Fluorescence in situ hybridization (FISH) technology is used to categorize cells in tissue under various conditions, such as in human colon cancer. Courtesy of Vizgen. The human body is made up of trillions of cells, and each one is as unique as the person it is part of. Each individual cell is arranged alongside its neighbors in a specific pattern that is essential to its systemic role within the tissue. In individual organs, numerous cell types — each with different physical characteristics, molecular signatures, and behaviors — act together as a cohesive unit. Fluorescence imaging has been put to use to capture these traits and behaviors in their spatial context. In recent years, scientists have undertaken the task of cataloging this immense complexity of cellular interdependency through an approach called cell atlasing, which profiles the individual identities of all the cells in a tissue. By mapping every cell in a system or an entire body, as well as the cells’ functions, innovations in the diagnosis and treatment of illness and disease may become possible. However, the techniques previously used to create cell atlases, such as single-cell sequencing, have not routinely captured spatial information (Figure 1). Since all cells are influenced by their neighbors and by their positions within the system, cell atlases without spatial context may be missing critical insight into the diverse range of cells present, as well as cell-cell interactions. Figure 1. Similarities and differences between genomics and spatial genomics. Genomics data is often gathered through bulk sequencing and single-cell sequencing. These approaches can profile a multitude of cells at once and distinguish cell types by genotype, but they destroy spatial context (left). A highly multiplexed single-cell spatial genomics approach, however, can profile individual cells based on the location and quantity of hundreds of gene species in situ, retaining spatial context (right). Courtesy of Vizgen. Now, evermore advanced techniques in spatial genomics are contributing to the biological cartography effort. The ultimate goal of this effort has been to create a cutting-edge technology that can profile individual cells based on the transcription of hundreds to thousands of genes at once. With new spatial genomics tools, scientists can dive deeper when exploring the diversity of cell types and states throughout the body, including identifying elusive rare species and subtle variations that affect the tissue as a whole. Imaging-based single-cell spatial genomics technologies are targeted methods to spatially resolve the expression profile for genes of interest in situ. These techniques offer high spatial resolution and sensitivity, which enables simultaneous profiling of genes at true single-cell resolution. Cell atlases can serve as a reference map to unify the results of biological studies from around the world. One example of an effort to compile this information is the international Human Cell Atlas consortium — founded in 2016 by a collaboration of researchers in the U.S. and the U.K. — which now consists of more than 2000 members from over 1000 scientific institutions. These cell atlases are of great value in research and medicine alike because they allow scientists to discover new cell types, characterize cell states, and understand how cells function and interact with each other. Furthermore, cell atlases of healthy and diseased tissue can be compared side by side, giving precise insight into the specific cell types whose abnormal behavior is associated with a disease — a task that too often feels like looking for a needle in a haystack to researchers and clinicians. By identifying the cell types responsible for an abnormality in functioning, scientists may uncover disease mechanisms and reveal therapeutic solutions. Gene transcription maps biology A cell atlas may include data from all facets of cell biology: the genome and the resulting transcriptome, proteome, metabolome, and metabolic flux. However, spatial context about cell position, shape, size, and the cell’s neighbors is also essential to create a true map in 3D. A human organ is typically made of billions of cells. So, to capture a biologically relevant snapshot, it is advantageous to use a technique that can capture data on a massive scale. To date, this capability has only been achieved by technologies designed to profile the transcriptome, or all the genetic readouts in a cell — a field of study sometimes called spatial genomics or spatial transcriptomics. Spatially resolved transcriptomic data can be built using several approaches that fall into one of two categories: either sequencing-based or imaging-based. In sequencing-based approaches, RNA transcripts (genetic copies) are first captured in situ using methods that retain their spatial information. Then the transcripts are sequenced ex situ. With this approach, the whole transcriptome can be captured. However, sequencing-based methods do not provide true single-cell resolution, and the sensitivity for RNA detection is often low, limiting the technique’s ability to identify rare species or lowly expressed genes that can be missed. Considering that individual cells have distinct identities and roles that cannot be generalized to their neighbors, high-resolution data is critical to understanding cell and tissue function. Imaging-based spatial genomics technologies are targeted methods to spatially resolve the expression profile for genes of interest in situ, and these technologies offer high spatial resolution and sensitivity, which enables simultaneous profiling of genes at true single-cell resolution. Popular approaches