Multimodal high-resolution embryonic imaging with light sheet fluorescence microscopy and optical coherence tomography

Abstract

A high-resolution imaging system combining optical coherence tomography (OCT) and light sheet fluorescence microscopy (LSFM) was developed. LSFM confined the excitation to only the focal plane, removing the out of plane fluorescence. This enabled imaging a murine embryo with higher speed and specificity than traditional fluorescence microscopy. OCT gives information about the structure of the embryo from the same plane illuminated by LSFM. The co-planar OCT and LSFM instrument was capable of performing co-registered functional and structural imaging of mouse embryos simultaneously.

Many diseases have congenital origins, and imaging developmental processes has provided invaluable insight towards understanding these diseases. Recent advances in optical imaging techniques have made it possible to investigate developmental processes at cellular, tissue, and organ levels in embryonic developmental biology [1], spurring growth in developmental biology research [2]. However, it is still challenging to image living biological specimens, such as murine embryos, at the mesoscopic scale with high resolution and sufficient depth penetration. Several techniques have been developed to acquire 3D structural information of biological specimens. Optical projection tomography (OPT) acquires high-resolution images [3], but requires samples to be fixed and cleared, which makes OPT severely restricted for live imaging. Micro-magnetic resonance imaging [4,5] has a resolution between 25 and 100 μm and has been utilized for imaging murine embryos, but requires long imaging times for such resolutions, which limits its use for live samples. Micro-coherence tomography (CT) imaging is capable of high resolutions as well, but utilizes ionizing radiation and requires external contrast agents and fixatives to obtain sufficient resolution and contrast for murine embryo imaging, limiting its applicability for live imaging [6].

Although it was originally developed for ophthalmic applications [7], optical CT (OCT) has since expanded into many other fields, particularly developmental biology [8,9]. It has become a preferred imaging modality in small animal embryonic imaging, especially for the mouse, due to its high spatial resolutions, rapid imaging speed, high contrast, depth penetration, and ability to image tissue structure and functional parameters without exogenous contrast labels [9]. OCT utilizes the interference of low coherent light, has micrometer-scale spatial resolution, has a few millimeters of penetration depth in tissues, and can perform volumetric imaging at video rates [10]. However, OCT lacks the specificity provided by fluorescence microscopy because it relies on intrinsic contrast using backscattered light.

Selective plane illumination microscopy (SPIM) overcomes limitations of traditional fluorescence microscopy [11]. SPIM provides illumination using a defocused light sheet, and image acquisition is performed orthogonally. Since this excitation method minimizes photobleaching and phototoxicity, SPIM has rapidly gained adoption in many fields such as developmental biology [12]. Light sheet fluorescence microscopy (LSFM) is also used for functional imaging [13,14]. Although the illumination and detection paths are decoupled in LSFM, extraneous illumination can cause photodamage [11]. Moreover, samples are often embedded in a clear semi-solid media (e.g., agarose) to prevent unwanted motion, which introduces blurring. [11,15].

OCT has been combined with several fluorescent imaging modalities such as two-photon fluorescence or fluorescence lifetime imaging [16,17]. However, OCT has not been combined with LSFM, which has the benefits over other fluorescent imaging techniques, as mentioned earlier. In this work, we developed a multimodal imaging system combining OCT and LSFM (both hardware and software). OCT generally utilizes low numerical aperture (NA) objectives for a long depth of focus, resulting in spatial resolutions generally worse than those of the microscopy techniques. To supplement the information acquired from the OCT structural images of tissue with molecular specificity, we integrated LSFM with OCT. The integrated system simultaneously images tissue structure and fluorescent-labeled cells. These multimodal images are acquired from the same plane at the same time, providing aligned and registered images of tissue structure and protein expression. The system was characterized, and we present a co-registered 3D image of a transgenic mouse embryo at developmental stage 10.5 (E10.5) with labeled red blood cell progenitors and vasculature.

