IL-2-mediated hepatotoxicity: knowledge gap identification based on the irAOP concept

Abstract

Drug-induced hepatotoxicity constitutes a major reason for non-approval and post-marketing withdrawal of pharmaceuticals. In many cases, preclinical models lack predictive capacity for hepatic damage in humans. A vital concern is the integration of immune system effects in preclinical safety assessment. The immune-related Adverse Outcome Pathway (irAOP) approach, which is applied within the Immune Safety Avatar (imSAVAR) consortium, presents a novel method to understand and predict immune-mediated adverse events elicited by pharmaceuticals and thus targets this issue. It aims to dissect the molecular mechanisms involved and identify key players in drug-induced side effects. As irAOPs are still in their infancy, there is a need for a model irAOP to validate the suitability of this tool. For this purpose, we developed a hepatotoxicity-based model irAOP for recombinant human IL-2 (aldesleukin). Besides producing durable therapeutic responses against renal cell carcinoma and metastatic melanoma, the boosted immune activation upon IL-2 treatment elicits liver damage. The availability of extensive data regarding IL-2 allows both the generation of a comprehensive putative irAOP and to validate the predictability of the irAOP with clinical data. Moreover, IL-2, as one of the first cancer immunotherapeutics on the market, is a blueprint for various biological and novel treatment regimens that are under investigation today. This review provides a guideline for further irAOP-directed research in immune-mediated hepatotoxicity.

Keywords:

• Adverse outcome pathway
• adverse event
• drug-induced liver injury
• immune safety
• immunotherapy
• interleukin-2
• hepatotoxicity
• preclinical model

Drug-induced liver injury

Drug-induced liver injury (DILI) is the most important cause for liver failure. It is an adverse reaction to a variety of drugs and other agents such as herbal products and dietary supplements (Hoofnagle and Björnsson 2019). The most common drugs leading to DILI include antimicrobials such as amoxicillin-clavulanate, central nervous system drugs such as valproate, cardiovascular drugs such as statins, and immunomodulatory agents such as tumor necrosis factor (TNF) inhibitors (Francis and Navarro 2022).

Diagnosis of DILI is difficult due to the facets of this condition and relies mostly on the exclusion of other entities. Commonly, it is associated with an increase in liver enzymes in the blood (Hoofnagle and Björnsson 2019). Other clinical symptoms may include weakness, poor appetite, jaundice, icteric sclera, dark urine, nausea, abdominal discomfort, fever, itchy skin, and ascites (Lu et al. 2016). It is crucial to detect DILI early and to halt drug exposure in order to avoid progression toward chronic disease or acute liver failure (David and Hamilton 2010).

Various studies report the epidemiology of DILI. Variations in epidemiological findings are evident, influenced by factors such as study design and the geographical location in which the research was undertaken. Altogether, the prevalence of DILI is rare in the general population, but it increases in hospitalized patients. The epidemiology of DILI has been thoroughly examined in previous literature (Li et al. 2022). In the general population of Western countries, incidences for DILI range from 2.3 cases per 100,000 persons/year in Sweden (retrospective cohort study (Valle et al. 2006)) or 2.7 per 100,000 persons/year in the United States (prospective cohort study (Vega et al. 2017)), to 19.1 per 100,000 persons/year in Iceland (prospective cohort study (Björnsson et al. 2013)). In Eastern countries, China estimates a rate of DILI cases in the general population of 23.8 per 100,000 persons/year (retrospective cohort study (Shen et al. 2019)) and in South Korea, the incidence of DILI patients that require hospitalization was estimated to be 12 per 100,000 persons/year (prospective case-series study (Suk et al. 2012)). There are various confounding factors, which complicate direct comparisons of epidemiological data (Li et al. 2022).

DILI can be classified into different patterns of injury, mostly hepatocellular, cholestatic, and mixed hepatocellular-cholestatic. These patterns refer to histologic features of injury with different location and degree of inflammation, necrosis, and apoptosis (Kleiner 2014). Mechanisms of DILI come in various shapes: direct (intrinsic), idiosyncratic, and indirect. The direct form of DILI is associated with the intrinsic toxicity of drugs, which causes liver injury by directly damaging liver cells or interfering with biliary excretion (Lu et al. 2016). It presents the most frequent DILI subtype, is dose-dependent, and predictable in animals as it is induced by compounds with immanent hepatotoxic effects, such acetaminophen (Gerussi et al. 2021). In contrast, idiosyncratic DILI is caused by the action of the drug and not by the toxic properties of the drug itself. Possible mechanisms for idiosyncratic DILI involve the host immune response and genetic and metabolic factors of the patients, for instance defects in detoxification pathways. It is often accompanied by allergic and autoimmune reactions. Therefore, idiosyncratic DILI is less common than direct DILI, unpredictable, and neither dose-dependent nor detectable in animal models (Fontana 2014; Hoofnagle and Björnsson 2019). Indirect DILI presents a recently emerging type of drug-induced hepatotoxicity. It is caused by the action of the active substance and both, induction of liver injury and aggravation of already existing conditions, are mechanisms of indirect DILI (Hoofnagle and Björnsson 2019). Immune-mediated liver injury is a subclass of indirect DILI which is primarily induced by immunomodulatory agents, TNF antagonists, and checkpoint inhibitors (Hoofnagle and Björnsson 2019). Although the mechanisms remain to be elucidated, blocking the immune checkpoints in a tolerogenic organ such as the liver presumably leads to immune-mediated adverse events. Important mechanisms discussed are the decrease of self-tolerance and activation of cytotoxic T-lymphocytes (Shojaie et al. 2021).

High dose recombinant human interleukin = (IL)-2, also known as the pharmaceutical aldesleukin, is applied to promote the immune response in the context of renal cell cancer and malignant melanoma (Boyman et al. 2006). It was observed to induce mild-to-moderate alterations in liver enzymes in at least half of the patients. In certain patients, elevations exceeding 5-=10-times the upper limit of normal was measured, resulting in the need for early cessation or postponement of further therapy (Bethesda 2018). We hypothesize that the liver damage observed in aldesleukin treated patients might be classified as indirect, immune-mediated liver DILI. For a better understanding of the liver immune system, the hepatic immune landscape will be described next.

DILI is a major risk in drug development and the main cause for withdrawal of drugs from the market (Meunier and Larrey 2019). Due to the small number of cases, it is difficult to rule DILI out with certainty before a new drug is approved for the market and widely available (Babai et al. 2021).

The hepatic immune landscape

The liver is constantly exposed to intestine-derived bacterial products and food-derived antigens. Under homeostatic conditions, a complex hepatic immune landscape with different cell types contributes to hepatic tolerance and prevents aberrant immune activation. Central immune cells in the healthy liver are myeloid macrophages, dendritic cells and myeloid-derived suppressor cells as well as lymphoid natural killer (NK), NKT, and T-cells (Robinson et al. 2016). Tissue-resident macrophages, also known as Kupffer cells (KC) represent the largest myeloid population in the human body and are implicated in both liver tolerance induction and in the pathogenesis of liver diseases. Other cell types which modulate the hepatic immune response include liver sinusoidal endothelial cells (LSEC) (Crispe 2003; Tsuchida and Friedman 2017; Kolios et al. 2006). The various roles of the LSEC layer in antigen presentation and tolerance induction and its morphology allowing transmigration of cells from sinusoids to the liver parenchyma, have been extensively reviewed (Guidotti and Iannacone 2013; DeLeve and Maretti-Mira 2017; Kubes and Jenne 2018; Wilkinson et al. 2020). Prominent non-parenchymal cell types are hepatic stellate cells (HSC). Activated HSC are major drivers of fibrosis during liver injury, but may also counteract CD8⁺ T-cell activation (Schildberg et al. 2011; Tsuchida and Friedman 2017).

Immune cell subsets differ both within the liver and in relation to those in the peripheral blood. In the healthy human liver, a reversed CD4⁺/CD8⁺ T cell ratio compared to the peripheral blood can be found (in peripheral blood 2:1, in liver 1:3.5), while the NK cell fraction is significantly enhanced (12.8% in the peripheral blood, in the liver 30.6%). CD3⁺ lymphocytes are predominantly located close to the portal tracts, others being dispersed throughout the parenchyma (Norris et al. 1998; Tu et al. 2007). In homeostasis, interaction of lymphocytes with liver-resident cells confers tolerance (see above), as regulatory T (T_reg)-cells are expanded and autoreactive CD8⁺ T-cells are inactivated or deleted. Under inflammatory conditions, both T-cell subtypes can be activated and contribute to both, immunity to fight infections and liver damage. There is little interaction between tissue resident NK cells and the liver non-parenchymal cells in homeostasis, but during chronic hepatic inflammation, resident NK cells release immunomodulatory cytokines, thus becoming liver-protective (Heymann and Tacke 2016).

The intrahepatic lymphocyte compartment is generally in a more activated state than the peripheral blood. For instance, intrahepatic lymphocytes show increased expression of activation markers such as the immune checkpoint CD27, the co-stimulatory molecule CD28, and the early activation antigen CD69. Further, intrahepatic CD4⁺ and CD8⁺ cells show elevated CD95 (Fas receptor) expression (Tu et al. 2007). The chemokine receptor CXCR6 is expressed on activated intrahepatic lymphocytes and its ligand, CXCL16, is constitutively expressed on LSECs (Heesch et al. 2014). In vitro IL-12/IL-18 stimulated CXCR6⁺ liver resident NK and T cells show enhanced expression of IFNγ and LAMP-1 (CD107a), a marker for cytotoxic T- and NK cell degranulation, compared to their CXCR6^- counterparts (Zhao, et al. 2020a). Reflecting a potential role of CXCR6⁺ liver resident NK and T-cells in Hepatitis B virus (HBV) associated liver disease, these cell types occur with increased frequency in HBV-infected livers. Moreover, single cell RNA sequencing identified distinct liver resident CD14⁺ monocyte, CD16⁺ monocyte and macrophage populations. These tissue-resident cell types may aid the maintenance of immune tolerance as gene set enrichment analysis showed downregulation of pro-inflammatory responses, the complement pathway, as well as IL-2-JAK-STAT5, IL-6-JAK-STAT3 and p53 signaling pathways (Zhao et al. 2020a). Intrahepatic lymphocytes play an important role in various liver diseases (Wang and Zhang 2019), as discussed below. In summary, the hepatic immune cell landscape differs at least partially from the peripheral blood immune cell compartment, especially in T- and NK cell composition, activation status, and surface marker expression.

