Single Cell Technologies

Received: 5 August 2021 Revised: 9 December 2021 Accepted: 20 December 2021
DOI: 10.1002/ctm2.694
REVIEW
Single-cell RNA sequencing technologies and applications:

A brief overview
Dragomirka Jovic1,2 Xue Liang1,2,3 Hua Zeng4 Lin Lin5,6 Fengping Xu1,2
Yonglun Luo1,2,5,6
1 Lars Bolund Institute of Regenerative
Medicine, Qingdao-Europe Advanced

Graphical Abstract
Institute for Life Sciences, Qingdao,
China
2 BGI-Shenzhen, Shenzhen, China
3Department of Biology, University of
Copenhagen, Copenhagen, Denmark
4 Nanjing University of Chinese Medicine,
Nanjing, China
5Department of Biomedicine, Aarhus
University, Aarhus, Denmark
6Steno Diabetes Center Aarhus, Aarhus
University Hospital, Aarhus, Denmark
Correspondence
Yonglun Luo and Dragomirka Jovic, Lars
Bolund Institute of Regenerative Medicine,
Qingdao-Europe Advanced Institute for
Life Sciences, Qingdao, China.
Email: alun@biomed.au.dk; dragomirka-
jovic@gmail.com
1. This review provides a concise summary of the single-cell RNA sequencing

technologies.
2. Overview and guidelines for planning experimental procedures are presented.
3. Bioinformatics tools for scRNA-seq data analysis are thoroughly discussed.
4. Applications and further development of scRNA-seq technology are high-
lighted.
Clin. Transl. Med. 2022;12:e694. wileyonlinelibrary.com/journal/ctm2

https://doi.org/10.1002/ctm2.694
Received: 5 August 2021 Revised: 9 December 2021 Accepted: 20 December 2021
DOI: 10.1002/ctm2.694
REVIEW
Single-cell RNA sequencing technologies and applications:

A brief overview
Dragomirka Jovic1,2 Xue Liang1,2,3 Hua Zeng4 Lin Lin5,6 Fengping Xu1,2
Yonglun Luo1,2,5,6
1Lars Bolund Institute of Regenerative

Medicine, Qingdao-Europe Advanced Single-cell RNA sequencing (scRNA-seq) technology has become the state-of-
Institute for Life Sciences, Qingdao, the-art approach for unravelling the heterogeneity and complexity of RNA tran-
China
scripts within individual cells, as well as revealing the composition of different
2 BGI-Shenzhen, Shenzhen, China
3
cell types and functions within highly organized tissues/organs/organisms. Since
Department of Biology, University of
Copenhagen, Copenhagen, Denmark its first discovery in 2009, studies based on scRNA-seq provide massive infor-
4Nanjing University of Chinese Medicine, mation across different fields making exciting new discoveries in better under-
Nanjing, China standing the composition and interaction of cells within humans, model animals
5Department of Biomedicine, Aarhus and plants. In this review, we provide a concise overview about the scRNA-seq
University, Aarhus, Denmark
6 Steno Diabetes Center Aarhus, Aarhus
technology, experimental and computational procedures for transforming the
University Hospital, Aarhus, Denmark biological and molecular processes into computational and statistical data. We
also provide an explanation of the key technological steps in implementing the
Correspondence
technology. We highlight a few examples on how scRNA-seq can provide unique
Yonglun Luo and Dragomirka Jovic, Lars
Bolund Institute of Regenerative Medicine, information for better understanding health and diseases. One important appli-
Qingdao-Europe Advanced Institute for cation of the scRNA-seq technology is to build a better and high-resolution cat-
Life Sciences, Qingdao, China.
alogue of cells in all living organism, commonly known as atlas, which is key
Email: alun@biomed.au.dk; dragomirka-
jovic@gmail.com resource to better understand and provide a solution in treating diseases. While
great promises have been demonstrated with the technology in all areas, we fur-
Funding information
Qingdao-Europe Advanced Institute for
ther highlight a few remaining challenges to be overcome and its great potentials
Life Sciences; Independent Research Fund in transforming current protocols in disease diagnosis and treatment.
Denmark, Grant/Award Number: 8048-
00072A KEYWORDS
atlas, big data, precision medicine, regenerative medicine, RNA sequencing, single cell
1 THE RISE OF SINGLE-CELL RNA were observed for the very first time in the 16th century,
SEQUENCING TECHNOLOGY and since then many signs of progress and advancement
of new technologies and methods evolved from elemen-
Humans are highly organized systems composed of tary to profound. Although the first microscope invented
approximately 3.72 × 1013 cells of various types form- by Zacharias Janssen and Hans Lippershey in the late 16th
ing harmonious microenvironments to keep proper organ century enabled Robert Hooke and Anton van Leeuwen-
functions and normal cellular homeostasis.1 Living cells hoek to spot the first living cell in the 17th century, it took
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the
original work is properly cited.
© 2022 The Authors. Clinical and Translational Medicine published by John Wiley & Sons Australia, Ltd on behalf of Shanghai Institute of Clinical Bioinformatics
Clin. Transl. Med. 2022;12:e694. wileyonlinelibrary.com/journal/ctm2 1 of 20

https://doi.org/10.1002/ctm2.694
2 of 20 JOVIC et al.
almost two centuries to re-define cells not only as of the the cost has been tremendously reduced. As such, scien-
structural but also functional unit of life.2 Since then, vari- tific publications using the scRNA-seq technology method
ous experiments and methods were conducted for the pur- each increases yearly due to fast and accurate scRNA-
pose of better understanding and investigating cells in het- seq technologies such as microfluidic- microwell-, droplet-
erogeneous multicellular systems.3,4 Despite that tremen- based, in situ barcoding and spatial transcriptome analy-
dous and revolutionary discoveries in cell biology field sis as summarized and illustrated in Figure 1.6–10 In this
have been made, the heterogeneity of cells remains to be section, we focus on highlighting a few key technological
further revisited. Almost all cells in the human body have and experimental steps in high throughput single-cell RNA
the same set of genetic materials, but their transcriptome sequencing technologies.
information in each cell reflects the unique activity of only The procedures of scRNA-seq mainly include single-
a subset of genes. Profiling the gene expression activity in cell isolation and capture, cell lysis, reverse transcription
cells is considered as one of the most authentic approaches (conversion of their RNA into cDNA), cDNA amplifica-
to probe cell identity, state, function and response. Huge tion and library preparation (Figure 2).11,12 Single-cell cap-
technological breakthroughs have been made in the single- ture, reverse transcription and cDNA amplification are
cell transcriptomics during the last decade. With single- the most challenging parts among the library preparation
cell RNA sequencing, it is now possible to analyse the tran- steps. With the development of many sequencing plat-
scriptome at single-cell level for over millions of cells in a forms, RNA-seq library preparation technologies have also
single study. This allows us to classify, characterize and dis- presented a rapid and diversified development. Thus, it is
tinguish each cell at the transcriptome level, which leads to important to know the features and applications of differ-
identify rare cell population but functionally important. ent single-cell RNA sequencing library preparation meth-
The first conceptional and technical breakthrough of the ods, in order to make appropriate choices in scientific
single cell RNA sequencing method was made by Tang research and better apply these techniques to clinical appli-
et al. in 2009, which sequenced the transcriptome of sin- cations.
gle blastomere and oocytes.5 The concept and technology Single-cell isolation and capture is the process of cap-
brought by this study open a new avenue to scale up the turing high-quality individual cells from a tissue, thereby
number of the cells and make compatible high-throughput extracting precise genetic and biochemical information
RNA sequencing possible for the first time. Since then, and facilitating the study of unique genetic and molecular
an increasing number of modified and improved single- mechanisms.13 Traditional transcriptome, epigenome or
cell RNA sequencing technologies were developed to intro- proteome from bulk RNA/DNA samples can only capture
duce essential modifications and improvements in sample the total level of signals from tissues/organs, which fail to
collection, single-cell capture, barcoded reverse transcrip- distinguish individual cell variations. The single-cell isola-
tion, cDNA amplification, library preparation, sequenc- tion and capture methods are largely different depending
ing and streamlined bioinformatics analysis. Most impor- on the organisms, tissues or cell properties.14 Cell isolation
tantly, cost has been dramatically reduced, while automa- can be accomplished by isolating whole cells, cell-specific
tion and throughput have been significantly increased. nuclei or cell-specific organelles, and even by separating
All these steps branch into more matured scRNA-seq the desired cells expressing specific marker proteins.15 The
methods, but the concept of the scRNA-seq remains the most common techniques of single-cell isolation and cap-
same. This review provides a comprehensive and concise ture include limiting dilution, fluorescence-activated cell
overview of the single cell technology development from sorting (FACS), magnetic-activated cell sorting, microflu-
its early stage and library constructions and its challenges idic system and laser microdissection. The key outcome
and data acquisition that transform our understandings of single capture, and particularly in high throughput, is
of RNA transcriptions into data output. We also discuss that each single cell is captured in an isolated reaction
applications of scRNA-seq, the potential of the scRNA-seq mixture, of which all transcripts from one single cell will
in spatial transcriptomics, cell atlases and future perspec- be uniquely barcoded after converted into complementary
tives. DNAs (cDNA).
However, the scRNA-seq has gradually revealed some
inherent methodological issues, such as ‘artificial tran-
2 HIGH THROUGHPUT SINGLE-CELL scriptional stress responses’. It means that the dissocia-
RNA SEQUENCING TECHNOLOGY: tion process could induce the expression of stress genes,
EXPERIMENTAL PROCEDURES which lead to artificially changes in cell transcription pat-
terns. This has been confirmed by a number of experi-
The throughput of scRNA-seq increases from a few cells ments. Brink et al. found that the process of protease dis-
per experiment to hundreds of thousands of cells, where sociation at 37°C could induce the expression of stress
JOVIC et al. 3 of 20
F I G U R E 1 Development of single-cell RNA sequencing technology. With the technological advances in single-cell RNA sequencing
(scRNA)-seq, (A) the number of analyzed cells increased, (B) the cost (in US dollar) was exponentially reduced, (C) the number of published
papers increased and (D) the history of technology evolution in the last decade using more sophisticated, accurate, high throughput analysis
was achieved. Part (D) is created with icons from BioRender with license for publication
genes, introduce technical error and cause inaccurate cell stress responses as compared to scRNA-seq.18 SnRNA-seq
type identification.16 Adam et al. also found that disso- becomes very useful in many tissue types, such as muscle
ciation at 37°C can cause ‘artificial changes’ of cell tran- tissue,19 heart,20 kidney,20 lung,21 pancreas22 and various
scriptome, resulting in inaccurate results.17 Dissociation of tumour tissues.23 It is particularly applicable in brain tis-
tissues into single-cell suspension at 4°C has thus been sues, which are difficult to be dissociated to obtain intact
suggested to minimize the isolation procedure-induced cells. Grindberg et al. demonstrated that single-cell tran-
gene expression changes. Single-nucleus RNA sequenc- scriptomic analysis can be done using the extremely low
ing (snRNA-seq) is an alternative single-cell sequencing levels of mRNA in a single nucleus of brain tissue.24 There-
method. Instead of sequencing all the mRNA in the cyto- fore, several compelling potential benefits of the method
plasm of cells, scRNA-seq only captures the mRNAs in emerge: firstly, compared with intact cells, the nucleus has
the nucleus of cells. The snRNA-seq solves the problems the advantage of being easily separated from complex tis-
related to tissue preservation and cell isolation that are not sues and organs, such as those in the central nervous sys-
easily separated into single-cell suspensions, applicable for tem. Secondly, snRNA-seq can be widely used for eukary-
frozen samples, and minimizes artificial transcriptional otic species, including species from different kingdoms.
F I G U R E 2 An overview of the single-cell RNA-sequencing procedures. (A) Isolation of the cells from tissue samples and capturing of
the single cells, wrapping of each individual cell with a bead inside a nanoscale droplet (each bead contains unique molecular identifiers), (B)
barcoding and amplification of complementary DNA (cDNA) and (C) library preparation procedure. After single-cell RNA sequencing (D),
the snapshot data would be analyzed to present and classify the landscape of gene expression in cells of a heterogeneous population (E).
Illustrative figure in (E) is generated with BioRender with license for publication
This method can also provide insights into the regulatory RNA, which results in additional 3′ coverage biases.25 Both
mechanisms of nuclear specificity.18 However, it should approaches can lead to amplification biases. To overcome
be noted that snRNA-seq only captures transcripts in the amplification-associated biases, unique molecular identi-
nucleus, which might fail to capture important biological fiers (UMIs) were introduced to barcode each individual
processes related to, that is, mRNA processing, RNA sta- mRNA molecule within a cell in the reverse transcription
bility and metabolism. Despite all these potential techni- step, thus improving the quantitative nature of scRNA-
cal limitations, both technologies have been demonstrated seq37 and enhancing the reading accuracy by effectively
by a large number of scientific publications in, for exam- eliminating PCR amplification bias. UMIs are adapted
ple, better understanding the cellular and biological pro- by the CEL-seq,35 MARS-seq,36 Drop-seq,29 inDrop-seq,9
cesses in organogenesis, gaining novel biomedical and cel- 10x Genomics,30 MATQ-seq31 Seq-Well32 and DNBelab C4
lular insights into disease pathogenesis. protocols.38
After the process of converting RNA into the first- Once the single cell-barcoded cDNAs are generated from
strand cDNA, the resulting cDNA are amplified by either single cells or single nucleus, the cDNA can be sequenced
polymerase chain reaction (PCR) or in vitro transcrip- using a number of deep sequencing platforms. In terms of
tion (IVT).25 PCR as a non-linear amplification process high throughput sequencing based on the DNA nanoballs
is applied in Smart-seq,26 Smart-seq2,27 Fluidigm C1,28 (DNBseq), the selected DNA fragments was repaired to get
Drop-seq,29 10x Genomics,30 MATQ-seq,31 Seq-Well 32 and a blunt end and modified at the three ends to obtain a dATP
DNBelab C4. Currently, there exist two PCR amplifi- overhang, then each end of the DNA fragment was ligated
cation strategies. One uses SMART technology, which by the dTTP tailed adapter sequence. The ligation prod-
takes advantage of transferase and strand-switch activ- uct was then amplified for a few cycles, and the following
ity of Moloney Murine Leukemia Virus reverse transcrip- single-strand cycle was carried out. One special strand of
tase to incorporate template-switching oligos as adap- the PCR product was reverse-complemented with a special
tors for downstream PCR amplification.33 This method molecule and was ligated with the single-strand molecule
was the mostly used cDNA amplification method. The by DNA ligase, finally obtaining a single-strand circular
other strategy connects the 5′ end of cDNA with either DNA library.39 Different scRNA-seq technology and exper-
poly(A) or poly(C) to build common adaptors in PCR imental protocols are summarized in Table 1.
reaction.34 IVT is another amplifying approach and a lin- In conclusion, we would like to highlight questions that
ear amplification process, which is used in CEL-seq,35 should be emphasized in the scRNA-seq library prepara-
MARS-Seq,36 and inDrop-seq9 protocols. It requires an tion: (1) how to capture the interesting RNA types from
additional round of reverse transcription of the amplified total RNA, also called RNA enrichment; (2) how to reverse
JOVIC et al.
TA B L E 1 Comparison of different single-cell RNA sequencing (scRNA-seq) technology and experimental protocols
Isolation Amplification Published

