1. Introduction

Structural and functional heterogeneities in one population are common phenomena for biological specimens from cells to tissues to model organisms, and need to be carefully evaluated in biomedical studies such as developmental biology and drug discovery. Although the heterogeneities have been widely addressed at the cell level [1], fewer studies were reported about the heterogeneities in model animals because of the dramatically increased expense compared to cell lines and lack of appropriate high throughput characterization techniques [2]. Zebrafish share 87% genome homology with human beings and are ideal for heterogeneous study as a vertebrate model. A female zebrafish can produce about one hundred eggs at one time and the embryos grow up to larvae in 3 days. In the last twenty years, zebrafish have been widely used for developmental studies and drug discovery [3–6]. For example, studies on zebrafish vasculature have significantly contributed to the understanding of the molecular mechanism of vascular formation and the screening of anti-angiogenic drugs [7–9]. However, heterogeneity was not well-quantified at a large scale in these studies due to the lack of high throughput imaging tool.

A benefit of using zebrafish as a model animal is the transparency of the larvae body which enables high-resolution imaging with various optical microscopy techniques [10,11]. A few attempts have been made to develop high throughput imaging instruments for zebrafish [12–17]. A straightforward high throughput imaging method was to embed multiple zebrafish samples in microfluidic wells and then image them one by one under a commercial microscope [18,19]. In 2010, Pardo-Martin et al. reported a vertebrate automated screening technology (VAST) that could automatically load each zebrafish larva to the field of view of a commercial microscope and rotate to a defined position for observation within 19s [12]. In 2017, Guo et al. reported the volume and surface distribution of up to 24 larvae from 3-5 days post-fertilization (dpf) using VAST [20]. Recently, VAST was used to characterize morphological and behavioral features of CRISPR-generated mutant zebrafish with autism phenotypes [21]. In 2021, Chen et al. developed an acoustofluidic rotational tweezing (ART) platform that enables contactless and rapid (∼1s/rotation) rotation of zebrafish larvae for three-dimensional (3D) multispectral imaging [14]. Phenotyping the effects of acute ethyl alcohol exposure on body morphology and liver size were characterized with ART for a total amount of 96 larvae. However, in the above studies, the imaging quality and efficiency were limited by the used commercial microscope. Usually, only bright field imaging or low-resolution fluorescent imaging were performed to provide rough features related to the shape and size of the embryos.

Developmental and drug response studies require the high throughput imaging technique with high-resolution 3D imaging capability and a whole-larva field of view. Light-sheet fluorescence microscopy (LSFM), also named selective plane illumination microscopy (SPIM), represents an ideal 3D imaging technique, owing to its high imaging speed and low photobleaching [22–24]. LSFM has been employed in numerous zebrafish studies, such as development and neuroscience [25,26]. However, when imaging with LSFM, the zebrafish were usually loaded in a specific chamber [22] or on a coverslip [27–29], which is not compatible with high throughput imaging of multiple samples. An alternative method is to drive samples through micro-fluidic devices. In 2015, Gualda et al. developed a SPIM-fluid setup in which the zebrafish were driven to an imaging module through a fluidic device [13]. SPIM-fluid can be operated in either Flow-Mode (F-Mode) for high-throughput imaging or Vertical-Mode (V-Mode) for high-resolution and time-lapse imaging. With F-Mode, 3D images of 12 anesthetized Fli:GFP or Hras:GFP zebrafish larvae were reported. However, the images with F-Mode had severe translational artifacts due to sample movement. So they had to be switched to V-Mode for high-resolution imaging and then lost high throughput capability. In 2018, Logan et al. integrated automatic fluidic control and image-based registration into a beam-scanned light-sheet imaging system and obtained a throughput of 30 volumes (volume size 666 μm*431 μm*1060 μm) per hour [15]. But the flow had to be stopped and the sample was driven by a slow piezo stage during light-sheet imaging, which limited their throughput to be further improved.

Here, we report a light-sheet flow imaging system (LS-FIS) for high throughput 3D imaging of zebrafish embryos. Different from traditional light-sheet microscopy, LS-FIS captures 3D images of multiple embryos by driving them through the light sheet illumination plane at a titled angle and then stacking the light sheet images. Hence in principle, there is no limitation of field of view (FOV) at the flow direction and no limitation of samples to be imaged. Bright-field images were simultaneously captured to calibrate the zebrafish translational distance and correct the image distortion. Automatic loading and dispensing of LS-FIS enable whole-larva 3D imaging of a hundred zebrafish embryos in 30 minutes. An algorithm for automatic registration of 3D structures to the same posture was also developed. With LS-FIS, we investigated the heterogeneity of the zebrafish vascular development from 3–9 dpf. Typical head and trunk vessels were extracted and their growth rates were measured quantitively from 3D images. The heterogeneity of vasculature development reviewed by LS-FIS emphasizes the importance of large-scale statistics for future studies in developmental biology.

