REVIEW

www.advancedscience.com

Spatial Transcriptomics: Technical Aspects of Recent

Developments and Their Applications in Neuroscience and
Cancer Research
Han-Eol Park, Song Hyun Jo, Rosalind H. Lee, Christian P. Macks, Taeyun Ku,
Jihwan Park, Chung Whan Lee, Junho K. Hur, and Chang Ho Sohn*

primarily through cell-to-cell interactions

Spatial transcriptomics is a newly emerging field that enables and subsequent signaling, which is gov-
high-throughput investigation of the spatial localization of transcripts and erned by the relative positions of cells
related analyses in various applications for biological systems. By within the embryo.[1,2] The spatial orga-
transitioning from conventional biological studies to “in situ” biology, spatial nization of tissues regulates the expres-
sion of transcription factors related to dif-
transcriptomics can provide transcriptome-scale spatial information.
ferentiation and ultimately generates a ro-
Currently, the ability to simultaneously characterize gene expression profiles bust organization of cellular structures re-
of cells and relevant cellular environment is a paradigm shift for biological lated to their functions.[3–5] Another exam-
studies. In this review, recent progress in spatial transcriptomics and its ple that illustrates the importance of spa-
applications in neuroscience and cancer studies are highlighted. Technical tial organization is cancer tissue, in which
cells actively interact with the surround-
aspects of existing technologies and future directions of new developments
ing tumor microenvironment to generate
(as of March 2023), computational analysis of spatial transcriptome data, suppressive conditions that block the ac-
application notes in neuroscience and cancer studies, and discussions tion of immune cells, thereby bypassing
regarding future directions of spatial multi-omics and their expanding roles in immune defense mechanisms and facili-
biomedical applications are emphasized. tating proliferation.[5,6] To understand the
complexity of biological systems ranging
from various physiological phenomena to
the pathological principles of diseases, it is
1. Introduction necessary to assess the functions of individual cells and their in-
teractions to orchestrate complex functions of tissues and organs.
1.1. Spatial Transcriptomics: Emerging Technology Strategies for examining these biological principles include ex-
ploring cells that exist in the tissue (cell-type inventory) and their
Each cell in a multicellular organism interacts with the surround- spatial arrangement and interaction with each other (understand-
ing environment. Stem cells differentiate during development ing their spatial organization).

H.-E. Park, C. P. Macks, C. H. Sohn H.-E. Park

Center for Nanomedicine School of Biological Sciences
Institute for Basic Science Seoul National University
Yonsei University Seoul 08826, Republic of Korea
Seoul 03722, Republic of Korea S. H. Jo, T. Ku
E-mail: chsohn@yonsei.ac.kr Graduate School of Medical Science and Engineering
H.-E. Park, C. P. Macks, C. H. Sohn Korea Advanced Institute of Science and Technology (KAIST)
Graduate Program in Nanobiomedical Engineering Daejeon 34141, Republic of Korea
Advanced Science Institute R. H. Lee, J. Park
Yonsei University School of Life Sciences
Seoul 03722, Republic of Korea Gwangju Institute of Science and Technology (GIST)
Gwangju 61005, Republic of Korea
C. W. Lee
The ORCID identification number(s) for the author(s) of this article Department of Chemistry
can be found under https://doi.org/10.1002/advs.202206939 Gachon University
Seongnam, Gyeonggi-do 13120, Republic of Korea
© 2023 The Authors. Advanced Science published by Wiley-VCH GmbH.
This is an open access article under the terms of the Creative Commons J. K. Hur
Attribution License, which permits use, distribution and reproduction in Department of Genetics
any medium, provided the original work is properly cited. College of Medicine
Hanyang University
DOI: 10.1002/advs.202206939 Seoul 04763, Republic of Korea

Adv. Sci. 2023, 10, 2206939 2206939 (1 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Molecular expression patterns of diverse biological states have most cases, these attempts were close to the preliminary, proof-
been analyzed by RNA sequencing owing to the efficient and sen- of-concept stage.[19] Initial efforts for multiplexed RNA profiling
sitive detection of RNA by the simple amplification of nucleic focused on improving the barcoding capacity using 1) spectral
acids by polymerase chain reaction (PCR) and high-throughput barcoding by super-resolution microscopy[20] and 2) sequential
readouts by next-generation sequencing (NGS). RNA sequencing barcoding using DNase I (Figure 1a).[21] Without barcoding, dif-
has served as one of the major approaches for basic biological re- ferent RNAs can be resolved near diffraction-limited spots; how-
search and clinical diagnosis because the transcriptome serves ever, the barcoding capacity can be further extended if one can
as a blueprint of the proteome, the actual functional proteins of resolve overlapping signals. Lubeck et al.[20] first demonstrated
the cell, and therefore reflects the current cellular state.[7] Fur- a multiplexed barcoding scheme for resolving the RNA expres-
thermore, recent advances in single-cell transcriptome analysis sion of 32 target genes in a single yeast cell by super-resolution
techniques allow researchers to analyze transcriptomes with un- microscopy. This approach, however, requires expert-level knowl-
precedentedly high throughput and single-cell resolution.[8,9] De- edge of optics for accurate implementation due to the techni-
spite the improvements in sequencing technologies, single-cell cal difficulties of its setup. Moreover, the super-resolution ver-
transcriptome analysis techniques are still limited in that all in- sion of the spectral barcoding method consumes a very long
formation related to the spatial organization of cells in the tissue time for imaging (i.e., a yeast cell takes 30 min to barcode 32
is permanently lost owing to tissue dissociation. genes), precluding the scalable application of this method to real-
Spatial transcriptomics provides information on the spa- world samples. An improved protocol for a sequential barcod-
tial distribution of gene expression profiles, thereby eluci- ing scheme by DNase I dramatically enhanced the performance
dating interesting features previously not revealed by single- of repeated probing for multiplexed RNA FISH and simplified
cell RNA sequencing methods that lack spatial information. the experimental procedure so that multiplexed barcoding ex-
Several high-quality review papers already exist on spatial periments could be executed with an ordinary epifluorescence
transcriptomics.[10–14] Therefore, in this review article, we instead microscope.[21] Although the initial report on the DNase I-based
focus on providing a comprehensive review of the technical dif- sequential barcoding scheme only presented its application to
ferences between sequencing- and imaging-based methodolo- yeast cells, the concept of multi-round imaging and registered
gies, recent advances in the spatial approach utilizing spatial data barcode calling for smFISH-based experiments was continuously
and their computational interpretation, and current research and adapted to the later versions of highly multiplexed FISH experi-
future perspectives in various fields of application in biological ments.
studies, including neuroscience and cancer, and in biomedical In 2015 and 2016, continued efforts to develop highly mul-
and clinical studies of disease. tiplexed FISH methods increased the number of detected
genes from a dozen to several hundred and then up to one
thousand transcripts per cell [multiplexed error-robust FISH
2. Imaging and Sequencing-based Spatial (MERFISH)[22] and sequential FISH (seqFISH),[23] Figure 1a,b].
Transcriptomics Methods To encode as many transcripts as possible, sequential barcode
schemes were implemented using repeated 3–4 color imaging
2.1. Imaging-Based Methods [Fluorescence In Situ Hybridization cycling platforms. The key to successful and highly multiplexed
and In Situ Sequencing-Based Methods] barcoding is to seek schemes for improving barcode calling rates
to distinguish probe signals from noise- and crosstalk-prone
To detect transcripts from cells, pioneering studies reported raw images and to minimize unwanted dye quenching during
single-molecule fluorescence in situ hybridization (smFISH) fast imaging cycles. However, the barcoding schemes for ever-
methods that employed in situ hybridization of reverse- increasing multiplexity are not always compatible with traditional
complementary oligo probes conjugated with fluorophores.[15,16] highly expressed marker genes because these high-copy genes
These smFISH methods facilitated the detection of target RNAs easily saturate the fluorescence field of view, resulting in poor
with high specificity and sensitivity at the single-molecule level. barcode calling rates due to signal overlap.[24,25] To evenly dis-
smFISH also provides single-cell and subcellular resolution with tribute the “loading” of gene expression in each fluorescence
optimized protocols for cell cultures and thin (<20 μm) tis- channel, careful estimation of expression from bulk sequenc-
sues respectively. Recently, spatial applications of smFISH have ing results is employed to design orders of imaging and to de-
demanded more scalable platforms in order to investigate bi- termine which genes can be co-imaged. Super-resolution imag-
ologically meaningful dimensions according to their detection ing by localization microscopy was employed to resolve more
targets.[17] Fast cycling of probe hybridization allows the spa- transcripts in each diffraction-limited spot, which finally en-
tial investigation of relatively large areas with high specificity abled true transcriptome-level detection in the order of ten thou-
and sensitivity when targeting 20–40 highly expressed marker sand transcripts per cell (seqFISH+) (Figure 1c).[26] Error-robust
genes (ouroboros smFISH, or osmFISH) without barcoding.[18] barcoding schemes, corrections for optical aberrations, lower-
However, the limited multiplexing capacity of smFISH, caused ing autofluorescence by tissue clearing, and semi-automated
by the spectral overlap of fluorescent dyes, precludes efficient microfluidics-based staining and imaging cycling systems fur-
transcriptome-level spatial investigations when creating a profile ther improved the quality of the resulting datasets in compari-
of the transcriptional status of cells in a tissue. son to those obtained by sequencing-based methodologies (dis-
Separating the fluorescence signals obtained from individual cussed in the next section) (Figure 1b).[27,28] From the point of
transcripts is the key to highly increasing the multiplexing capac- view of biomedical and translational/clinical applications, know-
ity. This idea has been previously implemented for DNA, but in ing whether a particular method can cover clinically meaningful

Adv. Sci. 2023, 10, 2206939 2206939 (2 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Figure 1. Imaging-based spatial transcriptomics methods. a) seqFISH by DNaseI-based digestion and sequential staining/imaging cycles to decode
transcripts in space. b) MERFISH employing error correction in barcode assignment for robust barcode calling in noisy FISH-based images. c) seqFISH+
for genome-scale transcriptome investigation by dilution of fluorescent signals, separating individual transcripts into fluorescent spectra, and employing
20 probes per each encoding round. d) in situ sequencing methods by sequencing by ligation (ISS, FISSEQ, STARmap) and sequencing by synthesis
(BaristaSeq). e) SOLiD sequencing of cDNA sequence by FISSEQ while cross-linking cDNA and amplicon generated by rolling-circle amplification (RCA)
to adjacent proteins. f) STARmap with SNAIL probes and SEDAL sequencing for identifying gene-specific identifiers. The polymerization of amplicons
with acrylamide moieties introduced by N-acryloxysuccinimide (NAS) into the hydrogel network and optical clearing by hydrogel-histochemistry enables
spatial transcriptome detection in thick tissues.

ranges of tissue size and thickness is critical. Although these mul- of a partial region of the cortex in a thin coronal tissue section
tiplexed FISH-based methods have excellent sensitivity and cov- of the mouse brain when performing 80 rounds of hybridiza-
erage of transcriptomes with subcellular resolutions, they all suf- tion/imaging to detect 10 000 transcripts. Even after this effort,
fer from substantially long imaging times, which limits practical a typical tissue section in these methods is only 10–20-μm thick,
tissue size and thickness. For example, in seqFISH+, an imag- and therefore, the cells at this thickness are mostly not intact. In
ing time of 1 week is required to image a single optical plain addition to the difficulty in registration and barcode calling from

Adv. Sci. 2023, 10, 2206939 2206939 (3 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

these large raw image datasets, the resulting data suffer from in- nals originating from rRNAs (Figure 1e). Additionally, random
complete analysis of single-cell expression profiles that are inher- hexamer priming results in a very poor yield after reverse tran-
ently non-intact in a spatial context. Future efforts should focus scription (0.2–1%), which inhibits the efficient detection of mR-
on developing fast imaging schemes by implementing simple NAs. FISSEQ-based methods also involve complex enzymatic re-
and easy signal amplification, which will eventually enable the actions, a week of long imaging time, and challenging data pro-
analysis of thick tissues by improving sample coverage. Easy and cessing owing to large raw data volumes. Therefore, despite map-
user-friendly interfaces in data acquisition and analysis for com- ping transcripts at the sub-cellular resolution, imaging-based
mercialization processes are crucial for facilitating widespread technologies suffer from several technical limitations that ham-
usage. Enhanced Electric FISH (EEL FISH) is an electrophoresis- per their clinical and biomedical applications for large tissues.
aided large tissue RNA sampling and multiplexed FISH study In contrast, hybridization-based ISS (HybISS)[37] and STARmap-
that attempts to reduce data acquisition time by transferring RNA based schemes allow the detection of signals by employing low
from 10-μm-thick tissue to the plane of an RNA capture slide for magnification objectives (20×, numerical aperture (NA) 0.8 air
imaging the transcriptome of thick tissues with an epifluores- for HybISS; 40×, NA 1.3 oil for STARmap), which enables the
cence microscope.[29] investigation of relatively large-sized tissues.[38] Directly target-
In contrast to sequential imaging of barcoded FISH probes, ing RNA with padlock probes to eliminate inefficient reverse
in situ sequencing (ISS) is an alternative approach for identify- transcription during cDNA synthesis can enhance the efficiency
ing a larger number of RNA-targeting probes by direct imaging of ISS [barcoded oligonucleotides ligated on RNA amplified for
of nucleotide sequences in situ (Figure 1d–f). In principle, both multiplexed and parallel insitu analyses (BOLORAMIS)[39] and
FISH and ISS provide similar transcriptome information at sub- hybridization-based RNA ISS (HybrISS)[40] ]. Moreover, commer-
cellular resolution. However, ISS can read nucleotide sequences cialized options for imaging-based spatial transcriptomics tech-
directly from tissues, which is a critical feature that allows the nologies will soon be available, including MERSCOPE[41] (Viz-
possibility for new applications such as in situ detection of sin- Gen, Cambridge, MA, United States; MERFISH), Xenium[42]
gle nucleotide polymorphisms. In 2013, the first ISS study em- (10X Genomics, Pleasanton, CA, United States; ISS and FIS-
ployed sequencing-by-ligation chemistry to read short sequences SEQ), products from Spatial Genomics (Pasadena, CA, United
of gene barcodes in situ.[30] In ISS, reverse-transcribed com- States; seqFISH), and GeoMX and CosMX from NanoString
plementary deoxyribonucleic acids (cDNAs) are hybridized with (Seattle, WA, United States).
padlock probes containing gene-specific barcode sequences, and
the padlock probe is ligated at the location of specific hybridiza-
tion before being amplified by rolling-circle amplification (RCA) 2.2. Sequencing-Based Methods
using a circularized padlock primer probe (Figure 1d). Sequen-
tial imaging by sequencing-by-ligation allows the identification NGS platforms have shown unprecedented success in produc-
of repeatedly amplified barcode sequences in situ. Fluorescentin ing massive genomic sequencing data because of their superior
situ sequencing (FISSEQ) detects RNA by employing sequencing sequencing capacity at a substantially reduced cost.[43] The lack
by oligonucleotide ligation and detection (SOLiD) chemistry to of spatial context in transcriptomic data, however, has led to the
directly read cDNA sequences synthesized by random hexamers development of new techniques capable of encoding spatial in-
and provide an unbiased examination of the whole transcriptome formation as nucleotide barcoding sequences. The spatial loca-
distribution (Figure 1e).[31,32] Spatially-resolved transcript ampli- tion of transcripts can be recovered using NGS by incorporat-
con readout mapping (STARmap) uses a novel in situ sequenc- ing spatial barcodes to construct RNA sequencing libraries. NGS-
ing chemistry called sequencing with error-reduction by dynamic based spatial transcriptomics approaches have superior through-
annealing and ligation (SEDAL) for the highly efficient synthesis put compared to imaging-based methodologies because the en-
of barcoded probe sequences (Figure 1f).[33] For volumetric inves- tire transcript information is detected and massively parallel-
tigations of in situ RNA distribution, STARmap uses CLARITY- processed. Additionally, NGS-based approaches do not require a
based hydrogel-tissue chemistry for sample processing to secure pre-targeted list of genes because the retrieval of transcript spatial
biomolecules and support the spatial architecture of tissues. Bar- distribution is accompanied by cDNA synthesis in an unbiased
code in situ targeted sequencing (BaristaSeq) utilizes Illumina and non-targeted manner.[35]
sequencing-by-synthesis chemistry to read barcode sequences in Early approaches for including the spatial context in NGS
situ with multiple rounds of imaging.[34] The optimized proto- performed microscopy-guided selection and isolation of re-
col has shown improved signal-to-noise ratios, thereby enabling gions of interest (ROIs) for analysis by full-read RNA se-
better detection during in situ synthesis of target-specific RCA quencing. Laser capture microdissection (LCM) is a widely
products (Figure 1d). used strategy for the physical dissection of tissue ROIs by
Since both FISH and ISS use pre-designed probes to label laser cutting.[44,45] LCM-seq,[44] and the more recent spatial-
target transcripts, only preset genes related to a certain experi- histopathological examination-linked epitranscriptomics con-
mental hypothesis are profiled, and this preset repertoire of tar- verged to transcriptomics with sequencing (Select-seq),[45] com-
get genes leads to biased detection of transcriptomes.[35,36] Al- bine LCM with Smart-seq2 for polyA+ RNA sequencing of select
though FISSEQ-synthesized cDNA uses random hexamers and cell populations that are identified for dissection by immunohis-
direct reading of cDNA sequences by SOLiD chemistry to avoid tochemistry (IHC). Tomo-seq uses a tomography-inspired sec-
target selection bias and enable de novo discovery of expressed tioning approach to resolve the transcriptome of anatomical
genes, it also shows low efficiency of gene detection owing to ROIs.[46] Through a combination of LCM-seq and Tomo-seq,
overwhelming occupancy of the fluorescence channels by sig- Geo-seq (geographical position sequencing) enables 3D tran-

