Background

Cerebrospinal fluid (CSF) is contained in the brain ventricles (chambers derived from the brain lumen) and contacts neuroepithelial progenitor cells during brain development and in adulthood. The CSF contains a unique composition of proteins (CSF proteome) [1–6], small molecules [5, 7] and lipid particles [8] for which limited studies have demonstrated function during brain development [1–5, 7–9]. Since the brain lumen forms before the choroid plexus (CP), ventricles exist with no requirement for a CP to be associated with each ventricle. In some cases early pre-CP fluid is called ependymal fluid [10]. The cohort of proteins in zebrafish CSF is largely overlapping with early amniote CSF prior to CP development [7] and this conservation has not been investigated at later stages. Once the CPs have become functional they secrete most of the ventricular fluid, although there is clear data that fluid continues to be produced from ependymal cells even after CP differentiation is complete [11–14]. The ventricular fluid is thus always a mix from different sources and the term CSF is most general and appropriate to refer to the ventricular fluid at all stages.

Restricted or directional CSF movement, either during production and drainage or coupled to heartbeat [15–19], could limit locally-produced factors to specific regions of the developing brain. Studies in Xenopus have investigated local intraventricular CSF dynamics in the developing brain [19, 20], but there has been no direct study of in vivo CSF movement dynamics between ventricles. Recent studies have suggested that cilia on neuroepithelial cells may detect a mechanical signal to regulate neuronal progenitor differentiation [21]. The nature of this signal is unknown but could be CSF movement. CSF fluid dynamics have been examined in zebrafish after CP formation [22], but have not been studied earlier in development during the major period of developmental neurogenesis. To address these deficits, we investigate CSF movement in the accessible larval zebrafish brain, prior to CP formation. Our data demonstrate directional movement of CSF, which may function during brain development.

Methods

Fish lines and maintenance

Danio rerio fish were raised and bred according to standard methods [23]. Embryos were kept at 28.5 °C and staged according to Kimmel et al. [24]. Times of development are expressed as hours post-fertilization (hpf) or days post-fertilization (dpf). Analyses were performed at early larva (pharyngula stage; 27–30 hpf) stage, late larva (hatching period; 51–54 hpf) stage, and some at 5 dpf. Lines used were: AB; Tg(βact:Arl13b–GFP) transgenic line [25]; troponin T type 2a (cardiac) mutants tnnt2a −/− (R14GBT0031/silent heart; obtained from The Zebrafish International Resource Center (ZIRC; Eugene, USA)) [26]. Larvae were incubated in 3 μg/mL 1-phenyl-2-thiourea (PTU) starting between 22 and 24 hpf to inhibit maturation of melanocytes and allow for clear visualization. All experimental procedures on live animals and embryos were reviewed and approved by the Institutional Animal Care and Use Committee of the Massachusetts Institute of Technology (Protocol# 0414-026-17) and were carried out in accordance with the recommendations in the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals.

Multiview SPIM brain ventricle imaging

Larvae were anesthetized in 0.1 mg/mL (0.38 mM) Tricaine (Sigma, St. Louis, USA) dissolved in embryo medium [23] prior to injection and imaging. At high doses, Tricaine can induce death or stop the heart [27–29], but these effects have not been shown at the concentration used here. Direct effects of Tricaine on CSF movement have not been investigated.

The rhombencephalic ventricle was microinjected with 2 nL of 2000 kDa dextran conjugated to fluorescein (ThermoFisher/Invitrogen Molecular Probes, Waltham, USA; D7137) diluted to 25 mg/mL in sterile ddH2O. Embryos were embedded in 1 % low-melting point agarose in a glass capillary, and selective plane illumination microscopy (SPIM) was performed using a Zeiss LightsheetZ.1 microscope within 60 min of injection. Six multi-view images were acquired (60° apart), and were processed using ZEN software (Zeiss, Oberkochen, Germany), FIJI [30, 31], and the Multiview reconstruction plugin [32, 33]. Images were analyzed using Imaris software (Bitplane, Belfast, UK) to visualize 3-D images and to generate volume measurements for ventricles at each age. The image threshold was set so that both the ventricle outline and the spinal canal were detected. For rhombencephalic ventricular volume measurements, the same threshold as above was used, but only the rhombencephalon was included in the region of interest. Injection does not affect development, survival or ventricle size over the time period studied ([34] and Additional file 1: Fig. S1).