2. Materials and methods

2.1 Zebrafish embryo preparation

A stable transgenic zebrafish model Tg(kdrl: EGFP), expressing the green fluorescent protein in vascular endothelial cells, was used for 3D imaging. The embryos were first raised until 24 hpf in an egg-water medium (0.6 g sea salt per 10 L purified water) and then transferred to egg-water containing 0.006% PTU (1-phenyl-2-thiourea, P7629, Sigma) to inhibit pigment formation. The embryos were maintained at 28 °C in a 14-h light/10-h dark cycle. The embryos at three dpf were treated with Pronase (Cat. NO. 10165921001, Sigma) to break down the chorion. Before the LS-FIS imaging experiment, all embryos were anesthetized with 0.4% tricaine (Sigma, A5040). The zebrafish preparation processes were carried out in accordance with the Guide for the Care and Use of Laboratory Animals (published by U.S. National Academy of Sciences, ISBN 0-309-05377-3), and were approved by the animal ethics committee of Suzhou Institute of Biomedical Engineering and Technology, CAS.

To characterize the vascular development of zebrafish embryos, 55 embryos from one spawning were imaged once per day for seven successive days. Thus a total of 385 embryo structures were captured in 3D. For one round of imaging in one day, it took an average of 23 min to complete all of the processes, including loading, imaging, and dispensing.

2.2 LS-FIS setup

LS-FIS mainly consisted of a light-sheet microscope and automatic sample-loading fluidics (Fig. 1 and Visualization 1). The optics of LS-FIS includes a light-sheet fluorescence excitation and bright-field illumination arm, a wide-field fluorescent detection arm, and a bright-field detection arm (Fig. 1(A)). A 100mW 488-nm laser was coupled to a single-mode fiber and then collimated by lens L1 (AC254-50-A, Thorlabs). A vertical slit (VA100C/M, Thorlabs) was used to block the stray light. The beam passed through a cylinder lens CL (ACY254-200-A, Thorlabs) and was then focused along the x-axis on the back focal plane of illumination lens L3 (AC254-50-A, Thorlabs), generating a light-sheet in the imaging chamber at an angle of 30° to a fluidic capillary. Fluorescence was collected by a water immersion objective O1 (10×, NA 0.3, Olympus) perpendicular to the light-sheet (i.e., at an angle of 60° to the fluidic capillary). A long-pass filter (LP02-488RU-25, Semrock) was used to reject the excitation light. The fluorescence image was focused by tube lens TL1 (AC254-150-A, Thorlabs) collected with an sCMOS camera (Prime 95B, Photometrics). The laser power at the sample region was typically 5mW during imaging.

 

Fig. 1. Principle of the high-throughput 3D imaging system, LS-FIS. (A) Sketch of the LS-FIS optics. L1-L3: lens; CL: cylinder lens; DM: dichroic mirror; M1-M3: mirror; O1-O2: objective; TL1-TL2: tube lens; CAM1-CAM2: camera; C: capillary. See Visualization 1. (B) Sketch of the LS-FIS fluidics. The fluidic flow was driven by a syringe pump between the buffer pool, sample reservoir, and dispense pool and controlled by two three-way valves and two photon-detectors. Val1-Val4: electromagnetic valve; LS: light sheet imaging; BF: bright filed imaging. See Visualization 2. (C) Timeline to reverse the pump, change the flow rate, open/close the valves with the signals from photon-detector PD1 (red) & PD2 (green), and the signal to trigger the imaging camera (blue). DET: detection of larva arrival; Trig: camera trigger. (D) Reconstruction of 3D embryo image by stacking light-sheet images with the correction of larva translational position.

Download Full Size | PDF

Bright-field imaging was used to calibrate the embryo's translational speed. A commercial white LED light source (Suzhou Keyi-Sky Inc.) with a liquid core fiber output was filtered by a bandpass filter (FB680-10, Thorlabs, not shown), collimated by lens L2 (AC254-30-A, Thorlabs), and finally combined with the fluorescence excitation arm by a short-pass dichroic mirror DM (SP650R, Thorlabs). The sample was imaged through objective O2 (4×, NA0.1, Olympus) at the opposite side. A bandpass filter (645QM75, Omega Optical, not shown) was used to block the light-sheet excitation laser beam, and a tube lens TL2 (AC254-200-A, Thorlabs) was placed before the bright-field camera (acA2000-165um, Basler).

A capillary made of Fluorinated ethylene propylene (FEP) with a refractive index of 1.338 was placed in the imaging zone at 30° to the light-sheet plane. The capillary has an inner diameter of 0.8 mm. Hence, an elliptical section with an area of 0.8 mm * 1.6 mm area was illuminated by the light-sheet and imaged by the imaging camera with a 600 × 1,200 pixels FOV. A rate of 140 frames per second can be achieved with an exposure time of 1 ms. All electrical devices were controlled by a workstation (Dell T5810) with a DAQ card (6353, National Instruments) as the master. Optical and fluidic parts in LS-FIS are listed in Appendix Table 1. The control and image processing software were developed with Python. The 3D images of LS-FIS were rendered using Imaris (Oxford Instruments, England). 2D MIP was performed with the ImageJ. 3D structures of the embryos were analyzed and measured with V.R. equipment (HTC Vivi Pro) and DIVA software [30]. Benefitting from the straightforward visualization of tiff-stack and convenient measuring tool in a user-friendly 3D virtual reality environment, the measurement of vascular structures in 3D space was conducted immersively. 2D structures from the MIP were analyzed and measured with ImageJ.