Figure 1 is a schematic of the combined system. The LSFM and OCT sub-systems have purple and orange backgrounds, respectively. LSFM excitation was at 488 nm (iChrome MLE, Toptica Photonics Inc., Farmington, NY) and detection was at 520 nm. The laser operated in continuous mode with an adjustable power up to 20 mW. The output beam diameter was expanded 3× using a beam expander (GBE03-C, Achromatic Galilean, Thorlabs Inc., Newton, NJ) and directed to a polarization beam splitter (Fig. 1). The excitation beam polarization was adjusted so it was reflected by the polarizing beam splitter to the galvanometer-mounted mirror scanners. The beam was then scanned through a telecentric scan lens (LSM03-BB, Thorlabs Inc., Newton, NJ) using only one of the two scanning mirrors (GVS002, Thorlabs Inc., Newton, NJ) to generate the light sheet (in the Y–Z plane). LSFM generally uses high-NA detection objectives for high lateral resolution; however, notable trade-offs are a short working distance, limited field of view (FOV), and small depth of focus. The LSM03-BB was selected to optimize light sheet thickness while providing a sufficiently large working distance, long depth of focus, and sufficient FOV for embryo imaging (Fig. 1). Telecentric scanning over a relatively large FOV (8 mm × 8 mm) enables its use for other imaging modalities such as OCT. The emitted fluorescent light from the sample was collected (in the X direction, shown in the inset) with a water-dipping 16× immersion objective (N16LWD-PF, Nikon Corp., Tokyo, Japan) with 0.8 NA and a 3 mm working distance. The emitted fluorescence light from the objective passes through a filter (520 ± 10 nm) and an infinity-corrected tube lens before it was imaged onto a digital camera (C11440–22CU, Hamamatsu, Hamamatsu City, Japan).

Fig. 1.

Schematic of the combined LSFM-OCT system. The orange background is the OCT sub-system, and the purple background shows the LSFM sub-system. A, aperture; BPD, balanced photo detector; BE, beam expander; C, collimator; DCC, digital CMOS camera; F, filter; FC, fiber coupler; G, galvanometer-mounted mirror scanners; Obj, objective; PBS, polarization beam splitter; PC, polarization controller; RM, reference mirror; SL, scan lens; TL, tube lens; TS, translation stage. The inset in the top left dashed box shows a side view of the light sheet, scan lens, and the orthogonal detection.

The OCT sub-system is based on a swept source laser (Model 1051 SSOCT, Axsun Tech., Billerica, MA) with a central wavelength of 1035 nm, bandwidth of 109 nm, and sweep rate of 100 kHz. The power of the OCT beam incident on the sample is about 6 mW. The OCT system is based on a Michelson interferometer, as show in Fig. 1. The OCT beam in the sample arm transmits through the polarization beam splitter, where it is combined with the LSFM beam. A polarization controller in the sample arm controlled the polarization state of the OCT beam to maximize transmission through the beam splitter. Both LSFM and OCT beams are co-aligned during scanning such that the LSFM excitation sheet and OCT beam were co-planar to enable trivial co-registration. In order to minimize differences in dispersion between the reference and sample arm of the OCT system, an identical polarization beam splitter was placed in the reference arm. The sensitivity of the OCT system was measured as 100.1 dB, and the sensitivity roll-off was 6.9 dB over 4.13 mm. The axial resolution was measured as 7.6 μm in air.

The OCT and LSFM beams were combined using the polarizing beam splitter, as shown in Fig. 1, which was mounted on a tilt-tip stage for alignment. The co-linearity of the two beams was checked over the OCT imaging depth (~3 mm) by a beam viewer (LaserCam-HR II 2/3-inch, Coherent Inc., Santa Clara, CA). To scan the sample in the vertical direction (i.e., the axis orthogonal to the imaged plane), the sample holder was mounted on a motorized translation stage (X-VSR20A, Zaber Tech., Vancouver, Canada). While OCT systems usually use two scanners for raster scanning, our combined LSFM-OCT system uses only one scanner to generate the light sheet, and the sample is stepped by the motorized stage so that the excitation light sheet remained at the focal plane of the detection objective.

We first characterized the system using a US Air Force resolution target to measure the transverse resolution of both the OCT and LSFM systems. Second, we measured the beam diameter in a fluorescein solution at the focus of the scan lens to determine the axial resolution of the LSFM system. Finally, we used both OCT and LSFM systems to image 1–5 μm fluorescent microspheres embedded in agarose (A4718, Sigma-Aldrich Inc., Saint Louis, MO). The panes in Fig. 2 illustrate the acquired images used to characterize the integrated LSFM + OCT system.

