Article Text

Original research
Therapeutic inhibition of monocyte recruitment prevents checkpoint inhibitor-induced hepatitis
  1. Cathrin L C Gudd1,
  2. Eoin Mitchell1,
  3. Stephen R Atkinson1,
  4. Marie-Anne Mawhin2,
  5. Samra Turajlic3,4,5,
  6. James Larkin4,5,
  7. Mark R Thursz1,
  8. Robert D Goldin1,
  9. Nick Powell1,
  10. Charalambos G Antoniades1,
  11. Kevin J Woollard2,
  12. Lucia A Possamai1 and
  13. Evangelos Triantafyllou1
  1. 1Department of Metabolism, Digestion and Reproduction, Imperial College London, London, UK
  2. 2Centre for Inflammatory Disease, Imperial College London, London, UK
  3. 3Cancer Dynamics Laboratory, The Francis Crick Institute, London, UK
  4. 4Renal and Skin Units, The Royal Marsden NHS Foundation Trust, London, UK
  5. 5Melanoma and Kidney Cancer Team, The Institute of Cancer Research, London, UK
  1. Correspondence to Dr Cathrin L C Gudd; c.gudd{at}
  • LAP and ET are joint senior authors.


Background Checkpoint inhibitor-induced hepatitis (CPI-hepatitis) is an emerging problem with the widening use of CPIs in cancer immunotherapy. Here, we developed a mouse model to characterize the mechanism of CPI-hepatitis and to therapeutically target key pathways driving this pathology.

Methods C57BL/6 wild-type (WT) mice were dosed with toll-like receptor (TLR)9 agonist (TLR9-L) for hepatic priming combined with anti-cytotoxic T lymphocyte antigen-4 (CTLA-4) plus anti-programmed cell death 1 (PD-1) (“CPI”) or phosphate buffered saline (PBS) control for up to 7 days. Flow cytometry, histology/immunofluorescence and messenger RNA sequencing were used to characterize liver myeloid/lymphoid subsets and inflammation. Hepatocyte damage was assessed by plasma alanine transaminase (ALT) and cytokeratin-18 (CK-18) measurements. In vivo investigations of CPI-hepatitis were carried out in Rag2−/− and Ccr2rfp/rfp transgenic mice, as well as following anti-CD4, anti-CD8 or cenicriviroc (CVC; CCR2/CCR5 antagonist) treatment.

Results Co-administration of combination CPIs with TLR9-L induced liver pathology closely resembling human disease, with increased infiltration and clustering of granzyme B+perforin+CD8+ T cells and CCR2+ monocytes, 7 days post treatment. This was accompanied by apoptotic hepatocytes surrounding these clusters and elevated ALT and CK-18 plasma levels. Liver RNA sequencing identified key signaling pathways (JAK-STAT, NF-ΚB) and cytokine/chemokine networks (Ifnγ, Cxcl9, Ccl2/Ccr2) as drivers of CPI-hepatitis. Using this model, we show that CD8+ T cells mediate hepatocyte damage in experimental CPI-hepatitis. However, their liver recruitment, clustering, and cytotoxic activity is dependent on the presence of CCR2+ monocytes. The absence of hepatic monocyte recruitment in Ccr2rfp/rfp mice and CCR2 inhibition by CVC treatment in WT mice was able to prevent the development and reverse established experimental CPI-hepatitis.

Conclusion This newly established mouse model provides a platform for in vivo mechanistic studies of CPI-hepatitis. Using this model, we demonstrate the central role of liver infiltrating CCR2+ monocyte interaction with tissue-destructive CD8+ T cells in the pathogenesis of CPI-hepatitis and highlight CCR2 inhibition as a novel therapeutic target.

  • Immune Checkpoint Inhibitors
  • CD8-Positive T-Lymphocytes
  • Immunotherapy
  • Inflammation

Data availability statement

Data are available upon reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Activation of CD8+ T cells and monocytes has been linked to the pathogenesis of checkpoint inhibitor (CPI)-hepatitis, however, their therapeutic targeting potential has not been examined.


  • This study uses a newly established mouse model which mimics human liver pathology and shows that CPI-hepatitis is driven by tissue-destructive CD8+ T cells. Furthermore, it mechanistically demonstrates that intrahepatic CCR2+ monocytes are required for CD8+ T-cell recruitment and activation. Therapeutic inhibition of CCR2-mediated monocyte liver recruitment using cenicriviroc prevents disease development.


  • This study highlights the CCR2 signaling pathways as a rational target for future investigation to treat liver inflammation in the context of CPI therapy.


Checkpoint inhibitors (CPIs) are a class of cancer immunotherapy with proven efficacy in a range of malignancies.1 2 These drugs are monoclonal antibodies targeting immune checkpoint molecules such as cytotoxic T lymphocyte antigen-4 (CTLA-4), programmed cell death-1 (PD-1) and its ligand programmed cell death-ligand-1 (PD-L1).1 Under normal physiological conditions, immune checkpoints are major regulators of immune homeostasis and tolerance, preventing the generation of autoreactivity and immune-mediated tissue damage.2 Blockade of these immune checkpoints has been shown to enhance cellular immunity and stimulate antitumor responses which result in improvements in long-term survival of patients with a range of cancer types.3 4

As checkpoint molecules are involved in pivotal regulatory pathways, CPI therapy is frequently complicated by immune-related adverse events (irAEs).5 6 These side effects may manifest in several different tissues.5 6 CPI-induced hepatitis (CPI-hepatitis) is one of the most common organ toxicities affecting up to 32% of patients on dual-agent CPI therapy.3 7 8 CPI-hepatitis is associated with increased serum alanine transaminase levels (ALT) levels and characterized histologically by lobular liver inflammation, with CD8+ T cell and macrophage inflammatory clusters associated with hepatocyte injury.9 The severity of the disease can range from incidental, mild liver function test abnormalities to fulminant hepatic failure and death.10–12 Current clinical care relies on cessation of CPI therapy and treatment with broadly immunosuppressive treatments, particularly corticosteroids, which have significant off-target side effects.13 A pressing clinical need therefore exists, to better understand what triggers liver inflammation in a proportion of CPI-treated patients and the immunopathology of this condition to develop targeted treatments.