IL-2 as a cancer immunotherapeutic

IL-2 is a human cytokine produced by macrophages and lymphocytes that induces the proliferation and maturation of T-cells. A recombinant form of IL-2 (aldesleukin) has been shown to be active against renal cell cancer and malignant melanoma (Bethesda 2018). The high dose IL-2 regimen for the treatment of cancer is usually 600,000 IU/kg, administered intravenously every 8 h for up to 14 doses over 5 d (Fisher et al. 1989; Atkins et al. 1999). Upon completion of infusion, approximately 30% of injected aldesleukin is detectable in the plasma. Common side effects include fatigue, fever, chills, nausea, diarrhea, and capillary leak syndrome leading to edema, hypotension, renal insufficiency, and pulmonary edema. Less common side effects include shock, anaphylaxis, severe infections, autoimmune conditions, and neurologic complications including somnolence, stupor, and coma (Bethesda 2018). In clinical trials it was also found to increase liver enzymes in serum and treatment might lead to cholestasis in certain patients (Bethesda 2018). Most of the adverse effects are reversible and manageable today (Atkins et al. 1999; Dutcher et al. 2014).

IL-2-mediated hepatotoxicity

Various clinical studies report adverse effects of high dose IL-2 on the liver during treatment of renal cell carcinoma or malignant melanoma. For renal cell carcinoma patients, hepatic dysfunctions have been reported in 11% of patients treated with high dose aldesleukin. These effects were not dose-limiting but reversible upon completion of the treatment (Fyfe et al. 1995). Grade 3 and 4 liver toxicities were similarity observed in a clinical study comparing high dose IL-2 treatment with IL-2/IFNγ combination treatment in patients suffering from metastatic renal cell carcinoma (McDermott et al. 2005). IL-2 therapy is linked to impaired liver function, manifesting in jaundice, right upper quadrant tenderness, hepatomegaly, nausea, anorexia, and pruritus (Fisher et al. 1989; Dutcher et al. 2001). The most common hepatic function abnormalities comprise hyperbilirubinemia with bilirubin levels rising to 10 mg/dl, and low albumin with levels dropping to 2 g/dl. Five- to ten-fold increases in hepatic enzyme levels have been reported and aldesleukin therapy is discontinued in some centers if such elevations are observed (Dutcher et al. 2014). Hepatobiliary scans of patients with different tumor entities treated with high dose IL-2 showed delayed liver clearance as determined by biliary excretion, quantitative analyses also pointing toward delayed uptake. Interestingly, these effects were not restricted to patients with a specific bilirubin level threshold. Ultrasound detected an increased distance of hepatobiliary and vascular structures from the parenchyma, suggesting hepatic edema or infiltrative processes (Fisher et al. 1989). For the management of high dose IL-2 related toxicities, therapy can be interrupted when bilirubin or hepatic transaminases are significantly augmented, albumin is dramatically decreased, and prothrombin time is prolonged. Upon cessation of high-dose IL-2 therapy, liver damage reverses as liver function parameters normally return to baseline within 5 to 6 days (Schwartz et al. 2002). In clinical trials, low-dose IL-2 treatment was well tolerated (Mahmoudpour et al. 2019) and no adverse effects on liver function were indicated (Todd et al. 2016).

The initially proposed mechanism for IL-2-mediated hepatotoxicity was activation of tissue-resident macrophages, which produce TNFα and trigger activation of the local endothelium leading to recruitment of peripheral immune cells that mediate liver damage (Nakagawa et al. 1996). This hypothesis has been extended to include various other IL-2 receptor (IL-2R) expressing cell types, as well as novel mechanisms induced by IL-2, as described below. In a literature-based approach, we identified key players and biorelevant molecules for IL-2-mediated hepatotoxicity. For this, we employed the Adverse Outcome Pathway (AOP) concept.

Immune-related Adverse Outcome Pathways

The AOP concept provides a framework for the collection, evaluation and integration of available data to understand the adverse events induced by specific mechanisms of action (Ankley et al. 2010). Extended to immunological processes, the immune-related AOP (irAOP) concept presents an overall AOP to describe immune-mediated adverse effects (Villeneuve et al. 2018). AOPs are anchored between two Key Events (KE), which are the Molecular Initiating Event (MIE) and the Adverse Outcome (AO). These are connected by a stringent order of KE depicting unique and measurable states of a system, linked to each other by KE relationships (KER). Starting with the MIE at the molecular level, the connecting KE ascend through cellular to organ levels, with the AO becoming apparent in vivo on the organ or organism level (Knapen et al. 2015). This approach enables, e.g., the identification of key molecules in biologically-relevant immune processes in the diagnosis or prognosis of hepatotoxicity and aid development of specific in vitro tests (Vinken 2020). Moreover, AOP can identify knowledge gaps to be filled with experimental evidence, rendering an AOP a living document (OECD Series on Adverse Outcome Pathways 2018). In a recent article, Ball et al. (2021) illustrate how AOP, knowledge drawn from assays, and structure activity relationship predictions can be combined to select predictive assays for compound toxicities and to link assay outcomes to AO (Ball et al. 2021).

In the past, the AOP concept was mainly used for answering toxicological questions of chemicals. For some time now, however, the AOP concept has been adapted to biomedical research questions and shown to have great value for biological research and drug discovery (Bal-Price and Meek 2017). Ultimately, the concept can help to understand the pathomechanisms of human diseases. This information can subsequently be used to develop novel and safer drugs. In this review, we aim to provide a prototype for further evaluation of immune-mediated liver damage evoked by novel IL-2-based therapies or of biological investigative approaches with a similar mode of action. For the generation of such a potential irAOP, published AOP may serve as template. There are already several AOP leading to liver injury published in the open-source AOP Wiki, and some of them account for immunotoxicity mechanisms. These include, for example, AOPs describing chemically induced immune-mediated hepatitis (AOP 362), immune-mediated liver fibrosis (AOP 38), AOP for liver cholestasis (e.g. AOP 27), inflammation (e.g. AOP 144), and cancer (e.g. AOP 220) (Gijbels and Vinken 2017).

In this review, we generate a putative irAOP for IL-2 induced DILI based on relevant literature data, to provide an overview of possible mechanisms, key players, and knowledge gaps in IL-2-mediated hepatotoxicity.

Overview of the model irAOP for IL-2-mediated hepatotoxicity

Based on literature review, we developed an irAOP framework for IL-2-mediated hepatotoxicity. For this, we collected both direct evidence for IL-2-mediated hepatotoxicity and for general mechanisms for immune-mediated injury.

High dose aldesleukin treatment supplies the circulation with a surplus of IL-2. Therefore, aldesleukin is not entirely consumed by T_reg cells, but presumably also bound by cells with a lower affinity toward IL-2 such as cytotoxic T- and NK cells, that are subsequently activated, which is the principle of IL-2 therapy (Abbas et al. 2018). Aldesleukin is also distributed to the liver (Proleukin^® Product Monograph 2018), where it may act on local IL-2R expressing cells. We hypothesize that this binding of IL-2 to liver resident cells depicts the starting point (MIE) of the model irAOP. We suggest that due to the availability of high IL-2 concentrations, the tolerogenic environment of the liver is overcome and cytotoxic cells are activated to create a pro-inflammatory environment (KE1) which leads subsequently to recruitment/infiltration of (activated) immune cells to the liver (KE2). Further, we theorize that the exacerbated inflammation then leads to damage of the liver (KE3), and finally to immune-mediated hepatitis with impaired liver function (AO, Figure 1). Within our irAOP model, we identified biorelevant molecules and mechanisms, which are listed in Table 1. More, several knowledge gaps were identified, which are highlighted in the appropriate section and summarized in Table 2 with a selection of models that could be used to fill the gaps and render the picture of the model irAOP more complete.

Figure 1. Illustrated potential immune-related Adverse Outcome Pathway for IL-2-mediated hepatotoxicity. Binding of IL-2 to IL-2R expressing cells initiates the AOP (MIE). IL-2 binding induces signaling and proliferation (a.), receptor upregulation (b.), cytokine expression (c.) or interaction with other immune cells (d.; KE1). Consequently, activated immune cells are recruited to the liver sinusoids and adhere to LSECs and/or transmigrate into the perisinusoidal space (KE2). Liver damage occurs by direct or indirect cytotoxic effects or interaction of recruited immune cells with hepatocytes or by impairing the blood flow in the sinusoid lumen (KE3). Finally, impaired liver function or killing of hepatocytes triggers the AO hepatotoxicity (not depicted). Black arrows mark interactions, dashed lines indicate possible interaction points which lack of literature evidence.

Table 1. Biorelevant molecules and mechanisms related to IL-2 according to the irAOP.