Tissue histology and hematoxylin and eosin staining

Larvae were anesthetized as described above and fixed overnight in aqueous Bouin’s fixative (RICCA Chemical, Pokomoke City, USA). Larvae were washed in 70 % EtOH until supernatant was clear (at least six washes of 1 h each at room temperature). Larvae were oriented in Richard Allan Scientific HistoGel (ThermoFisher, Waltham, USA). The HistoGel discs containing oriented larvae were fixed for 1 h in Bouin’s fixative, and subsequently washed in 70 % EtOH until supernatant was clear. Samples were infiltrated with paraffin using a Tissue-Tek VIP5 Vacuum Infiltration Processor (Sakura Finetek, Torrance, USA; model 5215). Sections were cut at 4–5 μm and slides were stained with Harris acidified hematoxylin (3 min) and eosin Y (10 s) using the “standard” H&E protocol [35] on a Shandon Varistain Gemini Slide Stainer (ThermoFisher, Waltham, USA). Slides were imaged using an Axioplan 2 (Zeiss, Oberkochen, Germany) equipped with a Q-Capture MicroPublisher 5.0 RTV camera/software (QImaging Scientific, Surrey, Canada) using 10× and 40× air objectives and 63× and 100× oil-immersion objectives.

Live time-lapse confocal imaging

Larvae were anesthetized as described above and the rhombencephalic ventricle was microinjected with 2 nL of Kaede, a photoactivatable protein with fluorescence that changes from green to red upon illumination with UV light [36, 37]. After injection, larvae recovered from injection and anesthesia for at least 30 min prior to imaging, mitigating any immediate effects of dye injection. Embryos were then anesthetized as described above and placed in wells of 1 % agarose (early larva), or embedded in 1 % low-melting point agarose on a coverslip (late larva) for confocal microscopy [37]. Confocal imaging and photoconversion was performed using an inverted Zeiss LSM710 laser-scanning microscope equipped with a 25× (0.8 NA) LD LCI Plan Apochromat water-immersion objective (Fig. 1a). Kaede photoconversion was performed on a circle (diameter of 8 pixels) in the center of the ventricle of interest using the 405 nm laser at 85–95 % strength (for 700 iterations at scan speed = 7, ~20 s per stack). Stacks were collected using simultaneous activation of green and red channels and continuous acquisition for 10–15 min (Fig. 1a).

Fig. 1
figure 1

Experimental workflow. a Schematic representation of rhombencephalic ventricle Kaede injection and confocal photoconversion and imaging of zebrafish ventricles. b Workflow for image processing. Raw images were collected and then thresholded in Imaris (Bitplane) software so that red (photoconverted) Kaede was only detected after photoconversion and was not detected in the unconverted target ventricle immediately after photoconversion. Then the volume of the unconverted target ventricle was calculated over time by Imaris (Bitplane) software. Dashed line indicates telencephalic-to-mesencephalic/diencephalic boundary. c Workflow for data analysis. From volume measurements of the target ventricle, volumetric flow rate (Q) was calculated using Eq. (1) (see “Methods” section). Q was smoothed using a 9-point moving-window average. Q was then converted to linear velocity by dividing by the cross-sectional area of the smallest part of the aqueduct between the two ventricles. These CSF dynamics were reported as the maximum velocity between the two ventricles (vmax). T telencephalic ventricle, D/M diencephalic/mesencephalic ventricle, R rhombencephalic ventricle, hpf hours post-fertilization, V volume, Q volumetric flow rate, v velocity. Scale bar: 50 μm

Full size image

Image processing and fluid movement quantification

Confocal time-lapse images were analyzed using Imaris software (Bitplane, Belfast, UK) to create thresholded volumetric surfaces of photoconverted Kaede. The red threshold was set for each image such that no signal was detected before photoconversion, and no signal was detected in the unconverted target ventricle at time = 0 (first scan after photoconversion) (Fig. 1b). Imaris generated volume (V) data from each time point. From these volume readings, the volumetric flow rate at time t (Qt) was calculated for each time point using Eq. (1):

$$Q_{t} = \frac{{V_{t} - V_{t - 1} }}{{T_{t} - T_{t - 1} }} = \frac{{\Delta {\text{V}}}}{\Delta T}$$
(1)

Q was then smoothed to negate technical noise using a 9-point moving window average. To control for individual variations in ventricle shape, Qt was converted to linear velocity (v) by dividing Q by the cross-sectional area of the aqueduct. Cross-sectional areas were measured using Imaris software (Bitplane, Belfast, UK) by creating surfaces of the unconverted (green) Kaede signal in the first scan (before photoconversion) at the narrowest point between the two ventricles being analyzed. The maximum velocity number (vmax) was used to analyze brain ventricle CSF movement (see Fig. 1c).

