Introduction

The mammalian liver is a complex organ consisting of diverse cell types that perform multiple physiological functions including digestion, synthesis, metabolism, and detoxification. It also acts as an immune organ and plays a crucial role in anti-infection, autoimmune stability, and anti-tumour effects. In particular, the liver has a unique immunological advantage as its parenchymal and nonparenchymal cells can exert immune functions to participate in immunoregulation for maintaining homeostasis. As a fundamental immune organ, the liver receives blood directly through the portal vein draining from the peritoneum, gastrointestinal tract, pancreas, and spleen, while several classic immune sentinel surveillance sites, such as the lymph nodes and spleen, are omitted [1] (Fig. 1A). Under physiological circumstances, the liver is more heavily exposed to microorganisms and endotoxins than other organs or tissues owing to its unique anatomical location. Therefore, the liver usually maintains a tolerogenic immune state to prevent needless immune activation and excessive autoimmune responses [2, 3]. However, liver tolerance also allows viruses, parasites, and other microorganisms to persist chronically and cause long-term damage [4, 5]. The maintenance of the immunologically balanced state in the liver is essential for human health and survival. During acute and chronic infections, the liver transforms from a tolerant state to an active immune state and constitutes an important line of defence against invading microorganisms [6,7,8,9]. The defence mechanism of the liver depends mainly on a complex network of immune cells, including Kupffer cells (KCs), dendritic cells (DCs), neutrophils, natural killer (NK) cells, and B and T lymphocytes [10] (Fig. 1B). Furthermore, non-immune cells such as hepatocytes, choanocytes, hepatic stellate cells (HSCs), and liver sinusoidal endothelial cells (LSECs) contribute to immunity by recruiting proinflammatory immune cells (Fig. 1B) and simultaneously producing inflammatory cytokines such as acute-phase proteins, complement factors, and cell adhesion molecules [11,12,13]. Owing to the crucial role played by the liver in defence against harmful microorganisms, a comprehensive liver cell atlas in both healthy and diseased states is needed for new therapeutic target development and disease intervention improvement.

Fig. 1
figure 1

Anatomical structure, composition, and function of the liver. A Arterial supply to the liver and portal venous drainage. B Anatomical structure and cellular composition of hepatic lobules. C Single-cell RNA sequencing has revealed gradients of gene expression along the three different zones of hepatic lobule in different cell lineages, enabling the differentiation of portal zone and central zone hepatocytes, LSECs, and HSCs. These markers are shown in Figure. D Main functions of the liver. ALB albumin, ADAMTSL2 ADAMTS-like protein 2, CYP2E1 cytochrome P450 2E1, DLL4 Delta-like protein 4, ECAD E-cadherin, EFNB2 ephrin-B2, GSS glutathione synthetase, HSC hepatic stellate cell, ITGB3 integrin β3, KIT mast/stem cell growth factor receptor Kit, LSEC liver sinusoidal endothelial cell, NGFR tumour necrosis factor receptor superfamily member 16, Rspo3 R-spondin-3, THBD thrombomodulin, WNT2 wingless-type MMTV integration site family member 2

Full size image

Conventionally, the liver landscape in a healthy or diseased state is studied using whole-tissue sequencing of a bulk and mixed population of cells within different zonations, which makes it difficult to interpret the uniqueness of individual cells and cell–cell crosstalk in maintaining cellular homeostasis [14]. With the development of high-throughput single-cell technology, we can now decipher heterogeneity, differentiation, and intercellular communication at the single-cell level in sophisticated organs and complicated diseases. For example, single-cell RNA sequencing (scRNA-seq) can characterise the complete transcriptome of a single cell on a massive scale. This novel technology usually involves tissue dissociation, single-cell capture, cell lysis, and messenger ribonucleic acid (mRNA) marked by cell barcodes (Fig. 2A, C). Following RNA capture, the next steps are reverse transcription and amplification, complementary deoxyribonucleic acid (cDNA) library construction, and sequencing using next-generation sequencing technology (Fig. 2C, D). Using the aforementioned cell barcode, the sequence reads are then navigated back to the original cells, producing complete transcriptome data for thousands of cells in a single experiment (Fig. 2). In many studies, scRNA-seq has been used for better understanding of cellular and pathological mechanisms and gaining novel cellular and biomedical insights into disease pathology [15, 16].

