EASI-FISH Spot Counting in Drosophila Brain Using RS-FISH v1

Claire Managan, Gudrun Ihrke2023 · DOI

Abstract

This protocol describes how to utilize the Radial Symmetry-FISH (RS-FISH) software for fluorescent spot detection and counting (Bahry, E., Breimann, L., Zouinkhi, M. et al., 2022) to analyze images of the adult Drosophila brain labeled by Expansion-Assisted Iterative Fluorescence In Situ Hybridization (EASI-FISH) (Wang, Y., Eddison, M., Fleishman, G. et al., 2021). In expansion microscopy, tissues are embedded and expanded isotropically in a swellable hydrogel to achieve a higher spatial resolution (Chen, F., Tillberg, P.W. & Boyden, E.S., 2015). EASI-FISH combines this method with RNA fluorescence in situ hybridization, utilizing the hybridization chain reaction (HCR). In images taken by light sheet or confocal microscopy, individual transcripts are detected as bright spots. These spots are sufficiently separated from each other after 2x or 3x expansion of the brains to perform spot counting, unless the gene in question is too highly expressed. To use this protocol, RS-FISH should be installed as a Fiji plugin according to the instructions detailed on the RS-FISH GitHub repository. The cell body masks were produced in Amira using the techniques described in the "Using Amira to generate cell body masks in fluorescent light sheet microscopy images of EASI-FISH samples" protocol.

Introduction

Mapping gene expression at the single-molecule level within an intact brain is a monumental challenge. In standard microscopy, RNA transcripts often appear as a crowded, blurry mess because they are packed closer together than the diffraction limit of light allows.

While Expansion Microscopy (ExM) solves physical crowding by embedding tissue in a hydrogel and physically expanding it (usually 2x or 3x), it creates a new data challenge: the resulting images are massive, 3D, and contain millions of points of interest. Researchers need a robust, scalable, and precise way to distinguish true biological signals from background noise and "halos" across these large volumes. This protocol addresses the need for a reproducible computational pipeline to turn those raw images into clean, cell-specific transcript counts.

Methods

The protocol utilizes a suite of open-source and specialized tools, primarily centered around Fiji (ImageJ) and the RS-FISH plugin. The workflow consists of several critical stages:

  1. Preprocessing: Images are cropped to specific regions of interest (around cell bodies) to save computational power. Background noise is mitigated using a "Rolling Ball" subtraction method.
  2. Anisotropy Correction: Because microscopy often stretches images along the Z-axis, spots can appear as ovals rather than spheres. The protocol calculates an "anisotropy coefficient" to mathematically "round out" the spots, ensuring the detection algorithm works accurately in 3D.
  3. Radial Symmetry-FISH (RS-FISH): This is the core algorithm. Unlike simple thresholding, it uses radial symmetry to find the center of fluorescent spots with sub-pixel precision. It employs RANSAC (Random Sample Consensus), an iterative method to ignore outliers and noise during the spot-fitting process.
  4. Intensity Thresholding: Users interact with a histogram of spot intensities to manually define the cutoff between real RNA molecules and background artifacts.
  5. Mask Filtering: To assign transcripts to specific cells, the protocol uses "cell body masks" (previously generated in software like Amira). This filters the global spot list down to only those contained within the boundaries of a specific neuron.
  6. Verification: The results are visualized using the Big Data Viewer to ensure the detected spots align perfectly with the original image data.

Key Insights

  • The Power of Expansion: The protocol emphasizes that 2x or 3x physical expansion is the "sweet spot" that allows individual transcripts to be separated enough for automated counting without losing tissue integrity.
  • Anisotropy is Non-Negotiable: A major technical requirement in the protocol is the calculation of the anisotropy coefficient. By ensuring spots are treated as spheres, the software avoids double-counting or missing spots that are skewed by the microscope's optics.
  • RANSAC for Robustness: By using RANSAC within the RS-FISH plugin, the workflow becomes much more resilient to the "halos" and non-specific staining that often plague FISH experiments.
  • Interactive Refinement: The protocol doesn't rely on a "black box" AI; it encourages researchers to use intensity histograms to manually validate the threshold, combining human intuition with algorithmic speed.

Results

The primary output of this protocol is a CSV file containing the precise 3D coordinates (X, Y, Z) of every detected RNA transcript within a specific cell or region.

  • Spatial Precision: The method allows for sub-pixel localization of transcripts.
  • Scalability: It provides a path to move from looking at a single image to quantifying gene expression across entire brain regions.
  • Comparability: By standardizing the "Sigma" and "Threshold" parameters, researchers can ensure that spot counts are comparable across different samples and experimental days.

Limitations and Future Work

  • High Expression Limits: The authors note that if a gene is too highly expressed, the spots will still overlap even after 3x expansion, making individual counting impossible.
  • Manual Steps: While the spot detection is automated, the initial cropping and final thresholding still require human intervention, which could be a bottleneck for massive, whole-brain datasets.
  • Software Dependencies: The workflow requires a specific ecosystem (Fiji, RS-FISH, and Amira for masking). Future iterations might benefit from a more unified, end-to-end software solution.
  • Masking Accuracy: The final counts are only as good as the cell body masks produced in Amira; any errors in defining the cell boundary will lead to incorrect transcript assignments.

Conclusion

This protocol provides a standardized workflow for quantifying individual RNA transcripts within the complex architecture of the adult Drosophila brain. By combining EASI-FISH with RS-FISH software, researchers can accurately detect and count millions of molecular "spots" in 3D space with high spatial resolution, enabling high-throughput transcriptomic mapping in intact tissues.

Summary updated

Abstract

This protocol describes how to utilize the Radial Symmetry-FISH (RS-FISH) software for fluorescent spot detection and counting (Bahry, E., Breimann, L., Zouinkhi, M.et al., 2022) to analyze images of the adult Drosophila brain labeled by Expansion-Assisted Iterative Fluorescence In Situ Hybridization (EASI-FISH) (Wang, Y., Eddison, M., Fleishman, G. et al., 2021). In expansion microscopy, tissues are embedded and expanded isotropically in a swellable hydrogel to achieve a higher spatial resolution (Chen, F., Tillberg, P.W. & Boyden, E.S., 2015). EASI-FISH combines this method with RNA fluorescence in situ hybridization, utilizing the hybridization chain reaction (HCR). In images taken by light sheet or confocal microscopy, individual transcripts are detected as bright spots. These spots are sufficiently separated from each other after 2x or 3x expansion of the brains to perform spot counting, unless the gene is question is too highly expressed. To use this protocol, RS-FISH should be installed as a Fiji plugin according to the instructions detailed on the RS-FISH GitHub repository. The cell body masks were produced in Amira using the techniques described in the "Using Amira to generate cell body masks in fluorescent light sheet microscopy images of EASI-FISH samples" protocol.
BibTeX
@misc{managan2023easi,
  author = {Claire Managan and Gudrun Ihrke},
  title = {{EASI-FISH Spot Counting in Drosophila Brain Using RS-FISH v1}},
  year = {2023},
  doi = {10.17504/protocols.io.ewov1opz7lr2/v1},
}
Figure