INCB024360

T cells and monocyte-derived myeloid cells mediate immunotherapy-related hepatitis in a mouse model

Heather P. Llewellyn1, Seda Arat2, Jingjin Gao3, Ji Wen3, Shuhua Xia2, Dalia Kalabat2, Elias Oziolor2, Richard Virgen-Slane1, Timothy Affolter4, Changhua Ji5,*

Abstarct

Background & Aims: Immune checkpoint inhibitors (ICIs) are associated with immune-related adverse events (irAEs) which are more severe when ICIs are used in combination. We aimed to use a mouse model to elucidate the molecular mechanisms of immune-related hepatitis, one of the common irAEs associated with ICIs.
Methods: Immune phenotyping and molecular profiling were performed on Pdcd1-/- mice treated with anti-CTLA4 and/or the IDO1 inhibitor epacadostat or a 4-1BB agonistic antibody.
Results: ICI combination-induced hepatitis and 4-1BB agonist- mediated hepatitis share similar features yet maintain distinct immune signatures. Both were characterized by an expansion of periportal infiltrates and pan-zonal inflammation albeit with different morphologic characteristics. In both cases, infiltrates were predominantly CD4+ and CD8+ T cells with upregulated T cell activation markers, ICOS and CD44. Depletion of CD8+ T cells abolished ICI-mediated hepatitis. Single-cell transcriptomics revealed that the hepatitis induced by combination ICIs is asso- ciated with a robust immune activation signature in all subtypes of T cells and T helper 1 skewing. Expression profiling revealed a central role for IFNc and liver monocyte-derived macrophages in promoting a pro-inflammatory T cell response to ICI combination and 4-1BB agonism.
Conclusion: We developed a novel mouse model which offers significant value in yielding deeper mechanistic insight into immune-mediated liver toxicity associated with various immunotherapies.
Lay summary: Hepatitis is one of the common immune-related adverse events in cancer patients receiving immune checkpoint inhibitor (ICI) therapy. The mechanisms of ICI-induced hepatitis are not well understood. In this paper, we identify key molecular mechanisms mediating immune intracellular crosstalk between liver T cells and macrophages in response to ICI in a mouse model.

Keywords: Liver injury; immune-related adverse events; ipilimumab; nivolumab; IDO1; 4-1BB; CTLA-4; PD-1.

Introduction

Immune checkpoint inhibitors (ICIs) targeting CTLA4 (cytotoxic T-lymphocyte-associated protein 4), PD1 (programmed cell death 1) or IDO1 (indoleamine 2,3-dioxygenase 1) demonstrated antitumor efficacy in preclinical models and humans across several types of cancers.1–3 Normally, these inhibitory pathways, or checkpoints, maintain the balance between T cell activation and inhibition. Disruption of these immune checkpoints can result in enhanced antitumor immune responses but may also cause undesirable off-target immune and inflammatory events known as immune-related adverse events (irAEs).4–6 Hepatitis is one such ICI-associated irAE. ICI-induced hepatitis occurs at incidence rates of 1–3% for PD1 inhibitors and 3–9% for CTLA4 inhibitors.7,8 Combining CTLA4 inhibitors with PD1 inhibitors substantially increases the risk of hepatotoxicity.5,9 A higher rate of hepatitis was also seen when the CTLA4 inhibitor ipilimumab was combined with an IDO1 inhibitor, epacadostat, in patients with cancer.10
CTLA4 is primarily expressed on CD4+ and CD8+ T cells in humans and mice.11 CTLA4 transmits an inhibitory signal to activated T cells by inhibiting the binding of CD28 on T cells to CD80/86 on antigen presenting cells (APCs), thereby blocking the second signal required for T cell activation. Genetic deletion of Ctla4 in mice leads to generalized lymphoproliferative disorder and multi-tissue (including the liver) accumulation of self- reactive T cells,15,16 suggestive of a loss of immune tolerance. Similar immunological changes and disease presentations were also observed in patients treated with CTLA4 block- ing antibodies.17
In contrast to CTLA4, PD1 is believed to play an important role in keeping T cell activity in check in peripheral tissues, thereby maintaining immunologic tolerance. Upregulation of PD1 negatively regulates T cell receptor signaling upon binding to one of its ligands, PD-L1 or PD-L2.12 PD-L1 is expressed on hepatocytes, hepatic stellate cells, liver sinusoidal endothelial cells and Kupffer cells, and PD-L2 is expressed on liver sinusoidal endothelial cells, Kupffer cells, and intra- hepatic leukocytes.11
The immune modulator IDO1 is an intracellular enzyme that degrades L-tryptophan and generates kynurenine. Decreased L- tryptophan can inhibit T cell activation and proliferation, and kynurenine promotes regulatory T cell (Treg) formation and ac- tivity. IDO1 can be induced in the liver by inflammatory stimuli.13 Hepatic stellate cells can induce tolerogenic dendritic cells (DCs) by inducing IDO1 expression.14 Furthermore, liver injury stimuli can promote inflammation in Ido1-/- mice.13,15
ICI-associated irAEs have phenotypic and morphologic char- acteristics that are generally reminiscent of autoimmune dis- eases. However, the exact immune mechanisms of irAEs have not been substantially elucidated.6 This is largely attributable to the lack of animal models exhibiting ICI-induced irAEs. Single-agent ICIs rarely cause irAEs in preclinical toxicology models. IrAEs have been reported in cynomolgus monkeys when a CTLA4 in- hibitor (ipilimumab) and a PD1 inhibitor (nivolumab) were co-administered.16,17
We have recently reported on a mouse model of ICI-induced hepatitis using a Pdcd1-/- mouse. This Pdcd1-/- mouse showed no hepatitis under physiological conditions, but developed mononuclear hepatitis following the administration of a mouse CTLA4 inhibitory antibody, clone 9D9, or an IDO1 inhibitor, epacadostat.18 To elucidate the immune mechanisms of ICI- induced hepatitis in the Pdcd1-/- mouse model, we conducted T cell depletion, bulk RNA sequencing (RNAseq) and single-cell RNASeq (scRNAseq) analyses on the livers of these mice.

