8, 186, 187). A connection between early ventricles and the spinal canal has also been shown in zebrafish by histology and three-dimensional ventricular labeling (182). In all species, an initially small ventricular lumen expands greatly with fluid, well before choroid plexus formation, consistent with CSF production by the neuroepithelium. Even after choroid plexus tissue appears, it continues to mature (84, 188). Thus, the neuroepithelium may continue to contribute to CSF production even later in development. An important component of the mechanism involved in expansion of the ventricular system lies in the specialized intercellular junctions (strap junctions) in the neuroepithelium lining the ventricles of the developing brain. These restrict the passage of all but the smallest hydrophobic molecules from CSF into brain interstitial fluid (131, 189, 190). It has been suggested that the high concentration of protein in CSF in the developing brain that remains within the ventricular system because of the presence of these strap junctions exerts a colloid osmotic pressure that contributes to ventricular expansion at a time when CSF secretion is at a low level (173).

Regulation of ventricular hydrostatic pressure is essential for normal brain development. Loss of ventricular pressure by intubation (185, 191) or CSF removal (192) in chick embryos has profound effects on cell proliferation of the developing brain. It results in considerable distortion of its gross morphology. On the other hand, increased CSF volume and/or pressure leads to hydrocephalus. This induces debilitating or even life-threatening signs including brain stem herniation, learning and memory deficits, and epilepsy (193, 194). However, the mechanisms sensing and regulating hydrostatic pressure in the brain remain largely unsolved. Some genetic studies suggest possible mechanisms contributing to congenital hydrocephalus. The genes Trim71, Smarcc1, Ptch1, and Shh, which are expressed in the neuroepithelium of the developing brain in mice, have also been identified in infants with congenital hydrocephalus (195). The authors identified mutations in these genes and suggest that it may be impaired neurogenesis rather than excess CSF accumulation that caused hydrocephalus. It was estimated that together the mutations in these four genes accounted for ∼10% of cases. However, the mechanistic link between abnormal function of these gene products in the neuroepithelium and hydrocephalus is unclear. Another genetic mechanism has been suggested by the identification of PTEN mutations in children with autism spectrum disorder and congenital hydrocephalus. Approximately 20% of children with congenital hydrocephalus are also diagnosed with autism spectrum disorders (196). It is suggested that the phosphatase and tensin homolog (PTEN)-phosphoinositide 3-kinase (PI3K)-mammalian target of rapamycin (mTOR) pathway is a common underlying mechanism for subsets of patients with these disorders (196). Just how these mutations produce such different outcomes is not understood.

Very little is known about how cells contacting the ventricles sense and respond to pressure. This makes it difficult to devise methods of controlling pathological increases of intraventricular pressure. A comprehensive review of the pathology and treatment of hydrocephalus is outside the scope of this review. The main causes of hydrocephalus are obstruction of CSF flow within the ventricular system or in the outflow tracts via the subarachnoid villi or normal-pressure hydrocephalus as occurs in neurodegenerative conditions such as Alzheimer’s disease. These are usually treated by draining CSF by a catheter from the lateral ventricles to the peritoneal cavity. In cases of overproduction of CSF, for example a choroid plexus papilloma, this can be treated surgically by removal of the tumor. CSF production in some cases is reduced by inhibiting carbonic anhydrase with acetazolamide (197). Idiopathic intracranial hypertension (pseudotumor cerebri syndrome) does not result in substantial ventricular enlargement but can be treated by acetazolamide or surgical shunting; this condition has been suggested to be due to abnormal CSF regulation such as reduced cerebrospinal compliance (198). Ventricular enlargement may be associated with subsequent development of schizophrenia (199); elevated volumes of CSF in the subarachnoid space in infants and very young children have been found in some cases to be associated with autism spectrum disorder later in development (200).

The composition of CSF is commonly analyzed for biomarkers of disease (201, 202). Also, the protein composition of early CSF has recently been used to study aspects of forebrain development (175, 176). These studies showed that there was a pronounced downregulation of ribosomal and mitochondrial proteins in the early-forming CSF in mouse embryos. Release of membrane particles into the CSF by progenitor cells of the neuroepithelium lining the ventricles has been reported (203). The adjacent forebrain appears to be a potential source of these proteins. This is suggested by concurrent downregulation of ribosome biogenesis and protein synthesis in the forebrain that occurs during neurulation (175, 176). This is accompanied by maturation of mitochondria and a metabolic shift in neuroepithelial cells from glycolysis to oxidative phosphorylation (175, 176, 204). Since CSF composition reflects normal developmental processes of the brain, future investigations of CNS development will greatly benefit from including CSF analyses. Mechanisms contributing to early-onset neurodevelopmental diseases (reviewed in Ref. 205) and/or hydrocephalus (206) remain poorly understood. However, analysis of the composition of CSF also represents a potential tool to understand progenitor changes that contribute to developmental defects.

5. IMMUNE AND CHEMICAL SURVEILLANCE AND RESPONSE TO INFLAMMATION AND PATHOLOGY

The epithelial cells of the choroid plexus, like many other nonneuronal cells derived from the neuroectoderm, have a remarkable ability to respond to a wide array of infections and diseases. Disease states involving an inflammatory response of the choroid plexuses have been summarized by others (207).These include inflammation and infection, neurodegenerative diseases, stroke, trauma, tumors, schizophrenia, and chronic stress. The inflammatory responses in the choroid plexuses largely involve interaction with circulating peripheral immune cells but can also include interaction with CSF-resident pathological proteins and immune cells. The CNS used to be described as an “immune-privileged” organ, largely because of historical reports of decreased responses to autoantigens present after allografts of non-CNS cells. This notion has been reconsidered, given our modern understanding of the different and controlled immune responses that occur in the CNS. These may involve resident microglia and astrocytes but also small numbers of infiltrating peripheral immune cells (208). Given the unique position of the choroid plexus as an interface between the periphery and the CNS, many important inflammatory pathways converge on, or traffic through, the choroid plexus epithelium. The vast majority of resident immune cells are macrophages (for review see Ref. 209); however, transcriptomic studies have revealed eight subsets of immune cells, including B cells, lymphocytes, macrophages, basophils, mast cells, dendritic cells, monocytes, and neutrophils, in the developing choroid plexus at baseline (63). Each of these cell types expresses specific cytokine, chemokine, and complement component genes. Many peripheral immune cells have also been visualized with a combination of immunohistochemistry and in situ hybridization or FACS isolation of cells from microdissected plexus tissue. These include CD4⁺ regulatory T cells [Th1 and Th2 cells (210)], nonmemory CD8⁺ T cells (211), and CD11c⁺ dendritic cells (212–214) (for review see Refs. 215, 216).