Acute inhibition of heartbeat

Heartbeat was acutely inhibited by soaking ventricle-injected embryos in 40 mM 2,3 butanedione monoxime (BDM; Sigma, St. Louis, USA) solubilized in embryo medium [23] for 10–30 min prior to imaging as previously shown [38]. This treatment stops heartbeat, but does not change the gross ventricular anatomy within 2 h of incubation (Additional file 1: Fig. S2). BDM prevents action of all muscles (and can alter activities of connexins [39], potassium channels [40, 41], and L-type calcium channels [42]) and is therefore less precise for disruption of heartbeat than the tnnt2a −/− mutant.

Cilia imaging

The Tg(βact:Arl13b–GFP) transgenic line was used for visualization of cilia. Embryos were treated with 40 mM BDM, as described above for 10 min to stop heartbeat and enable completely steady, immotile embryos suitable for high-speed imaging. They were prepared for SPIM as described above and incubated in 20 mM BDM and 0.1 mg/mL Tricaine (Sigma, St. Louis, USA) in embryo medium during imaging. Images were collected at 57.22 frames per second (fps) in the telencephalon, diencephalon/mesencephalon, rhombencephalon and spinal canal. Time-lapse movies were registered using FIJI [30, 31] and the StackReg plugin [43]. The ciliary beat frequency (CBF) was calculated using the following formula: [CBF = (number of frames per second)/(average number of frames for single beat)] [44]. Images were processed by overlaying sequential frames in pseudocolor using Photoshop (Adobe).

Results

The early zebrafish brain ventricular system is complex and dynamic

In order to obtain detailed structure of the zebrafish brain ventricular system, we filled the ventricles with 2000 kDa dextran conjugated to fluorescein and used SPIM to acquire images at 30 hpf (pharyngula or early larval stage; Fig. 2a, b) and 54 hpf (hatching period or late larval stage; Fig. 2d, e), prior to CP maturation but after ventricle inflation and spinal canal (SC) formation. We also examined ventricle anatomy at 5 dpf (Fig. 2g, h) after CP formation. Since the brain is still developing, the early ventricle anlagen are dynamic over these times. The telencephalic (T) ventricle is the most rostral. Early, the combined diencephalic/mesencephalic (D/M) superventricle is caudal to the T ventricle while later this middle ventricle subdivides into the diencephalic (D) and the tectal (Te) ventricles. The rhombencephalic (R) ventricle is at the most caudal aspect of the brain [22, 45] (Fig. 2). At these stages, the SC branches off from the ventral aspect of the D ventricle (Fig. 2b′, c–c′′, e, f–f′). This is the first demonstration of early connection from the diencephalon to the spinal canal in any species.

Fig. 2
figure 2

Zebrafish brain ventricle anatomy at early and late larval stages. Using 2 nL rhomencephalic ventricular injections of fluorescein-labeled dextran and imaging by Selective Plane Illumination Microscopy (SPIM, or Lightsheet Microscopy), ventricular system anatomy, volume, and connectivity was visualized in three dimensions. a Dorsal, b lateral view, 30 hpf zebrafish larvae with neuroepithelium labeled in red (mApple-caax) and the ventricular system labeled in green (fluorescein dextran injection). c H&E stain of 4 μm thick parasagittal section of 30 hpf zebrafish. aʹ Dorsal, bʹ Lateral view, 30 hpf larval zebrafish ventricular system. bʹ inset Magnification of boxed region in bʹ to show spinal canal (S) branching from the ventral aspect of the D/M ventricle (arrow). cʹ H&E stain of 4 μm thick section of 30 hpf zebrafish. cʹʹ magnification of boxed region in cʹ to show spinal canal (S) branching from D/M ventricle. d Dorsal, e Lateral view, 54 hpf larval zebrafish ventricular system. e inset Magnification of boxed region in d to show SC branching from the ventral aspect of the diencephalon (arrow). f H&E stain of 4 μm thick section of 54 hpf zebrafish. fʹ magnification of boxed region in f to show spinal canal (S) branching from diencephalon. g Dorsal, h Lateral view of the 5 dpf larval zebrafish ventricular system. h insets Magnification of boxed regions in h to show SC branching from the ventral aspect of the diencephalon (arrow) and rhombencephalon (arrow). i H&E stain of 4 μm thick section of 5 dpf zebrafish. iʹ magnification of boxed region in i to show spinal canal (S) branching from rhombencephalon. T telencephalic ventricle, D/M diencephalic/mesencephalic ventricle, D diencephalic ventricle, Te tectal ventricle, R rhombencephalic ventricle, S spinal canal, hpf hours post-fertilization, dpf days post-fertilization. Scale bar: bʹ, e, h insets, cʹʹ, fʹ, iʹ: 50 μm; all others: 100 μm