Fig. 2
figure 2

A concise overview of scRNA-seq workflow. CRISPR clustered regularly interspaced short palindromic repeats, CITE-seq cellular indexing of transcriptomes and epitopes by sequencing, GSVA gene set variation analysis, HVGs highly variable genes, PCR polymerase chain reaction, ScRNA-seq single-cell RNA sequencing, ScATAC-seq single-cell assay for transposase-accessible chromatin using sequencing, ScMT-seq single-cell methylome and transcriptome sequencing, UMI unique molecular identifier

Full size image

To maintain liver homeostasis in a healthy state while remaining responsive to exogenous and endogenous stress, a highly organised and specialised structure of liver cells, including hormones, cytokines, metabolites, and microbial products, is needed to integrate local information at the molecular level. Liver research has embraced and benefited from innovative approaches (high-throughput single-cell technologies and multiomics), with several liver scRNA-seq studies being published in the past 5 years [17,18,19,20,21,22,23]. Recently, several reviews have summarised the application of single-cell techniques in various types of liver diseases, especially liver cancer [79, 80]. Spatial barcoding data are usually subjected to deconvolution, while single-cell resolution spatial data (such as high-plex RNA imaging (HPRI) data) are typically subjected to map** to localise scRNA-seq subpopulations. Next, algorithms that evaluate the spatial arrangement of localised subpopulations further assess the ligand–receptor interactions predicted from scRNA-seq data. Given that much cellular crosstalk, notably juxtracrine and paracrine communication, is spatially restricted, ST data are well suited to evaluate the reliability of the ligand–receptor interactions computed from scRNA-seq [81, 82]. Finally, the integrated results can be visualised through the creation of heat maps, scatter plots, and two-dimensional graphs, among other methods. For more information on each module, please refer to a recent review article published in “Nature Reviews Genetics”, which explores the attempts to integrate scRNA-seq with ST techniques, including the use of emerging integrated computational methods, and proposes effective ways to combine current methods [83].

Single cell and ST in a normal liver

The human liver is the largest organ of the body. It is made up of parenchymal cells, such as hepatocytes and cholangiocytes, that constitute 80% of the liver mass [84], and nonparenchymal cells, such as HSCs, LSECs, and recruited and tissue-resident immune cells [85]. ScRNA-seq enables the simultaneous detection of parenchymal and nonparenchymal cells to generate an atlas in the normal liver, provides novel insights into the phenotypic zonation of hepatocytes, and improves our understanding of rare cell types, such as LSECs [73], cholangiocytes [73], and hepatic progenitor cells [86] (Fig. 1B). Additionally, the integration of scRNA-seq and ST can generate high-resolution maps of cell subpopulations in tissues. This combination can help us better understand the changes in cell types and gene expression in different regions of the liver.

A single-cell perspective of normal liver cell atlas

An important application of scRNA-seq is to generate a full-scale perspective of a normal liver cell atlas. Since 2018, several independent studies have performed single-cell analysis in healthy human livers at high-throughput resolution [15, 41, 59, 72, 73, 87,88,89,90]. MacParland et al. were the first to publish a study on the normal human liver utilising scRNA-seq technology [73]. Liver grafts were obtained from five neurologically deceased donors with systemic inflammation from brain death but normal liver histology, and 8444 liver cells were profiled. Twenty different cell populations were characterised, including hepatocytes, endothelial cells, cholangiocytes, HSCs, B cells, conventional and non-conventional T cells, and two distinct populations of resident macrophages [73]. Liver samples used to generate a normal liver cell atlas have also been acquired from individuals of both sexes across a wide range of age groups (from 21 to 65-years-old), and from those with diverse underlying medical conditions [72, 88, 90]

It is extremely difficult to obtain fresh, healthy human liver tissue of good quality in the field. Therefore, a recent study has compiled, integrated, and analysed available scRNA-seq data from 28 healthy human liver samples. Moreover, they have established an online cell browser that provides easy and open access to transcriptional data of a diverse range of annotated parenchymal and nonparenchymal cells [91]. The combined database includes 26 clusters of 36,188 human hepatic cells [91]. Notably, an R Shiny approach has been developed for interactive visualisation, and scientists can access the website (http://liveratlas-vilarinholab.med.yale.edu). This innovative approach offers a user-friendly experience and interactive visualisation of gene expression for each cell (sub)type, ranging from abundant to rare liver cell subpopulations. Furthermore, this pioneering web tool furnishes detailed resources on which and what proportion of cell (sub)types express a gene of interest.

In summary, using novel single-cell technologies, these studies provide new insights into the transcriptomic structure of the human liver in a physiological state, revealing previously unattainable information that is highly valuable for the liver research community worldwide. However, there are still many aspects of liver biology that have barely been studied and several problems that need to be addressed in the future. First, the normal livers are obtained from patients with a variety of underlying medical conditions, in other words, they are not really “normal”. Additionally, the dissociation and handling of liver tissue are thought to induce changes in gene expression in liver cells. Finally, the dynamics of the normal liver atlas due to gender and age need to be elucidated more thoroughly.

Landscape of intercellular crosstalk

ScRNA-seq has provided many innovative insights into cellular heterogeneity and function in liver physiology. However, pathological processes such as ischaemia, inflammation, anti-infection, fibrosis, regeneration, autoimmunity, and tumourigenesis are complicated, with a highly organised interactome between specific subgroups of multiple cell types that are likely to play crucial roles in disease progression. Therefore, to discover new and effective therapeutic targets, it is crucial to understand the complex interactions present in different liver diseases. Importantly, scRNA-seq data not only define the characterisation of different cell types but also provide information on which cell subgroups express specific genes of interest and to what extent. Accordingly, we can further analyse scRNA-seq data to study the underlying ligand–receptor interactions between distinct cell types, model the interactome, and highlight cellular and molecular mechanisms that might drive disease progression.

