Combined iDISCO and CUBIC tissue clearing and lightsheet microscopy for in toto analysis of the adult mouse ovary^†

Abstract

At any given time, the ovary contains a number of follicles in distinct growth stages, each with a set of identifying characteristics. Although follicle counting and staging using histological stains on paraffin-embedded ovary sections has been the gold standard in assessing ovarian health in fertility studies, the final counts rely on extrapolation factors that diverge greatly among studies. These methods also limit our ability to investigate spatial aspects of ovary organization. Recent advances in optical tissue clearing and lightsheet microscopy have permitted comprehensive analysis of intact tissues. In this study, we set out to determine the best clearing and imaging methods to generate 3D images of the complete adult mouse ovary that could be used for accurate assessments of ovarian follicles. We found that a combination of iDISCO and CUBIC was the best method to clear the immunostained ovary. Using lightsheet microscopy, we generated 3D images of the intact ovary and performed qualitative assessments of follicles at all stages of development. This study is an important step toward developing quantitative computational models that allow rapid and accurate assessments of growing and quiescent primordial follicles, and to investigate the integrity of extrinsic ovarian components including vascular and neuronal networks.

The combination of iDISCO and CUBIC tissue clearing methods allows in toto imaging of the ovarian follicle composition and extrinsic vascular and neuronal networks.

Introduction

At any given time during the reproductive lifespan of adult females, ovaries typically contain a number of follicles in each stage of growth. These include primordial follicles, which are quiescent and represent the ovarian reserve, primary follicles, which have just become activated and initiated growth, secondary and antral follicles, which represent steps in follicle maturation, Graafian follicles, which represent the stage just prior to ovulation, and corpora lutea, which are postovulatory follicles [1,2]. As the female ages, sequential activation of ovarian follicles eventually leads to the depletion of the ovarian reserve, which results in menopause and infertility [3,4]. Thus, the distribution of ovarian follicles within the ovary is a reliable marker of the health and fertility of the ovary, and follicle scoring and classification has become the gold standard in assessing fertility during ovarian biology studies. Current methods to count ovarian follicles rely mostly on manual counting and staging of each ovarian follicle using histological staining of paraffin-embedded ovary sections [5,6]. Although these methods have been essential to the advancement of the field of ovarian biology and fertility, they are extremely time-consuming and labor-intensive. In addition, classification criteria and extrapolation methods vary greatly among studies [5,7]. Although some efforts have been made to provide high throughput classification of ovarian follicles using machine learning-assisted feature segmentation [8,9], these studies relied on histological sections of the ovary. One way to avoid biases and speed up the process of scoring ovarian follicles is to determine a method that does not rely on histological sections but rather on in toto images of intact ovaries. The opacity of the adult ovary has previously made this difficult to implement, but with the recent expansion of optical clearing methods and microscopy techniques to image large samples, this strategy is now within our reach [10–13]. There are many optical clearing methods, but they rely on a common principle, which is to remove the molecules within biological tissues that scatter light and make tissues opaque, and replace these with a solution of constant refractive index, which strongly decreases light scatter and efficiently renders large biological samples transparent [14,15]. Our motivation in adapting these clearing methods for in toto imaging of the ovary was to provide the field of ovarian biology with a more reproducible and standardized method for assessing the state of growing follicles and the ovarian reserve in intact ovaries. The choice of the right tissue clearing method depends on the desired application, the nature of the biological tissue, and the availability of adapted microscopes. The intact adult ovary is difficult to clear, in part because of the amount of connective tissue that resides within the ovarian medulla. Another limitation is the lack of fluorescent reporter lines that label ovarian follicle cells. Thus, we set out to determine the optimal clearing method that would be compatible with fluorescent labeling of ovarian components using antibodies. Although recent studies have shown that the BABB [16] and CLARITY [17] clearing methods can successfully clear mouse ovaries, these did not provide reliable antibody labeling of both oocytes and granulosa cells, which is essential for accurate staging of follicles. Here, we describe the successful adaptation of a combined iDISCO [18] and CUBIC [19] tissue clearing method that is compatible with fluorescent immunohistochemistry to efficiently clear and label the adult mouse ovary for in toto imaging using lightsheet microscopy. We also provide an alternative imaging method using spinning disk confocal microscopy.

