DISCO-seq Opens the Door to 3D Single-Cell Transcriptomics

Single-cell transcriptomics studies begin by dissociating tissue into individual cells and reading, cell by cell, which genes are being expressed. This allows researchers to distinguish cell types, activation states, and how cellular populations change in disease environments. Yet once a tissue has been dissociated into a single-cell suspension or a single-nuclei suspension, the original spatial state of those cells within their biological microenvironment is lost.

Spatial transcriptomics was developed to restore that missing positional information, but it comes with another limitation: most current methods still rely on thin tissue sections. A section can preserve RNA within a local tissue region, but it captures only one slice of a much larger structure. If a lesion is scattered across different corners of an intact organ, appears only as a small number of cells in a remote region, or happens to lie outside the selected section, the most important biological signal may be missed.

One of the most common approaches in spatial transcriptomics. After tissue collection, the sample is prepared as a thin section and placed onto a spatial capture slide containing oligonucleotide probes. Each capture area on the slide carries a poly(dT) mRNA capture sequence, a UMI, a spatial barcode, and a sequencing primer. Once mRNAs are released from the tissue, they are captured by nearby probes, allowing each transcript to retain information about its original spatial location. The captured mRNA is then reverse-transcribed into cDNA, after which the probes are cleaved and the cDNA is collected. Next-generation sequencing is used to read the transcript sequence, UMI, and spatial barcode. Finally, computational analysis maps gene expression back to the original positions within the tissue section（Image source：Isnard P and Humphreys BD. (2025)， CC BY-NC-ND 4.0 ）

DISCO-seq was developed to address this problem. Its central idea is to first render an intact tissue—or even an entire mouse—transparent while preserving its three-dimensional anatomical architecture. Researchers can then use light-sheet microscopy to survey the positions of cells, lesions, or labeled proteins at the whole-organ or whole-body scale. Guided by these 3D images, they can precisely dissect regions that match the biological question of interest and perform single-cell or single-nucleus transcriptomic analysis.

The greatest technical challenge lies in RNA preservation. Many tissue-clearing methods use harsh organic solvents. These solvents can make opaque tissues optically transparent, but they often damage RNA, making downstream transcriptomic analysis unreliable. The research team therefore tested different perfusion reagents, fixation conditions, and clearing chemistries, eventually establishing a workflow that balances optical transparency with robust RNA preservation. The optimized protocol uses tetrahydrofuran/dichloromethane (THF/DCM) for the clearing step, followed by gradual removal of hydrophobic clearing residues so that the sample can be brought back from a hydrophobic organic solvent environment into an aqueous state suitable for RNA detection. Brain tissues processed with DISCO-seq still yielded sequencing-grade RNA, with purity and integrity comparable to fresh or fixed control samples. The method also did not substantially distort the original cellular composition of the tissue, meaning that the advantage of 3D imaging was not gained at the cost of molecular quality.

To test the power of DISCO-seq in disease research, the researchers applied it to a mouse model of glioblastoma. Conventional positron emission tomography can detect larger tumor locations and growth patterns, but its spatial resolution is not sufficient to capture tumor dissemination at single-cell resolution. By combining DISCO-seq with whole-brain clearing and 3D imaging, the researchers observed the primary tumor mass within the intact brain, as well as small accumulations of tumor cells distributed far from the main tumor across different brain regions. They also observed tumor cell trafficking toward the contralateral region via the corpus callosum. Although these microlésions represented only a small fraction of the total tumor burden, they were distributed across multiple cortical and subcortical regions, suggesting potential pathological significance and providing dissemination clues that conventional imaging could easily miss.