A number of bioinformatic methods have emerged to carry out interactome analysis on scRNA-seq data [92,93,94]. Unbiased CellPhone DB analysis [93] is an approach that infers inter- and para-cellular crosstalk from the combined expression of multi-subunit ligand–receptor complexes between cell clusters. To clarify the molecular mechanisms of an immune checkpoint inhibitor (ICIs)-related hepatitis, CellPhone DB analysis was carried out to identify the ligand–receptor communications between T cells and myeloids in the liver of an ICI-induced hepatitis murine model [95]. Significant ligand/receptor pairs with T cell activation cytokines TGFB, IL21, IL18, and IL15 were found between T cells and myeloids and between T cell subgroups [95]. The CellPhone DB algorithm has also been used to investigate the intra-scar activity of several pro-fibrogenic signalling pathways between endothelial cells, collagen-producing mesenchymal cells, and scar-associated macrophages in human fibrotic livers [90]. Notably, during fibrosis, high levels of the Notch ligands Delta-like ligand 4 (DLL4), Jagged 1 (JAG1), and Jagged 2 (JAG2) expressed by endothelial cells can signal through Notch receptors expressed by mesenchymal cells [90]. Single-cell secretome gene analysis was performed on hepatocytes and nonparenchymal cells isolated from the livers of an NAFLD mouse model [42]. By combining scRNA-seq data with a ligand–receptor interactome database [96], a highly correlated signalling network among major hepatic cell types was generated, indicating potential paracrine and autocrine connectivity with a high speciality in the mammalian liver [42].

In summary, compared with traditional bulk RNA sequencing, scRNA-seq data analysis provides a sensitive approach for delineating precise and complicated cellular sources of ligands and receptors and elucidating the roadmap of intercellular communication within the mammalian liver. Future challenges include functional assessments of the predicted signalling pathways and the precise elucidation of their biological functions in health and disease.

Liver zonation and function based on a combination of spatial and single-cell transcriptomes

The hepatic lobules, a fundamental functional unit in the mammalian liver, take on a hexagonal shape and contain hepatocytes arranged in plates along the radial lobule axis, with the central and portal veins situated at either end [97, 98]. In accordance with this spatial variation, versatile functions of the liver vary non-uniformly inside the hepatic lobules along this porto-central axis, a phenomenon termed “zonation” [84, 99]. Therefore, according to different hepatic functions and anatomical locations, the hepatic lobule is classically divided into three different zones [75]. Zone 1 surrounds the portal tract where oxygenated blood from the hepatic arteries enters. Zone 3 encircles the central veins, which is hypoxic but receives a nutrient-rich blood supply from the gastrointestinal tract. Zone 2 is located between Zone 1 and Zone 3 (Fig. 1C).

Our knowledge of hepatic cell organisation within these zones was deepened by spatially resolved transcriptomic analysis using scRNA-seq. By combining scRNA-seq with single-molecule RNA fluorescence in situ hybridisation (smRNA-FISH), Halpern et al. performed spatially resolved RNA sequencing to study liver zonation in mice [100]. This study indicated that liver zonation was mainly determined by the oxygen gradient and classical WNT signalling [101], and also by rat sarcoma protein (RAS) signalling to a lesser extent. Intriguingly, a substantial number of spatially zonated genes were not predicted to be downstream targets of these signalling pathways, suggesting the need for the identification of many other molecular signalling pathways important for liver zonation identification. Notably, the mouse interzonal hepatocyte marker genes identified are not applicable to human liver data [18, 73].

Building on the previous study [100], the concept of ST to dispel dependency on landmark genes or adherent cells was generalised by updated approaches. One of the current strategies is to fix MrnA molecules within the cell and introduce primers with barcode sequences, followed by rolling circle amplification (RCA) to increase the local copy number [102, 103]. Although a single-cell resolution is yet to be achieved, as recently reported by Sun et al., spatial transcriptomic techniques enable unbiased reconstruction of metabolic liver zonation [104]. Moreover, to highlight the spatiotemporal metabolic organisation of the liver, ST and scRNA-seq were performed in combination on individual liver cells obtained via perfusion from 10 ad libitum fed mice across the diurnal circadian cycle at different time points [22]. ST was also employed to resolve the spatial correlations of vascular parts involved in liver zonation and to explore previously unidentified structures in the mouse liver [20]. Tracking the expression of transcriptional markers associated with zonation along the lobular axis computationally enables the study of zonation gradients in physical positions and allows the prediction of vein-type identity according to the expression profiles of neighbouring spots. In addition, the presence of transcriptionally distinct structures, cluster 5, was discovered in liver tissues, consisting of a small number of spots with distinct spatial localisation, which express mesenchymal cell-marker genes such as Vim and Gsn [105]. This cluster has not been identified in previous transcriptomic studies, mainly owing to the cell rarity contributing to these structures [20].