taken with imaging-based spatial genomics technologies include in situ sequencing and fluorescence in situ hybridization (FISH)-based methods. Each approach comes with its own inherent advantages. Notably, select FISH-based techniques can generate some of the highest-volume, highest-resolution data on the genomics market, down to subcellular resolution. In recent years, these technologies have entered the market to help researchers perform true single-cell atlasing. As increasingly powerful tools become available and as the volume of data they capture increases, scientists are gaining an evermore direct window into biology. Variations on FISH Spatial genomics scientists using imaging-based methods rely on intricate testing strategies to maximize the volume of spatial genomic data that is captured in their analyses. These elaborate FISH-based assays are rooted in the method that has been a mainstay in molecular biology for over half a century. FISH represents a powerful and sensitive way to visualize gene transcripts within cells of intact tissue. In a typical assay, fluorescent probes are designed to bind to one or a few RNA targets of interest. The probes are allowed to hybridize to the RNA, or form double strands, within a tissue sample, and the sample is imaged to visualize the fluorescent signal, which signifies the transcripts’ expression pattern within the tissue, much like immunofluorescence. Variations on the FISH method have been developed over the years, including single-molecule (sm)FISH, a sensitive and quantitative technique for visualizing individual gene transcript copies within cells with single-molecule resolution1. To conduct a smFISH assay, multiple probes are designed to bind to a target RNA species. Therefore, when incubated with a fixed tissue, multiple fluorescent markers bind to each RNA molecule. When imaging the tissue under a high-resolution microscope, the probes on each RNA molecule generate a bright signal that can be localized within cells and accurately quantified. While powerful, this method by itself is still limited by multiplexing power. Because of the need to spectrally distinguish between fluorophores, transcripts from no more than four genes may typically be imaged per experiment. Therefore, on its own, smFISH is poorly suited to generating cell atlas information, and the method has been limited to quantifying specific genetic information within individual cells. Spatial genomics technologies rooted in FISH must work in conjunction with other techniques to capture data about many gene species at once. One such method is multiplexed error-robust fluorescence in situ hybridization (MERFISH), developed in the lab of Harvard University professor Xiaowei Zhuang2. This multiplexed approach pushes experimental boundaries when it comes to the number of gene species that can be captured, while retaining spatial context and the single-molecule sensitivity of smFISH. MERFISH uses combinatorial labeling with two or more fluorophores and error-robust barcoding, which allows the capture of measurement errors, to spatially profile hundreds of millions of RNA transcripts — representing hundreds to thousands of gene species — within hundreds of thousands of cells at once. The results enable scientists to spatially profile individual cells based on transcript expression, morphology, and cell-cell interactions, which can be used to understand cell type and state across a variety of complex tissue samples (Figure 2). Figure 2. A massively multiplexed FISH technology was used to spatially profile the cells within numerous tissue types, including human and animal. The method reveals the location and quantity of hundreds of key RNA species within individual cells of the tissue with high accuracy and sensitivity, elucidating the heterogeneity and interdependency of all the tissues — based on cell type, cell shape, cell-cell interactions, and specific regions within the tissue. Courtesy of Vizgen. Because the approach does not rely on any intermediate steps such as enzymatic amplification (the selective amplification of DNA or RNA targets using polymerase chain reaction), transcription detection is nonbiased. Detection efficiency is remarkably high: 70% in tissue3 and 95% in cell culture4. Furthermore, the technique produces images at subcellular resolution, since resolution is dictated by the fluorescence microscope used rather than the structure of an array. Several key tactics are combined in the experimental design to make these capabilities possible, including error-robust barcoding and sequential imaging and analysis. Error-robust barcoding For a massively multiplexed FISH experiment, a barcoding strategy is designed to target a panel of hundreds or even thousands of RNA species, each corresponding to a different gene. Each target species is assigned a unique barcode of ones and zeroes (Figure 3). Then, at least 20 probes are designed to bind to every RNA species in the panel, with each group of probes reflecting the unique barcode for each species. Because the probes bind to their RNA targets in a combinatorial manner, millions of target RNA molecules in the sample can be identified. Figure 3. In massively multiplexed FISH experiments, an error-robust barcoding strategy — in which several bits of each barcode are distinct from all the others — is used to account for single-molecule detection and labeling errors. Because errors can later be corrected, images are decoded computationally after the experiment concludes. Courtesy of Vizgen. Furthermore, the use of multiple probes per target raises detection efficiency by about an order of magnitude higher than other spatial genomics