Materials and methods

Mice

Unless otherwise stated, experiments were performed on wild-type C57BL/6 2-month old mice. The 129S.FVB-Tg (Amh-cre)8815Reb/J (Amh-Cre) mouse line was purchased from Jackson laboratory (Jax stock #007915) and maintained on the 129S1/SvImJ background. The Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J (Rosa-TdTm) line was purchased from Jackson Laboratory (Jax stock #007914) and maintained on the C57BL/6 background. To obtain mice with tomato fluorescence in growing granulosa cells, AmhCre^+/− males were crossed to Rosa-TdTm^fl/fl females. The resulting mice were on a mixed B6 (C57BL/6J), 129SVJ (129S1/SvImJ) background.

Mice were housed in a vivarium maintained at 72 °F with a light/dark cycle of 12 h dark/12 h light, with a humidity set point at 45%. All mice were housed in accordance with National Institutes of Health guidelines, and experiments were conducted with the approval of the Duke University Medical Center Institutional Animal Care and Use Committee.

Combined optical clearing and fluorescent immunostaining using CUBIC

For schematic overview of the experimental workflow, see Figure 1A. For detailed experimental procedures and solution recipes, see Supplemental protocol 1. Briefly, ovaries from 2-month-old female mice were dissected in PBS (Day 0), fixed for 1 h at room temperature (RT) in 4% PFA/PBS and gradually dehydrated into 100% methanol for storage at −20 °C. When ready for analysis (Day 1), samples were processed for CUBIC clearing and whole-mount immunofluorescent staining as follows. After gradual rehydration into PBS, tissues were incubated in a 1:1 solution of CUBIC Reagent 1 (CUBIC R1) and dH₂O for 6 h at 37 °C. Samples were then transferred into 100% CUBIC R1 and incubated at 37 °C for 72 h. On Day 4, following three 30-min washes in PBS, samples were bleached in chilled 5% H₂O₂ in PBS for 1 h at RT. Samples were then permeabilized in PBS 2% TritonX-100 for 3 h at RT and then transferred into blocking solution (PBS 1% TritonX-100, 10% horse serum) overnight at 37 °C. On Day 5, samples were incubated with primary antibodies diluted in blocking solution for 72 h at 37 °C. On Day 8, Following three 1 h-washes in PBS 1% TritonX-100, tissues were incubated with secondary antibodies diluted in blocking solution for 48 h at 37 °C. On Day 10, tissues were washed 3 times in PBS 1% TritonX-100 for 1 h each and first transferred into a 1:1 solution of CUBIC Reagent 2 (CUBIC R2) and PBS for 6 h to overnight at 37 °C, and then transferred into 100% CUBIC R2 until they appeared sufficiently clear. Information on the primary and secondary antibodies used in this study is listed in Supplemental Tables 1 and 2.

Figure 1.

Optimized CUBIC clearing for in toto imaging of the adult mouse ovary. (A) Schematic diagram of the experimental workflow for optical clearing and fluorescent immunostaining of intact mouse ovaries using the CUBIC method. (B) Photomicrograph of uncleared (left) versus CUBIC-cleared (right) 2-month-old ovaries (scale bar, 1 mm). (C) Epifluorescence images of uncleared (left) versus CUBIC-cleared (right) ovaries labeled with antibodies against oocyte marker MSY2 (scale bars, 300 μm). (D) Optical sections from lightsheet fluorescence microscopy (LSFM) images of CUBIC-cleared ovary from an Amh-Cre; Rosa-TdTm 2-month-old mouse labeled with endogenous Tomato fluorescence in red (red, left), Hoechst nuclear dye to visualize all cells (gray, middle), and antibodies against the vascular marker ENDOMUCIN (green, right). Insets are magnified views from the areas outlined in top row images. (scale bars, 300 μm). (E) 3D render of LSFM images of CUBIC-cleared ovary shown in D (scale bars, 300 μm). White arrows point to ovulation sites.

Figure 4.

Qualitative assessment of the ovarian reserve using iDISCO + CUBIC clearing and lightsheet microscopy. (A) 3D renders from LSFM images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against growing granulosa marker AMH (green, left), oocyte marker MVH (red, middle), and oocyte marker HuC/D (cyan, right). Far right is a merged 3D render of AMH and MVH stains. Insets in bottom row are magnified views of the areas outlined in top row images (scale bars, 400 μm). (B) Magnified views of 3D renders from LSFM images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against granulosa marker AMH (green, left) and oocyte marker HuC/D (red, right) (scale bars, 400 μm). Bottom row images are 3D surfaces generated in Imaris software using top row images of growing follicles (left) and oocytes (right). Surfaces are spectrum-coded by volume.

