Mol. Cells 2020; 43(7): 591-599
Published online June 4, 2020
https://doi.org/10.14348/molcells.2020.0020
© The Korean Society for Molecular and Cellular Biology
Correspondence to : kyoung.won@bric.ku.dk
Complex cell-to-cell communication underlies the basic processes essential for homeostasis in the given tissue architecture. Obtaining quantitative gene-expression of cells in their native context has significantly advanced through single-cell RNA sequencing technologies along with mechanical and enzymatic tissue manipulation. This approach, however, is largely reliant on the physical dissociation of individual cells from the tissue, thus, resulting in a library with unaccounted positional information. To overcome this, positional information can be obtained by integrating imaging and positional barcoding. Collectively, spatial transcriptomics strategies provide tissue architecture-dependent as well as position-dependent cellular functions. This review discusses the current technologies for spatial transcriptomics ranging from the methods combining mechanical dissociation and single-cell RNA sequencing to computational spatial re-mapping.
Keywords cellular communication, single-cell RNA, spatial transcriptomics, tissue architecture
Cell-to-cell communication is essential to maintain proper tissue homeostasis. Disruption of homeostatic cellular communication underlies many pathologic cellular transformations including cancer (Oktay et al., 2015). Studying the complexity of healthy tissue architecture and abnormal transformations both at the cellular and transcriptional level is important in improving the understanding of key pathways that can be targeted for therapeutic strategies. Recently, single cell RNA-sequencing (scRNAseq) technologies have revolutionized our understanding of gene expression by quantifying the transcriptome of individual cells. Moreover, the development of computational approaches to quantify large transcriptomic data alongside scRNAseq technology has provided transcriptomic information for previously uncharacterized cell types and has made it possible to study their dynamics at a population level (Grun et al., 2015; Patel et al., 2014; Pollen et al., 2014; Proserpio et al., 2016; Shalek et al., 2014; Trapnell et al., 2014).
However, scRNAseq technologies have the intrinsic limitation of losing positional information during tissue dissociation into single cells. Positional information is a critical aspect when studying tissue architecture to account for how physically interacting cells and signal exchanges maintain or alter homeostasis. Indeed, several strategies have been employed to provide positional information relative to transcriptomic data. In particular, tissue microdissection followed by RNA sequencing provides approximated positional information based on microdissected fragments (Combs and Eisen, 2013). Moreover, recent advances in a set of techniques collectively called spatial transcriptomics allow positional information to be identified at a single cell resolution.
By showing both expression and transcriptome at a single cell resolution, spatial transcriptomics can provide information such as tissue architecture and cell interactions. However, obtaining both position and expression information at a single cell resolution is still technically challenging. Historically, positional information has been obtained by physically taking cells positioned in a tissue, followed by expression profiling. Later, cells expressing specific genes were captured in the image to provide the position of cells as well as the gene expression levels (Chen et al., 2018). Also, a variety of barcode-based approaches to point the position of cells in the two-dimensional tissue section have also been introduced (Chen et al., 2018). Currently
One of the earliest methods to study spatial transcriptomics was through the use of tissue microdissection followed by RNA sequencing. In
Advancements such as laser capture microdissection (LCM) enabled a precise capture of targeted cells, or even single cells, while retaining intact tissue structure (Fig. 1B) (Datta et al., 2015). Subsequently, combining LCM and RNA sequencing was used to resolve spatially bound transcriptomic profiles of rare cell population (Nichterwitz et al., 2016). Comparably, geographical position sequencing (Geo-seq) is technique combining LCM with scRNAseq (Chen et al., 2017; Xue et al., 2019). Moreover, LCM has been used in various applications to provide position-based transcriptional information. For instance, LCM followed by RNAseq in mouse intestinal epithelium revealed the transcriptome of spatially zoned areas along the villus axis, which leads to spatial reconstruction of the tissue from scRNAseq data (Moor et al., 2018). LCM enables accurate separation of a small number of cells while preserving the tissue morphology. However, LCM procedures are labor-intensive and expensive to perform (Chung and Shen, 2015).
