A flexible and low-cost spatially resolved transcriptomics strategy making molecular tissue profiling a widely available solution

Description

Spatially-resolved transcriptomics methodologies are revolutionizing our understanding of complex tissues, but their elevated costs represent still a bottle-neck for their democratization. In this work we present a low-cost strategy for manufacturing molecularly double-barcoded DNA arrays, enabling large-scale spatially-resolved transcriptomics studies. We applied this technique to spatially resolve gene expression in several human brain organoids, including the reconstruction of a 3-dimensional view from multiple consecutive sections, revealing gene expression divergencies throughout the tissue.

Steps to reproduce

Spatial transcriptomics data has been generated with our in-house manufactured DNA arrays and dedicated library preparation procedure. NGS libraries were paired-end sequences (150nts length). Primary analysis has been performed with our in-house developed tool SysISTD (SysFate Illumina Spatial transcriptomics Demultiplexer: https://github.com/SysFate/SysISTD). SysISTD takes as entry paired-end sequenced reads (fastq or fastq.gz format), and two TSV files, the first containing the sequence of the molecular barcodes associated with the rows or columns in the printed arrays and the second file presenting the physical position architecture of the spatial barcodes. SysISTD search for the Gibson sequence (regex query), then for two neighboring barcodes. Paired reads presenting these features were aligned to the human genome (hg19) with Bowtie2, and known transcripts were annotated with featureCounts. As an outcome, SysISTD generated a matrix presenting read counts associated with physical coordinates in the array in columns and known transcripts ID in rows. To focus the downstream analysis to the physical positions corresponding to the analyzed tissue, we used an in-house R script taking as entry an image of the DNA array scanned with the TRICT filter after the reverse transcription step (https://github.com/SysFate/SysHMDAPr). Images used in this study are available herein. Indeed, this image reveals the presence of the fiducial borders and the cDNA within the tissue. Specifically, we upload to R a cropped image within the fiducials (imager package) and we use the “px.flood” function to retrieve the pixels associated with the tissue. Finally, we applied a pixel to gexel coordinates conversion prior to crossing this information with the outcome of SysISTD. The final matrix is available herein.

Institutions

Categories

Funding

Licence