Cell type	IL-2R expression (basal)	Cytokines/cytotoxic molecule expression	Receptor surface expression	Mechanisms induced
Neutrophils	Intermediate-affinity (Djeu et al. 1993; Girard et al. 1995; Abdel-Salam and Ebaid 2014)	TNF-α↑ (Wei et al. 1993)	CD11b↑, CD18↑ (Li et al. 1996; Abdel-Salam and Ebaid 2014)	Impaired chemotaxis, normal random migration (Klempner et al. 1990)
Monocytes and macrophages	Intermediate-affinity (Scheibenbogen et al. 1992)	IL-1β↑, −6↑, −8↑, TNF-α↑ (Bosco et al. 2000)	CFS-1R↑ (Espinoza-Delgado et al. 1990)
NK cells	General: intermediate-affinity; CD56^bright: high-affinity (Baume et al. 1992; David et al. 1998)	IFN-γ↑, TNF-α↑, IL-1β↔ (Young and Ortaldo 1987; Numerof et al. 1990)	KIR2DL4 ↑, NKG2D ↑, CD158a?, -b?, CD161 ↔, CCR2↑, CD11a↑, CD2↑, CD44↑, CD54 (ICAM-1)↑, CD58↑, CD18↑, TRAIL↔ (Mäenpää et al. 1993; Umehara et al. 1993; Polentarutti et al. 1997; Kogure et al. 1999; Kikuchi-Maki et al. 2003; Ishiyama et al. 2006; Konjević et al. 2010) liver NK cells: TRAIL↑, CD158a↔, -b↔, CD94↔ (Ishiyama et al. 2006)	Increased cytotoxic potential (Sharma and Das 2018) granule convergence (James et al. 2013) increased proliferation of CD56^bright (Baume et al. 1992) and intrahepatic NK (Martrus et al. 2017) Increased chemotactic response to MCP-1, −2, and −3 (Allavena et al. 1994)
CD8^+ T cells	Intermediate-affinity (David et al. 1998; Zhang et al. 1998)	Perforin↑, granzyme A↑, B↑, C↑ (Janas et al. 2005), IFN-γ↑ (Kasahara et al. 1983)	CC-CFR2 (MCP-1 receptor)↑, CC-CKR1 (RANTES receptor)↑ (Loetscher et al. 1996)	Reduced TCR threshold (Au-Yeung et al. 2017) sustained proliferation in periphery (D'Souza et al. 2002; D'Souza and Lefrançois 2003) restored effector function in absence of CD4^+ help, decreased apoptosis (Wolkers et al. 2011) increased cytolytic activity (Janas et al. 2005) induces FasL/Fas killing (Esser et al. 1997)
CD4^+ T cells	Intermediate-affinity (Zhang X et al. 1998)	IFN-γ↑, IL1β↑ (Kasahara et al. 1983; Numerof et al. 1990)		Priming for Th1 and Th2 differentiation (Zhou et al. 2003; Liao et al. 2008; 2011b) induces FasL/Fas killing in Th1 (Esser et al. 1997)
Treg cells	High-affinity (Setoguchi et al. 2005)	–	α4-integrin↑ β7- integrin↑ (Hsu et al. 2018)	Treg homeostasis (Zorn et al. 2006) low-dose IL-2 increases, high dose IL-2 decreases suppressive capacity (Moon et al. 2015)
NKT cells	High-affinity (Baev et al. 2004; Osada et al. 2005; Jukes et al. 2012; López et al. 2018)	GM-CSF↑, IFN-γ↑, IL-4, perforin↑ (Hou et al. 2003; Bessoles et al. 2008)	–	Increased proliferation (Hou et al. 2003)
Kupffer cells	High-affinity (Simone et al. 2021), IL-2Rβ undetectable in hepatocyte-KC co-cultures (Nguyen et al., 2015)	–	–	Enhanced response to LPS (Curran et al. 1988) suppression of hepatocyte protein synthesis (Curran et al. 1988) increased activity (Nakagawa et al. 1996)
LSECs	–	–	–	Overcomes tolerance induction (Schurich et al. 2010)
HSCs	–	–	–	Antagonizes inhibitory effect (Schildberg et al. 2011)
Hepatocytes	–	–	–
Liver (Organ)	All 3 subunits detectable (Fagerberg et al. 2013)	TNF-α↑, IL-1β↑, RANTES↑, MIP-1α↓, IFN-γ↔ (Anderson et al. 1996)	VCAM-1↑, ICAM-1↑ on parenchyma LFA-1 and VLA-4 on leukocytes (Anderson et al. 1996; Nakagawa et al. 1996; Lentsch et al. 1997)	Leukocyte, neutrophil, and platelet adhesion (Anderson et al. 1996; Nakagawa et al. 1996; Lentsch et al. 1997)

(↑) induced or upregulated; (↓) downregulated; (↔) not regulated; (?) contradicting/ambiguous literature; (-) no literature identified. Listed molecules and mechanisms were identified from the irAOP. Data derived from different species are marked as follows: Human (bold), murine (italic), both murine and human (bold + italic), rat (underligned).

Table 2. Knowledge gaps identified from the irAOP and suggested experimental approaches to fill the gaps.

Key event	Cell type(s)	Knowledge gap	Suggested models/methods to fill the gap
MIE	Purified primary human hepatic cells: Hepatocytes, LSECs, HSCs, KCs, intrahepatic lymphocytes, biliary cells; lymphocyte subpopulations	What is the IL-2R expression?	Quantitative PCR to determine IL-2R mRNA expression. Flow cytometry to determine IL-2R protein expression.
MIE	Treg cells	Could sCD25 released by Treg cells be a marker for IL-2 induced hepatotoxicity?	Correlation of diagnosed hepatotoxicity with concentration of plasma sCD25.
KE1	CD8^+ T cells	Is there a patient-specific TCR signaling threshold?	Isolation of CD8^+ T cells from blood, determination of TCR signaling threshold by TCR stimulation and analysis of reporter gene expression.
KE1	LSECs	Do LSECs directly respond to IL-2 stimulation?	Stimulation experiments of LSECs with IL-2, permeability studies, viability and surface marker assessment.
KE1	NK cells	Are there different IL-2 signaling thresholds in the different NK cell subpopulations?	Activation studies/ study of IL-2 signaling machinery upon IL-2 stimulation.
KE1	T cells	How is the proliferative response in human primary T cells from peripheral blood vs. the liver?	Proliferation assay with T cells isolated from liver tissue and from peripheral blood.
KE1	T helper cells	What role does exogenous IL-2 play in the activation or polarization of T helper cell subtypes?	Activation studies with T helper cell types; detection of transcription factor expression, surface marker and released cytokines.
KE2	CD8^+ T cells, NKT cells	Does IL-2 treatment increase adhesion marker (e.g. LFA-1 or VLA-4) expression on CD8^+ T or NKT cells?	Activation studies; detection of adhesion molecules and transmigration assay.
KE2	LSECs	How are LSEC adhesion and junctional molecules regulated upon IL-2 treatment?	Adhesion and transmigration experiments with LSECs. Co-culture adhesion and transmigration experiments LSECs and immune cells.
KE2	Neutrophils, NK cells, CD8^+ T cells, NKT cells, and LSECs	Does IL-2 affect cell migration in the liver?	Adhesion experiments with LSECs. Organ-on-a-chip model and assessment of transmigration.
KE2	Lymphocytes	Does IL-2 induce immune cell recruitment to the liver in humans?	Organ-on-a-chip model with human hepatocytes and perfused immune cells.
KE3	CD8^+ T cells and HSCs	Can exogenous IL-2 antagonize HSC-mediated inhibitory effect toward CD8^+ T cells?	Assess HSC effect on CD8^+ T cell activation under IL-2 stimulation conditions.
KE3	CD8^+ T cells and LSECs	Can exogenous IL-2 overcome LSEC tolerance induction in a human setting?	Assess LSEC effect on CD8^+ T cell activation under IL-2 stimulation conditions.
KE3	KCs and hepatocytes	Could CYP be an additional marker for hepatocyte damage following IL-2 administration?	Functional hepatocyte assay; detection of CYP activity/expression.
KE 3	KCs and lymphocytes	Are there synergistic effects of KCs and lymphocytes regarding cytokine expression upon IL-2 treatment?	Co-culture of KC and lymphocytes, detection of cytokine release.
KE3	NKT cells and hepatocytes	Does IL-2 induce hepatotoxicity via collateral damage by NKT cells?	Cytotoxicity assay with IL-2 activated NKT cells and hepatocytes.
KE3	Lymphocytes and hepatocytes	What are the effects of IL-2 activated lymphocytes on hepatocytes?	Co-culture or organ-on-a-chip model hepatocytes and immune cells; detection of hepatocyte function and viability; surface marker and cytokine release of immune cells.
KE3	(intrahepatic) lymphocytes and hepatocytes	What is the influence of supernatant from isolated and IL-2 stimulated immune cells?	Incubation of hepatocytes with supernatant derived from IL-2 stimulated immune cells, analysis of different marker (CYP expression, acute phase protein production, AST/ALT). Comparison of different immune cell pre-culture conditions.
KE3	(intrahepatic) lymphocytes and LSECs	Is there an IL-2 concentration limit in the liver where tolerance induction is overcome?	T cell suppression assay with cells isolated from liver tissue.
AO	Liver resections	How does the hepatic immune landscape differ in patients treated with IL-2 which exhibit hepatotoxicity from healthy control?	Immunohistochemistry or spatial transcriptomic profiling.

Human genetic polymorphisms and murine knock out experiments provide valuable information about the impact of IL-2 on immune-mediated disease and can aid to identify key players and mechanisms in IL-2-mediated hepatotoxicity. In a patient with IL-2Rα deficiency due to a frameshift mutation, lymphocytic infiltrates in various tissues, inter alia in the liver, were observed and the authors suspect this autoimmunity arises from the lack of self-tolerance (Roifman 2000). T_reg cell targeted IL-2Rα conditional knockout mice lack their regulatory function and thus, autoimmune features are observed, such as lymphocytic organ infiltration and liver inflammation (Toomer et al. 2019). Clear evidence for the importance of IL-2Rα on immune-mediated hepatotoxicity may be the occurrence of primary biliary cirrhosis, an autoimmune liver disease in which the host immune system attacks biliary cells, in a 5-year-old with IL-2Rα deficiency. Liver biopsy assessment showed lymphocyte infiltrates in the liver and no signs of infection. The authors speculate the disease to be promoted by autoreactive T-cells via two mechanisms: the up-regulation of the anti-apoptotic protein B-cell lymphoma 2 (bcl-2) during T-cell development and the lack of T_reg cells which normally suppress autoreactive T cells (Aoki et al. 2006). Studies with IL-2Rα deficient mice were applied to gain a closer insight into primary biliary cirrhosis. The livers of the animals showed CD4⁺ T- and predominantly CD8⁺ T-cell infiltrates, while NK and NKT cells were decreased compared to control animals. Regarding cytokines, pro-inflammatory T_H1 cytokines (TNFα, IL-2, IL-12p40, IFNγ) and IL-6 were increased in the serum (Wakabayashi et al. 2006). Regarding the IL-2Rβ chain, an IL-2RB missense mutation (L77P) hinders the egress of IL-2Rβ from the endoplasmic reticulum and consequently, humans with this mutation show autoimmune disease also affecting the liver. Patients with this mutation lack IL-2Rβ surface expression on T-cells while they retain moderate IL-2Rβ expression on NK cells. The CD56^bright NK cell compartment is enhanced and acts cytotoxic upon IL-2 stimulation. As patients lack high-affinity IL-2R expressing T-cells (T_reg cells), this loss of tolerance with concomitant NK cell activation may lead to the observed dysregulations of the immune system (Zhang et al. 2019). This renders NK cells and especially CD56^bright NK cells an interesting cell type to address in the context of aldesleukin-mediated hepatotoxicity. Deficiency in IL-2Rγ leads to X-linked severe combined immunodeficiency and about 200 different mutations have been characterized so far, leading to different severities of the disease (Lim et al. 2019; Belaid et al. 2021).