The liver is constantly exposed to microbial and non-microbial products, including toll-like receptor (TLR) ligands (TLR-L) from the portal bloodstream. Thus, robust regulatory mechanisms are required to promote liver tolerance and avoid excessive inflammatory responses in homeostasis.14 Immune checkpoint pathways play an essential role in mediating this tolerance.14–17

In the absence of immune checkpoint signaling, either through blockade (eg, anti-CTLA-4, anti-PD-1) or genetic deletion (eg, PD-1−/−), mice do not spontaneously develop hepatic inflammation.18–20 However, studies using mouse models of drug-induced liver injury (DILI) report the induction of more severe and persistent toxicity when hepatic tolerance is broken by interruption of CPI signaling.18 19 21 For example, Metushi et al reported hepatocyte injury and enhanced effector T-cell function in PD-1−/− mice treated with anti-CTLA-4 and amodiaquine, a known direct liver toxin.18 In addition, Affolter et al and Llewelyn et al have shown that administration of anti-CTLA-4 and an indoleamine 2, 3-dioxygenase 1 (IDO1) inhibitor in PD-1−/− mice induces enhanced liver toxicity, reproducing aspects of human CPI-hepatitis.19 22 Evidence from patients treated with sequential anticancer agents has suggested that drugs, which are minimally hepatotoxic when administered alone, can trigger significant liver injury in patients with prior CPI exposure.23 24 This suggests that the inhibition of tolerance pathways by checkpoint blockade promotes exaggerated immune-mediated liver damage in response to relatively minor direct hepatotoxic triggers.

Here, we demonstrate that TLR9 agonist (TLR9-L) challenge of anti-CTLA-4/anti-PD-1-treated mice reproduces key characteristic features of human CPI-hepatitis, including an inflammatory infiltrate of CCR2+ monocytes and cytotoxic CD8+ T cells associated with surrounding hepatocyte apoptosis. RNA sequencing further confirmed the highly inflammatory environment and highlighted a complex intrahepatic immune crosstalk. Using Ccr2rfp/rfp transgenic mice in this model, we show that CCR2+ liver-infiltrating monocytes are necessary for the recruitment and activation of tissue-destructive CD8+ T cells. Furthermore, we demonstrate that therapeutic inhibition of monocyte liver recruitment using cenicriviroc (CVC; CCR2/5 antagonist), prevented the development and reversed established experimental CPI-hepatitis.



All experimental protocols were approved by Imperial College London in accordance with UK Home Office regulations (PPL: P8999BD42). C57BL/6 wild-type (WT) male mice (8–12 weeks old) were purchased from Charles River Laboratories. Rag2−/− (B6(Cg)-Rag2tm1.1Cgn/J) and Ccr2rfp/rfp (B6.129(Cg)-Ccr2tm2.1Ifc/J) mice were kindly donated by Professor Marina Botto and Dr Kevin Woollard, respectively (Imperial College London). All mice were housed and bred under the same specific pathogen-free conditions at the Imperial College London animal facility.

Animal treatments

Anti-PD-1 (clone RMP1-14), anti-CTLA-4 (clone 9H10), anti-CD4 (clone GK1.5) and anti-CD8 (clone YTS169.4) monoclonal antibodies (mAbs) were purchased from Bio X Cell, diluted in sterile phosphate-buffered saline (PBS) and injected intraperitoneally (i.p.) at a concentration of 200 µg/mouse.

Single anti-PD-1 and anti-CTLA-4 or combination anti-PD-1/CTLA-4 (“CPI”) or sterile PBS as control (“PBS”) were given every 2 days starting on day 0 (D0). On D1, 20 µg/mouse of TLR9-L (CpG oligodeoxynuleotide 1668: 5-S-TCCATGACGTTC CTGATGCT-3) (TIB Molbiol, Germany) was administered i.p. to prime hepatic inflammation. Mice were sacrificed, and blood and liver tissue were collected on D1, D4, D7, D10 and D14 as described in online supplemental methods.

Supplemental material

Anti-CD4 or anti-CD8 mAbs were given on D-2 and D-1 before and on D3 after the first CPI injection. 100 mg/kg/day of CVC mesylate (MedchemExpress, USA) or vehicle control was administered in drinking water1 4 days before the first CPI injection (“prophylactic”) or2 4 days following the first CPI injection (“therapeutic”) and for the whole duration of the time course.25 Mice were sacrificed on D7 or D10 and blood and liver tissue were harvested.

Plasma screening

Liver function tests

Liver function was assessed by measuring ALT (service provided by MRC Harwell Institute, UK) and cytokeratin-18 (CK-18) levels in plasma. CK-18 was measured by ELISA (Abcam, UK). The optical density was measured at 450 nm using the Multiskan Go plate reader (Thermo Fisher Scientific, UK).

CXCL9 and CXCL10 levels

Levels of CXCL9 and CXCL10 were measured in plasma (diluted twofold) using the mouse cytokine release syndrome panel LEGENDplex system (BioLegend, UK), according to the manufacturer’s instructions. The acquisition was performed on the BD LSRFortessa.

Liver histology and immunofluorescence

Formalin-fixed paraffin-embedded (FFPE) liver sections stained with H&E were provided by the Research Histology Facility, Imperial College London. Hepatocyte apoptosis was confirmed in FFPE liver sections using TUNEL staining (Thermo Fisher Scientific, UK), according to manufacturer’s instructions and co-staining with albumin and CD8 (online supplemental table 1). Immunofluorescent staining of optimal cutting temperature compound (OCT)-fixed liver cryosections was performed to assess the expression of CD8, F4/80, CD11b, CCR2, granzyme B (GZMB) using fluorochrome-labeled mAbs listed in online supplemental table 1. Slides were mounted with fluoroshield with DAPI (Sigma-Aldrich, USA). Images were captured using the Leica DM4 B microscope and the LAS X software (Leica Camera AG, Germany).