Combined optical clearing and fluorescent immunostaining using iDISCO + CUBIC

For schematic overview of the experimental workflow, see Figure 2A. For detailed experimental procedure and solution recipes, see Supplemental protocol 2. Briefly, ovaries from 2-month-old female mice were dissected in PBS (Day 0), fixed for 1 h at RT in 4% PFA/PBS, and gradually dehydrated into 100% methanol for storage at −20 °C. When ready for analysis (Day 1), samples were incubated overnight in 33% methanol, 66% dichloromethane at RT. Samples were then treated with a solution of 5% H₂O₂/methanol overnight a 4 °C. On Day 3, after progressive rehydration to PBS 0.2% Triton X-100, (PTx.2) samples were permeabilized overnight at 37 °C in PTx.2 2.3% glycine, 20% DMSO and blocked for 6 h at 37 °C in PTx.2 10% DMSO, 3% horse serum solution. On Day 4, tissues were incubated in primary antibodies diluted in PTwH (PTx.2 with 0.001% heparin) 3% horse serum, 10% DMSO for 72 h at 37 °C. On Day 7, samples were washed three times for 1 h in PTwH and incubated for 48 h in secondary antibodies diluted in PTwH 3% horse serum at 37 °C. On Day 9, after washing three times for 1 h in PtWH, samples were progressively dehydrated into 100% methanol overnight and on Day 10 were incubated for 3 h in 33%methanol, 66% dichloromethane at RT. Following two 15-min washes in dichloromethane, samples were cleared in dibenzylether (DBE). Samples were left overnight in DBE to allow for sufficient clearing. On Day 11, samples were transferred to 100% methanol and progressively rehydrated to PBS and recleared following the CUBIC clearing protocol (Day 11–Day 15).

Figure 6.

The combined iDISCO and CUBIC method is applicable for imaging with spinning disk confocal microscopy. (A) Schematic diagram of the experimental workflow for optical clearing and fluorescent immunostaining of intact mouse ovaries using the combination of iDISCO and CUBIC methods and processed for confocal imaging. The whole ovary was mounted in a 3D-printed coverslip holder designed for imaging the same sample from each side. (B) Optical sections from top (left), middle (middle), and bottom (right) z-planes of spinning disk confocal images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against oocyte marker MVH (green) and growing follicle marker AMH (red). A 3D-printed coverslip holder was used to image the same ovary from both sides (side A and side B). Tiling junctions are improved but still evident (scale bars, 500 μm). (C) 3D renders generated from spinning disk confocal images of iDISCO + CUBIC-cleared ovary shown in B. The insets are magnified split-channel and merged views of the area outlined in the top image, SIDE A (scale bars, 500 μm).

Figure 2.

Combined iDISCO and CUBIC clearing for in toto imaging of the adult mouse ovary. (A) Schematic diagram of the experimental workflow for optical clearing and fluorescent immunostaining of intact mouse ovaries using the combination of iDISCO and CUBIC methods. (B) Photograph of ovaries cleared following the iDISCO protocol in DBE, visualized with halogen (top) or UV (bottom) light. White arrows point to cleared ovaries. (C) Photomicrograph of uncleared (top) versus iDISCO + CUBIC-cleared (bottom) 2-month-old ovaries (scale bar, 1 mm). (D) Optical sections from the surface (left) and center (right) z-planes of LSFM images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against interstitial marker αSMA (green) and growing follicle marker AMH (red). Insets are magnified views from the areas outlined in top row images (scale bars, 300 μm). (E) 3D render generated from LSFM images of iDISCO + CUBIC-cleared ovary shown in D. The inset is a magnified view of a single growing follicle from the area outlined in the top row image (scale bar, 50 μm).