approaches — so high, in fact, that this strategy is readily able to detect RNA transcripts expressed at extremely low frequencies (Figure 4). Figure 4. In parallel experiments detecting the same target molecules, the sensitivity and spatial context of massively multiplexed FISH technology (a, top row) is visibly higher and more refined compared to an array-based platform (a, bottom row). When quantifying the number of transcript molecules detected, massively multiplexed FISH technology detects 70× more transcripts (b). Courtesy of Vizgen. Sequential imaging and analysis Once the gene panel and probe library have been generated, data capture requires sequential imaging of a sample on a high-resolution fluorescence microscope. A sample of interest, which may be either a fixed tissue slice or a cell culture, is first stained with the library of encoding probes, which hybridize to the RNA species in the panel. Subsequently, a series of sequential steps illuminate each barcode. The sample is first stained with a fluorescent probe set. It is imaged and then washed to remove the fluorescent probe, and then it is stained again with a new readout probe set to initiate the cycle all over again. The cycle repeats until all bits of the barcode have been recorded, resulting in multiplexed FISH data (Figure 5). Each barcode is error-resistant because several bits of each one are unique from those of all others. Thus, a target molecule can still be accurately identified despite some binding or detection errors occurring during imaging. The optical signals visualized during sequential imaging appear precisely where each target RNA molecule is localized in the tissue or cell, and their intensity may be used to quantify the number of molecules in that location. Figure 5. Massively multiplexed FISH experiments are highly flexible and can be performed on virtually any tissue type. Their utility has been proven in many scientifically and medically relevant tissue types, including those from mouse and human. PBMC: peripheral blood mononuclear cells. Courtesy of Vizgen. Computation methods must be used to assemble and interpret the large data sets, such as those collected from a massively multiplexed FISH experiment. One challenge lies in the need to streamline these methods, to expand them to pull more information out of every data set and to improve data storage to accommodate data from many experiments over time. As the field of spatial genomics matures and scientists continue to explore what is possible with these experiments, the accompanying data science is likely to match pace. Bigger, better, faster There is an ongoing push in the spatial genomics community to build technologies that can capture more information about more cells and quickly capture data at higher resolution than ever before. Current technologies, led by massively multiplexed FISH experiments, are being developed to meet this challenge. With powerful single-cell spatial genomics technologies, cell atlases will soon be rendered with ease, spurring on the next generation of genomic discoveries. For research, this technological development may help to untangle complex biological pathways and answer longstanding questions about how cell identity and organization are determined. Moreover, it may revamp scientists’ understanding of cells’ roles as rare RNA species that influence cells’ functionality become more apparent. This kind of research may also have direct applications in medicine in the future. For example, single-cell spatial genomics instruments may enable a detailed comparison between healthy and diseased tissue to reveal new disease biomarkers and therapeutic targets. Companion diagnostics could allow doctors to capture a holistic view of a patient’s disease based on cellular phenotypes observed in diseased tissue. Results would enable a personalized approach to optimizing treatment, predicting prognosis, and evaluating treatment response. Overall, as technologies continue to evolve within the rapidly growing field of spatial genomics, the benefit they could bring to science and medicine is limited only by the ingenuity of the research community as it explores new experimental horizons. Currently, there are systems in commercial use that exploit these developments. In the future, as the multiplexing capabilities and resolution of these technologies become increasingly advanced, the data that spatial genomics experiments can capture will represent an ever-clearer, direct line into both cellular and systems biology. Meet the author Jiang He, Ph.D., is senior director of scientific affairs at Vizgen. A co-founder of the company, he has been recognized with numerous awards, including being listed in Forbes 30 Under 30 in health care and named as an Aspen Institute Spotlight Health Scholar and a STAT Wunderkind. He completed his postdoctoral research at MIT and his doctorate at Harvard University in Xiaowei Zhuang’s lab; email: jiang@vizgen.com. References 1. A.M. Femino (1998). Visualization of single RNA transcripts in situ. Science, Vol. 280, No. 5363, pp. 585-590, www.doi.org/10.1126/science.280.5363.585. 2. K.H. Chen (2015). Spatially resolved, highly multiplexed RNA profiling in single cells. Science, Vol. 348, No. 6233, p. aaa6090, www.doi.org/10.1126/science.aaa6090. 3. G. Wang (2020). Spatial organization of the transcriptome in individual neurons. BioRxiv reprint, www.doi.org/10.1101/2020.12.07.414060. 4. J.R. Moffitt (2016). High-throughput single-cell gene-expression profiling with multiplexed error-robust fluorescence in situ hybridization. Proc Natl Acad Sci USA, Vol. 113, No. 39, pp. 11046-11051, www.doi.org/10.1073/pnas.1612826113.