2.3 High-throughput 3D imaging

Embryos for LS-FIS imaging were stored in a 50-ml reservoir tube with a bubble mixer to prevent the embryos from sinking (Fig. 1(B) and Visualization 2). All loading and dispensing processes were controlled by a syringe pump and two three-way valves, Val1-2, and Val 3-4 (Fig. 1(C)). Two photodiodes (PD1 and PD2) were used to detect the arrival of each embryo. Before loading, the syringe pumper moved up and down to clear the air bubbles in the capillary (bubble clearing stage) with Val1 open and Val2 closed. A high flow rate v₀ (typically 167 μl/s, 334 mm/s) was applied to absorb one embryo from the sample reservoir to the capillary with Val2 and Val3 opened during the loading stage. When detected by PD1, Val3 was closed, Val4 was opened, and the syringe pump reversed the flow direction to push the embryo to the imaging module (sending stage). The embryo was first pushed with a fast flow rate v₁ (typically 83 μl/s, 166 mm/s) until detected by PD2, the flow rate was reduced to v₂ (typically 4.2 μl/s, 8.4 mm/s), ensuring that the embryo did not travel out of the FOV before imaging.

Once the bright-field camera detected an embryo, the flow rate was further reduced to a slow speed v₃ (typically 0.5 μl/s, 1.0 mm/s); simultaneously, the sCMOS camera was triggered with an exposure time of 1 ms and a frame rate of 140 Hz (imaging stage). Once the entire embryo passed through the imaging zone, the embryo was pushed out to the dispense pool with a fast flow rate v₄ (typically 500 μl/s, 1000 mm/s, unloading stage). The system then began the next cycle to load another embryo.

2.4 3D image reconstruction

Due to the unsteady translocation of the embryo during LS-FIS imaging, the raw 3D image stack needs to be reconstructed corresponding to the embryo position (Fig. 2). Thus, typical area is selected in the bright-field image, and template matching is performed between adjacent frames to determine the translation distance δ . The position of the embryo along the capillary was obtained according to the calculated position (Fig. 2(A)-(B)). The raw stacks were then interpolated in the axial direction and resampled evenly on a constant step of 3.99 μm. Meanwhile, because the flow direction was not perpendicular to the light-sheet plane, a spatial shift (δ*cos(30°)) along the x-axis between adjacent frames was applied (Fig. 2(C)).

 

Fig. 2. 3D image reconstruction with corrected translational distance and image shift. (A) Bright-field images captured by CAM2 to calibrate the embryo movement driven by the flow. Template matching (red box) is applied to determine the translational distance. (B) The calculated translational location of the embryo at each frame in an LS-FIS imaging sequence. (C) 3D reconstruction by resampling along z-axis to a constant distance of 3.99 μm and applying a shift along x-axis between adjacent frames for the 3D light-sheet data. (D) MIP image of an original 3D volume obtained without the correction of translational distance. (E) MIP image from the same set 3D image data with the correction of translational distance. Scale bars represent 50 μm for (D) and (E).

Download Full Size | PDF

The resulting images were cropped to remove the extra borders. Because the voxel size in the original data is not equal at the x-y plane (1.33 μm) and along the z-direction (3.99 μm), they were down-sampled by two times and up-sampled by 1.5 times respectively, yielding a final reconstructed 3D image with even voxel size of 2.66 μm. Typical results before and after correction are shown in Fig. 2(D)-(E).

2.5 3D image registration

Each embryo came into the FOV of LS-FIS with either “head first” or “tail first”, and with a random posture (rotation angle). The large number of 3D structures need to be registered to the same posture for analysis. Especially, 2D Maximum Intensity Projection (MIP) images at the dorsal view and lateral view are generally required for visual check and comparison to previously reported 2D structures. Hence, we developed an algorithm for automatic registration of the 3D zebrafish vasculature obtained with LS-FIS (Fig. 3).

 

Fig. 3. Auto registration of the 3D zebrafish vasculature structure. (A) A typical MIP image of a 3D structure obtained by LS-FIS and the corresponding intensity profile to determine the direction of the head. (B–D) MIP image of the tail region (red box in A) with three different rotation angles. (E) Variances of the serial MIP images with different rotation angles. The peak indicates the rotation angle in lateral view; the minimum indicates the rotation angle in dorsal view. (F, G) Registered MIP images of lateral (F) and dorsal (G) views for the whole zebrafish embryo. Scale bars represent 400 µm for (A, F, G) and 200 µm for (B, C, D), respectively.