Adv. Sci. 2023, 10, 2206939 2206939 (4 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

scriptome analysis of select ROIs and single cells.[47] For an al- rates spatial barcodes. Similar Illumina chemistry employed in
ternative to physical dissection, methods such as NICHE-seq,[48] polony-indexed library-sequencing (Pixel-seq)[61] includes a mod-
ZipSeq,[49] Light-seq,[50] and GeoMX Digital Spatial Profiling[51] ified polymer surface for clustering to minimize gaps between
instead use optical methods to mark regions or cells of inter- the barcoded pixels. This polymer-based surface can generate
est. The strength of these regioselective sequencing methods continuous features by tightly distributed mRNA capture oligos.
is easy implementation and sensitive sequencing with small Subsequently, the spatial barcode is incorporated into the cDNAs
sample sizes (<200 cells) where a clear spatial ROI is al- by reverse transcription, and the cDNA sequence and spatial bar-
ready defined by IHC staining. Additionally, LCM with SMART- code are simultaneously analyzed by NGS. Spatial enhanced res-
3Seq[52] and GeoMX[51] have demonstrated their applicability for olution omics sequencing (Stereo-seq)[62] also utilizes NGS de-
formalin-fixed paraffin-embedded (FFPE) samples. Recent in- vices to manufacture spatial barcode arrays using MGI’s DNA
vestigations employed GeoMX to study archived FFPE human nanoball sequencing (DNBseq) chemistry (Figure 2e). Stereo-seq
cancer samples.[53–55] Although there are notable advantages to can localize the position of spatial barcode oligo arrays with a su-
LCM methods, they typically have limited spatial resolution and perior resolution at a diameter of ≈0.22 μm with a 0.5-μm spacing
throughput compared to recent spatial barcoding NGS tech- in the DNBseq instrument. The UMI sequences are then ligated
niques. to mRNA-capturing oligos for quantification. This ultrafine res-
In general, spatial transcriptomics describes techniques used olution increases the cost of sequencing an enormous number
for analyzing the spatial information of transcriptomes. Accord- of pixel areas, thereby limiting large-scale tissue investigation. In
ing to a methodology reported earlier, spatial transcriptomics is a addition, at this ultrafine resolution, the lateral diffusion of mR-
non-generic term that has originated from the technique itself.[56] NAs could lead to incorrect localization of transcripts by blurring
This new method has been used to locate transcripts by transfer- their spatial locations.
ring RNA molecules to a glass slide coated with poly-T primers Therefore, platforms that do not rely on delivering mRNA to a
containing a unique molecular identifier (UMI) and a spatial bar- spatially barcoded oligo array anchored on the surface of chips or
code using ≈100 μm pixel size resolution (Figure2a). While cap- slides have been explored (Figure 2f–h). In deterministic barcod-
turing mRNAs with poly-T on the slide surface, the newly syn- ing in tissue for spatial omics sequencing (DBiT-seq), oligos en-
thesized cDNAs templated by these captured transcripts contain coding spatial barcodes are directly delivered to fixed tissues us-
pre-allocated spatial barcodes, which enable retrieval of the origi- ing a microfluidic chip[63] (Figure 2f). Oligos with predetermined
nal transcript locations. Through library construction and NGS spatial barcodes are loaded into each channel of the microfluidic
analysis, cDNAs and spatial barcodes can be sequenced to si- chip and delivered to a certain grid position of the tissue. The
multaneously identify and locate specific RNA transcripts in tis- same process is repeated along the axis of the fluidic channel
sues. However, this strategy suffers from low RNA-capturing ef- that is perpendicular to the initial axis, and the resulting oligo
ficiency and poor spatial resolution at a spot-to-spot distance of mixtures at each grid position are ligated into cDNA in a combi-
200 μm and consequently lacks single-cell resolution. Follow-up natorial manner to produce spatial barcodes. The location where
techniques have been developed with an improved spatial resolu- the two axes intersect is determined by examining the combina-
tion by reducing the pixel size of spatial barcodes (Figure 2b–e). tions of spatial barcodes by sequencing. Controlling the channel
Slide-seq[57,58] has achieved a spatial resolution of 10 μm using width of a microfluidic chip can also confer higher spatial res-
a random distribution of barcode-containing polystyrene beads olution. Compared with Slide-seq and HDST, each channel in
on a slide (Figure 2b). High-definition spatial transcriptomics[59] DBiT-seq contains a predetermined barcode sequence and does
(HDST) has subsequently demonstrated a spatial resolution of not require identifying the distribution of spatial barcodes us-
2 μm using a silicon wafer (Figure 2c). However, the improved ing time-consuming in situ sequencing. Furthermore, proteins
spatial resolution for Slide-seq and HDST requires random dis- can be labeled by treating fixed tissues with antibody-oligo con-
tribution of beads containing spatial barcodes in space, and jugates, which, with concomitant detection of nucleic acids, ren-
therefore, their exact distribution must be determined by time- ders quantitative analysis of both proteins and mRNAs with their
consuming imaging-based in situ sequencing. Additionally, the respective spatial distributions.
detection efficiency is negatively affected by the mechanism of Labeling mRNAs with spatially barcoded oligos in compart-
mRNA capture from even smaller sample volumes used with mentalized microfluidic channels is still insufficient to reflect
these methods. actual single-cell-level transcriptomes in space. Transcriptomes
To facilitate the fabrication of an mRNA-capturing oligo ar- of relatively large pixels only provide mixed information on
ray with encoded spatial barcodes, the Illumina NGS instrument transcripts of multiple cells. By extension, inferred transcrip-
can be employed to simplify the identification of spatial barcode tomes from multiple small pixels have inherent limitations
distributions (Figure 2d). Seq-Scope directly adopts the sequenc- in their correspondence to genuine single-cell-level data. The
ing process utilized by Illumina NGS devices to generate spa- sci-Space[64] and XYZeq[65] methods provide true single-cell or
tial barcode arrays at a separation distance of ≈0.6 μm.[60] Illu- single-nucleus-level spatial transcriptomes by dissociating cells
mina sequencing libraries are generated containing both spatial or nuclei while preserving their spatial origins at a certain well-
barcodes and oligo-dT, amplified by PCR, and dispersed on the or pixel-scale resolution (Figure 2g,h). After delivery of either
flow cell through clustering of the sequencing device. Spatial bar- hashing oligos containing spatial barcodes with poly(A) tail
codes are then identified through the actual sequencing steps. By (sci-Space, Figure 2h) or capturing mRNA by infiltrating tissues
treating tissue samples overlaid on the flow cell with digestive with spatially barcoded reverse transcription primers (XYZeq,
enzymes, the released mRNAs can be captured by the oligo-dT Figure 2g), the single-cell level transcriptome can be retrieved by
domain anchored at the surface of the flow cell that also incorpo- inserting additional cellular barcodes via a non-spatial version of

Adv. Sci. 2023, 10, 2206939 2206939 (5 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Figure 2. Next-generation sequencing (NGS)-based spatial transcriptomics methods. a) General workflow of the spatial transcriptomics (ST) method.
The spatial barcode-encoded oligos are immobilized on a functionalized surface to capture mRNA released from the mounted tissue or cells. Subsequent
cDNA synthesis, followed by sequencing libraries yield transcript sequences and their spatial locations, simultaneously. b–e) Developments of methods
for the spatial patterning of barcoded oligos with enhanced spatial resolution. Unlike imaging-based ST, these methods require barcode decoding after
patterning due to the random spatial distribution of spatial barcodes. b) Slide-seq employs random spatial bead spreading and in situ sequencing
decoding. c) HDST deposits beads with combinatorial barcodes on patterned wafers, followed by decoding with serial hybridization. d) Seq-Scope and
Pixel-seq utilize Illumina clustering for oligo patterning and directly read sequences using Illumina sequencers. e) Stereo-seq utilizes DNBSEQ chemistry
to generate DNA nanoballs with spatial barcodes, which are patterned on a flow cell, and barcode calling is performed by the MGI sequencer. f) DBiT-
seq delivers barcoded RT primers and ligation oligos through orthogonal microfluidic channels. The predetermined spatial distribution of overlapping

Adv. Sci. 2023, 10, 2206939 2206939 (6 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

combinatorial single-cell or single-nucleus indexing. However, structure, which improves their effective resolution by increas-
these techniques suffer from RNA integrity issues because these ing intermolecular distance. Following subsequent optimization
protocols require chemical fixation to maintain cellular integrity of ExM using swellable hydrogels and a nucleic acid crosslinker
during combinatorial indexing. (Label-X), expansion FISH (ExFISH) has demonstrated multi-
plexed FISH in expanded biological samples.[72] Further opti-
mization of the ExFISH method led to expansion-assisted iter-
2.3. 3D Spatial Investigation by Tissue Clearing and Expansion
ative FISH (EASI-FISH), which can quantitatively analyze rela-
tively large tissue volumes (up to 300 μm thickness before ex-
Recent spatial transcriptomic technologies are restricted to 2D
pansion) to enable true single-cell transcriptomics by investigat-
investigations of transcripts. Because NGS-based methods use
ing multiple layers of intact cells by imaging.[73] Development
spatially-barcoded oligo arrays to capture and label mRNAs to
of in situ sequencing using expansion methods yielded ExSeq[74]
preserve their spatial origin, the tissue mRNA or spatially bar-
for combining tissue expansion with in situ sequencing meth-
coded oligos must spread along the Z-axis of tissues toward the
ods such as FISSEQ. Expansion sequencing (ExSeq) has further
surface in contact with the array. As a result, 3D spatial distri-
improved the resolution of FISSEQ by adding a physical expan-
bution is analyzed in a projected form on the 2D sample plate.
sion, resulting in better coverage of transcript detection in vol-
In highly multiplexed FISH and in situ sequencing methods,
ume. Unified ExM (UniExM)[75] employs glycidyl methacrylate
the true 3D distribution of individual transcripts can be mea-
(GMA) as a nucleic acid crosslinker, which dramatically reduces
sured; however, imaging procedures involve time-consuming mi-
experimental costs. Using GMA improves nucleic acid retention
croscopic measurements with very limited fields of view, such as
in polyacrylamide-based hydrogels and provides excellent results
the single optical plane that is analyzed in seqFISH+. Therefore,
when applied in multiplexed FISH in situ sequencing with ex-
the collected data corresponds to a sequence of 2D samples, and
pansion. These 3D tissue investigation approaches may play sig-
local 3D spatial information is limited by the thickness of tissue
nificant roles in studying transcriptomes and multi-omes of tis-
sections. Although a 3D atlas can be constructed by serial sec-
sues and organoids as multi-cell layer spatial information is nec-
tioning, a systematic gap exists between the image data points
essary to understand the complex behavior of biological systems.
because not all tissue sections can be subjected to imaging owing
to the enormous volume of samples and the restricted through-
put imposed by the imaging methods. 3. Challenges and Opportunities of Current
Tissue clearing techniques enable direct observation of 3D Methods
structures from transparent thick-tissue blocks prepared by opti-
cal clearing and refractive index matching.[66] In cleared tissues, 3.1. Inherent Technical Difficulties in Imaging-Based Methods
lipids are removed to minimize light scattering inside tissues for
optical clearing. Target proteins are labeled by immunostaining, Highly multiplexed FISH and in situ sequencing methods are
however, owing to the harsh nature of tissue clearing protocols based on imaging using fluorescence microscopy to locate single
that involve organic solvents or highly concentrated aqueous so- transcripts at subcellular resolution. Although the spatial resolu-
lutions, tissue clearing methods may lead to damage or loss of tion varies among techniques, imaging-based methods can pro-
RNA transcripts and are typically incompatible with RNA de- vide sub-micron resolution when employing a high-NA objective
tection. For certain hydrogel-based clearing methods, smFISH during imaging. When additional information is obtained us-
is compatible after lipid removal by detergents [CLARITY[67] ing probes [e.g., proteins labeled with antibodies and nuclei with
and passive clarity technique (PACT)[68] ]. Additional RNA fix- nucleic acid-labeling dyes such as 4’,6-diamidino-2-phenylindole
ation helps to improve RNA retention during tissue clearing (DAPI)], the spatial data can be projected onto the cell contours
(EDC–CLARITY).[69] MERFISH, an imaging-based multiplexed by cell segmentation. Although these imaging-based methods
FISH method, is also compatible with hydrogel-embedded tis- provide high spatial resolution, this feature inevitably limits the
sues and was previously employed to reduce autofluorescence size of tissue coverage. The tissue area captured within a single
signals from tissues. The tissue clearing method stabilization un- imaging cycle is limited by the field of view of the objective, and
der harsh conditions via intramolecular epoxide linkages to pre- the imaging time depends on the signal-to-noise ratio and tar-
vent degradation (SHIELD)[70] utilizes a polyepoxy crosslinker get resolution. In addition, detecting broad features from various
as a fixative and was originally developed to protect multimodal transcriptional profiles requires multiple rounds of imaging fol-
biomolecules and tissue architecture. RNA molecules are easily lowed by probe hybridization or enzymatic reactions for ISS. Ac-
detected by FISH hybridization chain reaction (HCR) after tis- quiring multiple images includes time-consuming steps, such as
sue clearing by SHIELD; therefore, SHIELD could serve as a 3D probe replacement between each image, and requires a substan-
spatial multi-omic tissue processing and imaging platform. tial amount of time that limits the acquisition of information over
Tissue expansion techniques allow super-resolution imaging a large area.
with diffraction-limited light microscopy by physically expanding Optical crowding in raw images is another technical hurdle
hydrogel-embedded tissues. In expansion microscopy (ExM),[71] in imaging-based methods. Considering that one cell is typically
swellable hydrogel-embedded tissues expand and magnify their populated with thousands of transcripts, the point-spread func-

regions eliminates time-consuming steps for random spatial barcode sequencing procedures. g) True single-cell and single-nucleus resolutions with
regional spatial barcode printing in XYZeq. h) sci-Space delivers hashing oligos with spatial barcodes into tissue followed by additional fixation to retain
hashing oligos in the nuclei. After combinatorial barcoding for single-nucleus RNA sequencing, hashing oligos are sequenced as a transcriptome.