Image acquisition

Once sufficiently cleared, the samples were placed in CUBIC R2 containing Hoechst nuclear dye diluted 1:1000. The next day, samples were embedded in 1.8% agar/PBS in 1-mL syringes and stored in the dark at RT until imaged. Samples were imaged in CUBIC R2 using the Zeiss Lightsheet Z.1 microscope (Carl Zeiss, Inc., Germany). Lightsheet images were collected using either a 20x NA 1.0 Clarity objective (RI = 1.45) or a 5x NA 0.16 dry objective. Image acquisition was performed with Zen software. To image intact ovaries with the spinning disk confocal microscope (Andor Dragonfly, Oxford Instruments, UK), ovaries were mounted in 1.8% agar/CUBIC R2 onto a 3D-printed coverslip holder that can be flipped and imaged from both sides [23]. For imaging ovaries cleared with only iDISCO, samples were mounted onto the 3D-printed coverslip holder in a well of vacuum grease filled with diBenzyl Ether. The coverslips and coverslip holder were carefully sealed with nail polish to avoid the very corrosive DBE leaking onto the microscope. The 3D model for these coverslip holders is available at NIH 3D print exchange at https://3dprint.nih.gov/discover/3DPX-009765. Images were collected with a 10x NA 0.45 dry objective on a Leica DMi8 microscope stand attached to an Andor Dragonfly 505 unit using the 25-μm pinhole disk, 1x magnification relay lens, and Andor iXon 888 EMCCD camera. Image acquisition and multiple field of view tiling of images were performed with Andor Fusion software version 3.1. Videos of full stacks, 3D projections, and 3D surfaces were rendered in Imaris imaging software (Version 3.1; Bitplane, Inc., United Kingdom).

Results and discussion

CUBIC is not sufficient to clear the intact immunostained adult mouse ovary and does not result in penetration of antibodies to the center of the tissue

Several tissue clearing methods had already been described at the time we initiated this study, and the number of available methods has continued to increase [11]. However, there were several constraints that guided our selection of clearing methods for this project. First, due to the lack of fluorescent reporter lines specific to ovarian cells, we required a clearing method compatible with antibody penetration and fluorescent immunostaining. In addition, our final objective was to image the cleared ovaries in toto using the Zeiss Z.1 Lightsheet microscope, which is only compatible with aqueous imaging solutions. Taking these limitations into account, we tested several methods, including AbScale and Scale S [20] ACT-Presto using X-Clarity [21], and CUBIC [19]. With these initial trials, we determined that AbScale and X-Clarity were incompatible with our antibody stains and Scale S did not succeed in clearing the ovary (data not shown). The most promising method was CUBIC, as the ovaries were successfully cleared after only 7 days in the CUBIC solutions (Figure 1A and B). In addition, epifluorescent images of the surface of the ovary showed that oocyte labeling using antibodies against MSY2 in CUBIC-cleared ovaries was comparable to uncleared ovaries, demonstrating that the CUBIC protocol was compatible with antibody staining of the ovary (Figure 1C). We thus chose to optimize the CUBIC protocol, by modification of several steps of the original protocol. We found that a short fixation (1 h) in 4% PFA at RT followed by dehydration in a methanol gradient and storage at −20 °C in 100% methanol was ideal for preserving the integrity of the tissue and antigens while limiting autofluorescence. In addition, we incubated the ovaries in hydrogen peroxide (5% H₂O₂ in PBS) for 1 h at RT prior to antibody staining to avoid autofluorescence originating from red blood cells (Figure 1A). We also found that pretreating the tissue with CUBIC Reagent-1 (CUBIC R1) prior to immunostaining at 37 °C followed by incubation in CUBIC Reagent-2 (CUBIC R2) was the most efficient method in terms of antibody penetration and clearing quality (Figure 1A and Supplemental Protocol 1).

We next determined whether CUBIC-treated ovaries were sufficiently cleared and immunolabeled for in toto imaging of the ovary using lightsheet microscopy. For this purpose, we first tested ovaries from a 2-month-old AmhCre; Rosa-TdTm reporter mouse in which anti-Mullerian hormone (Amh)-expressing granulosa cells were labeled with endogenous Tomato fluorescence. While images of the Tomato signal confirmed that endogenous reporters survived our CUBIC protocol (Figure 1D), Hoechst DNA labeling showed that the ovary was not fully cleared. Furthermore, immunofluorescent staining of the ovarian vasculature using antibodies against ENDOMUCIN demonstrated that only the surface of the ovary was permeable to antibodies (Figure 1D). Nonetheless, we believe this method can be applied for qualitative assessments and visual representations of the intact adult mouse ovary, as demonstrated by the 3D-rendered image of the cleared and immunostained AmhCre; Rosa-TdTm ovary (Figure 1E and Movie S1).