These publications highlight the importance of a functional T_reg cell compartment to counteract self-targeted immune cells. Simultaneously, a concomitant downmodulation of regulatory function (immune tolerance) with consistent or even promoted effector cell (T-cell and/or NK cell) function might favor immune-mediated diseases such as immune-mediated hepatotoxicity. This is well exemplified by depletion or functional inhibition of T_reg cells in mice with a human immune system: Low dose IL-2 treatment of these animals caused weight loss and pulmonary edema, which are toxicities seen in high-dose IL-2-treated mice with functional T_reg cells (Li et al. 2017).

IL-2 binding (MIE)

IL-2 binds to the IL-2R. Other binding modalities are unknown. The IL-2R is generally expressed on various cell types such as lymphocytes, neutrophils, and monocytes (David et al. 1998; Malek 2008; Boyman and Sprent 2012) (Table 1). The IL-2R constitutes one to three subunits that are differentially expressed and exhibit different IL-2 binding affinities (Stauber et al. 2006). The low-affinity receptor consists of IL-2Rα (CD25; K_d ∼10⁻⁸ M) only and does not generate a signal due to lack of intracellular signaling domains. Both membrane-bound and soluble IL-2Rα (sCD25) can act as decoy receptors. The intermediate-affinity receptor is composed of IL-2Rβ (CD122) and -γ (CD132, K_d ∼10⁻⁹ M) and the high-affinity receptor of IL-2Rα, -β, and -γ (K_d ∼10⁻¹¹ M). Both intermediate- and high-affinity IL-2Rs have intracellular signaling domains, therefore enabling signaling capacity (Figure 2(A)). IL-2 binding to the intermediate-affinity receptor induces IL-2Rα production and subsequent formation of the high-affinity receptor (Baume et al. 1992; Grant et al. 1992; Spolski et al. 2018). Upon binding, the IL-2-IL-2R complex is internalized and the IL-2Rβ and -γ subunits are degraded in endosomes while the -α subunit is recycled to the cell surface (Hémar et al. 1995) (Figure 2(B)).

Figure 2. IL-2 receptor signaling and mechanisms of internalization. (A) The low-affinity receptor consists of IL-2Rα, the intermediate-affinity receptor consists of IL-2Rβ and -γ and the high-affinity receptor of all three subunits. While the low-affinity receptor lacking intracellular signaling domains can act as decoy receptor, the intermediate- and high-affinity receptors signal via MAPK, Pi3K and STAT pathways. (B) Upon IL-2 binding, the ligand-receptor complex is internalized and the IL-2Rβ and –γ subunits are targeted for endosomal degradation, while the –α subunit is recycled to the cell surface. Figure adapted after Spolski et al. (2018).

As described, important immune cells in the healthy liver are myeloid macrophages, lymphoid NK, NKT, and T-cells (Robinson et al. 2016) and for neutrophils, a role in cancer immunotherapy is attributed (Siwicki et al. 2021). Therefore, we give an overview of these cell types in the following.

Murine KCs express all three IL-2R subunits on a protein level and respond to IL-2 treatment in vitro (Simone et al. 2021). In human, neutrophils, monocytes, and differentiated macrophages express the intermediate-affinity IL-2R, while low-affinity IL-2Rα is absent (Scheibenbogen et al. 1992; Djeu et al. 1993; Girard et al. 1995; Abdel-Salam and Ebaid 2014). Notably, IFNγ is a more potent inducer of the IL-2Rγ chain than IL-2 in human monocytes (Bosco et al. 1994b).

Peripheral blood-derived human NK cells express the intermediate-affinity IL-2R, but only a small fraction expresses the α-chain. Expression of the high-affinity receptor occurs mostly in CD56^bright NK cells (Baume et al. 1992; David et al. 1998; Rudnicka et al. 2015) (Table 1). Upon lipopolysaccharide (LPS) activation, IL-2Rα is upregulated on CD56⁺ NK cells (Rudnicka et al. 2015).

NKT cells express both NK cell markers and T-cell receptor-αβ chains representing a bridge between innate and adaptive immunity. During liver inflammation, Type I and Type II NKT cells confer protective and damaging roles, respectively (Bandyopadhyay et al. 2016). Human peripheral blood NKT cells express surface IL-2Rβ more on the mature CD4^- than on the CD4⁺ subpopulation (Baev et al. 2004). In contrast, on in vitro expanded human peripheral blood-derived NKT cells, IL-2Rα expression was observed in a greater fraction of CD4⁺ than of CD4^-CD8^- double negative NKT cells. Analogous to CD4⁺CD25⁺ T_reg cells, a regulatory function for CD4⁺ NKT cells is suggested, as they express FoxP3 and inhibit antigen-specific CD8⁺ T-cells (Osada et al. 2005). Surface expression of IL-2Rα in peripheral blood NKT cells was confirmed, but NKT cell subtypes were not assessed (López et al. 2018). On resting murine NKT cells, all three subunits of the IL-2R were detectable (Jukes et al. 2012) (Table 1). These findings indicate that, as IL-2Rα is expressed at least on several NKT cell subtypes, this cell type might present a primary target of IL-2 and be implicated in the fine-tuning of IL-2 response by immune-activating and -inhibiting actions.

In human peripheral blood-derived T-cells, the intermediate-affinity IL-2R is more distinctly expressed on resting CD8⁺ than on CD4⁺ T-cells (David et al. 1998). The level of IL-2Rα expression influences the differentiation of CD8⁺ T-cells. While IL-2Rα^low CD8⁺ T-cells differentiate to memory-phenotype cells, IL-2Rα^high cells give rise to effector T-cells (Spolski et al. 2018). Further, murine memory-phenotype CD8⁺ T-cells require IL-2 signal for re-activation (Blachère et al. 2006). Taken together, IL-2R expression plays a critical role in CD8⁺ T-cell function, development, and reactivation and due to differing expression of the IL-2R subunits, potentially diverging responsiveness for aldesleukin treatment exist.

T_reg cells control T-cell-mediated immune responses and have important functions in preventing autoimmunity. They act by expressing anti-inflammatory cytokines such as IL-10 and IL-35, via cytolysis and by modulating dendritic cell responses (Vignali et al. 2008). T_reg cells are characterized as high-affinity IL-2R expressing cells (Malek 2008). T_reg cells expressing the trimeric IL-2R have the capability to consume local IL-2, thereby exerting a suppressive function termed metabolic disruption which involves the deprivation of the surrounding effector cells of IL-2 (Vignali et al. 2008). In contrast, in vitro, sCD25 containing supernatant does not interfere with the proliferation of CD4⁺CD25^- T-cells or deplete IL-2. The presence of T_reg cells may therefore be necessary to mediate effector T-cell suppression (Pedersen and Lauritsen 2009). Such a synergistic effect between T_reg cells and sCD25, especially at supraphysiologic concentrations of this molecule, is supported in patients with hepatocellular carcinoma (Cabrera et al. 2010a). In these patients, mean serum levels of sCD25 are increased 5-7-fold compared to healthy controls and high sCD25 levels are associated with low CD4⁺ T-cell reactivity as shown in ex vivo assays. These findings support a possible role for this soluble receptor in immune-inhibition (Cabrera et al. 2010b). In patients with chronic liver disease, sCD25 concentrations positively correlate with pro-inflammatory cytokine/chemokine expression as well as disease progression, thereby qualifying as a potential marker for immune cell activation in this condition (Seidler et al. 2012). Clearly, it would be helpful to understand the role of sCD25 in IL-2-induced hepatotoxicity, which seems not yet to have been addressed (Table 2).

In-depth knowledge about the IL-2 receptor expression in the human liver cell subtypes would be crucial to understand which cells possibly respond directly to high dose aldesleukin treatment. Transcriptomic analysis of human liver specimens showed all three subunits of IL-2R to be expressed (Fagerberg et al. 2013). However, no cell type specific values were indicated (Table 2). In murine KC, all three IL-2R subunits could be detected in flow cytometry analyses (Simone et al. 2021).

Activation of cells by IL-2 (KE1) and release of pro-inflammatory cytokines

Activation of Kupffer cells

In many cases, data about the IL-2 response of hepatic immune cells are missing. The liver contains a high density of myeloid cell types in homeostasis. Among them, KC as liver resident macrophages constitute the largest population of hepatic immune cells. High concentrations of IL-2 prime and trigger KCs to suppress protein synthesis (Curran et al. 1988). Murine KC can be activated by IL-2 to cross-present hepatic antigens to T-cells, thus promoting an immune response (Simone et al. 2021).

Activation of monocytes and neutrophils

Human monocytes are activated by IL-2, which leads to enhanced cytotoxicity (Espinoza-Delgado et al. 1990) as well as cytokine, growth factor, and growth factor receptor expression as previously reviewed (Espinoza‐Delgado et al. 1995). Furthermore, IL-2 induces upregulation of IL-2Rβ and -γ, but not of -α mRNA in these cells (Espinoza‐Delgado et al. 1992; Bosco et al. 1994a), indicating the promotion of intermediate-affinity IL-2R expression. Protein tyrosine kinase activation plays a central role in IL-2 signaling, mediating effector and secretory functions (Bosco et al. 2000). Human peripheral blood monocytes express the pro-inflammatory cytokines IL-1β, −6, −8, and TNFα after IL-2 stimulation (Bosco et al. 2000). In human neutrophils, IL-2 stimulation induces TNFα mRNA (Wei et al. 1993).

