Virus-specific memory T cell responses unmasked by immune checkpoint blockade cause hepatitis

Abstract

Treatment of advanced melanoma with combined PD-1/CTLA-4 blockade commonly causes serious immune-mediated complications. Here, we identify a subset of patients predisposed to immune checkpoint blockade-related hepatitis who are distinguished by chronic expansion of effector memory CD4⁺ T cells (T_EM cells). Pre-therapy CD4⁺ T_EM cell expansion occurs primarily during autumn or winter in patients with metastatic disease and high cytomegalovirus (CMV)-specific serum antibody titres. These clinical features implicate metastasis-dependent, compartmentalised CMV reactivation as the cause of CD4⁺ T_EM expansion. Pre-therapy CD4⁺ T_EM expansion predicts hepatitis in CMV-seropositive patients, opening possibilities for avoidance or prevention. 3 of 4 patients with pre-treatment CD4⁺ T_EM expansion who received αPD-1 monotherapy instead of αPD-1/αCTLA-4 therapy remained hepatitis-free. 4 of 4 patients with baseline CD4⁺ T_EM expansion given prophylactic valganciclovir and αPD-1/αCTLA-4 therapy remained hepatitis-free. Our findings exemplify how pathogen exposure can shape clinical reactions after cancer therapy and how this insight leads to therapeutic innovations.

Fig. 1. Individual predisposition to hepatitis after αPD-1/αCTLA-4 treatment.

a Individualised treatment of melanoma is guided by tumour staging, presence of B-RAF mutations and fitness-for-toxicity. b Colitis, hepatitis and thyroiditis are common immune-related complications of dual therapy with Nivolumab plus Ipilimumab; 31.4% of patients experienced two or more of these immune-related adverse reactions (n = 89). c Colitis, hepatitis and thyroiditis occurred independently and were not significantly associated with clinical response (n = 89; F.E). d–h Patients who developed hepatitis of any grade following dual therapy lacked biochemical signs of liver inflammation before treatment. In particular, no clinically meaningful differences in plasma levels of d aspartate transaminase (AST; n = 87; M.W.; Bonferroni-corrected p-value, m = 5), e alanine transaminase (ALT; n = 89; M.W.; Bonferroni-corrected p-value, m = 5), f gamma glutamyl transaminase (γ-GT; n = 89; M.W.; Bonferroni-corrected p-value, m = 5), g total bilirubin (n = 87; M.W.; Bonferroni-corrected p-value, m = 5), or h C-reactive protein (CRP; n = 85; M.W.; Bonferroni-corrected p-value, m = 5) were observed between patients who developed hepatitis and those who did not. Median values are indicated by a red line. i, j Biochemical markers of tumour burden were not different between patients who developed hepatitis and those who did not. i Pre-treatment levels of lactate dehydrogenase (n = 89; M.W.; Bonferroni-corrected p-value, m = 4). Median values are indicated by a red line. j Pre-treatment levels of protein S100 (n = 89; M.W.; Bonferroni-corrected p-value, m = 4). k–m No association was observed between seropositivity for k hepatitis B virus core antigen (HBcAg; n = 85; F.E.; Bonferroni-corrected p-value, m = 3), l hepatitis C virus (HCV; n = 85; F.E.; Bonferroni-corrected p-value, m = 3), or m hepatitis E virus (HEV; n = 67; F.E.; Bonferroni-corrected p-value, m = 3) and development of hepatitis following dual therapy. n No association was observed between rounds of αPD-1/αCTLA-4 administered and development of hepatitis (n = 89; M.W.).

