Boundary stain shows cell morphology better than the interior stain

Question:  Xenium Multimodal Cell Segmentation boundary stain shows cell morphology better than the interior stain, e.g. for elongated lung alveolar type cells. How can I rescue the transcripts and assign them to cells?

Answer: It is possible to rescue the transcripts with Xenium Ranger resegment by applying the interior segmentation logic on the boundary stain.

The solution in this article is not intended for low quality tissue samples where 18S rRNA signal is dim overall due to RNA degradation. Overall dim 18S rRNA signal accompanied by low transcripts suggests poor sample quality. In this case, write to support@10xgenomics.com for a workflow consultation.

The solution in this article is intended for the specific scenario where for select cell types, e.g. lung alveolar type 1 (AT1) cells, the 18S rRNA signal is comparatively diffuse and the ATP1A1|CD45|E-Cadherin boundary protein stain is strong. In such cases, clusters of transcripts continue outside of the cell as if a ghost cell and overlap with the boundary stain signal.

This article covers one context out of the multiple presented in Rescue Xenium transcripts outside cells. The article presents learnings from Xenium human and mouse lung samples that may extend to other tissue types.

Hidden options in xeniumranger-v2+ enable this unsupported mode. The following command has xeniumranger apply both the boundary stain logic and the interior stain logic on the ATP1A1|CD45|E-Cadherin stain. Because the interior segmentation model was not trained on boundary stain data, researchers should exercise caution in using the hidden option.

xeniumranger-xenium2.0/xeniumranger \
resegment \
--id updatedresults \
--xenium-bundle xeniumbundle \
--boundary-stain ATP1A1/CD45/E-Cadherin \ #default
--interior-stain ATP1A1/CD45/E-Cadherin \ 
--localcores 32 \
--localmem 128

The corresponding control command to use the same version of xeniumranger with the default Multimodal Cell Segmentation is as follows. The number of cells from xeniumranger can vary slightly from the number of cells from the Xenium Onboard Analysis pipeline when cell segmentation algorithm versions differ. The control command ensures an identical number of cells.

xeniumranger-xenium2.0/xeniumranger \
resegment \
--id controlresults \
--xenium-bundle xeniumbundle \
--localcores 32 \
--localmem 128

Discussion

The figure shows the cell segmentation result of running each command on a mouse lung sample. The cell definition coloring in the first column represents the segmentation method, where yellow is for interior logic and blue is for nuclear-expansion. The second column shows all transcripts, and the third column shows select transcripts that define AT1 cells (Ager), AT2 cells (Sftpd, Lamp3) and endothelial cells (Aqp1).

The default Multimodal Cell Segmentation uses 18S rRNA with the interior cell segmentation logic. Use of the ATP1A1|CD45|E-Cadherin stain with both the boundary segmentation logic and interior segmentation logic gives larger cells that are more contiguous and captures transcripts previously unassigned to a cell. The increase in transcripts within cells is modest but noticeable at 7.4% more transcripts in cells for the mouse lung sample. A human lung sample (not shown) similarly showed 6.3% more transcripts within cells.

The number of total cells does not vary, because the interior segmentation algorithm logic requires a nucleus to segment a cell. The number of cells segmented by boundary logic also does not vary, because both pipelines apply the boundary logic first that uses the default ATP1A1|CD45|E-Cadherin stain. The mouse lung sample shows an increase in the proportion of cells segmented by interior logic, while the human lung sample shows a slight decrease.

Product: Xenium In Situ Gene Expression
Last modified: September 13, 2024