Immune cells respond rapidly to infection, injury, and disease by releasing many proinflammatory cytokines. Choroid plexus epithelial cells interact with these immune cells, respond to immune cell-released cytokines, and may exacerbate CNS inflammatory responses. They can also directly respond by release of potentially noxious stimuli with associated disease pathology. For instance, the plexuses can release cytokine effectors such as interleukins (e.g., IL-1β) and tumor necrosis factor (TNF) in response to parasitic infection such as Trypanosoma brucei (217), Toxocara canis (218, 219), and Schistosoma mansoni (220, 221). Similarly, the innate immune response elicited by lipopolysaccharide (LPS) and mediated by toll-like receptor signaling pathways can directly stimulate the plexuses to upregulate many cytokines (222, 223). Notably, this response to inflammation can be attenuated in the chronic setting (160, 224). This suggests an ability of the choroid plexus to alter its reactive response to inflammation and perhaps respond by decreasing cytokine release during chronic inflammation and disease. In a model of maternal inflammation using polyinosine-polycytidylic acid (polyI:C), the choroid plexus exhibited an inflammatory response in the developing embryonic mouse brain, culminating in a proinflammatory CSF and accumulation of choroid plexus macrophages (225). Elevation of a single cytokine, CCL2, was sufficient to drive choroid plexus immune cell recruitment, activation, and proliferation. In vivo two-photon imaging of the embryonic choroid plexus showed that macrophages migrated away from their usual locations in the choroid plexus and entered the CSF through the “weakened” epithelial barrier, at the tips of choroid plexus villi. This suggests early immune surveillance by macrophages in the developing choroid plexus (225).

Cytokine production is also reported after Neisseria meningitidis infection. In turn, cytokines cause upregulation and secretion of interleukins-1β, -6, and -8, along with TNF and granulocyte-macrophage colony-stimulating factor (GM-CSF) (226, 227). It remains to be seen how many of these responses are due to direct interaction of infectious agents with the plexus epithelium in vivo. Some previous studies have used a human malignant choroid plexus papilloma cell line (228), which may not reflect the in vivo pathology. These immortalized cells may respond very differently from physiologically “normal” cells in vivo, owing to possible artifacts of culture systems or underlying disease biology associated with papilloma tumors.

An important example of an inflammatory response in the choroid plexus is that which occurs in human immunodeficiency virus (HIV) infection. There is evidence that HIV within macrophages may be carried into the plexus (229). Virus has been detected in dendritic-like cells and monocytes in the stromal tissue of choroid plexus in postmortem material from patients (230, 231). One report suggests that the incidence of stromal cell infection is higher in asymptomatic patients (231). Interpreting the nature of the inflammatory response in HIV and the relationship between the viral infection and specific features of the response in the choroid plexus and other tissues is complicated because many people with HIV infection have concurrent infections, presumably because of deleterious effects of the virus on the immune system rendering them more susceptible to other infections. Solár et al. (207) have provided a comprehensive list of inflammatory changes in the choroid plexus including those occurring in systemic inflammation and bacterial, viral (including HIV), and fungal infections. The influence of microbial infections (232), malaria (233), helminths, and tuberculosis (234) on the clinical progression of HIV has been documented. However, there are few reports of the effects of these concurrent infections on the inflammatory response in the choroid plexus. Some examples include toxoplasmosis, tuberculosis, and cryptococcal meningitis (229). Another study reported that in some HIV-positive patients cells in the choroid plexus stroma with a dendritic morphology were immunopositive for class II major histocompatibility complex (MHC) and were also infected with HIV (235).

Whether peripheral LPS acts directly on choroid plexus cells or the effects on choroid plexus are secondary via other primary targets elsewhere in the body (e.g., peripheral immune cells, gut microbiome) remains to be determined. Nevertheless, choroid plexus macrophages in the stromal space respond rapidly by elongating along the perivascular region of choroid plexus blood vessels (106). The direct response of the plexus epithelial cells to bacteria, parasites, and peripheral immune cells may contribute to further responses by other CNS cells such as astrocytes (236) and microglia (237). It is also possible that the plexus itself acts as an entry point for peripheral immune cells like lymphocytes into the CNS, as has been described in multiple sclerosis (238).

Kolmer or epiplexus cells are associated with the microvillous border on the ventricular surface of the plexus villi and also the ependymal lining of the ventricles (for review see Ref. 209). They were first described by Kolmer (239). They have since been identified in a wide variety of species including frog (240), lamprey (241), toad (242), pigeon (243), rat (213, 244), monkey (245), and human (246). The list here is far from comprehensive, as there are numerous studies describing various nonepithelial cells in the choroid plexuses. In a large number of veterinary species and humans it is estimated that Kolmer cells constitute <1% of choroid plexus cells (247). Early studies provided evidence that these cells have phagocytic activity (213, 242) and it was suggested that they are macrophages (242). After systemic inflammatory insult modeled by peripheral injection of the bacterial cell wall endotoxin LPS, epiplexus cells are immune reactive for complement receptor 3, MHC class I and II antigens, and leukocyte common antigen (CD45) (213), suggesting they may also play a role in immunological responses. Single-cell transcriptional studies have begun to examine the identities of these cells (248). However, more studies are needed to elucidate whether these cells constitute a pure population of macrophages and to clarify their exact functional roles. Interstitial macrophages appear to have a different lineage and later in development are replaced by monocytes (248, 249). Imaging studies in adult brain have shown that epiplexus cells home to the site of acute choroid plexus tissue injury (106).