Platforms strategies Tissue Cell numbers Targets UMI methods Region year Reference
26
Smart-seq FACS Dissociated cell Hundreds of / × PCR Full-length 2012
cells
27
Smart-seq2 FACS Dissociated cell Hundreds of / × PCR Full-length 2013
cells
28
Fluidigm C1 Micro-fluidic Dissociated cell Hundreds of No poly(A) minus × PCR Full-length 2013
cells RNA detection
29
Drop-seq Microdroplets Dissociated cell Large number of No poly(A) minus √ PCR 3′ end 2015
cell RNA detection
30
10x Genomics Microdroplets Dissociated cell Large number of No poly(A) minus √ PCR 3′ end 2016
cells RNA detection
31
MATQ-seq FACS Dissociated cell Hundreds of No poly(A) minus √ PCR Full-length 2017
cells RNA detection
32
Seq-Well Micro-fluidic Dissociated cell Large number of No poly(A) minus √ PCR 3′ end 2017
cells RNA detection
35
CEL-seq FACS Dissociated cell Hundreds of No poly(A) minus √ IVT 3′ end 2012
cells RNA detection
36
MARS-seq FACS Dissociated cell Hundreds of No poly(A) minus √ IVT 3′ end 2014
cells RNA detection
9
inDrop-seq Microdroplets Dissociated cell Large number of No poly(A) minus √ IVT 3′ end 2015
cell RNA detection
38
DNBelab C4 Microdroplets Dissociated cell Large number of No poly(A) minus √ PCR 3′ end 2019
cells RNA detection
Abbreviations: FACS, fluorescence-activated cell sorting; IVT, in vitro transcription; PCR, polymerase chain reaction; UMI, unique molecular identifier; MATQ-seq, multiple annealing and dC-tailing-based quantitative
single-cell RNA-seq; MARS-seq, massively parallel single-cell RNA-sequencing; DNB-seq, DNA Nanoball Sequencing .
5 of 20
transcribe RNA into cDNA fragment with appropriate 3.1 Data preprocessing
size; (3) how to connect the adaptor to the end of cDNA.
In addition, there are still some remaining challenges to Basic formats of raw sequencing data for single-cell tran-
overcome in preparation of scRNA-seq library. For exam- scriptome include FASTQ and BCL format, which depend
ple, high variability between cells often occurs in scRNA- on the data source and sequencing platform. Since only
seq data, which is caused by technical variations in RNA FASTQ files can be directly implemented for quality con-
capture and stochastic transcription in cells.34 Further- trol, once the raw data are not in FASTQ format, the first
more, due to high sequencing cost, previous scRNA-seq step is to convert it to FASTQ format with the appropri-
methods only concentrated on the 5′ or 3′ ends of the ate tools. FASTQ files can be generated from the BCL
transcriptome. In fact, the region sample for single cell files using cellranger mkfastq, a pipeline that has wrapped
RNA-seq should depend on the experimental purpose. bcl2fastq software. Importantly, a simple CSV matrix file
For example, even though 3′ end sequencing is cheaper including at least three columns (lane, sample and index)
than full-length sequencing and could provide the best should be provided in addition to the path of BCL files.
coding region data of 3′ with the addition of a non- Then FastQC can be applied to assess the quality of raw
templated poly(A) tail, it cannot sequence the entire tail single-cell RNA sequencing data.
and cannot specifically report the mRNA isoform to which High-quality reads need to be mapped to the specific ref-
tails are attached.40 In addition, the quality of scRNA- erence genome using an appropriate aligner (e.g., STAR
seq library preparation is affected by other factors, such or Tophat). The count is the most important function of
as technical noise and biological noise. The technical Cell Ranger, which has wrapped up the alignment, filter-
factors include the RNA capture efficiency and quality, ing, UMI counting and other practical steps internally.
random dropouts during library preparation, single-cell The Cell Ranger uses an aligner called STAR, which per-
amplification technology and experimental batch effects.41 forms splicing-aware alignment of reads to the genome and
As for biological noise, the nature and different genetic then uses transcriptional annotation general transfer for-
backgrounds of biological specimens (like cell sizes, gene mat (GTF) file to categorize these reads into exons, introns
expression) and the dynamic and random environment and intergenic based on whether the reads are aligned to
changes (various cell states, cell cycle states) are difficult the genome confidently.
to be controlled by experiment operation. Therefore, a
still crucial challenge in the scRNA-seq library prepara-
tion is to minimize RNA loss and maximize information 3.2 General analyses
Precision.
During the preparation of single-cell suspension, a viable
cell may experience death, cell membrane damage or
3 STREAMLINE scRNA-SEQ DATA multicellular adhesion due to unavoidable natural phe-
ANALYSIS nomena, experimental operations and technical barriers.
To eliminate the gene expression interference from low-
Analysis of scRNA-seq data is another key factor, and now quality cells, it is necessary to conduct a second round
the major need, to broaden the application of this technol- of quality control with suitable tools, such as Seurat,43
ogy in life and clinical sciences. To ensure the availability of scran44 and scanpy.45 In terms of citations, Seurat is the
single-cell transcriptome analysis tools, numerous devel- most popular one with built-in functions to handle low-
opers have made considerable efforts. Nearly 1000 differ- quality cell filtration. Basically, the following quality con-
ent bioinformatic tools have been developed and made trol (QC) indicators should be used to judge whether a cell
available by May 28th, 2021 (see Table S1).42 The increas- should be retained: the numbers of genes, the numbers of
ing number of tools for single-cell transcriptome analysis UMI (transcripts), the percentages of mitochondrial genes
has illustrated the importance of analytical methods in the and the percentages of ribosomal protein genes in each
field, but this also means more perplexity in choosing tools cell. There is no absolute standard for the setting of filter
for single-cell data analysis. In this section, we review basic thresholds, which usually depends on the type of cell and
single-cell transcriptome analysis processes according to tissue being analysed. Lambrechts et al. filtered out cells
key steps (Figure 3), and analysis module (Figure 4), which with ≤ 100 or ≥ 6000 expressed genes, ≤ 200 UMIs and
also covers exploratory analyses of the gene level and cellu- ≥ 10% mitochondrial genes as described in their study.46
lar level. For advance single-cell transcriptome data anal- Fan et al. retained good quality cells using the following
ysis, we refer readers to the specific tools summarized in parameters: (1) 200 < total number of expressed genes per
Table S1 and the original articles reporting the tools. cell (nGenes) < 2500; (2) 300 < total number of UMIs
F I G U R E 3 Roadmap for typical single-cell RNA sequencing data analysis. The classic roadmap for single-cell RNA sequencing
(scRNA-seq) data analysis mainly consists of data preprocessing (blue panel), general analyses (green panel) and exploratory analyses (yellow
panel). Data preprocessing includes quality control, alignment and quantification; general analyses include low-quality cell filtering,
normalization, HVG selection, dimension reduction, clustering and annotation of cell types; exploratory analyses include DEG analysis,
function enrichment, GSVA, TF prediction, cell trajectory, cell-cell interaction, cell cycle and spatial transcriptome analysis. The plot below
each box gives a schematic of the visualized results in each analysis step. HVG, highly variable gene; DEG, differentially expressed gene;
GSVA, gene set variation analysis; TF, transcription factor. Demo figures were generated with data set GSM4041174
F I G U R E 4 Overview of the analysis modules for single-cell RNA sequencing data analysis. The diagram shows a summary of analysis
modules in the actual analysis of single-cell RNA sequencing (scRNA-seq) data, which can be divided into four analysis modules; they are (A)
data preprocessing module, (B) general analysis module, (C) exploratory analysis module, and (D) optional analysis module, respectively.
More details about each module can be found in the “Streamline scRNA-seq Data Analysis” section
per cell (nUMIs) < 15000; and (3) percentage of UMIs stream analysis because the expression levels between
mapped to mitochondrial genes (MT%) < 10%.47 Adjust- cells are not comparable due to systemic errors or tech-
ing the above QC threshold flexibly according to the spe- nical noises, such as differences in sequencing depth
cific disease state and the diversity of tissue types is recom- and transcriptome capture rate for each cell. Normal-
mended. It should be noted the filtration of cells based on ization is intended to counteract technical noise or bias
mitochondrial genes should be carefully applied as some and to ensure comparability between each cell. In 2020,
cell types, such as cardiomyocytes, are biologically more Lytal et al. evaluated the effectiveness of seven normal-
abundant in expressing these genes. ization methods, including BASiCS, GRM, Linnorm, SAM-
Similar to the analysis of traditional bulk RNA-Seq strt, SCnorm, scran and Simple Norm.48 It is worth noting
data, each cell is treated as an independent sample when that the speed advantage of Linnorm and scran comes from
analysing the single-cell RNA sequencing data. The orig- being written in C++ and implemented in R, which is suit-
inal expression matrix cannot be directly used for down- able for large data sets. In contrast, BASiCS and SCnorm
take a longer time to generate more refined results. Over- scale data sets. And the time complexity and spatial com-
all, there exist large differences among these methods, and plexity of deepMNN algorithm are almost excellent. It
different tools perform optimally in different situations. took 17 min to complete the batch-effect correction for
The single-cell RNA sequencing data set is high- the large data set, while Harmony and Scanorama took
dimensional, with tens of thousands of cells in a sample about 35 and 77 min, respectively. In addition, it has a
and thousands of genes expressed in each cell. Most genes larger storage space than Seurat V4 and scGen. At the same
in each cell belong to housekeeping ones, as they are char- time, deepMNN can integrate multi batch data sets in one
acterized by no significant changes in the expression level step without multiple iterations. These characteristics of
between cells, and their presence tends to obscure the real deepMNN make it possible to be a new choice for large-
biological signals. The subsets of features that exhibit high scale single-cell gene expression data analysis.
cell-to-cell variation in the data set are also called highly In addition to feature selection, dimensionality reduc-
variable genes (HVGs). HVGs not only highlight biological tion is also one of the main strategies for processing
signals but also greatly accelerate the downstream analy- such high-dimensional data. For single-cell RNA sequenc-
sis of single-cell RNA sequencing data due to the signifi- ing data, two rounds of dimension reduction are gener-
cant reduction in the computation volume. A high-quality ally required, with principal component analysis (PCA)
HVGs should include genes that can distinguish different dimension reduction first, and then t-distributed stochas-
cell types, and the quality of HVGs has a significant effect tic neighbor embedding (t-SNE) or Uniform Manifold
on the precision of clustering. In 2018, Yip et al.49 evalu- Approximation and Projection (UMAP) dimension reduc-
ated seven methods for detecting HVGs, including BASiCS, tion for visualization. PCA is a mathematical linear dimen-
Brennecke, scLVM, scran, scVEGs, and Seurat, observed sion algorithm, which uses an orthogonal transforma-
large differences in the clustering results as well as in the tion to transform a series of potentially linearly related
run times of the different methods. Compared with other variables into new ones that are linearly unrelated, thus
methods, scran can detect a stable number of HVGs with using the new variables to show the characteristics of the
excellent running time and independence from the mean. data at a lower dimension. PCA has been widely used
Brennecke was proved to have stable and consistent perfor- in sRNA-seq studies to overcome the extensive technical
mances with a wide range of data sets. scran and Seurat noise in any single feature. Wu et al. conducted a system-
were shown to perform optimally with part of data sets. atic comparison of these two non-linear dimension reduc-
BASiCS and scLVM_LogVar were shown to be much slower tion methods in 2019. They pointed out the use of UMAP in