Adv. Sci. 2023, 10, 2206939 2206939 (7 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

tions of the fluorescence signals emitted from these transcripts points of optimization that continue to limit widespread imple-
often overlap when simultaneously detecting multiple species mentation. To our best knowledge, the combinatorial barcoding
in a single acquisition step. Highly multiplexed FISH methods scheme has not been applied for increasing multiplexity due to
can detect a large number of transcripts owing to their excel- technical challenges in registration of fluorescence signals from
lent hybridization efficiency; however, consequently, signal over- large 3D raw images. As a result, the processing time required for
lap makes the deconvolution of barcodes difficult. MERFISH and repeated 3–4 color imaging and staining/destaining cycles with
seqFISH+ can resolve optical crowding issues by reducing the linearly increasing numbers of gene targets is lengthy for cap-
number of transcripts detected in a single image providing in- turing large area samples. Additionally, the increased resolution
creased space for barcode combination. However, this approach and addition of volumetric analysis inherently result in immense
directly impacts imaging time by increasing the number of imag- data sets in biomedical applications. The current analysis plat-
ing rounds and time required to image large combinations of forms require considerable improvement before they are capable
barcodes. In ISS-based methods, optical crowding is less prob- of handling data volumes in the order of a hundred terabytes to
lematic because these methods exhibit low detection efficiency several petabytes. Following optimization of these aspects of tis-
when converting transcripts to cDNAs in situ. FISSEQ detects sue clearing and expansion techniques, large biomedical datasets
≈0.2–1% of transcripts owing to the low yield of in situ reverse with enhanced spatial resolution can be leveraged to uncover ad-
transcription and cDNA synthesis. Other methods utilizing pad- ditional transcriptomic information from complex and crowded
lock probes that hybridize to their target RNA species require en- environments.
zymatic ligation that results in low detection compared to those
of multiplexed FISH methods. 4. Challenges in Raw Data Processing
After successfully implementing spatial transcriptome experi-
3.2. Inherent Resolution Issues in NGS-Based Methods ments, interpreting large amounts of raw gene expression matri-
ces is challenging. Each method for spatially resolved transcrip-
NGS-based methods generate massive amounts of raw data ow- tomics generates data of various scales, resolutions, and modali-
ing to the high throughput of NGS devices. Initial capturing of ties, according to their working mechanisms, to capture and iden-
mRNAs using barcode oligos and subsequent cDNA synthesis tify transcripts. Therefore, analytical pipelines that process raw
does not require specialized equipment. An additional benefit is data should consider these differences for successful interpreta-
that NGS-based methods use poly-T oligos for unbiased capture tion. Additionally, limited information on current spatial tran-
of mRNAs, and therefore do not require predefined probe pan- scriptome data is sometimes analyzed together with the exist-
els to label target RNA. This feature allows de novo discovery of ing information. In this section, we introduce useful analysis
RNAs with statistically significant differences in spatial distribu- pipelines and algorithms for handling data from spatial transcrip-
tion. tomics (Figure 3). Since imaging-based and sequencing-based
Owing to the limited spatial resolution, deciphering spatial methods generate raw data in different classes, the workflows are
barcodes to infer the spatial origin of transcripts cannot provide depicted in two categories.
their cellular origin. The mRNA-capturing oligos contain specific
spatial barcode sequences that represent specific regions in 2D
4.1. Dealing with Error-Prone Raw Data
coordinates to distinguish the spatial origin of transcripts. A set
of transcriptomes sharing the same spatial barcode constitutes
Regardless of their working principles, all imaging-based meth-
transcriptome information for each tissue region, such as pixels
ods for spatial transcriptomics require three crucial steps in raw
in an image, and the arrangement of these pixels eventually rep-
data processings: 1) registration of barcoded fluorescence sig-
resents the spatial distribution of the entire tissue transcriptome.
nals from raw images, 2) barcode calling or processing to as-
However, each pixel specified by the same spatial barcode in the
sign target RNA reads, and 3) efficient cell segmentation to as-
2D coordinates does not provide the actual cellular boundaries
sign barcoded dots to each cell. Prior to further processing, raw
required to reconstruct single-cell-level information. Therefore,
images are stitched and maximum-projected along the z-axis to
different technologies require different information processing.
generate 2D xy plane images. These images are typically sub-
If the size of a pixel is greater than that of a single cell, the tran-
jected to various filters, such as Laplacian of Gaussian filters
script of one pixel would be a mixture of the transcripts of several
to remove noise and thresholding for dot detection. Initially,
cells and vice versa. Although sci-space and XYZeq enable single-
coarse-grained registration among color channels and imaging
nuclear distinction of the transcriptome by dissociating nuclei,
rounds is applied with reference to specified registration mark-
the spatial resolution of nuclear location is still limited by pixel
ers (e.g., fluorescence beads or blood vessel staining). One signif-
size.
icant challenge in assigning barcodes is to determine the accept-
able error range in terms of pixel distances for successful bar-
3.3. Processing Speed for 3D Spatial Investigation with Tissue code calling with minimal error. When the number of barcod-
Clearing and Expansion ing genes is small, less accurately registered fluorescence sig-
nals may be tolerated without substantially affecting the suc-
Tissue clearing and expansion techniques can enhance the ef- cess rate in barcode calling. Some methods such as MERFISH
fective resolution of 3D spatial transcriptomic investigations for have error correction schemes to reduce the effect of incorrect
future biomedical applications. Despite the valuable information barcode assignment. However, transcriptome-level barcoding re-
that can be extracted through these methods, there are remaining quires raw images obtained through super-resolution imaging

Adv. Sci. 2023, 10, 2206939 2206939 (8 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Figure 3. The workflows for preprocessing raw data in a) imaging-based and b) sequencing-based spatial transcriptomics methods. In the imaging-
based workflow, spatial transcriptomics data is predominantly analyzed independently, while single-cell sequencing data is occasionally integrated. In
contrast, sequencing-based methods more commonly employ both single-cell sequencing data and histological images simultaneously. The resulting
outputs of the two workflows differ in their celltype format. For sequencing-based methods, the spot size is usually larger than a single cell, so the cell
type of each spot is described proportionally. In contrast, imaging-based methods provide a cell-level gene count matrix, which directly labels the cell
type on each cell. These outputs are then employed for downstream analysis.

modalities to further resolve multiple transcripts located within [Figure 3b(i)]. Due to the low transcript-capturing efficiency of
diffraction-limited spots. With this level of complexity, barcode NGS-based spatial transcriptomic methodologies, raw data suf-
calling rates can be significantly reduced compared to those of fer from loss of gene expression information. Thus, data imputa-
experiments with dozens and hundreds of target genes, and the tion methods that are specifically tailored to the characteristics of
resulting analysis scheme inherently yields decreased RNA detec- the spatial data can enable the combination of spatial spots with
tion efficiency. Imaging-based methods also involve image pro- corresponding pathological images[77] and provide consolidated
cessing and error correction steps for chromatic correction be- analysis by spatial transcriptomics and pure single-cell sequenc-
tween fluorescent channels and image registration between mul- ing data[78,79] [Figure 3b(ii)].
tiple rounds. For example, initial MERFISH papers employed Normalization of spatial transcriptome data is crucial for com-
only the 647 channel to avoid chromatic aberration. paring gene expression between spots or genes [Figure 3b(iii)]
It is important to note that all of the correction steps men- and is performed using conventional methods (e.g., regularized
tioned earlier are typically performed with specific ad hoc param- negative binomial regression).[80] Recently, using deep-learning
eters optimized for a particular spatial transcriptomics method models, morphological features have been extracted using spa-
and setup, which makes the raw image data processing pipeline tial gene expression patterns in combination with hematoxylin
difficult to generalize. Therefore, it may be more practical for and eosin (H&E)-stained images. The extracted features can be
researchers to begin their imaging-based spatial transcriptomic compared among spatial data spots to complete the normaliza-
analysis with pre-processed data generated by commercial or pre- tion of spatial data, which is called Spatial Morphological gene
optimized protocols. The following sections will focus more on Expression Normalization (SME Normalization).[81]
reviewing data-processing techniques for NGS-based methods.
Due to technical limitations of currently available NGS-based
spatial transcriptomic methodologies, raw data may contain 4.2. Integration of Spatial and Single-Cell Transcriptome Data
noise from various sources and often suffer from signal loss. As
a result, a low signal-to-noise ratio (SNR) can compromise the 4.2.1. Predicting the Spatial Distribution of Transcripts
accuracy of data analysis. To address this issue, preprocessing
methods for noise reduction have been explored. Single-cell and single-nucleus sequencing databases have grown
The datasets generated by NGS-based methods, such as 10X rapidly in terms of higher throughput and data quality. When
Visium, contain spot swapping, where a certain spot may contain consolidating these rich databases, it is important to predict the
transcripts from nearby spots. These unwanted contaminants spatial distribution of transcripts by combining relatively crude
can be removed using a probabilistic model called SpotClean[76] quality spatial data [Figure 3b(iv)]. Spatially reconstructed tran-

Adv. Sci. 2023, 10, 2206939 2206939 (9 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

scriptomic data already provides valuable information, but single- ing deconvolution by substituting it as a domain adaptation task
cell sequencing data can be used to address some of the limita- is also plausible.[95,96] These methods initially define cell type as
tions of spatial transcriptomics methods. One drawback of FISH discrete or solid. Based on this hypothesis, they carry out cell-
or ISS data is that it only includes preselected genes and ex- type inference. However, tissues with cancer and inflammation
cludes others. These excluded genes can be imputed using spa- may contain transcriptome variations in each cell type. Deconvo-
tially reconstructed single-cell sequencing data because scRNA- lution of spatial transcriptomics profiles using variational infer-
seq can profile all types of transcripts using poly-T or random ence (DestVI) conducts cell-type inferences based on a continu-
primers [Figure 3a(ii)]. In the case of sequencing-based spatial ous cell-type model.[90] While most spot deconvolution methods
transcriptomics data, it often suffers from a low number of cap- estimate the proportion of different cell types within a given spot,
tured genes. Integrating single-cell sequencing data can improve the results do not provide single-cell level resolution. In contrast,
this, as single-cell sequencing data typically contain a sufficient cellular spatial positioning analysis via constrained expression
number of genes [Figure 3b(v)]. alignment (CytoSPACE) employs single-cell sequencing data to
Seurat[82] allows seamless integration between single-cell se- assign each cell to a specific spot in spatial transcriptomic data,
quencing data and multiplexed FISH data by comparing the ex- thereby achieving single-cell level resolution.[97]
pression of landmark genes in single-cell sequencing data with Detailed benchmarking of the performance of spatial
their spatial distribution in multiplexed FISH data. The expres- and single-cell transcriptome integration has recently been
sion data obtained from this process is represented by a bi- published.[98] Notably, performance benchmarking may be heav-
modal mixture model. Spatial backmapping[83] adopts a simi- ily affected by data volume, particularly in deep learning models.
lar approach, but comparisons are based on specificity-weighted Additionally, the spatial resolution of each method significantly
mRNA profiles, which indicate the expression of each gene in a affects the performance of the cell-type inference. For example,
specific cell relative to that of all other cells. robust cell type decomposition (RCTD)[91] and stereoscope[92]
Unlike these methods focusing on the spatial reconstruc- use a direct-count model for their inference; therefore, better
tion of single-cell sequencing data, spatial gene enhancement performance is expected on high-resolution spatial data.[99]
(SpaGE)[78] imputes missing genes in multiplexed FISH data us- When performing cell type detection, certain cell types may not
ing single-cell data via principal component analysis (PCA)-based be identified if the corresponding cells or transcripts are not cap-
domain adaptation and k-nearest-neighbor regression. Gene im- tured in the single-cell sequencing data or spatial transcriptomic
putation with Variational Inference (gimVI)[79] method also per- data. This can result in misinterpretation of the overall results,
forms imputation using a deep generative model. especially if the cell type of interest is rare. Therefore, to improve
Recently, data for the integration of spatial contexts is more di- the reliability of a spatial transcriptomics analysis, a power anal-
versified, and deep learning is widely employed. Seurat v3[84] inte- ysis framework has been proposed.[100]
grates single-cell and spatial data, as well as chromatin accessibil-
ity and immunophenotyping data. Integrative analysis of multi-
omics at single-cell resolution (GLUER)[85] integrates single-cell 4.3. Alignment and Integration of Multiple Spatial Data
sequencing data with transcript and protein spatial data captured
by highly multiplexed methods such as co-detection by indexing Currently, NGS-based methods may exhibit poor transcript-
(CODEX)[86] and MERFISH.[22] Both DEEPsc[87] and Tangram[88] capturing efficiency. The computational strategy used for im-
employ single-cell sequencing data for spatial reconstruction and provement is to collect multiple adjacent 2D spatial transcrip-
can consolidate with spatial data obtained by spatial transcrip- tome datasets from tissues and then perform alignment and
tomics methodologies without limitations; therefore, both multi- integration from multiple 2D data points [Figure 3b(iv)]. Here,
plexed FISH and NGS-based data can be used as inputs. GLUER, alignment means finding pairwise spots (i.e., overlapping zones)
DEEPsc, and Tangram have recently started using deep neural between 2D data points, and the integration constructs single-
network models for improved prediction and data integration. domain 2D data by consolidating multiple 2D data. As a re-
sult, these integrated spatial transcript data contain more gene
expression information than that of the source data. In STU-
4.2.2. Cell-Type Inference and Spot Deconvolution tility, histological images obtained from the same tissue used
for spatial transcriptome data acquisition are employed for 2D
Performing cell-type-based analysis is challenging when the spa- data alignment.[101] The probabilistic alignment of ST experi-
tial resolution of a method is lower than the size of a single ments (PASTE) method is effective without employing histolog-
cell because in this situation multiple cells can contribute to the ical images.[102] Instead, PASTE finds pairwise alignment be-
transcripts extracted from each spot. Therefore, it is essential to tween several 2D datasets based on probable transcriptomic and
estimate the proportion of cell types in each spot through de- spatial resemblance.
convolution using nonspatial single-cell sequencing data [Fig-
ure 3b(vi)]. Because the spot size of recent NGS-based spatial
techniques is larger than the size of a cell, deconvolution meth- 4.4. Cell Segmentation and Cell Typing of Imaging-Based
ods for spatial transcriptomic data have been actively proposed Methods
since 2020. On the other hand, Slide-seq, HDST, Seq-Scope, and
Pixel-seq have finer resolutions; therefore, the spot size is com- Analyzing data produced by imaging-based methods requires
parable to a cell or even smaller. For deconvolution, statistical a cell segmentation task for classifying transcripts to each as-
models or matrix decomposition can be employed,[89–94] but treat- signed cell [Figure 3a(i)]. Approaches to cell segmentation tasks