Fig. 2. Circulating CD4⁺ T effector memory cell frequency predicts hepatitis after Nivolumab plus Ipilimumab treatment.

a Peripheral blood samples were collected from melanoma patients with metastatic disease receiving αPD-1/αCTLA-4 therapy immediately before administration of the first dose (n = 89). Leucocyte subsets differentially represented in patients with or without hepatitis were identified in a randomly assigned training set (B.H.-corrected t-tests; n = 44; m = 50; FDR = 0.25). Red dots indicate significantly differently represented subsets. Example gating strategies for analysis of flow cytometry data are provided as Supplementary Figs. 2–8. b CD4⁺ T_EM % in training set patients with or without treatment-related hepatitis (n = 44; M.W.). Median values are indicated by a red line. c CD4⁺ T_EM % in validation set patients with or without treatment-related hepatitis (n = 45; M.W.). d ROC analysis of CD4⁺ T_EM % as a discriminatory marker for treatment-related hepatitis in the validation set (n = 45). e Comparison of the bimodal distribution of CD4⁺ T_EM % in patients with unresectable metastatic disease (n = 107) and the normal distribution (n = 49; K2 = 2.79; p = 0.248) of CD4⁺ T_EM % in patients with completely resected tumours. A cut-off of CD4⁺ T_EM ≥ 21% was set (indicated by a dashed red line) below which 99% of completely resected tumour cases should fall. Four pink points represent CD4⁺ T_EM^≥21% patients with metastatic disease who were electively treated with αPD-1 monotherapy. f In the validation set, 68.9% patients were correctly classified using a cut-off of CD4⁺ T_EM ≥ 21 %, whereas 55.6% were correctly classified under the no-information model (n = 45; F.E.). g CD4⁺ T_EM ≥ 21 % is not a marker of predisposition to αPD-1/αCTLA-4-related colitis (n = 89; F.E.). h CD4⁺ T_EM^≥21% patients did not experience more severe hepatitis than CD4⁺ T_EM^<21% patients (n = 38; F.E.). Dashed red line indicates a cut-off of CD4⁺ T_EM ≥ 21%. i Time-to-first presentation of hepatitis was not different between CD4⁺ T_EM^≥21% (n = 12) and T_EM^<21% (n = 26) patients (log-rank). j Twelve of 12 patients with unresectable metastatic melanoma and CD4⁺ T_EM ≥ 21% developed hepatitis after αPD-1/αCTLA-4 dual therapy. By contrast, 3 of 4 CD4⁺ T_EM^≥21% patients treated with αPD-1 monotherapy did not develop hepatitis (F.E.; p = 0.007).

Fig. 3. CD4⁺ T_EM frequency is an independent predictor of αPD-1/αCTLA-4-related hepatitis.

a CD4⁺ T_EM % did not correlate with age (n = 89; Pearson). Dashed red line indicates a cut-off of CD4⁺ T_EM ≥ 21%. b CD4⁺ T_EM % did not correlate with body mass index (n = 89; Pearson). c CD4⁺ T_EM % did not correlate with C-reactive protein (CRP) levels (n = 85; Pearson). d CD4⁺ T_EM % did not correlate with aspartate aminotransferase (AST) levels (n = 87; Pearson). e CD4⁺ T_EM % did not correlate with alanine aminotransferase (AST) levels (n = 89; Pearson). f CD4⁺ T_EM % did not correlate with gamma-glutamyltransferase (γ-GT) levels (n = 89; Pearson). g CD4⁺ T_EM % did not correlate with total bilirubin levels (n = 89; Pearson). h Glucose metabolism measured by ¹⁸F-fluoro-2-deoxy-d-glucose (¹⁸F-FDG) uptake in PET/CT studies is a quantitative marker of liver inflammation. Calculating standard uptake ratio (SUR_mean Liver) allows accurate quantification of ¹⁸F-FDG uptake in liver by correcting for differential clearance of tracer from blood and liver parenchyma. Higher SUR_mean Liver values indicate more severe inflammation. T₀ = reference time point at 75 min post-injection and T = actual scan time. i SUR_mean Liver was lower in CD4⁺ T_EM^≥21% patients than CD4⁺ T_EM^<21% patients who developed hepatitis indicating less inflammation at baseline in the CD4⁺ T_EM^≥21% subgroup (n = 28; Pearson). j CD4⁺ T_EM^≥21% was not associated with the number of organs containing one or more metastases ≥1.5 cm in diameter (n = 103; F.E.). k CD4⁺ T_EM^≥21% was not associated with the presence of hepatic metastases (n = 103; F.E.). l CD4⁺ T_EM % did not correlate with lactate dehydrogenase (LDH) levels (n = 88; Pearson). m CD4⁺ T_EM % did not correlate with protein S100 levels (n = 88; Pearson). n CD4⁺ T_EM^≥21% was not associated with any prior therapy (n = 89; F.E.).