Multiplexed image-based transcriptomics is an emerging technology for spatial detection of RNAs. Particularly,
Another FISH technology development includes a cyclic approach to single-molecule fluorescence
Alternatively, the multiplexed error-robust FISH (MERFISH) technology, instead of barcoding with different color combinations, determines the presence or absence of fluorescence using a two-stop hybridization protocol (Fig. 1C) (Moffitt and Zhuang, 2016). Each RNA specie is encoded with a N-bit binary word from N rounds of hybridization that shows the presence or absence of a color. MERFISH, as well as other multiplexed smFISH-based RNA profiling methods, requires non-overlapping signals from individual RNAs. Improvements in microscopic detection combined with MERFISH has been used to overcome technical drawbacks such as signal detection and background correction (Wang et al., 2018a). Spatial information resolution was increased when MERFISH was combined with high-content imaging of sub-cellular structures to accurately determine the compartmentalization of RNAs (Xia et al., 2019).
To maximize the features of
Recent strategies using barcodes on native tissue on a slide, namely Spatial transcriptomics and Slide-seq, have been developed for high-resolution spatial resolution (Rodriques et al., 2019; Stahl et al., 2016). This approach dissects a histological section with a grid, where each spot is labelled with barcoded oligonucleotide primers to capture adjacent tissue mRNA. The resulting transcripts are then reverse transcribed to cDNA and positional information is contained within their respective barcodes (Fig. 1D). The Visium Spatial Gene Expression Solution (10× genomics) is based on the barcode-based approach (https://www.10xgenomics.com/products/spatial-gene-expression/). The current resolution of commercial Spatial Transcriptomics is limited to 100 µm, capturing an average of 3 to 30 cells per regular grid. High-definition spatial transcriptomics (HDST) uses 2-μm beads to enhance the resolution (Vickovic et al., 2019) while the Slide-seq technology uses 10-μm beads containing individual position barcode (Rodriques et al., 2019).
Although FISH and barcode-based approaches provide cellular positioning within a tissue, approaches to use cell contact information (thus provide relative positional information) have also been developed. This approach utilizes trasncriptomic information obtained from physically-interacting cells. ProximID is an approach where cells are gently dissociated to retain cells that are in physical contact, and then subsequently processed for RNAseq to obtain transcriptomic information of doublets or triplets of interacting cells (Fig. 1E) (Boisset et al., 2018). In parallel, RNAs from single cells are also sequenced (scRNAseq), which provide the reference information of the doublets and triplets. Without microdissection, ProximID identified that Tac1+ enteroendocrine cell–Lgr5+ stem cells interact in the intestinal crypt (Boisset et al., 2018), thus providing insights to previously uncharacterized interactions that may play a role in stem cell driven regeneration. ProximID collects the interacting cells that survived mild dissociation therefor this method can potentially be biased towards collecting strongly interacting cells. Similarly, paired cell RNAseq (pcRNAseq) sequenced RNA from cellular doublets composed of liver endothelial cells (LECs) and hepatocytes generated from extensive flow cytometric sorting (Halpern et al., 2018). pcRNAseq and parallel scRNAseq allowed the identification of zone-dependent LEC gene expression relative to the hepatocyte gradient (Halpern et al., 2018). These approaches utilize the innate interactions between cells in the tissue. Similarly, PIC-seq sequences RNAs from physically interacting cells and identified the composition of the interacting cells and investigated genes that are differentially expressed when interacting with other cell types (Giladi et al., 2020).