Fig. 2.

(a) US Air Force resolution target image showing the lateral resolution of LSFM detection arm. (b) US Air Force resolution target image showing the OCT lateral resolution. (c) Image of the LSFM excitation beam waist (13.6 μm FWHM) at the scan lens focus. (d) Image of the OCT beam waist (14.5 μm FWHM) at focus of the scan lens. (e) 2D image of the green fluorescent microspheres (1–5 μm) embedded in 1% agarose using the LSFM system. The inset is a zoomed-in view of the small yellow dashed box of a single fluorescence microsphere. (f) 2D image of the green fluorescent microspheres (1–5 μm), shown in red, 20–27 μm non-FL particles, shown in white, embedded in 1% agarose, using the integrated system.

Figure 2(a) shows the image of the resolution target using the LSFM sub-system when illuminated by a white LED lamp. The detection microscope could resolve group 8 element 6, which is equivalent to a transverse resolution of ~2.1 μm. We measured the lateral resolution of the OCT sub-system by imaging the same resolution target. An en face projection is shown in Fig. 2(b). The transmission resolution target was mounted horizontally and illuminated from the bottom such that the objective can transmit the light towards the camera. In the image, we were able to distinguish the lines in group 6 element 1, which means the lateral resolution for the OCT system was at least ~14.9 μm.

Figure 2(c) shows that the LSFM illumination beam full width at half-maximum (FWHM) was ~13.6 μm (calculated by a Gaussian fit across the beam center), which was the light sheet thickness (i.e., axial resolution). This image was acquired by sending the light beam into a fluorescein solution, with the collection objective immersed so that there was no refractive index mismatch. We took an image of the OCT beam in the focus of the scan lens using the beam viewer and analyzed it assuming a Gaussian beam. The image in Fig. 2(d) shows the beam waist at focus and a graph of a Gaussian fit through its center. This image was acquired by magnifying the focal spot using a finite conjugate microscope objective to relay the spot onto the camera. In this manner, the image pixel size was 0.5 μm. Substituting these values, the OCT beam FWHM was ~11.5 μm.

As an another test of the LSFM sub-system, we imaged 1–5 μm fluorescent microspheres embedded in low melting point 1% agarose (w/w). Since 3D imaging was performed by moving the sample vertically, there was unwanted motion during acquisition. This was eliminated by gluing the agarose to the cuvette bottom. To minimize the refractive index mismatch in the emission path, the agarose was immersed in water. The cuvette and phantom were adjusted such that the center of the light sheet was at the focus of the detection objective. Figure 2(e) presents one slice from the image stack of the microsphere phantom. The inset in the dashed yellow box shows a magnified image of a single fluorescent microsphere inside the small yellow dashed square. The imaging time for one slice was ~0.6 s, with a total time of ~360 s for a stack of 600 images. The sample was stepped 6 μm per image, which was slightly less than half of the beam diameter. The contrast is high since a 520 ± 10 nm filter was used to image only the fluorescent emission. The maximum FOV of the LSFM sub-system was measured to be ~1.84 mm by 1.84 mm. We performed a similar experiment using the LSFM + OCT integrated system to simultaneously see OCT and LSFM images of microspheres with sizes near the resolution of the systems. We mixed 1–5 μm fluorescent with 20–27 μm non-fluorescent (non-FL) microspheres, embedded them in agarose, and acquired a stack of images, one slice of which is presented in Fig. 2(f). In the co-registered image, the white points show the non-FL particles imaged by OCT, and the red points are fluorescent particles from the LSFM image. Not all points are individual particles, as some microspheres aggregated.

Implementing the combined LSFM-OCT system, we acquired 3D images of a 10.5 embryo at 10 days post-coitum [strain Tg; e-globin-green fluorescent protein (GFP)]. The OCT and LSFM images were acquired simultaneously from the same region. The embryo was placed in 1 × phosphate buffered saline and fixed for 2 h in 4% paraformaldehyde (PFA) at 4°C. The embryo preparation procedure is detailed in previous work [18], except the embryos were only incubated in anti-platelet endothelial cell adhesion molecule-1 (PECAM1) antibody (1:200, Mec13.3, BD Pharmingen) diluted in blocking solution overnight at 4°C.

