2.1.1 Intravital optical imaging with direct exposure of the liver

The simplest and easiest way to image the liver in vivo is to directly image the liver lobe through a simple abdominal incision that is exposed and externalized. This imaging method causes great damage to the body and is only suitable for imaging within a short period of time (e.g., within 6 h) [19-21]. As early as 2004, Geissmann et al. studied the crawling behavior of NKT cells (labeled by green fluorescent protein [GFP]) in mouse liver with confocal imaging using direct exposure of the liver. Their study found that NKT cells stopped crawling when activated by an antigen, revealing a new type of intravascular immune surveillance behavior [22].

2.1.2 Long-term optical imaging based on the abdominal window

The implantable imaging window allows for extended imaging time and proper tissue function while maintaining the quality and depth of imaging. The pioneering work performed by Jacco van Rheenen's team at the Royal Netherlands Academy of Arts and Sciences in 2012 established an abdominal window imaging model using the “pouch suture” method combined with a titanium ring, which can be used to observe mouse liver over a long period (5–6 weeks) with less damage [23]. However, the problem of pulsing due to lung breathing and heart beating still exists during liver imaging. This has led to the development of many improved imaging window models. For example, Jacquemin et al. replaced the titanium alloy rings with organosilicon materials to improve the softness of materials and avoid inflammation caused by wear at contact sites [24]. The plastic 3D-printed abdominal imaging window developed by Kuss et al. is lightweight, quickly manufactured, and can be mounted directly on the microscope carrier scaffold without fixing the window step [25], which improves the elimination of cell motion artifacts caused by dither during confocal imaging. However, these methods all require the use of tissue glue to fix the liver to the window model, which limits the imaging area of the liver; additionally, its limited adhesive strength may lead to detachment of the liver from the window in long-term experiments.

2.1.3 Dual-mode fluorescence/photoacoustic imaging based on a drawer-type abdominal window

A drawer-type abdominal window with an acrylic/resin coverslip model (DAWarc) was developed by our group. This model not only solves the liver focal plane dither caused by breathing and heart beating through the drawer structure designed to physically fix the hepatic lobes, but also balances the penetration efficiency of light and sound signals with the acrylic/resin cover, making it compatible with long-term live fluorescence and photoacoustic imaging [26]. Using the DAWarc model, long-term intravital fluorescence/photoacoustic imaging was performed for several days to reveal the spatial location and movement behavior of invariant NKT (iNKT) cells, as well as the distribution and morphological information of KCs in the microenvironment of liver metastases [26]. Without the need to fix liver tissue with tissue glue, this model facilitates intravital imaging of the liver, which is conducive for long-term dynamic monitoring of changes in immune cells and liver blood vessels during disease progression and the relationship between the two.

2.2.1 Transgenic mouse models of liver immune cells labeled with fluorescent proteins

Fluorescent reporter mice are constructed by tapping reporter genes into marker genes, achieving the purpose of marking different types of cells in vivo. For example, Clec4f-Cre-tdtomato mice were constructed by expressing nuclear localization signal-tagged tomato fluorescent protein under the promoter of Clec4f, a KC-specific gene [27]. By inserting enhanced GFP (EGFP) cDNA into the CXCR6 coding region, CXCR-GFP mice with GFP-labeled iNKT cells in the mouse liver were constructed [22]. By expressing Cre recombinase under the guidance of the Lecithin retinol acyltransferase (Lrat) promoter, Lrat-Cre1Rshw/Mmjax mice containing retinol lipid droplet-labeled hepatic stellate cells (HSCs) were also constructed [28]. The development of different types of transgenic fluorescence reporter mice serves as a potent tool for visualizing cell behavior in vivo.

2.2.2 Models of liver immune cells labeled with fluorescent dye-conjugated antibodies

Mouse liver immune cells can be specifically labeled by injecting fluorescent dye-conjugated antibodies into mouse tail veins. The following lists a few of the many examples of this model type: liver macrophages labeled by fluorescent dye-conjugated F4/80 antibodies; different types of T cells labeled by CD4 or CD8 antibodies; liver sinusoidal endothelial cells (LSECs) labeled by CD31 antibodies; and NK cells labeled by NK1.1 antibodies. Fluorescent dye-conjugated antibodies are the most widely used for in vivo specific labeling, but they are expensive and may hinder some cell functions. In addition to antibodies, cells can also be specifically labeled by nanoparticles coated with fluorescent dyes. The pomegranate-type fluorescent nanoparticles (size, ~300 nm) that were developed by Deng et al. can be used to label KCs through nonspecific uptake [29] (Table 1).