Tissue architecture reconstruction is an algorithm-based strategy used for reverse identification of a cell’s spatial origin based on gene expression data and a positional reference (Fig. 1F). Tomo-Seq used slices of cryosectioned zebrafish embryo to reconstruct 3D expression patterns (Junker et al., 2014). Similarly, scRNAseq information was used to reconstruct zoned areas along the villus axis of the intestinal epithelium with respect to the location information provided by LCM (Moor et al., 2018). Computational algorithms have been developed to predict the positional origin of cells using a model trained using the
Cellular location is important to understand how cells interact with each other within a tissue. Current technologies in spatial transcriptomics such the ones mentioned in this review have provided vital information on the cellular identity, transcriptional changes, critical gene markers for cell communication and even previously uncharacterized cellular interactions relative to its position in the tissue. Spatial reconstruction allowed the detection of a comprehensive set of liver genes differentially expressed across the lobule (Halpern et al., 2017). The LECs, the interacting partner with hepatocytes, also exhibited differential gene expression across the lobule (Halpern et al., 2018). Moreover, proximity-based reconstruction of hematopoietic stem/progenitor cell’s (HPSC) microanatomy in bone marrow followed by scRNAseq post-transplantation allowed the retrieval of osteolineage cells (OLCs) that were proximal or distal to HPSCs (Silberstein et al., 2016). Through proximity-based profiling, the study identified differentially expressed genes in OLCs affected by nearby HSPC and niche factors regulating HSPC quiescence
Spatial transcriptomic methods can serve as a powerful tool in capturing cell-to-cell properties at a tissue-level resolution. Using spatial transcriptomics enabled the quantitative elucidation of important cellular dynamics. Besides the cell and the tissues that were used to develop the methodology in Table 1, the number of applications for spatial transcriptomics are increasing. For instance, spatial quantification of neural and mesodermal lineage trajectories from Tomo-Seq was associated with the anterior-posterior axis of gastruloids developmental (van den Brink et al., 2020). Barcode based spatial transcriptomics have been applied to detect spatial crosstalk in Alzheimer’s disease wherein novel cellular interactions were uncovered between microglia and astroglial cells in amyloid-β plaque niches (Chen et al., 2019). Also, these approaches have been applied to study spatial localization of cell population in pancreatic ductal adenocarcinoma (Moncada et al., 2018) and HER2+ breast tumors (Salmén et al., 2018).
Studies on cellular identity combined with their native location has been significantly advanced by scRNAseq and spatial transcriptomic technologies. This review highlights the current technological developments that have refined spatial transcriptomics. Pioneering image-based approaches use FISH to visualize cellular localization. However, overlaps between fluorescent signals have been a major limitation in image resolution (Burgess, 2019). This was surmounted by barcoding-based improvements that lead to the development of 2-μm beads for HDST to measure single cell transcriptomic profiles (Vickovic et al., 2019). Although combining scRNAseq and barcoding-based approaches can determine transcriptome-wide profiles, they are still currently limited by their lack of sensitivity towards RNA detection (Kolodziejczyk et al., 2015) thus providing opportunities for further technological refinements.
As cells constantly communicate with other cells within a tissue, the study of cell communication will become more popular using scRNAseq and spatial transcriptomics. Previous approaches to study cell communication relied on the co-expression of ligand-receptor pairs between cell types (Kumar et al., 2018; Skelly et al., 2018). Neighboring cells communicate via complex paracrine signaling networks (Roy and Kornberg, 2015). Studying paracrine signaling using single cell resolution spatial transcriptomics provides an opportunity to enhance our understanding of cell communication.
Subsequently, we can then begin to question spatio-temporal changes of cells during development or disease progression. The majority of current spatial transcriptomics required a section of a tissue and cannot reveal spatio-temporal changes effectively. Attempts such as 4D-seq (unpblished; https://techtransfer.universityofcalifornia.edu/NCD/30312.html) uses two-photon microscopy and DNA-labelling to capture cellular spatio-temporal variation within a tissue prior to scRNA-seq analysis. As technical resolution continually improves, it is possible to foresee studies integrating high-content 4D image data with scRNAseq in order to gain a deeper understanding of spatio-temporal cellular regulation.
We are grateful to Sung-Whan Bae for graphic illustration and Patrick Martin for proofreading of the manuscript. The Novo Nordisk Foundation Center for Stem Cell Biology is supported by a Novo Nordisk Foundation grant (NNF17CC0027852). This work is also supported by Lundbeck Foundation Experiment grant (R324-2019-1649) to K.J.W. and Marie Sklodowska Curie COFUND Fellowship to J.M.T.
Both J.M.T. and K.J.W. wrote the manuscript.
The authors have no potential conflicts of interest to disclose.
