Cilia-based flow network in the brain ventricles

See allHide authors and affiliations

Science  08 Jul 2016:
Vol. 353, Issue 6295, pp. 176-178
DOI: 10.1126/science.aae0450

Going with the flow

The interstitial spaces of the brain are filled with cerebrospinal fluid (CSF). Faubel et al. studied fluid transport in the third ventricle of the brain of mice, rats, and pigs. Sophisticated, state-of-the-art fluid dynamic studies revealed a complex pattern of cilia beating that leads to an intricate network of “highways” of CSF flow. This flow rapidly and efficiently transports and partitions CSF.

Science, this issue p. 176


Cerebrospinal fluid conveys many physiologically important signaling factors through the ventricular cavities of the brain. We investigated the transport of cerebrospinal fluid in the third ventricle of the mouse brain and discovered a highly organized pattern of cilia modules, which collectively give rise to a network of fluid flows that allows for precise transport within this ventricle. We also discovered a cilia-based switch that reliably and periodically alters the flow pattern so as to create a dynamic subdivision that may control substance distribution in the third ventricle. Complex flow patterns were also present in the third ventricles of rats and pigs. Our work suggests that ciliated epithelia can generate and maintain complex, spatiotemporally regulated flow networks.

The ventricular system of the brain consists of four interconnected, cerebrospinal fluid (CSF)–filled cavities, lined with ependyma whose apical surface bears bundles of motile cilia (1). CSF directional flow through the ventricles is driven by continuous CSF secretion into the ventricles and by orchestrated beating of the cilia. Motile cilia are eyelash-shaped cellular protrusions containing a microtubule-based axoneme that, by bending, confers motility. The direction of cilia beating depends on coupling of planar cell polarity signaling with hydrodynamic forces (2, 3). CSF flow toward the exit of each ventricle affords an efficient washout of waste (4) such as neurotoxins and protein aggregates. Additionally, CSF flow delivers nutrients, signaling molecules, microRNAs, and exosomes (512). How these CSF components reach their target sites is an important question that prompted us to investigate the CSF flow dynamics along the ventricular walls. We conducted this analysis in mice in the ventral part of the third ventricle (v3V; Fig. 1A) because of its simple geometry and close juxtaposition to potential targets, including the physiologically important periventricular hypothalamic nuclei that control circadian rhythms, hormone release, thermoregulation, blood pressure, satiety, and feeding (Fig. 1B).

Fig. 1 Anatomy of the third ventricle in the mouse and flow map of the v3V.

(A) Sagittal section through the third ventricle (blue), consisting of a dorsal (d3V) and a ventral (v3V) part connected by two ducts. As indicated by white arrows, CSF enters the v3V through the foramen interventriculare and exits through the aqueduct. (B) The position of periventricular hypothalamic nuclei [reconstructed from (22)]. Blue and red outlines differentiate the d3V and v3V, and hatching indicates the choroid plexus (blue) and median eminence (red). Medial preoptic nucleus, yellow; suprachiasmatic nucleus, gray; anterior hypothalamic area, brown; paraventricular nucleus, pink; ventromedial nucleus, green; arcuate nucleus, orange; dorsomedial nucleus, blue. (C) The flow map of the v3V, generated by particle tracking, shows that near-wall flow is subdivided into multiple flow domains. The associated eight major flow directions are indicated with turquoise arrows. (D) FITC-dextran, applied at the head of flow 6 (asterisk) from a femtopipette (p), follows the streamlines of the flow map shown in the first panel (t, time). Scales in (B) to (D) are Bregma levels in millimeters. a, anterior; ac, anterior commissure; aq, aqueduct; cx, cortex; d, dorsal; fiv, foramen interventriculare; mb, midbrain; my, myelencephalon; p, posterior; S, separatrix; th, thalamus; tz, tanycyte zone; v, ventral.

Organotypic v3V wall explant cultures were established, and cilia-generated flow was visualized by adding 1-μm fluorescent beads. Global flow was determined by particle tracking, and near-wall flows were extracted, yielding a highly complex flow map (Fig. 1C) that was consistent among animals (the number of animals used is given in table S1). Flow from the inflow duct fanned out after passing by the anterior commissure. A first flow was directed anteriorly toward the optic recess (Fig. 1C, flow 1). A second stream of the fan moved ventrally (flow 2). A third stream advanced posterodorsally (flow 3) and confronted, at Bregma level –0.9, an opposing flow (flow 4). When 70-kDa fluorescein isothiocyanate (FITC)–dextran was applied locally to the head of flow 6, it propagated as prescribed by streamlines of the flow map (Fig. 1D). Movie S1 shows the flow of FITC-dextran when applied to the heads of flows 1, 2, or 3. Injections to flows 1 and 2 resulted in an initially coherent flow that then fanned out, thereby dispersing the tracer over a sizable territory. Local application to the head of flow 3 revealed the ability of ciliated ependyma to produce a winding flow that replicates the shape of the streamlines seen in the flow map. Thus, the flow map may reflect the propagation of macromolecules in CSF.