Materials and Methods

Detailed material and methods can be found in the supplementary materials & methods and CTAT Table.

Pdcd1-/- mice have been described previously.19 Anti-CTLA4 antibody (clone 9D9) was administered intravenously at 300 lg/dose 2 to 4 days prior to the start of the study, then weekly. The IDO1 inhibitor, epacadostat, was administered by oral gavage at 600 or 300 mg/kg/dose, twice daily. Anti-(4-1BB) agonistic antibody (clone 3H3) was administered intraperitoneally to mice twice weekly. To deplete CD8+ T cells, anti-CD8 antibody (clone YTS 169.4) was administered intraperitoneally twice weekly. scRNAseq was performed using liver CD45+ cells. The Seurat object containing the cell identities, treatment groups and gene expression matrix was then used as an input for differential gene expression, CellPhoneDB and Monocle3.

Results

ICI-induced T cell infiltration, activation and CD8+ T cell- dependent liver injury

Our previous work demonstrated that Pdcd1-/- mice can be uti- lized as a model for ICI-induced liver injury.18 Pdcd1-/- mice have increased liver CD8+ T cells under physiological conditions18,20 and increased monocyte activation (Fig. S1), whereas short- term administration of anti-PD1 did not result in immune cell composition changes in the liver (Fig. S1). Additionally, we did not observe evidence of liver injury in Pdcd1-/- mice or with anti- PD1 administration to B6 mice, consistent with previous re- ports,18,21 suggesting that there are multiple pathways that maintain tolerance in the liver. Also consistent with prior find- ings,18 2 weeks of co-administration of a CTLA4 blocking anti- body, clone 9D9, and an IDO1 inhibitor, epacadostat, increased periportal infiltrates and pan-lobular inflammation in the livers (Fig. 1A, 1B) of Pdcd1-/- mice. GLDH, a specific and temporally sensitive liver injury biomarker,22,23 was significantly elevated in the anti-CTLA4 + epacadostat combination group relative to isotype control, anti-CTLA4 or epacadostat treatment groups (Fig. 1C). Combination treatment significantly increased the numbers of conventional CD4+ T cells, Tregs and CD8+ T cells in the liver (Fig. 1D). Representative gating strategy and FMO con- trols are shown in Fig. S1. The proportion of proliferating (Ki- 67+) CD4+ and CD8+ T cells was increased in both epacadostat- and combination-treated mice (Fig. 1E). These data suggest that blocking CTLA4 may promote T cell trafficking to the liver, whereas inhibition of IDO1 is consistent with enabling T cell proliferation.18,24 In addition, combination treatment increased the frequency of the activation marker ICOS and CD44 on CD4+ and CD8+ T cells in the liver and spleen (Fig. 1F, 1G).
To determine if CD8+ T cells mediate ICI-induced hepatotox- icity, a T cell depletion experiment was performed. Depletion of CD8+ T cells attenuated liver injury (Fig. 1H), periportal infiltra- tion and pan-lobular inflammation (Fig. 1I) in ICI combination- treated Pdcd1-/- mice.