Activation of NK cells

In primary human NK cells, IL-2 signaling is well documented. Upon IL-2 binding, high-affinity IL-2R expression and STAT1, −3, −4, and −5 phosphorylation are induced (Wang et al. 1999). Generally, IL-2 stimulation confers protection from apoptosis mediated by upregulation of the anti-apoptotic protein bcl-2 (Armant et al. 1995). IL-2 leads to an increased secretory and cytotoxic response in human NK cells by directly acting on intracellular signaling (Bonnema et al. 1994) which manifests in IFN-γ and TNF-α production, while IL-1β protein is not upregulated (Young and Ortaldo 1987; Numerof et al. 1990). In human NK cells, IL-2 favors proliferation of the immature CD56^bright compared to the cytotoxic CD56^dim population (Baume et al. 1992). IL-2 up-regulates the expression of the activating receptor killer cell immunoglobulin-like receptor 2DL4 (KIR2DL4) on CD56^high cells (Kikuchi-Maki et al. 2003). These findings suggest a high responsiveness of this NK cell sub-population toward IL-2. This may have relevance for hepatic toxicities after IL-2 treatment as in humans, immature NK cells constitute nearly 50% of intrahepatic NK cells (Hudspeth et al. 2016). Moreover, the presence of certain NK cell subpopulations might be predictive for liver injury: Comparison of NK cell subpopulation occurrence from healthy versus diseased liver and found a positive correlation between integrin α-1/IL-2Rα expressing (CD49a⁺CD25⁺) NK cells and alanine aminotransferase levels, indicating liver damage. In vitro, the CD49a⁺CD25⁺ NK cell population proliferates in response to low dose IL-2, indicating a lower activation threshold than in peripheral blood-derived NK cells (Martrus et al. 2017). Elucidation of IL-2 signaling threshold in hepatic NK cells would provide more information on the main IL-2 responders in the liver (Table 2).

Activation of T lymphocytes

IL-2 signaling in T-cells has been extensively reviewed (Benczik and Gaffen 2004; Beyer et al. 2011; Liao et al. 2011a, 2011b; Ross and Cantrell 2018). Rollings et al. (2018) compared the proteome of murine effector CD8⁺ T-cells maintained in IL-2 with samples deprived of IL-2. The authors state that an IL-2 dependent increase in effector molecules goes hand in hand with an upregulation of proteins that control these molecules, thereby regulating T-cell activity. Based on proteomic data gained on cytotoxic lymphocytes, IL-2 mediates cell cycle progression and proliferation. IL-2 sustains IL-2Rα expression in T-cell receptor activated effector CD8⁺ T-cells and IL-2 deprivation leads to downregulation of expression of all three IL-2R subunits. The authors highlight the fine-tuning of IL-2 responses and the critical role of IL-2 signaling not only for the maintenance of T-cell responses in an activated state but also in mediating responsiveness of quiescent T-cells and cell cycle progression (Rollings et al. 2018). With regard to T-cell activation, IL-2 reduces the T-cell receptor signaling threshold in CD8⁺ T- but not in CD4⁺ T-cells. The decreased T-cell receptor threshold in CD8⁺ T-cells may be enabled by increased expression of the cell cycle regulator c-Myc, possibly transforming sub-optimal T-cell receptor stimulation to an immune response to antigen (Au-Yeung et al. 2017). This increased response of CD8⁺ T-cells toward IL-2 may be crucial for the induction of cytotoxicity in the liver by this cell type.

Secondary antigen encounter of CD8⁺ T-cells primed in the absence of CD4⁺ T_H cells leads to CD8⁺ T-cell death. Addition of IL-2 rescues CD8⁺ T-cells by modulating the apoptosis relevant genes bcl-2, Flip, Trail, and DR-5, thereby downregulating the sensitivity to DR-5-mediated death. Moreover, IL-2 restores effector functions and enables secondary expansion. Therefore, IL-2 enables full effector function of CD8⁺ T-cells in the absence of CD4⁺ help (Wolkers et al. 2011). In a murine model in which T-cell priming against HBV is restricted to the liver, primed CD8⁺ T-cells led to dysfunctional cells without cytotoxic potential. In vivo administration of an IL-2/IL-2-antibody complex rescued the defective CD8⁺ T-cells by restoring gene expression levels, measured in wild type control mice. This gene regulation was only observed for antigen specific cells. In human CD8⁺ T-cells isolated from patients with chronic HBV infection, T-cells from immune tolerant patients but not T-cells from immune active patients, showed restored cytotoxic functions when expanded after IL-2 incubation, suggesting a similar mechanism to that in the mouse (Bénéchet et al. 2019). Thus, in the setting of HBV infection, exogenous IL-2 can shift the T-cell response from immune tolerant to immune active. The question remains whether such a mechanism might be implicated in IL-2-mediated hepatotoxicity in cancer patients in whom immune tolerance also arises (Table 2). Overall, the literature points toward a pro-immunogenic role of IL-2 by increasing not only proliferation, but also providing anti-apoptotic cues to helpless T-cells and up-regulating cytotoxic potential by decreasing T-cell receptor stimulation threshold and shifting the immune status from tolerant to active.

The balance of T_H cell subsets (T_H1 versus T_H2) is attributed a role in autoimmune hepatitis (Behfarjam et al. 2017). IL-2 has a multitude of roles in regulating differentiation of T_H1, T_H2, and T_H17 cells. IL-2 induces IL-12Rβ2, enhances responsiveness to IL-12, thereby promoting T_H1 cell differentiation, but down-regulates IL-6Rα and gp130, resulting in decreased IL-6 responsiveness and inhibition of T_H17 cells in primary murine cells. IL-2 regulates T_H2 cell differentiation in a STAT5-dependent manner, acting early at the Il4ra locus to induce IL-4Rα expression (Liao et al. 2008; 2011b). In primary murine CD4⁺ T-cells, IL-2 helps maintain transcriptional flexibility of gene expression in T-follicular helper cells (Oestreich et al. 2012). Therefore, it would be interesting to assess the role of exogenous IL-2 in T_H cell response or in polarization of local cell populations (Table 2).

IL-2 is central to T_reg cell homeostasis by maintaining lineage-specific FoxP3 expression (Zorn et al. 2006). IL-2R signals via STAT5 in T_reg cells. In vitro analysis of T_reg cell function with regard to suppression of effector T-cell proliferation reveals that low dose IL-2 enhances while high dose IL-2 abrogates suppressive capacity of T_reg cells, suggesting a concentration window for optimal T_reg cell stimulation (Moon et al. 2015). In a murine model of biliary fibrosis, depletion of intrahepatic T_reg cells increased intrahepatic CD8⁺ T- cells and aggravated fibrosis, and administration of low dose IL-2/anti-IL-2-antibody complexes led to T_reg cell expansion and suppressed proliferation of CD8⁺ T-cells, ameliorating hepatobiliary injury (Taylor et al. 2018). Whether there is an IL-2 concentration limit in the liver above which tolerance induction is overcome, remains unclear (Table 2).

In NKT cells isolated from PBMC, IL-2 signals via STAT3, −4, −5, and −6 as well as p38 mitogen-activated protein kinase (MAPK) pathway. IL-2 induces NKT proliferation and expression of IFNγ and IL-4, as well as granulocyte-macrophage colony-stimulating factor (GM-CSF) and perforin (Hou et al. 2003; Bessoles et al. 2008). In NKT cells isolated from murine spleen and liver, anergic NKT cells downregulate IL-2Rα expression. Exogenous IL-2 rescues proliferative capacity and reconstitutes expression of IL-4 but not IFNγ (Iyoda et al. 2010). This indicates that exogenous IL-2 is implicated in the fine-tuning of NKT cell response, which might be a critical point for this irAOP.

Hepatic cells. Binding of IL-2 to IL-2R⁺ cells induces IL-2R signaling (KE1). Such IL-2 binding potentially occurs directly in the liver as well, as indicated by rapid uptake of radiolabelled IL-2 in rats (Proleukin^® Product Monograph 2018). Although no specific literature on the effect of IL-2 on LSEC could be identified, information on other endothelial cell types is available providing potential hints for the situation in the liver. Assessment of human and bovine endothelial cell layers indicates enhanced permeability upon IL-2 incubation and up-regulation of IL-2Rα expression (Downie et al. 1992). The human brain microvascular endothelial cell line hCMEC/D3 expresses IL-2Rβ and -γ subunits (mRNA and protein) and -α is inducible upon IL-2 stimulation, leading to expression of pro-inflammatory IL-6 and Monocyte Chemoattractant Protein-1 (MCP-1) (Wylezinski and Hawiger 2016). As LSECs are specialized endothelial cells with distinct functions and properties, similar experiments to those on hCMEC/D3 would be crucial for understanding IL-2 effects in the liver (Table 2).

Inflammatory processes can suppress cytochrome P450 (CYP) enzymes, thereby interfering with drug metabolism and can depict an additional readout for liver damage (de Jong et al. 2020). In freshly isolated rat hepatocytes, IL-2 stimulation downregulates CYP2C11 and CYP3A2 mRNA (Tinel et al. 1995). Non-parenchymal cell fractions such as KC were not addressed in these studies. However, KC are presumably implicated in IL-2-mediated CYP downregulation, as shown in primary human hepatocyte/KC co-cultures (Sunman et al. 2004). Curran et al. (1988) showed that treatment with IL-2 enhances KC response to LPS but is also sufficient to activate KCs at doses > 1 x 10⁴ U/mL (Curran et al. 1988). In high dose IL-2 treated mice, increased activity of phagocytes was measured after 2 h, and these cells were identified to be KC (Nakagawa et al. 1996).

Lipid-lipid interactions laterally organize plasma membranes of cells into microdomains called lipid rafts which can influence membrane-bound receptor signaling by retaining or freeing them from lipid raft domains (reviewed in Varshney et al. (2016) and Levental et al. (2020)). In the murine T cell line CTLL-2, IL-2Rα locates to lipid rafts while IL-2Rβ and –γ are enriched in the soluble phase. The trimeric IL-2R locates to the soluble phase, indicating that IL-2R signaling occurs in the soluble membrane regions (Marmor and Julius 2001). In contrast, in PHA-activated primary human T-cells, the IL-2Rβ predominantly locates to lipid rafts while the –α and –γ subunit are enriched in the plasma membrane. Upon IL-2 stimulation, the IL-2Rβ translocates to the plasma membrane fraction, indicating that the formation of the trimeric receptor is induced (Goebel et al. 2002). These findings indicate that lipid rafts play a role in IL-2R signaling and, potentially, patient response to therapeutic IL-2. Additionally, the differing results obtained by Marmor and Julius (2001) and Goebel et al. (2002) emphasize that the correct cell source should be chosen to establish novel in vitro models translatable to the human setting.