In addition to MrnA, other cellular features in hepatic cells may have similar spatial heterogeneity, including proteins, metabolites, and regulatory molecules such as microRNAs (miRNAs). Understanding miRNA and protein space distribution are essential to deciphering the contributions of hepatic cells to liver development, metabolism, regeneration, fibrosis, infection, and cancer [106]. Itzkovitz et al. adopted “spatial sorting”, a generic method, to illustrate the zonation of hepatocyte proteins and miRNAs [104]. Their study provides a generalised spatial atlas of protein and miRNA zonation, identifying core hepatocyte-donated miRNAs, such as miR-30a-5p and miR-122-5p [104]. Through a combination of miRNA and target MrnA levels, they also identified potential regulatory interactions that could mediate the degradation of zonated MrnA. This provides an important resource for future studies because miRNAs are highly dynamic and zoned during disease progression of the liver, such as fibrosis, viral infection, and cancer [107,108,109]. Moreover, spatial identification of metabolic enzymes and signalling pathways using proteomics and in silico techniques will definitely increase our knowledge of hepatotoxicity and can greatly improve the prediction accuracy of drug absorption, distribution, metabolism, and excretion [104, 110].

Therefore, ST approaches have profoundly improved our understanding of the functional specialisation of human livers and could be used to establish how this division of labour may be restored after disruptions caused by liver injury [72, 73]. For further anticipation, ST will greatly benefit studies addressing liver development, sexual dimorphisms of liver zonation, immunity, and general pathology in mammals, particularly humans.

Spatial and single-cell transcriptomics in the liver during infection

The liver is the largest immune organ of the human body. Because of acute or chronic infection, liver inflammation results in necrotic hepatocyte death [111]. Persistent hepatic damage leads to progressive fibrosis, disrupted liver architecture, cirrhosis, and subsequent tumorigenesis. The identification of single-cell processes throughout the course of infection is hindered by restricted access to human liver tissue longitudinally over time and, until recently, limited options for investigating rare and potentially disease-driving cell populations. However, high-throughput spatial and single-cell technology has shed new light on the mechanisms underlying liver function during infection.

Hepatitis B virus (HBV) infection

HBV infection is a serious health and economic burden worldwide [112]. Currently, there is no radical cure for HBV infection because the virus remains in a latent state as stably inactivated covalently closed circular DNA (cccDNA) within host cells [113]. Therefore, shedding novel insights into the underlying mechanism for the maintenance of HBV cccDNA is of great significance in identifying new paths of HBV replication (Fig. 5). Using the scRNA-seq analysis, Hashimoto et al. characterised DOCK11 and DENND2A as key genes linked to the preservation of HBV cccDNA in 2325 primary human hepatocytes infected with HBV [21]. The amounts of HBV DNA and cccDNA both decreased below the limit of detection under the depleted expression of these two genes, which revealed an important role of these genes in HBV maintenance in human hepatocytes [21] (Additional file 1: Fig. S1A). However, the underlying mechanism remains poorly understood and needs further exploration. Furthermore, to obtain an unbiased and full-scale landscape of intrahepatic immunological characteristics and associations with disease status in HBV-infected patients, scRNA-seq was performed in the liver and blood during different stages of hepatitis B infection, including immune tolerance, immune activity, acute recovery, chronic resolved, and healthy controls [114] (Additional file 1: Fig. S1B). Diverse immune cell subgroups with different sources and cellular interactions have been observed in different disease statuses in HBV infection. This study highlighted the communication between exhausted CD8+ T (Tex) cells, regulatory CD4+ T cells, and FCGR3A+ macrophages. This communication contributes to immune failure during HBV maintenance through mediating HLA class I molecules together with their receptors such as leukocyte immunoglobulin-like receptor, and may be useful in guiding the progress of immunotherapy (Additional file 1: Fig. S1B). In addition to the adaptive immune response, the innate immune cells play a significant role in HBV infection. To delineate the function of KCs in HBV infection, live CD45+CD64+F4/80+ liver macrophages were sorted by flow cytometry from C57BL/6 mice and subjected to scRNA-seq analysis using the Smart-seq2 pipeline [23]. Two distinct clusters of KCs among liver-resident macrophages were identified, and the CD45+F4/80+CD11bint TIM-4+CD206+ESAM+ cells were poised to respond to IL-2 treatment and cross-present viral antigens contained within circulating virions or hepatocytes. Therefore, specific subgroups of KCs were identified as regulating the signalling pathways that lead to the hepatocellular priming of HBV-specific CD8+ T cells, which acquired pathogenic effector and antiviral functions following exogenous IL-2 treatment (Additional file 1: Fig. S1C).