Despite the steps to optimize the combination of immunofluorescence and CUBIC clearing, we concluded that CUBIC clearing was not sufficient to allow the penetration of antibodies to the center of the ovary for lightsheet imaging and quantitative in toto analysis of adult ovaries.

Combining iDISCO and CUBIC facilitates adult mouse ovary clearing and antibody penetration

Of all the clearing methods available in the literature, the one that seemed to reliably produce high-quality images of cleared and immunostained samples was iDISCO+ [18]. We first confirmed that this method was compatible with immunostaining on intact mouse ovaries and produced images using a spinning disk confocal microscope (Figure S1 A and B). These samples had the desired clarity and penetration of antibodies; however, the iDISCO+ clearing method is solvent-based and the final clearing agent, DBE, which is used for imaging, is a strong corrosive. This made the iDISCO+ method incompatible with the Zeiss Z.1 Lightsheet microscope. Since Lightsheet microscopy is at least four times faster than spinning disk confocal microscopy and allows high-throughput imaging of large intact volumes, we sought a way to solve this problem. We investigated whether we could add a CUBIC clearing step to the iDISCO+ method to transition samples to an aqueous solution at the end of the clearing procedure yet retain the desirable iDISCO results in terms of clearing and staining of the adult ovary for in toto lightsheet imaging. We chose to follow the iDISCO+ protocol from beginning to end, then rehydrate the samples through a methanol gradient to PBS, switching over to the CUBIC protocol at that point (Figure 2A and Supplemental Protocol 2). We found that ovaries cleared in iDISCO+ became fully transparent (Figure 2B), but once transferred through a methanol gradient back to PBS, it became opaque again. However, when iDISCO cleared ovaries were transferred from PBS through CUBIC Reagent 1 and Reagent 2, they recleared rapidly and efficiently (Figure 2C).

Comparing iDISCO + CUBIC to CUBIC alone, we found that higher clearing efficiency was associated with better antibody penetration into the tissue and thus higher quality lightsheet images. We immunostained ovaries from 2-month-old wild-type mice with antibodies against the interstitial marker Smooth Muscle Actin alpha (αSMA) and the growing follicle granulosa marker anti-Mullerian hormone (AMH). While samples cleared with only CUBIC failed to produce signal beyond the surface of the tissue (Figure S1 C and D), both labels were detectable throughout the tissue in samples cleared using the combined iDISCO + CUBIC method, as demonstrated by the clarity of the fluorescent signal in optical sections taken from the surface and the center of the lightsheet Z-stack (Figure 2D).

We initially used a 20X acquisition objective to image cleared ovaries. However, because the size of the field of view was smaller than the size of the tissue, this method required tiling—imaging the samples in several adjacent fields, then stitching them together. Higher laser powers were needed to achieve the penetration of the laser illumination through ovarian tissue as well as fast imaging time (thus short camera exposure times). However, at higher laser power, repeated illumination for image tiling caused fluorophore bleaching which made the borders between imaging fields much more apparent in the final stitched images and interfered with the continuity of the images (Figure 1D and E). In addition, tiled image collection required much longer acquisition times and resulted in larger data sets and associated data transfer and storage problems. Thus, for both image quality purposes and for data management and storage, we chose to use a 5X magnification objective, which permitted the imaging of the entire ovary in one frame, while still producing high-quality images with sufficient resolution to study individual ovarian follicles (Figure 2E and Movie S2).

