Spatially barcoding single nuclei to enable single-cell spatial transcriptomics Open Access

High-throughput single-cell transcriptome sequencing technologies were developed nearly a decade ago, kickstarting a revolution in identifying the constituent cell types of the body in health and disease. These various technologies essentially work by isolating single cells into a compartment and attaching a unique cell barcode (a string of oligonucleotide bases), to the mRNAs of the cell in the compartment. This was originally achieved using microtitre plates, where individual cells can be transferred by hand or fluorescence-activated cell sorting to a single well. Reaction volumes were later reduced by using nanowells, reverse-emulsion microfluidic droplets, and the nucleus itself as compartments. Despite these developments, a necessary step of these technologies is the generation of single-cell or single-nucleus suspensions from solid tissues or other biological samples, which are used as input. Therefore, although we get information on what the cell is, and how it is behaving, we lose important information on where the cell was in the tissue. This information can be helpful in identifying, which other cells a cell has interacted with, or could interact with. Additionally, being able to identify where a cell came from spatially, can help us understand what its function is.

To address this, spatially-resolved sequencing technologies were developed. These technologies generally work by placing a thin (10–20 micrometres) slice of tissue onto a specially designed array, which has a patterned surface of capture sites. These capture sites are analogous to the cell barcodes used in single-cell sequencing technologies, but since they barcode regions, they are often referred to as pixels. The spatial location and barcode are known in advance, and after laying the tissue on top, mRNA is captured by the nearest pixel, giving it a spatial address, revealing the spatial patterns of gene expression after library preparation and sequencing. These technologies were crucially able to reveal unbiased profiles of gene expression, since they often relied on the capture of polyadenylated mRNAs, rather than using gene-specific probes. However, these methods have three major drawbacks. The first of these is that the barcoded arrays impose an artificial regular structure on the tissue, meaning that one pixel can capture transcripts from multiple cells, partial cells, and information can be lost if there are gaps between the pixels. This is particularly pronounced when cells with very different expression levels are next to one another, as the cell with the higher expression level will be overrepresented on the pixel. Next, since the techniques rely on two-dimensional capture the efficiency of capture is reduced, reducing sensitivity compared to single-cell methods, which perform barcoding in three-dimensions. Finally, single-cell technologies have a mature suite of analysis tools, but spatially-resolved sequencing analyses often are bespoke in order to deal with the unique qualities of each tissue and the associated challenges.

Slide-tags is a technology, which aims to combine the best features of single-cell sequencing and spatially-resolved sequencing by spatially barcoding single nuclei within a tissue (Figure 1a). These nuclei can then be isolated and used as input into virtually any of the established single-cell sequencing technologies. This provides a dataset with two matrices: an expression matrix of genes x cells (in the case where single-cell transcriptomic technologies are used downstream of spatial barcoding), and spatial coordinates x cells. The data quality is indistinguishable from high-sensitivity single-cell sequencing data, but now we can also locate where the cell came from in the tissue which helps to contextualize the molecular measurements.

In order to barcode single nuclei in a tissue, a photocleavable barcoded-bead array is required. These are made by first synthesizing DNA onto the surface of solid 10 micron-diameter beads. Each bead contains, in the following order from 5’->3’: a linker moiety that is cleavable by ultraviolet (UV) light, a handle for library preparation, a bead barcode (later called a spatial barcode, since this encodes the spatial information), a unique molecular identifier (UMI), and a capture sequence (which is specific for single-cell sequencing technology to be used downstream). Each bead can contain millions of oligonucleotides. On each bead, the oligonucleotides will possess the same bead barcode sequence, which is achieved by using a split-pool synthesis method. However, they each have an individual UMI, which is a short string of random oligonucleotide bases, so that later, the number of molecules associated with each nucleus can be counted. Once the pool of beads is created, they are spread in a monolayer on a glass coverslip, which has been coated so that the beads stick to it. These bead arrays can be made in a variety of shapes and sizes depending on the tissue type and area to be profiled. At this point, the position of each bead barcode sequence is not known, since the beads have been randomly dispersed onto the surface, and so the bead barcode is read using in situ sequencing, generating a map of bead barcode sequences by x-y location on the array. The array is now ready to be used in an experiment.