To explore how the confrontation of flows 3 and 4 affects transport along the wall, beads were applied to v3V whole mounts (movie S2). Flow 4 carried beads anteriorly; near the site of confrontation with flow 3, beads moved into two mirror-symmetrical streams that intimately followed the ventricular wall and transported the beads down into the v3V. This flow dynamic created a separatrix that, together with flow 5, divided the v3V into two volumes. This flow-induced boundary hindered the passage of tracer. Flow 6 formed another intraventricular boundary that hindered flow 7 from drawing tracer away from flow 4. The flow pattern remained unchanged up to a distance of ~120 μm above the ependyma (fig. S1), and flows along the two walls were related by mirror symmetry (fig. S2). The velocity map showed that flow velocity at 21°C varied across the v3V wall within the range of 150 to 500 μm/s (fig. S3, A and B). Thus, it takes only a few seconds to move substances from the entrance of the v3V to its floor. These velocities are typical for cilia-driven flows (13).

We next analyzed cilia movement in the v3V and found that the flow map matched the cilia beating pattern. To show this, we captured movies of rectangular fields of beating cilia bundles across the entire v3V in a mosaic-like fashion. Using custom-made software, the cilia movement was extracted from each movie; results were then stitched together to generate a v3V-wide map of the time-averaged cilia beating pattern. At the separatrix, the flow pattern (Fig. 2A) corresponded to the beating orientation of underlying cilia (Fig. 2B and movie S4). At the entrance of the v3V, the tripartite flow also corresponded to the cilia beating orientation (fig. S4 and movie S5). Thus, cilia form distinct modules in which they beat coherently and, in this way, locally control the flow of the overlying fluid. As these two examples illustrate, the arrangement of cilia modules accounts for the complex flow map with its distinct flow domains.

Fig. 2 Cilia beating and fluid flow directions match.

(A) Flow map of and (B) time-averaged cilia beating orientation in the separatrix region [the flow map is color-coded as in Fig. 1C]. Computational analysis of cilia movement reveals obliquely opposing cilia orientation (movie S4) that corresponds to the pattern of the tracks shown in (A). (C) Flow map of two isolated fragments (I and II) of ependyma. (D and E) Time series show the flow of locally applied FITC-dextran across fragments I and II, which were recombined in vitro head to tail (D) or head to head (E). Red arrowheads flank the flow of FITC-dextran, turquoise arrows indicate cilia beating direction, and yellow dashes denote fragment borders. Scale bars, 30 μm [(A) and (B)] and 100 μm [(C) to (E)].

When the v3V wall was cut into pieces, the flow pattern in each fragment was virtually identical to that observed at the intact wall before cutting (fig. S5). Thus, cilia modules may determine flow independently of context. In keeping with this idea, we isolated two fragments (I and II) from the v3V wall (Fig. 2C) and recombined them at different angles. When the two fragments were combined head to tail, focally applied FITC-dextran translocated from one piece to the next (Fig. 2D). When the same two fragments were oriented with cilia beating in opposite directions, tracer applied to fragment I progressed toward the interfragment boundary, at which the flow was diverted upward by the counterflow originating from fragment II (Fig. 2E). Thus, transport along bent flows and formation of boundaries can be achieved by appropriate arrangement of cilia modules. Furthermore, cilia beating direction in each module is necessary and sufficient to sculpt the associated flow domains.

There are major physiological differences between an animal at rest (Zeitgeber time, ZT0 to ZT12) and an awake animal (ZT12 to ZT24) (14). The flow maps at ZT10 (Fig. 1C) and ZT23 (Fig. 3A) showed differences. At ZT23, we observed a prominent whirl ventrally of the separatrix. This whirl resulted from a circular arrangement of cilia beating (Fig. 3B). Movie S6 shows how this counterclockwise whirl relates to the separatrix. Recording the same region at ZT10 revealed a substantially different pattern devoid of a whirl (movie S7). At ZT10, separatrix and flow 5 formed a vertical intraventricular boundary. When a low dose of FITC-dextran was injected into flow 4, the tracer flowed ventrally, along the boundary, to reach the floor (white dashes in Fig. 3C, movie S8). This boundary withstood a high FITC-dextran dose (blue fields in Fig. 3C, movie S9). When low or high doses of FITC-dextran were injected into flows 3 or 4 at ZT23, tracer remained in the dorsal part of the v3V (Fig. 3D and movies S10 and S11), indicating the presence of a dorsoventral boundary in this region.