Mounting the adult mouse ovary for in toto imaging

To take full advantage of the rotating sample holder offered by the Zeiss Z.1 Lightsheet microscope, we determined that the optimal way to mount the ovary for imaging was to embed it in a cylinder of 1.8% agar. We initially used glass capillaries that were provided with the Zeiss Z.1 Lightsheet Microscope, but adult ovaries were often too large to mount in the widest capillaries (2.15-mm inner diameter). In addition, these capillaries were difficult and expensive to obtain. Instead, we opted to design our own mounting system using a 1-mL syringe (Figure 3A). Using this embedding method, we identified several problems. First, although it is possible to make 1.8% agar in the CUBIC Reagent-2 solution, the high concentration of sucrose in CUBIC R2 makes the agar difficult to dissolve and handle. To solve this problem, we made the 1.8% agar solution in PBS. Although we found using this method that the samples initially lost their transparency in PBS (Figure 3B), a 2-h RT re-incubation of the agar-mounted samples in CUBIC R2 after embedding led to reclearing (Figure 3C). Although CUBIC R2 can dissolve the agar, causing the samples to drop out of the syringe, this problem can be averted by allowing the 1.8% agar column containing the samples to solidify inside the syringe overnight before incubating in CUBIC R2 (Figure 3B). To facilitate the embedding process, we made a hole in the cap of a 15-mL conical tube to use as a syringe-holder to suspend the sample (Figure 3A). After drawing the samples in 1.8% agar into the syringe, we allowed the agar to solidify overnight, then partially ejected the agar column containing the samples from the syringe into the conical tube, which was filled with CUBIC R2 (Figure 3C). The samples remained in this solution until they were sufficiently cleared and ready for imaging (usually 1–2 h), at which point the syringe containing the samples could be introduced into the rotating sample holder of the Zeiss Z.1 (Figure 3D).

Figure 3.

Mounting the adult mouse ovary for in toto lightsheet imaging. (A) Photograph of modified 1-mL syringe and 15-mL conical tube used for mounting adult ovaries for LSFM imaging using the Zeiss Lightsheet Z.1 microscope. Narrowed end of the syringe is cut off to allow passage of agar-embedded samples, and cap of 15-mL conical tube is carved to hold the syringe. (B) Photograph of 2-month-old mouse ovaries embedded in 1.8% agar/PBS inside 1-mL syringe. Cleared ovaries are visualized using a UV light. (C) Photographs of cleared ovaries embedded in 1.8% agar inside 1-mL syringe inserted into the conical tube syringe holder. The agar including the ovaries is partially ejected into the conical tube, which is filled with CUBIC R2, to ensure reclearing of the ovaries after agar embedding. The cleared ovaries are visualized using a UV light. (D) Photographs of the 1-mL syringe including the cleared ovaries inserted into the Zeiss Lightsheet Z.1 sample holder. Left is a side view of the full montage; top middle is a view from the top showing how the syringe is inserted and locked into the holder (green arrows) and bottom middle is a view of the imaging setup, with samples placed in front of the capture objective (red arrow) and between the lightsheet objectives (yellow arrows). Far right is an image of the full montage with agar-embedded ovaries ejected into the CUBIC R2-filled imaging chamber inside the Zeiss Lightsheet Z.1 microscope. (E) 3D renders of lightsheet microscopy images of a 2-month-old ovary cleared using the combined iDISCO + CUBIC protocol and labeled with Hoechst nuclear dye. Each image is a render of z-stacks from the same sample rotated to capture four different angles: 0°; 90°; 180°; and 270° (scale bars, 400 μm).

Optimized lightsheet microscopy for in toto ovary imaging

Each lightsheet microscope is different, and there are several guides available online and in the literature describing how to set up the lightsheet for optimal imaging [13,22]. We will focus here on the specific settings that we adapted for imaging adult mouse ovaries. As mentioned earlier, we chose to use the 5X dry objective to image the adult ovaries, which allowed us to image the entire ovary in one orthogonal frame while still producing high-quality images (Figure 2E). In addition, using this approach, we were able to take advantage of the multiview component of the Z.1 Lightsheet microscope, which allowed us to image the ovary from multiple angles. Using this method, we took four images of each ovary, each 90° apart (Figure 3E), which provided us with four replicate images, useful for quantitative analysis. It is also important to note that image sharpness decreased 2-mm deep into the tissue. Since adult mouse ovaries average 2.5-mm wide, it was almost always impossible to get a clear image of the entire ovary, and thus the different angles provided more options for accurate data representation (Figure 3E). Another important setting for generating optimal 3D rendering of the ovary is the choice of step size for the Z-stack. One thing to keep in mind is that the interval between slices will contribute to the Z resolution of the 3D-rendered image. This means that once the 3D render is rotated out of the XY plane, the larger the Z-step, the lower the resolution. However, decreasing the step will increase the number of slices, which in turn leads to longer imaging times and larger files. With this in mind, we found that for our analysis, the optimal slice interval for the adult ovary is 1.2 μm. The first images we captured using the 20X objective and tiling took 4 h to image and approximated 700GB for a single view of one ovary. In comparison, with our optimized settings for the 5X capture objective on the Zeiss Z.1 Lightsheet microscope, we were able to produce four-angle, four-channel images of each ovary in an average of 1 h, and the final file size averaged 150GB. It is important before initiating a study that includes intense lightsheet imaging to consider data management and long-term storage.

