smiFISH and embryo segmentation for single-cell multi-gene RNA quantification in arthropods - PubMed (original) (raw)
Comparative Study
smiFISH and embryo segmentation for single-cell multi-gene RNA quantification in arthropods
Llilians Calvo et al. Commun Biol. 2021.
Abstract
Recently, advances in fluorescent in-situ hybridization techniques and in imaging technology have enabled visualization and counting of individual RNA molecules in single cells. This has greatly enhanced the resolution in our understanding of transcriptional processes. Here, we adapt a recently published smiFISH protocol (single-molecule inexpensive fluorescent in-situ hybridization) to whole embryos across a range of arthropod model species, and also to non-embryonic tissues. Using multiple fluorophores with distinct spectra and white light laser confocal imaging, we simultaneously detect and separate single RNAs from up to eight different genes in a whole embryo. We also combine smiFISH with cell membrane immunofluorescence, and present an imaging and analysis pipeline for 3D cell segmentation and single-cell RNA counting in whole blastoderm embryos. Finally, using whole embryo single-cell RNA count data, we propose two alternative single-cell variability measures to the commonly used Fano factor, and compare the capacity of these three measures to address different aspects of single-cell expression variability.
Conflict of interest statement
The authors declare no competing interests.
Figures
Fig. 1. smiFISH in different arthropod species and tissues.
smiFISH for the segmentation genes even-skipped (eve) and engrailed (en) are shown in early and later embryos from five different arthropod species, Drosophila melanogaster (D.mel), Drosophila virilis (D.vir), Tribolium castaneum (T.cas), Nasonia vitripennis (N.vit) and Parhyale hawaiensis (P.haw). Embryos are oriented with anterior to left. smiFISH for wingless (wg) and en is shown in the D.mel imaginal wing disc. Ovaries were stained for the maternally loaded RNAs bicoid (bcd) and nanos (nos), which accumulate at the anterior and posterior poles of the developing egg, respectively. A single stage 10 egg chamber is shown, oriented with nurse cells and the anterior of the developing egg to left. DAPI was used to stain cell nuclei. All images were acquired using a white light laser scanning confocal microscope with 40X or 100X objectives. White dashed boxes are magnified to the right. Single mRNAs are visible for all samples tested.
Fig. 2. smiFISH and white light laser confocal imaging to visualize all eight Drosophila Hox genes at single molecule resolution.
A stage 10 germband extended D. melanogaster embryo (lateral view, anterior left) with smiFISH staining for all 8 Hox genes, plus DAPI to show the nuclei. The X-flap sequence was used for all probes, with the following fluorophores: labial CalFluor 610, proboscipedia Quasar 570, Deformed AlexaFluor 488, Sex combs reduced Quasar 670, Antennapedia promoter 1 CalFluor 540, Ultrabithorax Quasar 705, abdominal-A CalFluor 590 and Abdominal-B CalFluor 635. The image stack was acquired using a Leica SP8 inverted confocal microscope, with 40X objective and a white light laser, enabling optimal excitation wavelengths for each fluorophore. Peak emissions were captured by narrow ~20 nm tunable collection windows, and the image spectrally unmixed in Leica LAS X software to correct any residual bleed-through. Large bright spots mark accumulations of nascent RNAs at transcriptional sites; smaller, fainter spots are single mRNAs.
Fig. 3. smiFISH with membrane immunofluorescence allows whole-embryo 3D segmentation and multi-gene single-cell RNA quantification.
a Stage 5 cellular blastoderm D. melanogaster embryo (lateral view, anterior left) with maximum projections of smiFISH for the pair rule gene even-skipped, and four gap genes: hunchback, knirps, giant and Kruppel. Cell membranes are stained by immunofluorescence, using mouse anti-Drosophila alpha Spectrin, and goat anti-mouse Alexa Fluor 488. The confocal image stack comprises 48 slices at 200 nm _z_-intervals (9.6 μm total depth), from the apical limit of mRNA spots, to the basal extent of membrane ingression, to capture all RNAs that could accurately be assigned to single cells. b The cells module in Imaris v9.2 software was used to automatically segment Spectrin staining in 3D through the confocal stack, creating individual cell volumes (1982 in total). The Imaris spots module was used to automatically identify mRNA spots for each gene; and automatically assign spots to cell volumes based on x, y, z coordinates. c Heatmaps displaying mRNA number per cell for each of the five genes. d Histograms of mRNA number per cell for each gene, using bins of five with zero excluded. The shape of histogram distributions is a product of the expression patterns of the genes.
Fig. 4. Automated identification of immediately neighbouring cells for single-cell variability analysis.
a 2D segmentation plane from a cellular blastoderm D. melanogaster embryo with anti-Spectrin membrane staining. The number of immediately neighbouring cells (defined as directly sharing a portion of membrane) was manually counted, for each cell in half of the embryo. A range of 2–8 immediate neighbours was found. b Low-threshold Imaris spot detection in the Spectrin channel, to generate dense representation of the membrane with spots, for 3D polygon generation. c A custom R code was developed to generate 3D polygons from Spectrin spot coordinates and assigned cell IDs, closely reproducing initial segmentation. Each polygon was slightly expanded in 3D by 0.6 μm (~10% of a cell diameter), generating intersections between borders of only directly neighbouring cells. The code detects all intersections, producing for each polygon a list of directly neighbouring cells. d Histogram summarizing the manual neighbour count (1028 cells) in (a). e Histogram summarizing the automated neighbour count (1982 cells) in (c), confirming close agreement with the manual count. f Histogram summarizing a direct cell-by-cell comparison of neighbour number between the manual count and the automated method, for the first 200 cell IDs, calculated as manual minus automated.
Fig. 5. Three alternative measures to express cell-to-cell mRNA variability.
a Formulae for Fano factor, a commonly used measure of cell variability, and NV and PV, two alternative measures designed to better capture individual cell variability. b Cells from the anterior stripe of eve expression in P. hawaiensis early germband embryo, highlighting that individual cells can differ greatly in expression from their immediate neighbours. The centre cell (white arrow) in the left panel is similar to all its neighbouring cells except one, whereas the centre cell in the right panel is highly different from all of its neighbours except one. c–f Hypothetical scenarios of neighbour-group mRNA variability, to highlight the capacity of each formula to capture the variability of the single centre cell (red dashed border). Here, 550 mRNA/cell is used as the population maximum. Fano factor incorrectly returns the same value for (c) and (d), and incorrectly finds (e) to be more variable than (c). NV correctly returns the same value for (c, e and f), and a low value for (d). PV correctly returns the same value for (c) and (e), and lower values for (d) and (f). g Heatmaps show the three different variability measures, calculated for each segmented cell of the embryo (1982 cells in total), for five genes: even-skipped, hunchback, knirps, giant and Kruppel. Dots representing cells are scaled in both size and colour by the variability value. Both NV and PV measures can range from minimum 0 to maximum 1. For PV heatmaps, cells were filtered on the criteria of neighbour-group mean mRNA number ≥1; cells failing this criterion were assigned a score of 0.
References
Publication types
MeSH terms
Substances
LinkOut - more resources
Full Text Sources
Other Literature Sources
Molecular Biology Databases