Fluorescence-based Single-cell Analysis of Whole-mount-stained and Cleared Microtissues and Organoids for High Throughput Screening Automatable organoid clearing and high-content analysis workflow and timeline

[Abstract] Three-dimensional (3D) cell culture, especially in the form of organ-like microtissues (“organoids”), has emerged as a novel tool potentially mimicking human tissue biology more closely than standard two-dimensional culture. Typically, tissue sectioning is the standard method for immunohistochemical analysis. However, it removes cells from their native niche and can result in the loss of 3D context during analyses. Automated workflows require parallel processing and analysis of hundreds to thousands of samples, and sectioning is mechanically complex, time-intensive, and thus less suited for automated workflows. Here, we present a simple protocol for combined whole-mount immunostaining, tissue-clearing, and optical analysis of large-scale (approx. 1 mm) 3D tissues with single-cell level resolution. While the protocol can be performed manually, it was specifically designed to be compatible with high-throughput applications and automated liquid handling systems. This approach is freely scalable and allows parallel automated processing of large sample numbers in standard labware. We have successfully applied the protocol to human mid- and forebrain organoids, but, in principle, the workflow is suitable for a variety of 3D tissue samples to facilitate the phenotypic discovery of cellular behaviors in 3D cell culture-based high-throughput screens. Graphic abstract:

[Background] Over the past years, three-dimensional (3D) cell culture systems, in particular stem cellbased organoids, have enabled novel insights into human biology and disease (reviewed by: Rossi et al., 2018; Schutgens and Clevers, 2020; Kim et al., 2020). However, working with 3D models requires more than new cell culture techniques. Quantifying cells in complex 3D tissues in fast, unbiased, and efficient workflows requires adapting the analyses as well. So far, many studies have relied on tissue sectioning followed by immunostaining to analyze the structures and composition of their 3D aggregates in detail (Lancaster et al., 2013;Pasca et al., 2015). While this method has been invaluable, it requires ample manual intervention, is time-intensive and cumbersome, and thus not ideally suited for largescale screening applications, including drug development campaigns. Moreover, sectioning provides only a view at a subset of sample tissue, often results in a loss of spatial information unless meticulous serial sections are prepared, and is challenging to use as a basis for 3D reconstruction.
More recently, the combination of whole-mount staining and tissue clearing allowed the analysis of entire 3D aggregates without the need for sectioning, as demonstrated by our group and others (Masselink et al., 2019;Dekkers et al., 2019;Renner et al., 2020). This approach allows rapid 3D acquisition of complex samples and preserves tissue context while providing single-cell-specific phenotypic information. Our workflow utilizes a customized immunostaining procedure, which we adapted and optimized for organoids based on a previous protocol (Lee et al., 2016). We combined this with benzyl alcohol benzyl benzoate (BABB)-based clearing (Dent et al., 1989), which clears human neural tissues quickly (minutes) and effectively (Renner et al., 2020). We specifically designed this procedure to be compatible with both manual pipetting and automated liquid handling systems, facilitating low-and high-throughput applications. Depending on the individual requirements, the samples can then be analyzed using a standard confocal microscope or a high-content imaging system.
In either case, the resulting optical tissue cross sections enable the quantitative analysis of entire 3D samples down to the single-cell level. This eliminates the need to freeze/section the organoids and enables parallel staining and analysis of many samples in 96-or 384-well formats. Though we developed the workflow for the analysis of neural organoids, it is not restricted to a particular organoid system or sample type but can be applied to any 3D sample of interest.   6. Important: Transfer the samples in 100% methanol from standard tissue culture plates to BABBresistant plates (e.g., "Screenstar" COC; BABB/methanol and BABB dissolve most standard tissue culture plastic within 30 min). For transfer, cut off the tip of a 1,000 μl pipette tip with ethanol-cleaned scissors. Aim to cut the conical tip about 5 mm from the opening at the narrow conical end (you can cut off more if your samples are large). This is designed to widen the pipet opening to avoid damaging large 3D samples; you can use wide-bore tips for a liquid handler. 7. Aspirate the 100% methanol and add 150 µl BABB/methanol (1:1, (v/v), see Recipes, Table 3) and incubate for 30 min at RT. BABB consists of benzyl alcohol/benzyl benzoate at a 1:1 (v/v) mixture (see Recipes, Table 4).