Other than immune interactions, which are common to many (if not all) neurological disorders, there are several notable pathologies specific to the choroid plexus. In neurodegenerative disease, changes in choroid plexus epithelial cells have been reported in addition to those in the brain itself. These reports describe inclusions of pathogenic proteins such as amyloid beta (250), Tau (251), and alpha-synuclein (252). Reports suggest that these inclusions lead to impaired oxidative stress and mitochondrial function (253). It is most likely they are taken up from the CSF and not produced locally by the choroid plexus epithelium. It has been difficult to unravel the exact functional changes in the plexus consequent on this uptake of amyloid, as current mouse models appear not to replicate the pathology seen in humans (250). Other rodent tools take advantage of genetic deletion of the amyloid beta-binding protein gelsolin. In the APP/PS1 mouse model of Alzheimer’s disease this can lead to epithelial cell death (254). These APP/PS1 mice express human transgenes for both amyloid (APP) bearing the Swedish mutation and a mutated presenilin-1 (PSEN1, PS1) gene (in this study an M146L mutation), both under the control of the Thy1 neuronal promoter, and express human APP approximately three times higher than endogenous murine App. An important gain-of-function experiment in this study highlights that overexpression of gelsolin in these mice reduces nitric oxide production and decreases cell death (254). Comparative studies also report that choroid plexuses obtained from Alzheimer’s disease patients corroborate amyloid presence and decreased gelsolin levels. Other interactions between the plexus and amyloid in Alzheimer’s disease are also reported. For instance, APP/PS1 mice have elevated amyloid levels when crossed with Ttr^−/− mice that lack transthyretin (255). Like gelsolin, transthyretin can bind amyloid taken up from the CSF (256) and may mediate its removal into the bloodstream. Other age-related pathologies reported in choroid plexus epithelial cells include Biondi ring tangles (257, 258). Although Biondi ring tangles are not found only in the Alzheimer’s disease brain, increased numbers are thought to be among the earliest indicators of the disease. In mice that express mutant Huntingtin (HTT), excess CSF production appears to drive hydrocephalus (259). Mutant HTT expression by plexus epithelial cells appears to drive this increased CSF production through unknown means, as cells are grossly morphologically intact and have no noticeable differences in aquaporin (namely AQP1) water channel numbers or localization.

Continued investigation of the differences in physiological and pathological roles of the choroid plexuses will no doubt continue to uncover their important functional roles in the earliest stages of disease. The choroid plexuses are a major entry point for multiple peripheral components. These components include serum proteins (see above), immune cells, and pathogens. The choroid plexuses are also positioned as an important link between resident CNS cells such as microglia and astrocytes and the peripheral mediators of their multiple “reactive” states. Indeed, recent evidence of interferon-responsive reactive astrocytes around the ventricles (236) adds to this mounting evidence. As interferon is largely not produced by CNS cells (260, 261), the strategic location of this reactive astrocyte substate implicates infiltration of either interferon or interferon-releasing peripheral immune cells. The cells involved may be either those infected with viruses (for type I interferon) or T cells, natural killer cells, and macrophages (for type II interferon) (262). Continued research on the routes of entry across the choroid plexus epithelium, and how this may alter in the early stages of infection and disease, will no doubt highlight novel druggable targets for many pathologies.

6. REGULATION OF CHOROID PLEXUSES FUNCTION BY SEX HORMONES AND OTHER FACTORS

Secretion as well as water and ion transfer are the functions of the choroid plexuses that have been most studied, particularly in adult choroid plexus epithelial cells, as summarized in sect. 3. The extensive studies by Brown and colleagues and by Praetorius, Damkier, and colleagues were comprehensively summarized in the excellent review by Damkier et al. (144). The subsequent independent review by Hladky and Barrand (130) adds a valuable assessment of the evidence available up to that time. Although these studies provide substantial detail of the mechanisms involved, as pointed out by Damkier et al. (144), there is little information available on how these mechanisms are controlled. Nilsson et al. (263) summarized the available data on potential endocrine regulators of choroid plexus function as well as the influence of sympathetic, parasympathetic, and peptidergic innervation. They propose that the available evidence suggests that the sympathetic nervous system has a tonic inhibitory influence on CSF production. This is probably due to a cyclic AMP-mediated decrease of Na⁺-K⁺-ATPase activity in the choroid plexus epithelium. They also highlight the high concentration of the 5-hydroxytryptamine (5-HT)2C receptor in the choroid plexus itself, which is believed to be the highest in the whole brain. Selective activation of the 5-HT2C receptor in choroid plexus epithelial cells leads to sustained intracellular calcium activity. This triggers increased vesicle fusion events in vitro or eventual apocrine secretion in vivo (106). Several neurotransmitter and endocrine mediator receptors have been identified in the choroid plexus. However, many studies to date have used nonphysiological concentrations of CSF ligands, and in many cases the actual CSF concentrations are not specified. Further studies are required to determine the functions of these receptors.

Santos et al. (264) have reviewed the evidence, including that from their own studies (265), for sex hormones that are expressed in the choroid plexuses and genes that appear to be differentially expressed in male and female plexuses. These include androgen receptor, three estrogen receptors, and three progesterone receptors. The authors also tabulate summaries of published gene expression data for choroid plexuses in functional categories. There were some differences in gene expression between sexes and in ovariectomized female and orchiectomized males compared with sham-treated controls. They point to the expression of aldosterone, angiotensin, and arginine vasopressin, which are known to regulate water and electrolyte homeostasis in other tissues. So far, there does not appear to be evidence on what their role in choroid plexus function may be. In an attempt to deal directly with the problem of studying control of choroid plexus function, Haoui et al. (266) have provided a detailed description of isolation and primary cultures of lateral ventricle choroid plexus epithelial cells. These can be used for patch-clamping individual cells and investigating possible functional control mechanisms at a cellular level, for example, steroid membrane signaling. There are limitations to such an approach. It is not clear to what extent isolated choroid plexus cells in vitro retain their in vivo functional capacity or to what extent findings on individual cells will translate to the function of the plexuses as a whole. Also, patch-clamping of choroid plexus cells in isolated fragments of plexus showed regulation of apical K⁺ and Cl⁻ channels by 5-HT (5-hydroxytryptamine) (267). In vivo 5-HT appears not to affect the mRNA expression of transferrin and transferrin receptor; nor does sympathetic nerve activity (268). Thus, although in vitro approaches may appear to be a start on showing what may be possible control mechanisms, they do not appear to have progressed very far. Catalogs of the data available from the literature of genes expressed in the choroid plexus also highlight putative functions more usually associated with olfactory organs (269, 270) and taste receptors (270, 271). Their survey of the literature shows that many of these genes are expressed in a wide range of tissues. Whether they are functional in any of these tissues, and in the choroid plexuses in particular, is unclear. Tomas et al. (272) have reported that the taste receptors TAS2R109 and TAS2R144 and the taste-related genes Plcb2 and Trpm5 were downregulated by 17β-estradiol and progesterone both in vivo and in vitro. They suggest that these receptors may monitor the composition of blood and CSF on the plexus epithelial cells and thus contribute to control of choroid plexus function.