Quantitative reverse transcription PCR

Total messenger RNA (mRNA) was extracted from snap-frozen liver tissue using the RNeasy Plus Mini Kit (Qiagen, Germany) and reversed transcribed to complementary DNA using SuperScript IV reverse transcriptase with Random Hexamers (Invitrogen, USA), according to manufacturer’s instructions. The 2×SensiMix SYBR Lo-ROX kit (Bioline, UK) was used for the quantification of Cxcl9, Cxcl10 and Ccl2 mRNA levels (online supplemental table 3). Quantitative PCR was performed using the Applied Biosystems ViiA7-fast block instrument and analyzed with the ViiA7 software (Thermo Fisher Scientific, UK). Fold-change in gene expression was calculated as 2−ΔΔCt, as compared with D1 control mice (n=3) of the respective strain.

RNA sequencing

RNA was extracted from snap frozen liver tissue of WT mice on D1, D4 and D7 (n=4/group) post administration of CPI or PBS, following the manufacturer’s instructions using the RNeasy Plus Mini Kit (Qiagen, Germany). RNA sequence data was aligned to GRCm39/Ensembl V.104 with transcript quantification by RSEM. Count data were quality controlled and analyzed in R V.4.0.4 using limma-voom, dorothea and gage. A detailed description of all related methods is provided in Online supplemental materials.

Flow cytometry of liver immune cells and absolute cell counts

Surface and intracellular staining of isolated hepatic mononuclear cells were carried out using fluorochrome-labeled mAbs listed in online supplemental table 4. The acquisition was performed on the BD LSRFortessa and data were acquired using the BD FACSDiva software (Becton Dickinson, UK). Data was analyzed with FlowJo V.10.9.0. Absolute cell counts were obtained using the 123count eBeads (Thermo Fisher Scientific, UK) by flow cytometry, according to the manufacturer’s instructions.

Statistical analysis

GraphPad Prism Software V.10.2.0 was used to test for statistical significance. Analyses were conducted to compare different groups at one given time point (D1, D4, D7, D10 or D14) or in comparison to D1 within each respective group. For non-parametric unpaired analysis, Mann-Whitney U tests were performed comparing two groups and Kruskal-Wallis tests were used for comparisons between three or more groups. For correlations, Spearman rank correlation coefficients were used. Data are expressed as mean±SEM and statistical significance is assumed for p values of <0.05.

Other details and additional experimental procedures are provided in the online supplemental methods.


CTLA-4/PD-1 blockade in combination with TLR9-L administration induces liver inflammation and hepatocyte damage

In line with previous reports,18–21 breaking immune tolerance by treating mice with a combination anti-CTLA-4/anti-PD-1 (CPI) was insufficient in inducing hepatitis in mice (online supplemental figure 1A). Here, we sought to examine whether administration of TLR9-L in WT mice, mimicking the microbial signal (unmethylated CpG rich DNA) from viral/bacterial infection or increased intestinal bacterial translocation into the portal bloodstream, in combination with CPI treatment would induce liver inflammation to model human disease (figure 1A).9

Supplemental material

Figure 1

Liver inflammation and hepatocyte damage following TLR9-L/CPI treatment of mice. (A) Set-up of experimental CPI-hepatitis time course, comparing CPI (blue) and PBS (gray)-treated mice across different time points. (B) Representative H&E stained liver sections of CPI or PBS-treated mice during the time course (n=8/group/time point). Magnification: 200×, 400×. (C) Measurement of liver damage indices (ALT, cytokeratin-18) in plasma. (D) Representative contour plots and absolute numbers of total liver CD8+ T cells. (E) Representative gating strategy and absolute numbers of total liver monocytes during the time course. (F) Representative pictures of liver cryosections stained for CD8 (white), CD11b (red) and DAPI (blue) on D7 of CPI and PBS-treated mice (n=8/group). Magnification: 400×. (G) Histograms and absolute number of CCR2+ monocytes and representative pictures of liver cryosections stained for CCR2 (red) and DAPI (blue) of CPI and PBS-treated mice on D7 (n=4/group). Magnification: 400×. (H) Fold change expression of total mRNA levels of Ccl2 in liver homogenates on D7, normalized to respective D1 controls (n=3). Each symbol represents an individual mouse. Results shown are representative of one to three independent experiments. *p<0.05, **p<0.01, ***p<0.001. ****p<0.0001. ALT, alanine transaminase levels; CPI, checkpoint inhibitor; D, day; FMO, fluorescence minus one; MHC, major histocompatibility complex; mRNA, messenger RNA; NK, natural killer; PBS, phosphate buffered saline; TLR-L, toll-like receptor 9 ligands.

Following a single dose of CPI or PBS and prior to TLR9-L administration (D1), mice showed normal liver histology (figure 1B). However, after TLR9-L dosing, CPI-treated mice showed panlobular or centrilobular granulomatous hepatitis on D4 (figure 1B), with even more pronounced inflammatory foci on D7, similar to human liver pathology.11 26 Notably, apoptotic and swollen hepatocytes were seen within hepatic immune clusters (figure 1B, online supplemental figure 1B).

To assess the severity of liver injury, ALT and CK-18 levels were measured in plasma. Although liver histology was normal on D1 (figure 1B), CPI-treated mice showed increased ALT levels, compared with PBS-treated mice on D1 (figure 1C). ALT levels increased further on D4 in CPI-treated mice but remained stable in PBS-treated mice. In addition, CK-18 was elevated in CPI-treated mice on D7, compared with D7 PBS and D1 CPI (figure 1C). On D10, liver inflammation was milder and resolved by D14 despite continuous CPI treatment for the duration of the time course (online supplemental figure 2A–C). Of note, hepatic TLR4-L priming induced milder liver inflammation on D7 following CPI treatment compared with TLR9-L priming, but had no effect on ALT levels (online supplemental figure 1C, D).

Supplemental material

The treatment of mice with TLR9-L in combination with single CPIs, anti-CTLA-4 but particularly anti-PD-1, only induced very mild levels of liver inflammation compared with combination CPIs on D7 (online supplemental figure 3A−C). While no changes in CK-18 levels were measured between single CPIs and PBS, both single CPIs caused histological patterns also observed in humans. These included mild lobular and bile duct inflammation in one anti-PD-1 treated mouse, with more severe changes and a granuloma found in the anti-CTLA-4 group (online supplemental figure 3C).

