Single-cell omics in tracing cellular heterogeneity of drug-induced liver injury: Technological landscape and prospective application

Abstract

Drug-induced liver injury (DILI) remains a serious problem in clinics for both diagnoses and treatment decisions. It is a result of accumulated drugs in human bodies metabolized into toxic constituents generating reactive metabolites, and then arise initial consequences of oxidative stress, organelle stress responses, and lethal consequences (liver necrosis or apoptosis). However, the idiosyncratic nature of DILI complexes its mechanistic studies and still little is known of its potential etiopathogenesis for certain. Single-cell omics technology and approaches serve as powerful tools for investigating cellular heterogeneity and relationships from measurements of up to millions of individual cells at an unprecedented resolution, which are achieved by advances in genome, epigenome, transcriptome, proteome, and metabolism technologies. As liver contains heterogeneous cell types of distinct spatial, molecular, and functional properties, they interact with each other to precede cell type-specific omics reprogramming and play an irreplaceable role in liver cells with heterogeneous properties upon encountering toxic insults. Single-cell omics, especially single-cell transcriptomics and single-cell proteomics, have been utilized for exploring the mechanisms of DILI and prediction for risk factors. In this review, we discuss the recent development and future perspectives of single-cell omics-based technologies for DILI-related research.

Graphical abstract

Highlights

• •Single-cell omics serve as powerful tools for investigating cellular heterogeneity in drug-induced liver injury (DILI).
• •Single-cell omics are essential for mechanism investigation and risk prediction in DILI.
• •Single-cell omics facilitate the precise diagnoses and clinical interventions for DILI.
• •The integration of multi-single omics has potentials in identifying cellular phenotypes for DILI.

Overview of Drug-induced Liver Injury

Drug-induced liver injury (DILI) is an uncommon but serious matter with consequence on both patients’ diagnosis and pharmaceutical regulatory decisions, although drugs undergo extensive in vitro and in vivo test systems prior to clinical development. It is induced by drugs, their metabolites, or by the reduced supersensitivity or tolerance of special constitutions to drugs, presenting as the leading cause of acute liver failure and major factors for liver transplantation [1]. Its incidence in general populations is around 14–19 per 100,000 inhabitants [2]. Over 1300 “classic” medicines, including medicinal plants, food supplements, and chemicals, may cause severe DILI. Clinical manifestations of DILI are diverse and complicated, varying extensively from sufferer with viral hepatitis-like and/or cholestatic syndrome to those who are asymptomatic with normal hepatobiliary imaging, among which acute hepatitis is up to now the most common manifestation that takes up more than 90%. Some patients present fluctuating biochemical indicators from normal level, and some even had multiple episodes in the follow-up without intervention of drug re-challenging and drug recurrence [3,4]. DILI is typically classified as intrinsic DILI (In DILI) and idiosyncratic DILI (IDILI), reflecting the distinctly dominant role of drug toxicity and host factors in liver injury. In DILI is generally dose-dependent occurring in a short incubation period (hours to days) upon a threshold dose is reached, featured with predictable toxicity and individual variation of thresholds [5]. The most well-known In DILI model is the hepatotoxicity of paracetamol [6]. While the incidence of IDILI is unpredictable within therapeutic doses in recommended situation of prescription that usually occurs in less than 1 of every 10,000 exposed individuals. It owns a longer latency time span (days to weeks) and is unexpected from the pharmacological action of agents and affected by the orchestration of causative agents (surpassing a threshold dose, physicochemical characteristics, metabolites formation, oxidative stress, and hepatic transporters inhibition), host-related conditions (individual conditions, toxicological responses, inflammation and immune responses, and imbalance between tissue injury and repair), and environmental factors (diets, toxins, exposure to tobacco, alcohol, pollutants, and oxidants) [7,8]. Traditionally, serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), and total bilirubin (TBiL) were used for early detecting and monitoring DILI, which reflect abnormities in normal liver function, as well as tissue and cell integrity [9]. However, these traditional biomarkers are not specific for distinguishing DILI from other etiologies of liver injury, and may also serve as the typical of muscle and cardiac damages. Currently, diagnostic biomarkers for assessing DILI include causality score using Roussel Uclaf Causality Assessment Method (RUCAM), determination of drug or one of these metabolites, as well as serum detection of reactive adducts and autoantibodies [10]. Prompt discontinuation of implicated agent remains the most important step to decrease risk of further liver injury. There exist two treatment approaches that have been endorsed to manage DILI patients, of which the one is through rapid depuration of toxic drugs for hampering its further absorption, and the other is administration of a drug-specific antidote (N-Acetylcysteine, corticosteroids, etc.) for countering hepatotoxic effects. Besides, close observations are necessary for DILI patients with repeated measurements of liver tests [11].

The pathogenesis of these two types of DILI share similarities and differences, and is so far scarcely clarified. A common prerequisite for In DILI and IDILI is the biotransformation of lipophilic drugs in the liver, generating reactive metabolites, and then arise initial consequences, such as oxidative stress and organelle stress responses, thereafter resulting in either lethal consequences (liver necrosis or apoptosis) or adaptive immune responses in genetically susceptible individuals. Intrinsic hepatotoxins such as acetaminophen are dose-dependent, while IDILI mostly appearing in clinics are not closely associated with drug features [12]. Current mechanistic investigations of DILI mainly focus on the roles of reactive metabolites, mitochondrial toxicity, immune system and hepatobiliary transporters, which can be concluded as two folds in drug/metabolite exposure to a threshold level, and in adaptive immune response [13]. Intrahepatic generation of neoantigens from drugs or corresponding metabolites evoke the recruitment and activation of T lymphocytes for triggering innate and adaptive responses, and the following tissue inflammation and injury for those with an insufficient adaptive response. Especially, the onset of an immunologic response to the metabolite appeared on the injured hepatocyte surface to a specific restricted human leukocyte antigen (HLA) T cell receptor, and then immune recognition triggers without adaptation [14]. Alternatively, non-immune mechanisms lie in the release of damage associated molecular pathway (DAMP's) proteins from necrotic cells and autophagy activation, resulting in the recruitment of localized tissue injury and disruption of lysosomal functions, during which various cytokines and co-stimulatory molecules participate in macrophage activation and DAMP response [15,16] (Figure 1). Abilities to repair tissue damage will further distinguish clinically DILI from mild liver injury.