Another aspect of control of choroid plexus function is the influence of plexus blood flow. This was extensively studied in the period from 1970s to the 1990s. In their comprehensive review, Damkier et al. (144) summarize some of the published information. They state that “The choroid plexuses are perfused with blood at a rate of 4 ml/min per gram of choroid plexus tissue, which is approximately 10 times higher than the rate of the blood supply to the brain parenchyma (157). This high rate of perfusion means that blood flow is not normally rate-limiting for the process of CSF secretion.” The paper cited (77) deals with the morphometry of the choroid plexus, not blood flow. This statement is repeated in Ref. 273, but without a reference. The value of 4 mL/min per gram of choroid plexus tissue perhaps comes from the studies of Maktabi et al. (274) or Szmydynger-Chodobska et al. (9). Evidence that the choroid plexus blood flow is ∼10 times higher than the rate of the blood supply to the brain parenchyma has been provided in Refs. 274, 275. The possible relation between plexus blood flow and rate of CSF secretion is an important physiological consideration. TABLE 1 summarizes results from a range of studies in different species using a variety of techniques. In most reports blood flow was estimated for the lateral ventricular choroid plexus, but in four papers the flow in the 4th ventricular was also measured. All four studies agreed in finding that flow was higher in the plexus of the 4th ventricle (TABLE 1). For both choroid plexuses there was a fourfold range of values. The differences appear to be mainly due to the different methods used. The microsphere results for rabbits were generally lower than those for rats with the same technique. In only two studies were measurements made in unanesthetized animals. The values in rat (288) were similar to those reported by Pollay and colleagues (290) but much less than those of Szmydynger-Chodobska et al. (9); the methods used were different in all three studies. In rabbits Faraci et al. (275) found similar values for chloralose-anesthetized and unanesthetized animals with the same microsphere technique, which is perhaps more convincing evidence that the anesthetic used did not have much effect on blood flow in the choroid plexuses. Faraci et al. (275) also found that although anesthesia did not have much effect on choroid plexus blood flow, cerebral blood flow was increased (0.30 ± 0.02, 0.33 ± 0.02 mL/g/min anesthetized and 0.53 ± 0.04, 0.47 ± 0.05 mL/g/min awake). The claim that the high rate of plexus blood flow could be taken as indicating that flow is not normally rate-limiting for the process of CSF secretion is debated. Cserr (291) considered that blood flow in the choroid plexus may limit CSF secretion. This view seems to have been based on a personal communication from Pappenheimer to Ames concerning his estimations of factors influencing CSF secretion (292) together with some observations on the vasoconstrictor activity of l-norepinephrine (noradrenaline), which reduced CSF flow in cats, although actual blood flow was not measured (293). In later studies in which both choroid plexus blood flow and CSF secretion rate were measured, it was reported that several agents affected both. Thus, Maktabi et al. (274) found that intravenously applied angiotensin II reduced choroid plexus blood flow in a dose-dependent manner in anesthetized rats. Chodobski et al. (294) showed that angiotensin II in low doses applied intrathecally reduced CSF formation; influence of specific inhibitors suggested that the effect was mediated by AT₁ receptors. They did not measure choroid plexus blood flow in these experiments and do not comment on the findings of Makabi et al. (274). Thus, the exact mechanisms relating effects of angiotensin II on choroid plexus blood flow and CSF production remain to be elucidated.

Table 1.

Choroid plexus blood flow in different species by a variety of methods

Species	Sex	Age	Anesthesia	Detection Method	4th Ventricle, mL/g/min	Lateral Ventricle, mL/g/min	References
Rabbit	M	Adult	Pentobarbital	Octanol bubble		2.86	(136)
Cat	M and F	Adult	Chloralose	Microspheres		3.01 ± 0.32	(276)
Cat		Adult	Pentobarbital	Microspheres		2.26 ± 0.28	(277)
Cat		Adult?	Pentobarbital	Microspheres		2.40 ± 0.56	(278)
Rat		Adult	Pentobarbital	Diffusible indicator	1.34 ± 0.09	0.85 ± 0.05	(279)
Rabbit		Adult	Ketamine, xylazine	Diffusible indicator	1.03 ± 0.03	0.82 ± 0.06	(279)
Dog		Adult	Pentobarbital	Hydrogen electrode		0.73 ± 0.12	(280)
Rabbit		Adult	Pentobarbital	Microspheres		2.78 ± 0.68	(281)
Sheep	F	Adult	Pentobarbital	Microspheres		6.01 ± 0.44*	(282)
Sheep	F	Adult	Pentobarbital	Microspheres		3.81 ± 0.44; 7.25 ± 0.75	(283)
Pig		Newborn	Halothane	Microspheres		2.81 ± 0.25	(284)
Rabbit (NZW)		Adult?	Chloralose	Microspheres		3.85 ± 0.73	(285)
Dog		Adult?	Pentobarbital	Microspheres		2.87 ± 0.26	(285)
Rabbit (NZW)		Adult?	Chloralose	Microspheres		3.24 ± 0.4; 3.30 ± 0.35	(275)
Rabbit (NZW)		Adult?	None	Microspheres		3.65 ± 0.54; 3.53 ± 0.59	(275)
Rabbit (NZW)		Adult	Chloralose	Microspheres		3.69 ± 0.26	(286)
Rabbit (NZW)		Adult	Chloralose	Microspheres		4.07 ± 0.82; 3.98 ± 0.60	(274)
Rat (Long Evans)	M	Adult	None	Autoradiography	4.25 ± 0.35	5.25 ± 0.39	(287)
Rat (SD)	M	Adult	None	Microspheres	1.56 ± 0.05	0.83 ± 0.01	(288)†
Rat		Adult	Pentobarbital	Indicator fractionation N-isopropyl-p-[125I]iodoamphetamine	4.12 ± 0.08	3.30 ± 0.07	(9)
Rat (WKY)		Adult	Urethane	Indicator fractionation		2.4 ± 0.08	(289)
Rat (SHR)		Adult	Urethane	Indicator fractionation		2.82 ± 0.21	(289)

Most studies were carried out in anesthetized animals under well-controlled physiological conditions (normal Po₂, Pco₂). In most studies, plexus blood flow was only measured in the lateral ventricular plexus. When it was also measured in 4th ventricular plexus it was generally higher than the lateral plexus. *Increased by 27% at Pco₂ 58 mmHg; much greater increase in cerebral blood flow at this Pco₂ (199%). †Hypercapnia (Pco₂ 62 mmHg) did not increase choroid plexus blood flow. In animals anesthetized with chloralose, anesthesia was induced with chloroform or thiopentone. F, female; M, male; ns, not significant; NZW, New Zealand White; SD, Sprague Dawley; SHR, spontaneously hypertensive rat; WKY, Wistar Kyoto.