D. Fluorescence-based single-cell analysis
High-throughput applications, including compound screening, must process and treat a large number of samples with fast readouts. The combination of our whole-mount immunostaining and clearing workflows outlined above with high-content confocal imaging enables the automated acquisition of entire 96-well plates in a screening-compatible manner. We have described the analysis workflows necessary to analyze this kind of high-content data in detail (Renner et al., 2020).
However, not all labs have access to the often highly specific hard-and software required for these kinds of analyses. Here, we detail a workflow using only standard confocal microscopes and freeware. As every 3D sample and staining has specific requirements for image analysis, we do not focus on specific parameters but rather on the general procedure, allowing anyone to apply the workflow to their own work.
D0. Acquire images, either single optical confocal slices or entire stacks, using a confocal microscope. Depending on your analysis needs, you may prefer to image at 8, 16, or 32-bit depth with lens magnifications that suit your samples and desired lateral resolution. For our needs, we imaged whole organoids with a 10× lens magnification and 16-bit dynamic range (see Figure 1 for representative images). Adjust the distance between successive Z-planes to your analysis needs.
A full 3D reconstruction will require more numerous and more closely spaced Z-planes. For quantitative comparison between samples, it is best to space the individual Z-planes further apart along the Z-axis, thus undersampling your 3D volume. This avoids data artifacts from doublecounting the same physical feature in adjacent Z-planes and reduces the amount of data acquired and the acquisition times. For example, to measure a representative number of nuclei, Z-planes should be placed apart further than twice the average diameter of a nucleus. In general, empirically choose spatial increments between focal planes so that you can prevent imaging the same sample features in adjacent Z-planes, preventing double-counting. Figure 1A)

D1. Analysis of nuclear markers (example Sox2, see
Note: We describe all analysis procedures using Fiji/ImageJ (Schindelin et al., 2012) as it is freely available and commonly used for processing and analyzing biological imaging data (the specific commands used for the analysis here and the path to find them are described at each step below). However, the same general steps can also be performed with similar software using similar steps and parameters. Here, our raw images consisted of 16-bit grayscale data and measured 1024 pixels by 1024 pixels. Parameters below are adjusted for these conditions. Other starting points likely require the adjustment of downstream image analysis parameters. c. Enable "select all" in ROI manager to show ROIs overlaid on the original image (if too many ROIs are selected, deselecting the checkbox marked "Labels" makes the image less crowded). d. Click "measure" in the ROI manager window to either measure single selected ROIs or all ROIs if none is selected.

D2. Analysis of cytoplasmic/filamentous markers (example TH, see Figure 1B)
For cytoplasmic/filamentous markers, clean segmentation of single cells can be challenging, especially in a dense 3D environment where they can span across several z-levels as thin cellular projections. Thus, it is often preferable to measure either the integrated or average intensity of positively identified structures of interest on every confocal plane. The signal for the whole organoid can then be summed for every plane within the organoid and quantitatively reflects the presence of each marker of interest contained in the 3D tissue.
1. Steps 1-5 can be repeated as described in the Sox2 nuclear analysis above.
2. Instead of separating the objects by watershedding, continue to define them as ROIs via the "analyze particles" function.
3. Use the "ROI manager" to apply the previously defined ROIs to the original image and measure the intensity with the "measure" function (after adjusting "set measurements" to the sample's specific requirements). Figure 1C for an example)

D3. Measurement of the sample area for normalization (see
Measuring 3D aggregates with varying diameters in different z-levels often requires normalization of the data to the total sample area to obtain comparable results between different samples. 2. Use thresholding to create a single object from the sample, separating background and sample. 3. Use the "measure" function to calculate the area of the object (after enabling "area" under "set measurements").
These procedures can be easily applied to entire aggregates by recording the workflow (with the parameters for the specific requirements of the samples) as a macro (e.g., via the "macro recorder").
These can iteratively process entire Z-stacks or even groups of Z-stacks contained in a common folder. (10% (v/v) in PBS) and sodium azide (10% (w/v) in PBS) as facilitates handling.