Figure 1.

General mechanisms of DILI

Intrahepatic generation of neoantigens from drugs or the corresponding metabolites evoke the recruitment and activation of Tlymphocytes for triggering innate and adaptive responses, and the following tissue inflammation and injury for those with an insufficient adaptive response. The onset of an immunologic response to the metabolites appeared on the injured hepatocyte surface to a specific restricted HLA T cell receptor, and then immune recognition triggers without adaptation. Non-immune mechanisms lie in the release of DAMP's proteins from necrotic cells and autophagy activation, resulting in the recruitment of localized tissue injury and disruption of lysosomal functions. Abbreviations: GSH, glutathione; ROS, reactive oxygen species; ATP, adenosine triphosphate; DILI, drug-induced liver injury; HLA, human leukocyte antigen; DAMP, damage associated molecular pathway.

It has been well recognized that the occurrence of DILI demands collaborative investigations and phenotype standardization. Recent progress on the research for DILI is set on three aspects: new mechanistic concepts of DILI that emphasize the role of unspecific “downstream” bioprocess instead of initial “upstream” hepatocyte events, as well as complex interplays of environmental factors and genetic risks that regulate aggressive and protective processes imbalance; newly developed high-throughput genotyping techniques for identifying genetic susceptibility markers and improving suboptimal performance of DILI prediction; multidisciplinary approaches based on various “-omics” to screen novel biomarkers for the prediction and diagnosis of DILI [17,18]. Ongoing of these studies will shed light into its mechanism-based treatments.

Technological Landscape of Single-cell Omics

Regulation of gene expression is executed within or between individual cells, which are the smallest functional and universal unit of organisms. Cell diversity in human body is complex due to sophisticated cell bioprocess, such as proliferation, differentiation and death, and even more complicated when considering their resided environment [19]. Traditionally, distinct cell types are defined by cellular morphology, location or its expressed proteins, while omitting distinction in molecular layers of cells, as well as cell unique niche around and their orchestration with adjacent or distant cells [20]. Heterogeneity of cells presents functional diversity in human biology [21]. Conventional bulk sequencing technologies severely ignore cell diversity by only studying the average of gene expression levels for all cells using biological samples, precluding cell-to-cell variability. Therefore, it is impressive to detect human biology at single-cell biology. Continuous advancement in high-throughput single-cell omics technologies of multiplexing, throughput, resolution and accuracy have accelerated the investigation of cellular heterogeneity and relationships from measurements of up to millions of individual cells at an unprecedented resolution, providing a desirable method for interrogation of complex disease biology [22].

Numerous single-cell omics sequencing technologies (single-cell surface functionalization, intracellular electrophysiology, single-cell isolation, and single-cell sequencing) have been applied to the corresponding tissues or peripheral blood in characterizing cellular properties at multi-molecular layers, including genome, epigenome, transcriptome, proteome, and metabolism (the summarization of these single-cell omics technologies is provided in Table 1), each pertaining to special portion of the central dogma of DNA–RNA–protein. Single-cell single modality methods make quantification of genomic, exomic, and epigenomic (DNA methylation, chromatin accessibility, and histone modifications) message from DNA, transcriptomic message from RNA, and proteomic, phosphoproteomic, and metabolomic information from proteins [23]. Contemporary flowcharts for sequencing are dissociating tissue into cells, labelling the content of each cell by cellular barcoding, modalities of sequencing, preparing sequencing library and sequencing, recording additional cellular metadata, analyzing data and visualizing. Up to now, a vast amount of single-cell omics data has been generated, such as the Human Cell Atlas, the Mouse Cell Atlas, the Mouse RNA Atlas, etc., serving not only as basic sources for thousands of bioinformatics algorithms development, but opening new avenues for cellular function related gene-regulatory and signaling networks construction when combining with computational analytical approaches. These advances lead to significant insights in the fields of oncology and immunity, assisted reproduction, embryonic development, and plant breeding [24]. Due to its enormous potential, single-cell omics sequencing has twice been chosen as the Method of the Year by Nat Methods [25,26]. Interestingly, this technology has been applied in enhancing the detection sensitivity of potential toxic effects and heterogeneous responses threshold in DILI at different cell masses, and many achievements have been obtained [27].

Table 1.

A brief introduction of single-cell omics-based technologies at multi-molecular layers, including genome, epigenome, transcriptome, proteome, and metabolism