than others. high-dimensional cytology and single-cell RNA sequenc-
Different scRNA-seq data may arise from different ing, with particular emphasis on the faster runtimes and
times and different sequencing platforms, and there are consistency of UMAP compared to t-SNE and the more
inevitable technological or non-biologically significant meaningful organization of cell clusters and preservation
batch effects between these data.50,51 The batch effect in of the continuum.58 In addition, UMAP has a clear advan-
the scRNA-seq data has plagued downstream analysis tage over t-SNE in the continuity of the cell subsets because
because it can disrupt gene expression patterns and then it preserves more of the global structure, although t-SNE is
lead to erroneous conclusions. As a consequence, batch- still applied in many single-cell studies, seemingly due to
effect correction is critical for the analysis of scRNA-seq better visual preferences.
data. Although a number of batch effect correction algo- The complexity of single-cell RNA sequencing data pro-
rithms have been proposed for scRNA-seq data,50 such as motes the development of a wide range of clustering meth-
Scanorama52 and Seurat V4,53 which can only merge two ods. Based on the ability to recover known subpopula-
data sets at a time and integrate multiple data sets by iterations, the stability and the run time and scalability, a recent
tion. Most of them consume a large amount of computing paper59 evaluated 14 clustering methods on a total of 12 dif-
memory and time, and this demand is likely to increase ferent data sets. Notably, SC3 and Seurat performed bet-
as the number of scRNA-seq data increases. Recently, Zou ter among these methods in a comprehensive view, with
et al. proposed a novel deep learning-based method, called Seurat being several orders of magnitude faster. When the
deepMNN, in order to correct batch effect in scRNA-seq number of clusters was the same, Seurat typically achieved
data.54 It compared the performance of deepMNN with the best consistency with the real partition, while Flow-
the most advanced batch correction methods, including SOM achieved better consistency with the real partition if
the widely used Harmony,55 Scanorama52 and Seurat V453 the number of clusters is higher than the real number.
methods, as well as the recently developed MMD-ResNet56 After clustering, assigning a biological annotation to
and scGen57 methods based on deep learning. The results54 each cluster is the basis of the subsequent analysis. Gen-
show that the accuracy of deepMNN is better than the erally, the workflow for annotating cells in scRNA-seq
existing common methods, especially in the case of large- data includes three main steps60 : automatic annotation,
manual annotation and validation with wet experiments. the KEGG pathway to make the results more biologically
Firstly, major automated annotation tools utilize a pre- explanatory.64
defined set of marker genes that are specifically expressed To identify the transcription factors enriched in each
in a known cell type to label clusters by matching their cell cluster from scRNA-seq data, Aibar et al.65 devel-
gene expression patterns to known cell types. The advan- oped SCENIC in 2017, which enabled inferring transcrip-
tage of the automated cell annotation method is that it tion factors because it firstly achieves the enrichment of
is fast and reproducible, and the results tend to be more TF motifs by searching the putative regulatory regions of
reliable in annotating common cell types. However, it is target genes. Then TF motif enrichment can realize the
unable to define rare and new cell types due to the limi- connection of candidate TF regulatory factors with candi-
tations of the reference marker gene set. In 2020, Huang date target genes. Although SCENIC can be implemented
et al.61 compared and assessed 10 cell-type annotation in both R and Python, pySCENIC is highly recommended
methods systematically, including Seurat, scmap, SingleR, for running big data sets due to its faster implementa-
CHETAH, SingleCellNet, scID, Garnett, SCINA, CP and tion of the SCENIC pipeline. Note that the latest version
RPC. They found that Seurat being the best method for of SCENIC supports Homo sapiens, Mus musculus and
annotating the major cell types among the top five meth- Drosophila melanogaster, with the possibility of manually
ods: Seurat, SingleR, CP, RPC and SingleCellNet. However, creating a custom database for other species.65,66 Although
Seurat performed relatively worse in predicting rare cell SCENIC was broadly used because of its outstanding scal-
types and distinguishing highly similar cell types. Sec- ability and robustness for a wide variety of databases, it
ondly, manual annotation is the gold standard method for ignored the dynamic changes in gene regulation mech-
annotating cells, although it is both subjective and labor- anisms in different cell types. In 2020, Ma et al. devel-
intensive by searching the relevant literature and mining oped IRIS3,67 an integrated cell-type-specific regulon infer-
existing scRNA-seq data. Finally, wet-lab experiments are ence server from single-cell RNA-Seq. In practical applica-
typically required to further validate the finding by scRNA- tions, IRIS3 was more suitable for the researchers without
seq. Traditional validation methods include immunoflu- substantial programming skills with its user-friendly web
orescence and immunohistochemistry, both of which are server. However, continuous improvement is required by
based on the principle of specific binding of antibodies to IRIS3 in accuracy and efficiency.
antigens (the surface proteins encoded by marker genes) Pseudo-time analysis can be used to infer the trajectory
to prove the true existence of the cell types obtained from of cells at the single-cell level, which is expected to discover
the data analysis. Besides, emerging spatial transcriptome rare cell types and cryptic states. Different types of analy-
sequencing technology can also be considered for increas- sis tools have been developed in the service of pseudo-time
ing the reliability of annotation, which can combine cell analysis. In 2019, Saelens et al. conducted a comprehen-
imaging and scRNA-seq to measure spatial transcript pat- sive comparison of 45 pseudo-time analysis tools and found
terns and cell morphology in one experiment.62 great complementarity of existing tools.68 Monocle is one
of the most broadly used tool for pseudo-time analysis,69
which learns an explicit principal graph to describe the
3.3 Exploratory analyses data and rebuilds single-cell trajectories by embedding
reversed graph to improve the robustness and accuracy of
To robustly reveal functional bias and biological signifi- predicted trajectories. Emphatically, the entire process of
cance of specific cell populations, it is necessary to per- establishing single-cell gene expression kinetics is largely
form functional enrichment analyses on a targeted dif- data-driven.
ferentially expressed gene set. Universal analysis strate- Organisms will self-regulate to maintain homeostasis
gies for function enrichment are also suitable for single- when stimulated, which must require the co-participation
cell data, such as gene ontology and Kyoto Encyclopedia and coordination of multiple cell types. With the rapid
of Genes and Genomes (KEGG) pathway. A large num- development of cell-cell communication research, the
ber of mature tools for functional enrichment analysis tools available to analyze cell-cell communication are no
have been developed. Huang et al. comprehensively com- longer limited, including CellChat, CellPhoneDB, Nich-
pared 68 enrichment analysis tools in 2009 after weighing eNet, SingleCellSignalR and iTalk,70–74 etc. Although each
respective advantages and disadvantages.63 In addition, of these tools works on the strength of the interaction of lig-
GSVA is also widely used in functional enrichment analy- ands and receptors on the cell surface, each has its advan-
sis and other standard analyses in a pathway-centered way. tages and weakness. Specifically, if the structural compo-
GSVA can calculate enrichment scores for different signal- sition of the ligand and receptor is expected to be con-
ing pathways in each sample to assess the causes of phe- sidered, CellPhoneDB should be preferred. If the regula-
notypic differences, which can be used as a supplement to tion of cofactors (such as promoters and antagonists) is
expected to be taken into account, CellChat can be selected including humans, animals and plants enabling more
to improve performance. It is also recommended to com- accurate, rapid identification of rare and novel cells in tis-
bine multiple cell-cell communication analysis tools flexi- sues like never before (Figure 5). Moreover, with the infor-
bly to avoid methodical bias. mation about gene expression at mRNA and protein lev-
Each cell in the single-cell suspension is at a specific els, metabolites, cell-cell communication and spatial land-
stage in the cell cycle: DNA synthesis prophase (G1 phase), scape, it becomes possible to solve the puzzle of cell com-
DNA synthesis phase (S phase), DNA synthesis anaphase position and functions in health and disease. Although
(G2 phase) or mitotic phase (M phase). There is a mix- the first findings and use of scRNA-seq were mostly done
ture of cells resining different cell cycles from each popula- on animal and later human cells, the sequencing in plant
tion. The CellCycleScoring function in the Seurat assigns science is still in its early stage and has many exciting
a quantitative score to each cell according to the expres- challenges remain to be overcome.84 To date, the appli-
sion of G2/M and S phase marker genes embedded in its cation of scRNA-seq remains limited to only few plants,
built-in package. In recent years, machine learning-based due to technical challenges or very limited information
methods have been developed to predict cell cycle stages on the cell types and discoveries in developmental biol-
from single-cell RNA sequencing data. In 2015, Scialdone ogy. Several plant research groups used the most used
et al. compared five established supervised machine learn- model plant in molecular genetics, Arabidopsis thaliana
ing methods as well as a custom-built predictor for assign- root for high throughput scRNA-seq and spatial transcrip-
ing cells to their cell cycle stages based on the tran- tomics analysis due to the relatively small number of
scriptome. Specifically, they indicate that only PCA-based cells, known gene markers and easy methods to isolate
methods and customized predictors perform best, which individual cells via enzymatic cell wall degradation.85–87
can robustly capture cell cycle signals.75 After successful proof of concept with A. thaliana root,
the studies have been increasingly applied to the study of
other parts of Arabidopsis and other plant species such
3.4 Optional analyses as rice leaf and root, tomato and maize.88–91 Moreover,
following the establishment of human cell atlases, the
Although we explained the major steps of the single-cell plant-based scientific community in 2019 initiated a plant
sequencing analysis process, there are still many other cell atlas consortium, aiming to collect more informa-
significant aspects that deserve more attention and explo- tion about various plant cell types, their nucleic acids,
ration, such as the combined application of scRNA-seq proteins and metabolites.92 Various web-based graphi-
and CRISPR screening,76 and the integrated analysis of cal information about plant scRNA-seq data is avail-
scRNAseq and multi-omics, including scATAC-sEquation able on (https://www.zmbp-resources.uni-tuebingen.de/
(single-cell chromatin accessibility and transcriptome timmermans/plant-single-cell-browser).93 Yet, the rapidly
sequencing),77 scMT-sEquation (single-cell methylome growing field of single-cell biology of the plants has a lot
and transcriptome sequencing),78,79 CITE-sEquation more to offer, including integration of sequencing scRNA-
(cellular indexing of transcriptomes and epitopes by seq, snRNA-seq and spatial transcriptomics, imaging tech-
sequencing)80,81 and spatial transcriptome.82 The combi- niques and omics will help to further understand changes
nation of these techniques enables a better and deeper in genotypes at single cells level.43
understanding of key biological processes and mecha- The scRNA-seq becomes a powerful tool to profile, iden-
nisms, which is an important direction for the develop- tify, classify and discover new or rare cell types and sub-
ment of single-cell technology in the future. In the field types from different human organs and tissues, giving
