In Situ Sequencing Data

Technical Note

Download ISS Data Technical Note

The HS Library Preparation Kit and the In Situ Sequencing Kit can sequence up to 600 genes in situ, simultaneously and at single cell resolution, while generating a high signal-to-noise ratio. Image data generated using the CARTANA Kits and fluorescent microscopy is analyzed to decode the identity of ISS spots, each representing one specific detected RNA transcript.
This technical note describes CARTANA ISS Images analysis and ISS Data
decoding and processing.


In Situ Sequencing Data:

• Validation of the library preparation with the HS Library Preparation Kit images

• Image Processing and Decoding of the ISS Kit Data

CARTANA In Situ Sequencing Workflow

Figure 1: Workflow of CARTANA ISS used to generate tissue maps. The imaging and data analysis steps are: library preparation images from the HS Library Preparation Kit and ISS Data from the ISS Kit.

CARTANA HS Library Preparation Kit images

Imaging the CARTANA HS Library Preparation Kit results is the first step in the CARTANA data analysis procedure. The library preparation generates ISS spots which are fluorescently labeled with the labeling mix (included in the HS Library Preparation Kit). All ISS spots in the tissue samples are visualized with one general fluorescent label, irrespective of their barcodes (in Cy3, Cy5 or Alexa Fluor® 750 channel, depending on the choice of labeling mix) together with DAPI counterstaining of the nuclei. Images are obtained using a standard epifluorescence microscope with 20x objective (Figure 2).

Imaging the library preparation results prior to the ISS
procedure has two purposes:
1. To validate the library preparation using the number of ISS spots and the fluorescent intensity of ISS spots. However, the image of the sequencing library does not reveal which transcript an ISS spot represents (this is done during the ISS procedure).
2. The image of the HS Library Preparation is the reference to align the sequencing cycles during the ISS image processing steps.

Figure 2: HS Library Preparation Kit result images using an epifluorescence microscope. Tissue: mouse Fresh Frozen brain section (10 µm-thick). Blue: DAPI, white: anchor. A. Fully zoomed-out image, scale bar: 1 mm, B. zoomed-in image with low and C. high contrast, scale bars: 50 µm.

CARTANA ISS Kit Data Processing and Decoding

Imaging the ISS cycles (with CARTANA ISS Kit) is the second step in the CARTANA data analysis procedure.

Raw Image Data

Using the CARTANA ISS Kit, ISS image data is generated at each cycle. For each cycle, ISS image data consists of images from 5 fluorescence channels (DAPI, Alexa Fluor® 488, Cy3, Cy5 and Alexa Fluor® 750) each taken as z-stack (for 20X images we recommend 1 z-stack per 0.8 µm interval)

with a lateral resolution of ~0.3 µm/ pixel. Images areflattened to 2D using maximum intensity projection. A total of 30 images (6 cycles x 5 fluorescence channels) are generated (Figure 3, DAPI not shown). For example, a coronal section of an adult mouse brain typically generates a total of ~1 TB of raw image data across all imaging rounds and z-stacks, which corresponds to ~100 GB of 16-bit flattened 2D images after maximum intensity projection (without any compression).

Figure 3: Raw images obtained with CARTANA ISS Kit of a mouse Fresh Frozen brain section (10 μm-thick) using an epifluorescence microscope in all fluorescence channels Alexa Fluor® 488, Cy3, Cy5 and Alexa Fluor® 750 without pre-processing.

Image Processing

From the 30 images (5 per cycle/6 cycles) generated with the ISS Kit, individual field-of-view images are stitched (parameters depend on the tissue sample size). The obtained whole-section images can then be pre-aligned using the DAPI counterstaining and an image registration algorithm (Figure 4). This step is optional, and only necessary if the slide re-positioning is not accurate.

The pre-aligned images of all 6 ISS cycles are then subjected to tiling (typical size: 2000×2000 pixels) for convenient handling of files. Prior to further analysis, images are filtered to remove background noise. Finally, a subpixel-resolution registration algorithm is applied to tiled images to accurately align all images and all corresponding ISS spots.