ICIs enhance immune cell trafficking and phenotypic changes specific to both pan-zonal and periportal areas

In this hepatitis mouse model, large lymphoid aggregates were primarily observed in the periportal region and ring-like clusters of lymphocytes and monocytes were primarily found in pan- zonal areas.18 To understand the specific spatial response to checkpoint blockade, foci of lymphocyte infiltration pan-zonally or in periportal regions were isolated from control- and ICI combination-treated Pdcd1-/- livers by laser capture microscopy followed by transcriptome analysis. Genes associated with DC activation (Fscn1, Ccr7, Cacnb3, Slco5a1) and the CCL19/CCR7 axis (Ccl19, Ccr7, Nsg2) are upregulated in periportal regions of treated mice (Fig. 2A), suggesting that these areas are function- ally similar to tertiary lymphoid structures. In the pan-zonal infiltration sites, higher Ltf expression indicates the presence of activated APCs25; increased Piezo2 expression suggests rapid cell migration26 throughout the liver parenchyma; and increased expression of Nrxn1, Reln, Fbln1, Atcay, Scin indicates potential extracellular matrix remodeling to facilitate leuko- cyte trafficking.
In the periportal infiltration foci, immune-related response and proliferation pathways were among the significantly upre- gulated pathways (adjusted p value <0.05), and liver metabolism-related pathways were downregulated in combination-treated mice relative to the isotype control group (Fig. 2B). T cell activation and chemokine/chemokine receptor genes were also increased in periportal areas following combi- nation treatment (Fig. 2C). Deconvolution of the RNAseq data for liver parenchymal and non-parenchymal cells suggests that in the periportal region, combination treatment increased the proportion of c/d T cells, stromal cells, CD4+ a/b T cells, plas- macytoid DCs, T helper (Th)2 cells, and Th1 cells and decreased the proportion of hepatocytes, neutrophils, and plasma cells (Fig. 2D). Altogether, these data suggest that the periportal re- gion is likely functioning as a tertiary lymphoid structure, allowing for DC and T cell differentiation, while lymphocytes (along with associated monocytes) rapidly migrate to the pan- zonal areas to form inflammatory foci, leading to variable he- patocyte necrosis.
Single-cell atlas of liver immune cells in ICI-induced hepatitis To gain deeper insight into the immune cell populations and their transcriptional changes during ICI treatment, scRNAseq was performed on CD45+ enriched liver cells. A total of15,651, 16,434, 18,298, and 18,651 cells were analyzed from isotype control, anti-CTLA4, epacadostat and anti-CTLA4 + epacadostat groups, respectively. Samples from each group were evenly distributed throughout all clusters (Fig. S2A), suggesting that each cluster has uniform representation of the samples. Clusters with high nFeature RNA and low percent mitochondria were subsequently analyzed (Fig. S2B). T-distributed stochastic neighbor embedding visualization of the aggregated samples revealed 10 major cell types corre- sponding to monocytes (includes monocyte-derived macrophages [MDMs]), T cells, Kupffer cells, B cells, neutro- phils, dendritic cells, endothelial cells, natural killer (NK) cells, mast cells and platelets (Fig. S2C). Clusters were iden- tified based on marker gene expression (Fig. S2D, 2E, 3, 4) and their top differentially expressed genes (Table S1). The T cell and monocyte cell types were expanded in the combi- nation group (Fig. S2C), prompting further exploration of these cell types.

ICI combination promotes CD8+ T cell proliferation and activation and IFNc gene signature

The T cell cluster can be further subdivided into CD4+ T cells, regulatory T cells, NKT cells and 4 subtypes of CD8 T cells (Fig. 3A, Fig. S3). The CD8+ effector (Eff) subtype expresses markers of effector T cells (Gzma, Lag3 and Tigit), the CD8+ Prolif subtype expresses makers of proliferating cells (Ki67, Top2a), the CD8+ effector memory (EM) subtype expresses markers of effector memory T cells (Cx3cr1) and CD8 central memory (CM)/naïve expresses markers of central memory/naïve T cells (Il7r, Tcf7, Ccr7, Lef2, Sell) (Fig. S3B, 3C, 3D). ICI combination-treated mouse liver contains a larger proportion of the proliferating CD8 T cells (Fig. 3B). To determine the relationship between the CD8+ T cell clusters, scRNAseq data was analyzed with the monocle package27 and the basal state was set to the naïve population (Fig. 3C, 3D). Pseudotime analysis suggests that the CD8+ T cell differentiation program progresses (purple to yellow), from naïve cells to EM or Eff T cells. Then, Eff T cells give rise to proliferating T cells in response to checkpoint inhibitors. Consistent with the mechanism of ICIs promoting T cell prolif- eration, these data suggest that CD8+ T cells are re-entering a proliferative state after differentiation.
Differential gene expression revealed distinct changes in gene expression in each T cell cluster with each treatment relative to isotype control (Fig. 3E). A signature of T cell effector and acti- vation genes including Gzmb, positive signaling molecules Tnfrsf18 (gene for GITR), and Tnfrsf9 (gene for 4-1BB) in the CD8+ EM cluster were synergistically upregulated with combination treatment relative to isotype control. Gzmb and Tnfrsf18 were also upregulated in the CD4+ T cell cluster. Negative signaling molecules (Ctla4, Pdcd1, Lag3, Tigit, Cd38) were also increased in the CD8+ EM and CD4+ T cell clusters with combination treat- ment. In the CD8+ CM/naïve cluster, Cd38, Pdcd1lg2 and Entpd1 (inhibitory signaling pathways) were synergistically decreased, while Eomes, Cd44, Ifng and Tbx21 (activation and differentiation pathways) were increased with combination treatment. Lastly, Ccr7 was downregulated in the Treg, CD8+ Prolif, and CD8+ Eff clusters, potentially as a mechanism to retain these cells in the liver. The oxidative phosphorylation pathway was the top enriched pathway in the Hallmark gene set collection in all clusters, except for Tregs with combination treatment (Fig. 3F), suggesting that checkpoint inhibitors boost metabolism to meet the energy demand for T cell activation. An IFNc gene signature associated with tumor response to ICIs28,29 was synergistically activated in each of the T cell subtypes with combination treat- ment (Fig. 3G), suggesting that this signature is also associated with ICI-induced hepatotoxicity. The CD4+ T cell cluster was also enriched for Th1 genes (Fig. S3E), consistent with the RNAseq data from laser capture microscopy sections (Fig. 2D), further suggesting that CD4+ T cells are skewed towards a Th1 phenotype.