of single-cell RNA transcriptome research, there is still more profound information about health and disease in
much more potential for analytical algorithms and tools to development, immunology, diabetes, microbiology, Covid-
improve the exploration of data and better understanding 19, cancer biology, vascular biology, neurobiology, clini-
of cell functions. Therefore, we also encourage readers to cal diagnosis and many other disciplines (Figure 5A-I).
read other excellent reviews that focus on various aspects With these new findings unlocking health and disease, we
of scRNA-seq analysis for more inspiration.83 are witnessing rapid progress and changes despite some
remaining experimental and bioinformatics challenges. In
this section, we highlight and discuss a few important
4 APPLICATIONS OF SINGLE-CELL scRNA-seq applications in biomedical and clinical inves-
RNA SEQUENCING tigations.
Every tissue/organ contains much morphologically and
To date, single-cell RNA expression profiling is rapidly functionally diverse population of cells in different states,
becoming an irreplaceable method for various research physiological transitions, differentiation trajectories and
F I G U R E 5 Application of single-cell RNA sequencing technology. Single-cell RNA sequencing has been employed in different species
(humans, animals, plants) to improve understanding of normal and disease models. A special note is placed on human health, and many
single-cell RNA sequencing (scRNA-seq) methods are focused on understanding (A) development, (B) immunology, (C) diabetes, (D)
microbiology, (E) SARS-CoV-2, (F) cancer biology, (G) vascular biology (H) neurobiology and (I) clinical diagnostics. Figure was created with
BioRender with license for publication
spatial position. This complex but well-synchronized majority of difficulties derive from the fact that tumour
microenvironment keeps homeostasis until extreme con- tissues are differently positioned in the body; thus their
ditions occur that might turn over the normal cell archi- microenvironment contains variety of tumour and non-
tecture into, for example, tumours. To understand the ini- tumour cells in different states and stages. Moreover, the
tiation of tumours, evolutional origin of cells, tumour pro- cells sample mixture and proportions even within the same
gression, metastasis and therapeutic responses, it is impor- section of a tumour might be very different if the biopsy
tant to advance our understanding of tumour microenvi- was taken under different times and conditions. In addi-
ronments with essential immune and stromal infiltrates.94 tion, single-cell gene expression data often contain a lot
A scRNA-seq analysis can distinguish functionally healthy of noises, and thus cells of the same type might end up in
cells from cancer cells at various developmental stages of different clusters, and cells of different types can be in the
tumours. This allows more precise prognoses and diag- same cluster due to batch effects.96 Therefore, it is neces-
noses through the identification and determination of sen- sary to carefully sort out high-quality cell clusters before
sitivity to different drugs and develop the most effective calculating cell-type-specific reference matrix. Although
treatment strategies for cancers. Initially, scRNA-seq tech- scRNA-seq is very useful, RNA expression measurements
nology was focused on analysing individual part of the do not always provide information about protein level
organ, its heterogeneity and cell types involved resulting or post-translational modifications. Recently, scRNA-seq
in producing comprehensive data.95 Although there are studies are supported with other techniques including
many scRNA-seq reports on the individual parts of the mass cytometry (cytometry by time-of-flight, CyTOF)
tumours, its heterogeneity and cell types involved, but where for example both studies confirmed that regula-
what are the biological functions of each individual cell tory T cells (T-reg) in the tumour express higher levels
type, and how the cells talk and work with each other of tumour necrosis factor receptor superfamily member 9
to accomplish their tasks is still largely challenging. The (TNFRSF9; encoding 4-1BB), inducible T cell co-stimulator
(ICOS) and cytotoxic T lymphocyte-associated antigen 4 pairs119,120,71,72,74 and CellPhoneDB v2.0, which predicts
(CTLA4) than T-reg cells in blood or adjacent normal tis- enriched signaling interactions between two cell popula-
sue, possibly reflective of an activated state.97 Further- tions by considering the minimum average expression of
more, by adding spatial information to scRNA-seq data, we the members of the heteromeric complex.73 Another plat-
are able to understand molecular, cellular and spatial tis- form CellChat predicts major signaling inputs and outputs
sue organization and cell-to-cell interactions in situ.98–100 for cells and signals coordinate for functions using net-
Tumour microenvironments are infiltrated with the work analysis and pattern recognition approaches (http:
immune cell types, that is, T lymphocytes cells, CD8+ //www.cellchat.org/).70
T cells,101,102 tumour-associated macrophages,103 cancer- Another important aspect of heterogeneity in tumours
associated fibroblasts, epithelial cells and cancer stem that can be investigated by scRNA-seq is evolutionary
cells,104–109 but the types of immune responses and process of tumour formation that has been found to play
their effects on tumour growth, metastasis and death a significant role in the tumour formation as well as
vary greatly between different cancers and individual acquisition of traits such as chemotherapy treatments
tumours.110 Immune cells have both anti-tumour effect and resistance.121,122 The continuous accumulation of
inhibiting and killing tumour cells and pro-tumour heterogeneity may reflect the evolution of cancer, and
activities that promote tumour growths and immune scRNA-seq can provide meaningful insights into the
escape. minor treatment-resistant cell populations inside complex
So far, there are more advanced approaches reported tumours, which can be used to select appropriate therapies
on alterations of immune cells in tumours such as in based on tumour type and more precisely treat the indi-
lung adenocarcinoma,111 breast cancer,112,113 head and vidual patient.123 Studies on melanoma,105 liver cancer124
neck squamous cell carcinoma (HNSCC),114 nasopharyn- hepatocellular carcinoma,125 glioblastoma,126–128 breast
geal carcinoma,115 head and neck cancer,116 pancreatic cancer112 and prostate cancer129 integrate a variety of infor-
cancer.117 Non-tumour cells have been investigated start- mation in a single cancer cell, deciphering the secrets of
ing from the first study on metastatic melanoma where cancer heterogeneity and evolution. Furthermore, another
almost 5000 cells both malignant and non-malignant were emerging technology like spatial transcriptome sequenc-
analyzed,105 following the study led by Lambrechts and ing incorporates information on the spatial location of
colleagues where they identified 52 subtypes of stromal cells, providing information on gene expression hetero-
cells, including new subsets of cells. The study found that geneity; organoids can mimic some tumour heterogeneity
fibroblasts expressed different collagen proteins, endothe- and 3D organization that can be used for drug screening.
lial cells downregulated immune cell homing and genes All these technologies are indeed needed to properly
co-regulated with established immune checkpoint tran- understand the tumour eco-system in 3D volume, the role
scripts and were associated with T cell activity. This study of the endothelial cells in tumour, and how angiogenesis
provided a comprehensive map of stromal cell types, and develops in tumours and how tumours react to different
their phenotypes and cooperative behavior, giving a deeper treatments. Therefore, vasculature system that compro-
insight into cancer biology that will help advance lung can- mises the endothelial cells, which line the interior surface
cer diagnosis and treatment.46 In addition, recently Pelka of blood and lymphatic vessels plays important roles not
and colleagues developed a systematic approach to apply only in cancer but also diabetes and neurodegeneration for
on two most common type of human colorectal cancer, instance. Endothelial cells play a vital role in maintaining
mismatch repair deficient (MMRd) and MMR proficient the homeostasis, metabolism and functions of all tissues
(MMRp) tumours, discovering cell types, their underlying and organs in the body, such as the exchange of oxygen,
programs and cellular communities. They found ‘hotspots nutrients, hormones, fluid and metabolites between blood-
of immune activity’ comprised of chemokine-expressing stream and surrounding tissues causing various types of
malignant and non-malignant cells adjacent to activated dysfunctions. Although this squamous cell is morpholog-
T cells.118 ical alike in all vessels, there exists a great heterogeneity
In tumours, there are active communications between of functionally distinct phenotypes of endothelial cells.
different cell types, including tumour cells, through var- The degrees of endothelial cells heterogeneity have
ious signaling pathways. Identifying the communications been well recognized for decades.130–132 For instance,
between tumour and non-tumour cells will provide impor- endothelial cells from different vascular beds (artery, cap-
tant insights into the development of novel therapeutic illary and vein), tissues and diseases exhibit substantial
strategies. There have been also several computational heterogeneity. Recently, the heterogeneity of endothelial
methods developed to infer cell–cell communication from cells was further extended at the single-cell transcriptional
scRNA-seq data, such as SingleCellSignalR, iTalk and Nich- level. Using single-cell RNA sequencing, Kalucka et al.
eNet that usually use only one ligand-one receptor gene profile the single-cell transcription of over 30 000 endothe-
lial cells from 11 mouse organs. Seventy-eight endothelial hallmark features in postnatal β cell proliferation and
cell phenotypes with distinct transcriptome profile are mass expansion i.e. increased amino acid metabolism,
identified and provide the first murine endothelial cell cat- reactive oxygen species (ROS) levels, and Srf/Jun/Fos
alogue for future research.133 The degree of endothelial cell (SRF) transcription factors.139 Another group character-
heterogeneity can be further illustrated with the increased ized the mechanisms governing pancreatic β and α cells
of cells numbers analyzed. One such example is renal generation, expansion, and maturation during pancreatic
endothelial cells, which is composed of endothelial cells development by scRNA-seq. The study found that the
adapted to different physiological conditions in kidney proliferation rate of β- and α -cells peaks at different
compartments. By analysing over 40 000 renal endothelial developmental time and distinct cell-type development
cells with single-cell RNA sequencing, Dumas et al. iden- regulatory pathways were enriched for maturation. Unlike
tified over 20 different phenotypes of endothelial cells in adult β cells, juvenile β cells are more heterogeneous
the kidney, which exhibit a high degree of plasticity when reflecting distinct maturation states.140 Other studies
exposed to dehydration and hypertonicity conditions.134 deepens the understanding on lineage-tracing, molecular
In adult tissues, endothelial cells are mostly quiescent heterogeneity of precursor cells and identified rare or
but metabolically active. However, under pathological putative multipotent cells present at stage E13.25 in ductal
conditions such as tumourigenesis, quiescent endothelial termini.141
cells are activated and involved in the generation of new The sequence of developmental events is highly
blood vessels and disease progression. Targeted inhibition conserved between species, for instance, NEUROG3
of angiogenesis is thus not surprising one of the most is transiently and robustly expressed, in two waves, in
broadly used strategies in cancer therapies. However, mice142 whereas human NEUROG3 expression occurs in
almost all anti-angiogenesis therapies in cancers lead single wave.143 The embryonic islet cells of mice are mostly
to the development of drug resistance, as new mecha- monohormonal, whereas a large proportion of human islet
nisms evolve to replace the drug-inhibited angiogenesis cells are initially polyhormonal.144 The scRNA-seq study
pathway. Single-cell RNA sequencing can reveal novel confirmed that mouse and human β- and α-cells have
endothelial cell types and mechanisms. By single-cell differential expression of multiple genes between these