Omics technology	Definition	Advantage	Disadvantage	Basic steps	Application	References
Single-cell genomics	Single-cell genomics is a method to detect gene expression in single cells, study cell heterogeneity and identify its clinical relevance. In a word, what can cells do-reflect the genotype.	Allow to identify the diversity and complexity of cells in the sample and can analyze the information of multi-cell or a large number of tissue samples to describe the type and state of cells, find unknown genes and therapeutic targets.	The amplification technology is challenging, and the cell separation technology has low throughput, easy allele loss, false positive and false negative phenomena.	Sample pretreatment → cell separation → deoxynucleotide extraction → genome amplification → library construction → genome sequencing → data analysis	Tumor, microbiology, embryonic development, and immunology	[[28], [29], [30], [31], [32]]
Single-cell epigenomics	Single-cell epigenomics aims to understand the epigenomic heterogeneity among different cells by measuring and analyzing the epigenome of a single cell. In a word, what can cells do-reflect the apparent modification.	Capture the DNA methylation, histone modifications, DNA-protein interactions, transcription factor binding, chromatin state, and three-dimensional (3D) genome architecture. Provide cell type-specific gene regulation information, which is important to discover gene regulation mechanisms.	The operation is complex, and the technology's throughput is low because only a small number of cells can be biopsied or circulated by capturing a small amount of cell DNA.	Preparation of single cell suspension → cell fragmentation → nuclear extraction → single cell nucleus sorting → library construction → computer sequencing → data analysis	Tumor, stem cell differentiation, botany, and embryonic development	[[33], [34], [35], [36]]
Single-cell transcriptomics	Single-cell transcriptomics is a technique for measuring mRNA, a product of gene expression in single cells. In a word, what can cells do-reflect the genotype.	Reveal gene expression at the single cell level, help discover new cell types and understand the mechanism of expression regulation during cell growth.	The analysis is expensive and complicated, and most of them pay attention to mRNA but ignore other RNA species.	Sample pretreatment → separating and capturing single cell → lysing cells → extracting ribonucleic acid → preparing complementary deoxyribonucleic acid → polymerase chain reaction amplification → sequencing → data processing	Stem cell differentiation, tumor, and early embryonic development	[28,37]
Spatial transcriptomics	Spatial transcriptome technology aims to quantitatively measure the gene expression of cells and provide specific location information of cells in tissue space. In a word, what can cells do-reflect the phenotype and location information.	Combine micro-imaging and sequencing technology to obtain gene expression data while preserve the spatial position information of samples. It is helpful to discover the information hidden by spatial structure.	The resolution and sensitivity are low, and it is impossible to achieve the resolution of a single cell.	Sample preparation → chip preparation → sequencing → data analysis	Developmental biology, immunology, oncology, and neuroscience	[38,39]
Single-cell proteomics	Single-cell proteomics refers to explore the expression and interaction of protein group at the level of single cell, realize the qualitative and quantitative analysis of protein, and understand how gene products interact to produce cell phenotype. In a word, what can cells do-reflect the phenotype.	Explore the expression level and interaction of proteins between cells, and understand how genes affect cell phenotype. Avoid the change in expression level caused by post-transcriptional regulation and post-translational regulation, which is closer to the regulatory essence of cells. Also, the sample preparation process is simpler.	The modification process is more difficult, and the content of protein in single cell is low. Also, the dynamic range is wide, and it is easy to omit the low-abundance proteins.	Sample separation → interference removal → protein extraction → treatment and digestion → mass spectrometry → data processing	Tumor, immunity, embryonic development, stem cell differentiation, and neurobiology	[28,[40], [41], [42]]
Single-cell metabolomics	Single-cell metabolomics is the study of small molecular metabolites with a relative molecular mass < 1000 and related to cell phenotype. In a word, what can cells do-reflect the phenotype.	Reflect the influence of environmental changes on cell phenotype and provide a sensitive dynamic picture for understanding cell function.	The absolute metabolite concentration is low, and the metabolite cannot be amplified. Also, the metabolite transformation speed is fast, and the metabolite concentration difference between cells is large.	Cell separation → sample processing → coupling analysis of single cell technology → data screening → statistical analysis → identification of metabolites → analysis of intracellular cell species or differential metabolites	Botany, neurobiology, and tumor	[28,[43], [44], [45]]

Single-cell Genomics

Cells are the basic unit of life, and DNAs serve as the heritable programme encoding protein formation that executes physiological functions and promotes the development of modern genetics and genomics. Inter-cellular heterogeneity within homogeneous cell populations has been regarded as the main reason for functional variation across different biospecimens. Of primary importance to human geneticists is to identify genes, pathways and corresponding mechanisms for disease phenotypes, detect disease occurrence, as well as monitoring its progression or response to treatments that may be confined to a specific subpopulation of cells [46]. Over the past decades, single-cell genomics, in combination with gene-editing technologies, has rapidly proved its value for studying human biology through determining gene expression at single-cell level that assesses the frequency and strength of transcriptional bursts [47]. It enormously increases the throughput and precision for detecting cell-to-cell variabilities within an organism or tissue, and makes it now possible to screen numerous genomic regions in a single experiment. Genomic DNA sequences are on behalf of genetic information of an organism. While diploid human cells own only two copies of genomic DNA, necessitating amplification prior to microarray- or next-generation sequencing assay. Several approaches, such as multiple annealing- and looping-based amplification cycles (MALBAC) [48], multiple displacement amplification (MDA) [49], linear amplification via transposon insertion (LIANTI) [50], and polymerase chain reaction (PCR), have been established for refining the uniformity, efficiency, precision, and coverage of the whole genome amplification at single-cell resolution. Plenty of genetic variants of different types, including single-nucleotide variations (SNVs), copy-number variations (CNVs), small indels, and structural variations (SVs), are often rare and unique, but can also be detected faithfully to present as intrinsic markers tracing the historical trajectory of single cell in human body [51]. Moreover, mitochondrial DNA (mtDNA), as another part of cellular genome, exerts 10-100-folder higher mutations scale than nuclear DNA, which can be analyzed via single-cell genomics for tracking cellular clones [52]. This technology has been successfully applied in compartmentalizing microbial dark matter, and identifying genetic mosaicism in multicellular organisms, especially for cancers in terms of studying tumor heterogeneity, analyzing tumor evolution and detecting drug resistance mechanisms, for the reason that cancer belongs to single-cell disease that caused by genome abnormity [53].

Developments in genomic technologies also drive the identification and elaboration of pathophysiological processes of liver diseases, including viral hepatitis, fatty liver disease, liver cirrhosis, and DILI [54]. In terms of etiology, DILI converges on the features of immune-mediated liver injury, and genetic background influences the risk of DILI hypersensitivity [55]. Genome-wide association studies for amoxicillin-clavulanate revealed a strong effect of human leukocyte antigen (HLA) class I and II genes in major histocompatibility complex regions that affect susceptibility to DILI, indicating an essential role of adaptive immune response in disease progression, as well as pharmacological interaction between drugs and immune receptors [56]. Another study has identified genetic variations for anti-tuberculosis DILI in the genomic regions within or near genes encoding proteins involved in drug-metabolizing enzymes, bile acids accumulation, immune adaptation, oxidant challenges, etc., contributing to design genetic tests and also act as biomarkers for predicting drug mediated DILI risks in personalized medicine [57]. Single-cell genomics of liver tissues or peripheral blood has not been applied in the research of DILI yet due to its novelty, although the onset of DILI is closely associated with the genomic DNA alterations or mutations of cell-to-cell viability. This technology can be further used in DILI via detecting drugs containing harmful compounds, identifying endogenous substance, and determining genetic susceptibility, thus serving as an efficient tool for providing important insights into the pathogenesis of DILI and pointing to new disease treatments (Figure 2).

