Multi-sample SPIM image acquisition, processing and analysis of vascular growth in zebrafish

ABSTRACT To quantitatively understand biological processes that occur over many hours or days, it is desirable to image multiple samples simultaneously, and automatically process and analyse the resulting datasets. Here, we present a complete multi-sample preparation, imaging, processing and analysis workflow to determine the development of the vascular volume in zebrafish. Up to five live embryos were mounted and imaged simultaneously over several days using selective plane illumination microscopy (SPIM). The resulting large imagery dataset of several terabytes was processed in an automated manner on a high-performance computer cluster and segmented using a novel segmentation approach that uses images of red blood cells as training data. This analysis yielded a precise quantification of growth characteristics of the whole vascular network, head vasculature and tail vasculature over development. Our multi-sample platform demonstrates effective upgrades to conventional single-sample imaging platforms and paves the way for diverse quantitative long-term imaging studies.

the embryos during imaging. The injection mix was prepared of 7.7 µl distilled water, 2 µl phenolred and 0.3 µl of stock solution containing alpha-bungarotoxin RNA at a concentration of 500 ng/µl. At around 15 hpf, the embryos expressing the fluorescent vascular marker of interest were prepared for imaging by dechorionation with sharp forceps under a stereomicroscope.

Fluorinated Ethylene Propylene (FEP) tubes
Fluorinated Ethylene Propylene (FEP) tubes with an inner diameter of 0.8 mm, and an outer diameter of 1.2 mm were cleaned according to established protocols (Kaufmann et al., 2012;Weber et al., 2014): First, the FEP tubes were flushed with 1M NaOH and then transferred to a 50ml centrifuge tube filled with 0.5 M NaOH and ultrasonicated for 10 min. After that, the tubes were flushed with distilled water and then with 70% ethanol. After flushing the tubes, they were ultrasonicated in 70% ethanol and cut to a length of about 4 cm. If desired, the tubes were straightened before cutting. For this, FEP tubes were inserted into steel tubes of 50 cm length. The outer diameter of the FEP tubes matched the inner diameter of the steel tube. The tubes were heated to 180 °C for 2 h in an autoclave and afterwards cooled at room temperature for at least 5 h.

Connectors
FEP tubes with an inner diameter of 1.1 mm and an outer diameter of 1.5 mm were used as connecting FEP tube (connector). The slightly smaller inner diameter of the connector compared to the outer diameter of the FEP tube containing the sample (1.2 mm) ensured a tight fit. The connector was prepared by mounting the FEP tube on a blunt ended needle (1.2 mm x 40 mm). On this, a hole of about 2 mm was cut in the middle of the connector and the connector cut to its final length of about 6 mm. The connectors were reused in several experiments.

3% methylcellulose for coating
3% methylcellulose was prepared for coating the FEP tube to prevent attachment of the growing fish to the FEP wall. E3 was heated to approximately 60-70 °C and methylcellulose powder was added to a final concentration of 3%. To ensure a homogenous solution, it was stirred at 4 °C overnight.

0.1% agarose in E3 and 3% agarose dish for plugging
The appropriate amount of agarose was dissolved in E3 and heated up in a microwave until the agarose solution appeared homogeneous. 1.2 ml aliquots of 0.1% agarose were prepared in 1.5 ml reaction tubes and stored at 4 °C. Similarly, the 3% low melting agarose was poured into small petri-dishes (60 mm x 15 mm) to a thickness of about 2 mm. After solidification, E3 medium was added on top of the low melting agarose layer to prevent it from drying. Also, the petri-dish with the low melting agarose was stored at 4 °C.

Preparation of the multi-sample tube
The FEP tubes containing one mounted zebrafish embryo (Fig. S2) were cut to a length of about 9 mm providing space for the plug (about 2-3 mm), the zebrafish embryo (at the end of the time-lapse around 4 mm), and the connector (about 2-3 mm). The individual tubes were connected to form one long tube (Fig. S3). Depending on the experimental requirements, the overlap of the connector with the FEP tubes containing the sample could be varied. If only parts of the embryo such as the head were imaged, the tubes could be connected with a higher overlap, resulting in an overall shorter tube and the possibility to image more samples (Fig. S3, B). The hole in the tube not only increased oxygen availability in the sample tube but also made the procedure easier as no air pressure built up when connecting the tubes.

Potential embedding difficulties
Mounting of zebrafish embryos can be challenging. Below, some pitfalls and suggestions how to solve them are listed (Table S1) Table S1: Suggestions for potential embedding problems

Problem Solution
Dying embryos It is important that the embedding material is at room temperature before embedding. Zebrafish growth and survival is affected by too cold and too warm temperatures. After embedding, the integrity of the embryo should be checked under the stereomicroscope. Ideally, the embryo lies just above the plug but does not touch it yet. If dying samples reoccur, also changing the suppliers of any of the materials should be considered. Some people have reported that FEP tubes from some companies might be more likely to decrease sample health. The same applies for agarose. Also check for pH and possible contamination of E3. We only used filtered E3.

