Nature Methods has selected spatially resolved transcriptomics as Method of the Year, but how is the field being developed, and what bottlenecks are currently being dealt with? Professor Joakim Lundeberg (SciLifeLab/KTH) sheds light on the future of the method in this Q&A.
Virtually all cells in the human body co-exist with neighboring cells and there is a cross-talk between these cells for survival, proliferation and differentiation. Where cells are, and what surrounds them, is therefore important to know when trying to understand their origin and function. One method that can do this, recently selected Method of the Year by Nature Methods, is spatially resolved transcriptomics.
To analyze where a cell and its neighbors reside in tissue, experimental and computational methods are used. Spatially resolved transcriptomics was originally developed at SciLifeLab by researchers from Karolinska Institutet and KTH Royal Institute of Technology, and is still constantly being developed. For the full origin story, head on over to the feature by Nature Methods and the methodology summary by Joakim Lundeberg et. al. Here, we dig a little deeper into the future of spatially resolved transcriptomics with the help of Joakim Lundeberg.
Deep learning can make it possible to get “super-resolved spatial transcriptomics”, according to your summary. When can we expect to see this?
“Super-resolved transcriptomics is just being realized at SciLifeLab. We use deep neural networks on current spatial transcriptomics data and we achieve subcellular resolution without any engineering of the glass slides. This is simply amazing, it will allow us to precisely predict gene expression on the tissue image alone.”
“Bead-based technology” is identified as a way forward for spatially resolved transcriptomics, but it currently does not provide enough data. Why is this, and how can it be improved?
“Initial efforts to increase resolution in spatial transcriptomics have been to use barcoded micron sized beads. The rationale has primarily been that a sphere has a larger surface than a flat surface, so that more transcripts can be spatially captured. We have demonstrated that we can reach subcellular resolution with a bead approach. However, it is challenging to decode the beads that are randomly distributed on a glass slide before using these for investigation of tissues. The next generation of spatially barcoded surfaces is most likely going to be based on ordered photolithography, that enable high precision synthesis of mRNA capture probes. This will indeed facilitate a more streamlined production of the core component of our methodology, the barcoded glass slides.”
You also state that “the conversion of RNA into cDNA in the first strand synthesis remains a bottleneck for most sequencing-based transcriptome”. Could you explain this further, and what could be a possible solution?
“Copying of mRNA into cDNA is executed by the enzyme reverse transcriptases (RT), and the efficiency of RT is well known to be suboptimal. Only a fraction of all mRNA molecules are copied to cDNA, so to get rid of this bottleneck we need to improve the enzymatic properties by protein engineering or sequence the RNA directly. In fact, the NGI infrastructure at SciLifeLab has access to nanopore technology that would, in principle, allow us to sequence RNA molecules directly from tissue sections, but the protocols need to be worked out.”
You foresee that “multimodal spatial profiling” will be a big thing for the future. Could you explain this further?
“An important component in our methodology is that we use DNA sequencing as read-put and its enormous capacity in providing genetic information. In a single experiment with a tissue section we can identify and position all active genes. Most other spatial technologies investigate only a fraction of all biomolecules. The frontier in this aspect is to take advantage of the sequencing capacity, but label other biomolecules than mRNA. This will be the next generation of spatial technology. Novel concepts to spatially barcode genomes, epigenomes, and proteins with short DNA barcode sequences would open up for multimodal analysis of single tissue sections, and I believe that simultaneous and parallel analysis of DNA, RNA and proteins in the context of morphology would contribute to a better understanding of health and disease states.”
What are the next big steps for spatially resolved transcriptomics?
“There are many opportunities on the technology side, such as the multimodal analysis, but also on the application side, in biology. For example, international efforts within the cancer community have established large databases of genomes and transcriptomes that have been widely used, but most of these efforts have used bulk tumor samples, a mixture of cells without spatial information. I believe that the next generation of databases for cancer will need to consider the spatial context to understand basic properties of tumor spread, tumor microenvironment in particular, immune cells, and so on.”
Read more about the origin of spatially resolved transcriptomics in the feature by Nature Methods, more about the method in the summary by Joakim Lundeberg et. al, and read the interview with Joakim Lundeberg by KTH Royal Institute of Technology.