Detection of discrete tumor cells and trafficking routes in a glioblastoma model. (A) Schematic of the glioblastoma mouse model and downstream analysis workflow. Researchers injected SB28 glioblastoma cells labeled with enhanced green fluorescent protein (eGFP) into the mouse brain, followed by tumor monitoring, whole-brain clearing, 3D image acquisition, and subsequent regional analysis. (B) Three-dimensional images of the cleared whole brain show that, in addition to a large primary tumor, small numbers of dispersed tumor cells or small tumor cell clusters can still be observed within the brain. The green signal comes from eGFP-labeled tumor cells, and the images also indicate that some tumor cells may traffic toward the contralateral brain region via the corpus callosum. (C) Positron emission tomography (PET), computed tomography (CT), and light-sheet microscopy images of the glioblastoma mouse whole brain were co-registered and aligned with Allen Brain Atlas-based anatomical annotations, allowing tumor distribution to be mapped onto specific brain regions（Image source：Bhatia HS et al. (2025)， CC BY-NC-ND 4.0 ）

The researchers then used the 3D images to select distinct brain regions, including the main tumor mass and remote regions containing sparse tumor cells, for single-cell RNA sequencing (scRNA-seq). The results showed a marked increase in microglia and blood-derived macrophages within the primary tumor mass, along with expression of glioblastoma-associated genes such as C1qa, Ctss, Ctsb, and Mpeg1. Neurons near the tumor also underwent clear transcriptional changes, including upregulation of genes related to synaptic signaling, vesicle trafficking, and neuronal activity, together with downregulation of some genes linked to stress regulation and homeostatic balance. These findings indicate that the tumor actively reshapes the states of nearby neurons, microglia, and immune cells. Remote regions showed molecular features that fell between normal tissue and the tumor core, suggesting that changes in the tumor microenvironment may extend into areas that are not yet clearly visible to the naked eye or by conventional imaging.

Top ten differentially expressed genes in glioblastoma brain tissue compared with sham control brain tissue. The upper dot plot shows differential gene expression in microglia, the middle dot plot shows glutamatergic neurons, and the lower dot plot shows GABAergic neurons（Image source：Bhatia HS et al. (2025)， CC BY-NC-ND 4.0 ）

Finally, the researchers pushed DISCO-seq to an even larger scale: the entire mouse. Thirty minutes after intravenous injection of fluorescently labeled SARS-CoV-2 spike S1 protein, they performed whole-body clearing and light-sheet imaging to observe the distribution of the S1 domain across different organs. S1 accumulated prominently in the liver, lung, intestine, and kidney, while the heart, skeletal muscle, and bone marrow showed relatively sparse accumulation. Further examination of the small intestine revealed distinct S1 distribution patterns between the duodenum and ileum, and these regions were selected for single-cell RNA sequencing.

Distribution of the SARS-CoV-2 spike S1 domain in the mouse body. The bright regions correspond to red light at 665–670 nm, which has been computationally recolored light blue for easier visualization（Image source：Bhatia HS et al. (2025)， CC BY-NC-ND 4.0 ）

Analysis of the small intestine showed that the presence of S1 for only 30 minutes did not dramatically alter the overall cellular composition, but it had already triggered region- and cell-type-specific transcriptional responses. In the duodenum, proximal enterocytes upregulated genes related to mitochondrial respiration, oxidative phosphorylation, and energy metabolism. Distal enterocytes activated programs associated with stress responses, barrier adaptation, and immune interaction. In the ileum, Paneth cells strongly expressed immune and secretory genes, macrophages upregulated inflammatory and antigen-presentation genes, and T cells showed markers associated with activation and cytotoxicity. These findings indicate that even adjacent intestinal segments exposed to the same foreign protein can respond very differently depending on their local microenvironment, immune architecture, and tissue-specific properties.

In the past, biologists often asked which cells express a given gene. With DISCO-seq, they can now begin from the opposite direction: first identify the truly abnormal or biologically meaningful locations within an intact organ or body, then ask which transcriptional programs are being activated by the cells in those precise regions. This makes it possible to capture rare lesions, remote infiltration, regional immune responses, and whole-body disease distributions in a more complete way. Although the method still relies on regional dissection and has not yet reached the point of directly reading every cell transcriptome within an intact animal, it has already connected 3D anatomical imaging and single-cell molecular analysis into a practical and powerful workflow.

Author: Shui-Ye You

References:

1. Bhatia HS et al. (2025). DISCO-seq: 3D single-cell transcriptomics of intact biological systems. bioRxiv.

2. Isnard P and Humphreys BD. (2025). Spatial Transcriptomics. The American Journal of Pathology.