sequencing of over 50 000 endothelial cells from lung species.145 These examples highlight the need to confirm
cancers,135 nearly 30 000 endothelial cells from choroidal any finding obtained in mice in humans. Therefore, the
neovascularization136 and over 20 000 from co-opted use of the human pluripotent cells is integrated part of
breast cancer endothelial cells,137 the heterogeneity and developmental biology that can mimic human pancreas
transcriptomic/metabolic plasticity of endothelial cells development in vitro that has been confirmed in several
are consistently revealed in these pathological conditions. studies.146–151
Most importantly, novel endothelial cell-targeting tar- Moreover, 3D cell culture microenvironment more
gets with therapeutic potential have been identified by closely resembles in vivo embryogenesis and organogen-
single-cell RNA sequencing approaches. esis as compared to monolayer (2D) cell culture. This
Another important application of the scRNA-seq tech- novel approach also increases the functionality of hiPSC-
nology is the better understanding of β cell development derived β cells.152,153 Notably, 3D organoids are used as
and pathology in diabetes. Cure of type 1 diabetes (T1D) a patient-specific cell model that offer alternative plat-
lies in the restoration of the β cells. However, to generate form to study transcriptomes of T1D. Clustered Regu-
functional β cells requires extensive understanding of pan- larly Interspaced Short Palindromic Repeats (CRISPR)
creas development, its molecular events and knowledge Cas9 gene editing technology has increased the acces-
about the cellular heterogeneity in health and disease. sibility of genetically engineered hiPSCs, allowing the
scRNA-seq studies of developing pancreas were performed manipulation of known or putative regulators of develop-
firstly on the mouse models, following recent studies on ment for their function assessment in human tissues. Fur-
human pluripotent models mostly on embryonic stem thermore, scRNA-Seq can be used in combination with
cells (ESCs) and induced pluripotent stem cells (iPSCs) 3D CRISPR or lineage tracing.154,155 Despite all these great
models. The studies performed on mouse models revealed promises, it is important to note that it is particularly chal-
several important aspects of the pancreas development. lenging to study individual cells in the pancreas due to
Several groups started on revealing basics in developmen- the high hydrolytic enzyme content of the exocrine cells.
tal stage identifying a novel α-cell specific marker Slc38A Protocols for overcoming these limitations are evolving,
on wild type model,138 another group with Zeng et al. including snap-freezing of the dissected pancreas followed
found diverse β-cell heterogeneity and transcriptomic by single-nucleus RNA-Seq.22 Moreover, single-cell omics
dynamics during post-natal maturation and postnatal techniques other than scRNA-seq have been developed
β cell proliferation. The study identified several novel such as Patch-Seq.156
In addition to the application of scRNA-seq in basic RNA sequencing technologies and the computational tools
life science investigations, this technology has also been make the technology accessible and applicable in almost
demonstrated as a powerful tool for understanding infec- all applications in life sciences. One important ground
tious diseases. The COVID-19 pandemic, caused by coro- knowledge to revisit by using the single-cell RNA sequenc-
navirus SARS-CoV-2, has affected more than 248 million ing technology is the construction of single cell atlas in
people worldwide (by 3 November 2021). Understanding tissues, organs and organism. Towards this, considerable
the pathogenesis of COVID-19 infection is of great impor- efforts have been made to investigate and establish cell
tance in preventing transmission, reducing the severity atlases. Just to mention a few, body-wide single-cell atlas
of the infection and developing novel therapeutic strate- has been revisited for Caenorhabditis elegans, planarian,
gies quickly and efficiently. To date, a number of stud- D. melanogaster, zebrafish, mouse, Macaca fascicularis and
ies using single-cell RNA sequencing technology have human7,161–166 (Figure 6). Particularly, Han and cowork-
been conducted to understand the immune cell land- ers profiled all major human organs, including 60 differ-
scape and response in COVID-19 patients157–159 and result- ent human tissue types, and constructed a scheme for the
ing in variations of clinical outcomes depending on the human cell landscape (HCL) for the very first time.164
age, sex, severity and COVID-19 disease stages.157 They They have uncovered a single-cell hierarchy for many tis-
found that COVID-19 induced a unique immune cell sig- sues that have not been well characterized. They profiled
naling in humans compared to healthy controls, partic- more than 599 000 cells using microwell-seq and estab-
ularly during the early recovery phase. Unique immune lished a ‘single-cell HCL analysis’ pipeline that helps to
cell signaling was found in infected humans compared define human cell identity, genetic networks, progenitor
to healthy controls.158 Differences in the composition of and adult cells. With the development of the technology
key immune cells between moderate, severe, convalescent and applications, more and more single-cell RNA sequenc-
COVID-19 patients and the control group by performing ing data are expected to be generated and integrated in a
scRNA-seq on peripheral blood from COVID-19 patients publicly accessible database to facilitate the understanding
and healthy individuals were identified. Most cell types gene and cell functions in health and diseases.
in COVID-19 patients showed a robust interferon alpha Combining scRNA-seq and other large scale-genetic
response and an acute immune response.159 In addition screening tools will be further expanding the applications
to single-cell sequencing of immune cells from periph- of the technology. One such combinational technology
eral blood mononuclear cells (PBMC) and bronchoalveo- is combing scRNA-seq and CRISPR-based genome-scale
lar lavage (BAL), single nuclei RNA sequencing of COVID- genetic screening, such as Perturb-seq that enables the
19 tissues/organs have also provided important patholog- assessment of transcriptional effects of knocking out sev-
ical insights in the disease severity and progression. By eral genes with CRISPR,167 and LinTIMaT that integrates
single-cell sequencing of 24 lung, 16 kidney, 16 liver and 19 single-cell transcriptome data and mutation data for lin-
heart autopsy tissue samples and spatial transcriptomics eage tracing.168 In addition to CRISPR-mediated mutagen-
sequencing of 14 lung samples from COVID-19 patients, esis, it is also possible to combine scRNA-seq and CRISPR-
Delorey et al.160 reveal the biological effects of severe mediated gene activation or interference.169,170 These com-
SARS-COV-2 infection and remodeling of lung epithelial, binational applications allow us to investigate the genetic
immune and stromal compartments in patients. The pan- effect on the cellular transcriptome and functions in a large
demic is still far from complete dissolution, and scRNA-seq scale. With the continuous development of both single-
would certainly remain an important pipeline to properly cell RNA sequencing and CRISPR gene editing, such as
puzzle immune responses on different variants across the prime editing,171 more such combinational technologies
globe. Taken together, single-cell RNA sequencing tech- and applications are expected to be arrived and contributed
nology gained more scientific insights in the fight against to the better understanding of gene and cell functions.
COVID-19 and can be used in the future for detecting not In this review, we focus on the single-cell RNA sequenc-
only current SARS-CoV-2 but also the other pathogens in ing technologies and its applications. However, it should
combination with conventional methods. be noted that single-cell sequencing technology has
been developed to measure nearly all OMICS, such as
single-cell whole-genome sequencing,172 single-cell copy
5 CONCLUSION REMARKS AND number variation sequencing,173 single-cell epigenetic
FUTURE PERSPECTIVES markers (i.e., DNA methylation, chromatin accessibility)
sequencing,174 single-cell proteinomics175 and single-cell
Single-cell RNA sequencing has proven as one of the metabolomics.176 More and more multiomics studies and
transforming technologies in life sciences over the past analyses are expected to be carried out to fully character-
decade. The development of high throughput single-cell ize the gene regulatory processes, functions, molecules and
F I G U R E 6 Cell atlases of model organisms. First cell atlases of model organism Caenorhabditis elegans (A), planarian (B), Drosophila
melanogaster (C), zebrafish (D), mouse (E), monkey (F), and human (G). Year of published data, cell number and cell type analyzed by
single-cell RNA sequencing (scRNA-seq) were indicated. Icons of model organisms are created with BioRender with license for publication
interactions for cell types in healthy tissues/organs and in seq data processing, analysis, presentation and interpre-
diseased conditions.177 tation. Automatic scRNA-seq data analysis pipelines with
Despite all these great promises, one major disadvan- user friendly interphase, and most importantly, which
tage of single cell RNA sequencing is the loss of histo- can be used by personnel without any bioinformatic
logical information as both single cell and single nuclei skills and background, are needed to further broaden the
suspensions have to be prepared from tissues. Although scRNA-seq-based clinical applications. One such example
trajectory analysis can help with projecting the associa- is the single-cell omics workbench from the Galaxy Com-
tion and transition between different cell types, other con- munity (https://galaxyproject.org/use/singlecell/), which
founding factors associate with tissue digestion, cell iso- integrates more than 20 bioinformatics tools. Since a large
lation and preservation could alter gene expression and number of open-source tools have been developed for this
cell representation. Spatial dimension of single-cell tran- purpose (see Table S1), more streamlined and automatic
scriptomics represents an essential step and breakthrough scRNA-seq data analysis and visualization platforms are
in the field to investigate whole organism architecture at expected to generate and be available in the future. In con-
the molecular level. Several spatial transcriptomics meth- clusion, we have presented a brief and concise overview
ods have been developed and demonstrated in proof-of- of single-cell RNA sequencing technology and its appli-
concept studies, such as barcoded array-based capture of cations. The continuous development of the technology
transcripts on microdissected tissues and in situ sequenc- will broaden its applications in clinical and personalized
ing. In 2020, spatially resolved transcriptomics technology medicine.
has been selected as the Method of the Year by Nature
Methods. Spatially and temporally revealing the single- AC K N OW L E D G E M E N T S
cell transcriptions in a complex tissues and organs will We would like to thank Fred Dubee, Lars Bolund
be the rising transforming tools to understand composi- and Huanming Yang for their critical comments to the
tion, complexity, interaction and functions of cells in tis- manuscript. The single-cell project was partially supported
sues/organs/organisms. by the Qingdao-Europe Advanced Institute for Life Sci-
Another promising application in the future is inte- ences. L.L. is supported by the Independent Research Fund
gration of the scRNA-seq technology into routine clini- Denmark (DFF, Sapere Aude Starting grant 8048-00072A).
cal diagnoses and personalized medicine. However, cur- We thank the China National GeneBank for the support
rently most scRNA-seq-based clinical studies are still at of executing the project under the framework of Genome
their exploratory phases, mainly focusing on revisiting and Read and Write.
better understanding the disease processes and identifica-
tion of diagnosis and therapeutic markers. Although the
cost per cell has been reduced significantly, the cost per CONFLICT OF INTEREST
sample (including the library preparation and sequenc- The authors declare no conflict of interest.
ing) is still substantially high (Figure 1). This remains one
limiting factor to use the scRNA-seq as a routine diag- ORCID
nostic tool. Other remaining challenges are the scRNA- Yonglun Luo https://orcid.org/0000-0002-0007-7759
REFERENCES 21. Koenitzer JR, Wu H, Atkinson JJ, et al. Single nucleus