3.1.1 Liver resident macrophages—KCs

The liver contains the largest population of tissue-resident macrophages, the KCs, which develop from the yolk sac and account for approximately 20% of nonparenchymal liver cells [34]. Located in the hepatic sinuses, KCs have the amazing ability to quickly clear pathogens from the blood flowing through these sinuses, thereby performing effective immune surveillance. Because of their special morphology, strong phagocytosis, and easy labeling, an abundance of intravital imaging studies have been conducted on KCs. Zeng et al. labeled KCs with intravenous injection of Alexa Fluor 750-conjugated anti-mouse F4/80 antibody and visualized their ability to capture GFP-labeled bacteria in hepatic sinuses using spinning disk confocal microscopy. This study reported that 60%–80% of intravenous Staphylococcus aureus bacteria were captured and isolated from liver sinusoids by KCs within 5 min [35] and that this bacterial capture relied on complement receptor of the immunoglobulin superfamily (CRIg). CRIg was found to bind directly to lipteichoic acid, a cell wall component of Gram-positive bacteria, indicating that it is a functional pattern recognition receptor that can directly recognize bacterial components. This relatively independent process mediated by CRIg enables KCs to capture and prevent systemic bacterial transmission with high efficiency. Additionally, Sun et al. revealed the mechanism by which liver KCs actively filter the transmission of fungal pathogens in blood using confocal microscopy to image the dynamic interaction between KCs and fungi in the liver [36]. The imaging results visually showed that GFP-labeled yeast cells flowed in through the blood of infected mice stagnated during capture by KCs labeled with phycoerythrin (PE)-conjugated anti-F4/80 antibody (Figure 2a), suggesting that the clearance process of the fungus was related to the active filtration by KCs. In another study, the clearance rate of Cryptococcus neoformans was significantly reduced in KC-depleted mice compared with that in normal mice, further demonstrating the filtration function of KCs on fungi (Figure 2b) and suggesting that enhancing KC phagocytosis could improve the clinical status of patients with invasive fungal infection [36].

In addition to working alone, KCs can also act in concert with LSECs to effectively remove particles from circulation. When applied systematically, 30%–99% of nanoparticles are captured in the liver, mainly by KCs and LSECs [37]. By adjusting the size of the nanoparticles, the path of uptake can also be adjusted. In terms of nonspecific uptake, single nanoparticles with a diameter of >100 nm or polymerized nanoparticles with a diameter of <100 nm, interact more strongly with KCs, while other smaller nanoparticles have a higher proportion of uptake by LSECs [37]. Knowing the nonspecific uptake ability of KCs, our group developed a self-assembled “off/on” nanopomegranate particle (size ~300 nm) that was efficiently taken up by KCs (>98%) in the liver following intravenous injection. By applying this nanoparticle, fluorescence/photoacoustic dual-mode imaging was executed to reveal the strategic arrangement and function of KCs in living liver [29].

In addition to KCs, there are monocyte-derived macrophages in the liver. When KCs are depleted by diseases or artificial means, monocytes will quickly infiltrate the liver and transform into KCs in volume and shape (Figure 2c). Such behaviors enable monocytes to do the following: interact with other cells in the liver sinusoids by extending pseudopodia; promote changes in factors that drive the differentiation and development of KCs in the microenvironment; and eventually become macrophages imprinted with the KC identity, including the expression of transcription factors ID3 and LXR-α, and cell surface markers F4/80 and Clec4f [38, 39].

3.1.2 iNKT cells

Accounting for approximately 30% of intrahepatic lymphocytes, iNKT cells can reside in tissues for a long time [40]. Intravital imaging of CXCR6-GFP mice has revealed that iNKT cells in the liver crawl along the hepatic sinuses to perform intravascular patrol work (Figure 3a) and quickly stop moving in response to homologous antigens [22]. Under treatment with iNKT cell agonists or inflammatory cytokines, 70%–90% of iNKT cells were observed to become stationary for a few minutes [22], suggesting that the rapid arrest behavior of iNKT cells is a feature of activation. Thus, by monitoring the spatiotemporal characteristics of iNKT cells, researchers can better understand their activation in vivo. Wang et al. observed the dynamic relationship between iNKT cells (CXCR6-GFP cells) and KCs (labeled by Alexa Fluor 647-conjugated anti-F4/80 antibody) through real-time confocal imaging of the liver in nonalcoholic fatty liver disease (NAFLD) model mice established by feeding with a methionine- and choline-deficient (MCD) diet [41]. The imaging results showed that iNKTs were recruited to the liver and slightly aggregated on the first day of MCD feeding, while KCs began to gather together slightly on the second day (Figure 3b). Upon removal of KCs with liposome disodium chlorophosphite, iNKT cells did not accumulate until the fifth day of feeding. Multiphoton microscopy was then used to image the colocalization contact sites among iNKT cells, KCs and Nile red dye-labeled free lipids in the livers of NAFLD mice. This confirmed that rather than KCs, which are mainly involved in scavenging and phagocytosis, the main cells that phagocytose free lipids released by necrotic hepatocytes are activated iNKT cells [41], providing new insights into the pathogenesis of NAFLD and inspiring new therapeutic targets.