Figure 2.

The technological landscape of single-cell genomics and its prospective application in tracing cellular heterogeneity of DILI

Abbreviations: PCR, polymerase chain reaction; MDA, multiplex displacement amplification; MALBAC, multiple annealing and looping-based amplification cycles; HLA, human leukocyte antigen; DILI, drug-induced liver injury.

Single-cell Epigenome

Epigenetic regulation stands in the core of gene regulation network, participating in gene transcription and differentiated progenies of stem cells. Single-cell epigenome aims to understand the epigenomic heterogeneity among different cells by measuring and analyzing the epigenome of a single cell, which mainly includes epigenetic regulation such as DNA methylation, histone modifications, DNA-protein interactions, transcription factor binding, chromatin state, and three-dimensional (3D) genome architecture [58]. These modifications affect gene expression and transmission of genetic information, while they do not alter DNA sequence itself. DNA methylation, mainly occurring in mammalian genome, is the covalent addition of a methyl group to the 5′ carbon of cytosine. It enormously influences genome imprinting, X-chromosome inactivation, repetitive-element silencing, transposon and retroviral repression, and response to external stress [59,60], and performs mainly based on reduced representation bisulfite sequencing (RRBS) and post-bisulfite adaptor tagging (PBAT) bisulfite sequencing strategies [61]. Histone modifications are post-translational modifications that occur at histone loci and include more than a dozen chemical modifications such as acetylation, phosphorylation, methylation, and ubiquitination [62]. Chromatin immunoprecipitation (ChIP) is commonly used to map histone modifications, as well as to detect transcription factor binding sites and DNA-protein interactions that result in non-specific antibody binding. Chromatin state has the potential to dynamically regulate gene expression, and open chromatin state can be assessed by the accessibility of enzymes [63]. Chromatin state captures information on chromatin open regions, while 3D genome structure provides message on interactions between entire genomes in a cell and even that on spatial and structural organization. The latest single-cell Hi-C (scHi-C) method can detect approximately 1.9 million contacts in a single cell with high sensitivity [64]. This technology will help to reveal the epigenome and epigenetic heterogeneity of rare cells, and has been broadly applied in studying the highly heterogeneous nature of tumor cells at a single-cell level [65].

The pathogenesis of DILI has been suggested to be involved in genetic polymorphism and epigenetic modifications, which can be responsible for modulating toxicity degree through strengthening pro-toxicant pathways of either impairing defense mechanism or weakening immune tolerance. Such epigenetic alterations can occur due to indirect responses in DILI, and these genes might be epigenetically silenced by the trimethylation of lysine 27 of histone 3 and methylation of cytosine residues in CpG dinucleotides in gene regulatory regions [66]. A study also revealed an epigenetic memory in the liver enriched transcription factor HNF4α promoter region in reprogrammed cells derived from human hepatocytes [67]. At present, little is known about epigenetic characteristics of DILI, and further researches using single-cell epigenomic sequencing based on liver tissues or peripheral blood are needed to explore chromatin changes and epigenetic heterogeneity in cellular reprogramming during DILI (Figure 3).

Figure 3.

The technological landscape of single-cell epigenomics and its prospective application in tracing cellular heterogeneity of DILI

Abbreviation: DILI, drug-induced liver injury.

Single-cell Transcriptomics

Tremendous advances have been made in the throughput, accuracy, automation, and commercialization of single-cell transcriptomics, promoting it to be the most widely applied single-cell omics in the fields of discovering new and rare cell types, detecting cellular heterogeneity within tissues and cellular states, as well exploring potential biological mechanisms with low cost and high accessibility. Single-cell RNA sequencing measures expression levels of thousands of genes simultaneously, and can deal with the stochastic nature of gene expression [68]. It can be mainly divided into two categories of microfluidics droplet-based transcriptomics and microwell plate-based transcriptomics, of which the former prefers to detect rare cell types with high throughput properties, and the latter tends to capture full-length messenger RNA (mRNA) and detect more genes, transcripts with low abundance and alternative splicing events [69,70]. The most commercially available platform is the chromium controller from 10× genomics of isolating single cells or nuclei in droplets via emulsion, and mRNA capture, reverse transcription, and molecules barcoding proceed [71]. Integration of single-omics modalities for transcriptome with genome [72], epigenome [73], and proteome [74] not only preserves the progress in highlighting transcriptomic information, but simultaneously enhances the throughput by cell multiplexing, cell tagging, etc. Meaningfully, the value of high-resolution transcriptomic data of individual cells has endowed profound knowledge of human tissues for Human Cell Atlas (HCA) global consortium [75]. Currently, single-cell transcriptomics has been applied across a breadth of medical areas, such as in vitro cellular models, disease therapies and mechanism detection [76].