Summary of selected technologies for spatial profiling of cells
Approach | Technology | Input material | Experimental method | Quantification method | Representative detection sensitivity | Detection rangea | Reference |
---|---|---|---|---|---|---|---|
LCM-seq | LCM-seq | Primary mouse brain and spinal cord tissue; | Laser capture microdissection | NGS data analysis (DESeq2 + GO analysis) | ~1,743 to 14,893 genes per 0.1 RPKM | SG to HT | (Combs and Eisen, 2013) |
FISH-based | smFISH | A549 and CHO cell line; | Fluorescence imaging + photo-bleaching on fixed cells | Probe-based computational identification of mRNA targets | ~3 mRNA species per cell | SG | (Halpern et al., 2017) |
MERFISH | IMR90 cell line; | Multiplexed fluorescence imaging of target probes on fixed cells (+ clearing) | Probe-based encoding + GO analysis | ~100 to 1,000 RNA species per cell | SG to MT | (Salmén et al., 2018) | |
seqFISH+ | NIH3T3 cell line; | Sequential fluorescence of pseudocolor probes | Probe-based encoding + scRNA-seq-based spatial localization mapping | ~10,000 genes per cell | SG to HT | (Kumar et al., 2018) | |
In situ sequencing (ISS) barcode-based | FISSEQ | HeLa, 293A, COS1, U2OS, iPSC, primary fibroblasts and bipolar neurons cell lines; | Reverse transcript probes + sequence-by-ligation | Probe-based calling + 3D image deconvolution | ~200 to 400 mRNA per cell; scalable 5X | MT | (McFaline-Figueroa et al., 2019) |
STARmap | Primary mouse cortical neuron cells; | Hydrogel-based isolation of target probes + SEDAL sequencing | Probe-based calling + 2D/3D cell segmentation + differential gene expression analysis | ~160 to 1,020 genes simultaneously; scalable to ~30,000 cells | MT to HT | (Moffitt and Zhuang, 2016) | |
Spatial and single-cell sequencing-based | Spatial reconstruction from single-cell transcriptomics (Seurat) | Danio rerio embryo tissue | Tissue dissociation + strand-specific, scRNA-seq modified from SMART protocol | NGS analysis + spatial location inference | Spatial reconstruction from 851 single-cell reference | HT | (Patel et al., 2014) |
Spatial transcriptomics | Primary mouse olfactory bulbs and brain tissue; | Spatial oligodT barcode array + cDNA synthesis + RNA-seq | Transcriptome analysis | On surface: 9.6 M unique transcripts per 400 M reads | HT | (Moncada et al., 2018) | |
Spatially-interacting cells | pcRNAseq (to detect liver zonation) | Primary mouse liver tissue; Primary mouse hepatocyte cells | Mild dissociation + single-cell and paired cell sorting + MARS-Seq | ScRNAseq analysis + landmark gene identification + spatial zonation inference | ~70 landmark genes | MT to HT | (Pollen et al., 2014) |
ProximID (to detect interacting cell network) | Primary mouse bone marrow and fetal liver tissue; Primary mouse intestinal crypts | Mild dissociation + single-cell sorting or pipette dissociation + CEL-Seq | ScRNAseq analysis + cell interaction inference + network analysis | Example: ~17 to 78 simulated preferential cell interactions | SG to HT | (Nitzan et al., 2019) | |
PIC-Seq (to detect physical interactions) | Primary mouse spleen tissue | Mild dissociation + population-based low cytometric sorting + MARS-seq single-cell RNA sequencing | NGS analysis + computational calling of physically interacting cells + differential analysis | Example: ~348 differential genes from ~2,389 simulated physically interacting cells | HT | (Okamura-Oho et al., 2012) |
A549, adenocarcinomic human alveolar basal epithelial cells; CHO, epithelial cell line derived from the ovary of the Chinese hamster; IMR90, human foetal lung cells; U2OS, Human Bone Osteosarcoma Epithelial Cells; NIH3T3, mouse embryonic fibroblast cells; HeLa, human cervical cancer cells; COS1, African green monkey kidney fibroblast-like cell; 293A, human embryonic kidney cells; iPSC, induced pluripotent stem cells.
a Detection range: SG, single gene; MT, medium throughput (targeted transcript capture sequencing); HT, high throughput (non-targetted transcriptome-wide sequencing).
Mol. Cells 2020; 43(7): 591-599
Published online July 31, 2020 https://doi.org/10.14348/molcells.2020.0020
Copyright © The Korean Society for Molecular and Cellular Biology.