1. Bianconi E, Piovesan A, Facchin F, et al. An estimation of the RNASeq profiling of mouse lung: reduced dissociation bias and
number of cells in the human body. Ann Hum Biol. 2013;40:463- improved detection of rare cell types compared with single
471. cell RNASeq. bioRxiv. 2020. https://doi.org/10.1101/2020.03.06.
2. Arendt D, Musser JM, Baker CVH, et al. The origin and evolu- 981407
tion of cell types. Nat Rev Genet. 2016;17:744-757. 22. Tosti L, Hang Y, Debnath O, et al. Single-nucleus and in situ
3. Coons AH, Creech HJ, Jones RN. Immunological properties of RNA–sequencing reveal cell topographies in the human pan-
an antibody containing a fluorescent group. Proc Soc Exp Biol creas. Gastroenterology. 2021;160:1330-1344.e11.
Med. 1941;47:200-202. 23. Slyper M, Porter CBM, Ashenberg O, et al. A single-cell and
4. Murphy D. Gene expression studies using microarrays: princi- single-nucleus RNA-Seq toolbox for fresh and frozen human
ples, problems, and prospects. Adv Physiol Educ. 2002;26:256- tumors. Nat Med. 2020;26:792-802.
270. 24. Grindberg RV, Yee-Greenbaum JL, McConnell MJ, et al.
5. Tang F, Barbacioru C, Wang Y, et al. mRNA-Seq whole- RNA-sequencing from single nuclei. Proc Natl Acad Sci.
transcriptome analysis of a single cell. Nat Methods. 2009;6:377- 2013;110:19802-19807.
382. 25. Kolodziejczyk AA, Kim JK, Svensson V, et al. The technol-
6. Brennecke P, Anders S, Kim JK, et al. Accounting for tech- ogy and biology of single-cell RNA sequencing. Mol Cell.
nical noise in single-cell RNA-seq experiments. Nat Methods. 2015;58:610-620.
2013;10:1093-1095. 26. Goetz JJ, Trimarchi JM. Transcriptome sequencing of single
7. Cao J, Packer JS, Ramani V, et al. Comprehensive single-cell cells with Smart-Seq. Nat Biotechnol. 2012;30:763-765.
transcriptional profiling of a multicellular organism. Science. 27. Picelli S, Björklund ÅK, Faridani OR, et al. Smart-Seq2 for sen-
2017;357:661-667. sitive full-length transcriptome profiling in single cells. Nat
8. Fan HC, Fu GK, Fodor SPA. Combinatorial labeling Methods. 2013;10:1096-1098.
of single cells for gene expression cytometry. Science. 28. Xin Y, Kim J, Ni M, et al. Use of the fluidigm C1 platform for
2015;347(6222):1258367. https://doi.org/10.1126/science.1258367 RNA sequencing of single mouse pancreatic islet cells. Proc Natl
9. Klein AM, Mazutis L, Akartuna I, et al. Droplet barcoding Acad Sci. 2016;113:3293-3298.
for single-cell transcriptomics applied to embryonic stem cells. 29. Macosko EZ, Basu A, Satija R, et al. Highly parallel genome-
Cell. 2015;161:1187-1201. wide expression profiling of individual cells using nanoliter
10. Rodriques SG, Stickels RR, Goeva A, et al. Slide-seq: a scalable droplets. Cell. 2015;161:1202-1214.
technology for measuring genome-wide expression at high spa- 30. Zheng GXY, Terry JM, Belgrader P, et al. Massively parallel
tial resolution. Science. 2019;363:1463-1467. digital transcriptional profiling of single cells. Nat Commun.
11. Hawrylycz MJ, Lein ES, Guillozet-Bongaarts AL, et al. An 2017;8:1-12.
anatomically comprehensive atlas of the adult human brain 31. Sheng K, Cao W, Niu Y, et al. Effective detection of variation
transcriptome. Nature. 2012;489:391-399. in single-cell transcriptomes using MATQ-seq. Nat Methods.
12. Ziegenhain C, Vieth B, Parekh S, et al. Comparative analysis 2017;14:267-270.
of single-cell RNA sequencing methods. Mol Cell. 2017;65:631- 32. Gierahn TM, Wadsworth MH, Hughes TK, et al. Seq-Well:
643.e4. portable, low-cost RNA sequencing of single cells at high
13. Macaulay IC, Voet T. Single cell genomics: advances and future throughput. Nat Methods. 2017;14:395-398.
perspectives. PLoS Genet. 2014;10:e1004126. 33. Zhu YY, Machleder EM, Chenchik A, et al. Reverse transcrip-
14. Sanz E, Yang L, Su T, et al. Cell-type-specific isolation of tase template switching: a SMART™ approach for full-length
ribosome-associated mRNA from complex tissues. Proc Natl cDNA library construction. BioTechniques. 2001;30:892-897.
Acad Sci. 2009;106:13939-13944. 34. Hedlund E, Deng Q. Single-cell RNA sequencing: technical
15. Zeb Q, Wang C, Shafiq S, et al. An Overview of Single-Cell Isola- advancements and biological applications. Mol Aspects Med.
tion Techniques. Elsevier; 2019. 2018;59:36-46.
16. Van Den Brink SC, Sage F, Vértesy Á, et al. Single-cell sequenc- 35. Hashimshony T, Wagner F, Sher N, et al. CEL-Seq: single-
ing reveals dissociation-induced gene expression in tissue sub- cell RNA-Seq by multiplexed linear amplification. Cell Rep.
populations. Nat Methods. 2017;14:935-936. 2012;2:666-673.
17. Adam M, Potter AS, Potter SS. Psychrophilic proteases dramat- 36. Jaitin DA, Kenigsberg E, Keren-Shaul H, et al. Massively paral-
ically reduce single-cell RNA-seq artifacts: a molecular atlas of lel single-cell RNA-seq for marker-free decomposition of tissues
kidney development. Development. 2017;144:3625-3632. into cell types. Science. 2014;343:776-779.
18. Ding J, Adiconis X, Simmons SK, et al. Systematic compari- 37. Fu GK, Hu J, Wang P-H, et al. Counting individual DNA
son of single-cell and single-nucleus RNA-sequencing meth- molecules by the stochastic attachment of diverse labels. Proc
ods. Nat Biotechnol. 2020;38:737-746. Natl Acad Sci. 2011;108:9026-9031.
19. Zeng W, Jiang S, Kong X, et al. Single-nucleus RNA-seq of dif- 38. Liu C, et al. A portable and cost-effective microfluidic system for
ferentiating human myoblasts reveals the extent of fate hetero- massively parallel single-cell transcriptome profiling. Biorxiv.
geneity. Nucleic Acids Res. 2016;44:e158-e158. 2019; https://www.biorxiv.org/content/10.1101/818450v3
20. Wolfien M, Galow A-M, Müller P, et al. Single-nucleus sequenc- 39. Huang J, Liang X, Xuan Y, et al. A reference human
ing of an entire mammalian heart: cell type composition and genome dataset of the BGISEQ-500 sequencer. Gigascience.
velocity. Cells. 2020;9:318. 2017;6:gix024.
40. Legnini I, Alles J, Karaiskos N, et al. FLAM-seq: full-length 61. Huang Q, Liu Y, Du Y, et al. Evaluation of cell type annotation
mRNA sequencing reveals principles of poly (A) tail length con- R packages on single-cell RNA-seq data. Genomics Proteomics
trol. Nat Methods. 2019;16:879-886. Bioinf. 2020. https://doi.org/10.1101/827139
41. Hicks SC, Townes FW, Teng M, et al. Missing data and techni- 62. Chen KH, Boettiger AN, Moffitt JR, et al. RNA imaging. Spa-
cal variability in single-cell RNA-sequencing experiments. Bio- tially resolved, highly multiplexed RNA profiling in single cells.
statistics. 2018;19:562-578. Science. 2015;348:aaa6090.
42. Zappia L, Phipson B, Oshlack A. Exploring the single-cell RNA- 63. Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrich-
seq analysis landscape with the scRNA-tools database. PLoS ment tools: paths toward the comprehensive functional analysis
Comput Biol. 2018;14:e1006245. of large gene lists. Nucleic Acids Res. 2009;37:1-13.
43. Stuart T, Butler A, Hoffman P, et al. Comprehensive integration 64. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set varia-
of single-cell data. Cell. 2019;177:1888-1902.e21. tion analysis for microarray and RNA-seq data. BMC Bioinf.
44. Lun AT, McCarthy DJ, Marioni JC. A step-by-step workflow for 2013;14:1-15.
low-level analysis of single-cell RNA-seq data with Bioconduc- 65. Aibar S, González-Blas CB, Moerman T, et al. SCENIC: single-
tor. F1000Res. 2016;5:2122. cell regulatory network inference and clustering. Nat Methods.
45. Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell 2017;14:1083-1086.
gene expression data analysis. Genome Biol. 2018;19:1-5. 66. Van De Sande B, Flerin C, Davie K, et al. A scalable SCENIC
46. Lambrechts D, Wauters E, Boeckx B, et al. Phenotype molding workflow for single-cell gene regulatory network analysis. Nat
of stromal cells in the lung tumor microenvironment. Nat Med. Protoc. 2020;15:2247-2276.
2018;24:1277-1289. 67. Ma A, Wang C, Chang Y, et al. IRIS3: integrated cell-type-
47. Fan X, Bialecka M, Moustakas I, et al. Single-cell reconstruction specific regulon inference server from single-cell RNA-Seq.
of follicular remodeling in the human adult ovary. Nat Com- Nucleic Acids Res. 2020;48:W275-W286.
mun. 2019;10:1-13. 68. Saelens W, Cannoodt R, Todorov H, et al. A comparison
48. Lytal N, Ran D, An L. Normalization methods on single-cell of single-cell trajectory inference methods. Nat Biotechnol.
RNA-seq data: an empirical survey. Front Genet. 2020;11:41. 2019;37:547-554.
49. Yip SH, Sham PC, Wang J. Evaluation of tools for highly vari- 69. Qiu X, Mao Q, Tang Y, et al. Reversed graph embedding resolves
able gene discovery from single-cell RNA-seq data. Brief Bioin- complex single-cell trajectories. Nat Methods. 2017;14:979-982.
form. 2019;20:1583-1589. 70. Jin S, Guerrero-Juarez CF, Zhang L, et al. Inference and anal-
50. Tran HTN, Ang KS, Chevrier M, et al. A benchmark of batch- ysis of cell-cell communication using CellChat. Nat Commun.
effect correction methods for single-cell RNA sequencing data. 2021;12:1-20.
Genome Biol. 2020;21:1-32. 71. Browaeys R, Saelens W, Saeys Y. NicheNet: modeling intercel-
51. Goh WWB, Wang W, Wong L. Why batch effects matter lular communication by linking ligands to target genes. Nat
in omics data, and how to avoid them. Trends Biotechnol. Methods. 2020;17:159-162.
2017;35:498-507. 72. Cabello-Aguilar S, Alame M, Kon-Sun-Tack F, et al. Single-
52. Hie B, Bryson B, Berger B. Efficient integration of heteroge- CellSignalR: inference of intercellular networks from single-
neous single-cell transcriptomes using Scanorama. Nat Biotech- cell transcriptomics. Nucleic Acids Res. 2020;48:e55-e55.
nol. 2019;37:685-691. 73. Efremova M, Vento-Tormo M, Teichmann SA, et al. Cell-