Fig. 5
figure 5

Underlying mechanism in HBV infection using ScRNA-seq. cccDNA covalently closed circular DNA, FVH fulminant viral hepatitis, HCC hepatocellular carcinoma, HBV hepatitis B virus, IL-2 interleukin-2, KC Kupffer cell, MHV-3 murine hepatitis virus strain-3, NTCP Na+-taurocholate cotransporting polypeptide, NETs neutrophil extracellular traps, rcDNA relaxed circular DNA

Full size image

Indeed, scRNA-seq can provide abundant transcriptomic data to study molecular events and cell heterogeneity at the single-cell level. However, due to hereditary deficiency, high-expression genes in the library are sequenced more frequently, and rare genes can be easily ignored. Therefore, it is challenging to acquire transcript information from low-abundance or rare genes in the presence of highly abundant genes [115]. Tyrrell et al. developed a CRISPR-CRISPR-associated nuclease 9 (CRISPR-Cas9) assay to deplete high-abundance transcripts, resulting in preferential enrichment of low-abundance genes, such as HBV genes in human hepatocytes. By incorporating CRISPR-Cas9 with scRNA-seq in hepatocytes infected with HBV, the three most abundant transcripts were successfully depleted, which enabled selective enrichment of the HBV transcript and subsequent HBV RNA sequencing in more than 74% of the cells, compared to the detection of HBV RNA in only 0.6% of the cells through direct sequencing without CRISPR-mediated enrichment [116] (Additional file 1: Fig. S1D). The technical improvement further enabled the study of HBV infection and interferon treatment of the HBV-infected Huh7.5-NTCP cell model.

The single-cell assay for transposase-accessible chromatin using sequencing (ScATAC-seq) is a novel technique that detects open regions and identifies regulatory regions of chromatin at the single-cell level, making it a broad region of single-cell high-throughput technologies [117]. The corresponding transcription factors or other regulatory proteins bind to open regions of chromatin and directly affect intracellular gene expression [118]. To better understand the HBV-associated cell types, different open regions of chromatin, and specific gene expression regulatory networks in patients with chronic hepatitis B (CHB), Dai et al. performed ScATAC-seq on 8016 peripheral blood mononuclear cells (PBMCs) derived from normal control (NC) individuals and CHB patients [119]. Specific leukocytic subgroups associated with CHB were found to be located in B-0 and T-3 clusters. Additionally, the potential mechanisms of the transcription factor (TF) motifs IRF2 and FOXC2, which are associated with B-0 and T-3, respectively, were elucidated in the occurrence of CHB (Additional file 1: Fig. S1D). Therefore, a more systematic and distinctive gene regulatory network was constructed by combining scRNA-seq and ScATAC-seq [119].

HBV can also cause fulminant viral hepatitis (FVH), a deadly disease that lacks an effective treatment, and its pathogenesis is not fully understood [120]. Immune response plays a crucial role in the pathological progression of liver failure, yet the entire immune microenvironment landscape of infected livers is important for effective strategy discovery [121]. Wang et al. constructed an scRNA-seq study of a mouse model of FVH due to murine hepatitis virus strain-3 (MHV-3) infection. ScRNA-seq revealed that neutrophils were the only innate immune cells exhibiting a significant increase in the liver in response to MHV-3 infection when the mice had severe systemic inflammation. Neutrophils extracellular traps (NETs) exacerbated liver damage by promoting fibrin deposition and inflammation, which were regulated by the FGL2-mucolipin 3-autophagy axis (Additional file 1: Fig. S1E). In summary, targeting NETs could offer a novel approach for treating FVH.

The underlying mechanisms of HBV-induced liver damage are multifarious; one of the well-known mechanisms is the HBV genome integration into the genome of human hepatocytes [122, 123]. Virus–human genome integration results in genomic instability and is a high-risk factor for the development of liver tumorigenesis (Fig. 5). In the context of hepatocellular carcinoma (HCC), the heterogeneity of HBV integration has been investigated using single-cell genome sequencing [124]. HBV integration was remarkably consistent in all cells of a single-nodular HCC with portal vein tumour thrombosis, supporting the notion that HBV integration is an early initiating event in hepatotumorigenesis. Moreover, a potential driver gene ZNF717 was specifically identified in HCC. This gene exhibits a noteworthy mutation rate at both the single-cell and population levels, and acts as a tumour suppressor by modulating the IL-6/STAT3 pathway [124] (Fig. 5).

Hepatitis C virus (HCV) infection

Although HCV is the world’s most common blood-borne virus, there is currently no vaccine to protect against it [125]. HCV persistence has been established in most infected patients, suggesting that it successfully evades innate and adaptive immune surveillance at multiple levels. Previous standard treatment depends on interferon (IFN)-based therapy; however, it has many adverse side effects and only 50% patients respond. Since 2015, highly effective direct-acting antiviral (DAA) therapy has revolutionised the treatment of chronic HCV infection, with standard cure rates exceeding 95% [126, 127]. Many studies have shown that, following IFN-free DAA treatment, both innate and adaptive immune homeostasis may be partially restored [127,128,129]. To clarify the role of DAA-mediated HCV eradication in global T cell immune function, Rosen et al. utilised scRNA-seq to characterise the transcriptome of circulating T cells before, during, and after the DAA-mediated HCV cure [130]. They revealed that rapid normalisation of IFN signalling was achieved in T cells with DAA therapy and was well maintained after therapy termination. Several T cell subtypes, including CD4 central memory and effector cells, CD8 effector memory and TEMRA cells, as well as Hi-interferon stimulated gene (ISG) populations, exhibit the most notable transcriptional alterations following this HCV cure. This study identified previously uncharacterised shifts in innate immune and interferon signalling within T cell subpopulations induced by DAA therapy in chronic HCV infection at different time points, providing an abundant data source for exploring the effects of DAA treatment on bulk T cells.