Fig. 4. CD4⁺ T_EM cell enrichment is associated with chronic activation and increased numbers of effector memory CD4⁺ T cells.

a–d Increased CD4⁺ T_EM frequency before treatment reflects a decrease in circulating naïve CD4⁺ T cells and an increase in circulating numbers of CD4⁺ T_EM cells. a Baseline naïve CD4⁺ T cell counts (n = 103; M.W.; Bonferroni correction: m = 4, p = 5.2 ⨯ 10⁻⁵). b Baseline CD4⁺ T_EM cell counts (n = 103; M.W.; Bonferroni correction: m = 4, p = 8.0 ⨯ 10⁻⁴). c Baseline CD4⁺ T_EMRA cell counts (n = 103; M.W.; Bonferroni-corrected p-value, m = 4). d Baseline CD4⁺ T_CM cell counts (n = 103; M.W.; Bonferroni-corrected p-value, m = 4). Median values are indicated by a red line. e Pairwise correlations between CD4⁺ T_EM cell frequency and other leucocyte subset frequencies in the training and validation sets (n = 103; Pearson). Example gating strategies for analysis of flow cytometry data are provided as Supplementary Figs. 2–8. f Baseline CD27⁺ CD4⁺ T cell frequencies (n = 103; M.W.; Bonferroni correction: m = 60, p = 4.2 ⨯ 10⁻¹⁰). g Baseline CD57⁺ CD4⁺ T cell frequencies (n = 103; M.W.; Bonferroni correction: m = 60, p = 2.0 ⨯ 10⁻⁹). h Baseline HLA-DR⁺ CD4⁺ T cell frequencies (n = 103; M.W.; Bonferroni correction: m = 60, p = 4.1 ⨯ 10⁻⁵). i Baseline CD279⁺ CD4⁺ T cell frequencies (n = 103; M.W.; Bonferroni-corrected p-value, m = 60). j Baseline CD160⁺ CD244⁺ CD8⁺ T cell frequencies (n = 103; M.W.; Bonferroni-corrected p-value, m = 60). k Baseline Torque Teno Virus (TTV) loads. Uninfected patients were censored from analysis (n = 74; M.W.; Bonferroni-corrected p-value, m = 60). Median values are indicated by a red line.

Fig. 5. CD4⁺ T_EM cell expansion correlates with immunity to cytomegalovirus.

a Seasonal presentation of CD4⁺ T_EM^≥21% patients between 2017 and 2020 (n = 103; F.E.). Dashed red line indicates a cut-off of CD4⁺ T_EM ≥ 21%. b CD4⁺ T_EM^≥21% status was associated with high serum levels of anti-CMV IgG antibodies (n = 100; F.E.; p = 1.2 ⨯ 10⁻⁵). Dashed red line indicates a cut-off of CD4⁺ T_EM ≥ 21%. c CD4⁺ T_EM^≥21% status was associated with CMV-reactivity in pp65 ELISPOT (n = 53; F.E.; p = 9.0 ⨯ 10⁻⁵). Dashed red line indicates a cut-off of CD4⁺ T_EM ≥ 21%. d Development of hepatitis was associated with CMV-seropositivity and CD4⁺ T_EM^≥21% status (n = 89). Median values are indicated by a red line. e ROC analysis showing CD4⁺ T_EM % is a superior discriminator of patients at risk of hepatitis when considering only CMV IgG⁺ cases as opposed to all cases. f Classification of patients with unresectable metastatic melanoma who did or did not develop αPD-1/αCTLA-4-related hepatitis according to CMV IgG status and baseline CD4⁺ T_EM cell frequency using a revised cut-off of CD4⁺ T_EM ≥ 16%. Red boxes indicate 34 of 40 (85%) cases correctly classified by our model (n = 40; F.E.; p = 1.4 ⨯ 10⁻⁵). Green box indicates 2 of 17 (11.7%) cases predicted to develop hepatitis who did not. Pink box indicates 4 of 23 (17.4%) cases predicted not to develop hepatitis who did. Blue box indicates 19 of 49 CMV IgG⁻ patients not considered by our model who developed hepatitis.