Download Full Size | PDF

The algorithm first detects the “head” or “tail” by calculating the fluorescent intensity profile along the flow direction. Since the head of the zebrafish always exhibits a stronger fluorescence than the trunk, the position of the centroid relative to the image center will indicate the “head” or “tail” (Fig. 3(A)). Then the lower half of the larva (caudal portion) was extracted and rotated along the z-axis with an angle $\alpha$ from 0° to 180° to determine the posture of the zebrafish (Fig. 3(B)-(D)). Since the intersegmental vessels (ISVs) are primarily distributed laterally and are more concentrated in the dorsal-ventral aspect, the variance in the portion of the MIP images serves as a good indication of the lateral view or dorsal view:

It should be noted that although LS-FIS takes less than 20s for a zebrafish larva, the 3D image reconstruction and registration takes about 3 min with a Dell T5810 computer (Intel Xeon @ 3.7 GHz CPU and 32 GB RAM). This image processing time can be largely reduced with a GPU accelerated computation in the future.

2.6 3D image visualization and analysis

To quantify trunk vessels at the yolk extension region during development, polygon selection in ImageJ was performed on the lateral view to extract ten pairs of ISVs with the first pair located at the posterior of caudal vein (CV). The secondary sprout vessels at the same region were also selected. The resulting image stacks were binarized in 3D using the Otsu method and skeletonized with the Skimage package in Python. The data were then analyzed through the ImageJ plugin “Analyze Skeleton (2D / 3D)” [31]. Counting of secondary sprout vessels was performed manually with Imaris.

To quantify the size of hyaloid baskets, measurements were performed with a VR platform, DIVA, in which the specific vessels from dense head structures are easy to be identified. Brightness and contrast may be adjusted in desk mode to show the apparent hyaloid basket shape if necessary. The measured data points in DIVA were exported as CSV files and the subsequent counting was performed in Python.

3. Results

3.1 Characterization of LS-FIS

Figure 4(A) shows the intensity profile of the light-sheet in LS-FIS measured with the bright-field imaging arm and replacing the long pass filter with an appropriate neutral density filter. The thickness of the light-sheet is 9.8 μm (Fig. 4(B)) at the center, and 14.1 μm at halfway away from the center (Fig. 4(C)), respectively. Note that the resolution along the flow direction depends on not only on the thickness of the light sheet, but also the translational speed of the embryos. Fluorescent beads with size of 100 nm were used to calibrate the lateral resolution of LS-FIS. Figure 4(D) shows a raw image captured by the sCMOS with a pixel size of 1.33 μm. Fitting of the intensity profile of 30 randomly selected beads with Gaussian function shows a mean Full-Width-at-Half-Maximum (FWHM) of 2.93 μm (Fig. 4(E)).

 

Fig. 4. Characterization of thickness and resolution of LS-FIS. (A) Light-sheet in LS-FIS captured by the bright-field imaging arm with the long pass filter replaced by an appropriate neutral density filter; (B) Intensity along the blue line at the center of light-sheet in (A) is fitted by Gaussian function showing a FWHM of 9.8 μm. (C) Intensity along the red line at halfway away from the center of light-sheet in (A) is fitted by Gaussian function showing a FWHM of 14.1 μm; (D) Raw image of 100 nm fluorescent beads captured by LS-FIS for calibration of the lateral resolution; (E) Intensity profile along the center of the bead marked in (D). The black curve shows the Gaussian fitting to the profile with FWHM of 2.9 μm. Scale bars represent 200 µm for (A) and 150 µm for (D), respectively.

Download Full Size | PDF

The unlimited image size along the flow direction of LS-FIS enables 3D imaging of zebrafish whole larva. Figure 5(A) and Visualization 3 shows a typical 3D vasculature of a zebrafish at 5 dpf. Zoomed image in Fig. 5(B) shows the vessels at the trunk, including caudal vein (CV), intersegmental vessels (ISV), dorsal longitudinal anastomotic vessel (DLAV), vertebral artery (VTA), dorsal aorta (DA), and posterior cardinal vein (PCV). Especially, the tubular structure of DLAVs and DA can be seen from the section view (Fig. 5(C)) with their inner diameter measured to be 7.5 μm and 17 μm. At the head of the zebrafish, more complex vessels exist and some of them are hard to be distinguished due to light scattering, as shown in Fig. 5(D). However, the dorsal midline junction (DMJ), formed by the dorsal longitudinal vein (DLV), mesencephalic vein (MsV) and middle cerebral vein (MCeV), can be always distinguished at the top of the head. The structure of the DMJ is consistent with the previously established blood vessel atlas [32]. The pair of hyaloid baskets locate on both sides of the head, but only the one facing the illumination could be seen while the other one was smeared out because of scattered light. In the following, the vessels at the trunk and the resolved hyaloid basket are selected to evaluate the vasculature development process.

 

Fig. 5. Whole-larva 3D image of zebrafish vasculature captured with LS-FIS at 5 dpf. (A) 3D rendering of the whole larva. See also Visualization 3. (B) Zoomed-in of the box in (A) showing vessels at the truck. CV: caudal vein; ISV: intersegmental vessels; DLAV: dorsal longitudinal anastomotic vessel; VTA: vertebral artery; DA: dorsal aorta; PCV: posterior cardinal vein. (C) Sectioned view along the dashed line of (B). (D) Zoomed-in of the box in (A) showing the vessels at the head. DLV: dorsal longitudinal vein; MCeV: middle cerebral vein; MsV: mesencephalic vein; HB: hyaloid basket. Scale bars represent sent 100 µm for (B, D) and 50 µm for (C), respectively.