Adv. Sci. 2023, 10, 2206939 2206939 (10 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

for imaging-based methods can be classified into three major cat- by the connections between functional neurons, a comprehen-
egories: 1) manual, 2) supervised image segmentation, and 3) un- sive understanding of this system requires elucidation of the
supervised approaches. It is worthwhile to note that conventional wiring between different types of neurons by dissecting neuronal
image processing approaches without machine-learning models communication. Spatial transcriptomics can provide important
show inferior accuracy compared to supervised image segmenta- information regarding the molecular states of cells, which can be
tion techniques and are rarely studied nowadays. With this con- integrated with various physiological features by precise assign-
sideration, conventional approaches have not been discussed fur- ment of individual cells (Figure 4).
ther in this paper.
Manual and supervised image segmentation approaches first
fluorescently stain cell bodies and/or nuclei, and then recog-
nize cell and/or nuclei boundaries by manual or supervised algo- 5.1. Brain Transcriptomic Atlas
rithms. Manual approaches require huge amounts of labor and
the results are often unsatisfactory. Commonly, cell boundaries The brain is the most segmented and annotated organ according
are not clearly distinguishable from multiplexed FISH or ISS im- to its spatial architecture. The brain is precisely divided into re-
ages. Supervised image segmentation requires a manually anno- gions or areas associated with specialized functions and connec-
tated dataset for training machine learning models,[103–106] often tivity. However, as brain regions become more fragmented, the
resulting in inaccurate segmentation for other datasets due to the accuracy of defining their entities becomes controversial.[111] Spa-
low performance of the generalized models. tial transcriptomic analysis on a series of coronal sections across
Unsupervised approaches utilize transcript distribution to the whole brain provides new borders for region annotation de-
cluster transcripts for assignment to each cell. These ap- fined by molecular characteristics. The hierarchical clustering of
proaches could be adequate alternatives or complements to pre- spatial transcriptomic data points is highly correlated with con-
viously discussed cell segmentation approaches because they are ventional neuroanatomical annotation and provides novel spatial
annotation-free and often show better performance on cell-level borders to segment molecular subregion candidates.[112] To es-
transcripts clustering as transcript distribution can be a better tablish a single-cell spatial atlas with cell-type annotation, Zhang
indicator of segmentation than visually recognizable cell bound- et al.[113] profiled ≈300 000 cells in the primary motor cortex of
aries in messy fluorescence images.[107] Also, supervised image mouse brain using MERFISH with a selected gene panel de-
processing and unsupervised transcript clustering approaches rived from the results from previous single-cell RNA sequencing
can be used together to complement each other.[108,109] (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq).
By analyzing the imaging-based spatial transcriptomics data, Manno et al.[38] demonstrated the potential to create a spatial at-
cell types can be also identified [Figure 3a(iii)]. This can be con- las of developing mouse embryos by integrating scRNA-seq data
ducted with data obtained purely from imaging-based methods with in situ sequencing of transcripts using HybISS applying a
without aids from sequencing data. In these approaches, cell deep-learning-based method.[88] STARmap PLUS established a
types are defined within the cell segmentation results.[107] Hence, spatial molecular atlas with single-cell resolution by successful
inaccurate cell segmentation results in poor cell type identifica- segmentation from transcript annotation,[114] from which the im-
tion. This issue can be resolved by employing a cell segmentation- puted gene expression pattern of a brain section was found to be
free method, which uses spots as a unit of cell-type inferences.[110] comparable to the in situ hybridization (ISH) database of Allen
Additionally, single-cell sequencing data can be combined to Mouse Brain Atlas.[115] STARmap PLUS also enables the study of
identify cell types from FISH or ISS data more precisely.[36] engineered recombinant adeno-associated virus (rAAV) tropism
across whole mouse brain regions by capturing transcripts pack-
4.5. Future Works for Analyzing Spatial Transcriptomics Data aged in AAVs. Given that the mouse brain atlas is not merely an
architectural map of the brain but also contains various func-
Spatial data processing studies focus on single-cell or tional annotations, integrating functionality with the molecular
supracellular-scale data, probably obtained using NGS- atlas is noteworthy for further improvements. The Brain Initia-
based methods. However, imaging-based methods can obtain tive Cell Census Network has initiated the integration of multiple
subcellular-level details; therefore, subcellular-level processing transcriptome data generated by different methods, samples, and
of the spatial distribution of transcripts will be highly important. laboratories to construct a multimodal atlas of the primary motor
Currently, the database may be too limited to employ deep cortex while tracking the laminar distribution of diverse neuronal
learning approaches, as found in earlier studies by Seurat and types using MERFISH.[116] Allen Institute for Brain Science re-
gimVI for <2000 cells. Therefore, the performance of analysis is cently announced the creation of an extensive spatial transcrip-
not guaranteed, and more experimental databases are expected tomic atlas covering the entire adult mouse brain. This achieve-
to be available to develop algorithms. ment was made possible by integrating data from scRNA-seq
analysis of 7 million cells and MERFISH analysis of 4 million
5. Application Note in Neuroscience cells, resulting in a total of nearly 11 million cells analyzed.[41]
The resulting atlas is an unprecedented achievement that sheds
The nervous system of higher organisms has a complex struc- light on the spatial organization of individual cell types and val-
ture. By orchestrating organism-level responses to internal or ex- idates the role of transcription factors in determining cell type
ternal stimuli, the nervous system governs complicated commu- through an integrated hierarchical classification of spatial tran-
nications across the peripheral and central nervous systems. Be- scriptomics. Building a comprehensive reference atlas is a critical
cause the function of the nervous system is primarily determined endeavor that will facilitate the generation and validation of new

Adv. Sci. 2023, 10, 2206939 2206939 (11 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Figure 4. Applications of spatial transcriptomics in neuroscience. The transcriptomic atlas is established by cell-type identification with the spatial context
of precisely compartmentalized brain structure. Transcriptome-based neuronal cell types can be further dissected by integrative studies employing tools
for functionality and connectivity investigations. Accurate neural classification will lead to circuit-level studies to investigate individual circuits to elucidate
the comprehensive function of the entire brain.

hypotheses as spatial transcriptomics continues to uncover novel The landscape of isoform expression is modulated by alter-
transcriptional relationships with functions and architectures. native splicing as post-transcriptional regulation of gene expres-
sion, thereby contributing to the phenotype of individual cells.
Therefore, different isoforms can provide various criteria for sub-
dividing cell types identified by gene expression profiling. The
5.2. Neural Classification conjunction of high-throughput single-cell gene expression pro-
filing by 3′-end sequencing and full-length RNA-seq could sub-
The classification of neuronal types can facilitate investigating divide clusters that were previously identified by RNA profiling
the comprehensive relationship between the structure and func- based on their isoform expression patterns. The gradient dis-
tion of individual cells. Various criteria are applied to dissect neu- tribution of isoform landscape in the primary motor cortex of
rons for proper classification based on their morphology, physiol- the mouse brain was observed using MERFISH, with marker
ogy, and molecular features.[117] scRNA-seq has been widely em- genes correlated with isoform expression.[126] Joglekar et al.[127]
ployed in the molecular profiling of neurons to identify specific directly observed the spatial distribution of isoforms by long-read
molecular characteristics that correlate with their physiology and sequencing of RNAs captured by 10X Genomics Visium.
function.[118–123] Spatially resolved transcriptomics can provide
new insights for criteria to tabulate ambiguous cell types into dif-
ferent categories according to their origins. 5.3. Studying Neural Circuits
Connectivity, another criterion for neuronal classification that
was merely accessible previously owing to the lack of effi- Neural circuits are composed of substantially intermingled neu-
cient methodologies,[117] can be analyzed through a spatially re- ronal connections across the entire nervous system. Dissecting a
solved transcriptomic investigation. In barcode analysis by se- specific circuit to study the functions of individual neurons is a
quencing (BARseq),[124] the anterograde tracing virus packaged fundamental approach to comprehending the entire system. The
with random RNA barcodes was injected into the cortical area, functional connectivity of neural circuits is being explored with
and the barcodes were sequenced by in situ sequencing using developed genetic tools for dissecting neural circuits according to
BaristaSeq.[34] The single-cell projection pattern was identified their phenotypic properties.[128,129] For instance, in vivo calcium
by matching RNA barcodes and barcode reads using bulk-RNA imaging enables functional observation of neurons even within
sequencing in multiple projection areas. The gene expression freely moving animals.[130,131] Spatially resolved neuronal func-
pattern from smFISH was integrated to identify the cell types tionality and transcriptomics integrate different modalities of
of projection-mapped neurons. BARseq2,[125] an improved ver- neuronal phenotypes to further dissect neuronal identities.[132,133]
sion of BARseq, allows multiplexed gene detection with padlock Post-hoc molecular profiling of the recorded cells can be car-
probes and in situ sequencing, followed by mapping of projection ried out using IHCof ex vivo brain sections of the same ani-
patterns to broad molecular profiles with single-cell resolution. mals, enabling the detection of marker genes to identify neuronal

Adv. Sci. 2023, 10, 2206939 2206939 (12 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

subtypes.[134–136] In this regard, rather than simply detecting a the brain and cause autoimmune multiple sclerosis and observed
few marker proteins by IHC, multiplexed FISH can further en- the residence of T-cells inside the brain during disease progres-
hance the performance of molecular profiling by analyzing the sion using spatial transcriptomics. The same group identified pu-
higher number of markers to classify neurons of interest accord- tative drug targets for progressive multiple sclerosis with inter-
ing to their expression patterns.[137–139] Bugeon et al.[140] devel- actome analysis based on transcriptome distance derived from
oped combinatorial padlock-probe-amplified FISH (coppaFISH) spatial transcriptomics.[149]
for multiplexed FISH to detect 72 genes in the primary visual cor-
tex of mouse brain for posthoc transcriptomic profiling after in
vivo calcium imaging during visual stimulation. This increased 5.5. Future Perspectives in Neuroscience Application
number of detected transcripts enabled the precise classification
of individual cells based on the previous scRNA-seq transcrip- Spatial transcriptomics has impacted various fields of neuro-
tome. The dependency of neuronal responses to visual stimuli science by providing additional spatial dimensions to other exist-
according to the cortical state was related to the transcriptional ing research modalities. As the application of spatial transcrip-
profile of cells. tomic technology in neuroscience matures, spatially resolved
Moffit et al.[25] characterized the molecular profiles of neu- transcriptome data will be further consolidated with separate an-
rons activated by social behavior using MERFISH. Represen- alytical modalities. Applying spatial transcriptomics to the in-
tative marker genes for MERFISH analysis were selected after vestigation of the brain during different developmental stages
scRNA-seq of the preoptic area, which is a subregion of the hy- has enabled spatiotemporal tracking of transcriptomic changes
pothalamus, and MERFISH analyzed the spatial organization of in specific developmental structures, allowing us to impute the
each cell type in this area. By integrating neuronal activities de- intermediate cellular and molecular profiles during neurodevel-
rived from immediate early gene expression with cellular tran- opment processes.[150] By integrating rich scRNA-seq data with
scriptome, the transcriptomic profile of behavior-related neurons limited gene profiles generated by seqFISH, entire gene expres-
and their locations were identified. The analysis subdivided pre- sion patterns were successfully inferred, and this provided addi-
defined cell types into precise classes and provided anatomical tional gene profiles related to the organization of developmen-
information for socially relevant neural circuits. tal processes, specifically those associated with the formation of
the midbrain-hindbrain boundary.[151] By integration of data from
spatial transcriptomics and imaging of dynamics of chromatin
5.4. Pathology structure, it has been suggested that chromatin condensation is
a predictive criterion for Alzheimer’s disease progression.[152] As
Pathology of the neural system is extremely complex owing to we found in these examples, spatial transcriptomics has the po-
cellular heterogeneity and complex interactions among resident tential to provide high-dimensional molecular profiling to con-
cells. Spatially resolved transcriptomics offers new perspectives ventional neurodevelopmental and neuropathological investiga-
on how these interactions influence the study of pathological tions. Since neuroscience is a multidisciplinary field with various
mechanisms. In mechanistic studies with animal models, spa- tools available for research, the addition of spatial information us-
tial transcriptomics can validate specific hypotheses of causal- ing spatial transcriptomics is expected to dramatically expand the
ity of diseases with spatial context[141–143] or directly character- boundaries of existing knowledge.
ize a new concept to find disease-related features from spatial Spatial transcriptomics is still a new approach in neuro-
correlation.[144,145] Chen et al. is a representative study demon- science research, particularly for clinical studies. By enabling
strating how to generate and validate new hypotheses concerning high-throughput molecular profiling with spatial contexts, it will
correlations between genetic signatures and pathological struc- offer a unique opportunity to comprehend complex biological
tures related to Alzheimer’s disease. Novel genetic signatures for systems composed of intricate cell-to-cell interactions. Addition-
Alzheimer’s disease were suggested by the identification of gene ally, spatial transcriptomics holds promise as a valuable screen-
network alterations in the periphery of 𝛽-amyloid plaques us- ing tool to enhance disease diagnosis accuracy, which will be fur-
ing low-resolution spatial transcriptomics and characterization ther discussed in the following section.
of cellular identity for cells possessing altered gene expression
using ISS.[146] 6. Application Note in Cancer Studies
Studying pathology in the human brain requires precise target-
ing, as its physical scale generally exceeds the currently available Heterogeneity increases as cancer transitions from a benign to
biochemical tools used for small animals, and these procedures malignant state. The increasing heterogeneity hinders the pre-
require pathologists who can distinguish between physiologically vention and treatment of cancer. Cellular components of the tu-
normal and diseased regions within a specimen. This encour- mor immune microenvironment (TIME) and communication
ages the application of spatial transcriptomics to clinical speci- between cells in the TIME are associated with cancer progno-
mens to characterize the biological features that can be discrim- sis and response to therapies.[153] Spatial transcriptomics can
inated from a diseased partition of tissues compared to adjacent help clinicians and scientists discover new cell types and co-
healthy tissues. Candidate genes associated with hereditary amy- localization patterns, characterize the TIME, and monitor tumor
otrophic lateral sclerosis can be identified by detecting cerebel- response to therapy (Figure 5).
lar granule cell layer-specific transcripts in H&E-stained sections By linking spatial transcriptomes with scRNA-seq data, new
of post-mortem tissue specimens.[147] Kaufmann et al.[148] identi- cell types can be identified, and cellular interactions can be in-
fied the T-cell subtype crossing the blood-brain barrier to colonize ferred from cell-type co-localization patterns. In a study of human

Adv. Sci. 2023, 10, 2206939 2206939 (13 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Figure 5. Applications of spatial transcriptomics in cancer research. Spatially resolved transcriptomics can be used to define the cell type compositions
and discover new cell–cell interactions of specific tumor ecosystems, profile and characterize these ecosystems by utilization and integration in multi-
modal/omics analyses, and help fuel the joint renewal of current histopathological standards to accommodate these new findings. (Figure created with
icons and redesigned templates provided by Biorender.com).

cutaneous squamous cell carcinoma (cSCC), researchers iden- The spatial results help to weigh in on the speculation that tumor
tified a specific cell type, tumor-specific keratinocytes (TSKs), cells could exploit TIGIT-Nectininteraction to evade the immune
crowding the leading edge of the tumor and confirmed a sur- response.
rounding fibrovascular niche consisting of cancer-associated fi- Spatial transcriptomics is emerging and gradually gaining
broblasts (CAFs) and endothelial cells.[154] This confirmation of recognition as an essential tool for sifting through the tumor mi-
spatial transcriptomics colocalization was used to support in- croenvironment. A spatial transcriptome constructed from 21 tis-
ferences on cell interaction drawn from scRNA-seq, and subse- sues progressing from non-tumor to ecotone, to tumor regions
quently, TSKs were found to be a pillar of intercellular commu- of primary lung cancer revealed that the tumor capsule was in-
nication. Some immune cells were also found to possibly hin- volved in transcriptome complexity, immune cell infiltrations,
der effector lymphocytes from accessing the tumor. These dis- and continuity of intratumor spatial clusters; intratumor archi-
coveries showcase the potential of spatial transcriptomics for dis- tecture was supported by bidirectional ligand-receptor signaling
covering targets for therapeutic intervention by immune con- at the cluster perimeters; and PROM1+ , CD47+ cancer stem cell
trol. Overall, the cell subpopulations constituting the cSCC tu- niches contributed to the remodeling of the tumor microenvi-
mor and stroma, and interaction among their spatial niche could ronment and metastasis.[157] Hunter et al. described an “inter-
be characterized. Another study reinforced previous knowledge face” cell state located at the ecotone of tumor and normal tis-
that epithelial-to-mesenchymal transition (EMT) and prolifera- sues, including the possibility that this new cell state induces
tion are inversely correlated—for example, an exit from the EMT tumor invasion into surrounding tissues.[158] Although the se-
is required to enter proliferation, by reporting the cellular rela- quencing resolution and depth of spatial transcriptomic data may
tionship of cells of the two states. Cell states related to the respec- not be comparable to those of scRNA-seq data, sufficient spa-
tive processes were segregated into their own distinct zones, the tial data are preserved for linking and deconvoluting the spatial
regions being mutually exclusive.[155] In the case of pancreatic transcriptomic data with scRNA-seq data. Analysis of spatial tran-
neoplasms, spatial transcriptomics profiling directly confirmed scriptome data can indicate specific transcriptomic signals orig-
the initial annotations of pancreatic intraepithelial neoplasia con- inating from the location of cellular aggregation and enables ad-
structed with scRNA-seq data.[156] Furthermore, receptor-ligand ditional insights into cell types and interactions that may assist
analyses suggested the interaction of the T cell immunorecep- TIME formation. This can help determine whether specific cell–
tor with Ig and ITIM domains (TIGIT) receptor of lymphocytes cell interactions are enhanced in segregated niches or separately
with Nectinligands, TIGIT being highly expressed in tumor cells. organized structures.[155,159,160]

