Spatial transcriptomics technologies opened the door for new kinds of biological measurements, allowing scientists to generate detailed maps of where genes are expressed in tissue. But most methods rely on expensive and time-intensive imaging that requires specialized equipment.

A new method developed by researchers at the Broad Institute promises to make spatial transcriptomics easier for scientists to use. The approach eliminates the need for imaging and instead uses computational methods to reconstruct the spatial locations of gene expression.

The team showed that by using their method, they could map larger sections of tissue more quickly and cheaply than with previous methods. They add that, more importantly, the new approach requires no specialized equipment and can be used by more researchers around the world. The work appears in Nature Biotechnology.

“Our work converts imaging into molecular biology—just a reaction in a test tube,” said Fei Chen, who is a senior author on the study, a core institute member at the Broad, and an assistant professor in the Department of Stem Cell and Regenerative Biology at Harvard University. “That means anybody can use this approach if they have the algorithm and some common materials.”

“When biologists think about spatial locations, they might think they need to look at samples with light microscopy or electron microscopy,” added Chenlei Hu. Hu is the first author on the work and a Harvard graduate student in Chen’s lab. “But we’ve found that we can computationally infer physical locations instead.”

Mapping with beads

The new findings build on a technique called Slide-seq, which was developed by Chen, Broad core institute member Evan Macosko, and colleagues in 2019. The method generates high-resolution maps of gene expression across tissue. Researchers first collect images of an array of DNA-barcoded beads on a slide, creating a reference that tells the location of each bead.

Next, they place a tissue section on the beads and dissolve it, leaving messenger RNA from the tissue bound to the barcoded beads. They then load the beads into a sequencer and use specialized software to create a map of gene expression across the tissue.

In the past, Chen’s lab made the arrays and imaged them for other scientists. Their microscope was under near-constant use. But lingering in their minds was the possibility that they could infer the location of each bead with sequencing alone, eliminating the need for imaging.

They thought that if you knew the distance between every pair of beads on the array, you could reconstruct their spatial positions—in the same way you can locate a cell phone by knowing its distance from satellites.

When she joined Chen’s lab, Hu thought it might be possible to pinpoint the location of the beads by measuring how much molecules diffuse between them. She and her colleagues built a new kind of bead array containing both “transmitter” and “receiver” beads, each with DNA barcodes.

When exposed to UV light, the barcodes cleave from the transmitter beads, diffuse away, and are captured by the receiver beads. Receiver beads that are closer to transmitters will capture more DNA barcodes as they diffuse from the transmitters.

Slide-seq’s sequencing step can measure the level of these captured barcodes, providing information not only about gene expression, but also the location of the beads. Hu then used an algorithm commonly used in single-cell analysis called Uniform Manifold Approximation and Projection (UMAP) to reconstruct the beads’ original locations on the slide.

When the researchers used their method and the image-based Slide-seq to analyze the same sample, they found very little difference. Without the time-intensive imaging step, Chen’s team was able to map gene expression across larger sections of tissue than before: areas up to 1.2 centimeters wide of mouse embryo tissue (previous maps covered only about 3 millimeters). The Chen group is now working with the Macosko lab to map areas as large as 7 centimeters, close to the size of entire organs in people.

“We’re no longer limited by how long it takes us to image something,” Chen said. “Eventually we’d like to analyze the whole human brain. That just wasn’t possible with other technologies.”