53. Hao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of PhoneDB: inferring cell–cell communication from combined
multimodal single-cell data. Cell. 2021;184:3573-3587.e29. expression of multi-subunit ligand–receptor complexes. Nat
54. Zou B, Zhang T, Zhou R, et al. deepMNN: deep learning-based Protoc. 2020;15:1484-1506.
single-cell RNA sequencing data batch correction using mutual 74. Wang Y, Wang R, Zhan S, et al. iTALK: an R package to charac-
nearest neighbors. Front Genet. 2021;12:708981. terize and illustrate intercellular communication. BioRxiv. 2019.
55. Lake BB, Chen S, Hoshi M, et al. A single-nucleus RNA- https://doi.org/10.1101/507871
sequencing pipeline to decipher the molecular anatomy and 75. Scialdone A, Natarajan KN, Saraiva LR, et al. Computational
pathophysiology of human kidneys. Nat Commun. 2019;10:1-15. assignment of cell-cycle stage from single-cell transcriptome
56. Shaham U, Stanton KP, Zhao J, et al. Removal of batch effects data. Methods. 2015;85:54-61.
using distribution-matching residual networks. Bioinformatics. 76. Datlinger P, Rendeiro AF, Schmidl C, et al. Pooled CRISPR
2017;33:2539-2546. screening with single-cell transcriptome readout. Nat Methods.
57. Lotfollahi M, Wolf FA, Theis FJ. scGen predicts single-cell per- 2017;14:297-301.
turbation responses. Nat Methods. 2019;16:715-721. 77. Ranzoni AM, Tangherloni A, Berest I, et al. Integrative single-
58. Wu D, Yeong Poh Sheng J, Tan Su-En G, et al. Comparison cell RNA-seq and ATAC-seq analysis of human developmental
between UMAP and t-SNE for multiplex-immunofluorescence hematopoiesis. Cell Stem Cell. 2021;28:472-487.e7.
derived single-cell data from tissue sections. BioRxiv. 2019. 78. Hu Y, Huang K, An Q, et al. Simultaneous profiling of tran-
https://doi.org/10.1101/549659 scriptome and DNA methylome from a single cell. Genome Biol.
59. Duò A, Robinson MD, Soneson C. A systematic performance 2016;17:1-11.
evaluation of clustering methods for single-cell RNA-seq data. 79. Hu Y, An Q, Guo Y, et al. Simultaneous profiling of mRNA tran-
F1000Res. 2018;7:1141. scriptome and DNA methylome from a single cell. Methods Mol
60. Clarke ZA, Andrews TS, Atif J, et al. Tutorial: guidelines for Biol. 2019;1979:363-377.
annotating single-cell transcriptomic maps using automated 80. Golomb SM, Guldner IH, Zhao A, et al. Multi-modal single-
and manual methods. Nat Protoc. 2021;16:2749-2764. cell analysis reveals brain immune landscape plasticity
during aging and gut microbiota dysbiosis. Cell Rep. 2020;33: 100. Asp M, Giacomello S, Larsson L, et al. A spatiotemporal organ-
108438. wide gene expression and cell atlas of the developing human
81. Stoeckius M, Hafemeister C, Stephenson W, et al. Simultane- heart. Cell. 2019;179:1647-1660.e19.
ous epitope and transcriptome measurement in single cells. Nat 101. Guo X, Zhang Y, Zheng L, et al. Global characterization of T
Methods. 2017;14:865-868. cells in non-small-cell lung cancer by single-cell sequencing.
82. Baccin C, Al-Sabah J, Velten L, et al. Combined single-cell and Nat Med. 2018;24:978-985.
spatial transcriptomics reveal the molecular, cellular and spa- 102. Sato E, Olson SH, Ahn J, et al. Intraepithelial CD8+ tumor-
tial bone marrow niche organization. Nat Cell Biol. 2020;22:38- infiltrating lymphocytes and a high CD8+/regulatory T cell
48. ratio are associated with favorable prognosis in ovarian cancer.
83. Luecken MD, Theis FJ. Current best practices in single-cell Proc Natl Acad Sci. 2005;102:18538-18543.
RNA-seq analysis: a tutorial. Mol Syst Biol. 2019;15:e8746. 103. Zhang Q, He Y, Luo N, et al. Landscape and dynamics of single
84. Shaw R, Tian X, Xu J. Single-cell transcriptome analysis in immune cells in hepatocellular carcinoma. Cell. 2019;179:829-
plants: advances and challenges. Molecular Plant. 2020;14:115- 845.e20.
126. 104. Tirosh I, Venteicher AS, Hebert C, et al. Single-cell RNA-
85. Denyer T, Ma X, Klesen S, et al. Spatiotemporal developmen- seq supports a developmental hierarchy in human oligoden-
tal trajectories in the Arabidopsis root revealed using high- droglioma. Nature. 2016;539:309-313.
throughput single-cell RNA sequencing. Dev Cell. 2019;48:840- 105. Tirosh I, Izar B, Prakadan SM, et al. Dissecting the multicellu-
852.e5. lar ecosystem of metastatic melanoma by single-cell RNA-seq.
86. Shulse CN, Cole BJ, Ciobanu D, et al. High-throughput single- Science. 2016;352:189-196.
cell transcriptome profiling of plant cell types. Cell Rep. 106. Harper J, Sainson RCA. Regulation of the anti-tumour immune
2019;27:2241-2247.e4. response by cancer-associated fibroblasts. Semin Cancer Biol.
87. Zhang T-Q, Xu Z-G, Shang G-D. A single-cell RNA sequenc- 25, 2014: 69-77.
ing profiles the developmental landscape of Arabidopsis root. 107. Bartoschek M, Oskolkov N, Bocci M, et al. Spatially and func-
Molecular plant. 2019;12:648-660. tionally distinct subclasses of breast cancer-associated fibrob-
88. Nelms B, Walbot V. Defining the developmental program lead- lasts revealed by single cell RNA sequencing. Nat Commun.
ing to meiosis in maize. Science. 2019;364:52-56. 2018;9:1-13.
89. Tian C, Du Q, Xu M, et al. Single-nucleus RNA-seq resolves 108. Calon A, Espinet E, Palomo-Ponce S, et al. Dependency of col-
spatiotemporal developmental trajectories in the tomato shoot orectal cancer on a TGF-β-driven program in stromal cells for
apex. bioRxiv. 2020. https://doi.org/10.1101/2020.09.20.305029 metastasis initiation. Cancer Cell. 2012;22:571-584.
90. Wang Y, Huan Q, Li K, et al. Single-cell transcriptome atlas of 109. Turley SJ, Cremasco V, Astarita JL. Immunological hallmarks
the leaf and root of rice seedlings. J Genet Genom. 2021;48:881- of stromal cells in the tumour microenvironment. Nat Rev
898. Immunol. 2015;15:669-682.
91. Song Q, Ando A, Jiang N, et al. Single-cell RNA-seq anal- 110. Thorsson V, Gibbs DL, Brown SD, et al. The immune landscape
ysis reveals ploidy-dependent and cell-specific transcriptome of cancer. Immunity. 2018;48:812-830.e14.
changes in Arabidopsis female gametophytes. Genome Biol. 111. He D, Wang D, Lu P, et al. Single-cell RNA sequencing reveals
2020;21:1-18. heterogeneous tumor and immune cell populations in early-
92. Rhee SY, Birnbaum KD, Ehrhardt DW. Towards building a stage lung adenocarcinomas harboring EGFR mutations. Onco-
plant cell atlas. Trends Plant Sci. 2019;24:303-310. gene. 2021;40:355-368.
93. Ma X, Denyer T, Timmermans MCP. PscB: a browser to 112. Chung W, Eum HH, Lee H-O, et al. Single-cell RNA-seq enables
explore plant single cell RNA-sequencing data sets. Plant Phys- comprehensive tumour and immune cell profiling in primary
iol. 2020;183:464-467. breast cancer. Nat Commun. 2017;8:1-12.
94. Levitin HM, Yuan J, Sims PA. Single-cell transcriptomic analy- 113. Azizi E, Carr AJ, Plitas G, et al. Single-cell map of diverse
sis of tumor heterogeneity. Trends Cancer. 2018;4:264-268. immune phenotypes in the breast tumor microenvironment.
95. Cao J, Spielmann M, Qiu X, et al. The single-cell transcriptional Cell. 2018;174:1293-1308.e36.
landscape of mammalian organogenesis. Nature. 2019;566:496- 114. Qi Z, Liu Y, Mints M, et al. Single-cell deconvolution of head
502. and neck squamous cell carcinoma. Cancers. 2021;13:1230.
96. Haghverdi L, Lun ATL, Morgan MD, et al. Batch effects 115. Zhao J, Guo C, Xiong F, et al. Single cell RNA-seq reveals the
in single-cell RNA-sequencing data are corrected by match- landscape of tumor and infiltrating immune cells in nasopha-
ing mutual nearest neighbors. Nat Biotechnol. 2018;36:421- ryngeal carcinoma. Cancer Lett. 2020;477:131-143.
427. 116. Yu X, Chen YA, Conejo-Garcia JR, et al. Estimation of immune
97. Lavin Y, Kobayashi S, Leader A, et al. Innate immune land- cell content in tumor using single-cell RNA-seq reference data.
scape in early lung adenocarcinoma by paired single-cell anal- BMC Cancer. 2019;19:1-11.
yses. Cell. 2017;169:750-765.e17. 117. Pan Y, Lu F, Fei Q, et al. Single-cell RNA sequencing reveals
98. Ji AL, Rubin AJ, Thrane K, et al. Multimodal analysis of com- compartmental remodeling of tumor-infiltrating immune cells
position and spatial architecture in human squamous cell car- induced by anti-CD47 targeting in pancreatic cancer. J Hematol
cinoma. Cell. 2020;182:1661-1662. Oncol. 2019;12:1-18.
99. Chen W-T, Lu A, Craessaerts K, et al. Spatial transcriptomics 118. Pelka K, Hofree M, Chen JH, et al. Multicellular immune hubs
and in situ sequencing to study Alzheimer’s disease. Cell. and their organization in MMRd and MMRp colorectal cancer.
2020;182:976-991.e19. bioRxiv. 2021;184(18):4734-4752.e20.
119. Wang S, Karikomi M, Maclean AL, et al. Cell lineage and com- 139. Zeng C, Mulas F, Sui Y, et al. Pseudotemporal ordering of single
munication network inference via optimization for single-cell cells reveals metabolic control of postnatal β cell proliferation.
transcriptomics. Nucleic Acids Res. 2019;47:e66-e66. Cell Metab. 2017;25:1160-1175.e11.
120. Raredon MSB, Adams TS, Suhail Y, et al. Single-cell con- 140. Qiu W-L, Zhang Y-W, Feng Y, et al. Deciphering pancreatic
nectomic analysis of adult mammalian lungs. Sci Adv. islet β cell and α cell maturation pathways and characteris-
2019;5:eaaw3851. tic features at the single-cell level. Cell Metab. 2017;25:1194-
121. Klein CA. Selection and adaptation during metastatic cancer 1205.e4.
progression. Nature. 2013;501:365-372. 141. Sznurkowska MK, Hannezo E, Azzarelli R, et al. Defining lin-
122. Chisholm RH, Lorenzi T, Lorz A, et al. Emergence of drug tol- eage potential and fate behavior of precursors during pancreas
erance in cancer cell populations: an evolutionary outcome of development. Dev Cell. 2018;46:360-375.e5.
selection, nongenetic instability, and stress-induced adaptation. 142. Gradwohl G, Dierich A, Lemeur M, et al. neurogenin3 is