Adv. Sci. 2023, 10, 2206939 2206939 (14 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Spatial transcriptomic data along with scRNA-seq have be- detect and classify specific tumor subregions, streamlining the
come routine procedures in exploring and characterizing tis- laborious task of manual annotation. Additionally, they can be
sues, cell types, and diseases. Consequently, spatially resolved quickly trained on large datasets and minimize potential errors
multi-omics studies will be increased in cancer biology. A com- and variations that may arise from human annotation. These ad-
bination of scRNA-seq, spatial transcriptomics, genomics, and vantages will aid in making better decisions regarding the spe-
metabolomics was used to reveal spatially segregated transcrip- cific treatment type for matching specific disease stages or sub-
tion patterns, which revealed distinct genomic alterations and types. With the support of future research, existing histopatho-
identified hypoxia as a driver for genomic instability.[160] logical diagnosis criteria may need to undergo minor to major
The inference of copy number alterations (CNAs) using spatial updates to accommodate the new findings presented by spatial
transcriptomics is consistent with bulk whole-exome sequenc- transcriptomics.
ing (WES) data, thereby reinforcing its practicality.[157] Distinct The use of spatial transcriptomics holds promise in cancer de-
clonal patterns of spatial CNAs are found near and within tu- tection and subtype classification, however at its current state is
mors and could be used to distinguish tumors from normal and in need of more case studies, data, and process standardization
transitional cell states. These clonal CNA patterns located in can- before its use as a clinical classifier.[166,171] With more spatial can-
cer drivers are not associated with immediately visible morpho- cer data acquired through movements such as the Human Tumor
logical transformation; therefore, they are strongly proposed as a Atlas Network launched by the US National Cancer Institute in
measure for early diagnosis of cancer.[161] 2018 to construct cellular, morphological, and molecular features
Novel transcripts, such as fusion transcripts and alternative of progressing cancer, and with higher resolution, researchers
splicing variants, are frequently observed in various cancers. Spa- will be more equipped to identify the key genes and regulators in
tial transcriptomics fusion (STfusion)[162] can infer the existence the processes of mutation acquisition, factors differentiating hot
of fusion transcripts from spatial transcriptomic data. By apply- and cold tumors, and tumor invasion.[172] In summary, spatial
ing this method to prostate cancer, cis-SAGe SLC45A3-ELK4 has transcriptomics technology enables the dissection of tumor het-
been detected mostly in physiologically altered, inflamed, or neo- erogeneity, and its ongoing developments will reshape the can-
plastic areas. Long-read sequencing that enables the profiling of cer research landscape and open up new possibilities for precise
full-length transcripts has been applied to spatial transcriptomics prognosis and treatment of cancer.
and scRNA-seq.[163] Full-length spatial transcriptomics can be
used to explore alternative splicing events during tumorigenesis. 7. Conclusion and Future Perspectives
Recently, base-specific in situ sequencing (BaSISS) of muta-
tion branches that were obtained from whole genome sequenc- Here, we have provided a review of the recent technical aspects
ing was utilized to simultaneously track the cancer-specific so- of spatial transcriptomics and related fields of application. Spa-
matic mutations arising in breast cancer.[164] Spatial transcrip- tial transcriptomic technologies are rapidly evolving. This review
tomics was integrated to generate maps that quantitate these ge- provides historical background and practical examples in both
netically distinct subclones. Genetically similar subclones were imaging- and NGS-based spatial transcriptomics.
found to exhibit similar co-localization patterns and histological Using spatial transcriptomics is challenging and the scale of
features regardless of their spatial vicinity, in the same way, dis- studies should be carefully considered. This sample scale is di-
tinct subclones localized with different groups of immune cells, rectly connected to the method of choice, and one may still strug-
possibly mediating clone-specific immune interactions. gle to use imaging-based techniques for whole-brain slices. In-
The discovery of new factors influencing tumorigenesis and stead, gross expression patterns at a voxel of 10–20 μm are prac-
characterizing malignancy using spatial transcriptomics will tically available on NGS-based commercial platforms. To reduce
bring new diagnosis criteria and standards for cancer research. experimental cost and improve single-cell clustering quality, sam-
As the niches of TIME, cell-type interactions, and cancer ex- ples can be split into parts, one for non-spatial regular single-cell
pression regions become unveiled, they deviate from tradition- RNA-seq and the other for the spatial version of the technique.
ally annotated regions of cancer.[165] The future of automating Computational analysis is key to investigating spatial informa-
the process of histopathology-based diagnosis is near, especially tion, and established commercial platforms with modification
with the advent of machine learning technology and artificial will soon be available for ease of use.
intelligence.[166–170] These technology-based methods are trained Spatial technologies are bringing new horizons in investigat-
on pathologist-annotated histology slides and spatial transcrip- ing epigenomes, chromosome accessibility, chromosomal struc-
tomics data, then are able to distinguish healthy and diseased ar- tures, proteomes, CRISPR gene perturbation, lipid nanoparticles
eas of the tissue, infer gene expression levels to spatially char- for gene therapy and mRNA vaccine development, AAV serotype
acterize tumor heterogeneity, and detect expression-supported optimization, and T-cell receptor (TCR) sequencing.
morphological patterns that are indistinguishable to the human Recent reports have demonstrated the spatial version of assay
eye. The application of a deep learning method trained on whole for transposase-accessible chromatin using sequencing (ATAC-
slide images to breast and lung cancer slides deduced a statisti- seq).[173–175] By combining RNA expression profiling, spatial in-
cally significant link between high tumor heterogeneity and poor vestigation of chromosome accessibility can precisely define cell
survival.[167] A neural network-based method trained to relate his- types in neuroscience and cancer research. Updated protocols
tological morphology to the underlying gene expressions was not have been reported for cleavage under targets and tagmentation
only able to generally match the manual assessments by pathol- (CUT&Tag) methods optimized for investigating specific histone
ogists but also provided more detailed interpretation, taking the mark distribution, and such methods for the spatial version are
actual expression patterns into account.[168] These techniques can expected to be multiplexed.[176–178] Epigenomic MERFISH also

Adv. Sci. 2023, 10, 2206939 2206939 (15 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

achieved spatial analysis of single-cell level epigenetic features acids in human tissue specimens are readily degraded prior to
by in situ tagmentation followed by antibody-driven capturing of cryopreservation. Certain clinical institutions, such as university
epigenetic features.[179] hospitals, have limited resources for deep freezing to preserve
Cellular indexing of transcriptomes and epitopes (CITE- nucleic acids from nuclease digestion. Moreover, histological di-
seq)[180] and RNA expression and protein sequencing (REAP- agnosis and research are based on FFPEsamples; therefore, next-
seq)[181] showed that it is possible to simultaneously analyze generation technologies should be effective in analyzing heav-
single-cell transcriptome and proteome by employing DNA- ily fixed samples. 10X Genomics has recently announced two
barcoded antibodies. NGS-based spatial transcriptomics takes platforms for NGS-based and imaging-based spatial transcrip-
this a step further by incorporating antibody-derived tags during tomes for FFPE samples.[201] NanoString Technologies reported
the library generation, allowing for the analysis of spatial multi- the compatibility of their CosMx Spatial Molecular Imager (SMI)
omics data. Spatial multiomics methods such as CITE-seq with platform with FFPE samples.[202] Based on these commercial
ST,[182] 10X Visium,[183] and DBiT-seq[184] have facilitated the so- platforms, spatial transcriptomics can be more easily applied in
phisticated classification of cell types by integrating transcrip- clinical research.
tome and proteome data, yielding insights into the spatial orga- Finally, spatial information obtained from clinical samples
nization of these molecules in tissue. should yield new classes of biomarkers that are often underes-
Advances in clustered regularly interspaced short palindromic timated because of the lack of appropriate investigative technolo-
repeats (CRISPR) technology have provided versatile tools for gies. Therefore, seamlessly integrating multiple sources of tran-
genome editing and gene therapy.[185] Genome-wide CRISPR scriptome data that may or may not include spatial or molecular
knockout screens to cell models have helped identify im- information is important. Computational efforts to analyze mas-
portant genes in pathogenic pathways.[186,187] Pooled CRISPR- sive spatial transcriptomic data may help improve the success of
based screen with single-cell RNA-sequencing readouts (Perturb- clinical applications for marker discovery and drug development.
seq)[188–190] has been developed by combining CRISPR screen-
ing with single-cell sequencing methods and enables system-
atic investigation of the knock-out effect of individual genes at
the single-cell level. Additionally, Perturb-seq can explore high- Acknowledgements
throughput intracellular and cell-extrinsic spatial features at C.H.S. acknowledges that this work was supported by the Insti-
the single-cell level.[191] The recent spatial version of Perturb- tute for Basic Science (IBS), Center for Nanomedicine (IBS-R026-
seq, Perturb-map, utilizes combinatorial protein barcoding for D1 to C.H.S.), the Yonsei University Research Fund (2022-22-0093
a detailed spatial context of interactions between the “CRISPR- to C.H.S.), the collaborative research grant funding provided by the
perturbed” and surrounding cells. These toolkits will facilitate Halle Institute for Global Research at Emory University and Yon-
research on genomic and functional aspects of model organisms sei University (2021-12-0006 to C.H.S.), the National Research Foun-
dation (NRF) of Korea, funded by the Ministry of Science (no.
and organs with spatial molecular information on cell–cell inter- 2021R1C1C1004092 to C.H.S, 2021H1D3A2A02096517 to C.H.S., C.P.M,
actions. 2021R1A2C2005294 and 2021M3A9H3015690 to T.K, 2019R1C1C1005403
Lipid nanoparticles (LNPs) are evolving for improved RNA- to J.P., 2021R1F1A1047198 to C.W.L., and 2021M3A9I4024452 to J.K.H.).
based therapeutics[192] and gene-editing tools for delivery.[193] Bar-
coded LNPs have been utilized to assess heterogeneity in cellular
expression and its effect on LNP-mediated mRNA delivery.[194]
The spatial version of this technology may open a new era for Conflict of Interest
gene therapy and its efficacy assay with massive data to better
The authors declare no conflict of interest.
understand the process of LNP formulation and mRNA inserts.
AAV virus serotypes have been actively optimized for target-
specific gene therapy.[195,196] Organ and cell-type specificities have
been optimized by screening capsid proteins of AAV serotypes at Author Contributions
single-cell resolution.[197,198] Spatial and temporal information ac-
quired from spatial transcriptomics of optimized candidates with S.H.J. and R.H.L. contributed equally to this work. H.-E.P. and C.H.S. con-
AAV serotype barcodes can be employed, and by reading these ceptualized the contributions of all technologies. H.-E.P. and C.H.S. wrote
the part for imaging and NGS-based methods, application notes in neuro-
barcodes from each tissue and cell type, the efficiency of AAV in- science, and future directions and conclusions. S.H.J. and T.K. wrote the
fection can be determined in detail, which should be useful in computational analysis part. R.H.L. and J.P. wrote the application notes
further optimization. for cancer studies. C.P.M., C.W.L., and J.K.H. reviewed and revised the
Spatial transcriptomics and TCR sequencing are essential for manuscript and provided critical comments.
tracing and dissecting T-cell infiltration and their interactions in
metastatic cancer tissues.[45,199] A recent product from 10X Ge-
nomics allows simultaneous investigation of mRNA and TCR in-
formation from H&E-stained tissue slides.[200] High resolution Keywords
and coverage are crucial in understanding the tumor microenvi- bioinformatics, cancer, neuroscience, RNA, spatial transcriptomics
ronments and metastases; therefore, the sensitivity and coverage
of sequences need further improvement. Received: November 25, 2022
For clinical applications, spatial transcriptomics should be ef- Revised: March 10, 2023
fective in clinical practice and sampling environments. Nucleic Published online: April 7, 2023

Adv. Sci. 2023, 10, 2206939 2206939 (16 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