This technology has been applied in researching the etiology of liver homeostasis, progression, regeneration, chronic or accurate liver disease, hepatocellular carcinoma, and even describes the cellular landscape of liver, offering insights into the transcriptomic architecture and stability of liver organ with high value [77]. It benefits discovery of uncharacterized cell types and states in the liver, observation of intercellular communication and signaling in liver microenvironment, dissecting of cellular heterogeneity and molecular dynamics, and detection of gene regulatory mechanisms that are involved in tissue regeneration following liver injury [78]. As liver is the most important detoxification organ, while how these substances alter cell composition and intercellular communication within liver microenvironment, as well affect specific cell types (hepatocytes, sinusoidal endothelia, and intrahepatic biliary epithelial cells) and corresponding genes haven't been clarified. Fortunately, the single-cell RNA sequencing makes it come true. For example, Wu et al. [79] use single-cell RNA sequencing based on mouse liver tissues and immunofluorescent staining to generate the first RNA related atlas of glyphosate-exposed mouse liver, revealing cell-to-cell viability of activation and upregulated phagocytosis in macrophages, and proliferation and extracellular matrix over excretion in hepatic stellate cells, as well contribute to detect the acting mechanism of glyphosate induced hepatotoxicity. They also employ this technology to dissect the changes in liver gene expression and related signaling pathways after tripterygium glycosides tablet administration in mouse models, and a total of fifteen cells in mouse liver tissues are identified, such as endothelial cells, hepatocytes, hepatic stellate cells, etc., of which inflammatory response in endothelial cells, inflammatory reaction, apoptosis, and fatty acid metabolism dysfunction in hepatocytes, phagocytosis in liver capsular macrophages cells are observed, offering references for further identification of therapeutic targets for alleviating drug toxicity [80]. Shen et al. [81] detected gene expression profiles of hepatocytes and non-parenchymal cells in bavachin treated and normal livers using single-cell RNA sequencing and discovered three types of hepatocytes subsets in response to bavachin, all of which suffer from bursts in lipid metabolism. Scd1 overexpression in these hepatocytes facilitated monounsaturated fatty acids synthesis and led to lipid accumulation and hepatocyte steatosis, meanwhile tenascin signals from immune cells modulated downstream pathways after receiving bavachin, suggesting the potential of Scd1 serving as a hub molecule for predicting early liver injury in clinics. In DILI, auto protection occurs upon body manifests as resistance to re-exposure to the same toxicant, however, its key cell populations and corresponding biological functions during protective adaption remain unclear. Yu et al. [82] utilized single-cell RNA sequencing of mouse liver tissues to explore the dynamics and functions of hepatic non-parenchymal cells in auto-protection against DILI, and found out that the proportion of identified cells, such as T cells, B cells, dendritic cells, macrophages, neutrophils, and natural killer cells, varied significantly between acetaminophen and its resistant groups. Especially, dendritic cells subpopulations facilitate auto protection against acetaminophen hepatotoxicity by inhibiting autoimmune response and promoting angiogenesis, natural killer cells by regulating interferon-γ(IFN-γ) pathways, and T cells via amplifying IFN-γ response, highlighting the key dominant role of non-parenchymal cells in resistance to hepatotoxicity. Liver organ carries an enormous ability to regenerate and preprogramme promptly for taking over the pericentral hepatocyte functions after acute zonal damage, during which multiple cell types exhibit both zone-independent and zone-specific gene expression reactions.

Although single-cell transcriptomics entails targeted organs and tissues dissociating into single-cell suspensions, it leads to the complete omission of spatial information and additional risk of undesirable transcriptional alterations. Recent emerging spatial transcriptome technology, first developed by Stahl et al. [135], mainly contains two types of technologies that include single-molecule fluorescence in situ hybridization (FISH)-based ones (sequential FISH and multiplexed error-robust FISH) using combinatorial barcodes for single-molecule FISH, and in situ sequencing-based ones (in situ sequencing, fluorescence in situ sequencing, spatially resolved transcript amplicon readout mapping, and expansion sequencing) using rolling cycle amplification for signal amplification before in situ sequencing analysis. Both of these methods enable imaging mRNAs for up to 10,000 genes in a single cell with high spatial resolution [83]. It tends to elucidate the cell type composition of a population in a defined domain of tissue or organ, characterized with gene expression profiles whilst retaining message on spatial context of expression, which significantly impairs the influence of environmental conditions and interactive signals on cells at certain locations [76], and therefore plays a vital role in comprehending hepatic structure and function in both homeostasis and disease states due to spatial properties of liver organ. Ben-Moshe et al. [84] applied spatially resolved single-cell RNA sequencing of liver tissues to explore the spatial and temporal dynamics of orchestrated response of liver cells in the entire zonal regeneration bioprocess after acute acetaminophen intoxication. Hepatocytes are revealed to perform continuous proliferation, and contribute to repopulating the necrotic pericentral zone rapidly. Hepatocytes that are located at regenerating front modulate reprogram-related fetal specific genes. While hepatic stellate cell and macrophages participate in immune recruitment, proliferation, and matrix initiation, endothelial cells exist in zonated proliferation patterns, which are more robust to toxic microenvironment. These results highlight the details of liver plasticity during zonal regeneration and present the zone-dependent coordinated programs that benefit accurate healing of liver tissues. In clinics, hepatotoxicity is typically not identified until clinical trials, endowing great financial burden in drug development, and increasing safety risks for clinical trial participants. To overcome this problem, researchers invent the human liver organoids (HLOs) as physiologically organotypic for high-throughput DILI risk prediction. HLOs derived from induced-pluripotent stem cells respond to known DILI-causing drugs, and integrate with multi-omics platform such as biomarker detection, high-content imaging enabled phenotyping, and single-cell RNA sequencing to dissect DILI with scalability and consistency [85].

In the current, the most widely used spatial transcriptome technology still cannot reach single-cell resolution, although preserving spatial information in situ. Some researchers prefer to integrate single-cell and spatial transcriptome analysis to examine cellular heterogeneity, thus offering a deeper understanding of the function and interaction of liver cells under both physiological and pathological conditions, as well as opening new opportunities to investigate fundamental questions in liver development, regeneration, and disease, respectively [86]. For instance, Uzun et al. [87] applied whole-transcriptome and spatial transcriptome sequencing based on liver biopsy samples from patients with the coronavirus disease 2019 (COVID-19) vaccine-induced liver injury, and those with autoimmune hepatitis, conjugated with histomorphology evaluation and multiplex immunofluorescence, to reveal their shared similarities in immune response, while the former shows varied differences in histomorphology, pathway activation, cellular immune infiltrates, and T-cell receptor usage, suggesting its distinct disease entity. Besides, this combined technology has been used for other liver related diseases, such as detecting genetic and epigenetic changes in human liver cancer stem-like cells [88], identifying tumor immune barrier in the hepatocellular carcinoma microenvironment that determines the efficacy of immunotherapy [89], revealing fetal liver hematopoietic stem cell niche architecture [90], etc.