In general, OCT imaging of embryos requires minimal preparation. An embryo can be placed in a culture dish, submerged in media, and imaged in 3D using a pair of galvanometer-mounted mirrors [8]. In our case, scanning with a second mirror would change the distance between the light sheet and objective lens, resulting in a blurred image since the excitation lightsheet would move out of the detection objective focus. Our platform mitigates this issue by moving the sample instead. Although a second scanner is integrated into the system, it is used only for precise alignment and is otherwise stationary during imaging. Moreover, the refractive index between the detection objective and the sample must be matched. Thus, the sample was submerged in water and imaged with a water-dipping objective. Another challenge was unwanted motions caused by the motorized stage stepping. We addressed this issue by embedding the embryo in the same agarose as the microspheres and glued the agarose to the glass cuvette. This way, the embryo was fixed in place with respect to the translation stage. Development of a custom embryo holder is underway for live imaging.

Figure 3 shows the co-registered LSFM and OCT images. Here, 600 images were acquired with a step size of 5 μm. The power incident on the sample from the LSFM beam was approximately 18 mW, and the camera exposure time was 120 ms per image. This was spread over the entire lightsheet, which scanned not only the LSFM FOV (1.8 mm × 1.8 mm), but also the OCT FOV (4.7 mm × 2.5 mm). Therefore only 38% of the illuminated sheet was actually imaged by the LSFM system. The LSFM and OCT images were acquired simultaneously using a home made LabVIEW (NI, Austin, TX) interface, processed using Matlab R2020b (MathWorks, Inc., Natick, MA), and co-registered and displayed in Amira (Thermo Fisher Scientific, Waltham, MA).

Fig. 3.

(a) Co-registered OCT and LSFM 3D image of an E10.5 embryo taken by the combined LSFM-OCT system. (b) Single plane of the 3D data ~130 μm inside the embryo. (c) Zoomed-in view of the LSFM image of erythroblasts shown in an orange–yellow map. The vasculature and individual cells are recognizable.

In the orientation shown in Fig. 3(a), the embryo has dimensions of ~3.5 mm by 4.5 mm. Imaging was performed from the right side of the embryo in Fig. 3(a). The OCT penetration depth was sufficient to image the entire embryo at this stage. The OCT image is shown in gray, and the red blood cell progenitors are shown in orange. The LSFM data (shown in orange) is only a fraction of the OCT image (gray color), because the LSFM detection objective has a much higher NA (0.8) with a much smaller FOV. Figure 3(b) shows a single plane of the 3D data ~130 μm inside the embryo from the same region. To portray the capabilities of the LSFM system, Fig. 3(c) presents a zoomed in view of the LSFM FOV. Even though only the erythroblasts were imaged, the vasculature is also visible from the organization of the erythroblasts. However, some out of focus light was captured, which was due to the curved shape of the embryo, resulting in the blurring around the periphery in Fig. 3(c).

Our LSFM + OCT integrated system enables acquisition of high-resolution multimodal optical images of the same plane acquired at the same time in murine embryos. By acquiring images from the same plane, we were able to see both LSFM and OCT features in real time and search for specific targets. Co-registration was trivial and did not require computationally complex methods since both imaging systems are optically coplanar. However, there were some limitations. For example, the LSFM detection microscope objective limited the FOV of the LSFM images, but one of the next steps of our work is to incorporate mosaicing to increase the FOV. The 3D imaging speed was limited by camera exposure and stage stepping, since OCT imaging is much faster than LSFM acquisition. Penetration of the fluorescence excitation was limited in tissue, but could be further enhanced with multi-photon imaging, which is the next step of our work. The incident power of both systems is compromised because of the beamsplitter cube. However, the LSFM excitation power can be increased, and the OCT system sensitivity was 100 dB, which is sufficient for embryo imaging. The embryo imaged in this work was fixed, but imaging live transgenic embryos is underway to understand the development of the umbilical artery and vein and the impact of any associated pathologies on the embryo development.

Data Availability.

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.