Full size image

Ventricular shape changes over the first 2 days of development; with the combined diencephalic/tectal (D/Te) ventricle developing a more complex shape as development proceeds allowing the two regions to be discernible (Fig. 2d, e). The ventricular volume increases over development and the total volume of the ventricular system is ~4 nL at early larval stage, ~5.5 nL at late larval stage and ~6.5 nL at 5 dpf. The rhombencephalic ventricular volume increases from ~1.5 to ~3.5 nL from 30 to 54 hpf, but stabilizes at ~3.5 nL from 2 to 5 dpf. At the early larval stage, the isthmic (D/M-to-R) canal is significantly narrower than the anterior (T-to-D/M) canal (cross-sectional area: anterior: 1667 ± 131.3 μm2; n = 18 vs. isthmic: 1159 ± 61.9 μm2; n = 17; p = 0.0016). At the late larval stage, the isthmic (D-to-R) and anterior (T-to-D) canals are smaller than at early larval stage, but are equal to one another (cross-sectional area: anterior: 665.3 ± 118.0 μm2; n = 17 vs. isthmic: 716.2 ± 129.5 μm2; n = 21; p = 0.78). By 5 dpf, the adult connection from the R ventricle to the SC is forming (Fig. 2g–iʹ). These data add detail to the anatomical understanding of the developing ventricular system and ventricular dimensions may also play a role in influencing CSF movement.

CSF moves from anterior to posterior in early larval zebrafish brain ventricles, but partially reverses direction at late larval stages

To quantify CSF movement during the major period of developmental neurogenesis, we filled the ventricles with photoactivatable Kaede protein. Kaede was injected into zebrafish brain ventricles and activated by a local UV pulse at 405 nm in the T, D (D/M) or R ventricle at early or late larval stages. CSF movement was subsequently determined and vmax calculated (see “Methods” section, Fig. 1). At 27–30 hpf (early larva), the average vmax for CSF movement from T to D/M ventricle is significantly greater than for CSF movement in the reverse direction, from D/M to T ventricle (Fig. 3a–c). Preferential anterior-to-posterior movement is also observed in the D/M to R ventricle direction in early larvae (Fig. 3d–f). However, at late larval stages (51–54 hpf) the average vmax for CSF movement from T to D/Te ventricle is less than that of CSF movement in the D to T ventricular direction (Fig. 3g–i), indicating a reversed direction of movement between these two ventricles comparing the stages examined. The anterior-to-posterior directionality is still observed between D and R ventricles at these later stages (Fig. 3j–l).