Download Full Size | PDF

3.2 Development of zebrafish trunk vessels

Vasculature development plays an essential role in the growth of embryos and is closely related to many diseases in the coronary heart and nervous system. We used LS-FIS to study the vascular development process of zebrafish from 3–9 dpf. Approximately 50 zebrafish embryos with EGFP labeled in the vascular system were run through LS-FIS one time every day.

We first examine the development of zebrafish trunk vessels. Vessels at the yolk extension (YE) regions were defined as trunk vessels, including DLAVs, ten pairs of ISVs, and other primary vessels such as DA, caudal vein (CV), and subintestinal veins (SIVs) (Fig. 6(A), Visualization 4). The most apparent trunk vessels are ISVs, which have a well-ordered periodic structure from 3 dpf. Pendicular to the ISVs, the secondary sprout vessels begin to sprout and gradually anastomosed from 4 dpf (Fig. 6(B)–(H)).

 

Fig. 6. Development process of zebrafish trunk vessels from 3 to 9 dpf. (A) A typical whole larva 3D image at 3 dpf with a red box indicating the trunk vessels at the yolk extension region. See also Visualization 4. (B–H) 3D render of representative trunk vessels from 3 dpf to 9 dpf, respectively. White, yellow, and red arrows indicate the ISV, the sprouting, and anastomosis of the secondary sprout vessel, respectively. (I) The total length of ISVs at the yolk extension region measured from the 3D volume data (blue) and the 2D MIP (red). (J) Number of ISVs with the sprouted secondary vessel. (K) Percentage of larvae with the secondary sprout. (L) The number of anastomosed secondary sprout vessels. Error bars represent the corresponding standard deviation.

Download Full Size | PDF

To analyze the trunk vessels quantitatively, ten pairs of ISVs, as well as the secondary sprout vessels, were extracted from the 3D images. Then, the extracted region was binarized and analyzed to quantify the vascular development process from the 3D structure. As a comparison, measurements were also performed on 2D MIP images in lateral view. Both curves (Fig. 6(I)) show rapid growth before 6 dpf, while the growth rate measured from the 3D volume data is higher than that from the 2D MIP images. After 7 dpf, the total length remains relatively constant from the 3D volume measurements. Whereas the measured length from 2D MIP images decreases slowly, indicating that extracted vessels become curved in the 3D space. The differences in measured length also suggest that traditional 2D imaging is insufficient to quantify the embryo developmental process. The average coefficients of variation (ratio of variance to the mean value of the ISV length) were 0.34 and 0.23 for 3D volume measurement and 2D MIP measurement, respectively, indicating obvious heterogeneity of ISV development among individual embryos.

Figure 6(J)-(K) shows the number of secondary sprout vessels over the ten pairs of ISVs and the percentage of larvae with at least one secondary sprout. Only 30% of larvae began to sprout at 3 dpf, while 90% of larvae exhibited secondary sprouts at 4 dpf, indicating the onset of the sprouting between 3 dpf to 4 dpf. After that, the vessel began to be anastomosed. The number of anastomoses (Fig. 6(L)) shows a steady increase before 8 dpf with approximately half of the ISVs anastomosed through secondary sprout vessels thereafter.

3.3 Development of hyaloid basket vasculature

Next, we evaluated the development of hyaloid vasculature, which is a vascular network that nourishes the embryonic eye as it develops. Hyaloid vasculature is located in the anterior of the eye between the retina and lens. Failure to develop the hyaloid vascular leads to many eyes disease such as ocular coloboma [33,34]. Previous studies with a time-lapse confocal microscope show that the formation of hyaloid vasculature can be divided into three distinct stages and is fully enclosed by 5 dpf [35]. However, the difference in hyaloid development between different embryos was not studied so far.

We segment out the head regions from the same 3D imaging dataset (Visualization 5). The hyaloid vasculature at one side of the head could be distinguished as a basket-like shape from 3-8 dpf (Fig. 7(A)-(D), Visualization 6, Visualization 7). The other hyaloid was smeared in most cases because of strong light scattering by the head. At 9 dpf, the heads of many zebrafish become too big that both hyaloid cannot be distinguished. Hence, we measured the diameter and depth of the clearly-distinguished hyaloid basket from 3 to 8 dpf (Fig. 7(E)-F), using a Virtual-Reality (VR) based display and measurement platform. As shown in Fig. 7(G), the steady increase of the diameter and depth indicate that the hyaloid continuously grow even after they are fully enclosed. Variances of the diameter and depth are constant, about 10% of their average value. In contrast, the ISVs stop their development at 7 dpf and the variance of ISVs’ length is about 34% of their average value (Fig. 6(I)). This implies that the heterogeneity of zebrafish hyaloid development is less compared to the vasculature in the trunk, probably because of restriction of head structure. It also suggests that proper visual function requires a small structural difference of the eyes.