ICI combination promotes an IFNc response via monocyte- derived macrophages

We further investigated the monocyte/monocyte-derived cells as they were also expanded with ICI treatment (Fig. 4A, 4B, Fig. S2C). The monocyte/monocyte-derived clusters are distinct from conventional DC populations (cDC1, cDC2, plasmacytoid DC, Ccr7+ DC) and Kupffer cell clusters (Fig. S4). Relative to the other monocyte subtypes, the monocyte-derived macrophage-1 (MDM-1) cluster expressed high levels of Cxcl9, Cxcl10, MHCII genes, and C1q genes (Fig. S4A, 4D, 4E), suggesting this cluster may be a hybrid monocyte-derived macrophage/DC population. Furthermore, the MDM-1 cluster expressed Csf1r, Vcam1, but also Itgax (CD11c) and exhibited low expression of F4/80 (Adgre1) (Fig. S4A, 5B). MDM-2 shared a similar gene expression profile with MDM-1 but also expressed high levels of Itgax, Mertk, Vcam1, Saa3, Cd63, Ccl5 and Nos2 (Fig. S4D, 4E), suggesting MDM-2 may be a more differentiated state of MDM-1. MDM-1 and MDM-2 are enriched for oxidative phosphorylation, glycol- ysis and antigen processing and presentation pathways, and low for phagocytosis (Fig. S5A). The classical monocyte (class. Mo) subtype expressed high levels of classical monocyte markers Chil3, Fn1, Ly6c2 as well as Ccr2. The non-classical monocyte (non-class. Mo) cluster expressed markers of non-classical monocytes (Fcgr4, Cx3cr1, and Treml4) and are Ly6c2- (Fig. S4D, 4E). ICI combination treatment increased the percentage of the MDM-1 subtype (Fig. 4B).
MDM-1 and MDM-2 express Ifngr2 (Fig. S5) and are enriched for the Hallmark IFNc response pathway (Fig. 4C). IFNc response was increased with combination treatment (Fig. 4D), suggesting that the MDM-1 and MDM-2 cells respond to IFNc produced by T cells during ICI therapy and are consistent with the phenotype of a classically activated MDM.30 Similar to T cells (Fig. 3F), the oxidative phosphorylation gene set was upregulated with com- bination treatment in the MDM-1 cluster (Fig. 4D). Consistent with IDO1 inhibition, the AHR pathway was downregulated in MDM-1 in epacadostat- and combination-treated mice (Fig. S6). Cytokine, co-stimulatory, and chemokine genes from the IFNc response pathway (Batf2, Ccl5, Cd40, Cxcl9, Fcgr1) and other IFNc responsive genes (Il18, Cxcl16, Ifnar2) were upregulated with combination treatment in the MDM-1 cluster (Fig. 4E), sugges- tive of an enhanced APC phenotype.
CellPhoneDB analysis, a tool that infers inter- and para- cellular communication from combined expression of multi- subunit ligand-receptor complexes between cell clusters, was performed to determine the ligand and receptor interactions between myeloid and T cells in the liver. Analysis of the major immune populations identified the monocyte/monocyte-derived cells as clusters that had a high number of interactions with all other cell types (Fig. 5A). Displayed interactions (Fig. 5B) were selected based on their known biology in relation to T cell acti- vation, and if there was a significant difference between control and combination treatment according to the variance analysis (Table S3). Significant ligand/receptor pairs with T cell activation cytokines TGFB, IL21, IL1831 and IL1532 were found between myeloid and T cells and between T cell subsets in the combina- tion group mice. These data suggest that myeloid cells promote T cell activation in the context of checkpoint inhibition. Recipro- cally, the CSF/CSF1r interaction is significantly upregulated by T cell and monocyte clusters, suggesting that T cells may be pro- moting monocyte-derived cell activation and/or differentiation. Combination treatment also significantly increased the ligand/ receptor pair interactions for co-inhibitory pathways between myeloid cells and T cells, including TIGIT and CD20033 (Fig. 5B). Overall, our data suggest that close interactions between monocytes/monocyte-derived cells and T cells plays a key role in the mouse model of ICI-induced hepatitis.