Fig. 3
figure 3

Bulk movement of CSF at early and late larval stages. a, d, g, j Schematics of brain ventricular regions analyzed. b At 27–30 hpf, the average vmax is significantly higher in the anterior to posterior (red) direction through the telencephalic-to-diencephalic/mesencephalic aqueduct than in the posterior to anterior direction (blue). c Overlaid thresholded volumes generated by Imaris (Bitplane) software of a representative 27–30 hpf T → D/M dataset. e At 27–30 hpf, the average vmax is significantly higher in the anterior to posterior (red) direction through the diencephalic/mesencephalic-to-rhombencephalic aqueduct than in the posterior to anterior direction (blue). f Overlaid thresholded volumes generated by Imaris (Bitplane) software of a representative 27–30 hpf D/M → R dataset. h At 51–54 hpf, through the telencephalic-to-diencephalic aqueduct, directionality is reversed such that the vmax is significantly higher in the posterior to anterior (blue) direction through the telencephalic-to-diencephalic aqueduct than in the anterior to posterior direction (red). i Overlaid thresholded volumes generated by Imaris (Bitplane) software of a representative 51–54 hpf T → D dataset. k At 51–54 hpf, the average vmax is significantly higher in the anterior to posterior (red) direction through the diencephalic-to-rhombencephalic aqueduct than in the posterior to anterior direction (blue). l Overlaid thresholded volumes generated by Imaris (Bitplane) software of a representative 51–54 hpf D → R dataset. Red circles indicate photoconversion region. Lines represent average vmax and error bars denote SEM. p value calculated using unpaired Student’s t test, **p < 0.001; ***p < 0.0005; ****p < 0.0001. T telencephalic ventricle, D/M diencephalic/mesencephalic ventricle, Te tectal ventricle, R rhombencephalic ventricle, hpf hours post-fertilization. Scale bar: 50 μm

Full size image

CSF directional movement is dependent on heartbeat at early larval stages

Adult CSF pulses and mixes during the cardiac cycle and with breathing [15–18], zebrafish ventricle expansion and early Xenopus CSF movement is dependent on heartbeat [19, 34]. We therefore investigated whether directionality of CSF movement is dependent on heartbeat at early larval stages (27–30 hpf). Heartbeat was prevented genetically using the tnnt2a −/− fish that never develop a heartbeat. In tnnt2a −/− fish, directionality was maintained (albeit with a less significant difference), but vmax from D/M → T ventricle in tnnt2a −/− animals is significantly higher than in WT and is unchanged from T → D/M ventricle (Fig. 4a). However, loss of tnnt2a function eliminated directional movement between the D/M and R ventricles and vmax was significantly lower from D/M → R ventricle than in WT, but unchanged from R → D/M (Fig. 4b). To acutely disrupt heartbeat, we treated WT fish with 40 mM 2,3 butanedione monoxime (BDM). In BDM-treated animals vmax directionality was eliminated in both T → D/M and D/M → R ventricular directions (Additional file 1: Fig. S3). BDM has a wide variety of targets that are not associated with heartbeat disruption (see “Methods” and “Discussion” sections), which might account for differences observed in CSF movement between BDM-treated and tnnt2a −/− animals. These data suggest that the heartbeat is partially required to maintain directional CSF movement, specifically between D/M and R. In tnnt2a −/− fish, the ventricles develop normally until 27 hpf but do not fully inflate at 36 hpf [34], suggesting brain abnormalities that could be attributed to disruption in CSF movement.

Fig. 4
figure 4

Directionality of CSF movement is partially dependent on heartbeat at early larval stage. a Average vmax for movement between the telencephalon and diencephalon/mesencephalon at 27–30 hpf in wild type (WT; red) and tnnt2a −/− (pink) from the telencephalic-to-diencephalic/mesencephalic aqueduct or the posterior to anterior direction [blue (WT), light blue (tnnt2a −/−)]. b In tnnt2a −/− mutant fish there is no directionality of CSF movement observed through the diencephalic/mesencephalic -to-rhombencephalic aqueduct. WT data are from Fig. 3. Horizontal lines represent average vmax and error bars denote SEM. p value calculated using unpaired Student’s t test, **p < 0.001; ****p < 0.0001; n.s. not significant, T telencephalic ventricle, D/M diencephalic/mesencephalic ventricle, R rhombencephalic ventricle, hpf hours post-fertilization

Full size image

Brain ventricle cilia are non-motile at 27 hpf and cannot contribute to directional CSF movement

It has been suggested that ciliary beating in the neuroepithelium bordering the brain ventricular lumen drives CSF movement [46]. Monociliated cells are present in the zebrafish neuroepithelium at the stages examined [25], so we investigated whether these cilia beat in a concerted fashion. Early central nervous system cilia in the zebrafish spinal canal beat at ~12 Hz [38], but motility of the cilia in the early zebrafish brain has not been characterized. Using the Tg(βact:Arl13b–GFP) transgenic line (Fig. 5a, c) to visualize CNS cilia, we confirmed that a subset of spinal canal cilia beat at early (27 hpf) and late (51 hpf) larval stages (Fig. 5; Additional file 2: Movie S1, Additional file 3: Movie S2). At the early larval stage, cilia in the neuroepithelium do not beat (Fig. 5b, e; Additional file 2: Movie S1). However in late larvae, motile brain cilia are observed in telencephalon and are especially active in the diencephalon (Fig. 5d, f; Additional file 3: Movie S2). We conclude that at early larval stages it is unlikely that the immotile cilia contribute to CSF movement.