In summary, there are rich data for IL-2-mediated immune cell activation in the periphery and there is information about KC response to IL-2. For a better understanding of the processes in the liver, addressing hepatic immune cells specifically would be crucial (Table 2). We presume that immune cells recruited from the periphery might play a role in IL-2-mediated hepatotoxicity. Therefore, in the next section mechanisms for the recruitment of peripheral immune cells to the liver are addressed.

Recruitment of immune cells to the liver (KE2)

We speculate that IL-2 promotes the infiltration of leukocytes into the liver by locally upregulating chemokines and adhesion molecules (Figure 1). In mice, IL-2 treatment induced significant leukocyte infiltration (Lentsch et al. 1997).

In humans, IL-2 signaling and adhesion processes are presumably closely connected. In the human T-lymphoma cell line HuT-102-B2 that constitutively expresses the high-affinity IL-2R, ICAM-1 and IL-2R are situated in close proximity (Burton et al. 1990) and in super antigen activated CD4⁺ T cells, ICAM-1 co-stimulation induces IL-2 expression (Labuda et al. 1998).

The liver presents unique mechanisms of leukocyte adhesion because selectin-dependent rolling does not precede adhesion in the sinusoids but plays a role in post-sinusoidal venules. In contrast, concomitant P-selectin and ICAM-1 deficiency in mice abolishes lymphocyte adhesion to the liver sinusoids. This not only highlights the differences from other endothelial cell types, but also argues against the hypothesis that trapping of lymphocytes in the sinusoids solely follows sheer physical trapping (Wong et al. 1997).

Immune cell recruitment to the liver has been addressed in various murine settings. Murine in vivo studies showed that under non-inflamed conditions primarily NK, NKT, and memory T-cells migrate to the liver, mirroring populations identified as intrahepatic lymphocytes under homeostatic conditions (Klugewitz et al. 2004). Upon intraperitoneal IL-2 injection in mice, leukocytes, neutrophils, and platelet adhesion occur to liver venules besides leucocytosis and increased neutrophil counts in the circulation. Moreover, in IL-2 treated mice, VCAM-1 and ICAM-1 were upregulated and their binding partners, LFA-1 and VLA-4, on lymphocytes increased (Anderson et al. 1996; Nakagawa et al. 1996; Lentsch et al. 1997). In vivo immunotoxic effects of high dose IL-2 were evaluated histologically in humanized mice and a massive leukocyte infiltration in the liver was found, as well as systemic immune cell activation (Li et al. 2017; Yong et al. 2020). CD4⁺ and CD8⁺ T-cells might also be important in IL-2-mediated toxicities, as depletion of these cell types increases survival (Li et al. 2017). Translation of these effects to humans and dissection of the implicated immune cell types would enhance understanding of the hepatotoxic mechanisms of IL-2 (Table 2).

In liver sinusoids of mice treated with IL-2, adherent lymphocytes, platelets, and neutrophils can be found, which presumably leads to impaired blood flow in the liver (Nakagawa et al. 1996). Generally, the infiltration of neutrophils is associated with liver inflammation (Heymann and Tacke 2016). In the literature, L-selectin and CD11b (integrin α-M)/CD18 (integrin β-2) expression on neutrophils are attributed to an important role in the adhesion and migration of the cells across endothelial cell layers. Human neutrophils upregulate CD11b and CD18 expression upon incubation with IL-2 and adhere to human umbilical vein endothelial cells (HUVECs) (Li et al. 1996; Abdel-Salam and Ebaid 2014). Interestingly, neutrophils isolated from patients after the first and second course of IL-2 therapy showed normal random migration but impaired chemotaxis toward the chemotactic peptide FMLP or zymosan-activated serum. One week after completion of the IL-2 treatment, values reverted to normal. Superoxide production, phagocytosis, bactericidal activity and granule secretion were not influenced (Klempner et al. 1990). These results point toward a potential role of neutrophils in vascular injury (including liver), supporting the findings in the murine setting, but there is a knowledge gap because experiments with LSEC are missing (Table 2).

In IL-2-activated human PBMC, the NK cell fraction adheres to endothelial cells. This effect is directly attributed to IL-2, as blocking of the pro-inflammatory secondary cytokines TNFα, IL-1α or -β by antisera does not inhibit this adhesion (Aronson et al. 1988). Analysis of adhesion molecule expression on IL-2 treated human NK cells indicated up-regulation of CD11a, CD2 (LFA-3R), CD44, CD54 (ICAM-1), and CD58 (LFA-3) (Mäenpää et al. 1993) with phosphorylation and thus activation of CD18 signaling (Umehara et al. 1993). IL-2 activated primary human NK cells also exhibit chemotaxis to MCP-1 and, to a less extent, MCP-2 and MCP-3 (Allavena et al. 1994). This finding is underpinned by increased expression of the MCP-1 receptor CCR2 on IL-2 stimulated NK cells (Polentarutti et al. 1997).

In primary human memory-phenotype CD45RO⁺ lymphocytes, IL-2 induces expression of the chemokine receptors MCP-1 and RANTES and increases chemotactic responses toward MCP-1, −2, −3, MIP-1α, -β, and RANTES (Loetscher et al. 1996).

Tlymphocytes. While trapping of activated CD8⁺ T-cells in the hepatic sinusoids in the context of inflammation is widely documented, there is little known about recruitment of IL-2 activated CD8⁺ T-cells to the liver (Table 2). Murine models indicate that in the context of inflammation, activated CD8⁺ T-cells are preferentially trapped in the liver and undergo consequent apoptosis, thus clearing them from any immune responses (Mehal et al. 1999, 2001). In an adoptive transfer model using transgenic T-cell receptor (TCR) OT1 cells, bearing an ovalbumin (OVA) peptide specific TCR restricted to MHC I, ICAM-1/LFA-1, and VCAM-1/VLA-4 interactions enable CD8⁺ T-cell trapping in the liver. In contrast to ICAM-1-mediated trapping of CD8⁺ T-cells, in which antigen presentation by hepatocytes or LSEC is central, VCAM-1-induced trapping occurs independently of antigen presentation. The authors hypothesize that antigen recognition might enhance affinity of T-cell LFA-1 for ICAM-1 via T-cell receptor signaling. They suggest a model for trapping activated T-cells in the liver initiated by passive interactions of VCAM-1/VLA-4 or VAP-1/VAP-1R, following increased adhesion mediated by antigen stabilized ICAM-1/LFA-1 interaction (John and Crispe 2004). An additional mechanism for CD8⁺ T-cell arrest in the liver sinusoids is provided by non-antigen specific binding to adherent platelets which provide a docking site for T-cells (Guidotti et al. 2015). This might involve a link to IL-2, as increased platelet adhesion to liver sinusoids was measured in IL-2 treated mice (Nakagawa et al. 1996).

In human T_reg cells, IL-2 promotes tissue homing by increasing α4- and β7-integrin expression (Hsu et al. 2018). Assessment of surface marker expression on liver-infiltrating T-cells in inflamed human livers showed that nearly all CCR10⁺CD4⁺ T-cells expressed α4β1- integrin and subpopulations expressed α4β7- integrin (Eksteen et al. 2006). A further role for α4- integrin in liver inflammation involving T_H cells has been shown in murine Concanavalin A liver inflammation. T_H1 cells adhere to post sinusoidal venules via α4 integrin while T_H2 cells rely on VAP-1 (Bonder et al. 2005). In vitro polarized and in vivo injected murine T_H1 and, to a less extent, T_H2 cells adhere to the liver sinusoids in large fractions and locate preferentially to the periportal field (Klugewitz et al. 2002).

Little is known regarding IL-2-induced recruitment of NKT cells to the human liver (Table 2). In this process, LFA-1 expression on non-lymphoid cells plays an essential role, as suggested by LFA-1-deficient mice (Emoto et al. 1999). Excluding ICAM-1 as an interaction partner of LFA-1 on NKT cells, the authors did not provide evidence for alternative LFA-1 ligands. CD1d-reactive NKT cells patrolling the liver sinusoids also express CXCR6, which binds to CXCL16 expressed on LSEC. CXCR6-deficient mice exhibit reduced NKT numbers in the liver sinusoids and diminished severity of Concanavalin A-mediated hepatitis (Geissmann et al. 2005). IL-2 was shown to up-regulate CXCR6 surface expression on CD3⁺ PBMCs (Perna et al. 2009).

Adhesion to LSEC. As described above, on lymphocytes, the adhesion molecules LFA-1, VLA-4 and on neutrophils, CD11b and CD18 are up-regulated upon IL-2 stimulation (Anderson et al. 1996; Abdel-Salam and Ebaid 2014). Consequently, it is essential to address the expression of important immune cell binding partners in the liver, specifically in the LSEC layer. To our knowledge, no information exists on direct effects of IL-2 on LSEC adhesion marker expression. However, LSEC adhesion markers have been described. Human LSEC express VAP-1, while VAP-1 is not present on parenchymal or bile duct cells (Salmi et al. 1993; Lalor et al. 2007). Additionally, upon activation of LSEC with pro-inflammatory TNFα, VCAM-1 and E-selectin are induced, and ICAM-1 is increased. Under static conditions, CD11a, CD18, and VAP-1 mediate adhesion of peripheral lymphocytes, while VCAM-1, ICAM-1, and E-selectin are implicated to a less extent. Upon activation of LSEC with TNFα, additionally ICAM-1 and VCAM-1 confer adhesion. Under physiological flow conditions, few lymphocytes adhere to sinusoidal cells. Pro-inflammatory activation of LSEC leads to increased lymphocyte adhesion with shear-stress dependent increase in adhesion. This adhesion process is only marginally mediated by rolling, rather than immediate arrest, which is most likely conferred by VAP-1, with its monoamine oxidase activity (Lalor et al. 2002). Upon binding of the VAP-1 enzymatic substrate benzylamine, the NF-κB pathway is induced and CXCL8, E-selectin, and ICAM-1 are up-regulated (Lalor et al. 2007).