Faraci et al. (286) showed that vasopressin reduced both plexus blood flow and CSF secretion in rabbits, an effect that appeared to be mediated by vasopressin V1 receptors. Norepinephrine (noradrenaline) has been found to decrease choroid plexus blood flow in dogs (285) and production of CSF in rabbits (295). As pointed out by Faraci et al. (286), these observations are compatible with a causal relation between blood flow and CSF production. Nilsson et al. (296) found that intraventricularly administered vasoactive intestinal peptide (VIP) decreased CSF production (30%) but increased choroid plexus blood flow (20%). They interpreted this as indicating that CSF production and plexus blood flow are not directly coupled, but an alternative explanation is that VIP is acting on different mechanisms.

Szmydynger-Chodobska et al. (9) carried out a detailed developmental study of choroid plexus blood flow in anesthetized postnatal and adult rats. They showed that the values were similar at 2 and 3 wk of postnatal age, with a substantial increase by 5 wk in both lateral and 4th ventricular plexuses. There was a further increase in 4th ventricular flow by adulthood. Williams et al. (287) found slightly higher values and did not find a difference between flows in the different plexuses (TABLE 1). The lack of difference may have been due to the different methods used. Szmydynger-Chodobska et al. (9) comment on published findings on CSF production, which increases progressively during the first 3 wk of postnatal development in the rat (72). Given the many developmental changes occurring during this period, it is hard to determine whether the increase in blood flow can account for the increase in CSF production.

The only direct evidence that there may be a direct effect of blood flow to the choroid plexus and CSF production comes from experiments in which the plexus of adult sheep was perfused at different rates. These experiments showed a reduction in the rate of vascular perfusion from 0.63 ± 0.08 to 0.25 ± 0.04 mL/min and decreased CSF secretion from 21.37 ± 2.69 to 9.15 ± 1.51 µL/min (297). In addition, these authors also showed that a decrease in plexus perfusion directly affected some specific plexus epithelial cell transport functions including ion and sugar transport (297, 298). The studies outlined above on factors that influence choroid plexus blood flow and CSF production in vivo indicate that the mechanisms linking these functions are complex and further work is required to clarify them. Since the turn of the century, there have been almost no in vivo animal studies reported on choroid plexus blood flow and CSF production. However, several clinical studies using modern imaging techniques have been published (299–301).

The increasing emphasis on developing in vitro choroid plexus models and application of molecular techniques over the past 20–30 years may provide some insights into choroid plexus mechanisms. However, in vitro approaches may not be particularly suitable for providing answers to physiological questions about control of choroid plexus function raised by the molecular screening studies discussed above. The consequent loss of expertise in in vivo experimental techniques will make it more difficult to make progress on understanding control of blood flow and CSF production and their possible links.

7. POSSIBLE ROLE OF CHOROID PLEXUSES IN CIRCADIAN RHYTHMS AND SLEEP

CSF composition and fluid dynamics change throughout the day, and some of these changes match circadian rhythms or diurnal behavior. CSF production volume varies across the 24-h light-dark cycle, and CSF distribution between brain parenchyma interstitial fluid, ventricles, and cervical lymph nodes also varies across the 24-h day (302–304). Concentrations of hormones such as cortisol and corticotropin-releasing hormone (CRH) cycle in CSF (305), and levels of melatonin in the CSF better match pineal gland activity than do levels in the serum (306–308). Furthermore, results from over a century of research suggest that CSF composition not only contains biomarkers of circadian rhythmicity but also is likely to play key roles in relaying diurnal information to brain tissues. Early CSF transplant studies suggest that CSF can carry circadian cues for drowsiness (309, 310) and satiety (311). Experimental models indicate that diffusible factors released from the suprachiasmatic nucleus of the hypothalamus into the CSF can mediate circadian locomotion rhythms (312–316). These diffusible signals include the neuropeptides VIP, arginine vasopressin (AVP), and gastrin-releasing peptide (GRP) (313), but others may also exist.

The specialized choroid plexus epithelial cells that directly modulate CSF content also display their own cell-autonomous circadian rhythmicity as evidenced by oscillating expression of molecular clock genes including Bmal1, Clock, Cry2, Per1, and Per2 (317–320). The oscillations are affected by sex hormones (318, 319, 321). The possibility that these oscillations influence choroid plexus output is explored more extensively elsewhere (322). However, if choroid plexus rhythmicity alters the CSF environment, it may be a key component of the brain circadian axis. This role is suggested by explant experiments whereby choroid plexus-supplied conditioned media could reset the rhythmicity of a suprachiasmatic nucleus explant (317). Systemically, the overall circadian locomotor period of mice in constant darkness is lengthened by conditional knockout of the essential clock gene Bmal1 from all multiciliated cells (with Foxj1^Cre), a population that includes plexus epithelial cells (317). Efforts have been made to identify downstream effectors of this circadian rhythmicity on choroid plexus output. However, mRNA quantification has either identified no change in potential targets, as for the water channel Aqp1 (320), or smaller changes for Clu (APOJ) and Ttr in rats (321). Consideration must also be made for the fact that diurnal changes are not always exclusively regulated by the circadian clock, as evidenced by the example of Ttr (transthyretin), which is also induced in cultured choroid plexus epithelial cells by glucocorticoids and is upregulated in liver, plexus, and CSF by acute and sometimes chronic stress (323) or by nicotine (324). Thus, further mechanistic study is needed to determine behavioral influences on choroid plexus output and to define functional roles for daily changes in CSF composition.

Understanding the choroid plexus-CSF-circadian axis is of vital importance, as abnormal CSF components have been associated with a number of neurological conditions that copresent with circadian disruptions. These conditions include hydrocephalus, autism spectrum disorder, Alzheimer’s disease, bipolar disorder, and schizophrenia (325–331). For example, sleep disruption is a diagnostic criterion for major depression, bipolar disorder, posttraumatic stress disorder, generalized anxiety, and other mood disorders (332). CSF components, including those that are diurnally regulated, often originate from outside the choroid plexus. Examples are peripheral adrenal glands (cortisol), hypothalamus (corticotropin-releasing hormone), suprachiasmatic nucleus (vasoactive intestinal peptide, arginine vasopressin, gastrin-releasing peptide), and pineal gland (e.g., melatonin) (305–308, 313, 315). Daily changes in choroid plexus function, including its barrier function, could also lead to altered CSF concentrations of these peripherally derived contents.