Joji Marie Teves1,2 and Kyoung Jae Won1,2,*
1Biotech Research and Innovation Centre (BRIC), University of Copenhagen, DK-2200 Copenhagen, Denmark, 2Novo Nordisk Foundation Center for Stem Cell Biology (DanStem), Faculty of Health Sciences, Faculty of Health and Medical Sciences, University of Copenhagen, DK-2200 Copenhagen, Denmark
Correspondence to:kyoung.won@bric.ku.dk
Complex cell-to-cell communication underlies the basic processes essential for homeostasis in the given tissue architecture. Obtaining quantitative gene-expression of cells in their native context has significantly advanced through single-cell RNA sequencing technologies along with mechanical and enzymatic tissue manipulation. This approach, however, is largely reliant on the physical dissociation of individual cells from the tissue, thus, resulting in a library with unaccounted positional information. To overcome this, positional information can be obtained by integrating imaging and positional barcoding. Collectively, spatial transcriptomics strategies provide tissue architecture-dependent as well as position-dependent cellular functions. This review discusses the current technologies for spatial transcriptomics ranging from the methods combining mechanical dissociation and single-cell RNA sequencing to computational spatial re-mapping.
Keywords: cellular communication, single-cell RNA, spatial transcriptomics, tissue architecture
Cell-to-cell communication is essential to maintain proper tissue homeostasis. Disruption of homeostatic cellular communication underlies many pathologic cellular transformations including cancer (Oktay et al., 2015). Studying the complexity of healthy tissue architecture and abnormal transformations both at the cellular and transcriptional level is important in improving the understanding of key pathways that can be targeted for therapeutic strategies. Recently, single cell RNA-sequencing (scRNAseq) technologies have revolutionized our understanding of gene expression by quantifying the transcriptome of individual cells. Moreover, the development of computational approaches to quantify large transcriptomic data alongside scRNAseq technology has provided transcriptomic information for previously uncharacterized cell types and has made it possible to study their dynamics at a population level (Grun et al., 2015; Patel et al., 2014; Pollen et al., 2014; Proserpio et al., 2016; Shalek et al., 2014; Trapnell et al., 2014).
However, scRNAseq technologies have the intrinsic limitation of losing positional information during tissue dissociation into single cells. Positional information is a critical aspect when studying tissue architecture to account for how physically interacting cells and signal exchanges maintain or alter homeostasis. Indeed, several strategies have been employed to provide positional information relative to transcriptomic data. In particular, tissue microdissection followed by RNA sequencing provides approximated positional information based on microdissected fragments (Combs and Eisen, 2013). Moreover, recent advances in a set of techniques collectively called spatial transcriptomics allow positional information to be identified at a single cell resolution.
By showing both expression and transcriptome at a single cell resolution, spatial transcriptomics can provide information such as tissue architecture and cell interactions. However, obtaining both position and expression information at a single cell resolution is still technically challenging. Historically, positional information has been obtained by physically taking cells positioned in a tissue, followed by expression profiling. Later, cells expressing specific genes were captured in the image to provide the position of cells as well as the gene expression levels (Chen et al., 2018). Also, a variety of barcode-based approaches to point the position of cells in the two-dimensional tissue section have also been introduced (Chen et al., 2018). Currently
One of the earliest methods to study spatial transcriptomics was through the use of tissue microdissection followed by RNA sequencing. In
Advancements such as laser capture microdissection (LCM) enabled a precise capture of targeted cells, or even single cells, while retaining intact tissue structure (Fig. 1B) (Datta et al., 2015). Subsequently, combining LCM and RNA sequencing was used to resolve spatially bound transcriptomic profiles of rare cell population (Nichterwitz et al., 2016). Comparably, geographical position sequencing (Geo-seq) is technique combining LCM with scRNAseq (Chen et al., 2017; Xue et al., 2019). Moreover, LCM has been used in various applications to provide position-based transcriptional information. For instance, LCM followed by RNAseq in mouse intestinal epithelium revealed the transcriptome of spatially zoned areas along the villus axis, which leads to spatial reconstruction of the tissue from scRNAseq data (Moor et al., 2018). LCM enables accurate separation of a small number of cells while preserving the tissue morphology. However, LCM procedures are labor-intensive and expensive to perform (Chung and Shen, 2015).