T-cell exhaustion may serve to limit excessive immunopathology during prolonged antigen stimulation; nevertheless, the potential outcomes of T cell exhaustion may include viral persistence and tumour progression [131,132,133,134]. In order to characterise the transcriptional profile, trajectory, and ultimate fate of these exhausted HCV-specific CD8 + T cells during and after the cessation of chronic antigen stimulation in the context of chronic HCV infection, Hofmann conducted scRNA-seq analysis in a well-defined cohort of patients receiving DAA therapy for chronic HCV infection [135]. It demonstrated that memory-like T cells persisted while terminally exhausted T cells diminished after DAA-mediated HCV treatment, leading to memory polarisation of the overall HCV-specific CD8+ T cell response. However, a core signature of exhausted memory-like CD8+ T cells was still detectable, including HCV-specific CD8+ T cells targeting variant epitopes to a lesser extent (Fig. 6A). The aforementioned data delineated a fresh hallmark of T cell exhaustion, which persists as a chronic blemish in CD8+ T cells specific to HCV, even subsequent to the discontinuation of chronic antigenic stimulation. Consequently, in order to elicit complete effector potential, therapeutic targeting of these chronic blemishes, in addition to antigen withdrawal, is imperative.

Fig. 6
figure 6

Single-cell perspective of the liver in other infection diseases. DAA direct-acting antiviral agents, DCs dendritic cells, HCV hepatitis C virus, HLJ1 human liver DnaJ-like protein, INF-γ interferon-γ, LPS lipopolysaccharide, MP mononuclear phagocytes, MPCC micropatterned co-cultures, NKs natural killer cells, P. vivax Plasmodium vivax, PIH proinflammatory hepatocyte, YBP ring1 and YY1 binding protein, RM recruited macrophage, TGFB1 transforming growth factor B1

Full size image

In addition to adaptive immune response, myeloid cells from liver fine needle aspirates (FNAs) in HCV patients before and after DAA treatment were deeply profiled by scRNA-seq [136]. Subpopulations of liver neutrophils, classical monocytes, non-classical monocytes, macrophages, eosinophils, conventional dendritic cells (cDCs), plasmacytoid dendritic cells (pDCs), and mast cells were comprehensively characterised. Notably, the upregulation of MCM7+STMN1+ proliferating CD1C+cDCs and expected downregulation of ISGs were observed post-cure (Fig. 6A). Programmed cell death-ligand 1 (PD-L1) + ISG-high neutrophils and IDO1-high eosinophils were identified as being crucial for immunoregulatory during chronic infections (Fig. 6A). In addition, three recurring gene programmes, ISG, MHC-II, and S100 were shared by multiple cell types, providing immunotherapeutic insights into HCV infections.

Plasmodium infection

Malaria remains a grave global health concern, afflicting approximately 421 million individuals and resulting in 627,000 fatalities in the year 2020 [137]. Plasmodium vivax is responsible for the majority of non-African cases of malaria [137]. There is an obligatory developmental stage of all Plasmodium parasites that infect humans in the liver, where the parasite experiences asexual reproduction known as schizogony within a human hepatocyte before releasing thousands of merozoites into the blood [138]. In the life cycle of P. vivax in the liver, some parasites known as hypnozoites give up immediate division in hepatocytes and persist in the liver for days, weeks, months, or even years before initiating schizogony, resulting in a relapse of blood infection [139] (Fig. 6B). Therefore, eliminating P. vivax without therapy that directly targets hypnozoite reservoirs is challenging [140]. However, it remains unclear whether the mechanism regulating the formation and dormancy of hypnozoites is the same in different relapsing malarial parasites.

Previously, transcriptome-wide profiling of the Plasmodium during liver stages was performed using traditional bulk RNA sequencing strategies [141]. Although informative, the study only provided insights into average transcriptional expression originating from either mixed (schizont and hypnozoite) or hypnozoite-only populations, instead of any transcriptional variations that might be present between individual parasites. Moreover, the potential impact of the parasite on host cells was not considered, impeding the investigation of host–pathogen interactions. These deficiencies emphasise the need to better understand the differences between individual plasmodia, as well as how the Plasmodium alters hepatic cells. Progress in these fields could facilitate the characterisation of parasite and host factors important for the development of the parasite liver stage.