ICI- and 4-1BB agonist-induced liver injury exhibit similar yet distinct immune signatures

4-1BB is expressed on antigen-experienced T cells and agonistic antibodies are known to cause liver injury in mice by promoting T cell activation.21 We next sought to compare the immune characteristics between ICI combination and anti-41BB-mediated liver injury (Fig. 6A). The incidence and severity of liver in- filtrates and inflammation was similarly increased in the ICI combination- and anti-41BB-treated mice (Fig. 6B, 6C). In anti- 41BB-treated livers, the inflammatory pattern and necrosis was more haphazardly arranged and lacked the annular association of mononuclear cells surrounding necrotic hepatocytes exhibited with the ICI combination (Fig. 6B). ICI combination-treated livers displayed prominent portal lymphoproliferation and the previ- ously described annular lesions associated with necrotic cells, monocytes, and occasional foamy macrophages. GLDH activity was significantly increased with ICI combination at day 25 (Fig. 6D). Flow cytometric analysis demonstrated that 4-1BB agonism induces several activation markers (CD69, EOMES, KLRG1) on CD4+ and CD8+ T cells, consistent with previous re- ports.21 However, these activation markers were not induced by ICI combination treatment (Fig. 6E). ScRNAseq was performed on CD45+ enriched liver cells (Fig. S7). While ICI combination treatment increased the percentage of cells in the CD8+ Eff cluster, 4-1BB agonism increased the percentage of cells in the CD8+ Prolif and CD8+ EM clusters (Fig. 6F). 4-1BB agonism increased distinct T cell activation genes in the CD8+ EM cluster (Eomes, Gzma, Tox, Tnfrsf18, Vsir, Klrg1) and in the CD8+ naïve/CM cluster (Tnfrsf18, Ctla4, Ifng, Tnfsf4, Cd44, Icos), whereas a different profile of T cell activation genes were increased in the CD8+ EM cluster (Havcr2, Gzmb, Lag3, Pdcd1, Ctla4, Tnfrsf4), the CD8+ naïve/CM cluster (Vsir, Cd274, Cd38) and the CD4+/Treg cluster (Tigit, Tnfrsf9, Havcr2, Gzmb, Lag3, Pdcd1, Tox, Prf1, Tnfrsf18, Il2rb, Pdcd1lg2) in ICI combination-treated mice (Fig. 6G). Furthermore, 4-1BB agonism increased the IFNc gene signature in all T cell subsets by a greater magnitude than ICI combination treatment (Fig. 6H).
We next evaluated the monocyte, MDM and Kupffer cell (KC) transcriptional response in ICI combination- and 4-1BB agonist- treated mice at 4 weeks. In 4-1BB agonist-treated Pdcd1-/- mice, the proportion of MDM and Class. Mo cells were increased, whereas there was no change in these populations between ICI combination-treated and isotype control-treated mice (Fig. 7A). These monocyte clusters had enrichment of the IFNc response gene set, which was more substantial in anti-41BB-treated mice compared with ICI combination-treated mice (Fig. 7B). IFNc response genes shown in Fig. 7C suggest that the MDM and KC-1 clusters are responding similarly to anti-41BB treatment whereas genes such as Cd40, Cxcl9, Cxcl10, and Cxcl16 that were signifi- cantly increased at 2 weeks (Fig. 4E), were no longer upregulated in the MDM cluster at 4 weeks. 4-1BB agonism has been shown to activate macrophages in the liver.34 Consistent with their findings, Il27, MHCII, Ccr5 genes were upregulated in the MDM cluster and B2m was upregulated in the KC-1 cluster (Fig. 7D). Tnfrsf9 and Ifng were not expressed or had very low expression in these clusters. Altogether, transcriptomic analysis suggested a role for MDMs in orchestrating the T cell response in ICI-induced liver injury.

