Skip to main content
Premium Trial:

Request an Annual Quote

Harvard Researchers Publish Light-Activated Spatial Transcriptomics Method


NEW YORK – Researchers from the Wyss Institute and Harvard University are commercializing a new method for spatial genomics, which uses light to select areas of interest in fixed tissue samples for both imaging- and sequencing-based analyses.

Dubbed Light-Seq, the method builds on several molecular chemistries developed by members of the group, amounting to the ability to "perform light-directed in situ spatial barcoding of target molecules in desired regions of interest for ex situ NGS without sample destruction," as the group wrote in their proof-of-concept paper published this week in Nature Methods.

The key steps are using UV light to crosslink pre-designed barcodes onto cDNAs made from transcripts in the tissue sample and then using a special synthesis reaction to bridge the Y-shaped junction in between the transcript and the barcode, creating a library that can be read out with short-read sequencing.

The workflow is nondestructive, and the researchers showed they were able to revisit the sample for further analysis, such as protein staining. It captures approximately 4 percent of transcripts compared to single-molecule fluorescence in situ hybridization, the gold-standard for transcriptomics sensitivity. Moreover, Light-Seq can pull out lowly expressed genes, said co-first author Emma West. "We're capturing as many as 24,000 genes. That's the whole transcriptome; it's a huge amount of information."

The optical setup described in the paper yielded "practical resolution" of less than 2 µm, but theoretically it could reach the diffraction limit of light for subcellular resolution.

"Light-Seq’s unique combination of features fills an unmet need: the ability to perform imaging-informed, spatially prescribed, deep-sequencing analysis of hard, if not impossible-to-isolate cell populations or rare cell types in preserved tissues, with one-to-one correspondence of their highly refined gene expression state with spatial, morphological, and potentially disease-relevant features," senior author Peng Yin, a professor at the Wyss Institute, said in a statement.

Yin and two co-first authors of the paper, Jocelyn Kishi and West, founded Digital Biology, a startup to commercialize the technology, in August 2021. Kishi is president and CEO while West is chief technology officer.

Kishi declined to provide more information about the company but said Light-Seq was developed to be a "game changer."

"You hit some limits by just imaging" tissue samples, she said. "RNAs are quite crowded, and you have to do many cycles."

The Light-Seq group used work from two earlier methods to create their workflow: Action-PAINT, a high-resolution labeling method, Primer Exchange Reactions, and SABER-FISH (Signal Amplification by Exchange Reaction FISH), a method for imaging gene expression directly in intact tissues.

"Some of the ideas we can trace back to 2015," Kishi said, though much of the brainstorming happened in 2017. "Just as we were getting going, COVID hit. It has been a long journey, but we're very excited to have gotten this far."

The method starts with in situ reverse transcription to create cDNAs from the RNAs in the sample. Adding in the barcodes and a 1 second pulse with UV light fixes those barcodes to the molecules. The light can be directed to highlight particular cells in a sample, such as those with interesting morphology identified under a microscope or stained with the markers one would use in fluorescence-activated cell sorting.

"This is sort of like being able to do FACS without destroying the sample," West said.

Finding a way to get the barcode and the transcript sequence on the same molecule was a big challenge. "Polymerase can't go through a photocrosslink," Kishi said. To solve this, they designed the sequences around the crosslink to have identical domains "so that the extended primer can reach across the junction and be templated on the opposing strand through a branch migration competition," they wrote. They then treat with RNase H to release the single-stranded DNA molecule, amplify it, and prepare it for NGS.

"The study demonstrated high-resolution spatial barcoding and single-cell barcoding, and overall good assay specificity," said Chongyuan Luo, a professor at the University of California, Los Angeles, who is working on his own light-directed spatial transcriptomics method, photonic indexing sequencing, which he presented on at this year's Advances in Genome Biology and Technology meeting in June.

"The main challenge for the light-based strategy is that throughputs have been relatively low," he said in an email. "Multiplexed RNA in situ-based methods can easily profile millions of cells, and the several light-guided barcoding strategies have shown data generated from dozens of cells, so there is a clear throughput gap here."

But sequencing captures a wider array of RNA molecules, not just known transcripts, Luo added. The study showed detection of not only protein-coding genes but also short transfer RNAs and long noncoding RNAs.

In studies of HEK and 3T3 cells, the study authors reported an average of nearly 2,000 and 1,200 UMIs per 100 square µm — roughly the area analyzed by other sequencing-based spatial methods such as Slide-seq or deterministic barcoding in tissue sequencing (DBiT-seq) — at sequencing depths of 30 million reads per replicate.

Kishi noted that the method only creates amplicons from cells receiving the barcode treatment. "We're only paying to sequence the cells we’re interested in," she said. Library preparation per slide totals approximately $35 per sample, excluding Illumina library preparation, which about doubles that. Sequencing costs depend on how deeply one wants to sequence, she said, which can vary.

"It offers answers to really targeted questions without having to break the bank or jump through a lot of hoops for sample prep," West said.

Sinem Saka, a coauthor who is now at the European Molecular Biology Laboratory in Heidelberg, Germany, noted that Light-Seq enables discovery research because it doesn't require a special sample preparation from the beginning. "You don't need to know in advance you'll do all of this," she said.

In the study, the researchers performed analyses of rare cells in the retina that are scattered throughout space and different layers of the tissue. They would have been nearly impossible to single out with other methods. "You can't just take a big section if you want this individual cell addressability," West said.

While mostly a proof-of-concept paper, she said their data suggest interesting new biology. She and others have tried and failed to capture the transcriptomes of these cells with sorting. "It's a rabbit hole someone should go down; very little is known about what they do," she said.

The paper heralds a new wave of technologies that provide both sequencing and spatial data from tissue slides.

In addition to the forthcoming Xenium in situ spatial genomics platform from 10x Genomics, other groups, like Luo's, are working on methods with a similar concept to Light-Seq.

Such strategies allow "arbitrary selection of regions of interest … which is a unique feature," Luo said.