To comprehensively examine Plasmodium-specific host interactions and potential responses in bystander hepatocytes, Dr. Sangeeta N. Bhatia conducted dual transcriptional profiling of P. vivax infection and surveyed host- and state-dependent gene expression patterns in both parasites and hepatocytes [142]. This research utilised a bioengineered human microliver platform to culture patient-derived P. vivax, generating a single-cell liver atlas of relapsing human malaria through scRNA-seq analysis. The study identified distinct features of early-, dormant-, mid-, and late-stage P. vivax, as well as subpopulations of sexually committed forms in the liver that were previously thought to emerge only during erythrocytic infection [143] (Fig. 6B). P. vivax infection inhibits the transcription of functional genes in key hepatocytes and induces an anti-P. vivax innate immune response, which includes the dysregulation of IFN and inflammatory signalling pathways. This study provides a foundation for understanding parasite–host interactions and sheds light on the biology of P. vivax dormancy and transmission. Moreover, the technology developed in this study can be applied to research other intracellular microorganisms, where low expression of pathogen transcripts or host contamination makes it difficult to perform single-cell studies.

At around the same time, Kyle et al. utilised the scRNA-seq platform of 10X Genomics to characterise the transcriptomic features of both the host and parasite in a P. vivax liver-stage model in vitro [144]. Differential gene expression was observed between replicating hypnozoites and schizonts, and previously unobservable variation by bulk sequencing was revealed between individual hypnozoites. Moreover, multiple host genes linked to energy metabolism and antioxidant stress were exclusively upregulated during hypnozoite infection in infected hepatocytes. This study not only elucidated the transcriptional signatures among infected hepatocytes but also provided potential mechanisms by which P. vivax can exploit host cells for rapid proliferation or quietly maintain persistence (Fig. 6B), which can inform therapeutic targets against P. vivax liver-stage infection.

Previous ex vivo investigations have suggested that the rate of Plasmodium infection may differ among hepatocytes situated in distinct regions (pericentral and periportal) within the liver, owing to spatial heterogeneity of hepatocytes [145, 146]. To identify heterogeneous host and parasite responses, scRNA-seq and single-molecule transcript imaging were combined to characterise the host and parasite temporal expression programmes in a zonally controlled manner for the rodent malaria parasite Plasmodium berghei ANKA [147]. The study revealed differences in parasite gene expression in distinct zones, including potentially co-adaptive programmes related to iron and fatty acid metabolism (Fig. 6B). Parasites were found to develop more rapidly in the pericentral lobule zones, and a subpopulation of ‘abortive hepatocytes’ were identified periportally; these appeared predominantly with high parasite inoculum, upregulated immune recruitment, and key signalling programmes. This study provides an ideal application of combined spatial single-cell analysis for understanding the liver stage of Plasmodium infection at high spatial resolution and highlights the heterogeneous behaviour of both the parasite and the host hepatocyte [148].

Single-cell technology has also boosted vaccine research and development against malaria. Dr. Alexandra J. Spencer and her colleagues conducted an in-depth analysis of immune phenoty** and scRNA-seq on the kinetics of peripheral blood samples and thin needle puncture of liver tissue to investigate CD8+ tissue-resident memory (TRM) cells and their circulating counterparts in volunteers who received a novel malaria vaccine treatment called prime-target vaccination [149]. By exploring the heterogeneity among liver CD8+ TRM cells at the single-cell level, two main subpopulations were identified, each sharing expression profiles with blood T cells. This study uncovered the potential of liver TRM-like cells as a protected object by liver-stage malaria vaccines, especially those adopting a prime-target approach. Phase III clinical trials of liver-stage malaria vaccines are currently underway, and a reproducible correlation of protection would be particularly valuable (Fig. 6B).

Schistosomiasis

Schistosomiasis, whose pathogenesis remains unclear, is a widespread helminth disease that causes acute and chronic injury and poses a heavy social and economic burden on people worldwide [13]. High-throughput single-cell transcriptomics enables us to decipher cell heterogeneity, differentiation, and cell–cell communication during the pathological process occurring in the liver during systemic inflammation. In a study conducted by Cai et al., the cellular landscape of hepatic cells at three different time stages in a “double-hit” endotoxemia mouse model was elucidated. Based on single-cell transcriptomic analyses, the study provided the first evidence that the phenotypic transition occurring in hepatocytes during endotoxemia plays a vital role in the recruitment of circulating monocytes to generate a recruited macrophage population, as well as in the inhibition of T lymphocytes through the CCL2-mediated pathway and upregulation of PD-L1 [157]. This study enhances the comprehension of endotoxemia pathology through a liver-centric perspective and establishes a crucial basis for develo** effective therapeutic strategies for acute infections. Human liver DnaJ-like protein (HLJ1), a molecular chaperone belonging to the heat shock protein 40 (HSP40) family, is a potential target for sepsis treatment. ScRNA-seq analysis elucidated the immune profile influenced by HLJ1 in a mouse model challenged with lipopolysaccharides (LPS). HLJ1 plays a critical role in regulating the IL-12/IFN-γ axis-dependent sepsis progression and may serve as a potential molecular target for novel antisepsis or immunomodulatory therapies [158] (Fig. 6D). Additionally, a mouse model of LPS-challenged endotoxemia was used to explore the zonation and spatial heterogeneity of innate immune function in the liver, revealing that NF-KB (p50) activation and c-reactive protein expression in response to endotoxemia are zone-specific [159]. However, this exploration was conducted using traditional RNAscope and immunohistochemistry methods instead of high-throughput spatial transcriptome analysis.

