Creating Tissue Atlases with Spatial Transcriptomics

New methods of tissue analysis are furthering our understanding of disease mechanisms, allowing researchers to generate gene expression datasets of tissues and organs. In order to retain spatial information, these methods may use next-generation sequencing (NGS) to encode position onto RNA transcripts before sequencing, or instead may employ imaging-based approaches using in situ sequencing or hybridization. Alternatively, some approaches use a combination of methods. The result is large datasets of RNA expression that researchers are organizing into tissue atlases as they use the information to identify and annotate cell types.

Because technology continues to evolve, most atlases are works in progress, with spatial resolution, coverage, and annotation levels ever-improving. Brain atlases are helping researchers understand connectivity, while atlases of tumors, or healthy versus diseased tissues, are relied upon for studying disease mechanisms. This article discusses advances and limitations of spatial transcriptomics technologies in developing tissue atlases.

Generating an atlas

Spatial transcriptomics, in which gene expression is measured in many cells in situ simultaneously, is a powerful method of cataloging and spatially mapping individual cell types of tissues and organs. It can profile the transcriptome across entire tissues or organs, but lacks single-cell resolution and sensitivity. Single-cell RNAseq can profile individual cells but is much lower in throughput. Another option, imaging-based transcriptomic methods such as multiplexed error-robust fluorescence in situ hybridization (MERFISH), can measure copy numbers and spatial distributions of RNA molecules in individual cells, but with its own caveat. “As cell atlas [generation] requires individual cell types to be identified and characterized within tissue, single-cell resolution is required to perform true cell atlasing,” says Jiang He, Co-Founder and Senior Director of Scientific Affairs at Vizgen. “Imaging-based technologies offer true single-cell resolution and have much higher sensitivity, but can only image a targeted panel of genes.”

Vizgen and other companies offering spatial transcriptomics technologies are supporting the rapid evolution of spatial transcriptomics methods used to generate state-of-the-art tissue atlases. These atlases aim to identify the types of cells within each tissue, in addition to the location and function of each cell type, and their interactions with other cells. “With the advent of single-cell spatially resolved transcriptomic tools, scientists can not only resolve individual cell types and map out their spatial location, but also characterize the gene expression within individual cells,” says He. Scientists at Vizgen recently used MERFISH on the MERSCOPE™ platform to create a molecularly defined atlas with single-cell resolution along the anterior-posterior axis of mouse brain.

Vizgen has also generated 3 publicly available datasets mapping out cell types, including the expression of 483 genes for receptors in mouse brain, and the expression of 347 genes in mouse liver; and a tumor atlas with 500 genes characterizing cells in 8 major tumor types, including human liver, lung, skin, prostate, uterine, ovarian, colon, and breast cancer. “It is also possible to profile more than one type of analyte within the same sample, and perform spatial proteomics assays on the tissue in parallel,” explains He. For this purpose, Vizgen recently released a protein co-staining kit for simultaneous imaging of protein and RNA.

A tissue atlas can detail an entire organism, or only one tissue or organ. For example, the Human Cell Atlas (HCA) is an international collaborative consortium that aims to create freely available reference maps of all cell types in the human body. Recent HCA contributors published work that used transcriptomics and machine-learning to annotate fine distinctions among many immune cell types of varied tissues; created comprehensive atlases of developing organs, which can aid researchers in regenerative medicine and cell engineering; developed methods to optimize RNA sequencing of single nuclei, and created a cross-tissue atlas to study cells from banked frozen tissues; and performed live, single-cell RNA sequencing from over 400 cell types in the same donor, allowing large-scale alternative gene splicing analysis, and cross-tissue comparisons in a data set called “the Tabula Sapiens”.

Many atlases focus on individual organs or tissues, such as human liver or mouse brain. For example, Zizhen Yao, Assistant Investigator at the Allen Institute for Brain Science, recently presented a transcriptomic cell type atlas of whole mouse brain. Her group created the atlas by integrating several large-scale single-cell RNAseq datasets with millions of cells, and annotating the precise location of each cell type using MERFISH.

Limitations

Scientists creating tissue atlases face limitations that technology is continually trying to overcome. For example, while single-cell RNAseq provides better detection and resolution of RNA transcripts, its use is limited to fresh tissue, and some cell types can be damaged by the procedure. “If a certain cell population is less likely to survive the protocol, then the atlas we build may not faithfully represent the cell type proportions—that’s a problem that we’re still tackling,” says Yao. “On the other hand, single nuclei isolation protocols can be applied to frozen tissues, which can be gentler for vulnerable cell types, but this provides lower resolution.”

The size of MERFISH’s field of view can be a limiting factor—suitable for studying a mouse (but not primate) brain section, for example. A specimen larger than the field of view requires division into smaller grids for processing, followed by stitching the data together. This becomes expensive and time-consuming for large, complicated organs. “Stereo-seq, another recently developed technology, provides a larger field of view, but limited cellular resolution and detection sensitivity,” says Yao.

Cost can become prohibitive when performing high-resolution, high-coverage transcriptomics. “We have to make compromises by finding some middle ground, or just wait until the technologies develop further to do another round of similar efforts,” says Yao. “Technology is moving so quickly that what we’re doing today, we didn’t even know that it would be doable last year.” Technology’s fast pace also makes planning for a multi-year project tricky, as committing to one platform or technology is like trying to see into the future.

Advances and applications

Beyond mere cell location, current tissue atlases are reference maps with information about gene expression levels and cell-cell interactions, often with data from multiple modalities. For example, in a large collaboration funded by the Brain Initiative Cell Census Network, Yao and others generated a mouse motor cortex atlas by integrating single-cell and nuclear RNAseq data, single nuclei ATACseq data, and MERFISH data.

Beyond cell typing, researchers are developing assays to study function more directly. “For example, Patchseq technology can profile the transcriptome, electrophysiology, and morphology of a single neuron simultaneously,” says Yao. “Newly developed barcode connectivity assays also allow profiling RNA and connectivity at single-cell resolution simultaneously while preserving single-cell spatial locations.” While there is still a lot of room for improvement, she adds, these technologies show great promise.

Molecular tools for tissue analysis are also evolving as atlases become more detailed, further paving the way for advanced therapeutics. “At the Allen Institute, we have a dedicated genetics tool team trying to integrate single-cell RNAseq and single-cell ATACseq datasets to identify cell type specific enhancers,” says Yao. “Many of these enhancer tools that we’re building will be useful in therapeutic studies because they enable targeting of the particular cell type associated with a disease.”

In addition to deepening our understanding of cell functions and connectivity, tissue atlases will be important tools in developing better therapeutics. “Scientists will be able to pinpoint which individual cells are responding to certain drug treatment in a complex tissue, characterize the drug response, and study the mechanism of action for different therapeutics, which could pave the way for companion diagnostics in the future,” says He.