In addition, manifestations of disease morbidities, including psychiatric symptoms, occur at specific times of day, with 6 AM being the time with highest risk for all causes of death (333). Blood-brain barrier permeability, including carrier-mediated transport, cell-cell junction permeability, and efflux pump activity, also varies diurnally. Some barrier mechanisms have been reported to be more permeable during resting phases (334). Recently, neuronal activity (higher during waking hours) was shown to activate endothelial cell ATP-binding cassette (ABC) efflux transporter expression at the blood-brain barrier through a circadian mechanism (335). Although the blood-brain and blood-CSF barriers each comprise some unique components, many key permeability mechanisms are similar and some are disrupted in CNS disease (336). The close connection between circadian rhythmicity and barrier properties suggests that taking advantage of rhythmic symptoms and drug efflux and availability could lead to more effective drug dosing. Because the choroid plexus is a key brain barrier that can modulate drug efflux (336), it could also be a target for CNS drug access at appropriate circadian times.

8. ENTRY OF DRUGS INTO THE BRAIN VIA THE CHOROID PLEXUSES

8.1. Adult: Potential Drug Delivery Systems

A major impediment to discovering effective treatments for neurological and neuropsychiatric disorders is poor understanding of their biology and of brain barrier interface mechanisms. The latter limit the entry of drugs and toxins into the brain (337, 338). Pardridge (339) has estimated that 98% of small-molecule drugs and 100% of larger-molecule therapeutics developed by pharmaceutical companies to treat neuropsychiatric disorders are ineffective. He ascribed this to studies being conducted in vitro with no account being taken of their penetrability across the brain barrier interfaces. Despite a huge investment of time and resources in developing in vitro blood-brain barrier models (e.g., Refs. 340–342), this approach has so far not resulted in any new drugs for treatment of neurological and psychiatric conditions. These problems, and the lack of suitable models for specific brain disorders, have not yet led to development of many clinically useful drugs (337).

In some clinical circumstances (e.g., CNS infections, cerebral tumors, and intractable pain), brain barriers can be circumvented by direct injection of drugs into the CSF (343, 344), but this technique can carry significant risks for patients (344).

Targeting drugs to the CNS via the choroid plexuses seems to have received much less attention than via the blood-brain barrier itself. Thus a simple search in PubMed for the term “drug targeting CNS choroid plexus” provides 34 entries, but if “choroid plexus” is replaced by “blood-brain barrier” there are 1,308 entries (PubMed May 25, 2022). Gonzalez et al. (345) comment that the choroid plexuses have largely been ignored as a route for bypassing the blood-brain barrier, a view with which Strazielle and Ghersi-Egea (346) concur. Gonzalez et al. (345) used phage display to identify peptides that might serve as ligands for drugs and other therapeutic agents for delivery from blood to CSF via the choroid plexuses. Saunders and Habgood (347) and Strazielle and Ghersi-Egea (346) have provided comprehensive reviews of biological transport mechanisms in the choroid plexus that might be harnessed to promote drug entry into the CNS. Strazielle and Ghersi-Egea (346) point out that most small-molecule candidates are lipid soluble and would be expected to enter the CNS. However, they rarely do so because their entry is limited or prevented by ABC transporters and solute carrier (SLC) transporters. Their review (346) focuses on a large range of naturally occurring biological mechanisms including carrier-mediated influx transporters, transcytosis, and receptor-mediated transcytosis. Many of these transporters have also been studied in relation to drug delivery via the blood-brain barrier but in neither case have so far yielded a clinically effective system.

Strazielle and Ghersi-Egea (346) also review transcytosis mechanisms specific to the choroid plexus, namely folate and plasma proteins (with albumin the most studied). It has been shown in several species that a transport mechanism across choroid plexus epithelial cells is species dependent. For example, in fetal sheep human albumin is transported into CSF to a steady-state level that is around half that of endogenous albumin (99). In rat fetuses the CSF plasma ratios for human albumin were 4–5 times those for bovine albumin (348). In mice the CSF plasma ratios were only about half for human albumin compared to endogenous albumin in postnatal and adult animals (98). Early studies of computed hydropathy profiles showed species-related differences in molecular structure that could explain the species differences in albumin transport (349). The molecular structures of many species of albumin are now known (https://www.rcsb.org/). Molecular modeling and comparative hydropathy plots should provide the possibility of identifying the part of the albumin molecule involved in transport across the choroid plexus epithelial cells. Synthesis of a peptide of the relevant part of the albumin molecule would provide a vector to which drugs could be attached. Appropriate binding sites on the albumin molecule would also have to be considered when designing a transport peptide (350). A potential receptor and transcellular pathway for albumin transfer across choroid plexus epithelial cells has been identified. A molecular screen of fetal mouse choroid plexus identified three candidates with plasma protein binding properties (161). These were Secreted protein acidic and rich in cysteine (Sparc) and Glycophorin A (Gypa) and C (Gypc). A combination of single-cell PCR and immunohistochemistry showed a significant number of plexus epithelial cells that were positive for both SPARC and GYPA and plasma protein/albumin immunoreactivity. SPARC, GYPA, and GYPC were identified in choroid plexus epithelial cells in the embryo but not in the adult. The subcellular distribution was consistent with transport of albumin from blood to CSF (161). This finding was confirmed by using in situ proximity ligation assay to localize albumin and the putative receptor SPARC (98). A model of albumin transport across a subpopulation of choroid plexus epithelial cells in the fetus based on these studies is shown in FIGURE 7.

FIGURE 7.

Proposed transepithelial pathway for albumin through choroid plexus epithelial cells. Glycophorins A/C (GYPA/C) in the endothelial cells deliver albumin to the basement membrane (1) that then can be taken up into plexus epithelial cells by GYPA/C or Secreted protein acidic and rich in cysteine (SPARC) (2). From here, albumin moves along a SPARC-specific pathway through tubulocisternal endoplasmic reticulum (3) and Golgi (4a) or via a VAMP-mediated pathway in vacuoles, lysosomes, or multivesicular bodies (4b). On the apical surface of the plexus epithelium, GYPA/C may be involved in efflux of protein from the cell into the cerebrospinal fluid (CSF) (5), as validated by extensive GYPA immunoreactivity in embryonic plexus. In the adult, the lack of immunoreactivity in the endoplasmic reticulum and Golgi along with increased expression of gene products for VAMP molecules suggest that the majority of transport occurs via VAMP-mediated vesicular/lysosomal transport (4b). Adapted from Ref. 161, with permission from PLoS One.