Multiplexed image-based transcriptomics is an emerging technology for spatial detection of RNAs. Particularly,
Another FISH technology development includes a cyclic approach to single-molecule fluorescence
Alternatively, the multiplexed error-robust FISH (MERFISH) technology, instead of barcoding with different color combinations, determines the presence or absence of fluorescence using a two-stop hybridization protocol (Fig. 1C) (Moffitt and Zhuang, 2016). Each RNA specie is encoded with a N-bit binary word from N rounds of hybridization that shows the presence or absence of a color. MERFISH, as well as other multiplexed smFISH-based RNA profiling methods, requires non-overlapping signals from individual RNAs. Improvements in microscopic detection combined with MERFISH has been used to overcome technical drawbacks such as signal detection and background correction (Wang et al., 2018a). Spatial information resolution was increased when MERFISH was combined with high-content imaging of sub-cellular structures to accurately determine the compartmentalization of RNAs (Xia et al., 2019).
To maximize the features of
Recent strategies using barcodes on native tissue on a slide, namely Spatial transcriptomics and Slide-seq, have been developed for high-resolution spatial resolution (Rodriques et al., 2019; Stahl et al., 2016). This approach dissects a histological section with a grid, where each spot is labelled with barcoded oligonucleotide primers to capture adjacent tissue mRNA. The resulting transcripts are then reverse transcribed to cDNA and positional information is contained within their respective barcodes (Fig. 1D). The Visium Spatial Gene Expression Solution (10× genomics) is based on the barcode-based approach (https://www.10xgenomics.com/products/spatial-gene-expression/). The current resolution of commercial Spatial Transcriptomics is limited to 100 µm, capturing an average of 3 to 30 cells per regular grid. High-definition spatial transcriptomics (HDST) uses 2-μm beads to enhance the resolution (Vickovic et al., 2019) while the Slide-seq technology uses 10-μm beads containing individual position barcode (Rodriques et al., 2019).
Although FISH and barcode-based approaches provide cellular positioning within a tissue, approaches to use cell contact information (thus provide relative positional information) have also been developed. This approach utilizes trasncriptomic information obtained from physically-interacting cells. ProximID is an approach where cells are gently dissociated to retain cells that are in physical contact, and then subsequently processed for RNAseq to obtain transcriptomic information of doublets or triplets of interacting cells (Fig. 1E) (Boisset et al., 2018). In parallel, RNAs from single cells are also sequenced (scRNAseq), which provide the reference information of the doublets and triplets. Without microdissection, ProximID identified that Tac1+ enteroendocrine cell–Lgr5+ stem cells interact in the intestinal crypt (Boisset et al., 2018), thus providing insights to previously uncharacterized interactions that may play a role in stem cell driven regeneration. ProximID collects the interacting cells that survived mild dissociation therefor this method can potentially be biased towards collecting strongly interacting cells. Similarly, paired cell RNAseq (pcRNAseq) sequenced RNA from cellular doublets composed of liver endothelial cells (LECs) and hepatocytes generated from extensive flow cytometric sorting (Halpern et al., 2018). pcRNAseq and parallel scRNAseq allowed the identification of zone-dependent LEC gene expression relative to the hepatocyte gradient (Halpern et al., 2018). These approaches utilize the innate interactions between cells in the tissue. Similarly, PIC-seq sequences RNAs from physically interacting cells and identified the composition of the interacting cells and investigated genes that are differentially expressed when interacting with other cell types (Giladi et al., 2020).