Fig. 6. CMV-reactive CD4⁺ T cells are enriched in patients with CD4⁺ T_EM cell expansion.

CMV-reactive CD4⁺ T cells in patients with unresectable metastatic melanoma were assayed by in vitro stimulation with CMV lysates (blue boxes). Neutralising antibodies (red hatching) were used to detect T cells unable to respond to CMV antigens owing to expression of PD-1 or CTLA-4. T cell responses were quantified by flow cytometry analysis of cytokine expression. a Example data from a CMV IgG⁺ CD4⁺ T_EM^≥16% patient who developed hepatitis. Gating of live, singlet, IFN-γ-producing CD4⁺ T cells is illustrated in Supplementary Fig. 11. b Frequencies of IFN-γ-producing CD4⁺ T cells in patients categorised according to CMV IgG serostatus and baseline CD4⁺ T_EM frequency (n = 42; two-way ANOVA with Tukey correction for multiple comparisons; n.s. = not significant). Boxplots represent the median, 25th and 75th percentiles, and Tukey whiskers. c Absolute numbers of circulating IFN-γ-producing CD4⁺ T cells patients categorised according to CMV IgG serostatus and baseline CD4⁺ T_EM frequency (n = 42; two-way ANOVA with Tukey correction for multiple comparisons; n.s. = not significant). Boxplots represent the median, 25th and 75th percentiles, and Tukey whiskers.

Fig. 7. Case 1: valganciclovir treatment of late-onset αPD-1/αCTLA-4-related hepatitis.

A 54-year-old male patient who received αPD-1 (Nivolumab) plus αCTLA-4 (Ipilimumab) dual therapy for metastatic melanoma presented with late-onset hepatitis. a Course of treatment. b Change in hepatitis-related parameters over time and their association with introduction, withdrawal and re-introduction of valganciclovir treatment. c The patient presented at the end of week 40 with recrudescent hepatitis and was treated with 900 mg/day valganciclovir for 2 days prior to liver biopsy. Histopathological image of the liver biopsy (H&E staining; scale bar 500 µm). d Generally, the liver parenchyma appeared normal (H&E staining; scale bar 100 µm). e Only a few lymphocytes and sparse necrotic hepatocytes were observed in portal areas with small bile plugs (arrows) with no signs of hepatitis (H&E staining; scale bar 100 µm).

Fig. 8. Case 2: valganciclovir treatment of late-onset αPD-1/αCTLA-4-related hepatitis.

A 49-year-old male patient who received αPD-1 (Nivolumab) plus αCTLA-4 (Ipilimumab) dual therapy for metastatic melanoma presented with late-onset hepatitis. a Course of treatment. b Change in hepatitis-related parameters over time and their association with introduction of valganciclovir treatment. c Histopathological image of a liver biopsy taken at the start of week 72 showing signs consistent with drug toxicity, autoimmunity or viral infection (H&E staining; scale bar 500 µm). d Extensive centrilobular necrosis (arrows) involving 30–40% of hepatocytes was observed (H&E staining; scale bar 100 µm). e Dense inflammatory infiltration of lymphocytes, eosinophils and neutrophils was seen in portal areas (H&E staining; scale bar 100 µm).

Fig. 9. Valganciclovir prophylaxis prevents αPD-1/αCTLA-4-related hepatitis.

a Heatmaps showing peak values of AST, ALT and bilirubin expressed as multiples of sex-adjusted upper limit of normal ranges. Fifteen of 17 (88.2%) patients classified as CMV IgG⁺ CD4⁺ T_EM^≥16% at baseline subsequently developed hepatitis. In contrast, 4 of 4 CMV IgG⁺ CD4⁺ T_EM^≥16% patients treated with 900 mg/day prophylactic valganciclovir remained hepatitis free. b In four initial cases, prophylactic valganciclovir prevented development of hepatitis (n = 21; F.E.).