Qualitative assessment of the ovarian reserve using lightsheet microscopy

Our objective was to design a method that could lead to the rapid and accurate assessment of the ovarian follicle reserve in the adult mouse ovary. We first tested standard antibodies used for labeling ovarian follicles using our combined iDISCO + CUBIC clearing method followed by in toto imaging of the ovary using lightsheet microscopy (Supplementary Table 1). We found that the standard antibody against the granulosa cell marker AMH provided high-quality images of growing follicles in 2-month-old ovaries with very little background (Figure 4A). We also tested the standard antibody against oocyte marker mouse vasa homolog (MVH). We found that MVH labeling in the cleared 2-month-old ovary was of good quality, and oocytes of all classes appeared to be labeled (Figure 4A). However, this antibody displayed a significant level of bright speckled fluorescent background, which could lead to difficulties in computer-assisted quantitative assessments of the follicle reserve. In contrast, the newly described oocyte antibody against HuC/D [23] provided high-quality images of the ovarian oocytes with less background compared with MVH (Figure 4A and Movie S3).

We next demonstrated that it is possible to segment ovarian follicles using 3D surfaces generated using Imaris software (Figure 4B). This provided proof-of-concept for the use of imaging analysis software to generate automated quantitative assessments of the ovarian follicle reserve using 3D images produced by our method. However, we found that the accuracy of the surfaces generated using Imaris software was subpar and could lead to limitations in accurate assessments. This is in part due to the high levels of background fluorescence detected using lightsheet technology, which make it difficult to classify primordial follicles over background speckles. In addition, the spatial distribution of follicles within the ovary leads to errors in image segmentation, where two close growing follicles will often be segmented as one large object. Alternative image segmentation software should be tested to determine whether a better option is available. For example, Faire and colleagues used a combination of Volocity software and MATLAB programs to assess the spatial dynamics of follicle growth [16]. However, we believe that the most powerful computational method for providing accurate automated assessments of the entire ovarian reserve will include a machine learning model. Although this idea has been explored in recent studies [8,9], this option has not, to our knowledge, been investigated using training data sets composed of in toto 3D images of the whole ovary, which, if achieved, will be a significant breakthrough for the field.

Although it was initially thought that oocyte volume alone could allow an accurate classification of quiescent versus growing follicles [16], recent evidence suggests that granulosa cells display the signs of activation prior to oocyte growth [24]. Thus, any high-throughput method developed to classify follicles should include the markers of granulosa growth in addition to oocyte markers. AMH is one such marker, as expression is initiated upon follicle activation. However, the AMH antibody used in this study labeled only follicles in the primary and secondary stages, and the widely used antibody against FOXL2, which labels all granulosa cells in the ovary, did not provide images of analyzable quality, as shown by the superficial and heterogenous labeling of granulosa cells in the 2-month-old ovary (Figure S2). Further research is required to define new antibodies to specifically and efficiently label the granulosa cells of all ovarian follicles.

Imaging vascular and neural networks inside the intact ovary using lightsheet microscopy

Because of their network structure weaving in and out of a single imaging plane or histological section, the dynamics and nature of the vasculature and neurons that invade the ovary have remained elusive. We provide here a method that will allow accurate and reliable mapping of ovarian vasculature and innervation.

Here, we show that our method is applicable to imaging intact vascular and neural networks inside the ovary. We first labeled ovarian vasculature using antibodies against ENDOMUCIN in combination with AMH to label growing follicles in ovaries from 2-month-old mice (Figure 5A). We found that ENDOMUCIN was labeled throughout the ovary, including microvasculature around each follicle (Figure 5A and Movie S4). This method could be extremely helpful in providing an accurate and reliable map of ovarian vasculature and in the assessment of vascular phenotypes in the ovary. Next, we used antibodies against the pan-neuronal marker TUJ1 to label the innervation of the 2-month-old ovary, in combination with antibodies against αSMA, which labels interstitial cells in the ovary (Figure 5B and Movie S5). TUJ1 staining revealed the presence of an intricate network of neural projections in the ovarian medulla that extend outward into the cortex (Figure 5B). We found that TUJ1 staining led to speckled background fluorescence. However, we determined that these speckles could be masked using Imaris-generated surfaces of the TUJ1 channel (Figure 5B). While masking the original imaging data using these computer-generated surfaces could be considered as misrepresentative data adjustments, we believe that methods such as these deserve to be considered in the generation of clean renders of lightsheet images, as long as raw data are provided to confirm that the images provide a fair representation of the biological data.