Discussion

ICI combination immunotherapy has improved efficacy compared to alternative systemic treatments for a number of cancer indications, yet its use is limited by increased toxicity. The frequency and severity of immune-related hepatitis was exac- erbated when IDO1, PD1, and or CTLA4 checkpoint inhibitors were co-administered to patients.5,10 The mechanism of ICI- induced hepatitis and the reasons for enhanced hepatitis with ICI combinations are largely unknown. Recently, we reported on a mouse model of ICI-induced hepatitis in which the combina- tory blockade of PD1, CTLA4, and IDO1 enhanced hepatitis.18 In the current study, we further investigated the immune mecha- nisms of ICI-induced hepatitis in this mouse model.
In this mouse model of ICI-induced hepatitis, the number of MDMs increased significantly in the livers of ICI combination-treated mice. Our data indicate the existence of close interactions between monocyte subtypes and T cells (Fig. 5D). Increased expression of IFNc response pathway genes in MDMs suggest a classical activation phenotype and pro-inflammatory response to IFNc produced by T cells (Fig. 4C). Batf2 was upregulated in the MDM cluster by ICI com- bination and is a master transcriptional regulator of pro- inflammatory macrophages involved in gene regulation of IFNc-activated classical macrophages.35 These MDMs produce the chemokines CXCL9, CCL5, and CXCL16 to recruit T cells and produce IL-18 to enhance Th1 responses and IFNc pro- duction, thus serving as a positive T cell–MDM feedback loop. In this study, we showed similar but distinct characteristics between ICI- and 4-1BB agonist-induced hepatitis. Both ICIs and the 4-1BB agonist induced increased hepatic infiltration of T cells and increased the activation and proliferation phenotypes. The 4-1BB agonist exhibited a greater effect (Fig. 6C), apart from in CD8+ T effector cells for which an increase was only found in the ICI combination-treated mouse livers (Fig. 6D). Both 4-1BB agonist-36 and ICI- induced hepatitis are dependent on CD8+ T cell activity since depletion or deficiency of CD8+ T cells eliminated liver injury.
Analogous to ICIs that remove the brakes on tumor-targeting effector T cells, 4-1BB co-stimulation may result in pressing of the accelerator or increasing immune effector activities. A clinical study of urelumab (a 4-1BB agonist) was terminated due to the incidence of fatal hepatitis.37 Liver toxicity was also observed in mice using anti-mouse 4-1BB agonistic antibodies.21 Interest- ingly, unlike ICIs that cause irAEs that are autoimmune-like diseases, 4-1BB agonism can ameliorate autoimmune dis- eases.38 This suggests that the immune mechanisms of 4-1BB agonist- and ICI-induced hepatitis may not be the same. Morphologic association between infiltrating mononuclear cells and necrotic hepatocytes was much less consistent with 4-1BB in comparison to the ICI-associated lesions. Additionally, foci of inflammation were generally much less discrete with 4-1BB and demonstrated great variability in size and shape even within a given individual sample. Bartkowiak et al. showed that activation of 4-1BB on liver myeloid cells is essential to initiate hepatitis and activated myeloid cells produce IL-27 which engages CD8+ T cells to cause liver injury.34
In the current study we demonstrated that treatment with an ICI combination or a 4-1BB agonist increased infiltration of pri- marily T cells into the liver. Trafficking of T cells to tissues re- quires the expression of chemokine receptors on T cells and the expression of the corresponding chemokines in tissues. The expression of multiple chemokine and chemokine receptors was increased in the treated mouse liver tissues or infiltrates, espe- cially the CXCR3, CCR2, and CCR5 chemokine axis molecules (Fig. 2C, 4E and 7D). This is consistent with findings in the mouse model of 4-1BB-induced hepatitis.34 Interestingly, CXCR3-CXCL9/ CXCL10 and CCR5/CCL5 chemokine axes have also been shown to be required for T cell trafficking to tumors.39
In patients treated with ipilimumab and nivolumab, case re- ports displayed more diverse, yet similar, liver pathological changes and immune cell features to our mouse model. Focal necrosis throughout the liver parenchyma with CD8+ T cell-rich mixed mononuclear infiltration was commonly observed in pa- tients treated with ICIs.40,41 These human cases characterized in the literature are likely representative of the more severe cases of immune-mediated hepatotoxicity associated with ICI treat- ment, and therefore it is not surprising that the histopathologic findings reported from those lesions are most similar morpho- logically to the most severe cases we identified in our mouse model. This includes the coalescence of smaller inflammatory and necrotic foci into larger regions of necrosis, at which time structural impairment leads to further non-specific histopatho- logic features of injury. Granulomatous foci have been described in ICI patient livers which represent mixtures of lymphocytes and histiocytes.42 In our mouse model, less severe lesions were morphologically characterized by an admixture of monocytes within the annular arrangements of lymphocytes surrounding small foci of necrosis, as well as frequent foamy macrophage aggregates further promoting the alternative morphologic description of microgranuloma, similar to what has been described in human liver lesions. The fibrinous and endothelial changes reported in the human liver were minimally present in mice, likely due to species differences in sensitivity, the acute nature of liver injuries in the mouse model, and corticosteroid therapy in humans.
This mouse model can be very useful for studying ICI hepatitis albeit with some limitations. It represents a less severe disease phenotype and is a more acute disease model. Although the key lesion phenotypes are generally similar to human liver lesions, there are some subtle differences as described above. Nonethe- less, this mouse model is of value for investigating immune mechanisms, pathogenesis, and treatment, as well as for evalu- ating potential liver toxicities associated with new ICIs or combinations.
Taken together, we developed a mouse model of ICI-induced hepatitis with in-depth molecular characterization of the hepatic immune microenvironment. We showed that ICI-induced hepa- titis is mediated by CD8+ T cells through close interactions with monocyte-derived myeloid cells.