Although these transcriptomics techniques have been proven useful in experimental design ranging from atlas generation, chemical compounds optimization, and mechanistic research in DILI, limitations of mRNA capture efficiency, number of omics data, sample integrity, and experimental objectives quality are still unmet to be improved in further investigation (Figure 4).

Figure 4.

The technological landscape of single-cell transcriptomics/spatial transcriptomics and their prospective application in tracing cellular heterogeneity of DILI

Abbreviations: COVID-19, the coronavirus disease 2019; VILI, COVID-19 vaccine-induced liver injury; AIH, autoimmune hepatitis; ECM, extracellular matrix; HSC, hepatic stellate cells; DILI, drug-induced liver injury.

Single-cell Proteomics

Most single-cell studies shed spotlights on nucleic acids, especially transcriptome that measure expressed genes in a cell, while proteins are indeed recognized as “worker bees” for cell, for its encoded by the mRNA instead of mRNA itself, and act as the effector molecules of cellular processes. Single-cell proteomics has been an area of study for over 30 years, and provides an opportunity to conduct casual protein inference and a platform of direct and indirect protein interactions, which fill in gaps in omitting heterogeneity for global mass spectrometry (MS), as well as refined specificity and a small number of preselected proteins per cell for antibody-based approaches [91]. This technology is mainly based on the application of fluorescence flow cytometry or fluorescence-activated cell sorting to characterize cellular phenotypes at an unprecedented scale. Indicators associated with light scatter discriminate cell types via size, and others associated with distinct membrane proteins that are marked with fluorescent antibodies show a phenotypic classification of cell types. Recent developed methods of mass cytometry technique optimize affinity with adherent surface, chemical derivatization, and manipulation of ions promote this analysis at a depth of 1000–1500 proteins and 1500–2500 proteins across all cells, especially for differentiating stem cells, monocytes, and cardiomyocytes [92]. It conjugates with antibodies to enable deeper proteomic analysis of cell phenotype at single-cell level with minimal protein loss, and identification and quantification of peptides [93]. There are two particularly promising MS platforms of the nanodroplet processing in one pot for trace samples platform (NanoPOTS) and the single-cell ProtEomics by MS platform (SCoPE-MS). NanoPOTS demands MS-ready samples at a volume less than 200 nL to minimize unnecessary protein loss, and allow the identification of 1500–3000 proteins from 10 to 140 cells when combined with ultrasensitive liquid chromatography-MS [94]. SCoPE-MS incorporates tandem mass tags for reforming the efficiency of data throughput and single-cell peptide quantification measuring over 3000 proteins from more than 1000 individual monocytes and macrophages [95]. As the bioprocess of cells functions around proteins, it is essential to study a comprehensive protein profile in the field of islet cell subpopulations. In practical application, protein and mRNA omics studies are co-profiled to confer single-cell phenotypes through leveraging DNA polymerase activity of reverse transcriptase to conduct proximity extension and complementary DNA synthesis in the same reaction [96].

Although invasive liver biopsy is considered to be the gold standard method for diagnosing hepatic fibrosis and cirrhosis, it can't represent the whole state of liver and causes intolerable pain and complications in patients. Outpatient monitoring of risks for DILI using non-invasive measurement of biomarkers in blood or urine would help patients monitor DILI risk to defeat the limitations of post-marketing surveillance [97]. Bae et al. [98] conducted a single cell-based, high-dimensional mass investigation of the peripheral blood mononuclear cells from mice of three groups (normal chow, thioacetamide induced fibrosis, and 11β-hydroxysteroid dehydrogenase inhibitor-treated) to explore the pathogenesis of drug induced liver fibrosis using noninvasive biomarkers, and then orchestrate with dimensionality reduction and two-dimensional (2D) visualization techniques, as well automated clustering approaches to qualitative detect major cells distribution changes and cell types/subsets. Results showed a significant alteration in the populations of manually gated B cells, natural killer cells, effector CD4⁺ T cell, and memory CD8⁺ T cells. Clustering analysis reveals that subsets of B cell play an essential role in both disease development and recovery after treatment, while that of natural killer cells dominate in the progression of thioacetamide induced fibrosis and recovery from this disease. This study provides insight into the classification of DILI related cells and reference for the development of a non-invasive, accurate, and reliable method for DILI diagnosis. However, only a few researches applied single-cell proteomics in studying DILI currently, due to the lack of intracellular cytokine markers in the mass cytometry panel, so further functional studies that involve more comprehensive intracellular markers are demanded (Figure 5).

Figure 5.

The technological landscape of single-cell proteomics and its prospective application in tracing cellular heterogeneity of DILI

Abbreviations: CyTOF, cytometry by time-of-flight; HSC, hepatic stellate cells; DILI, drug-induced liver injury.

Single-cell Metabolomics

Metabolites, with varied masses ranging from a few Daltons to several thousand Daltons, belong to precursors, intermediates, and end products of cellular bioprocesses. These metabolites are important for the majority of biological reactions and body functions, accurately reflecting samples' response to genetic or experimental changes. As cells’ metabolites are dynamic and shift promptly in accordance with cellular activity and environmental alterations, it is difficult to predict the composition and dynamic features of cellular metabolome using transcriptomics and proteomics data [99]. Single-cell metabolomics has become the leading fields of identifying phenotypic heterogeneity among individual cells and detecting subpopulations of cells for clarifying basic mechanism of drug resistance, cell differentiation and disease progression. Current challenges in this technology include low picolitre volume of available analyzed samples, rapid metabolism changes, metabolism properties, and incomplete state of current databases [100]. High diversity in abundances and physicochemical parameters of various metabolites in a single cell requires multistage technologies. The two most common methods for measuring untargeted metabolome of samples are based on MS and nuclear magnetic resonance (NMR) spectroscopy, enabling detection of all ionizable metabolites without labelling or preselection. Especially, sample collection strategies in MS can efficiently ionize metabolites in a solid surface that include high-throughput secondary ion mass spectrometry (SIMS), matrix assisted laser desorption (MALDI), and laser ablation electrospray ionization (LAESI), and in liquid phase using electrospray ionization (ESI) [101]. Then, cell handling and metabolite sampling, metabolites quantification, as well as data throughput and analysis are conducted. This technology has successfully turned once theoretical speculation into present-day applications in terms of bacteria, protists, fungi, plants, and animals [102].