Additionally, using CXCR6-GFP mice, Liew et al. performed intravital imaging through spinning disk confocal microscopy, showing three stages of iNKT cells during aseptic liver injury repair [42]: (1) repulsion, during which iNKT cells did not enter the site of injury and even performed a 180° U-turn when patrolling near the site; (2) retention, which occurred at approximately 8 h after injury, when self-antigens and arrest cytokines such as interleukin-12 (IL-12) and IL-18 remained in production by iNKT cells at the boundary of the injury site; and (3) infiltration, which occurred at the injury site 48–72 h after injury. By delving into the behavior of iNKT cells that differ from that of neutrophils, macrophages and platelets recruited at the initial stages of injury, this research discovered that self-antigen-driven iNKT cells act as detectors and coordinators of the transformation from inflammation to tissue recovery, playing an indispensable role in the liver injury repair process [42]. While damage to the liver causes changes in iNKT movement behavior, damage to distant organs can also affect the dynamics of iNKT behavior [43]. Wong et al. used CXCR6-GFP mice for intravital imaging through spinning disk confocal microscopy and found that iNKT crawling behavior in the liver of mice with transient cerebral artery embolization was initially no different from that of normal mice; however, with cerebral ischemia reperfusion, the number of crawling cells was significantly reduced. This movement behavior pattern was not observed during cytokine or antigenic stimulation, or in posterior limb ischemia‒reperfusion model mice, suggesting that this behavior is mediated by neurotransmitters [43] and providing new insights into the cross-talk between the central nervous system and the immune system.

3.1.3 Effector T cells in the liver

The liver has been regarded as an organ conducive to the induction of strong immune tolerance [44], which is also considered to be the main reason for the persistence of chronic liver disease [45]. Despite this, the liver can also establish a rapid and potent T-cell response in acute inflammatory infections [46].

T-cell activation in acute infections is characterized by complex mechanisms. Under activation by different antigen-presenting cells, CD8⁺ T cells may exhibit different behaviors. Benechet et al. showed that when hepatitis B virus (HBV) antigen was expressed by hepatocytes (hepatocyte antigen-presenting group), antigen-specific CD8⁺ T cells formed loose but persistent intravascular clusters around the portal vein bundle, with a significantly increased proportion of motile cells. These cells differed from effector T cells in the transcription level and epigenetics, secreting little or no interferon (IFN)-γ and showing no cytotoxicity. Instead, they expressed programmed cell death protein 1. However, when the antigen was presented by KCs (KC antigen-presentation group), CD8⁺ T cells fully differentiated into effector T cells with the ability to secrete IFN-γ and cytolytic-like granzyme, forming dense extravascular clusters with most cells at rest and scattered among the hepatic lobules. Long-term continuous imaging monitoring showed that the T-cell clusters in the KC antigen-presenting group began to depolymerize, with migration of the cells from the liver on the 5th to 7th day of antigen presentation, while the cell clusters in the hepatocyte antigen-presenting group remained, suggesting persistence of the antigen-presenting effect in the hepatocyte antigen-presenting group. This study revealed differences in the roles of different antigen-presenting mechanisms in the initiation of CD8⁺ T cells in the liver [47].

As the main effector of tumor elimination, cytotoxic T cells (CTLs) recognize the major histocompatibility complex polypeptide antigen complex on the surface of tumor cells via the T-cell receptor (TCR). After making stable contact, the CTLs secrete IFN-γ and perforin to kill the tumor cells [48]. Our research group established a method that quantitatively evaluates the ability of CTLs to kill tumor cells in vivo and was used to prove that the ability of CTLs to kill tumor cells is limited in the immunotolerant environment of the liver [49].