Figure 4: Stitched DAPI images from two sequencing cycles A. before and B. after pre-alignment prior to decoding. Scale bars: 1 mm.

Barcode Decoding of ISS spots

After alignment and processing, images are decoded. First, all ISS spots in each image tiles are segmented, i.e. each ISS spot is recognized by the image analysis software based on its shape, size and intensity.

Second, the sequencing color within each ISS spot in each ISS cycle image is determined, yielding a color code for each targeted RNA. Finally, a transcript name search for the matching code is performed (e.g. ISS cycle1: Cy5; ISS cycle 2: AF750; ISS cycle 3: Cy5; ISS cycle 4: Cy3; ISS cycle 5: AF488; ISS cycle 6: Cy5 = transcript X) (Figure 5).

Figure 5: Images of different fluorescence channels are visualized in pseudo-color and the assignment of gene name to an ISS spot is shown. Example: Cy3-AF750-Cy3-Cy5-AF488-Cy3= transcript X.

CARTANA ISS Result visualization and analysis

The CARTANA ISS data enables the analysis of tens to hundreds of transcripts in more than a million reads, in one single tissue section. Each individual detected RNA transcript is represented as a colored symbol vector to which coordinates are assigned (Figure 6 and 7).

ISS results are summarized in a csv file with three columns: transcript name, X coordinate and Y coordinate (Figure 6.E).

The data is then plotted on top of the DAPI image usingMATLAB (matching the background image in pixel unit, (Figure 6.D).

The data is then ready to be explored in different ways: by visualizing individual genes, or several selected transcripts or all transcripts at once (Figure 6.D).

Figure 6:CARTANA ISS results visualized on top of the DAPI image (grey background image of size 33K x 21K pixels). A-C. At different zoom levels. D. MATLAB figure file (provided by CARTANA) with only one gene toggled on. E. Example view of the result csv file.

CARTANA ISS Result visualization and analysis

Spatial single cell gene expression analysis

Various single-cell analyses can be performed depending on project requirements. Individual cells can be segmented with standard methods using binary threshold and watershed algorithm. The segmentation can also be improved by using single-cell RNA-sequencing data to better define cell boundaries, which is currently an active field of research. Once the segmentation has been defined, reads are assigned to cells allowing classification and statistical analysis.

Figure 7 shows an example of cell segmentation with a probabilistic method based on single-cell RNA-sequencing data. The reads were assigned to cells accordingly (Figure 7.A.). Based on this analysis, cells were classified and statistics performed (Figure 7.B.). The interactive viewer is available here:

Figure 7:Example of result analysis in mouse hippocampal CA1 area [2]. Screenshots from . A. A hippocampo-septal cell (a type of Sst-positive inhibitory neuron) is selected and all reads assigned to this cell are highlighted. B. The probability of the cell being different cell types is visualized as bar plot as well as a donut pie chart.

Software requirements

CARTANA uses the following software for ISS data analysis:
• Image stitching:
microscope vendor’s native software*, MATLAB* or FIJI/
• Image registration:
MATLAB*, OpenCV (Python bindings) or FIJI/ImageJ.
• Object segmentation and batch processing:
CellProfiler ( or MATLAB*.
• Decoding: Python or MATLAB*.
* may require a license

Relevant publications

Detailed description about image analysis and basecalling of the original pipeline:
[1] Pacureanu, A., Ke, R., Mignardiv, M., Nilsson, M.
and Wählby, C. Image based in situ sequencing for
RNA analysis in tissue. 2014 IEEE 11th International
Symposium on Biomedical Imaging (ISBI), Beijing,
2014, pp. 286-289.
Detailed description about improved image analysis
pipeline and single-cell analysis:
[2] Qian, X., Harris, K.D., Hauling, T. et al. Probabilistic
cell typing enables fine mapping of closely related
cell types in situ. Nat Methods 17, 101–106 (2020).

Contact Us
Technical Support: [email protected]
Information and Pricing: [email protected]

© 2 0 2 0 C A R T A N A A B .
F O R R E S E A R C H U S E O N L Y .

Download ISS Data Technical Note