Metabolome is regarded to provide a direct readout of dynamic biochemical states of human bodies and has been widely applied to study liver-related diseases of DILI, non-alcoholic fatty liver disease, fibrosis, cirrhosis, hepatocellular carcinoma, etc. [103]. Hepatocyte endo-metabolome tends to be reflected in the liver exo-metabolome. A study using ultraperformance liquid chromatography-MS based metabolomic approaches revealed that 31 metabolites were associated with DILI and alterations of primary bile acid biosynthesis and α-linolenic acid metabolism pathways occurred during disease progression [104]. Metabolomics analysis has been applied in studying the metabolic profiling and mechanism for several DILI, such as emodin [105], extracts of Fallopia multiflora (Thunb.) Harald [106], tripterygium glycosides [107], acetaminophen [108], etc. In addition, metabolomics analysis can also be used for discriminating DILI related phenotypes [109], discovering potential urinary biomarkers and predicting risk profiling [110,111]. However, there is no single-cell metabolomics research in studying DILI up to now, due to incomplete technology of handling in cell and sample, quantification, and data analysis for ensuring cell integrity and thereafter establishing metabolic status of individual cells. Further bioanalytical technologies are necessary for aspects of distinct analyte extraction, separation, and detection for enlarging single-cell metabolome coverage, especially, DILI included to discriminate heterogeneity and cell interactions in metabolites (Figure 6).

Figure 6.

The technological landscape of single-cell metabolomics and its prospective application in tracing cellular heterogeneity of DILI

Abbreviations: GC-MS, gas chromatography and mass spectrometry; UPLC-MS, ultra-performance liquid chromatography and mass spectrometry; MS, mass spectrometry; DILI, drug-induced liver injury.

Concluding Remarks and Perspectives

DILI belongs to unpredictable and potentially serious adverse reactions, involving the liver after administrating western drugs, herbs, and dietary supplements. It is among the most challenging acute or chronic liver conditions by physicians and hepatologists, for limited accurate identifying markers and its diagnosis remains one of exclusion [112]. DILI can be classified into vascular injury, hepatocyte injury, cholestatic injury, and hepatocellular-cholestatic mixed injury according to target cell types [113]. Approximately 23% of DILI patients require hospitalization, and account for 11% of acute liver failure, among which 8% progress to chronic liver injury, thus adding an enormous burden to both patients and society [114]. Multifactorial mechanisms are regarded to underlie DILI, of which the immune system predominantly mediates, and associated molecules such as neoantigens, reactive metabolites, DAMPs, and CD4⁺ helper T cells are involved. Liver cells are capable of reducing antigens transported from blood into intestinal lumen for maintaining homeostasis. DILI causing drugs participate in this biological process through inhibiting hepatic bile acid efflux, and its intracellular accumulation can lead to hepatocyte death once immune tolerance doesn't respond promptly [115]. Although management of discontinuation of suspicious drug, drug treatment (hepatoprotective drugs, anticholestatic drug, immunosuppressants, and specific treatment agents), and compatibility administration with other drugs can exert positive therapeutic effects in DILI, insufficient studies on potential pathogenesis of DILI and lack of its proposed model and clinical outcomes promote researchers to seek deeper insights in the pathophysiological changes and cellular mechanisms of DILI [116,117].

Methodological and technological advancements speed up the simultaneous profiling of genome, epigenome, transcriptome, proteome, metabolomics, and other omics for clarifying genotype-to-phenotype relationships in DILI. For instance, genome-wide association studies reported associations between several human leukocyte antigens (HLA) (HLA-A∗33:01, HLA-B∗57:01, HLA-DRB1∗1501) with DILI, indicating an immune system component to liver injury [118]. The enzyme N-acetyltransferase 2, phase I drug metabolizing enzymes (CYP450s) and drug transporters (particularly the bile salt export pump and the multidrug resistance associated protein 2) have been nominally associated with DILI [119]. Epigenetic modifications of H3K9me3 signal are responsible for the repression of metabolic-associated genes, while H3S10ph vitals in liver regeneration after injury [120]. Hepatic transcriptome analyses revealed DILI involved genes were significantly enriched in inflammation, endoplasmic reticulum stress, mitochondrial dysfunction, and hepatic lipid accumulation, which is consistent with disease etiopathogenesis [121]. MS-based quantitative proteomic approach was used for exploring protein expression levels in serum samples from DILI patients, and found that related proteins are components of inflammation, immune system activation, and hepatotoxicity-specific pathways [122]. Metabolomics allows detecting altered cell metabolism in the course of DILI. Altered lipid metabolism and accumulation, as well as bile acids occurred in DILI, accompanied by the inability to oxidise Acetyl-CoA in the tricarboxylic acid cycle (TCA) together with abnormal excretion of TCA metabolites [123]. Although many achievements have been obtained in the current, while as organisms and tissues are in arrangement of diverse cell types (hepatocytes, hepatic stellate cells, macrophages, endothelial cells, etc.), cell-to-cell variability in genotypic or phenotypic traits are ignored by these conventional analysis methods in DILI. It is desirable to explore complicated biological processes systematically at higher resolution with multi-dimensions at single-cell level.