To establish a tumor visualization model with multicolor labeling in vivo, Liu et al. screened a melanoma cell line that stably expressed the genetically encoded caspase-3 molecular probe (C3) and used it to generate liver metastasis model mice. Next, the tumor-bearing mice received in vitro-activated tdTomato fluorescent protein-labeled ovalbumin T-cell receptor I CTLs by adoptive transfer. The process of CTL-specific recognition and contact with tumor cells in the liver, as well as the temporal and spatial dynamics of caspase-3 activation, were dynamically monitored by intravital microscopic molecular imaging (Figure 4a). The results showed that the cumulative contact time required for CTLs to induce apoptosis of tumor cells was 98.9 ± 6.3 min. Most (82.6%) apoptotic tumor cells had contact with multiple CTLs, and at most six CTLs made successive contact with a single tumor cell (Figure 4b) causing the apoptosis of tumor cells. Therefore, the killing of tumor cells in living liver tissue, which leads to tumor cell apoptosis, is the result of long-term cumulative contact among multiple CTLs acting synergistically [49]. Quantitative evaluation and comparison of the tumor cell killing capacity of CTLs revealed that their ability to kill tumor cells in the liver was significantly limited compared with that in the spleen and was related to the high expression of the immunosuppressive IL-10 in liver tissue. Thus, intravital immuno-optical molecular imaging can be used to accurately evaluate the effectiveness of tumor immunotherapy in vivo and help understand the effect of organ immune status on cellular immunotherapy.

3.1.4 Neutrophils in the liver

Infiltration of neutrophils into the liver is one of the early responses to tissue injury or systemic inflammation [50]. Damaged cells in the liver release endogenous damage-related molecular proteins, such as formyl peptide, to activate innate immunity [51]. Formyl peptide receptor (FPR) signaling is important for guiding neutrophils to migrate to the site of inflammation [52]. Using a two-photon excitation microscope, Honda et al. investigated the number, movement velocity, and morphological changes of EGFP-labeled neutrophils recruited during inflammatory reactions in LysM-EGFP mouse liver following laser irradiation or ischemia reperfusion (I/R) of the liver [53]. After induction of regional necrosis by laser irradiation of the LysM-EGFP mouse livers, time-lapse videos demonstrated that neutrophils treated with FPR antagonist carried out nondirectional random migration around the injury area at 2 h post laser irradiation, with a lower meandering index compared with the untreated control group. At 3 h post laser irradiation, the accumulation of neutrophils in the necrotic area in the control group was significantly inhibited in the FPR antagonist-treated group. In I/R model mice treated with an FPR antagonist, the crawl velocity of neutrophils decreased (1.76 ± 0.28 vs. 2.70 ± 0.25 μm/min) and the ischemic necrotic area of the liver was significantly reduced, suggesting that inhibiting the inflammatory activation caused by neutrophil aggregation can alleviate acute liver injury.

3.2.1 NAFLD

NAFLD is a chronic liver disease that is rapidly increasing worldwide, along with hepatic steatosis, inflammation, fibrosis, and severe liver failure [56, 57]. However, due to the limited understanding of the pathophysiological mechanism of NAFLD, the assessment and diagnosis of NAFLD are insufficient and effective treatment strategies for NAFLD have not yet been established. An ideal research approach to clarify the cellular and molecular mechanisms of NAFLD pathogenesis in the liver microenvironment would be to perform repeat and long-term intravital imaging monitoring of the liver at the steatosis stage in a mouse model of NAFLD. Seoul-Fluor 44 (SF44), a newly developed fluorescent dye, can selectively label lipid droplet adipocytes in vivo within minutes, with a very low background [58, 59]. Peter Kim's research team successfully recorded the process of lipid droplet accumulation in the liver of NAFLD mice using a customized frame rate laser scanning confocal microscopy imaging system [60] developed by their research group and combining it with SF44 [61]. Over 30% of lipid droplets in NAFLD model mice that were fed an MCD diet had a diameter >3 μm at day 2, and up to 22% had a diameter >9 μm at day 21, whereas all of the observed lipid droplets in the control group, which was fed a conventional diet, had a diameter <3 μm (Figure 5a). Subcellular changes in the position of nuclei due to the accumulation of large lipid droplets in individual hepatocytes were also clearly observed. During the initial phase of the MCD diet, the positions of two nuclei observed in a single hepatocyte remained centered, and there were multiple micro-vesicles in the cytoplasm. After 21 days of the MCD diet, very large lipid droplets formed in the cytoplasm and shifted the position of the nuclei to the periphery (Figure 5b) [61]. Similar imaging observations of the progression of NASH using fluorescent dyes have contributed to the understanding of its pathological mechanisms and improved prognosis.