Fig. 5
figure 5

Brain cilia are non-motile at early larval stage; some are motile at late larval stage. a, c Arl13b: GFP is expressed in cilia throughout the developing fish including the neuroepithelium and spinal canal. b, d Quantification of ciliary beat frequency (Hz) for motile cilia in each region at early (27 hpf) and late (51 hpf) larval stages, if any. e, f Visualization of ciliary movement for cilia in each region at 27 and 51 hpf, if any. Each of the three panels is an overlay of two sequential image acquisition frames (images were taken in the same region at distinct times at 57.22 fps, so each overlay represents a time difference of ~0.017 s). See Additional file 2: Movie 1 and Additional file 3: Movie 2 for full dataset. Colors were selected such that unmoved pixels appear black (Frame 1: green + magenta; Frame 2: red + cyan). Horizontal lines represent average vmax and error bars denote SEM. T telencephalic ventricle, D/M diencephalic/mesencephalic ventricle, R rhombencephalic ventricle, SC spinal canal ventricle analysis region, hpf hours post-fertilization, N.A. quantification of cilia frequency is not applicable because no cilia are motile. Scale bars: a, c: 100 μm; e, f: 5 μm

Full size image

Discussion

In this study, we expanded understanding of CSF dynamics through quantification of CSF movement. Using the zebrafish model, we focused on the major period of developmental neurogenesis at two stages: early larva [pharyngula stage; 27–30 h post-fertilization (hpf)] and late larva (hatching period; 51–54 hpf). Our study leads to four conclusions: (1) CSF moves preferentially anterior to posterior from the diencephalic (D) (or combined diencephalic/mesencephalic superventricle (D/M)) to rhombencephalic (R) ventricles. (2) Direction of CSF movement between the telencephalic (T) and diencephalic (D) ventricles reverses from early to late larval stages. (3) Directional CSF movement is partially dependent on heartbeat. (4) At early larval stages no motile cilia are present in the brain and therefore are not likely to direct the CSF movement observed. Together these data indicate that locally restricted CSF factors or physical signals associated with fluid movement may contribute to specialization of the early brain.

The primarily unidirectional CSF movement observed between D (or D/M) and R ventricles was associated with a large vmax from D (D/M) to R ventricles. Reverse movement was present, but with about 25 % of the observed forward vmax. This raises two questions: (1) where does the excess volume of CSF entering the R ventricle go and (2) what mechanisms underlie directional CSF movement? First, with respect to CSF “destination”, the brain-CSF barrier is quite tight at early developmental stages [47], however small molecules (<70 kDa) can pass between the ventricles and the neuroepithelium [36]. Water is also likely selectively transported since aquaporins 1a.1, 3a, 7 are expressed by the neuroepithelium [48]. Therefore some CSF components could pass in or out of the ventricular system. Additionally, we note that R ventricle volume more than doubles over this timeframe of development, perhaps due to CSF influx from more anterior ventricles as movement proceeds from D → R ventricles. Second, with respect to potential mechanisms of directional CSF movement at the stages examined, the choroid plexuses (CP) are not yet mature and there is unlikely to be active secretion from CP precursors. It is not known at what stage the CPs become mature and secrete CSF in zebrafish. Ventricles inflate at approximately 20 hpf [34] with no evidence of CPs. Henson et al. [49] claim that the zebrafish CP are mostly mature by 3 dpf, but Garcia-Lecea et al. [50] report more complete maturation and coalescing of the CP in the R ventricle by 5 dpf. However, even without secretory CP, different regions of the neuroepithelium could produce CSF at different rates to drive directional movement. Alternately, osmolarity differences between the ventricles could drive directional CSF movement, a mechanism that contributes to diffusion gradients in adult brain [51].