Since the first report of single-cell transcriptomic sequencing in 2009, this technology has developed rapidly and types of single-cell analyses, including single-cell genomics, transcriptomics, epigenomics, proteomics, and metabolomics, have emerged, with enormous advantages in detecting the variation between individual cells, and scaling significant deviation of individual cells from the population average that induced by fluctuations of gene regulators [124]. It is in line with the cellular characteristic determined by the dynamic states of its 3D chromatin conformation, chromatin remodeling, regulatory element interaction, epigenetic modification, transcriptional regulation, gene expression, proteome, and metabolome, all of which form the foundation for single-cell research to explore the diversity of cell types and subsets [125]. Single-cell omics has widely been used for decoding several liver diseases and contributes to a detailed understanding of different cells in the liver by obtaining gene expression states to disclose the whole functions of liver in both normal and pathological conditions, thus providing reference to specific diagnosis marker identified and accurate treatment strategies in liver associated diseases such as non-alcoholic fatty liver disease, liver fibrosis, hepatocellular carcinoma, DILI, etc. Liver organ contains heterogeneous cell types of distinct spatial, molecular, and functional properties, among which cells interact with each other via a network of ligands and receptors to proceed cell type-specific omics reprogramming in unhealthy conditions [126]. It plays an irreplaceable role in body metabolisms, and cell damage as well as liver function impairment happens to liver cells with heterogeneous properties upon encountering toxic insults [127]. Single-cell omics, especially single-cell transcriptomics and single-cell proteomics, are utilized for exploring the mechanisms of DILI and predicting the risk factors using the corresponding bio-samples collected from various DILI animal models (In DILI: acetaminophen-, carbon tetrachloride-induced liver injury model. IDILI: inflammatory model, Uetrecht-Pohl model, and subclinical mitochondriopathy model) [128]. For instance, single-cell transcriptomics has been performed to discover gene profile alterations of various cells in the liver before and after drug administration (acetaminophen, tripterygium glycosides tablet, glyphosate, bavachin, etc.), and found that DILI related genes are notably involved in immune-inflammatory imbalance, apoptosis, and fatty acid metabolism dysfunction [[79], [80], [81], [82]]. Interestingly, this approach combined with human liver organoids to form a liver-on-chip system with high-throughput characteristics, and therefore provides a potential platform for DILI risk assessment [85]. Single-cell proteomics of the peripheral blood mononuclear cells from biological samples revealed distinct physical functions of B cells, CD4⁺ T cells, memory CD8⁺ T cells, and natural killer cells during DILI, contributing to developing a non-invasive and reliable approach for DILI diagnosis [97]. However, in the current, genomics, epigenomics, and metabolomics sequencing for DILI at single-cell resolution haven't been employed for the following several limitations, such as bias during sample preparation, technical noise, drop-out events, batch effects, sample size, interpatient and intra-patient variation, spatial and temporal state of the cell, etc. [129].

Overall, single-cell omics technologies have changed dramatically from sequencing-based low throughput to high throughput, while its application in DILI is still in infancy and only a few researches have been reported so far. Existing limitations in technology itself, such as unrefined computer analysis tool for data integration, unavailability of different omics layers for DILI standard datasets, unsatisfied temporal and spatial resolution, high sequencing costs, and insufficient coverage of single cells need to be overcome in the next step. It is also challenging for researchers to prepare suspensions directly from fresh tissue to isolate intact single cells with high integrity, purity, quality, and sensitivity, in particular, taking into consideration short-term responses to environmental changes or long-term maintenance of cell fates. Technologies have increased greatly in terms of throughput, automation, and speed, while they haven't met clinical requirements. Vitally, current bio-samples for single-cell omics analysis are mainly obtained from in vivo animal models (liver tissues for DILI) or organoids that exert clinical resemblance and mechanistic resemblance due to experience convenience, while they still cannot present the real pathological changes of human bodies.

Moving forward, the number of related new techniques, data analysis methods, and chances for integrating single-cell omics data with images will continue to grow and mature. The integrated multimodal omics of genomes, transcriptomes, epigenomes, proteomes, and metabolism will help connect multiple aspects of cellular identity in DILI and promoting new cell types or new mechanisms discovery, contributing to elucidate the complexity of biological processes through capturing more informative and complete omics data [24]. For instance, transcriptomes, acting as a functional readout of global transcriptional activity, allow for identifying and separating different types of cells, followed by the organization of genome and transcriptional activity of each gene that is modified by various layers of epigenomes. Then, genome information is applied to constructing lineage relationship. Up to now, assays for detecting transcriptome/genome (linking variation in genotype with their corresponding variation in transcriptional responses) simultaneously, transcriptome/DNA methylome (studying how DNA methylation and histone modification alter with changes in transcription in a single cell) simultaneously, and chromatin accessibility/transcriptome (capturing proteins at different cellular locations and throughputs) simultaneously have been reported, and have enormous potential in applying in DILI in the future to reveal cellular heterogeneity at multiple molecular layers within a population of liver cells and clarify whether this variation is coupled or uncoupled between the captured omics layers [73,[130], [131], [132], [133]]. Spatial localization is important to determine cellular fate. It will be novel perspectives for analyzing etiology and pathogenesis of diseases through exploring spatial multi-omics and spatial dynamics to measure characteristics in the same cell with spatial resolution, such as RNA, DNA, chromatin, protein, and epigenetic modification, therefore proposing significant potentials in identifying context-specific, functionally relevant relationships and their influence on cellular phenotypes for DILI [134].

By integrating properly with other sets of conventional wet-lab experiments for in vivo and in vitro procedures, these improvements in single-cell omics will eventually permit us to understand the pathogenesis of DILI in a more complex dimension and also facilitate its precise diagnosis and interventions.

Acknowledgements

This study was supported by grants from National Natural Science Foundation of China (Nos. 82330124 & 82104467), Beijing Municipal Natural Science Foundation (No. 7212186), Scientific and technological innovation project of China Academy of Chinese Medical Sciences (Nos. CI2021A03807 & CI2021A01508), Fundamental Research Funds for the Central public welfare research institutes (No. ZXKT22019), Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (No. ZYYCXTD-C-202002), Key Laboratory of Beijing for Identification and Safety Evaluation of Chinese Medicine, Institute of Chinese Materia Medica, China Academy of Chinese Medical Sciences, and Fundamental Research Funds for the Central Public Welfare Research Institutes (No. ZXKT22042). No study sponsors are involved in the research process of this project.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Author Contributions

Xia Mao: writing – original draft, conceptualization. Mingzhu Xu: writing – original draft, investigation. Xiangying Yan: visualization. Lin Chen: visualization. Na Lin: writing – review & editing, supervision, funding acquisition, and conceptualization. Yanqiong Zhang: writing – review & editing, supervision, funding acquisition, and conceptualization.