References

[1] Prendergast GC, Malachowski WP, DuHadaway JB, Muller AJ. Discovery of IDO1 Inhibitors: from bench to bedside. Canc Res 2017;77:6795–6811.
[2] Ribas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade. Science 2018;359:1350–1355.
[3] Chen DS, Mellman I. Elements of cancer immunity and the cancer- immune set point. Nature 2017;541:321–330.
[4] Postow MA, Sidlow R, Hellmann MD. Immune-related adverse events associated with immune checkpoint blockade. N Engl J Med 2018;378:158–168.
[5] Larkin J, Chiarion-Sileni V, Gonzalez R, Grob JJ, Cowey CL, Lao CD, et al. Combined nivolumab and ipilimumab or monotherapy in untreated melanoma. N Engl J Med 2015;373:23–34.
[6] Michot JM, Bigenwald C, Champiat S, Collins M, Carbonnel F, Postel- Vinay S, et al. Immune-related adverse events with immune checkpoint blockade: a comprehensive review. Eur J Canc 2016;54:139–148.
[7] Weber JS, Kahler KC, Hauschild A. Management of immune-related adverse events and kinetics of response with ipilimumab. J Clin Oncol 2012;30:2691–2697.
[8] Robert C, Ribas A, Wolchok JD, Hodi FS, Hamid O, Kefford R, et al. Anti- programmed-death-receptor-1 treatment with pembrolizumab in ipilimumab-refractory advanced melanoma: a randomised dose- comparison cohort of a phase 1 trial. Lancet 2014;384:1109–1117.
[9] Postow MA, Chesney J, Pavlick AC, Robert C, Grossmann K, McDermott D, et al. Nivolumab and ipilimumab versus ipilimumab in untreated mela- noma. N Engl J Med 2015;372:2006–2017.
[10] Gibney GT, Hamid O, Gangadhar TC, Lutzky J, Olszanski AJ, Gajewski T, et al. Preliminary results from a phase 1/2 study of INCB024360 combined with ipilimumab (ipi) in patients (pts) with melanoma. J Clin Oncol 2014;32. 3010-3010.
[11] Makarova-Rusher OV, Medina-Echeverz J, Duffy AG, Greten TF. The yin and yang of evasion and immune activation in HCC. J Hepatol 2015;62:1420–1429.
[12] Boussiotis VA. Molecular and biochemical aspects of the PD-1 checkpoint pathway. N Engl J Med 2016;375:1767–1778.
[13] Ito H, Hoshi M, Ohtaki H, Taguchi A, Ando K, Ishikawa T, et al. Ability of IDO to attenuate liver injury in alpha-galactosylceramide-induced hepa- titis model. J Immunol 2010;185:4554–4560.
[14] Sumpter TL, Dangi A, Matta BM, Huang C, Stolz DB, Vodovotz Y, et al. Hepatic stellate cells undermine the allostimulatory function of liver myeloid dendritic cells via STAT3-dependent induction of IDO. J Immunol 2012;189:3848–3858.
[15] Nagano J, Shimizu M, Hara T, Shirakami Y, Kochi T, Nakamura N, et al. Effects of indoleamine 2,3-dioxygenase deficiency on high-fat diet- induced hepatic inflammation. PloS One 2013;8:e73404.
[16] Ji C, Roy MD, Golas J, Vitsky A, Ram S, Kumpf SW, et al. Myocarditis in cynomolgus monkeys following treatment with immune checkpoint in- hibitors. Clin Canc Res 2019;25:4735–4748.
[17] Selby MJ, Engelhardt JJ, Johnston RJ, Lu LS, Han M, Thudium K, et al. Preclinical development of ipilimumab and nivolumab combination immunotherapy: mouse tumor models, in vitro functional studies, and cynomolgus macaque toxicology. PloS One 2016;11:e0161779.
[18] Affolter T, Llewellyn HP, Bartlett DW, Zong Q, Xia S, Torti V, et al. Inhibition of immune checkpoints PD-1, CTLA-4, and IDO1 coordinately induces immune-mediated liver injury in mice. PloS One 2019;14:e0217276.
[19] Keir ME, Freeman GJ, Sharpe AH. PD-1 regulates self-reactive CD8+ T cell responses to antigen in lymph nodes and tissues. J Immunol 2007;179:5064–5070.
[20] Charlton JJ, Chatzidakis I, Tsoukatou D, Boumpas DT, Garinis GA, Mamalaki C. Programmed death-1 shapes memory phenotype CD8 T cell subsets in a cell-intrinsic manner. J Immunol 2013;190:6104–6114.
[21] Chen S, Lee LF, Fisher TS, Jessen B, Elliott M, Evering W, et al. Combination of 4-1BB agonist and PD-1 antagonist promotes antitumor effector/ memory CD8 T cells in a poorly immunogenic tumor model. Canc Immunol Res 2015;3:149–160.