Supplemental material

Liver infiltration of CD8+ T cells and CCR2+ monocytes in TLR9-L/CPI-treated mice

Studies on human CPI-hepatitis from our group and others report increased recruitment and aggregation of CD8+ T cells with CCR2+ monocytes in histological observations.9 11 26 These inflammatory structures are associated with surrounding hepatocyte death and are a distinct pathological feature of CPI-hepatitis.26 To investigate whether this is replicated in our model, we used flow cytometry and immunofluorescence microscopy to identify and quantify the liver myeloid and lymphoid populations.17

No differences in natural killer (NK) cell, NKT cell, B cell, and CD4+ T-cell populations were observed between CPI and PBS-treated mice (online supplemental figure 4A–C). However, liver CD8+ T cells were significantly increased in CPI-treated animals on D7, compared with PBS (figure 1D, online supplemental figure 2D). Similarly, Ly6Chigh inflammatory monocytes were expanded on D7 compared with control (figure 1E). In addition, D7 CPI-treated mice showed significantly increased numbers of resident Kupffer cells (KC) and monocyte-derived macrophages (MoMF), compared with D7 PBS, and neutrophils in D7 CPI-treated mice compared with D1 (online supplemental figure 5A, B).

Supplemental material

Supplemental material

To identify topographically where myeloid cells accumulate in liver tissue, cryosections were stained for F4/80 and CD11b. CD11b+ cells formed lobular inflammatory clusters on D7 in CPI-treated mice (online supplemental figure 5C, online supplemental figure 6A), whereas D7 PBS-treated mice showed evenly distributed CD11b+ cells. In contrast, F4/80+ cells were distributed evenly across lobules and portal tracts in D7 CPI-treated mice with typical KC morphology (online supplemental figure 5C). Furthermore, mirroring the human liver histology, CD11b/CD8 and F4/80-CD8 double staining revealed that clusters of CD11b+ also contained CD8+ T cells (figure 1F, online supplemental figure 6A,B), while F4/80 cells were distributed across hepatic lobules and not localized to the same areas (online supplemental figure 5D). These data suggest that recruited CD11b+ monocytes, rather than F4/80+ KC, contribute to the formation of inflammatory foci in CPI-hepatitis.

Supplemental material

In line with these findings and similar to human CPI-hepatitis, increased numbers of monocytes expressing the recruitment chemokine receptor CCR2, were measured on D7 following CPI treatment and CCR2 was expressed within inflammatory clusters (figure 1G, online supplemental figure 2D, online supplemental figure 6C). Notably, monocytes are the main expressors of CCR2 within the liver out of other myeloid cells (online supplemental figure 5E). This liver homing phenotype of monocytes was reflected in significantly increased total liver mRNA Ccl2 levels (ligand for CCR2) on D7 in CPI-treated mice compared with D7 PBS (figure 1H). Intracellular cytokine staining revealed that KCs were a potential source of CCL2 (online supplemental figure 5F).

Liver transcriptome demonstrates exaggerated and perpetuated inflammatory responses in TLR9-L/CPI-treated mice

Our previous human transcriptional analysis of circulating monocytes and CD8+ T cells revealed the activation of transcription factors such as JUN, GATA3, NFκB and JAK signaling, and inflammation-associated markers (eg, TNFα and IL-6).9 To gain a deeper insight into the inflammatory environment of the liver throughout the CPI-hepatitis time course, bulk RNA sequencing was performed. Principal component analysis (PCA) and clustered heatmap of differentially expressed (DE) genes across all samples demonstrated similar gene expressions on D1 in PBS and CPI-treated mice and on D4 between both groups (figure 2A,B). D7 CPI-treated mice demonstrated persistent alterations in their transcriptional profile with augmentation of expression of inflammation-associated genes. In contrast, PBS-treated mice demonstrated only a transient shift in gene expression profile on D4, which resolved by D7. The D7 PBS samples showed clustering with D1 PBS both in PCA space and in DE genes (figure 2A,B). In addition, DoRothEA predicted transcription factor (TF) activation analysis showed a modest activation of various inflammation-associated TFs on D4 in both CPI and PBS-treated mice, whereas the activation of such TFs (eg, Jun, Stat1, Nfkb1, Irf9 and Gata3) was further enhanced in D7 CPI-treated mice (figure 2C), replicating aspects of the reported human transcriptional profile.9 In contrast, TF expression patterns had returned to near baseline in D7 PBS-treated mice (figure 2C).

Figure 2

Liver transcriptome demonstrates perpetuated inflammatory responses with combined TLR9-L/CPI treatment. RNA sequencing analysis of n=4 mice/group for D1, D4, D7. (A) A schematic summary of principal components analysis (PCA) with a transient shift induced by TLR9-L in the absence of CPI (dashed arrow) but sustained in its presence (solid arrow). (B) Heatmap of differentially expressed genes across all samples. (C) Activity of selected transcription factors inferred using dorothea and summarized by time point and treatment group. (D) Volcano plot of differential expression analysis between mice after 7 days of treatment with, or without, CPI. (E) Significantly dysregulated KEGG pathways tested using Generally Applicable Gene set Enrichment analysis for D7 CPI compared with D7 PBS as baseline. CPI, checkpoint inhibitor; D, day; DE, differentially expressed; KEGG, Kyoto Encyclopedia of Genes and Genomes; PBS, phosphate buffered saline; TLR-L, toll-like receptor ligand.

These data suggested that the observed changes on D4 are largely TLR9-L treatment effects while D7 changes are due to CPI treatment; therefore, our subsequent analyses focused on the comparisons between D7 TLR9-L/CPI and D7 TLR9-L/PBS mice. RNA sequencing confirmed flow cytometry protein expression analyses (Ly6c2, Ccr2, Cd8a and Gzmb) and increased mRNA levels of Ccl2 and further showed differential expression of T-cell recruiting chemokines (Cxcl9/10) and markers of inflammation (eg, Ifnγ, Tnfα) (figure 2D and online supplemental materials/excel spreadsheet). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of this D7 comparison showed the downregulation of pathways involved in metabolism, particularly associated with anti-inflammatory conditions (eg, tryptophan metabolism, oxidative phosphorylation) and the upregulation of pathways demonstrating a highly activated, pro-inflammatory profile (figure 2E). These included pathways important for cell migration (chemokine signaling, leukocyte endothelial migration), cell interactions (eg, cytokine–cytokine receptor, T-cell receptor signaling, MAPK signaling, antigen presentation), inflammation (eg, NFκB signaling, JAK-STAT signaling) and cytotoxicity (figure 2E).

