Automated Analysis of Cerebrospinal Fluid Flow and Motile Cilia Properties in The Central Canal of Zebrafish Embryos.

Circulation of cerebrospinal fluid (CSF) plays an important role during development. In zebrafish embryo, the flow of CSF has been found to be bidirectional in the central canal of the spinal cord. In order to compare conditions and genetic mutants across each other, we recently automated the quantification of the velocity profile of exogenous fluorescent particles in the CSF. We demonstrated that the beating of motile and tilted cilia localized on the ventral side of the central canal was sufficient to generate locally such bidirectionality. Our approach can easily be extended to characterize CSF flow in various genetic mutants. We provide here a detailed protocol and a user interface program to quantify CSF dynamics. In order to interpret potential changes in CSF flow profiles, we provide additional tools to measure the central canal diameter, characterize cilia dynamics and compare experimental data with our theoretical model in order to estimate the impact of cilia in generating a volume force in the central canal. Our approach can also be of use for measuring particle velocity in vivo and modeling flow in diverse biological solutions.

. One challenge now is to generalize this approach in order to compare a variety of genetic animal models and experimental conditions. This is of special interest for investigations on 2 www.bio-protocol.org/e3932 cilia-defective mutants in which motility defects are partial and the consequences on flow are not fully understood.
The goal of this protocol is to guide the computation of CSF flow profiles from fluorescence measurements. We developed a user-friendly interface program to generate flow profiles from collected data. We additionally provide a protocol to compare experimentally measured profiles of embryonic CSF flow to a theoretical profile. Our theoretical model relies on the assumption that the average flow rate is null. In this case, the volume force that gives rise to CSF flow can be computed and compared between different conditions. The volume force depends on different cilia parameters: = ℎ , where is the average cilia beating frequency, ℎ the width of the region occupied by the cilia, the viscosity, and a dimensionless parameter. We finally show how to quantify the main cilia frequency using transgenic embryos with cilia labeled by fluorescent proteins.    c. Pull microinjection needles from borosilicate glass capillaries with a 2-step needle puller.

Materials and Reagents
Adjust temperature and pulling force to produce a long and sharp funnel shape needle with an approximate tip diameter of 1-3 µm (equivalent to egg injection pipettes).  iii. Imaging could be performed with any imaging system as long as the signal to noise ratio (SNR) is high enough, and the imaging speed is above a few frames/s. Upright or inverted spinning disk can be chosen. Spinning disk imaging seems the most adequate but widefield microscope could be suitable as well with bright fluorescent beads. Confocal or two-photon microscopes could also be used, although the imaging speed might be limited with classical 6 www.bio-protocol.org/e3932 v. Because imaging was mostly performed in the sagittal plane at the embryonic and larval stage using visible laser for excitation, pigmentation was not an issue and did not therefore require the use of PTU.
a. As the central canal shape and cilia properties may differ along the rostrocaudal axis, we recommend imaging always at the same rostro-caudal position. In our case, we focused on 3 segments above the yolk extension ( Figure 1). Because spinning disk microscopes perform sharp optical sectioning, we advise using either the Differential Interference  c. Use the program cilia analysis to extract cilia beating frequency, length and angle. See Data analysis section for more details.

Data analysis
On top of the experimental procedures, we detail below two independent analysis workflows.
The first analysis (Section A) enables to obtain the CSF flow profile from the time series of bead trajectories acquired in Part Procedure-A. This also allows measuring the total CSF flow rate (Section B), which is expected to be null in WT embryos (Thouvenin et al., 2020). If adequate (see conditions 8 www.bio-protocol.org/e3932 below), the experimentally measured CSF flow can be fitted to a bidirectional flow model (Thouvenin et al., 2020) in order to extract the volume force generated by the motile cilia.
The second analysis workflow (Section C) uses the cilia beating movies (Procedure-B) to extract cilia parameters, including each cilia main beating frequency, length, and angle.
If appropriate, the last section (Section D) aims to combine the outputs from the two analysis workflows and extract a parameter we called α, an ad hoc coupling parameter that measures how multiple cilia efficiently work together to generate a flow.