[36] X. Qian, K. D. Harris, T. Hauling, D. Nicoloutsopoulos, A. B. Muñoz-

[1] J. Lewis, Science 2008, 322, 399. Manchado, N. Skene, J. Hjerling-Leffler, M. Nilsson, Nat. Methods
[2] A. D. Lander, Science 2013, 339, 923. 2020, 17, 101.
[3] F. Cremisi, A. Philpott, S. Ohnuma, Curr. Opin. Neurobiol. 2003, 13, [37] D. Gyllborg, C. M. Langseth, X. Qian, E. Choi, S. M. Salas, M. M.
26. Hilscher, E. S. Lein, M. Nilsson, Nucleic Acids Res. 2020, 48, e112.
[4] E. Fuchs, T. Tumbar, G. Guasch, Cell 2004, 116, 769. [38] G. La Manno, K. Siletti, A. Furlan, D. Gyllborg, E. Vinsland, A. Mossi
[5] P. Rompolas, K. R. Mesa, V. Greco, Nature 2013, 502, 513. Albiach, C. Mattsson Langseth, I. Khven, A. R. Lederer, L. M. Dratva,
[6] S. P. Kerkar, N. P. Restifo, Cancer Res. 2012, 72, 3125. A. Johnsson, M. Nilsson, P. Lönnerberg, S. Linnarsson, Nature 2021,
[7] Z. Wang, M. Gerstein, M. Snyder, Nat. Rev. Genet. 2009, 10, 57. 596, 92.
[8] R. Sandberg, Nat. Methods 2014, 11, 22. [39] S. Liu, S. Punthambaker, E. P. R. Iyer, T. Ferrante, D. Goodwin, D.
[9] V. Svensson, R. Vento-Tormo, S. A. Teichmann, Nat. Protoc. 2018, Fürth, A. C. Pawlowski, K. Jindal, J. M. Tam, L. Mifflin, S. Alon, A.
13, 599. Sinha, A. T. Wassie, F. Chen, A. Cheng, V. Willocq, K. Meyer, K.-
[10] M. Asp, J. Bergenstråhle, J. Lundeberg, BioEssays 2020, 42, 1900221. H. Ling, C. K. Camplisson, R. E. Kohman, J. Aach, J. H. Lee, B. A.
[11] A. Rao, D. Barkley, G. S. França, I. Yanai, Nature 2021, 596, 211. Yankner, E. S. Boyden, G. M. Church, Nucleic Acids Res. 2021, 49,
[12] L. Moses, L. Pachter, Nat. Methods 2022, 19, 534. e58.
[13] C. G. Williams, H. J. Lee, T. Asatsuma, R. Vento-Tormo, A. Haque, [40] H. Lee, S. M. Salas, D. Gyllborg, M. Nilsson, Sci. Rep. 2022, 12,
Genome Med. 2022, 14, 68. 7976.
[14] L. Tian, F. Chen, E. Z. Macosko, Nat. Biotechnol. 2022, https://doi. [41] Z. Yao, C. T. J. van Velthoven, M. Kunst, M. Zhang, D. McMillen, C.
org/10.1038/s41587-022-01448-2. Lee, W. Jung, J. Goldy, A. Abdelhak, P. Baker, E. Barkan, D. Bertag-
[15] A. M. Femino, F. S. Fay, K. Fogarty, R. H. Singer, Science 1998, 280, nolli, J. Campos, D. Carey, T. Casper, A. B. Chakka, R. Chakrabarty,
585. S. Chavan, M. Chen, M. Clark, J. Close, K. Crichton, S. Daniel, T. Dol-
[16] A. Raj, P. van den Bogaard, S. A. Rifkin, A. van Oudenaarden, S. beare, L. Ellingwood, J. Gee, A. Glandon, J. Gloe, J. Gould, J. Gray,
Tyagi, Nat. Methods 2008, 5, 877. et al., bioRxiv: 2023.03.06.531121 2023.
[17] F. Wang, J. Flanagan, N. Su, L.-C. Wang, S. Bui, A. Nielson, X. Wu, [42] S. M. Salas, P. Czarnewski, L. B. Kuemmerle, S. Helgadottir, C.
H.-T. Vo, X.-J. Ma, Y. Luo, J. Mol. Diagn. 2012, 14, 22. Matsson-Langseth, S. Tismeyer, C. Avenel, H. Rehman, K. Tiklova,
[18] S. Codeluppi, L. E. Borm, A. Zeisel, G. La Manno, J. A. van Lunteren, A. Andersson, M. Chatzinikolaou, F. J. Theis, M. D. Luecken, C.
C. I. Svensson, S. Linnarsson, Nat. Methods 2018, 15, 932. Wählby, N. Ishaque, M. Nilsson, bioRxiv: 2023.02.13.528102 2023.
[19] M. R. Speicher, S. G. Ballard, D. C. Ward, Nat. Genet. 1996, 12, 368. [43] S. Goodwin, J. D. McPherson, W. R. McCombie, Nat. Rev. Genet.
[20] E. Lubeck, L. Cai, Nat. Methods 2012, 9, 743. 2016, 17, 333.
[21] E. Lubeck, A. F. Coskun, T. Zhiyentayev, M. Ahmad, L. Cai, Nat. Meth- [44] S. Nichterwitz, G. Chen, J. Aguila Benitez, M. Yilmaz, H. Storvall,
ods 2014, 11, 360. M. Cao, R. Sandberg, Q. Deng, E. Hedlund, Nat. Commun. 2016, 7,
[22] K. H. Chen, A. N. Boettiger, J. R. Moffitt, S. Wang, X. Zhuang, Science 12139.
2015, 348, aaa6090. [45] A. C. Lee, Y. Lee, A. Choi, H.-B. Lee, K. Shin, H. Lee, J. Y. Kim, H. S.
[23] S. Shah, E. Lubeck, W. Zhou, L. Cai, Neuron 2016, 92, 342. Ryu, H. S. Kim, S. Y. Ryu, S. Lee, J.-H. Cheun, D. K. Yoo, S. Lee, H.
[24] G. Wang, J. R. Moffitt, X. Zhuang, Sci. Rep. 2018, 8, 4847. Choi, T. Ryu, H. Yeom, N. Kim, J. Noh, Y. Lee, I. Kim, S. Bae, J. Kim,
[25] J. R. Moffitt, D. Bambah-Mukku, S. W. Eichhorn, E. Vaughn, K. W. Lee, O. Kim, Y. Jung, C. Kim, S. W. Song, Y. Choi, J. Chung, et al.,
Shekhar, J. D. Perez, N. D. Rubinstein, J. Hao, A. Regev, C. Dulac, Nat. Commun. 2022, 13, 2540.
X. Zhuang, Science 2018, 362, eaau5324. [46] J. P. Junker, E. S. Noël, V. Guryev, K. A. Peterson, G. Shah, J. Huisken,
[26] C.-H. L. Eng, M. Lawson, Q. Zhu, R. Dries, N. Koulena, Y. Takei, J. A. P. McMahon, E. Berezikov, J. Bakkers, A. van Oudenaarden, Cell
Yun, C. Cronin, C. Karp, G.-C. Yuan, L. Cai, Nature 2019, 568, 235. 2014, 159, 662.
[27] J. R. Moffitt, J. Hao, D. Bambah-Mukku, T. Lu, C. Dulac, X. Zhuang, [47] J. Chen, S. Suo, P. P. Tam, J.-D. J. Han, G. Peng, N. Jing, Nat. Protoc.
Proc. Natl. Acad. Sci. USA 2016, 113, 14456. 2017, 12, 566.
[28] J. R. Moffitt, J. Hao, G. Wang, K. H. Chen, H. P. Babcock, X. Zhuang, [48] C. Medaglia, A. Giladi, L. Stoler-Barak, M. De Giovanni, T. M.
Proc. Natl. Acad. Sci. USA 2016, 113, 11046. Salame, A. Biram, E. David, H. Li, M. Iannacone, Z. Shulman, I.
[29] L. E. Borm, A. Mossi Albiach, C. C. A. Mannens, J. Janusauskas, C. Amit, Science 2017, 358, 1622.
Özgün, D. Fernández-García, R. Hodge, F. Castillo, C. R. H. Hedin, [49] K. H. Hu, J. P. Eichorst, C. S. McGinnis, D. M. Patterson, E. D. Chow,
E. J. Villablanca, P. Uhlén, E. S. Lein, S. Codeluppi, S. Linnarsson, K. Kersten, S. C. Jameson, Z. J. Gartner, A. A. Rao, M. F. Krummel,
Nat. Biotechnol. 2023, 41, 222. Nat. Methods 2020, 17, 833.
[30] R. Ke, M. Mignardi, A. Pacureanu, J. Svedlund, J. Botling, C. Wählby, [50] J. Y. Kishi, N. Liu, E. R. West, K. Sheng, J. J. Jordanides, M. Serrata,
M. Nilsson, Nat. Methods 2013, 10, 857. C. L. Cepko, S. K. Saka, P. Yin, Nat. Methods 2022, 19, 1393.
[31] J. H. Lee, E. R. Daugharthy, J. Scheiman, R. Kalhor, J. L. Yang, T. C. [51] C. R. Merritt, G. T. Ong, S. E. Church, K. Barker, P. Danaher, G. Geiss,
Ferrante, R. Terry, S. S. F. Jeanty, C. Li, R. Amamoto, D. T. Peters, B. M. Hoang, J. Jung, Y. Liang, J. McKay-Fleisch, K. Nguyen, Z. Nor-
M. Turczyk, A. H. Marblestone, S. A. Inverso, A. Bernard, P. Mali, X. gaard, K. Sorg, I. Sprague, C. Warren, S. Warren, P. J. Webster, Z.
Rios, J. Aach, G. M. Church, Science 2014, 343, 1360. Zhou, D. R. Zollinger, D. L. Dunaway, G. B. Mills, J. M. Beechem,
[32] J. H. Lee, E. R. Daugharthy, J. Scheiman, R. Kalhor, T. C. Ferrante, R. Nat. Biotechnol. 2020, 38, 586.
Terry, B. M. Turczyk, J. L. Yang, H. S. Lee, J. Aach, K. Zhang, G. M. [52] J. W. Foley, C. Zhu, P. Jolivet, S. X. Zhu, P. Lu, M. J. Meaney, R. B.
Church, Nat. Protoc. 2015, 10, 442. West, Genome Res. 2019, 29, 1816.
[33] X. Wang, W. E. Allen, M. A. Wright, E. L. Sylwestrak, N. Samusik, S. [53] L. Brady, M. Kriner, I. Coleman, C. Morrissey, M. Roudier, L. D. True,
Vesuna, K. Evans, C. Liu, C. Ramakrishnan, J. Liu, G. P. Nolan, F.-A. R. Gulati, S. R. Plymate, Z. Zhou, B. Birditt, R. Meredith, G. Geiss,
Bava, K. Deisseroth, Science 2018, 361, eaat5691. M. Hoang, J. Beechem, P. S. Nelson, Nat. Commun. 2021, 12, 1426.
[34] X. Chen, Y.-C. Sun, G. M. Church, J. H. Lee, A. M. Zador, Nucleic [54] K. Pelka, M. Hofree, J. H. Chen, S. Sarkizova, J. D. Pirl, V. Jorgji, A.
Acids Res. 2018, 46, e22. Bejnood, D. Dionne, W. H. Ge, K. H. Xu, S. X. Chao, D. R. Zollinger,
[35] L. N. Waylen, H. T. Nim, L. G. Martelotto, M. Ramialison, Commun. D. J. Lieb, J. W. Reeves, C. A. Fuhrman, M. L. Hoang, T. Delorey, L.
Biol. 2020, 3, 602. T. Nguyen, J. Waldman, M. Klapholz, I. Wakiro, O. Cohen, J. Albers,

Adv. Sci. 2023, 10, 2206939 2206939 (17 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

C. S. Smillie, M. S. Cuoco, J. Wu, M. Su, J. Yeung, B. Vijaykumar, A. [74] S. Alon, D. R. Goodwin, A. Sinha, A. T. Wassie, F. Chen, E. R. Daugh-
M. Magnuson, et al., Cell 2021, 184, 4734. arthy, Y. Bando, A. Kajita, A. G. Xue, K. Marrett, R. Prior, Y. Cui, A.
[55] H. Bergholtz, J. Carter, A. Cesano, M. Cheang, S. Church, P. Divakar, C. Payne, C.-C. Yao, H.-J. Suk, R. Wang, C.-C. (Jay) Yu, P. Tillberg,
C. Fuhrman, S. Goel, J. Gong, J. Guerriero, M. Hoang, E. Hwang, H. P. Reginato, N. Pak, S. Liu, S. Punthambaker, E. P. R. Iyer, R. E.
Kuasne, J. Lee, Y. Liang, E. Mittendorf, J. Perez, A. Prat, L. Pusztai, J. Kohman, J. A. Miller, E. S. Lein, A. Lako, N. Cullen, S. Rodig, K.
Reeves, Y. Riazalhosseini, J. Richer, Ö. Sahin, H. Sato, I. Schlam, T. Helvie, et al., Science 2021, 371, eaax2656.
Sørlie, D. Stover, S. Swain, A. Swarbrick, E. Thompson, et al., Cancers [75] Y. Cui, G. Yang, D. R. Goodwin, C. H. O’Flanagan, A. Sinha, C.
2021, 13, 4456. Zhang, K. E. Kitko, D. Park, S. Aparicio, I. Consortium, E. S. Boy-
[56] P. L. Ståhl, F. Salmén, S. Vickovic, A. Lundmark, J. F. Navarro, J. Mag- den, bioRxiv: 2022.06.19.496699 2022.
nusson, S. Giacomello, M. Asp, J. O. Westholm, M. Huss, A. Moll- [76] Z. Ni, A. Prasad, S. Chen, R. B. Halberg, L. M. Arkin, B. A. Drolet,
brink, S. Linnarsson, S. Codeluppi, Å. Borg, F. Pontén, P. I. Costea, M. A. Newton, C. Kendziorski, Nat. Commun. 2022, 13, 2971.
P. Sahlén, J. Mulder, O. Bergmann, J. Lundeberg, J. Frisén, Science [77] Y. Wang, B. Song, S. Wang, M. Chen, Y. Xie, G. Xiao, L. Wang, T.
2016, 353, 78. Wang, Nat. Methods 2022, 19, 950.
[57] S. G. Rodriques, R. R. Stickels, A. Goeva, C. A. Martin, E. Murray, C. [78] T. Abdelaal, S. Mourragui, A. Mahfouz, M. J. T. Reinders, Nucleic
R. Vanderburg, J. Welch, L. M. Chen, F. Chen, E. Z. Macosko, Science Acids Res. 2020, 48, e107.
2019, 363, 1463. [79] R. Lopez, A. Nazaret, M. Langevin, J. Samaran, J. Regier, M. I. Jor-
[58] R. R. Stickels, E. Murray, P. Kumar, J. Li, J. L. Marshall, D. J. Di Bella, dan, N. Yosef, arXiv: 1905.02269 2019.
P. Arlotta, E. Z. Macosko, F. Chen, Nat. Biotechnol. 2021, 39, 313. [80] C. Hafemeister, R. Satija, Genome Biol. 2019, 20, 296.
[59] S. Vickovic, G. Eraslan, F. Salmén, J. Klughammer, L. Stenbeck, D. [81] D. Pham, X. Tan, J. Xu, L. F. Grice, P. Y. Lam, A. Raghubar, J. Vukovic,
Schapiro, T. Äijö, R. Bonneau, L. Bergenstråhle, J. F. Navarro, J. M. J. Ruitenberg, Q. Nguyen, bioRxiv: 2020.05.31.125658 2020.
Gould, G. K. Griffin, Å. Borg, M. Ronaghi, J. Frisén, J. Lundeberg, [82] R. Satija, J. A. Farrell, D. Gennert, A. F. Schier, A. Regev, Nat. Biotech-
A. Regev, P. L. Ståhl, Nat. Methods 2019, 16, 987. nol. 2015, 33, 495.
[60] C.-S. Cho, J. Xi, Y. Si, S.-R. Park, J.-E. Hsu, M. Kim, G. Jun, H. M. [83] K. Achim, J.-B. Pettit, L. R. Saraiva, D. Gavriouchkina, T. Larsson, D.
Kang, J. H. Lee, Cell 2021, 184, 3559. Arendt, J. C. Marioni, Nat. Biotechnol. 2015, 33, 503.
[61] X. Fu, L. Sun, R. Dong, J. Y. Chen, R. Silakit, L. F. Condon, Y. Lin, S. [84] T. Stuart, A. Butler, P. Hoffman, C. Hafemeister, E. Papalexi, W. M.
Lin, R. D. Palmiter, L. Gu, Cell 2022, 185, 4621. Mauck, Y. Hao, M. Stoeckius, P. Smibert, R. Satija, Cell 2019, 177,
[62] A. Chen, S. Liao, M. Cheng, K. Ma, L. Wu, Y. Lai, X. Qiu, J. Yang, J. 1888.
Xu, S. Hao, X. Wang, H. Lu, X. Chen, X. Liu, X. Huang, Z. Li, Y. Hong, [85] T. Peng, G. M. Chen, K. Tan, bioRxiv: 2021.01.25.427845 2021.
Y. Jiang, J. Peng, S. Liu, M. Shen, C. Liu, Q. Li, Y. Yuan, X. Wei, H. [86] Y. Goltsev, N. Samusik, J. Kennedy-Darling, S. Bhate, M. Hale, G.
Zheng, W. Feng, Z. Wang, Y. Liu, Z. Wang, et al., Cell 2022, 185, 1777. Vazquez, S. Black, G. P. Nolan, Cell 2018, 174, 968.
[63] Y. Liu, M. Yang, Y. Deng, G. Su, A. Enninful, C. C. Guo, T. Tebaldi, D. [87] F. Maseda, Z. Cang, Q. Nie, Front. Genet. 2021, 12, 636743.
Zhang, D. Kim, Z. Bai, E. Norris, A. Pan, J. Li, Y. Xiao, S. Halene, R. [88] T. Biancalani, G. Scalia, L. Buffoni, R. Avasthi, Z. Lu, A. Sanger, N.
Fan, Cell 2020, 183, 1665. Tokcan, C. R. Vanderburg, Å. Segerstolpe, M. Zhang, I. Avraham-
[64] S. R. Srivatsan, M. C. Regier, E. Barkan, J. M. Franks, J. S. Packer, P. Davidi, S. Vickovic, M. Nitzan, S. Ma, A. Subramanian, M. Lipinski,
Grosjean, M. Duran, S. Saxton, J. J. Ladd, M. Spielmann, C. Lois, P. J. Buenrostro, N. B. Brown, D. Fanelli, X. Zhuang, E. Z. Macosko, A.
D. Lampe, J. Shendure, K. R. Stevens, C. Trapnell, Science 2021, 373, Regev, Nat. Methods 2021, 18, 1352.
111. [89] V. Kleshchevnikov, A. Shmatko, E. Dann, A. Aivazidis, H. W. King,
[65] Y. Lee, D. Bogdanoff, Y. Wang, G. C. Hartoularos, J. M. Woo, C. T. T. Li, R. Elmentaite, A. Lomakin, V. Kedlian, A. Gayoso, M. S. Jain,
Mowery, H. M. Nisonoff, D. S. Lee, Y. Sun, J. Lee, S. Mehdizadeh, J. J. S. Park, L. Ramona, E. Tuck, A. Arutyunyan, R. Vento-Tormo, M.
Cantlon, E. Shifrut, D. N. Ngyuen, T. L. Roth, Y. S. Song, A. Marson, Gerstung, L. James, O. Stegle, O. A. Bayraktar, Nat. Biotechnol. 2022,
E. D. Chow, C. J. Ye, Sci. Adv. 2021, 7, eabg4755. 40, 661.
[66] H. R. Ueda, A. Ertürk, K. Chung, V. Gradinaru, A. Chédotal, P. Toman- [90] R. Lopez, B. Li, H. Keren-Shaul, P. Boyeau, M. Kedmi, D. Pilzer, A.
cak, P. J. Keller, Nat. Rev. Neurosci. 2020, 21, 61. Jelinski, I. Yofe, E. David, A. Wagner, C. Ergen, Y. Addadi, O. Golani,
[67] K. Chung, J. Wallace, S.-Y. Kim, S. Kalyanasundaram, A. S. Andal- F. Ronchese, M. I. Jordan, I. Amit, N. Yosef, Nat. Biotechnol. 2022,
man, T. J. Davidson, J. J. Mirzabekov, K. A. Zalocusky, J. Mattis, A. 40, 1360.
K. Denisin, S. Pak, H. Bernstein, C. Ramakrishnan, L. Grosenick, V. [91] D. M. Cable, E. Murray, L. S. Zou, A. Goeva, E. Z. Macosko, F. Chen,
Gradinaru, K. Deisseroth, Nature 2013, 497, 332. R. A. Irizarry, Nat. Biotechnol. 2022, 40, 517.
[68] B. Yang, J. B. Treweek, R. P. Kulkarni, B. E. Deverman, C.-K. Chen, E. [92] A. Andersson, J. Bergenstråhle, M. Asp, L. Bergenstråhle, A. Jurek,
Lubeck, S. Shah, L. Cai, V. Gradinaru, Cell 2014, 158, 945. J. Fernández Navarro, J. Lundeberg, Commun. Biol. 2020, 3, 565.
[69] E. L. Sylwestrak, P. Rajasethupathy, M. A. Wright, A. Jaffe, K. Deis- [93] R. Dong, G.-C. Yuan, Genome Biol. 2021, 22, 145.
seroth, Cell 2016, 164, 792. [94] M. Elosua-Bayes, P. Nieto, E. Mereu, I. Gut, H. Heyn, Nucleic Acids
[70] Y.-G. Park, C. H. Sohn, R. Chen, M. McCue, D. H. Yun, G. T. Drum- Res. 2021, 49, e50.
mond, T. Ku, N. B. Evans, H. C. Oak, W. Trieu, H. Choi, X. Jin, V. [95] S. Bae, K. J. Na, J. Koh, D. S. Lee, H. Choi, Y. T. Kim, Nucleic Acids
Lilascharoen, J. Wang, M. C. Truttmann, H. W. Qi, H. L. Ploegh, T. Res. 2022, 50, e57.
R. Golub, S.-C. Chen, M. P. Frosch, H. J. Kulik, B. K. Lim, K. Chung, [96] K. D. Yang, A. Belyaeva, S. Venkatachalapathy, K. Damodaran, A. Kat-
Nat. Biotechnol. 2019, 37, 73. coff, A. Radhakrishnan, G. V. Shivashankar, C. Uhler, Nat. Commun.
[71] F. Chen, P. W. Tillberg, E. S. Boyden, Science 2015, 347, 543. 2021, 12, 31.
[72] F. Chen, A. T. Wassie, A. J. Cote, A. Sinha, S. Alon, S. Asano, E. R. [97] M. R. Vahid, E. L. Brown, C. B. Steen, W. Zhang, H. S. Jeon, M. Kang,
Daugharthy, J.-B. Chang, A. Marblestone, G. M. Church, A. Raj, E. A. J. Gentles, A. M. Newman, Nat. Biotechnol. 2023, https://doi.org/
S. Boyden, Nat. Methods 2016, 13, 679. 10.1038/s41587-023-01697-9.
[73] Y. Wang, M. Eddison, G. Fleishman, M. Weigert, S. Xu, T. Wang, K. [98] B. Li, W. Zhang, C. Guo, H. Xu, L. Li, M. Fang, Y. Hu, X. Zhang, X.
Rokicki, C. Goina, F. E. Henry, A. L. Lemire, U. Schmidt, H. Yang, Yao, M. Tang, K. Liu, X. Zhao, J. Lin, L. Cheng, F. Chen, T. Xue, K. Qu,
K. Svoboda, E. W. Myers, S. Saalfeld, W. Korff, S. M. Sternson, P. W. Nat. Methods 2022, 19, 662.
Tillberg, Cell 2021, 184, 6361. [99] Y. Ma, X. Zhou, Nat. Biotechnol. 2022, 40, 1349.