Supplemental material

Increased cytotoxicity of CD8+ T cells is associated with hepatocyte damage

Based on reports of enhanced cytotoxicity particularly within inflammatory foci in human CPI-hepatitis9 and our transcriptional data showing upregulation of genes associated with cytotoxicity, we next assessed the cytotoxic activity of CD8+ T cells. Functional analysis showed an expansion of CD8+ T cells producing GZMB, perforin and interferon (IFN)-γ on D7 in CPI-treated mice (figure 3A). This increase in cytotoxic activity correlated positively with disease severity (CK-18/GZMB: r=0.65, p=0.008; CK-18/perforin: r=0.57, p=0.02) (figure 3B). Moreover, tissue staining revealed that CD8+ T cells produce GZMB within the center of these immune clusters (figure 3C).

Figure 3

Enhanced cytotoxic CD8+ T-cell activity in TLR9-L/CPI-treated mice. (A) Representative histograms and frequencies of liver GZMB+, perforin+ and IFN-γ+CD8+ T cells in CPI and PBS-treated mice on D1 compared with D7. (B) Spearman correlations of absolute numbers of perforin+ (purple) and GZMB+ (blue) CD8+ T cells with CK-18 levels in D7 CPI-treated mice. (C) Representative pictures of liver cryosections stained for CD8 (white), GZMB (red) and DAPI (blue) on D7 of CPI and PBS-treated mice (n=4/group). Magnification: 400×. Each symbol represents an individual mouse. Results shown are representative of one to three independent experiments. *p<0.05, **p<0.01, ***p<0.001. ****p<0.0001. CK-18, cytokeratin-18; CPI, checkpoint inhibitor; D, day; FMO, fluorescence minus one; GZMB, granzyme B; IFN, interferon; PBS, phosphate buffered saline.

CD8+ T cells are required for the development of CPI-hepatitis

To identify whether liver injury following TLR9-L/CPI treatment is in fact driven by the lymphoid compartment, we performed the CPI-hepatitis time course using Rag2–/– mice (which lack mature lymphocytes) (figure 4A). While monocyte numbers were similarly increased among groups, liver injury index CK-18 plasma levels were significantly reduced in D7 CPI Rag2–/– mice compared with D7 CPI WT mice (figure 4B) while Rag2–/– mice also showed normal liver histology (figure 4C).

Figure 4

CD8+ T cells mediate hepatocyte damage following TLR9-L/CPI treatment. Comparison of experimental CPI-hepatitis time course in WT, Rag2–/– and CD4 or CD8 depleted mice. (A) Representative contour plots pre-gated on CD3+ confirming the absence of CD8+ and CD4+ T cells in Rag2−/− mice. (B) Absolute numbers of total liver monocytes (top) and measurement of CK-18 plasma levels (bottom) in WT versus Rag2−/− mice on D1 compared with D7. (C) Representative H&E stained liver sections of D7 WT versus Rag2−/− mice (n=4/group). Magnification: 100×. (D) Depletion schedule of CD4+/CD8+ T cells during experimental CPI-hepatitis in WT mice. Representative contour plots pre-gated on CD3+ confirming the absence of either CD4+ or CD8+ T cells in WT mice. (E) Absolute numbers of CD8+ T cells and monocytes and CK-18 plasma levels during the CD4/CD8 depletion time course. (F) Representative H&E stained liver sections of mice on D7 following CPI±anti-CD4/anti-CD8 treatment (n=4/group). Magnification: 100×. (G) Representative pictures of D7 liver sections stained left: for CD8 (white) and DAPI (blue) and right: CD11b (red) and DAPI (blue) of D7 CPI±anti-CD4/anti-CD8 treated mice (n=4/group). Magnification: 100×. (H) Frequencies of liver GZMB+, perforin+ and IFN-γ+CD8+ T cells in CPI and PBS-treated mice and D7 CPI+anti-CD4-treated mice. Each symbol represents an individual mouse. Results shown are representative of one to three independent experiments. *p<0.05, **p<0.01, ***p<0.001. ****p<0.0001. CK-18, cytokeratin-18; CPI, checkpoint inhibitor; D, day, GZMB, granzyme B; IFN, interferon; PBS, phosphate buffered saline; WT, wild type.

To further pinpoint the subset of T-cell mediating CPI-hepatitis, we carried out targeted CD4 or CD8 depletion in WT mice during the time course (figure 4D,E). In the absence of CD4+ T cells (D7 CPI+anti-CD4), absolute numbers of CD8+ T cells and monocytes were similar to D7 CPI WT mice (figure 4E). This was reflected by elevated CK-18 levels, histology showing hepatocellular injury with inflammatory foci of CD8+ T cells and CD11b+ monocytes (figure 4E–G) and enhanced cytotoxicity of CD8+ T cells similar to D7 CPI WT mice (figure 4H). In contrast, mice without CD8+ T cells (D7 CPI+anti-CD8) (figure 4D,E), despite similarly increased absolute numbers of liver infiltrating monocytes compared with D7 CPI WT mice, were protected from liver injury and showed normal liver histology and distribution of monocytes (figure 4E–G).