[22] Church RJ, Kullak-Ublick GA, Aubrecht J, Bonkovsky HL, Chalasani N, Fontana RJ, et al. Candidate biomarkers for the diagnosis and prognosis of drug-induced liver injury: an international collaborative effort. Hepatol- ogy 2019;69:760–773.
[23] Thulin P, Hornby RJ, Auli M, Nordahl G, Antoine DJ, Starkey Lewis P, et al. A longitudinal assessment of miR-122 and GLDH as biomarkers of drug- induced liver injury in the rat. Biomarkers 2017;22:461–469.
[24] Liu X, Shin N, Koblish HK, Yang G, Wang Q, Wang K, et al. Selective in- hibition of IDO1 effectively regulates mediators of antitumor immunity. Blood 2010;115:3520–3530.
[25] de la Rosa G, Yang D, Tewary P, Varadhachary A, Oppenheim JJ. Lactoferrin acts as an alarmin to promote the recruitment and activation of APCs and antigen-specific immune responses. J Immunol 2008;180:6868–6876.
[26] Tolar P, Wack A. Monocytes work harder under pressure. Nat Immunol 2019;20:1422–1424.
[27] Trapnell C, Cacchiarelli D, Grimsby J, Pokharel P, Li S, Morse M, et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 2014;32:381–386.
[28] Ayers M, Lunceford J, Nebozhyn M, Murphy E, Loboda A, Kaufman DR, et al. IFN-gamma-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest 2017;127:2930–2940.
[29] Ribas A, Robert C, Hodi FS, Wolchok JD, Joshua AM, Hwu W-J, et al. As- sociation of response to programmed death receptor 1 (PD-1) blockade with pembrolizumab (MK-3475) with an interferon-inflammatory im- mune gene signature. J Clin Oncol 2015;33. 3001-3001.
[30] Aras S, Zaidi MR. TAMeless traitors: macrophages in cancer progression and metastasis. Br J Canc 2017;117:1583–1591.
[31] Timperi E, Focaccetti C, Gallerano D, Panetta M, Spada S, Gallo E, et al. IL- 18 receptor marks functional CD8(+) T cells in non-small cell lung cancer. Oncoimmunology 2017;6:e1328337.
[32] Knudson KM, Hodge JW, Schlom J, Gameiro SR. Rationale for IL-15 superagonists in cancer immunotherapy. Expert Opin Biol Ther 2020;20:705–709.
[34] Bartkowiak T, Jaiswal AR, Ager CR, Chin R, Chen CH, Budhani P, et al. Activation of 4-1BB on liver myeloid cells triggers hepatitis via an inter- leukin-27-dependent pathway. Clin Canc Res 2018;24:1138–1151.
[35] Roy S, Guler R, Parihar SP, Schmeier S, Kaczkowski B, Nishimura H, et al. Batf2/Irf1 induces inflammatory responses in classically activated macrophages, lipopolysaccharides, and mycobacterial infection. J Immunol 2015;194:6035–6044.
[36] Niu L, Strahotin S, Hewes B, Zhang B, Zhang Y, Archer D, et al. Cytokine- mediated disruption of lymphocyte trafficking, hemopoiesis, and induc- tion of lymphopenia, anemia, and thrombocytopenia in anti-CD137- treated mice. J Immunol 2007;178:4194–4213.
[37] Segal NH, Logan TF, Hodi FS, McDermott D, Melero I, Hamid O, et al. Re- sults from an integrated safety analysis of urelumab, an agonist anti- CD137 monoclonal antibody. Clin Canc Res 2017;23:1929–1936.
[38] Lee J, Lee EN, Kim EY, Park HJ, Chang CY, Jung DY, et al. Administration of agonistic anti-4-1BB monoclonal antibody leads to the amelioration of inflammatory bowel disease. Immunol Lett 2005;101:210–216.
[39] Idorn M, Thor Straten P. Chemokine receptors and exercise to tackle the inadequacy of T cell homing to the tumor site. Cells 2018;7.
[40] Kleiner DE, Berman D. Pathologic changes in ipilimumab-related hepatitis in patients with metastatic melanoma. Dig Dis Sci 2012;57:2233–2240.
[41] Zen Y, Yeh MM. Hepatotoxicity of immune checkpoint inhibitors: a his- tology study of seven cases in comparison with autoimmune hepatitis and idiosyncratic drug-induced liver injury. Mod Pathol 2018.
[42] De Martin E, Michot JM, Papouin B, Champiat S, Mateus C, Lambotte O, et al. Characterization of liver injury induced by cancer immunotherapy using immune checkpoint inhibitors. J Hepatol 2018;68:1181–1190.