 

Fig. 7. Development of zebrafish hyaloid vasculature. (A) Typical head vessel structure; red arrow indicates the hyaloid vessel. See Visualization 5. (B) Schematic of the hyaloid basket from ventral-dorsal view. (C–D) Typical hyaloid vessels at 3 dpf and 7 dpf, respectively. See Visualization 6 and Visualization 7. (E) Projection along the lens axis of a typical hyaloid at 5 dpf showing an ellipse shape (red); blue arrows indicate its long axis and short axis, and their mean value is taken as the diameter of the hyaloid basket. (F) Projection along the V-D axial of the hyaloid showing a basket shape. The blue dashed circle indicates the lens, and brown arrows indicate the depth of the basket from the lower edge to the bottom. (G) Diameters (red) and depths (blue) of hyaloid baskets dependent on time show a continuous growth from 3 to 8 dpf. Scale bar: 40 μm for (E) and (F). D: dorsal, V: ventral, N: nasal, T: temporal, La: lens anterior, Lp: lens posterior. Error bars represent the corresponding standard deviation.

Download Full Size | PDF

4. Discussion

During their development, all organs of zebrafish have more or less structural differences even in one population. This heterogeneity is important for quantitative analysis of their development. However, by now there is a lack of high throughput imaging tools to characterize their 3D structures. For this purpose, LS-FIS was developed with automatic flow-driven sample-loading and high-resolution 3D imaging capabilities. LS-FIS can load, image, and dispense each embryo in 20 s. Notably, the time spent on the light-sheet imaging is about 4s for one embryo, most other time is spent on loading the sample to the imaging module. The loading efficiency varies with the concentration of embryos in the reservoir pool. Practically, 3D imaging of 100 zebrafish embryos can be finished in about half an hour.

3D imaging with high resolution (2.9 μm at x-y plane, about 10 μm along the flow direction) could be achieved with LF-FIS so that the vasculature can be studied in detail. The light sheet in the current LS-FIS is generated with a cylinder lens and illuminated from one side of the sample. To compromise with the large FOV of 0.8 mm * 1.6 mm, a relatively thick light sheet with an FWHM of 9.8μm at the center was shaped to illuminate the sample. In comparison, the classic SPIM by Huisken et al. used a light sheet with a thickness of 6 μm to achieve the FOV of 660 μm [22]. Deterioration of the image quality was observed at the side of the image away from illumination, especially at the zebrafish head region due to stronger light scattering (Fig. S1). In the future development of LS-FIS, a better 3D resolution could be achieved by optimization of the embryo translocation through the FOV, beam-scanned light sheet generation, and dual side illumination. Simultaneous dual-color imaging can also be incorporated to achieve more than one type of tissue structure, which will greatly benefit the developmental and drug response study.

LS-FIS was used for high throughput 3D imaging of zebrafish larvae from 3 dpf to 9 dpf. The vasculature developments from 3 to 9 dpf show a large variance of trunk vessel development but relatively less variance for hyaloid vessels. The heterogeneity indicates the corresponding gene expression or activation is not well coordinated among individuals. In the future, a deep-learning-based processing pipeline may be developed to identify each vessel from the 3D images and automatically track their developments over time.

Besides the zebrafish vasculature, LS-FIS can also be used for high throughput 3D imaging of other organs, including the neuron system, immune system, bones, and muscles. Considering there are thousands of transgenic zebrafish models representing different labeled structures and mutations, they are all worthy to perform high throughput imaging to characterize the structural and functional heterogeneity. In addition to development biology, LS-FIS could also be widely use to pharmacological studies for drug screen, in which massive test of drugs’ effects are required.

In summary, LS-FIS provides a high throughput 3D imaging platform that can examine the 3D structure of zebrafish in whole larva size. LS-FIS can provide significantly more information than conventional microscopes, thus enabling large-scale evaluation. We expect that LS-FIS can be a powerful tool for the heterogeneity study in developmental biology and to increase the efficiency of drug discovery.

Appendix

Table 1. List of optical and fluidic parts in LS-FIS

View Table

Funding

Suzhou Basic Research Pilot Project (SJC2021013); Key Research and Development Program of Jiangsu Province (BE2020664); Strategic Priority Research Program of Chinese Academy of Sciences (XDC07040000).

Disclosures

The authors Guang Yang, Xin Jin, Yong Liang, Xiaohu Chen, Linbo Wang and Hui Li have filed a patent application, and they stand to benefit should the patent be awarded.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplemental document

See Supplement 1 for supporting content.

References

1. L. Li, Q. Zhou, T. C. Voss, K. L. Quick, and D. V. LaBarbera, “High-throughput imaging: Focusing in on drug discovery in 3D,” Methods 96, 97–102 (2016). [CrossRef]  

2. J. Giacomotto and L. Segalat, “High-throughput screening and small animal models, where are we?” Br. J. Pharmacol. 160(2), 204–216 (2010). [CrossRef]  

3. L. I. Zon and R. T. Peterson, “In vivo drug discovery in the zebrafish,” Nat. Rev. Drug Discov. 4(1), 35–44 (2005). [CrossRef]  

4. G. J. Lieschke and P. D. Currie, “Animal models of human disease: zebrafish swim into view,” Nat. Rev. Genet. 8(5), 353–367 (2007). [CrossRef]  