Bryniarski et al. (351) have provided a comprehensive summary of these protein and other choroid plexus transport mechanisms that might be harnessed for drug delivery to the CSF. They also consider the current knowledge of movement of CSF and extracellular fluid within the brain, which would be essential for effective delivery of a drug to its therapeutic target following entry into the CSF.

A mechanism for delivering a drug into the CSF via the choroid plexus is, however, only part of the problem. In terms of a therapeutic effect, the drug would only be effective if it reached its target. There is a difference of opinion on how well a drug delivered to the CSF would distribute in the brain. A long-standing view has been that because movement of molecules within the brain parenchyma is by diffusion in the brain extracellular space, drugs in the CSF would not reach most therapeutic targets (352). However, many years ago Cserr showed that a variety of molecules, including albumin, when injected into the caudate nucleus were cleared at a much faster rate than could be accounted for by diffusion. The clearance was not related to molecular size, suggesting convective movement. Clearance via the CSF was only 10–20% of the total (e.g., Ref. 353). In recent years there have been extensive studies of movement of endogenous molecules in the perivascular spaces using modern imaging techniques including two-photon microscopy. There are divergent views on how the results of such experiments should be interpreted (130, 354, 355), but it seems clear that there is movement of endogenous molecules of various sizes within perivascular spaces at much faster rates than by diffusion. Nevertheless, it remains to be shown whether any of the proposed drug targeting strategies across either the blood-brain barrier or choroid plexus can be translated to clinical use.

8.2. Fetus and Newborn: Potential Hazards of Drugs Taken by Pregnant Women

A key question for pregnant women and their clinical advisors is whether a drug that is required for maternal health during pregnancy will reach the fetus and, in particular, will reach the fetal brain. If so, may there be detrimental effects on subsequent brain development and behavior? The background to this problem has been reviewed elsewhere (356). The relevance of the choroid plexuses to this problem is that they develop well in advance of much of the vascularization of the brain. The plexuses are thus likely to be the main route of entry for drugs that get into the brain early in development (8). None of the few experimental studies on this problem involved measurement of drug in fetal blood or CSF after administration to a pregnant animal. This means that the relative role of the blood-brain and blood-CSF barriers in restricting or permitting drug entry into the developing brain cannot be assessed; nor can the role of the placenta be determined. This problem has begun to be rectified in a series of recent experiments in pregnant rats. The approach was to inject pregnant rats, their fetuses, or neonatal animals with drugs. The levels of the drugs were measured in plasma, CSF, and brain with liquid scintillation for radiolabeled drugs and/or liquid chromatography-mass spectrometry for unlabeled drugs (357–360). Entry into each compartment is represented by ratios, in the case of CSF as CSF-to-plasma ratios, as this allows for variations in plasma levels of the drug. This is illustrated in FIGURE 8A for digoxin, cimetidine, lamotrigine, valproic acid, ivacaftor, and paracetamol/acetaminophen at E19 (2 days before birth), P4, and adult. It is clear from these results that for these drugs there is no relation between lipid solubility and level of entry into CSF. Also, many of the drugs are at a low level in CSF as early as E19, indicating that entry-limiting mechanisms (most likely ABC efflux transporters or related enzymes) are already active early in brain development. In experiments in which animals were administered drugs over 4–5 days this chronic treatment increased the CSF-to-plasma ratios for cimetidine, digoxin, and particularly acetaminophen. This was perhaps because of saturation of efflux mechanisms (FIGURE 8B).

FIGURE 8.

Mean cerebrospinal fluid (CSF)-to-plasma ratios for ³H-labeled drugs compared to lipid solubility (log₁₀D_octanol partition coefficient) following acute and chronic treatment. A: permeability studies in rats at embryonic day (E)19, postnatal day (P)4, and adult. Intraperitoneal drug injection to fetuses (except cimetidine via mother) samples taken at 30 min (acute treatment). Drugs were compared with 3 molecules of increasing lipid solubility (sucrose, l-glucose, glycerol) that enter the CSF by diffusion. The similarity of results for sucrose and l-glucose at the 3 ages shows that the marked increase in CSF turnover that occurs during brain development (13) does not account for the decline in CSF-to-plasma ratios with age for all the drugs tested. Ratios for all drugs were <100%, indicating restricted entry into CSF that was different for each drug and not related to lipid solubility. Age-related changes likely to be due to development of efflux mechanisms (ABC transporters or related metabolic enzymes). Log₁₀D_octanol computed from Tetko et al. (361). B: chronic measurements of CSF penetration of same drugs in A. Thirty-minute intraperitoneal drug exposure with ³H-labeled drugs after 4- to 5-day treatment with clinically relevant doses (chronic treatment). All ratios were <100%, indicating some degree of restriction. Compared with acute doses (A) at E19 there was an increase in ratios for acetaminophen/paracetamol, cimetidine, and digoxin, suggesting that the capacity of the efflux system was exceeded. At P4, the drugs had similar ratios in the 2 treatment groups. In adults the ratio for acetaminophen/paracetamol was reduced, suggesting upregulation of an efflux transporter in response to longer-term treatment. There was a significant increase in expression of Abcb1b (P-glycoprotein) in adult choroid plexus, which could account for this (362). Note change of the x-axis to log₁₀D_octanol positive values only in right-hand column (chronic). Data from Ref. 357, with permission from Scientific Reports; additional data from Ref. 360, with permission from F1000Research, and Ref. 363, with permission from ACS Pharmacology & Translational Science.

An important question is whether the apparent effectiveness of mechanisms restricting entry and increase with age can be accounted for by known ABC efflux transporters and their related enzyme systems. A summary of the ABC transporters that have been identified in choroid plexus is shown in FIGURE 9, with examples of the modes of efflux highlighted in FIGURE 10. The most studied of the efflux transporters is P-glycoprotein (ABCB1 in humans, ABCB1A and ABCB1B in rodents). Digoxin is well established as a substrate for this transporter. P-glycoprotein expression was detected at E19, P5, and adult and did not change with chronic treatment (357). The much smaller CSF-to-plasma ratios for digoxin at E19, P5, and adult compared with the other drugs suggests that this efflux transporter was effective in the choroid plexus even at fetal stages of development. In the case of acetaminophen at E19, none of the eight genes (Abcb1a, Abcb1b, Abcg2, Abcc3–5) whose expression was estimated by RT-qPCR changed expression in response to chronic treatment. The substantial increase in CSF-to-plasma ratio at E19 is probably a reflection of efflux mechanisms being saturated. At P4, expression of Abcc1 was decreased to nearly half the control value and Abcb1a to a third. This may indicate that acetaminophen is a substrate for these transporters but does not account for the small decrease in ratio with chronic treatment at P4. The decrease in ratio in the adult in response to chronic treatment did correlate with an upregulation of Abcb2. In a detailed RNA sequencing (RNA-seq) study of lateral ventricular choroid plexus at E19, P5, and adult, two transporters previously identified as highly expressed in adult choroid plexus (357) were confirmed and found to be highly expressed at E19 with expression levels that were increased in adults. A third transporter, Abcc5, was the highest expressed at E19, and its expression was approximately halved in adult female choroid plexus (357). Analysis of the RNA-seq data on the effect of acetaminophen on gene expression of these ABC transporters and their related enzymes showed complex changes in regulation. These will require further investigation to determine their contribution to permeability changes to acetaminophen in developing brain.

