NEW YORK – A new method that adds spatial position data to single-cell sequencing experiments may help explain why cells of the same type still show differences in gene expression.
Sci-Space is a riff on spatial transcriptomics that incorporates elements of combinatorial indexing, a way to provide statistical uniqueness that enables high throughput experiments, and cellular hashing, a way to label cells. The method, developed by researchers at the University of Washington in the labs of Cole Trapnell and Kelly Stevens, uses microarray slides to introduce barcoded, transcript-like synthetic DNA sequences into a cell. These sequences convey a position on the microarray so that after sequencing, transcripts can be mapped back to a location in a tissue section.
In a paper published earlier this month in Science, the researchers used sci-Space to analyze single nuclei from mouse embryos, capturing approximately 120,000 whole single-nucleus transcriptomes —about 164 nuclei per mm2 of tissue, on average, resulting in a sampling of just over 2 percent of the total estimated nuclei present. The microarray slide contained 7,056 uniquely barcoded spots, covering an 18 mm2 area, with spatial resolution of approximately 200 microns.
Their proof-of-concept studies suggested that spatial context helped explain about half of the variance in global gene expression that was not attributable to sampling, or about 2.5 percent of overall variance.
"The paper shows convincingly that, particularly for some cell types, variation in spatial position contributes significantly to expression heterogeneity," Rahul Satija, a single-cell expert at New York University and the New York Genome Center, said in an email. He was not involved in the study; however, the researchers used his lab's Seurat package for single-cell data analysis. "These signals may be subtle in comparison to variation across entirely different cell types, but they are clearly important, and challenging to discover in traditional single-cell RNA-seq experiments where the spatial context of each cell is lost," he said.
"Sci-Space fills a niche not addressed by any other methods, yet," Trapnell said. "It's the only technology that gives you a map of gene expression at single-cell resolution over very large fields of view, like whole embryos."
The method is part of a "family" of methods that draw on combinatorial indexing methods developed in Jay Shendure's UW lab, in collaboration with Illumina and Trapnell. Perhaps its closest relative is sci-Plex, a single-cell perturbation screening assay that uses cellular hashing to track the conditions each cell was exposed to. "Sci-Space is repurposing the same trick of molecular biology that allows sci-Plex to work, but instead of looking at thousands of samples, we're looking at thousands of locations in the same tissue specimen," Trapnell said.
Work on sci-Space started around 2018, according to co-first author Sanjay Srivatsan. "The hard part was, 'How do you actually deposit these oligos?'" he said. The team turned to the old standby of microarray technology. "It's funny because it seems, obvious," he said, noting that finding a printer took some searching, though they didn't have to look too far to find one at the Fred Hutchinson Cancer Research Center.
Srivatsan credited co-first author Mary Regier, a postdoc at UW, with figuring out how to prevent the barcodes from wandering from their initial position upon contact with tissue. "If they were just on glass, the second they touch tissue they go everywhere," he said. The solution was to add a gel layer and embed the oligos into that.
The use of grids to encode spatial information and a gel to keep order make sci-Space a cousin of two recent spatial techniques: deterministic barcoding in tissue sequencing, or DBiT-seq, from Rong Fan's lab at Yale University, and polony-indexed library sequencing, or PIXEL-seq, from Liangcai Gu's lab, also at UW.
Sci-Space's limitations come in the form of resolution and cell capture rate. With resolution of 200 microns, the method can't catch specific cell-cell interactions and isn't even as highly resolved as the original spatial transcriptomics method, commercialized by 10x Genomics as Visium, which offers approximately 55-micron resolution. The authors also noted that the method is "limited by recovery of only a fraction of cells from each serial section, such that we obtain a 'survey' rather than achieving dense measurements."
But sci-Space is a true single-cell or single-nucleus method, meaning it meshes nicely with those types of technologies.
"One of the most exciting elements of sci-Space is that it uses established technologies," Satija said. "This means the resulting data is not only spatially resolved, but also (by definition!) maintains single-cell resolution. Experimentally, the method does not require high-end microscopy or instrumentation and can potentially be widely adopted across the community. Analytically, the resulting data can be easily integrated with existing single-cell RNA-seq datasets, as the authors show … even those generated by different labs and with different technologies."
In the paper, the authors integrated their data with single-cell experiments performed with 10x's droplet-based Chromium platform as well as the RNA-seq3 protocol.
The paper's authors noted that sci-Space data "can serve as a spatial 'scaffold' for conventional, nonspatial single-cell RNA-seq atlases, which may be more challenging to map onto tissues profiled by spatial profiling methods that lack single-cell resolution."
"Think about Google Earth," Srivatsan said, which combines images from satellites, planes, and street-level cameras. "At some point we're going to need to create scaffold that resolves the different spatial levels [of biology]. I think this is a first step towards that."
Trapnell said he could not discuss whether there are plans to commercialize the assay. "We are some ways away from demonstrating sci-Space in a pathology setting," he said, adding that it would be "an exciting target to strive for."
In the immediate future, the method could be used to find patterns of gene regulation as a function of location, as opposed to cell type, Trapnell said. "There's a lot of signal, so there's a lot to be discovered, probably," he said. "I think we're going to try to find a lot."