Figure 5.

Vascular and neural networks can be imaged in toto in the iDISCO + CUBIC cleared ovary using lightsheet microscopy. (A) 3D renders from LSFM images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against granulosa marker AMH (red, left) and vascular marker ENDOMUCIN (green, right). Far right is a merged 3D render of AMH and ENDOMUCIN stains. Insets are magnified views of the areas outlined in top row images. (scale bars, 400 μm). (B) 3D renders from LSFM images of iDISCO + CUBIC-cleared ovary from a 2-month-old mouse labeled with antibodies against interstitial marker αSMA (green, left) and neural marker TUJ1 (red, right). Far right is a merged 3D render of αSMA and TUJ1 stains. Bottom row images are 3D surfaces generated in Imaris software using top row images of ovarian interstitium (red, left) and innervation (green, right).

Qualitative assessment of the ovarian reserve using confocal microscopy

Recognizing that the lightsheet microscope is not yet widely available, we tested whether we could acquire images of adult ovaries cleared using our combined iDISCO + CUBIC method with traditional microscopes (Figure 6A). As mentioned previously, we were able to image intact ovaries cleared with iIDISCO+ using spinning disk microscopy (Figure S1 A and B), which presents the advantage of being faster and less phototoxic than traditional laser scanning confocal. However, this requires mounting and imaging the samples in DBE, which is challenging due to its strong corrosive nature. We thus investigated the possibility of using our combined iDISCO + CUBIC method for imaging intact ovaries with spinning disk microscopy. We cleared ovaries immunostained for AMH and MVH following the combined iDISCO + CUBIC protocol (Supplemental Protocol 1) and mounted them on a 3D-printed coverslip holder that could be flipped to image the same sample from both sides [23]. The ovaries were mounted between two No 1.5 coverslips in a thin layer of 1.8% agar made in CUBIC R2. Once the agar had solidified, we used a spinning disk confocal microscope to generate images from each side of the same ovary. We used a 10X NA 0.45 objective for this experiment, which required us to image each side of the ovary in a set of six tiles that were then stitched together to generate images of the whole ovary. Each image set was generated in under an hour, and the imaging depth (Figure 6B) and quality of 3D-rendered images was comparable to that obtained with lightsheet microscopy (Figure 6C). This experiment demonstrated that this imaging method allows the generation of 3D renders of the entire adult ovary and could be used for automated quantitative assessments of the ovarian reserve. Although we took advantage of the imaging speed and reduced photobleaching provided by spinning disk technology, we believe this strategy is also applicable using a laser scanning confocal microscope. While confocal microscopy is a decent alternative if a lightsheet microscope is not available, lightsheet technology has the advantage of being at least four times faster than spinning disk confocal. In addition, while the spinning disk microscope requires flipping the sample to image the entire volume, the objectives used for lightsheet microscopy have been optimized for longer imaging distances, which, along with the orthogonal single plane illumination, allow imaging entirely through large volumes with reduced photobleaching.

Conclusion

We provide a rapid, efficient, and affordable method to image optically cleared adult mouse ovaries in toto. We hope that the community will develop additional compatible antibodies to label individual structures within the mouse ovary. We believe that this study is an important first step toward developing computational models that will allow rapid and accurate assessments of growing and primordial follicles, and to investigate the integrity of extrinsic ovarian components using 3D images of the intact ovary. In future work, we intend to develop machine learning methods to automate these processes.

Grant support: This work was supported by a grant from the National Institutes of Health (grant # 1R01HD090050-0) to BC; JM was supported by postdoctoral fellowships from the Fondation ARC pour la Recherche contre le Cancer (award # SAE20151203560) and from the American Cancer Society (award # 130426-PF-17-209-01-TBG). Lightsheet microscopy was supported by a grant from the National Institutes of Health (grant # 1S10OD020010-01A1) to the Duke Light Microscopy Core Facility and by a voucher from the Duke University School of Medicine.

Conflict of interest

The authors have declared that no conflict of interest exists.