5. S. Cassar, I. Adatto, J. L. Freeman, J. T. Gamse, I. Iturria, C. Lawrence, A. Muriana, R. T. Peterson, S. Van Cruchten, and L. I. Zon, “Use of zebrafish in drug discovery toxicology,” Chem. Res. Toxicol. 33(1), 95–118 (2020). [CrossRef]  

6. P. Letrado, I. de Miguel, I. Lamberto, R. Díez-Martínez, and J. Oyarzabal, “Zebrafish: speeding up the cancer drug discovery process,” Cancer Res. 78(21), 6048–6058 (2018). [CrossRef]  

7. D. Li, W. Xue, M. Li, M. Dong, J. Wang, X. Wang, X. Li, K. Chen, W. Zhang, S. Wu, Y. Zhang, L. Gao, Y. Chen, J. Chen, B. O. Zhou, Y. Zhou, X. Yao, L. Li, D. Wu, and W. Pan, “VCAM-1 + macrophages guide the homing of HSPCs to a vascular niche,” Nature 564(7734), 119–124 (2018). [CrossRef]  

8. W. Zhao, L. Cao, H. Ying, W. Zhang, D. Li, X. Zhu, W. Xue, S. Wu, M. Cao, C. Fu, H. Qi, Y. Hao, Y. C. Tang, J. Qin, T. P. Zhong, X. Lin, L. Yu, X. Li, L. Li, D. Wu, and W. Pan, “Endothelial CDS2 deficiency causes VEGFA-mediated vascular regression and tumor inhibition,” Cell Res. 29(11), 895–910 (2019). [CrossRef]  

9. S. P. Herbert, J. Huisken, T. N. Kim, M. E. Feldman, B. T. Houseman, R. A. Wang, K. M. Shokat, and D. Y. R. Stainier, “Arterial-venous segregation by selective cell sprouting: an alternative mode of blood vessel formation,” Science 326(5950), 294–298 (2009). [CrossRef]  

10. M. Mickoleit, B. Schmid, M. Weber, F. O. Fahrbach, S. Hombach, S. Reischauer, and J. Huisken, “High-resolution reconstruction of the beating zebrafish heart,” Nat. Methods 11(9), 919–922 (2014). [CrossRef]  

11. T. V. Truong, W. Supatto, D. S. Koos, J. M. Choi, and S. E. Fraser, “Deep and fast live imaging with two-photon scanned light-sheet microscopy,” Nat. Methods 8(9), 757–760 (2011). [CrossRef]  

12. C. Pardo-Martin, T. Y. Chang, B. K. Koo, C. L. Gilleland, S. C. Wasserman, and M. F. Yanik, “High-throughput in vivo vertebrate screening,” Nat. Methods 7(8), 634–636 (2010). [CrossRef]  

13. E. J. Gualda, H. Pereira, T. Vale, M. F. Estrada, C. Brito, and N. Moreno, “SPIM-fluid: open source light-sheet based platform for high-throughput imaging,” Biomed. Opt. Express 6(11), 4447 (2015). [CrossRef]  

14. C. Chen, Y. Gu, J. Philippe, P. Zhang, H. Bachman, J. Zhang, J. Mai, J. Rufo, J. F. Rawls, E. E. Davis, N. Katsanis, and T. J. Huang, “Acoustofluidic rotational tweezing enables high-speed contactless morphological phenotyping of zebrafish larvae,” Nat. Commun. 12(1), 1118 (2021). [CrossRef]  

15. S. L. Logan, C. Dudley, R. P. Baker, M. J. Taormina, E. A. Hay, and R. Parthasarathy, “Automated high-throughput light-sheet fluorescence microscopy of larval zebrafish,” PLoS One 13(11), e0198705 (2018). [CrossRef]  

16. X. Lin, V. W. T. Li, S. Chen, C.-Y. Chan, S.-H. Cheng, and P. Shi, “Autonomous system for cross-organ investigation of ethanol-induced acute response in behaving larval zebrafish,” Biomicrofluidics 10(2), 024123 (2016). [CrossRef]  

17. L. Liu, G. Yang, S. Liu, L. Wang, X. Yang, H. Qu, X. Liu, L. Cao, W. Pan, and H. Li, “High-throughput imaging of zebrafish embryos using a linear-CCD-based flow imaging system,” Biomed. Opt. Express 8(12), 5651–5662 (2017). [CrossRef]  

18. J. Akagi, K. Khoshmanesh, C. J. Hall, J. M. Cooper, K. E. Crosier, P. S. Crosier, and D. Wlodkowic, “Fish on chips: Microfluidic living embryo array for accelerated in vivo angiogenesis assays,” Sens. Actuators, B 189, 11–20 (2013). [CrossRef]  

19. W. Connacher, N. Zhang, A. Huang, J. Mei, S. Zhang, T. Gopesh, and J. Friend, “Micro/nano acoustofluidics: materials, phenomena, design, devices, and applications,” Lab Chip 18(14), 1952–1996 (2018). [CrossRef]  

20. Y. Guo, W. J. Veneman, H. P. Spaink, and F. J. Verbeek, “Three-dimensional reconstruction and measurements of zebrafish larvae from high-throughput axial-view in vivo imaging,” Biomed. Opt. Express 8(5), 2611–2634 (2017). [CrossRef]  