Tissue architecture reconstruction is an algorithm-based strategy used for reverse identification of a cell’s spatial origin based on gene expression data and a positional reference (Fig. 1F). Tomo-Seq used slices of cryosectioned zebrafish embryo to reconstruct 3D expression patterns (Junker et al., 2014). Similarly, scRNAseq information was used to reconstruct zoned areas along the villus axis of the intestinal epithelium with respect to the location information provided by LCM (Moor et al., 2018). Computational algorithms have been developed to predict the positional origin of cells using a model trained using the
Cellular location is important to understand how cells interact with each other within a tissue. Current technologies in spatial transcriptomics such the ones mentioned in this review have provided vital information on the cellular identity, transcriptional changes, critical gene markers for cell communication and even previously uncharacterized cellular interactions relative to its position in the tissue. Spatial reconstruction allowed the detection of a comprehensive set of liver genes differentially expressed across the lobule (Halpern et al., 2017). The LECs, the interacting partner with hepatocytes, also exhibited differential gene expression across the lobule (Halpern et al., 2018). Moreover, proximity-based reconstruction of hematopoietic stem/progenitor cell’s (HPSC) microanatomy in bone marrow followed by scRNAseq post-transplantation allowed the retrieval of osteolineage cells (OLCs) that were proximal or distal to HPSCs (Silberstein et al., 2016). Through proximity-based profiling, the study identified differentially expressed genes in OLCs affected by nearby HSPC and niche factors regulating HSPC quiescence
Spatial transcriptomic methods can serve as a powerful tool in capturing cell-to-cell properties at a tissue-level resolution. Using spatial transcriptomics enabled the quantitative elucidation of important cellular dynamics. Besides the cell and the tissues that were used to develop the methodology in Table 1, the number of applications for spatial transcriptomics are increasing. For instance, spatial quantification of neural and mesodermal lineage trajectories from Tomo-Seq was associated with the anterior-posterior axis of gastruloids developmental (van den Brink et al., 2020). Barcode based spatial transcriptomics have been applied to detect spatial crosstalk in Alzheimer’s disease wherein novel cellular interactions were uncovered between microglia and astroglial cells in amyloid-β plaque niches (Chen et al., 2019). Also, these approaches have been applied to study spatial localization of cell population in pancreatic ductal adenocarcinoma (Moncada et al., 2018) and HER2+ breast tumors (Salmén et al., 2018).
Studies on cellular identity combined with their native location has been significantly advanced by scRNAseq and spatial transcriptomic technologies. This review highlights the current technological developments that have refined spatial transcriptomics. Pioneering image-based approaches use FISH to visualize cellular localization. However, overlaps between fluorescent signals have been a major limitation in image resolution (Burgess, 2019). This was surmounted by barcoding-based improvements that lead to the development of 2-μm beads for HDST to measure single cell transcriptomic profiles (Vickovic et al., 2019). Although combining scRNAseq and barcoding-based approaches can determine transcriptome-wide profiles, they are still currently limited by their lack of sensitivity towards RNA detection (Kolodziejczyk et al., 2015) thus providing opportunities for further technological refinements.
As cells constantly communicate with other cells within a tissue, the study of cell communication will become more popular using scRNAseq and spatial transcriptomics. Previous approaches to study cell communication relied on the co-expression of ligand-receptor pairs between cell types (Kumar et al., 2018; Skelly et al., 2018). Neighboring cells communicate via complex paracrine signaling networks (Roy and Kornberg, 2015). Studying paracrine signaling using single cell resolution spatial transcriptomics provides an opportunity to enhance our understanding of cell communication.
Subsequently, we can then begin to question spatio-temporal changes of cells during development or disease progression. The majority of current spatial transcriptomics required a section of a tissue and cannot reveal spatio-temporal changes effectively. Attempts such as 4D-seq (unpblished; https://techtransfer.universityofcalifornia.edu/NCD/30312.html) uses two-photon microscopy and DNA-labelling to capture cellular spatio-temporal variation within a tissue prior to scRNA-seq analysis. As technical resolution continually improves, it is possible to foresee studies integrating high-content 4D image data with scRNAseq in order to gain a deeper understanding of spatio-temporal cellular regulation.
We are grateful to Sung-Whan Bae for graphic illustration and Patrick Martin for proofreading of the manuscript. The Novo Nordisk Foundation Center for Stem Cell Biology is supported by a Novo Nordisk Foundation grant (NNF17CC0027852). This work is also supported by Lundbeck Foundation Experiment grant (R324-2019-1649) to K.J.W. and Marie Sklodowska Curie COFUND Fellowship to J.M.T.
Both J.M.T. and K.J.W. wrote the manuscript.
The authors have no potential conflicts of interest to disclose.
. Summary of selected technologies for spatial profiling of cells.