In the context of IL-2, lymphocyte migration across the LSEC layer has not been addressed to our knowledge (Table 2). Patten et al. (2016) have provided a comprehensive assessment of lymphocyte migration through the primary human endothelial cell layer in vitro. They demonstrated that intracellular crawling represents an important mechanism during transendothelial migration of lymphocytes across the LSEC layer, migration of lymphocytes being facilitated by formation of intracellular vesicles. Antibody blocking experiments showed that ICAM-1 and stabilin-1, as well as PD-L1 but not CXCR3, have important roles in intracellular migration of lymphocytes. Regarding junctional molecules, LSEC express VE-cadherin, JAM-A, and ZO-1, while occludin, claudin-1, and E-cadherin are undetectable. Combination treatment with IFNγ and TNFα leads to a reduction in junctional molecule levels as seen by increased permeability upon concurrent IFNγ and TNFα stimulation. Between healthy and chronic-disease liver tissue, junctional molecule expression is comparable but differs strongly from HUVEC. Expression analyses confirm the differences between LSEC and HUVEC at both baseline and upon IFNγ and TNF-α treatment (Patten et al. 2016).

Taken together, these data provide a rich source of information on lymphocyte recruitment and adhesion to, as well as migration through the liver sinusoids that differs from mechanisms observed in HUVEC. Further research is needed to assess the roles of the respective molecules secondarily induced upon IL-2 stimulation (Table 2).

Immune-mediated hepatocyte damage (KE3)

IL-2 induced damage of hepatocytes constitutes the final key event of the irAOP that precedes the adverse outcome hepatotoxicity, which has already been described.

Inflammatory cytokines TNFα is a pleiotropic cytokine implicated in liver homeostasis and contributes to both hepatocyte injury and proliferation. In rats, high amounts of TNFα, produced during LPS-induced liver injury, act pro-apoptotically, and low concentrations released during acetaminophen-induced liver injury promote cell survival. The authors identified the Hippo/Yap pathway as a central regulator of these pleiotropic effects (Zhao et al. 2020b). Moreover, in mice challenged with different pro-inflammatory stimuli, both the presence and the cell type producing the TNFα determine its pathophysiological effects. Thus, TNFα produced by macrophages and neutrophils, but not from T- and B-cells, was detrimental in septic shock (Grivennikov et al. 2005). IL-6 also has pleiotropic roles in the liver, promoting liver regeneration but also inflammation and cancer (Schmidt-Arras and Rose-John 2016). IL-8 is a pro-inflammatory cytokine and chemoattractant and is up-regulated systemically and locally in the livers of patients with chronic liver diseases (Zimmermann et al. 2011). In human hepatocytes, both IL-6 and IL-8 induce the expression of acute phase proteins and show additive effects, while IL-1β and TNFα lead to a weaker response (Castell et al. 1989; Wigmore et al. 1997). IL-1 type cytokines rather modulate than induce an acute phase response (Bode et al. 2012). Moreover, it was proposed that IL-1β plays a role in liver damage activating local leukocytes and recruiting further immune cells (Barbier et al. 2019) (Figure 3). IFNγ induces cell cycle arrest and p53-independent apoptosis of primary murine hepatocytes (Kano et al. 1997).

Figure 3. Immune-mediated hepatocyte damage induced by IL-2. All depicted mechanisms follow IL-2 stimulation. Neutrophils produce the cytokines IL-1β, -6, and -8 which increase acute phase protein production. TNFα can act pro-apoptotically but also promotes liver regeneration under certain conditions (see text). TNFα is produced by neutrophils, CD4⁺ T-, CD8⁺ T-, NKT, and NK cells. IFNγ inhibits cell cycle progression and induces apoptosis and is produced by CD4⁺ T-, CD8⁺ T-, and NK cells. NK cells induce hepatocyte apoptosis by TRAIL-TRAIL-R interaction. CD8⁺ T-cells induce hepatocyte damage by FasL-Fas interaction and by production of perforin and granzymes. Exogenous IL-2 can both overcome the tolerance induction and impair the inhibitory effect of HSCs. KCs downmodulate protein synthesis and CYP activity. Impaired bilirubin and biliary acid metabolism are an AO of IL-2 therapy in humans, but the key players are not evident from literature. The figure compiles human, murine, and rat data.

PBMCs. Early studies by Mier et al. (1987) on incubation of Hep3B cells with supernatants of IL-2 stimulated PBMCs showed an IL-2 concentration-dependent increase in acute phase protein synthesis, only with unfractionated PBMC and not when incubating Hep3B cells with IL-2 alone. IL-1β was identified as the mediator of an indirect effect of IL-2 on acute phase protein production in liver cells, while roles of IL-1α and TNFα could be excluded by incubation with antisera (Mier et al. 1987). Contrasting to the results obtained by Mier et al. (1987), Wigmore et al. (2002) did not find significant influences of IL-2 incubation on IL-6 and TNFα increase in PBMC supernatants and no effect of these supernatants on hepatic acute phase protein synthesis (Wigmore et al. 2002). Despite comparable IL-2 concentrations and incubation times, three apparent experimental differences may explain the different outcomes. Firstly, Mier et al. (1987) incubated PBMC at a 2-fold higher density (2 x 10⁶ cells/mL compared to 1 x 10⁶ cells.ml). Culturing PBMC at high density has been shown to enhance T- cell responses, for instance toward the CD28 superagonist TGN1412 (Römer et al. 2011). This might also apply to IL-2 stimulated cells. Secondly, Mier et al. (1987) incubated the Hep3B cell line with 1:3 diluted PBMC supernatant, while Wigmore et al. (2002) diluted supernatant 1:10. Lastly, the use of cell lines by Mier et al. (1987) versus primary human hepatocytes by Wigmore et al. (2002) could be a factor. Comparison of these data shows that there is a need to identify the correct culture and stimulation conditions for modeling immune-mediated hepatotoxicity. Additionally, the impact of immune cell subtypes on the observed effects remains elusive (Table 2).

Tlymphocytes. In both primary human CD4⁺ and CD8⁺ T-cells, IL-2 induces IFNγ expression in a dose-dependent manner (Kasahara et al. 1983). Activated CD8⁺ T-cells express the cytotoxic surface molecule FasL and binding FasL on activated T-cells induces apoptosis on Fas expressing hepatocytes. For this antigen-independent hepatocyte killing mechanism, the term collateral damage was proposed (Ando et al. 1997; Crispe et al. 2000; Mehal et al. 2001). In murine naïve CD8⁺ T-cells, there is a dose-dependent increase in perforin and granzyme A, B, and C mRNA with IL-2 stimulation, accompanied by increased cytolytic activity (Table 1). The up-regulation of mRNA expression is independent of the anti-apoptotic and proliferative effects of IL-2 (Janas et al. 2005) (Figure 3).

Also, in antigen-specific responses, Fas/FasL-mediated apoptosis by infiltrating CD8⁺ T-cells plays an important role in hepatotoxicity. In OT-1 mice, Fas-expressing mice showed hepatotoxicity upon OVA peptide activation, while in Fas-deficient OT-1lpr mice, lymphocyte infiltration was comparable but hepatocyte damage was nearly undiscernible. Isolated liver CD8⁺ T-cells are cytotoxic against the hepatoma cell line Hepa1.6 and express high levels of TNFα, perforin, and FasL, highlighting the enhanced cytotoxic activity of liver infiltrating immune cells (Kennedy et al. 2001). The Fas/FasL system supposedly plays a major role in inflammatory conditions such as sepsis. In cecal ligation and puncture induced sepsis in mice, inhibition of Fas/FasL signaling by a Fas fusion protein restored hepatic blood flow, lessened liver damage, and decreased IL-10 production (Chung et al. 2001). These data emphasize the role of CD4⁺ T-cell-derived TNF-α and IFN-γ, and CD8⁺ T-cell FasL-mediated apoptosis in IL-2-induced hepatic injury.

NKT cells contribute to liver injury by influencing the local immune milieu through IL-4, −5, and −13 release, directly acting on hepatocytes via IFNγ and TNFα production, as well as by killing hepatocytes through cell-cell contact dependent mechanisms (Gao et al. 2009). FasL/Fas interaction plays an important role in NKT cell-mediated hepatocyte killing, as shown in murine Concanavalin A-induced liver injury (Takeda et al. 2000). Notably, NKT cells can be non-specifically activated by DMSO to express granzyme B and IFNγ and locate to the liver (Masson et al. 2008). These findings point toward a contribution of antigen-independently activated NKT cells to liver damage, comparable to the collateral damage hypothesis mentioned earlier which remains to be tested for IL-2 (Table 2).

Kupffer cells. KC themselves influence hepatocyte functions, as high IL-2 concentrations added to primary rat KC suppress protein synthesis in rat hepatocytes in co-culture (Curran et al. 1988). In primary human hepatocyte cultures, there is a concentration-dependent decrease in CYP activity in hepatocyte/KC co-cultures stimulated with IL-2, without direct effects of IL-2 (Kashuba et al. 2000; Sunman et al. 2004). Therefore, it would be favorable to assess, whether CYP activity could be an additional marker for IL-2-mediated hepatic injury (Table 2). In hepatocyte/KC co-cultures, IL-2 only produces negligible effects on pro-inflammatory cytokines (Nguyen et al., 2015). Furthermore, data from in vitro co-culture experiments of primary murine KC and lymphocytes under Concanavalin A stimulation suggest that these cell types act synergistically and enhance cytokine expression compared to the isolated cell types (Gantner et al. 1996), but it is not clear how these findings translate to IL-2 (Table 2). These results point toward a complex interaction of pro-inflammatory responses in the liver resident immune cell populations. It is probable that lessons can be learned from interaction of macrophages with lymphocytes since KCs are specialized macrophages and antigen-presenting cells. Therefore, the KC activation state would likely affect the outcome of any IL-2 exposure.