Adv. Sci. 2023, 10, 2206939 2206939 (18 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

[100] E. A. G. Baker, D. Schapiro, B. Dumitrascu, S. Vickovic, A. Regev, Güntürkün, K. Hagihara, V. J. Hall, M. Helmstaedter, S. Herculano-
Nat. Methods 2023, 20, 424. Houzel, et al., Nat. Neurosci. 2020, 23, 1456.
[101] J. Bergenstråhle, L. Larsson, J. Lundeberg, BMC Genomics 2020, 21, [123] K. Morarach, A. Mikhailova, V. Knoflach, F. Memic, R. Kumar, W. Li,
482. P. Ernfors, U. Marklund, Nat. Neurosci. 2021, 24, 34.
[102] R. Zeira, M. Land, A. Strzalkowski, B. J. Raphael, Nat. Methods 2022, [124] X. Chen, Y.-C. Sun, H. Zhan, J. M. Kebschull, S. Fischer, K. Matho,
19, 567. Z. J. Huang, J. Gillis, A. M. Zador, Cell 2019, 179, 772.
[103] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, K. H. Maier-Hein, [125] Y.-C. Sun, X. Chen, S. Fischer, S. Lu, H. Zhan, J. Gillis, A. M. Zador,
Nat. Methods 2021, 18, 203. Nat. Neurosci. 2021, 24, 873.
[104] S. Mandal, V. Uhlmann, in 18th IEEE International Symposium on [126] A. S. Booeshaghi, Z. Yao, C. van Velthoven, K. Smith, B. Tasic, H.
Biomedical Imaging, IEEE, Piscataway, NJ 2021, p. 1082. Zeng, L. Pachter, Nature 2021, 598, 195.
[105] C. Stringer, T. Wang, M. Michaelos, M. Pachitariu, Nat. Methods [127] A. Joglekar, A. Prjibelski, A. Mahfouz, P. Collier, S. Lin, A. K.
2021, 18, 100. Schlusche, J. Marrocco, S. R. Williams, B. Haase, A. Hayes, J. G.
[106] M. Pachitariu, C. Stringer, Nat. Methods 2022, 19, 1634. Chew, N. I. Weisenfeld, M. Y. Wong, A. N. Stein, S. A. Hardwick, T.
[107] Y. He, X. Tang, J. Huang, J. Ren, H. Zhou, K. Chen, A. Liu, H. Shi, Hunt, Q. Wang, C. Dieterich, Z. Bent, O. Fedrigo, S. A. Sloan, D.
Z. Lin, Q. Li, A. Aditham, J. Ounadjela, E. I. Grody, J. Shu, J. Liu, X. Risso, E. D. Jarvis, P. Flicek, W. Luo, G. S. Pitt, A. Frankish, A. B.
Wang, Nat. Commun. 2021, 12, 5909. Smit, M. E. Ross, H. U. Tilgner, Nat. Commun. 2021, 12, 463.
[108] V. Petukhov, R. J. Xu, R. A. Soldatov, P. Cadinu, K. Khodosevich, J. R. [128] T. N. Lerner, L. Ye, K. Deisseroth, Cell 2016, 164, 1136.
Moffitt, P. V. Kharchenko, Nat. Biotechnol. 2022, 40, 345. [129] L. Luo, E. M. Callaway, K. Svoboda, Neuron 2018, 98, 256.
[109] S. Prabhakaran, Bioinf. Adv. 2022, 2, vbac048. [130] C. Grienberger, A. Konnerth, Neuron 2012, 73, 862.
[110] J. Park, W. Choi, S. Tiesmeyer, B. Long, L. E. Borm, E. Garren, T. N. [131] W. Yang, R. Yuste, Nat. Methods 2017, 14, 349.
Nguyen, B. Tasic, S. Codeluppi, T. Graf, M. Schlesner, O. Stegle, R. [132] E. Lein, L. E. Borm, S. Linnarsson, Science 2017, 358, 64.
Eils, N. Ishaque, Nat. Commun. 2021, 12, 3545. [133] J. L. Close, B. R. Long, H. Zeng, Nat. Methods 2021, 18, 23.
[111] C. Ortiz, M. Carlén, K. Meletis, Annu. Rev. Neurosci. 2021, 44, [134] A. M. Kerlin, M. L. Andermann, V. K. Berezovskii, R. C. Reid, Neuron
547. 2010, 67, 858.
[112] C. Ortiz, J. F. Navarro, A. Jurek, A. Märtin, J. Lundeberg, K. Meletis, [135] A. G. Khan, J. Poort, A. Chadwick, A. Blot, M. Sahani, T. D. Mrsic-
Sci. Adv. 2020, 6, eabb3446. Flogel, S. B. Hofer, Nat. Neurosci. 2018, 21, 851.
[113] M. Zhang, S. W. Eichhorn, B. Zingg, Z. Yao, K. Cotter, H. Zeng, H. [136] T. Geiller, B. Vancura, S. Terada, E. Troullinou, S. Chavlis, G.
Dong, X. Zhuang, Nature 2021, 598, 137. Tsagkatakis, P. Tsakalides, K. Ócsai, P. Poirazi, B. J. Rózsa, A. Loson-
[114] H. Shi, Y. He, Y. Zhou, J. Huang, B. Wang, Z. Tang, P. Tan, M. Wu, czy, Neuron 2020, 108, 968.
Z. Lin, J. Ren, Y. Thapa, X. Tang, A. Liu, J. Liu, X. Wang, bioRxiv: [137] S. Xu, H. Yang, V. Menon, A. L. Lemire, L. Wang, F. E. Henry, S. C.
2022.06.20.496914 2022. Turaga, S. M. Sternson, Science 2020, 370, eabb2494.
[115] E. S. Lein, M. J. Hawrylycz, N. Ao, M. Ayres, A. Bensinger, A. Bernard, [138] C. Condylis, A. Ghanbari, N. Manjrekar, K. Bistrong, S. Yao, Z. Yao,
A. F. Boe, M. S. Boguski, K. S. Brockway, E. J. Byrnes, L. Chen, L. T. N. Nguyen, H. Zeng, B. Tasic, J. L. Chen, Science 2022, 375,
Chen, T.-M. Chen, M. Chi Chin, J. Chong, B. E. Crook, A. Czaplinska, eabl5981.
C. N. Dang, S. Datta, N. R. Dee, A. L. Desaki, T. Desta, E. Diep, T. A. [139] Q. Zhao, C. D. Yu, R. Wang, Q. J. Xu, R. Dai Pra, L. Zhang, R. B.
Dolbeare, M. J. Donelan, H.-W. Dong, J. G. Dougherty, B. J. Duncan, Chang, Nature 2022, 603, 878.
A. J. Ebbert, G. Eichele, et al., Nature 2007, 445, 168. [140] S. Bugeon, J. Duffield, M. Dipoppa, A. Ritoux, I. Prankerd, D. Ni-
[116] BRAIN Initiative Cell Census Network (BICCN), Nature 2021, 598, coloutsopoulos, D. Orme, M. Shinn, H. Peng, H. Forrest, A. Vid-
86. uolyte, C. B. Reddy, Y. Isogai, M. Carandini, K. D. Harris, Nature
[117] H. Zeng, J. R. Sanes, Nat. Rev. Neurosci. 2017, 18, 530. 2022, 607, 330.
[118] A. Zeisel, A. B. Muñoz-Manchado, S. Codeluppi, P. Lönnerberg, G. [141] E. F. Willis, K. P. A. MacDonald, Q. H. Nguyen, A. L. Garrido, E. R.
La Manno, A. Juréus, S. Marques, H. Munguba, L. He, C. Betsholtz, Gillespie, S. B. R. Harley, P. F. Bartlett, W. A. Schroder, A. G. Yates,
C. Rolny, G. Castelo-Branco, J. Hjerling-Leffler, S. Linnarsson, Sci- D. C. Anthony, S. Rose-John, M. J. Ruitenberg, J. Vukovic, Cell 2020,
ence 2015, 347, 1138. 180, 833.
[119] A. Zeisel, H. Hochgerner, P. Lönnerberg, A. Johnsson, F. Memic, [142] P. Hasel, I. V. L. Rose, J. S. Sadick, R. D. Kim, S. A. Liddelow, Nat.
J. van der Zwan, M. Häring, E. Braun, L. E. Borm, G. La Manno, Neurosci. 2021, 24, 1475.
S. Codeluppi, A. Furlan, K. Lee, N. Skene, K. D. Harris, J. Hjerling- [143] G. M. Welch, C. A. Boix, E. Schmauch, J. Davila-Velderrain, M. B.
Leffler, E. Arenas, P. Ernfors, U. Marklund, S. Linnarsson, Cell 2018, Victor, V. Dileep, P. L. Bozzelli, Q. Su, J. D. Cheng, A. Lee, N. S. Leary,
174, 999. A. R. Pfenning, M. Kellis, L.-H. Tsai, Sci. Adv. 2022, 8, eabo4662.
[120] A. Saunders, E. Z. Macosko, A. Wysoker, M. Goldman, F. M. Krienen, [144] T. Kiss, Á. Nyúl-Tóth, J. DelFavero, P. Balasubramanian, S. Tarantini,
H. de Rivera, E. Bien, M. Baum, L. Bortolin, S. Wang, A. Goeva, J. J. Faakye, R. Gulej, C. Ahire, A. Ungvari, A. Yabluchanskiy, G. Wiley,
Nemesh, N. Kamitaki, S. Brumbaugh, D. Kulp, S. A. McCarroll, Cell L. Garman, Z. Ungvari, A. Csiszar, GeroScience 2022, 44, 661.
2018, 174, 1015. [145] J. F. Navarro, D. L. Croteau, A. Jurek, Z. Andrusivova, B. Yang, Y.
[121] R. D. Hodge, T. E. Bakken, J. A. Miller, K. A. Smith, E. R. Barkan, Wang, B. Ogedegbe, T. Riaz, M. Støen, C. Desler, L. J. Rasmussen,
L. T. Graybuck, J. L. Close, B. Long, N. Johansen, O. Penn, Z. Yao, T. Tønjum, M.-C. Galas, J. Lundeberg, V. A. Bohr, iScience 2020, 23,
J. Eggermont, T. Höllt, B. P. Levi, S. I. Shehata, B. Aevermann, A. 101556.
Beller, D. Bertagnolli, K. Brouner, T. Casper, C. Cobbs, R. Dalley, N. [146] W.-T. Chen, A. Lu, K. Craessaerts, B. Pavie, C. Sala Frigerio, N.
Dee, S.-L. Ding, R. G. Ellenbogen, O. Fong, E. Garren, J. Goldy, R. P. Corthout, X. Qian, J. Laláková, M. Kühnemund, I. Voytyuk, L. Wolfs,
Gwinn, D. Hirschstein, et al., Nature 2019, 573, 61. R. Mancuso, E. Salta, S. Balusu, A. Snellinx, S. Munck, A. Jurek, J.
[122] R. Yuste, M. Hawrylycz, N. Aalling, A. Aguilar-Valles, D. Arendt, R. Fernandez Navarro, T. C. Saido, I. Huitinga, J. Lundeberg, M. Fiers,
Armañanzas, G. A. Ascoli, C. Bielza, V. Bokharaie, T. B. Bergmann, B. De Strooper, Cell 2020, 182, 976.
I. Bystron, M. Capogna, Y. Chang, A. Clemens, C. P. J. de Kock, [147] J. M. Gregory, K. McDade, M. R. Livesey, I. Croy, S. Marion de
J. DeFelipe, S. E. Dos Santos, K. Dunville, D. Feldmeyer, R. Fiáth, Proce, T. Aitman, S. Chandran, C. Smith, Neuropathol. Appl. Neu-
G. J. Fishell, A. Foggetti, X. Gao, P. Ghaderi, N. A. Goriounova, O. robiol. 2020, 46, 441.