Directional CSF movement from D (D/M) to R ventricles could result in factors from the D (D/M) ventricle entering the R ventricle, but factors unique to the R ventricle may remain largely distinct from those of D (D/M). The stages examined encompass a major period of neurogenesis, when the detailed patterns of the diencephalon, mesencephalon, and rhombencephalon begin to read out as distinct neuronal and glial differentiation programs. Thus differentially localized CSF factors could modulate differentiation of the optic tectum (which is derived from the mesencephalic epithelium) or the cerebellum and medulla (which is derived from the rhombencephalic epithelium). At the stages examined, the spinal canal connects to the D (D/M) ventricle. We observed no converted Kaede protein in the spinal canal (data not shown), indicating that the vast majority of D (D/M) CSF movement is into the R ventricle, consistent with the smaller diameter of the spinal canal connector relative to the isthmic canal.

A surprising contrast to the consistent direction of movement between D (D/M) and R ventricles was the reversal of direction between T and D (D/M) ventricles when early and late larval stages were examined. What would lead to this reversal? One possibility is that at late larval stages, CP cells present along the D ventricle, which arise later but complete differentiation first, have begun to coalesce and are producing CSF, while the CP cells in the R and T ventricles have not yet completed differentiation [50, 52]. This could lead to greater pressure from in the D ventricle from more mature CP cells and promote CSF entry into the T ventricle, where the CP is the last to form [31]. Therefore, posterior-to-anterior CSF movement between D and T ventricles could result from active CSF secretion and greater D ventricular CSF pressure. The strong ciliary movement in D neuroepithelium relative to T neuroepithelium at late larva stage may also contribute to reversal of flow. Another possibility is that the reverse flow could result from gradually increasing pressure in the D ventricle as result of ventricular shape change and the subsequent reduction in D ventricular volume as development proceeds. Whatever the mechanism, the outcome of this change in direction would be that at early larval stages CSF in the D ventricle would contain factors derived from the T, whereas at late larval stages CSF in the T ventricle would contain some components derived from the D. Thus distinct cohorts of factors in T or D ventricles could be present at each stage.

Heartbeat plays a role in CSF dynamics early in brain development, as has also been shown in Xenopus [19]. After tnnt2a loss of function, a mechanism independent of heartbeat, perhaps CSF production, maintains directional CSF movement from T and D/M ventricles, but not for D/M to R ventricular movement, although there is a trend toward less significant directionality. Consistently, acute loss of heartbeat with BDM treatment disrupts all directional CSF movement. However, BDM prevents action of all muscles (and can alter activities of connexins [39], potassium channels [40, 41] and l-type calcium channels [42]) and some of these side effects may contribute to the phenotypes observed.

At early larval stages no motile cilia are present in the neuroepithelium and are therefore unlikely to contribute to the CSF movement observed. The non-motile cilia at this time may be able to detect CSF movement, potentially to activate planar cell polarity signaling described by Ohata and colleagues [21], or to guide migrating neurons or extending axons [54]. At late larval stages, cilia along the D ventricle are highly motile and may indicate differentiating CP, as CP cells have motile cilia [52]. As discussed above, these ciliary movements or CSF production may drive CSF movement from D to T ventricles. Mutants that disrupt the primary cilia present with hydrocephalus [53], which would preclude clear comparison of CSF movement. However, future studies acutely disrupting cilia are needed to fully address potential roles of cilia in CSF movement.

In summary, this study demonstrates directional CSF movement during early brain development. One outcome of this may be differential concentrations of CSF factors in different ventricles, which may contribute to regional specialization of the brain. Directional CSF movement may also act as a physical signal to modulate neuroepithelial fates. The zebrafish offers an excellent system with which to address these intriguing possibilities.

Conclusion

Embryonic CSF in the zebrafish model moves directionally during the major period of developmental neurogenesis. Specifically, CSF moves preferentially from the diencephalic into the rhombencephalic ventricle. In addition, direction of CSF movement between telencephalic and diencephalic ventricles reverses from early [pharyngula stage; 27–30 h post-fertilization (hpf)] to late (hatching period; 51–54 hpf) larval stages. CSF movement is partially dependent on heartbeat. At early larval stages, the absence of motile cilia indicates that cilia likely do not direct CSF movement. These data suggest that CSF components may be compartmentalized and could contribute to specialization of the early brain and CSF movement may also provide directional mechanical cues.