NK cells. IL-2 induces lytic granule convergence in human NK cells - an early step in NK cell cytotoxicity - in a Src kinase dependent manner (James et al. 2013) (Table 1). TNF-Related Apoptosis-Inducing Ligand (TRAIL) binds to the death receptor DR5 which induces apoptosis (Yuan et al. 2018). Liver NK cells, but not peripheral blood NK cells, express low TRAIL, which can be induced with IL-2. Concomitantly, IL-2 stimulated liver NK cells exert cytotoxicity against the HepG2 cell line. Stimulation with IL-2 leads to maintenance of the Killer-cell inhibitory receptors CD158a, -b, and the C-type lectin receptor CD94 in liver and peripheral blood NK cells. Well-differentiated hepatocellular carcinoma cells weakly express the death inducing receptors TRAIL-R1/DR4 and -R2/DR5 as well as the non-signaling –R3/DcR1 and –R4/DcR2, while normal hepatocytes lack these receptors. At the same time, endothelial cells express these receptors more abundantly. Poorly differentiated hepatocellular carcinoma and HepG2 cells express the TRAIL-R1/–DR4 and TRAIL-R2/–DR5 but lack expression of the non-signaling death receptors (Ishiyama et al. 2006). As described above, CD56^bright NK cells are a major NK cell sub-population in the human liver (Hudspeth et al. 2016). In chronic HBV-related liver cirrhosis, TRAIL⁺CD56^bright NK cells are associated with the degree of fibrosis and CXCR3-related chemokine levels (Jiang et al. 2021) highlighting the hepatotoxic relevance of this cell type (Figure 3). These findings indicate that IL-2 may activate liver NK cells to promote hepatic inflammation and kill hepatocytes, mainly via TRAIL.

LSEC exert antigen-presenting cell functions and despite being highly capable of taking up antigens, they are alone poor stimulators of an immune response (Katz et al. 2004). LSEC rather promote liver tolerance by PD-L1 interaction with PD-1 on T-cells. This tolerance can be overcome by anti-CD28 antibody addition (Diehl et al. 2008). In primary murine OT-1 restricted T-cells, exogenous IL-2 can overcome the tolerance induction by LSECs during first contact with naïve CD8⁺ T-cells and during T-cell priming, even in the absence of co-stimulation by CD80/CD86, resulting in CD8⁺ T-cell effector function, such as IFNγ production (Figure 3) (Schurich et al. 2010). Whether IL-2 can break the LSEC tolerance in humans remains unclear (Table 2). Furthermore, with regard to CD4⁺ T-cells, murine LSEC induce differentiation of naïve CD4⁺ T-cells to T_H0 cells but not to T_H1 cells, thereby contributing to liver immune tolerance. This effect is not dominant as it can be overcome by other antigen-presenting cells, for instance in the case of non-specific lymphocyte recruitment to the liver (Knolle et al. 1999). In a murine antigen-specific T_H1 cell-inducible liver injury model, the late effector phase of acute liver injury by T_H1 cells is mediated by TNFα and IFNγ. Fas and perforin were not found to play a role in this hepatitis model (Ohta et al. 2000). Translation of these effects to IL-2 would indicate whether T_H1 cell-mediated liver injury also plays a role in this irAOP (Table 2).

HSC down-modulate CD8⁺ T-cell activation by trans-interaction (i.e., as third-party modulators) of ICAM-1 on HSC with its ligand LFA-1 on CD8⁺ T-cells. Knockout experiments in mice indicate that ICAM-1 is up-regulated on HSC and mediates down-regulation of IL-2Rα expression and thus, the IL-2 response. Exogenous IL-2 antagonizes this inhibitory effect of HSC in a dose-dependent manner (Schildberg et al. 2011) (Figure 3). These results indicate that IL-2 can contribute to breaking the tolerogenic potential of the liver, but further research remains to provide evidence in the human setting (Table 2).

In conclusion, there are a variety of possible mechanisms for IL-2-induced immune-mediated hepatic damage (Figure 3). Research is needed to characterize specifically the effect of IL-2 in these processes, which would aid to validate or reject the hypotheses made in this putative irAOP and to sharpen the picture of the irAOP. Regarding the mechanisms retrieved from the literature, IL-2-mediated hepatotoxicity can be assigned to indirect DILI. Possibly, the elucidation of mechanisms in IL-2-induced liver damage can be extrapolated to checkpoint-inhibitor-mediated indirect DILI and vice-versa.

Identification of knowledge gaps and biorelevant molecules with the irAOP approach

The development of a hypothetical irAOP for IL-2-mediated hepatotoxicity unveiled several knowledge gaps of central importance. Addressing these gaps is essential, as it would not only enhance our comprehension of the underlying mechanisms but is also pivotal for validating this potential irAOP. In Table 2, the identified gaps and potential methods that could be applied to fill the gaps, are summarized. Primarily, there is a lack of data that can only be generated with experiments on human hepatocytes and/or liver resident cells (e.g., LSEC, KC), which is however difficult to achieve due to the low availability of these primary human cell types. Central experiments to the validation of the irAOP would be the elucidation of IL-2R expression on non-parenchymal cells in the liver to identify cells capable of responding directly to IL-2 signaling. Secondly, the literature does not provide sufficient information on IL-2-mediated recruitment and migration of immune cells to liver sinusoids and into the perisinusoidal space. To achieve this, use of human LSEC and IL-2-activated immune cells, especially in flow-based settings, could provide valuable information on the situation in the liver. For this, more sophisticated experimental approaches recapitulating the physiology of the liver would be desirable, such as organ-on-a-chip models with perfused immune cells. Setups are already available that model liver sinusoids (Du et al. 2017) or that have been used for determination of liver dysfunction upon treatment with a sepsis-related protein (Blaurock-Möller et al. 2019). These models highlight the capacity of liver microphysiological systems to model liver toxicities. However, such organ-on-a-chip approaches would need adaptation to fill the knowledge gaps by selecting adequate readouts provided by the irAOP (Table 2) and by application of primary human material - ideally, the use of induced pluripotent stem cells would allow fully autologous microphysiological systems (Ramme et al. 2019). Regarding the latter KE of the irAOP, it would be helpful to compare healthy human liver samples to resections from IL-2 treated patients who have clear hepatotoxicity. High-definition spatial transcriptomic profiling combines single cell RNA sequencing with positional information in tissues in situ (Vickovic et al. 2019). This method could provide valuable information, not only on the transcriptional landscape, but could also aid identification of the location of immune cell sub-populations in the liver. Overall, it seems that irAOPs are a suitable tool to identify knowledge gaps and biorelevant molecules and to link AO to both their MIE and the associated KE for IL-2-mediated hepatotoxicity (Table 1, Table 2).

Besides IL-2-specific mechanisms, the hypothetical irAOP presented here also highlights common hepatotoxicity mechanisms (e.g., acute phase protein production, hepatocyte damage by cytotoxic molecules, impaired bilirubin or bile acid metabolism). Hence, models suggested here can potentially be used as blueprints for the evaluation of novel IL-2 drug candidates or IL-2 combination therapies with similar AOs. This irAOP concept may also be applicable for other biologicals, low molecular-weight therapeutics, or RNA-based drugs, if these molecules share common KEs with those described in this putative irAOP. The next step would be to translate this irAOP to other drugs known to cause immune-mediated hepatotoxicity, such as TNFα inhibitors (French et al. 2016) or checkpoint inhibitors (Zen and Yeh 2018) and to evaluate to what extent transferability is possible. The overall aim would be to generate a compound-agnostic irAOP for immune-mediated DILI.

Conclusion

IL-2 is a well-characterized molecule with known mechanisms of action. Nevertheless, many missing links hinder the extensive understanding of IL-2-mediated hepatotoxicity. The irAOP concept described here helps to identify these knowledge gaps and provides starting points for further research. The relevance of such attempts becomes clear as currently, modifications of IL-2 and/or IL-2 combination therapies are under investigation to increase efficacy and reduce toxicity of conventional IL-2 therapy (Overwijk et al. 2021). To augment the chances of success for such therapies, understanding and correct modeling of IL-2-mediated hepatic toxicities is crucial. Therefore, the inclusion of relevant immune cell fractions and exploitation of key event-defining molecules as readouts would provide meaningful and targeted in vitro hepatotoxicity models which could facilitate reduced use of in vivo preclinical models.

Abbreviations
AO	=	Adverse Outcome
Bcl-2	=	B-cell lymphoma 2
CYP	=	Cytochrome P450
DILI	=	Drug-induced liver injury
Eomes	=	Eomesodermin
FPKM	=	Fragments Per Kilobase Million
GM-CSF	=	Granulocyte-macrophage colony-stimulating factor
HBV	=	Hepatitis B virus
HSC	=	Hepatic Stellate Cell
HUVEC	=	Human Umbilical Vein Endothelial Cell
IL-2	=	Interleukin-2
IL-2R	=	IL-2 receptor
imSAVAR	=	Immune Safety Avatar
irAOP	=	Immune-related Adverse Outcome Pathway
KE	=	Key Event
KIR2DL4	=	Killer cell immunoglobulin-like receptor 2DL4
KC	=	Kupffer Cell
LPS	=	Lipopolysaccharide
LSEC	=	Liver Sinusoidal Endothelial Cell
MAPK	=	Mitogen-Activated Protein Kinase
MCP-1	=	Monocyte Chemoattractant Protein-1
MIE	=	Molecular Initiating Event
NKG2D	=	Natural Killer Group 2 Member D
NKT	=	Natural Killer T Cell
OVA	=	Ovalbumin
PBMC	=	Peripheral blood mononuclear cell
sCD25	=	soluble IL-2Rα
TCR	=	T cell receptor
Th	=	T helper
TNF-α	=	Tumor necrosis factor alpha
TRAIL	=	TNF-Related Apoptosis-Inducing Ligand
Treg	=	CD4+ regulatory T

Acknowledgements

The JU receives support from the European Union′s Horizon 2020 research and innovation programme and EFPIA and JDRF INTERNATIONAL. The project was supported by the Landesoffensive zur Entwicklung wissenschaftlich-ökonomischer Exzellenz (LOEWE) Centre Novel Drug Targets against Poverty-Related and Neglected Tropical Infectious Diseases (DRUID), the LOEWE Centre Translational Biodiversity Genomics (TBG) and the Fraunhofer Cluster of Excellence Immune-mediated diseases (CIMD).

Disclosure statement

MJP is an employee of EpiEndo Pharmaceuticals ehf.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/1547691X.2024.2342736)

Additional information

Funding

This project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement [No 853988].