Tissues are routinely preserved by freezing, and 20 micrometre thick sections of this tissue can be obtained by using a cryostat. A section is subsequently placed on the sequenced barcoded bead array and melted onto the surface (Figure 1a). A buffer is carefully added to prevent the section from drying out and the assembly is placed under a UV-light source to release the spatial barcodes from the bead array. Barcodes then diffuse into the tissue and stably associate with nuclei. Each nucleus will receive 100 to 1000 of spatial barcodes, these can be counted and plotted in 2-dimensional space per nucleus, forming a 2D gaussian distribution. The mean of this distribution is the spatial location of the nucleus. This feature of the technology allows higher resolution of spatial positioning (~4 micrometres, measured as the standard error of distribution) than the bead size of 10 micrometres. After barcoding, nuclei are isolated from the bead array and washed to remove debris and excess spatial barcodes. Finally, these nuclei can be loaded into a single-cell sequencing technology of choice and profiled. The resultant sequencing libraries are sequenced and data are processed to yield single-cell sequencing data and spatial positions (Figure 1b–e).

The field of single-cell sequencing is now over 15 years old, and so there are many established technologies that can measure a variety of macromolecules present in a cell, including, but not limited to: DNA mutations, DNA methylation, histone marks, open chromatin, transcripts, and proteins. These measurements can often be performed singly or in combination with other macromolecules. Therefore, Slide-tags allows many features of a cell to be measured with spatial context that were not possible previously. In the original publication, we showed that it is possible to measure open chromatin and transcriptomes of single cells with spatial positions. This, for example, allows the measurement of certain transcription factor activities, which are difficult to measure with transcriptomics alone due to their low expression level. Slide-tags produces single-cell data, so the advanced suite of analysis tools can also be leveraged to study the resultant data. For example, we showed that you can use a tool called copy number inference on transcriptomic data to predict patterns of DNA mutations in a human melanoma sample.

Existing spatial transcriptomics methods that capture transcripts on pixel-based arrays allow researchers to plot the spatial distribution of gene expression and capture both nuclear and cytoplasmic mRNAs. Furthermore, they give a continuous picture of gene expression, in contrast to Slide-tags, where only nuclear transcripts are currently captured, and some nuclei are necessarily lost during isolation, leading to a sparser spatial representation of the tissue gene expression. Additionally, single molecule fluorescent in situ hybridization and In situ sequencing approaches can also be used to read out tissue transcriptomes. However, these can be limited to certain targets, rather than transcriptome-wide, and it can be challenging to assign individual transcripts to their cell of origin, a process called segmentation.

Slide-tags has since been commercialized and the number of tissues profiled is expanding every day along with the types of macromolecular measurements. The in situ sequencing step can now be obviated by using computational reconstruction, which will further democratize the technology. We envision that the tool will be important in atlasing efforts, where immediate contextualization of cell types in their spatial environment will aid in annotation and inference of function. Overall, the protocol adds an additional ∼10 minutes of barcoding time to any single-nucleus sequencing experiment, and so we believe it will become intrinsic to single-nucleus sequencing workflows on frozen tissues.

This article details a novel technology, Slide-tags, which allows spatial single-cell sequencing.

Andrew Russell obtained an MBiochem in Biochemistry from the University of Oxford in 2016, and a PhD in Biological Sciences from the University of Cambridge and Wellcome Sanger Institute in 2021. He is currently a Postdoctoral Associate at the Broad Institute of MIT and Harvard and the Harvard University Department of Stem Cell and Regenerative Biology. He was awarded an EMBO Postdoctoral Fellowship for this research and was also appointed as a BroadIgnite Principal Investigator in 2024. His research focuses on the development and application of spatial single-cell genomics technologies. Bluesky: ajcrussel.bsky.social, Email: [email protected]. Twitter: @AndyRusss