3.2.2 Liver infection

The liver is the main organ for removing infected bacteria from the blood, and KCs play a major role in this process. Unlike bacteria, parasites are not immediately captured by KCs. Ute Frevert's group used mice expressing Tie2-GFP and lys-EGFP-ki mice (expressing GFP in LSECs and KCs, respectively) to generate Plasmodium sporophyte (GFP labeling) infection models and then performed intravital imaging to study sporophyte behavior. They found that some Plasmodium sporozoites in circulating blood suddenly adhered to the endothelium of the hepatic sinuses and then slid downstream or upstream to the KCs; however, instead of being phagocytic, they passed over the KCs and moved into the hepatic parenchyma, where they eventually invaded hepatocytes (Figure 6) [62]. The movement of this parasite can permanently damage the membrane integrity of the transfected cell, so the ability to cross KCs may be one of the immune-escape strategies that it uses to improve survival during liver infection [63]. In addition to parasitic infections, visual imaging studies of oncolytic viruses have been conducted. Naumenko et al. performed dynamic imaging tracing of the interactions between Alexa Fluor dye-labeled oncolytic viruses and host cells in vivo to explore their systemic distribution. Their results showed that, in the liver, AF647-labeled viruses were preferentially ingested by GFP-labeled KCs over other hepatocyte populations [64]. Notably, current research has focused on infection caused by a single type of pathogen, such as a bacterial, fungal, or HBV infection [65]; however, clinical disease outcomes can be highly complicated and exacerbated by simultaneous or consecutive infection with two or more different microorganisms [66]. The ability to image different microbes during multimicrobial infections to determine how they affect colonization and clearance of each other in the body could inspire new clinical treatments [67].

3.2.3 Liver tumors

The rich blood flow of the liver facilitates the metastasis of tumor cells in circulating blood through the hyperpermeable hepatic sinuses, making the liver an organ prone to metastasis [68]. Intravital imaging is a widely used and necessary tool to study the mechanisms of liver metastasis and regression of liver tumors [69]. In the process of tumor development, genetically programmed changes occur in tumor cells and surrounding cells [70]. Bolei Dai et al. established a liver metastasis mouse model with a customized triple-fixed liver window imaging model and tetramer far-red fluorescent protein (tfRFP)-labeled B16 cells expressing apoptosis-related probes, including reactive oxygen species (ROS), caspase-3, and calcium probes. Employing this model, they observed the immune response in liver metastasis using intravital optical molecular imaging [71]. They found that, at the early stage, B16 tumor cells injected into hepatic sinuses via the spleen were influenced by blood flow shear forces and metabolic changes and that the ROS level was gradually increased. This upregulation of ROS induced the activation of intracellular caspase-3 (Figure 7), resulting in cell membrane perforation and the release of intracellular antigen tfRFP, which combined with an extracellular antibody to form an antigen–antibody complex. Further activation of the complement system accelerated and amplified the cell membrane perforation effect, while recruiting neutrophils and F4/80⁺ macrophages to the liver for synergistic antitumor effects [71].

Anti-CD20 therapy for tumor progression can lead to the depletion of B cells and thus tumor regression, but the exact mechanism remains unclear. Using in vivo liver imaging of mice receiving anti-CD20 treatment, Montalvao et al. revealed that KCs effectively consumed malignant B cells by mediating a sudden arrest and subsequent phagocytosis of circulating B cells in the sinuses of the liver [72]. Dynamic imaging of mice with RFP-labeled B cells after the injection of anti-CD20 antibody showed that circulating B cells begin to stop moving within minutes after injection of the antibody. After 10–30 min, the rounded morphology of the B cells is lost, and the RFP signal appears more diffuse, possibly indicating B-cell death and/or degradation. Subsequently, anti-CD20 therapy was performed on macrophage Fas-induced apoptosis mice with GFP-labeled KCs following adoptive transfer of yellow fluorescent protein-labeled B cells. Almost all of the B cells within the imaging horizon in the antibody-injection group were stagnated on KCs specifically and were phagocytosed by contact KCs a few minutes later, as was also demonstrated by the detection of occasional B-cell fragments in KCs. The results of this study identified the specific antibody-mediated phagocytosis of B cells by KCs as the main mechanism of action in anti-CD20 therapy [72].

Intravital imaging is also a productive method to visually observe the distribution of drugs (e.g., immune checkpoint-blocking drugs) or adoptive immune cells (e.g., chimeric antigen receptor T-cell immunotherapy) and their interactions with target cells in vivo providing valuable dynamic spatiotemporal information for the optimization of liver cancer immunotherapy regimens [73].