Fig. 3 Temporal changes in flow.

(A) Flow map of the v3V at ZT23. A whirl is circumscribed by the ellipse. The scale shows Bregma levels in millimeters. Solid and dashed arrows indicate the direction of flow. (B) Time-averaged cilia beating orientation in the whirl, showing a counterclockwise orientation (movie S6). Scale bar, 50 μm. (C and D) Dynamics of FITC-dextran propagation at ZT10 (C) and ZT23 (D), as shown in movies S8 to S11, superimposed on the outline of the v3V. Injection points are marked by red asterisks. Near-wall flow of low-dose injections is shown with white dashes. Nested color fields trace the dye propagation for high-dose injections from a tip placed into the field of most intense color. Gray arrows represent flow directions taken from Figs. 1C (C) and 3A (D). A black dashed line represents the separatrix. The red circle in (D) represents a reference mark from movie S10.

The appearance of a whirl at ZT23 leads to a major change in the global flow pattern: The vertical intraventricular boundary that was present at the end of the resting phase was absent at the end of the awake phase. This raises the possibility that the distribution of CSF varies with the time of the day. Our FITC-dextran tracer experiments show that mixing of the flow is very weak. This is also supported by comparing the time scale of mixing, estimated from the flow gradient, with the time that it takes the flow to transverse the ventricle.

Mammalian third ventricles differ in size and morphology (15), yet both the modular organization and the flow domains are largely conserved between mice and rats (compare Fig. 1C with Fig. 4A). The ventricular recesses in the rat are more prominent than in the mouse. Accordingly, when we examined cilia-generated flow flow in the v3V of rats, flow 1, directed at the supraoptic recess, was S-shaped, and a new flow (flow 9) emerged from the posterior floor of the v3V, to which the infundibular and inframammillary recesses are connected. The v3V of pigs differs in size and architecture from those of rodents (16), yet we also observed flow domains in pig v3V (Fig. 4B).

Fig. 4 Flow maps of rat and pig v3Vs.

(A) The flow map of the rat v3V strongly resembles that of the mouse v3V, except for an anteroventral whirl (ellipse), an S-shaped flow (1), and a posterior flow (9) that emerges from the floor of the posterior part of the v3V. (B) The flow map of the pig v3V shows that it is highly organized and has cilia modules, separatrices, and whirls. Scale bars, 1 mm. Color coding is as in the previous figures.

A combination of straight and bent flows guides molecules across several hundred microns. For example, flow initiating at the entrance of the v3V extends as far as the arcuate nucleus. This flow may play a role in delivery or uptake of substances in this region, which is characterized by an abundance of chemosensory tanycytes (17). Tanycyte cell bodies are in contact with CSF, and their processes extend several hundred micrometers deep into the brain parenchyma, thus acting as bridges between the ventricle and neurons (18).

The similarity in design of the transport networks across species suggests a genetic regulation of network development. Hypothalamic nuclei arise from radial glia cells (19) that acquire their fate by coexpressing a combination of transcription factors (20). The later descendants of these radial glia cells also form the ependymal cell sheet (21), and thus cilia modules therein might also be specified by a particular combination of regionally expressed transcription factors. Such orchestrated development would establish a direct relationship between potential targets, the hypothalamic nuclei, and the structure of the transport network.

Fluid dynamics has long described complex types of movement in other systems. Now we show that complex fluid movements are also present in the brain in the form of a transport network driven by coordinated cilia beating patterns. In addition to targeted substance delivery, cilia generate separating flows and whirls that may establish and modulate intraventricular boundaries capable of concentrating locally released compounds and preventing their entry into off-target regions. Transient local changes in the beating pattern evoked a major change in ventricular subdivision. The cellular and structural bases that underly such changes in beating direction are unknown but may be driven by transient changes in cell-cell interactions and in planar cell polarity.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Table S1

References (2325)

Movies S1 to S11

References and Notes

  1. Acknowledgments: We thank A. Bae, A. Pumir, and C. Lo for valuable comments and suggestions; C. Thaller for technical assistance; J. Schröder-Schetelig for providing multirecorder software; and D. Hornung and T. Baig for providing euthanized minipigs. This work was supported by the Max Planck Society.
View Abstract

Stay Connected to Science

Navigate This Article