Monocytes mediate liver recruitment, clustering and cytotoxic activity of CD8+ T cells following TLR9-L/CPI treatment

Given our data demonstrating that cytotoxic CD8+ T cells drive liver injury in CPI-hepatitis, we next questioned what mediates their liver infiltration. Our transcriptomic data revealed an upregulation of the chemokine receptor Cxcr3 and its ligands (eg, Cxcl9, Cxcl10) important for T-cell recruitment (figure 2B,D). These findings were confirmed by flow cytometry showing an expansion of CXCR3+CD8+ T cells on D7 in CPI-treated WT mice (figure 5A) as well as significantly increased CXCL9/10 plasma levels and total liver mRNA expression (figure 5B). Cytokine staining identified monocytes as the substantial source of CXCL9 on D7 during experimental CPI-hepatitis (figure 5C). Notably, the number of liver infiltrating CCR2+ monocytes also correlated positively with the number of CD8+ T cells producing GZMB and perforin (GZMB: r=0.74, p=0.0015; perforin: r=0.79, p=0.0005) (figure 5D).

Figure 5

Monocytes mediate recruitment, clustering and activation of tissue destructive CD8+ T cells following TLR9-L/CPI treatment. (A) Representative histograms and frequency of liver CXCR3+CD8+ T cells in CPI and PBS-treated mice on D1 compared with D7. (B) Fold change of plasma CXCL9 and CXCL10 levels (left) and fold change expression of total mRNA levels of Cxcl9 and Cxcl10 in liver homogenates (right) in D7 CPI and PBS-treated mice compared with respective D1 controls (n=3–5/group). (C) Representative histograms and fold change of CXCL9+ liver myeloid cell frequencies on D7 CPI normalized to PBS (n=4/group). Frequency of CXCL9+ monocytes in CPI and PBS-treated mice on D1 compared with D7. (D) Spearman correlations of absolute numbers of perforin+ (purple) and GZMB+ (blue) CD8+ T cells with total number of CCR2+ monocytes in D7 CPI-treated mice. (E) Comparison of experimental CPI-hepatitis time course in WT and Ccr2rfp/rfp mice. Representative contour plots pre-gated on MHC-II cells showing few Ly6Chigh monocytes in Ccr2rfp/rfp mice, compared with WT mice. (F) Absolute numbers of total liver CD8+ T cells (left) and fold change expression of total mRNA levels of Cxcl9 and Cxcl10 in liver homogenates in D7 CPI and PBS-treated WT and Ccr2rfp/rfp mice compared with respective D1 controls (n=3) (right). (G) Data showing frequencies of GZMB+ and perforin+CD8+ T cells during the time course. (H) Representative H&E stained liver sections of D7 WT versus Ccr2rfp/rfp mice during the time course (n=4/group; magnification: 100×) and liver cryosections on D7 stained for CD8 (white), CD11b (red) and DAPI (blue) (n=4/group; magnification: 400×). (I) Measurement of cytokeratin-18 in plasma. Each symbol represents an individual mouse. Results shown are representative of one to three independent experiments. *p<0.05, **p<0.01, ***p<0.001. ****p<0.0001. CPI, checkpoint inhibitor; D, day; FMO, fluorescence minus one; GZMB, granzyme B; KC, Kupffer cells; MHC, major histocompatibility complex; MoMF, monocyte-derived macrophages; mRNA, messenger RNA; PBS, phosphate buffered saline; WT, wild type.

Since these data, together with observations of dense clusters of CD8+ T cells with monocytes, suggest that monocytes may play a role in CD8+ T-cell recruitment, we next explored the effects of CCR2+ monocyte deficiency using Ccr2rfp/rfp transgenic mice (figure 5E). In Ccr2rfp/rfp mice, the numbers of liver CD8+ T cells on D7 did not increase from D1 following TLR9-L/CPI treatment and were significantly reduced compared with D7 CPI-treated WT mice (figure 5F). This was accompanied by significantly lower levels of Cxcl9 and Cxcl10 mRNA (figure 5F).

The absence of liver infiltrating monocytes, further led to reduced frequencies of CD8+ T cells producing GZMB and perforin on D7, compared with CPI-treated WT mice (figure 5G). Moreover, these mice presented normal liver histology, with evenly distributed CD11b+ and CD8+ cells (figure 5H) and baseline CK-18 levels (figure 5I).

Therapeutic inhibition of hepatic monocyte recruitment ameliorates CPI-hepatitis

Based on our data indicating that CCR2+ monocytes are required for CD8+ T-cell liver recruitment, clustering and activation, we explored the therapeutic potential of monocyte inhibition in preventing CPI-hepatitis development using CVC (a small molecule inhibitor against CCR2/CCR5) in drinking water (figure 6A,B).25 Following prophylactic CVC treatment, significantly fewer CCR2+ monocytes were measured on D7 in CPI-treated mice, compared with those receiving vehicle control or normal drinking water (figure 6B). In line with our previous results, in the absence of CCR2+ monocytes, CD8+ T-cell numbers were significantly reduced on D7, compared with vehicle control or normal drinking water, and were not altered compared with D1 CPI mice (figure 6C). This was reflected in the reduced frequency of CXCR3+CD8+ T cells, as well as GZMB and perforin-producing CD8+ T cells in CVC-treated mice (figure 6D). The loss of CCR2+ monocytes and the associated reduction in CD8+ T-cell recruitment and cytotoxic activity further resulted in the prevention of the development of liver injury in CVC-treated mice (figure 6E).

Figure 6

Therapeutic inhibition of monocyte liver recruitment is protective of experimental CPI-hepatitis. (A) Set-up of experimental CPI-hepatitis time course and CVC/vehicle treatment schedule. (B) Representative contour plots pre-gated on MHCII cells showing few monocytes in CVC, compared with vehicle-treated mice and absolute numbers of total liver CCR2+ monocytes during the time course. Representative liver sections of D7 vehicle versus CVC-treated mice stained for CD11b (red) and DAPI (blue) (n=4/group; magnification: 100×). (C) Absolute numbers of total liver CD8+ T cells (left) and representative liver sections of D7 vehicle versus CVC-treated mice stained for CD8 (white) and DAPI (blue) (n=4/group; magnification: 100×). (D) Frequencies of total liver CXCR3+, granzyme B+ and perforin+CD8+ T cells during the time course. (E) Measurement of cytokeratin-18 in plasma (left) and representative H&E stained liver sections (n=4/group; magnification: 100×). Each symbol represents an individual mouse. Results shown are representative of one to three independent experiments. *p<0.05, **p<0.01, ***p<0.001. ****p<0.0001. CPI, checkpoint inhibitor; D, day; MHC, major histocompatibility complex; PBS, phosphate buffered saline; TLR9-L, toll-like receptor 9 ligand.