Adv. Sci. 2023, 10, 2206939 2206939 (19 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

[148] M. Kaufmann, H. Evans, A.-L. Schaupp, J. B. Engler, G. Kaur, A. Will- [164] A. Lomakin, J. Svedlund, C. Strell, M. Gataric, A. Shmatko, G.
ing, N. Kursawe, C. Schubert, K. E. Attfield, L. Fugger, M. A. Friese, Rukhovich, J. S. Park, Y. S. Ju, S. Dentro, V. Kleshchevnikov, V. Vask-
Med 2021, 2, 296. ivskyi, T. Li, O. A. Bayraktar, S. Pinder, A. L. Richardson, S. Santagata,
[149] M. Kaufmann, A.-L. Schaupp, R. Sun, F. Coscia, C. A. Dendrou, A. P. J. Campbell, H. Russnes, M. Gerstung, M. Nilsson, L. R. Yates,
Cortes, G. Kaur, H. G. Evans, A. Mollbrink, J. F. Navarro, J. K. Sonner, Nature 2022, 611, 594.
C. Mayer, G. C. DeLuca, J. Lundeberg, P. M. Matthews, K. E. Attfield, [165] B. Zheng, L. Fang, J. Exp. Clin. Cancer Res. 2022, 41, 179.
M. A. Friese, M. Mann, L. Fugger, Nat. Neurosci. 2022, 25, 944. [166] N. Yoosuf, J. F. Navarro, F. Salmén, P. L. Ståhl, C. O. Daub, Breast
[150] X. Wei, S. Fu, H. Li, Y. Liu, S. Wang, W. Feng, Y. Yang, X. Liu, Y.-Y. Cancer Res. 2020, 22, 6.
Zeng, M. Cheng, Y. Lai, X. Qiu, L. Wu, N. Zhang, Y. Jiang, J. Xu, X. [167] A. Levy-Jurgenson, X. Tekpli, V. N. Kristensen, Z. Yakhini, Sci. Rep.
Su, C. Peng, L. Han, W. P.-K. Lou, C. Liu, Y. Yuan, K. Ma, T. Yang, X. 2020, 10, 18802.
Pan, S. Gao, A. Chen, M. A. Esteban, H. Yang, J. Wang, et al., Science [168] E. Chelebian, C. Avenel, K. Kartasalo, M. Marklund, A. Tanoglidi, T.
2022, 377, eabp9444. Mirtti, R. Colling, A. Erickson, A. D. Lamb, J. Lundeberg, C. Wählby,
[151] T. Lohoff, S. Ghazanfar, A. Missarova, N. Koulena, N. Pierson, J. A. Cancers 2021, 13, 4837.
Griffiths, E. S. Bardot, C.-H. L. Eng, R. C. V. Tyser, R. Argelaguet, C. [169] S. Maniatis, J. Petrescu, H. Phatnani, Curr. Opin. Genet. Dev. 2021,
Guibentif, S. Srinivas, J. Briscoe, B. D. Simons, A.-K. Hadjantonakis, 66, 70.
B. Göttgens, W. Reik, J. Nichols, L. Cai, J. C. Marioni, Nat. Biotechnol. [170] B. He, L. Bergenstråhle, L. Stenbeck, A. Abid, A. Andersson, Å.
2022, 40, 74. Borg, J. Maaskola, J. Lundeberg, J. Zou, Nat. Biomed. Eng. 2020, 4,
[152] X. Zhang, X. Wang, G. V. Shivashankar, C. Uhler, Nat. Commun. 827.
2022, 13, 7480. [171] L. Zhang, D. Chen, D. Song, X. Liu, Y. Zhang, X. Xu, X. Wang, Signal
[153] P.-H. Li, X.-Y. Kong, Y.-Z. He, Y. Liu, X. Peng, Z.-H. Li, H. Xu, H. Luo, Transduction Targeted Ther. 2022, 7, 111.
J. Park, Mil. Med. Res. 2022, 9, 52. [172] A. Coulson, BioTechniques 2022, 73, 215.
[154] A. L. Ji, A. J. Rubin, K. Thrane, S. Jiang, D. L. Reynolds, R. M. Meyers, [173] C. A. Thornton, R. M. Mulqueen, K. A. Torkenczy, A. Nishida, E. G.
M. G. Guo, B. M. George, A. Mollbrink, J. Bergenstråhle, L. Lars- Lowenstein, A. J. Fields, F. J. Steemers, W. Zhang, H. L. McConnell,
son, Y. Bai, B. Zhu, A. Bhaduri, J. M. Meyers, X. Rovira-Clavé, S. T. R. L. Woltjer, A. Mishra, K. M. Wright, A. C. Adey, Nat. Commun.
Hollmig, S. Z. Aasi, G. P. Nolan, J. Lundeberg, P. A. Khavari, Cell 2021, 12, 1274.
2020, 182, 497. [174] Y. Deng, M. Bartosovic, S. Ma, D. Zhang, P. Kukanja, Y. Xiao, G. Su,
[155] S. Z. Wu, G. Al-Eryani, D. L. Roden, S. Junankar, K. Harvey, A. Ander- Y. Liu, X. Qin, G. B. Rosoklija, A. J. Dwork, J. J. Mann, M. L. Xu, S.
sson, A. Thennavan, C. Wang, J. R. Torpy, N. Bartonicek, T. Wang, Halene, J. E. Craft, K. W. Leong, M. Boldrini, G. Castelo-Branco, R.
L. Larsson, D. Kaczorowski, N. I. Weisenfeld, C. R. Uytingco, J. Fan, Nature 2022, 609, 375.
G. Chew, Z. W. Bent, C.-L. Chan, V. Gnanasambandapillai, C.-A. [175] E. Llorens-Bobadilla, M. Zamboni, M. Marklund, N. Bhalla, X. Chen,
Dutertre, L. Gluch, M. N. Hui, J. Beith, A. Parker, E. Robbins, D. J. Hartman, J. Frisén, P. L. Ståhl, Nat. Biotechnol. 2023, https://doi.
Segara, C. Cooper, C. Mak, B. Chan, S. Warrier, et al., Nat. Genet. org/10.1038/s41587-022-01603-9.
2021, 53, 1334. [176] M. Bartosovic, M. Kabbe, G. Castelo-Branco, Nat. Biotechnol. 2021,
[156] D. Cui Zhou, R. G. Jayasinghe, S. Chen, J. M. Herndon, M. D. Iglesia, 39, 825.
P. Navale, M. C. Wendl, W. Caravan, K. Sato, E. Storrs, C.-K. Mo, J. [177] Y. Deng, M. Bartosovic, P. Kukanja, D. Zhang, Y. Liu, G. Su, A. En-
Liu, A. N. Southard-Smith, Y. Wu, N. Naser Al Deen, J. M. Baer, R. ninful, Z. Bai, G. Castelo-Branco, R. Fan, Science 2022, 375, 681.
S. Fulton, M. A. Wyczalkowski, R. Liu, C. C. Fronick, L. A. Fulton, A. [178] T. Stuart, S. Hao, B. Zhang, L. Mekerishvili, D. A. Landau, S. Mani-
Shinkle, L. Thammavong, H. Zhu, H. Sun, L.-B. Wang, Y. Li, C. Zuo, atis, R. Satija, I. Raimondi, Nat. Biotechnol. 2022, https://doi.org/
J. F. McMichael, S. R. Davies, et al., Nat. Genet. 2022, 54, 1390. 10.1038/s41587-022-01588-5.
[157] R. Wu, W. Guo, X. Qiu, S. Wang, C. Sui, Q. Lian, J. Wu, Y. Shan, Z. [179] T. Lu, C. E. Ang, X. Zhuang, Cell 2022, 185, 4448.
Yang, S. Yang, T. Wu, K. Wang, Y. Zhu, S. Wang, C. Liu, Y. Zhang, B. [180] M. Stoeckius, C. Hafemeister, W. Stephenson, B. Houck-Loomis, P.
Zheng, Z. Li, Y. Zhang, S. Shen, Y. Zhao, W. Wang, J. Bao, J. Hu, X. K. Chattopadhyay, H. Swerdlow, R. Satija, P. Smibert, Nat. Methods
Wu, X. Jiang, H. Wang, J. Gu, L. Chen, Sci. Adv. 2021, 7, eabg3750. 2017, 14, 865.
[158] M. V. Hunter, R. Moncada, J. M. Weiss, I. Yanai, R. M. White, Nat. [181] V. M. Peterson, K. X. Zhang, N. Kumar, J. Wong, L. Li, D. C. Wilson,
Commun. 2021, 12, 6278. R. Moore, T. K. McClanahan, S. Sadekova, J. A. Klappenbach, Nat.
[159] D. Barkley, R. Moncada, M. Pour, D. A. Liberman, I. Dryg, G. Werba, Biotechnol. 2017, 35, 936.
W. Wang, M. Baron, A. Rao, B. Xia, G. S. França, A. Weil, D. F. Delair, [182] S. Vickovic, B. Lötstedt, J. Klughammer, S. Mages, Å. Segerstolpe,
C. Hajdu, A. W. Lund, I. Osman, I. Yanai, Nat. Genet. 2022, 54, 1192. O. Rozenblatt-Rosen, A. Regev, Nat. Commun. 2022, 13, 795.
[160] V. M. Ravi, P. Will, J. Kueckelhaus, N. Sun, K. Joseph, H. Salié, L. [183] N. Ben-Chetrit, X. Niu, A. D. Swett, J. Sotelo, M. S. Jiao, C. M. Stew-
Vollmer, U. Kuliesiute, J. von Ehr, J. K. Benotmane, N. Neidert, M. art, C. Potenski, P. Mielinis, P. Roelli, M. Stoeckius, D. A. Landau,
Follo, F. Scherer, J. M. Goeldner, S. P. Behringer, P. Franco, M. Khiat, Nat. Biotechnol. 2023, https://doi.org/10.1038/s41587-022-01536-
J. Zhang, U. G. Hofmann, C. Fung, F. L. Ricklefs, K. Lamszus, M. 3.
Boerries, M. Ku, J. Beck, R. Sankowski, M. Schwabenland, M. Prinz, [184] Y. Liu, M. DiStasio, G. Su, H. Asashima, A. Enninful, X. Qin, Y. Deng,
U. Schüller, S. Killmer, et al., Cancer Cell 2022, 40, 639. J. Nam, F. Gao, P. Bordignon, M. Cassano, M. Tomayko, M. Xu, S.
[161] A. Erickson, M. He, E. Berglund, M. Marklund, R. Mirzazadeh, Halene, J. E. Craft, D. Hafler, R. Fan, Nat. Biotechnol. 2023, https:
N. Schultz, L. Kvastad, A. Andersson, L. Bergenstråhle, J. Bergen- //doi.org/10.1038/s41587-023-01676-0.
stråhle, L. Larsson, L. A. Galicia, A. Shamikh, E. Basmaci, T. Díaz De [185] J. A. Doudna, E. Charpentier, Science 2014, 346, 1258096.
Ståhl, T. Rajakumar, D. Doultsinos, K. Thrane, A. L. Ji, P. A. Khavari, F. [186] T. Wang, J. J. Wei, D. M. Sabatini, E. S. Lander, Science 2014, 343, 80.
Tarish, A. Tanoglidi, J. Maaskola, R. Colling, T. Mirtti, F. C. Hamdy, [187] O. Shalem, N. E. Sanjana, E. Hartenian, X. Shi, D. A. Scott, T. S.
D. J. Woodcock, T. Helleday, I. G. Mills, A. D. Lamb, et al., Nature Mikkelsen, D. Heckl, B. L. Ebert, D. E. Root, J. G. Doench, F. Zhang,
2022, 608, 360. Science 2014, 343, 84.
[162] S. Friedrich, E. L. L. Sonnhammer, BMC Med. Genomics 2020, 13, [188] B. Adamson, T. M. Norman, M. Jost, M. Y. Cho, J. K. Nuñez, Y. Chen,
110. J. E. Villalta, L. A. Gilbert, M. A. Horlbeck, M. Y. Hein, R. A. Pak, A.
[163] E. Boileau, X. Li, I. S. Naarmann-de Vries, C. Becker, R. Casper, J. N. Gray, C. A. Gross, A. Dixit, O. Parnas, A. Regev, J. S. Weissman,
Altmüller, F. Leuschner, C. Dieterich, Front. Genet. 2022, 13, 912572. Cell 2016, 167, 1867.

Adv. Sci. 2023, 10, 2206939 2206939 (20 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

[189] A. Dixit, O. Parnas, B. Li, J. Chen, C. P. Fulco, L. Jerby-Arnon, N. D. [197] B. E. Öztürk, M. E. Johnson, M. Kleyman, S. Turunç, J. He, S. Ja-
Marjanovic, D. Dionne, T. Burks, R. Raychowdhury, B. Adamson, T. balameli, Z. Xi, M. Visel, V. L. Dufour, S. Iwabe, L. F. L. Pompeo
M. Norman, E. S. Lander, J. S. Weissman, N. Friedman, A. Regev, Marinho, G. D. Aguirre, J.-A. Sahel, D. V. Schaffer, A. R. Pfenning, J.
Cell 2016, 167, 1853. G. Flannery, W. A. Beltran, W. R. Stauffer, L. C. Byrne, Elife 2021, 10,
[190] P. Datlinger, A. F. Rendeiro, C. Schmidl, T. Krausgruber, P. Traxler, e64175.
J. Klughammer, L. C. Schuster, A. Kuchler, D. Alpar, C. Bock, Nat. [198] M. J. Jang, G. M. Coughlin, C. R. Jackson, X. Chen, M. R. Chuapoco,
Methods 2017, 14, 297. J. L. Vendemiatti, A. Z. Wang, V. Gradinaru, Nat. Biotechnol. 2023,
[191] M. Dhainaut, S. A. Rose, G. Akturk, A. Wroblewska, S. R. Nielsen, https://doi.org/10.1038/s41587-022-01648-w.
E. S. Park, M. Buckup, V. Roudko, L. Pia, R. Sweeney, J. Le Berichel, [199] S. Liu, J. B. Iorgulescu, S. Li, M. Borji, I. A. Barrera-Lopez, V. Shan-
C. M. Wilk, A. Bektesevic, B. H. Lee, N. Bhardwaj, A. H. Rahman, mugam, H. Lyu, J. W. Morriss, Z. N. Garcia, E. Murray, D. A. Rear-
A. Baccarini, S. Gnjatic, D. Pe’er, M. Merad, B. D. Brown, Cell 2022, don, C. H. Yoon, D. A. Braun, K. J. Livak, C. J. Wu, F. Chen, Immunity
185, 1223. 2022, 55, 1940.
[192] X. Hou, T. Zaks, R. Langer, Y. Dong, Nat. Rev. Mater. 2021, 6, [200] W. H. Hudson, L. J. Sudmeier, STAR Protoc. 2022, 3, 101391.
1078. [201] A. Janesick, R. Shelansky, A. D. Gottscho, F. Wagner, M. Rouault, G.
[193] P. Kazemian, S.-Y. Yu, S. B. Thomson, A. Birkenshaw, B. R. Leavitt, Beliakoff, M. F. de Oliveira, A. Kohlway, J. Abousoud, C. A. Morrison,
C. J. D. Ross, Mol. Pharmaceutics 2022, 19, 1669. T. Y. Drennon, S. H. Mohabbat, S. R. Williams, S. E. B. Taylor, bioRxiv:
[194] C. Dobrowolski, K. Paunovska, E. Schrader Echeverri, D. Loughrey, 2022.10.06.510405 2022.
A. J. Da Silva Sanchez, H. Ni, M. Z. C. Hatit, M. P. Lokugamage, Y. [202] S. He, R. Bhatt, C. Brown, E. A. Brown, D. L. Buhr, K. Chantranu-
Kuzminich, H. E. Peck, P. J. Santangelo, J. E. Dahlman, Nat. Nan- vatana, P. Danaher, D. Dunaway, R. G. Garrison, G. Geiss, M. T. Gre-
otechnol. 2022, 17, 871. gory, M. L. Hoang, R. Khafizov, E. E. Killingbeck, D. Kim, T. K. Kim,
[195] D. Brown, M. Altermatt, T. Dobreva, S. Chen, A. Wang, M. Thomson, Y. Kim, A. Klock, M. Korukonda, A. Kutchma, Z. R. Lewis, Y. Liang,
V. Gradinaru, Front. Immunol. 2021, 12, 730825. J. S. Nelson, G. T. Ong, E. P. Perillo, J. C. Phan, T. Phan-Everson, E.
[196] R. Colón-Thillet, K. R. Jerome, D. Stone, Virol. J. 2021, 18, 85. Piazza, T. Rane, Z. Reitz, et al., Nat. Biotechnol. 2022, 40, 1794.

Taeyun Ku is an assistant professor in Medical Science and Engineering at KAIST in South Korea. He
received his M.D. in Medicine from Yonsei University, and his Ph.D. in Medical Science and Engineer-
ing from KAIST. He co-invented various tissue clearing and expansion methods such as MAP, eMAP,
and ELAST during his postdoctoral training at MIT. Currently, he develops novel approaches in tissue
and material engineering to visualize complex biological architecture.

Jihwan Park obtained his Ph.D. in Life Science from Pohang University of Science and Technology in
South Korea. He previously studied the single-cell transcriptome of mouse kidneys (Park et al. Sci-
ence, 2018) during his postdoctoral training at U. Penn. He is currently an associate professor at the
School of Life Sciences, GIST, South Korea. His research interest is in single-cell biology and epige-
nomics of human diseases. His group explores the molecular mechanisms of disease development
using single-cell analysis, genomics technologies, computational analysis, as well as conventional
molecular biology experiments.

Adv. Sci. 2023, 10, 2206939 2206939 (21 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advancedscience.com

Chang Ho Sohn is an assistant professor at the Advanced Science Institute and Center for
Nanomedicine, Institute for Basic Science at Yonsei University, South Korea. He received his B.S.
from Seoul National University and his Ph.D. from Caltech, both in Chemistry. He conducted post-
doctoral training for the development of the early phase of multiplexed FISH methods (seqFISH) at
Prof. Long Cai’s lab at Caltech, and for tissue clearing technology, SHIELD at Prof. Kwanghun Chung’s
lab at MIT. His current research is focused on the development of next-generation single-cell and spa-
tial transcriptomics technologies using both imaging- and sequencing-based modalities for clinical
investigations.

Adv. Sci. 2023, 10, 2206939 2206939 (22 of 22) © 2023 The Authors. Advanced Science published by Wiley-VCH GmbH