Approach | Technology | Input material | Experimental method | Quantification method | Representative detection sensitivity | Detection rangea | Reference |
---|---|---|---|---|---|---|---|
LCM-seq | LCM-seq | Primary mouse brain and spinal cord tissue; | Laser capture microdissection | NGS data analysis (DESeq2 + GO analysis) | ~1,743 to 14,893 genes per 0.1 RPKM | SG to HT | (Combs and Eisen, 2013) |
FISH-based | smFISH | A549 and CHO cell line; | Fluorescence imaging + photo-bleaching on fixed cells | Probe-based computational identification of mRNA targets | ~3 mRNA species per cell | SG | (Halpern et al., 2017) |
MERFISH | IMR90 cell line; | Multiplexed fluorescence imaging of target probes on fixed cells (+ clearing) | Probe-based encoding + GO analysis | ~100 to 1,000 RNA species per cell | SG to MT | (Salmén et al., 2018) | |
seqFISH+ | NIH3T3 cell line; | Sequential fluorescence of pseudocolor probes | Probe-based encoding + scRNA-seq-based spatial localization mapping | ~10,000 genes per cell | SG to HT | (Kumar et al., 2018) | |
In situ sequencing (ISS) barcode-based | FISSEQ | HeLa, 293A, COS1, U2OS, iPSC, primary fibroblasts and bipolar neurons cell lines; | Reverse transcript probes + sequence-by-ligation | Probe-based calling + 3D image deconvolution | ~200 to 400 mRNA per cell; scalable 5X | MT | (McFaline-Figueroa et al., 2019) |
STARmap | Primary mouse cortical neuron cells; | Hydrogel-based isolation of target probes + SEDAL sequencing | Probe-based calling + 2D/3D cell segmentation + differential gene expression analysis | ~160 to 1,020 genes simultaneously; scalable to ~30,000 cells | MT to HT | (Moffitt and Zhuang, 2016) | |
Spatial and single-cell sequencing-based | Spatial reconstruction from single-cell transcriptomics (Seurat) | Danio rerio embryo tissue | Tissue dissociation + strand-specific, scRNA-seq modified from SMART protocol | NGS analysis + spatial location inference | Spatial reconstruction from 851 single-cell reference | HT | (Patel et al., 2014) |
Spatial transcriptomics | Primary mouse olfactory bulbs and brain tissue; | Spatial oligodT barcode array + cDNA synthesis + RNA-seq | Transcriptome analysis | On surface: 9.6 M unique transcripts per 400 M reads | HT | (Moncada et al., 2018) | |
Spatially-interacting cells | pcRNAseq (to detect liver zonation) | Primary mouse liver tissue; Primary mouse hepatocyte cells | Mild dissociation + single-cell and paired cell sorting + MARS-Seq | ScRNAseq analysis + landmark gene identification + spatial zonation inference | ~70 landmark genes | MT to HT | (Pollen et al., 2014) |
ProximID (to detect interacting cell network) | Primary mouse bone marrow and fetal liver tissue; Primary mouse intestinal crypts | Mild dissociation + single-cell sorting or pipette dissociation + CEL-Seq | ScRNAseq analysis + cell interaction inference + network analysis | Example: ~17 to 78 simulated preferential cell interactions | SG to HT | (Nitzan et al., 2019) | |
PIC-Seq (to detect physical interactions) | Primary mouse spleen tissue | Mild dissociation + population-based low cytometric sorting + MARS-seq single-cell RNA sequencing | NGS analysis + computational calling of physically interacting cells + differential analysis | Example: ~348 differential genes from ~2,389 simulated physically interacting cells | HT | (Okamura-Oho et al., 2012) |
A549, adenocarcinomic human alveolar basal epithelial cells; CHO, epithelial cell line derived from the ovary of the Chinese hamster; IMR90, human foetal lung cells; U2OS, Human Bone Osteosarcoma Epithelial Cells; NIH3T3, mouse embryonic fibroblast cells; HeLa, human cervical cancer cells; COS1, African green monkey kidney fibroblast-like cell; 293A, human embryonic kidney cells; iPSC, induced pluripotent stem cells..
a Detection range: SG, single gene; MT, medium throughput (targeted transcript capture sequencing); HT, high throughput (non-targetted transcriptome-wide sequencing)..