A. CSF flow profile generation
Specifically for this protocol, we developed a user-interface platform to allow users to generate CSF flow profiles as easily as possible. Here, we present the analysis workflow ( Figure 2) and how to generate a first CSF flow profile from the fluorescent beads measurements. More subtle fine tuning of parameters is available within the user interface to adapt to variable imaging conditions, and is fully described in the document ManualGeneProfile.pdf file that can be found with the software.
As input, the analysis takes 2D time lapses of beads flowing in the central canal ( Figure 2A1). In order to generate kimographs, we swap for a given dorsoventral position the axes so that the X axis corresponds to the rostrocaudal position and the Y axis to time. Then, the beads trajectories appear as lines whose slopes reflect the direction and speed of the particles along the rostrocaudal axis. In order to build the flow profile, the program filters each kymograph and performs automatic segmentation of all lines in each kymograph ( Figure 2B2). It then extracts the slope of each line, and converts it into the particle velocity, in order to build a histogram of velocities for each dorsoventral position ( Figure 2B3). By calculating the average velocity at each position, we generate the CSF flow profile ( Figure 2C1). 9 www.bio-protocol.org/e3932  b. Alternatively, download and install the standalone application. Once it is installed, go to the command window and navigate to the installed folder. Run: application\GeneProfile. Figure 2A2 opens.

The user interface window in
3. Select .tif files to analyze. Multiple files can be selected at once, and they will be processed one by one. The volume force generated by motile cilia, 2) the pressure gradient that is established in the canal to oppose the cilia beating, 3) the width of the region bearing motile cilia, and 4) the diameter of the central canal. In order for the fit to be meaningful, the two assumptions of our bidirectional flow model should be respected: a cylindrical geometry for the central canal and the "no net flux" condition (Thouvenin et al., 2020). As a reminder, under these two assumptions, we showed that the averaged velocity profile can be fairly described by a piecewise second-order polynomial, defined as: Here is the diameter of the channel and the viscosity of the CSF, is the pressure gradient and is the average force per unit volume generated by the cilia. The latter two parameters can be measured experimentally from the CSF velocity profiles processed with the GeneProfile interface.
Our theoretical model describes a symmetric bidirectional flow for which the net total flow rate is null. In the user interface, before launching the fitting tool, we provide the user an estimate of the bidirectionality of the flow, called that is defined as: varies between 0% for a purely monodirectional flow and 100% for a purely bidirectional flow (the flow rate advected caudally equals the flow rate advected rostrally).
We advise users not to perform the fit of velocity profiles for values of < 70%, because under this arbitrary threshold, the "no flux" condition is no longer valid, and therefore the parameters of the fit are meaningless.
1. Once the flow profile is calculated (Data analysis Section A), the "Fit Model" button on the right ( Figure 2A2) turns green. The fit can be performed by pushing this button.
2. Two possibilities can arise: a. If < 70%, we estimate that the flow is not bidirectional enough to fit the velocity with our model, and display the warning message "We advise the user not to go further". If the two assumptions for our model are not respected, we advise to click on the "Stop here" button in order to stop the fitting process. If the measured flow profile was robustly measured, having a low β means that the flow could not be simply explained by the action of motile cilia in a closed cylindrical geometry and that another model should be developed by taking into account other physical effects.
b. If > 70%, the flow can be reasonably fitted with the simple model, and a verification imprecise.

C. Cilia frequency measurement
This section aims to describe the analysis protocol to estimate the main beating frequency of cilia (see Thouvenin et al., 2020), from the fluorescence cilia time lapse acquisitions described in the section Procedure B.
Similarly to section Data Analysis A, we describe here the principle of the analysis workflow ( Figure 3), as well as key instructions to perform a first analysis. Detailed instructions, as well as descriptions of fine-tuning parameters are available in an external document ManualCilia.pdf that can be found with the shared code.
The program first loads the imaging data with cilia dynamics versus time ( Figure 3A), and applies a local average filter (of size 4 by default) to increase the cilia SNR. For each pixel in the filtered data, the time series is extracted and Fourier transformed ( Figure 3B). The 5 maximal peaks of the Fourier spectrum are extracted, but, by default, only the first one is used. The frequency of the other peaks can be used for validation (e.g., if sampling errors are made, the sum of the frequencies of the first and second peak is equal to the acquisition frequency). A 2D image with the main frequency found at each pixel is thus created ( Figure 3C). In noisy regions, it outputs a random frequency, but in cilia regions it draws regions of interests of a given frequency that we considered to be single cilium. Each of these regions of interest containing more than 40 pixels (7.5 μm 2 ) is finally segmented and analyzed. The parameters frequency, diameter, eccentricity, area, angle, and major axis length are extracted and associated to their corresponding cilia parameters. 13 www.bio-protocol.org/e3932 If a comparison between dorsal and ventral cilia is of interest (Thouvenin et al., 2020), the program allows to manually draw a line at the center of the central canal and classify cilia as dorsal or ventral with respect to their relative position from the central line. This procedure is not described further here, but can be found in the document "ManualCilia.pdf" located in the same folder as the shared Matlab code.