21. A. Colón-Rodríguez, J. M. Uribe-Salazar, K. B. Weyenberg, A. Sriram, A. Quezada, G. Kaya, E. Jao, B. Radke, P. J. Lein, and M. Y. Dennis, “Assessment of Autism Zebrafish Mutant Models Using a High-Throughput Larval Phenotyping Platform,” Front. Cell Dev. Biol. 8, 586296 (2020). [CrossRef]  

22. J. Huisken, J. Swoger, F. D. Bene, J. Wittbrodt, and E. H. K. Stelzer, “Optical Sectioning Deep Inside Live Embryos by Selective Plane Illumination Microscopy,” Science 305(5686), 1007–1009 (2004). [CrossRef]  

23. H.-U. Dodt, U. Leischner, A. Schierloh, N. Jährling, C. P. Mauch, K. Deininger, J. M. Deussing, M. Eder, W. Zieglgänsberger, and K. Becker, “Ultramicroscopy: three-dimensional visualization of neuronal networks in the whole mouse brain,” Nat. Methods 4(4), 331–336 (2007). [CrossRef]  

24. P. J. Keller, A. D. Schmidt, J. Wittbrodt, and E. H. K. Stelzer, “Reconstruction of zebrafish early embryonic development by scanned light sheet microscopy,” Science 322(5904), 1065–1069 (2008). [CrossRef]  

25. R. K. Chhetri, F. Amat, Y. Wan, B. Höckendorf, W. C. Lemon, and P. J. Keller, “Whole-animal functional and developmental imaging with isotropic spatial resolution,” Nat. Methods 12(12), 1171–1178 (2015). [CrossRef]  

26. M. B. Ahrens, M. B. Orger, D. N. Robson, J. M. Li, and P. J. Keller, “Whole-brain functional imaging at cellular resolution using light-sheet microscopy,” Nat. Methods 10(5), 413–420 (2013). [CrossRef]  

27. Y. Wu, A. Ghitani, R. Christensen, A. Santella, Z. Du, G. Rondeau, Z. Bao, D. Colon-Ramos, and H. Shroff, “Inverted selective plane illumination microscopy (iSPIM) enables coupled cell identity lineaging and neurodevelopmental imaging in Caenorhabditis elegans,” Proc. Natl. Acad. Sci. U. S. A. 108(43), 17708–17713 (2011). [CrossRef]  

28. Y. Wu, P. Wawrzusin, J. Senseney, R. S. Fischer, R. Christensen, A. Santella, A. G. York, P. W. Winter, C. M. Waterman, Z. Bao, D. A. Colón-Ramos, M. McAuliffe, and H. Shroff, “Spatially isotropic four-dimensional imaging with dual-view plane illumination microscopy,” Nat. Biotechnol. 31(11), 1032–1038 (2013). [CrossRef]  

29. Y. Wu, A. Kumar, C. Smith, E. Ardiel, P. Chandris, R. Christensen, I. Rey-Suarez, M. Guo, H. D. Vishwasrao, J. Chen, J. Tang, A. Upadhyaya, P. J. La Riviere, and H. Shroff, “Reflective imaging improves spatiotemporal resolution and collection efficiency in light sheet microscopy,” Nat. Commun. 8(1), 1452 (2017). [CrossRef]  

30. M. El Beheiry, C. Godard, C. Caporal, V. Marcon, C. Ostertag, O. Sliti, S. Doutreligne, S. Fournier, B. Hajj, M. Dahan, and J.-B. Masson, “DIVA: Natural Navigation Inside 3D Images Using Virtual Reality,” J. Mol. Biol. 432(16), 4745–4749 (2020). [CrossRef]  

31. I. Arganda-Carreras, R. Fernandez-Gonzalez, A. Munoz-Barrutia, and C. Ortiz-De-Solorzano, “3D reconstruction of histological sections: application to mammary gland tissue,” Microsc. Res. Tech. 73(11), 1019–1029 (2010). [CrossRef]  

32. S. Isogai, M. Horiguchi, and B. M. Weinstein, “The vascular anatomy of the developing zebrafish: An atlas of embryonic and early larval development,” Dev. Biol. 230(2), 278–301 (2001). [CrossRef]  

33. S. S. Kitambi, K. J. McCulloch, R. T. Peterson, and J. J. Malicki, “Small molecule screen for compounds that affect vascular development in the zebrafish retina,” Mech. Dev. 126(5-6), 464–477 (2009). [CrossRef]  

34. M. L. Weaver, W. P. Piedade, N. N. Meshram, and J. K. Famulski, “Hyaloid vasculature and mmp2 activity play a role during optic fissure fusion in zebrafish,” Sci. Rep. 10(1), 10136 (2020). [CrossRef]  

35. A. Hartsock, C. Lee, V. Arnold, and J. M. Gross, “In vivo analysis of hyaloid vasculature morphogenesis in zebrafish: A role for the lens in maturation and maintenance of the hyaloid,” Dev. Biol. 394(2), 327–339 (2014). [CrossRef]