Skip to main content
Premium Trial:

Request an Annual Quote

Yale University, Karolinska Institute Researchers Advance Spatial Chromatin Accessibility Profiling


NEW YORK – A new method that adds a spatial aspect to chromatin accessibility assays provides more resolution on the fundamental epigenetic drivers of cell states than ever before.

The method combines ATAC-seq (assay for transposase-accessible chromatin by sequencing) with deterministic barcoding, a method developed by Rong Fan of Yale University. Dubbed spatial-ATAC-seq, the assay offers the potential for single-nucleus resolution of chromatin accessibility.

In a proof-of-concept paper published Wednesday in Nature, the Yale team and its collaborators in Gonçalo Castelo-Branco's lab at Sweden's Karolinska Institute showed the ability to get data from thousands of nuclei per sample in tissue slides taken from mice and humans. The grid-like barcoding scheme provides a tunable range of resolutions, from 10 µm to 50 µm.

The new method can help researchers "go one step beyond" what single-cell and spatial transcriptomics methods can reveal, said Castelo-Branco, senior author of the paper. "We can see the cell in the context."

"We studied development here, but we're also interested in disease," he said. "To define the disease niche and see how chromatin changed in the disease state, it really opens a lot of doors." 

"This is very exciting for the field of epigenomics," Claudia Lalancette, managing director of the Epigenomics Core at the University of Michigan Medical School, said in an email, adding that she has "been on the lookout for this publication" for over a year.

"Single-cell ATAC-seq methods have shown that chromatin organization is crucial to better understand cellular gene expression," she said. "Therefore, this spatial-ATAC-seq method will be in high demand!"

Fan and co-first author Yanxiang Deng have applied for a patent related to this work, and spatial-ATAC-seq is already available as a service through AtlasXomics, a Fan lab spinout.

Spatial-ATAC-seq is one of several methods developed as part of a collaboration between the Fan and Castelo-Branco labs.

By late 2020, Fan had developed deterministic barcoding in tissue sequencing (DBiT-seq), a spatial transcriptomics method that uses microfluidics to overlay tissue samples with barcodes in a lattice with 10-µm resolution.

His lab was working on extending the barcoding scheme to other assays. "I never studied neuroscience," Fan said. "But I knew this was the best system to apply the tool. There are so many different tissue types developing at the very beginning of organogenesis, all tightly controlled by epigenetic alterations. If there was one person who was the best collaborator, it was Gonçalo."

The labs also developed a spatial version of CUT&Tag (Cleavage Under Targets & Tagmentation), a sequencing-based method for chromatin modifications, published in February in Science (Castelo-Branco's lab recently developed a new, non-spatial method that builds on CUT&Tag, called nano-CT).

Spatial-ATAC-seq follows a method called sciMAP-ATAC, developed in Andrew Adey's lab at Oregon Health & Science University, which initially offered spatial resolution of 200 microns with 200 micron spacing with the use of punch-holed tissue samples.

One of the keys to unlocking the new workflow was to simply add the Tn5 transposase directly to the tissue section. "No one had done direct tagmentation," Fan said. "It turns out you don't have to get high-quality nuclei first."

One could simply digest the tissue after direct tagmentation for bulk assay, he suggested, but to add the spatial dimension, the microfluidic platform he developed adds X- and Y-coordinate barcodes, via bulk reverse-transcriptase and in situ ligation.

Areas where the barcodes overlap create "pixels." In the paper, the researchers used a 50-by-50 setup, for 2,500 pixels per slide; however, Fan said they have since upgraded the method to do 100-by-100 barcodes, for 10,000 pixels per slide. In the paper, they used 10-µm and 50-µm pixels for mouse embryos and 20-µm pixels for brain and tonsil tissues.

"We do get single-cell data," Fan said. "You don't always have pixels covering one cell, but if a nucleus is in the pixel, you're getting single-cell ATAC-seq data." Some pixels, however, may cover more than one nucleus, which would require deconvolution algorithms that don't yet exist for spatial ATAC-seq datasets.

The sample preparation workflow takes about one and a half days, and costs approximately $500 per sample, inclusive of sequencing at a commercial NGS provider, Deng estimated. Much of the prep cost is driven by the Tn5 transposase, about $100 per sample.

In the paper, the researchers applied spatial-ATAC-seq in a variety of mouse and human tissues. Fan suggested that it will work in just about any tissue from mice or humans.

Using 20-µm pixels in mouse brain and human tonsil tissues, the researchers reported median counts of 7,647 and 14,939 unique fragments per pixel, respectively. Of those, approximately 18 percent fell within transcriptional start site regions and 24 percent and 14 percent were in ATAC-seq peaks. For comparison, 10x Genomics' non-spatial single-cell ATAC-seq assays obtained a median of 17,321 unique fragments per cell with 23 percent TSS fragments, they said.

"We used it to validate many things we had seen in the literature," Castelo-Branco said, such as gene regulators involved in increasing or decreasing chromatin accessibility and the organization of immune cell types in lymphoid follicles, adding that it was "a showcase of the potential of the technology."

"I am already looking to integrate this tool for the researchers I serve," Lalancette said. "However, we must not forget that chromatin accessibility is the 'end result' of many layers of epigenetic regulation involved in chromatin organization." She noted that the authors' spatial-CUT&Tag workflow can be run on the same microfluidics barcoding platform and that she is looking forward to seeing other assays built on top of that.

As single-cell analysis goes increasingly multi-modal, spatial will soon follow. "I really think that's where the most powerful data analysis can be achieved," Fan said. "Not just using your own spatial-ATAC-seq data but adding single-cell RNA-seq data."

New bioinformatics strategies for this type of data will need to be developed, though. "Pixels can have two or three nuclei," Deng said. "In principle, you can do decomposition or deconvolution to get the epigenetic profile from each different nucleus."

Computational methods to do that already exist for spatial transcriptomics data. "I hope someone can do the same deconvolution for spatial-ATAC-seq," he said.