Formation of air bubbles
To avoid air bubbles, empty the syringe after sucking up the methylcellulose by disconnecting it from the cannula. Put the syringe and cannula back together, wipe the methylcellulose from the exterior of the FEP tube and rinse it several times with E3. Before mounting the embryo in the tube, also suck up some 0.1% agarose. Moreover, fill the dish containing the 3% low melting agarose to produce the plug with a bit of E3. This procedure reduces the formation of air bubbles during plugging. The 0.1% agarose might contain air bubbles. To remove them, vortex the reaction tube shortly before putting the fish into it. Zebrafish is far away from the plug after embedding Before plugging, tap the zebrafish several times on the 3% low melting agarose without plugging. This produces a suction force if the FEP tube is long enough. A length of 3.5 -5 cm proved to be ideal for this.

Zebrafish moves during imaging
If you apply Tricaine as described in the original sample mounting paper (Kaufmann et al., 2012), zebrafish move at around 1 dpf. Therefore, we used alpha-bungarotoxin RNA in our long-term imaging studies. Also make sure that the embryo is close to the plug before starting an imaging experiment and that the FEP tubes have not been bent before the imaging was started. Development: doi:10.1242/dev.173757: Supplementary information
(iv) now you can start your Fiji plugin from the command line: sbatch start_yourplugin.sh Note: If the .sh file is not found, be sure to be in the right folder where you also saved your .sh file Development: doi:10.1242/dev.173757 Comparison of old and new scheme. In the old scheme, data transfer (gray) did not start until data acquisition (black) was complete. With this acquisition scheme, there was a risk of filling the hard disks (F) before the end of the planned acquisition time. In the new scheme, after every acquisition (black), data is transferred (gray). Ideally, all data acquired was transferred, and therefore the available hard disk space did not decrease. Moreover, a new experiment could be started right at the end of first experiment leading to higher throughput. Development: doi:10.1242/dev.173757: Supplementary information Overview of processing pipeline

Fig. S7: Overview of data processing pipeline
The dedicated data processing pipeline was composed of modular parts for data compression (grey), projections (green), stitching of the data (blue), registration (brown), analysis (white). Full arrows indicated data flow, dashed arrow indicated the input of parameters from the microscope such as stage positions to the processing steps. Data quality was evaluated visually (à) to ensure proper processing of the data. (A) Stitching of two images (magenta and cyan) zebrafish expressing the vascular marker Tg(kdrl:EGFP) (Jin et al., 2005). The overlap between the two images is highlighted in violet. (B) To automatically calculate the stitching of the two images in (A), an initial offset was set to initialize the calculation. Varying this offset revealed that it was robust over a large range of initialization values. Nevertheless, stitching became unreliable if the value were too far off from the assumed offset given by the translational stages (red line). Moreover, the stitching result indicated that the calculated translation (black line) was close to the physical coordinate given by the stage position (blue line) but not exact (0.06 mm / 53 pixels away). Therefore, using the stage position was a reasonable estimate for initialization. (C) In our implementation of stitching, stage positions determined which tiles to combine and then based on the initialization, the stitching generated a final output image.
Here, a zebrafish expressing the vascular marker Tg(kdrl:Hsa.HRASmCherry) (Chi et al., 2008) is shown. Scale bar 0.5 mm. Development: doi:10.1242/dev.173757 The Gompertz model (Gompertz, 1825), Weibull model (Ratkowsky, 1983), Richards model (Richards, 1959;Tjørve and Tjørve, 2010), a normalized log-logistic growth model modified from (Bennett, 1983), the logistic model (Verhulst, 1938), and a normalized log-normal growth model modified from (Johnson et al., 1994) were compared and their residual sum of errors (RSS) calculated. A low RSS indicated a good fit. For each growth model, the predicted values from the model (blue line) were overlaid with the experimental data (dashed line -mean, gray ribbon 95% confidence interval). Moreover, for each growth model, the residuals were plotted over time. Small residual values combined with a random distribution of them around the 0 value (gray line) indicated good description of the data by the growth model. The log-normal and log-logistic growth models represented the data the best of all the tested models.

Fig. S12: Growth characteristics of the caudal vein plexus
The volume of the caudal vein plexus in the tail grew until around 73 hpf, after which its size started to decrease. The mean of the measurements is depicted with a blue line and the 95% confidence interval (t-statistics, n=7) as a ribbon in the light turquoise.

Movie 1
Zebrafish embryos expressing the fluorescent vascular marker Tg(kdrl:EGFP) in cyan and the red blood cell marker, Tg(GATA1a:dsRed) in magenta imaged from three different angles over several days. Scale bar: 0.5 mm.

Movie 2
Three different zebrafish embryos expressing the fluorescent vascular marker Tg(kdrl:EGFP) in cyan and the red blood cell marker, Tg(GATA1a:dsRed) in magenta imaged in the same experiment simultaneously over several days. Scale bar: 0.5 mm.

Movie 3
3D-rendered segmentation of the zebrafish vasculature with annotations of the head vasculature in orange, the tail in turquoise with its caudal vein plexus highlighted with light-turquoise and unannotated vasculature in grey.