Cancer Res. 2015;75:930-939. required for the development of the four endocrine cell lineages
123. Guo M, Peng Y, Gao A, et al. Epigenetic heterogeneity in cancer. of the pancreas. Proc Natl Acad Sci. 2000;97:1607-1611.
Biomark Res. 2019;7:1-19. 143. Lyttle BM, Li J, Krishnamurthy M, et al. Transcription factor
124. Su X, Zhao L, Shi Y, et al. Clonal evolution in liver cancer expression in the developing human fetal endocrine pancreas.
at single-cell and single-variant resolution. J Hematol Oncol. Diabetologia. 2008;51:1169-1180.
2021;14:1-5. 144. Bocian-Sobkowska J, Zabel M, Wozniak W, et al. Polyhormonal
125. Ma L, Wang L, Khatib SA, et al. Single-cell atlas of tumor cell aspect of the endocrine cells of the human fetal pancreas. His-
evolution in response to therapy in hepatocellular carcinoma tochem Cell Biol. 1999;112:147-153.
and intrahepatic cholangiocarcinoma. J Hepatol. 2021;75:1397- 145. Xin Y, Kim J, Okamoto H, et al. RNA sequencing of single
1408. human islet cells reveals type 2 diabetes genes. Cell Metab.
126. Dirkse A, Golebiewska A, Buder T, et al. Stem cell-associated 2016;24:608-615.
heterogeneity in glioblastoma results from intrinsic tumor 146. Augsornworawat P, Maxwell KG, Velazco-Cruz L, et al.
plasticity shaped by the microenvironment. Nat Commun. Single-cell transcriptome profiling reveals β cell maturation
2019;10:1-16. in stem cell-derived islets after transplantation. Cell Rep.
127. Darmanis S, Sloan SA, Croote D, et al. Single-cell RNA-seq 2020;32:108067.
analysis of infiltrating neoplastic cells at the migrating front of 147. Maxwell KG, Augsornworawat P, Velazco-Cruz L, et al. Gene-
human glioblastoma. Cell Rep. 2017;21:1399-1410. edited human stem cell–derived β cells from a patient with
128. Patel AP, Tirosh I, Trombetta JJ, et al. Single-cell RNA-seq high- monogenic diabetes reverse preexisting diabetes in mice. Sci
lights intratumoral heterogeneity in primary glioblastoma. Sci- Transl Med. 2020;12:eaax9106.
ence. 2014;344:1396-1401. 148. Sharon N, Vanderhooft J, Straubhaar J, et al. Wnt signaling sep-
129. Horning AM, Wang Y, Lin C-K, et al. Single-Cell RNA-seq arates the progenitor and endocrine compartments during pan-
reveals a subpopulation of prostate cancer cells with enhanced creas development. Cell Rep. 2019;27:2281-2291.e5.
cell-cycle–related transcription and attenuated androgen 149. Veres A, Faust AL, Bushnell HL, et al. Charting cellular
response. Cancer Res. 2018;78:853-864. identity during human in vitro β-cell differentiation. Nature.
130. Aird WC. Phenotypic heterogeneity of the endothelium: II. Rep- 2019;569:368-373.
resentative vascular beds. Circ Res. 2007;100:174-190. 150. Balboa D, Saarimäki-Vire J, Borshagovski D, et al. Insulin muta-
131. Aird WC. Phenotypic heterogeneity of the endothelium: I. tions impair beta-cell development in a patient-derived iPSC
Structure, function, and mechanisms. Circ Res. 2007;100:158- model of neonatal diabetes. Elife. 2018;7:e38519.
173. 151. Krentz NAJ, Lee MYY, Xu EE, et al. Single-cell transcriptome
132. Conway EM, Carmeliet P. The diversity of endothelial cells: a profiling of mouse and hESC-derived pancreatic progenitors.
challenge for therapeutic angiogenesis. Genome Biol. 2004;5:1- Stem cell reports. 2018;11:1551-1564.
5. 152. Velazco-Cruz L, Song J, Maxwell KG, et al. Acquisition of
133. Kalucka J, De Rooij L, Goveia J, et al. Single-cell transcriptome dynamic function in human stem cell-derived β cells. Stem cell
atlas of murine endothelial cells. Cell. 2020;180:764-779.e20. reports. 2019;12:351-365.
134. Dumas SJ, Meta E, Borri M, et al. Single-cell RNA sequencing 153. Russ HA, Parent AV, Ringler JJ, et al. Controlled induction
reveals renal endothelium heterogeneity and metabolic adapta- of human pancreatic progenitors produces functional beta-like
tion to water deprivation. J Am Soc Nephrol. 2020;31:118-138. cells in vitro. EMBO J. 2015;34:1759-1772.
135. Goveia J, Rohlenova K, Taverna F, et al. An integrated gene 154. Rooijers K, Markodimitraki CM, Rang FJ, et al. Simultaneous
expression landscape profiling approach to identify lung tumor quantification of protein–DNA contacts and transcriptomes in
endothelial cell heterogeneity and angiogenic candidates. Can- single cells. Nat Biotechnol. 2019;37:766-772.
cer Cell. 2020;37:21-36.e13. 155. Raj B, Wagner DE, Mckenna A, et al. Simultaneous single-cell
136. Rohlenova K, Goveia J, García-Caballero M, et al. Single- profiling of lineages and cell types in the vertebrate brain. Nat
cell RNA sequencing maps endothelial metabolic plasticity in Biotechnol. 2018;36:442-450.
pathological angiogenesis. Cell Metab. 2020;31:862-877.e14. 156. Camunas-Soler J, Dai X-Q, Hang Y, et al. Patch-seq links single-
137. Teuwen L-A, De Rooij L, Cuypers A, et al. Tumor vessel co- cell transcriptomes to human islet dysfunction in diabetes. Cell
option probed by single-cell analysis. Cell Rep. 2021;35:109253. Metab. 2020;31:1017-1031.e4.
138. Stanescu DE, Yu R, Won K-J, et al. Single cell transcriptomic 157. Ren X, Wen W, Fan X, et al. COVID-19 immune features
profiling of mouse pancreatic progenitors. Physiol Genomics. revealed by a large-scale single-cell transcriptome atlas. Cell.
2017;49:105-114. 2021;184:1895-1913.e1819.
158. Wen W, Su W, Tang H, et al. Immune cell profiling of COVID- 170. Genga RMJ, Kernfeld EM, Parsi KM, et al. Single-Cell
19 patients in the recovery stage by single-cell sequencing. Cell RNA-sequencing-based CRISPRi screening resolves molecu-
Discovery. 2020;6:1-18. lar drivers of early human endoderm development. Cell Rep.
159. Zhang J-Y, Wang X-M, Xing X, et al. Single-cell landscape 2019;27:708-718 e710.
of immunological responses in patients with COVID-19. Nat 171. Anzalone AV, Randolph PB, Davis JR, et al. Search-and-replace
Immunol. 2020;21:1107-1118. genome editing without double-strand breaks or donor DNA.
160. Delorey TM, Ziegler CGK, Heimberg G et al. COVID-19 tis- Nature. 2019;576:149-157.
sue atlases reveal SARS-CoV-2 pathology and cellular tar- 172. Gawad C, Koh W, Quake SR. Single-cell genome sequencing:
gets. Nature 2021;595:107–113. https://doi.org/10.1038/s41586- current state of the science. Nat Rev Genet. 2016;17:175-188.
021-03570-8 173. Zhang C, Zhang C, Chen S, et al. A single cell level based
161. Cusanovich DA, Hill AJ, Aghamirzaie D, et al. A single- method for copy number variation analysis by low coverage
cell atlas of in vivo mammalian chromatin accessibility. Cell. massively parallel sequencing. PLoS One. 2013;8:e54236.
2018;174:1309-1324.e18. 174. Clark SJ, Lee HJ, Smallwood SA, et al. Single-cell epigenomics:
162. Cusanovich DA, Reddington JP, Garfield DA, et al. The cis- powerful new methods for understanding gene regulation and
regulatory dynamics of embryonic development at single-cell cell identity. Genome Biol. 2016;17:72.
resolution. Nature. 2018;555:538-542. 175. Shahi P, Kim SC, Haliburton JR, et al. Abseq: ultrahigh-
163. Fincher CT, Wurtzel O, De Hoog T, et al. Cell type transcrip- throughput single cell protein profiling with droplet microflu-
tome atlas for the planarian Schmidtea mediterranea. Science. idic barcoding. Sci Rep. 2017;7:44447.
2018;360:eaaq1736. 176. Duncan KD, Fyrestam J, Lanekoff I. Advances in mass spec-
164. Han X, Zhou Z, Fei L, et al. Construction of a human cell land- trometry based single-cell metabolomics. Analyst. 2019;144:782-
scape at single-cell level. Nature. 2020;581:303-309. 793.
165. Lau X, Munusamy P, Ng MJ, et al. Single-cell RNA sequencing 177. Perkel JM. Single-cell analysis enters the multiomics age.
of the cynomolgus macaque testis reveals conserved transcrip- Nature. 2021;595:614-616.
tional profiles during mammalian spermatogenesis. Dev Cell.
2020;54:548-566.e7.
166. Wagner DE, Weinreb C, Collins ZM, et al. Single-cell map-
S U P P O RT I N G I N F O R M AT I O N
ping of gene expression landscapes and lineage in the zebrafish
Additional supporting information may be found in the
embryo. Science. 2018;360:981-987.
167. Replogle JM, Norman TM, Xu A, et al. Combinatorial single- online version of the article at the publisher’s website.
cell CRISPR screens by direct guide RNA capture and targeted
sequencing. Nat Biotechnol. 2020;38:954-961.
168. Zafar H, Lin C, Bar-Joseph Z. Single-cell lineage tracing by inte-
grating CRISPR-Cas9 mutations with transcriptomic data. Nat How to cite this article: Jovic D, Liang X, Zeng
Commun. 2020;11:3055. H, Lin L, Xu F, Luo Y. Single-cell RNA sequencing
169. Alda-Catalinas C, Bredikhin D, Hernando-Herraez I, et al. A technologies and applications: A brief overview.
single-cell transcriptomics CRISPR-activation screen identifies Clin Transl Med. 2022;12:e694.
epigenetic regulators of the zygotic genome activation program. https://doi.org/10.1002/ctm2.694
Cell Syst. 2020;11:25-41 e29.

Single Cell Technologies

Uploaded by

Copyright:

Available Formats

Single Cell Technologies

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Single Cell Technologies

Uploaded by

Copyright:

Available Formats

Received: 5 August 2021 Revised: 9 December 2021 Accepted: 20 December 2021

Single-cell RNA sequencing technologies and applications:

1 Lars Bolund Institute of Regenerative

Medicine, Qingdao-Europe Advanced

1. This review provides a concise summary of the single-cell RNA sequencing

Clin. Transl. Med. 2022;12:e694. wileyonlinelibrary.com/journal/ctm2

Single-cell RNA sequencing technologies and applications:

1Lars Bolund Institute of Regenerative

Clin. Transl. Med. 2022;12:e694. wileyonlinelibrary.com/journal/ctm2 1 of 20

Isolation Amplification Published

REFERENCES 21. Koenitzer JR, Wu H, Atkinson JJ, et al. Single nucleus

You might also like