COVID-19

Since December 2019, the corona virus disease 2019 (COVID-19) and its pathogen, a novel coronavirus named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), have rapidly spread worldwide [160]. The essential factors for SARS-CoV-2 invasion into host cells are angiotensin-converting enzyme 2 (ACE2), a receptor binding to the spike glycoprotein (S) of the virus, and type II transmembrane serine protease (TMPRSS2), which mediates membrane fusion vital to release the viral contents into the infected cell cytosol [161]. Thus, the underlying mechanism of COVID-19-induced organ injury in the human body is related to these two factors (Fig. 7A). Although the lungs are the most frequently affected organ in COVID-19, SARS-CoV-2 can also cause damage to other organs, such as the brain, gut, and liver. Recent studies have shown that liver abnormalities have been observed in more than 50% of patients with COVID-19; in 20% of the patients, the abnormalities subsequently progressed to liver injury [162,1: Fig. S2). Prognostic modelling of tumour cell stemness and intratumoural heterogeneity signatures will be relevant in the future for patient stratification of treatments, particularly in liver cancer [182, 183] (Additional file 1: Fig. S2). Deeper comprehension of diseased liver tissue using single-cell multiomics has highlighted diagnostic biomarkers and gene signatures, pathogenic cell subgroups, and potential targetable signalling pathways [184, 185]. To realise this potential, further investigation into preclinical lineage-tracing and gene manipulation research with clinical trials is required to identify druggable targets for future therapeutic applications.

High-throughput single-cell technology has been used in various clinical settings. Single-cell genomics has already been applied to track disease trajectories before and after HCV therapy and tumour immunotherapy, highlighting future clinical applications as a precision medicine approach [130, 135, 186] (Additional file 1: Fig. S2). Moreover, these studies were aimed at investigating whether antiviral therapy facilitates the restoration of T cell responsiveness. Single-cell genomics can also be applied to evaluate the therapeutic efficacy and disease progression in clinical trials of a variety of liver diseases to generate frameworks for genetic, epigenetic, transcriptional, protein, and cellular landscape-based therapy strategies [187] (Additional file 1: Fig. S2). For instance, novel mechanisms and biomarkers of drug resistance have been identified to stratify patients and guide personalised therapeutic decisions via the application of scRNA-seq in clinical trials of multiple myeloma patient cohorts [188, 189]. In the future, patient-specific single-cell analysis of liver biopsies can reveal opportunities for personalised immunology by aiding the evaluation of patient outcomes and treatment effects. The extensive clinical application of single-cell technologies will revolutionise our capacity to track efficacy and tailor treatment strategies for understudied liver pathologies.

Future directions

In the era of precision medicine, higher-resolution data are needed to identify heterogeneous tissues and complicated diseases, such as microorganism infections. Currently, technologies that aid in the advancement of multiomics single-cell techniques that will allow the characterisation of genetic, epigenetic, transcriptional, proteomic, and metabolic changes in the same cell are being developed. Single-cell multiomics will allow an even more comprehensive understanding of liver biology and disease at a single-cell resolution. Future research should focus on reducing the expense of high-throughput single-cell technologies and identifying molecular, cellular, histological, or radiological deputy biomarkers that facilitate the characterisation and stratification of liver diseases, which will help predict drug efficacy or patient prognosis instead of conducting full single-cell analysis of samples per patient. As high-throughput single-cell techniques are further developed, they will result in a more comprehensive analysis of single cells in healthy and diseased livers, potentially producing novel therapeutic targets. These approaches, which generate more high-dimensional, complicated, and informative resources, necessitate the advancement of cutting-edge bioinformatic algorithms to extract the optimal amount of related information from these integrated datasets. We need to continuously evolve computational analysis pipelines synchronously with methodological advancement to offer new strategies for visualising, analysing, and interpreting single-cell omics datasets (Additional file 1: Fig. S2).

Conclusion

In summary, high-throughput single-cell and spatial omics are revolutionising our understanding of liver physiology and pathology in healthy and diseased states. These ever-advancing techniques have rapidly and profoundly accelerated the identification of rare- and disease-driving cellular subpopulations within the liver microenvironment. As we continue to become proficient in this technology, expand applications to map initial stages of infection in human tissue, and sample liver tissue longitudinally alongside disease progression, we will unveil novel and effective therapeutic targets and treatment options for patients suffering from various infections.