FIGURE 9.

Localization of ABC efflux transporters in choroid plexus epithelial cells of the fetus and adult. Data for localization of the proteins and developmental gene expression changes are from Refs. 63, 161, 364–370. See also FIGURE 10 for an overview of efflux mechanisms at the blood-cerebrospinal fluid (CSF) barrier. Most transporters require movement of the molecule through the entire phospholipid bilayer, either under control of diffusion gradients or by energy (ATP)-mediated transfer. An exception to this is PGP (Abcc1), which collects molecules for efflux directly from the bilayer itself before entry into the cell cytoplasm.

FIGURE 10.

Efflux mechanisms in the apical membrane of a choroid plexus cell. Multiple mechanisms exist for removal of molecules out of the cerebrospinal fluid (CSF). Example transporters given [e.g., PGP (Abcb1), MRPs (ABCC family transporters, see FIGURE 9), BCRP (Abcg2)]. “Others ”covers organic anion and cation acid transporters in FIGURE 9. Based on efflux transporter functions in brain barrier cells and adapted from Ref. 12, with permission from Frontiers in Pharmacology. ADP, adenosine diphosphate; ATP, adenosine triphosphate.

A particular clinical problem that arises in the care of women with epilepsy is whether or not to continue or change medication required to control their seizures. It is usually continued, perhaps with some modification, as discontinuation poses hazards for both the mother and her fetus (371). However, antiepileptics are known to cause problems in a proportion of fetuses. In particular, there is high incidence of congenital malformations with valproate (372), and this drug and other antiepileptics have been reported to cause developmental and behavioral problems in the postnatal offspring (373). Valproate continues to be used in regions, Nordic countries and Australia for example, as it is the most effective treatment for some forms of epilepsy (374). A key question for studying the potential deleterious effects of antiepileptic drugs is measuring the extent to which the drugs enter the fetus, its brain, and CSF. An example of such a study is illustrated in FIGURE 11. Here results for two antiepileptic drugs, valproate and lamotrigine, are shown for experiments using embryonic, postnatal, and adult rats. There were strikingly different age-related patterns of entry for these two drugs. Valproate shows a marked decline from ∼70% CSF-to-plasma ratio at E19 to ∼10% in the adult. In contrast, there was a doubling in the CSF-to-plasma ratio for lamotrigine between E19 and P4. This increase was followed by a slight decline in the adult. There are conflicting data in the literature on which ABC transporters valproate and lamotrigine are substrates for. On balance, it seems that lamotrigine is a low-affinity substrate for P-glycoprotein (ABCB1) and BCRP (ABCG2); in the case of valproate, the published results are too conflicting to come to any conclusion (360). This question may be resolved by further RNA-seq studies, particularly those of single-cell and cross-species design that compare heterogeneity of drug effects and transporters in rodents and humans. A different factor that needs to be taken count of is binding of drugs to plasma proteins, as these are known to have different concentrations at different ages (101) and their binding properties might be different. It is the free fraction of a drug in plasma that determines how much transfers across the placental and brain barriers.

FIGURE 11.

Age comparison of drug entry into cerebrospinal fluid (CSF). CSF-to-plasma concentration ratios (%) of valproate (A) or lamotrigine (B) at embryonic day (E)19 and postnatal day (P)4 and in nonpregnant female adult rats collected 30 min after a single intraperitoneal injection of 100 mg/kg valproate or 20 mg/kg lamotrigine. Points are results from a single animal. Means ± SD; n = 3–7 individual animals. **P < 0.01, 1-way ANOVA with multiple comparisons. ns, Not significant. Data replotted from Ref. 360, with permission from F1000Research.

9. CONCLUDING STATEMENTS

In summarizing the most recent advances in choroid plexus biology in health and disease, we appreciate that we will have missed some exciting and relevant discoveries. Indeed, the numbers of manuscripts that will continue to be published given the explosion of new sequencing, imaging, and genetic manipulation tools will warrant continued monitoring of this research field. With the advent of such tools and their uptake by the field we envisage that many new questions will now be able to be addressed. Do choroid plexus epithelial cells exhibit “reactive” phenotype heterogeneity as do other immune and glial cell populations? Are dysfunctions in plexus function common across diseases or specific for each condition? As is now being discovered in other fields, here too we can refer back to the words of Weed (1): “…the study of structure must proceed hand in hand with the study of function.” We look forward to these discoveries and can only imagine that the combination of tried and true methods with the newest scientific advances will provide answers to how the choroid plexus and CSF are active contributors to disease but also solutions on how to more effectively treat these diseases.

GRANTS

We acknowledge the following support: National Institutes of Health (NIH) Grant RO1NS088566; OFD/BTREC/CTREC Faculty Development Fellowship Award (R.M.F.) and NIH R01 NS088566 and the New York Stem Cell Foundation (M.K.L.). M.K.L. is a New York Stem Cell Foundation–Robertson Investigator. Cass Foundation Grant 9309 to N.R.S. and K.M.D. enabled the start of some of the studies described in sect. 8.

DISCLOSURES

S.A.L. maintains a financial interest in AstronauTx, Ltd., and is on the Scientific Advisory Board of RM Global.

AUTHOR CONTRIBUTIONS

N.R.S., K.M.D., M.K.L., and S.A.L. conceived and designed research; R.M.F. and S.A.L. prepared figures; N.R.S., K.M.D., R.M.F., M.K.L., and S.A.L. drafted manuscript; N.R.S., K.M.D., R.M.F., M.K.L., and S.A.L. edited and revised manuscript; N.R.S., K.M.D., R.M.F., M.K.L., and S.A.L. approved final version of manuscript.