Finally, we assessed whether therapeutic CVC treatment could be used to reverse established CPI-hepatitis. For this, CVC was administered from D4 and tissues were harvested on D10 (online supplemental figure 7A); this allowed sufficient time for CVC to be therapeutically effective (online supplemental figure 7B,C). On D10, similarly to prophylactic administration of CVC, therapeutic administration reduced the absolute number of liver infiltrating CD8+ T cells and the frequency of GZMB and perforin-producing CD8+ T cells (online supplemental figure 7D,E). This was further associated with a reduction in CK-18 levels and normal liver histology (online supplemental figure 7F).

Supplemental material


In this work, we established a new mouse model of CPI-hepatitis. In line with previous studies, durable hepatitis failed to be induced in the context of CPI alone.18 19 Recently, Hutchinson et al, proposed an association of cytomegalovirus reactivation with the development or worsening of CPI-hepatitis, suggesting that the triggered T cell-mediated immunity in the context of CPI might drive liver inflammation directly or promote tissue-damaging bystander T-cell responses.27 In our study, administration of combination CPIs with a TLR9 stimulating agent replicating a viral/bacterial trigger induced a model closely resembling human CPI-hepatitis, thus providing a new platform for further in vivo mechanistic studies. Parallel to human liver pathology and circulating immune cell phenotype, mice in our model presented with mixed cytotoxic GZMB+perforin+CD8+ T cell and CCR2+ monocyte inflammatory foci with surrounding hepatocyte apoptosis and elevated plasma markers of liver injury.9 Liver RNA sequencing replicated many aspects of the human transcriptomic profile of circulating monocytes/CD8+ T cells and identified an exaggerated tissue-damaging inflammatory response characterized by increased expression of key pro-inflammatory mediators (NfKb, Stat1, Tnfα, Ifnγ) and chemokines (Cxcl9, Cxcl10 and Ccl2). Using this model, we demonstrate that cytotoxic CD8+ T cells are drivers of hepatocellular injury, while their recruitment, clustering and activation are mediated by interactions with liver CCR2+ monocytes. The absence of hepatic monocyte recruitment following genetic deletion of CCR2 or through its therapeutic inhibition using CVC prevented the development and reversed established experimental CPI-hepatitis.

While the treatment of mice with only CPIs led to an initial rise of ALT after 1 day, this was not sustained beyond day 1. This may be due to initial low levels of immune activation in the liver following the first CPI treatment. However, due to the highly tolerogenic environment and the high threshold of stimulation required to trigger a prolonged inflammatory response in the liver, the CPI-mediated response was not sufficient to induce hepatitis.14 The administration of a physiologically plausible priming stimulus in the context of checkpoint inhibition generated hepatitis with a similar histological characteristic of human disease. This differs from other published models of CPI-hepatitis using Pdcd1−/− mice, which are known to have increased CD8+ T-cell liver infiltration and monocyte activation during homeostasis17 19 22 and supports the theory that CPI-hepatitis arises when a subclinical “trigger” occurs in the context of checkpoint blockade. While the mild TLR9-L-induced inflammation resolves by day 7 in PBS-treated mice, loss of important liver tolerance mechanisms in the context of checkpoint inhibition allows the perpetuation of unchecked inflammatory responses, ultimately causing hepatocellular damage. This damage, however, was self-limiting and resolved after 14 days despite continuous CPI treatment without TLR9 restimulation. Although crucial regulatory receptors were blocked using CPIs, compensatory liver tolerance mechanisms including other immune checkpoint pathways or the release of anti-inflammatory cytokines may be activated in response to toxicity and continuous exposure to CPIs.28 Similarly, there is increasing evidence that human CPI-hepatitis is also self-limiting.29

Liver single-cell RNA sequencing data have demonstrated that TLR9 is predominantly expressed by monocytes and other antigen-presenting cells (dendritic cells, B cells) whereas TLR4, for example, is more widely expressed across all cells in the liver.30 Recent murine studies focusing on priming of hepatic immune responses using TLR9-L as an immunotherapeutic strategy to overcome the exhaustion of CD8+ T cells, showed that TLR9-L administration induced local activation and expansion of CD8+ T cells within cocoon-like structures of myeloid cells in the liver in the context of chronic viral liver infection.31 Moreover, Cebula et al showed that the systemic treatment of mice with TLR9-L is beneficial in therapeutic vaccination for immunization against hepatotropic viruses and the development of memory-recall responses in CD8+ T cells.32 TLR9-L administration was also shown to promote the control of tumor growth in murine hepatomas by inducing the proliferation of effector CD8+ T cells within immune clusters formed with monocytes in the liver.33 Similarly, in the context of TLR9 stimulation with CPI treatment, the interaction between CCR2+ monocytes and CD8+ T cells within inflammatory clusters may lead to the expansion and enhanced cytotoxicity of CD8+ T cells.

Using Rag2−/− mice, as well as selective depletion of CD4+ and CD8+ T-cell populations, we reveal that CD8+ T cells are the main drivers of this damage. This is further supported by recently published data by Llewellyn et al, demonstrating that selective CD8+ T-cell depletion abrogated hepatotoxicity in another model of CPI-hepatitis.22 Moreover, our analysis identified cytolytic mediators (eg, GZMB, perforin), IFN-γ and CXCL chemokines as key drivers of experimental CPI-hepatitis, providing additional support for a CD8-mediated mechanism of cytotoxicity directed towards hepatocytes. These pathways have been associated with a number of pathologies including other CPI irAEs.34–38 However, direct targeting of CD8+ T cells or the identified relevant mediators would significantly compromise the CPI response, as functional CD8+ tumor infiltrating lymphocytes are essential in mediating antitumor immunity.

Our data further suggests a complex chemokine/cytokine-mediated crosstalk leading to the recruitment and activation of the tissue-destructive CD8+ T cells by monocytes. The inhibition of monocyte recruitment to the liver may therefore present an alternative target upstream of crucial antitumor functions. Both previously reported human9 and our most recent mouse data highlighted an involvement of the CCR2/CCL2 axis in monocyte recruitment during the pathogenesis of CPI-hepatitis. Genetic deletion of CCR2 in our model prevented the CD8-mediated hepatocyte damage, implicating anti-CCR2/CCL2 as a potential therapeutic avenue. In fact, the role of monocytes in promoting CD8+ T cell-associated liver injury and the beneficial effects CCR2/CCL2 targeting has recently been described in a model of in CD8-dependent idiosyncratic DILI.18 In the study, Mak et al reported that anti-CCL2 treatment of mice reduced hepatic monocyte infiltration attenuated liver injury.18 39 Moreover, CCR2/CCL2 signaling has been implicated in various other liver conditions, such as acute liver injury, fibrosis/cirrhosis, and tumor progression in hepatocellular carcinoma (HCC). Experimental and early clinical investigations have demonstrated the safety of the CCR2/CCR5 small molecule inhibitor, CVC, in patients with established liver disease, both non-alcoholic steatohepatitis and alcohol-induced steatohepatitis.40–42 Additionally, CVC has also been trialed in acutely unwell patients with COVID-19 pneumonia, without safety concerns.43 As a result, targeting the CCR2/CCL2 axis may represent a therapeutic approach in CPI-hepatitis.

Moreover, in the tumor microenvironment, CCR2/CCL2 was shown to contribute to tumor growth and metastasis in a number of tumors. Tumors secret CCL2 to attract monocytes/macrophages, which in turn polarize to an “M2-like” phenotype and promote local immunosuppression and the development of metastasis.44 45 In preclinical models, CCR2/CCL2 inhibition, including CVC treatment, was shown to impede the growth and metastasis of HCC, colon carcinoma, prostate and breast cancer by regulating the recruitment and function of tumour-associated macrophages, ultimately leading to improved survival of mice.44 46–48 In addition, CCR2/CCL2 inhibition in combination with anti-PD-1 was shown to further enhance tumor responses over monotherapy across murine tumor models.49 Similarly, in patients with breast, cervix, prostate, HCC and colon cancer, elevated CCL2 levels in plasma were shown to be associated with poor prognosis and metastatic disease.44 46 48 50 51 Thus, CVC or other antagonists targeting CCR2, might not only have the potential to attenuate CPI-hepatitis but could also have beneficial effects on the antitumor immune response. Currently, ongoing phase 2 clinical trials investigate the beneficial effects of a CCR2/CCR5 inhibitor in combination with nivolumab in HCC and non-small cell lung cancer and locally advanced pancreatic cancer ( Identifier: NCT04123379, NCT03496662). Results from a completed clinical trial assessing the safety and tolerability of a CCR2 inhibitor in combination with FOLFIRINOX chemotherapy for the treatment of pancreatic cancer demonstrated a local tumor control in 97% of patients, in comparison to 80% in the FOLFIRINOX alone group.52 However, a different phase 1b study of a CCR2 inhibitor combined with chemotherapy as first-line treatment of metastatic pancreatic ductal adenocarcinoma raised safety concerns due to pulmonary toxicity and did not demonstrate an efficacy above chemotherapy alone.53 Thus, despite the promising results from preclinical models, further clinical studies are needed to assess the potential of CCR2 inhibitors to safely and effectively treat cancers in humans, particularly in combination with immunotherapy.

One of the limiting factors of this study is that mice in our model do not bear tumors and we cannot predict how the tumor burden would alter hepatic immune responses during experimental CPI-hepatitis. However, our data mirrors human CPI-hepatitis pathology9 11 26 suggesting that our observations of CPI-hepatitis are likely to be similar in tumor-bearing mice. Future work will focus on using our model further and assessing the effects of CCR2 inhibition during experimental CPI-hepatitis on a cancer background (eg, B16 melanoma model). This work will provide important insights into the effectiveness of CCR2 inhibition in treating CPI-hepatitis, but also shed light on its potential to boost antitumor immunity. Moreover, mice in our model show mild liver injury compared with other acute liver injury models (eg, paracetamol-induced liver injury).17 Human CPI-hepatitis is often characterized as a mild disease, which is generally treated with immunosuppressants, but can lead into fulminant hepatic failure and death.11 54 Our in vivo model should therefore be tuned to investigate different severity levels according to the Common Terminology Criteria for Adverse Events grading system and can be used to compare the effects of immunosuppressive therapies to CVC treatment.

In summary, using our experimental model, we have shown that liver infiltration of CCR2+ monocytes is necessary for the development of CD8+ T-cell-driven hepatocyte injury during CPI-hepatitis. Inhibition of CCR2-mediated monocyte recruitment is a rational target for future investigation. This work provides important insights into how CPI-hepatitis develops and highlights potential therapeutic pathways to treat liver inflammation in this condition.

Supplemental material

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

Acknowledgments was used to create the graphical abstract and figures of experimental set-up.


Supplementary materials


  • X @TriantafyllouE_

  • Contributors Conceptualization: CLCG, CGA, LAP, ET. Investigation, Data curation, Formal analysis: CLCG, EM, SRA, M-AM, ET. Interpretation of data: CLCG, SRA, RDG, NP, KJW, LAP, ET. Writing—original draft: CLCG. Writing—critical revision of manuscript: CLCG, SRA, MRT, NP, KJW, ALP, ET. Funding acquisition: CLCG, MRT, ST, JL, CGA, LAP, ET. Guarantor: CLCG.

  • Funding This project was supported by the Royal Marsden Cancer Charity, Rosetrees Trust (A1783, M439-F1, and CF2\100002), Academy of Medical Sciences (Clinical Lecturer Starter Grant to LP), Imperial College London (ICRF fellowship award to CG), UKRI Medical Research Council (MR/X009904/1 to ET) and Wellcome Trust (WT101159 and 225875 to NP), UK National Institute for Health Research (NIHR) and NIHR Imperial Biomedical Research Centre (BRC) and the Imperial College Wellcome Trust Strategic Fund.

  • Competing interests No, there are no competing interests.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.