PD-1 Modulates Radiation-Induced Cardiac Toxicity through Cytotoxic T Lymphocytes

Abstract

Introduction:

Combined immune checkpoint blockade has led to rare autoimmune complications, such as fatal myocarditis. Recent approvals of several anti–programmed death 1 (anti–PD-1) drugs for lung cancer treatment prompted ongoing clinical trials that directly combine PD-1 inhibitors with thoracic radiotherapy for locally advanced lung cancer. Overlapping toxicities from either modality have the potential to increase the risk for radiation-induced cardiotoxicity (RICT), which is well documented among patients with Hodgkin’s disease and breast cancer.

Methods:

To investigate cardiotoxicity without the compounding pulmonary toxicity from thoracic radiotherapy, we developed a technique to deliver cardiac irradiation (CIR) in a mouse model concurrently with PD-1 blockade to determine the presence of cardiac toxicity by using physiological testing and mortality as end points along with histological analysis.

Results:

We observed an acute mortality of 30% within 2 weeks after CIR plus anti–PD-1 antibody compared with 0% from CIR plus immunoglobulin G (p = 0.023). Physiological testing demonstrated a reduced left ventricular ejection fraction (p < 0.01) by echocardiogram. Tissue analyses revealed increased immune cell infiltrates within cardiac tissue. Depletion of CD8-positive lymphocytes with anti-CD8 antibody reversed the acute mortality, suggesting that the toxicity is CD8-positive cell-mediated. To validate these findings using a clinically relevant fractionated radiotherapy regimen, we repeated the study by delivering five daily fractions of 6 Gy. Similar mortality, cardiac dysfunction, and histological changes were observed in mice receiving fractionated radiotherapy with concurrent anti–PD-1 therapy.

Conclusions:

This study provides strong preclinical evidence that radiation-induced cardiotoxicity is modulated by the PD-1 axis and that PD-1 blockade should be administered with careful radiotherapy planning with an effort of reducing cardiac dose.

Introduction

Radiation-induced cardiac toxicity (RICT) is a well-documented event that may occur after excessive radiation to the thorax. Although the mechanisms of radiation toxicity are multifactorial, damage to the cardiac tissue and subsequent fibrosis seem to be the underlying events.¹ Damage can occur to any portion of the heart, including the pericardium, myocardium, coronary arteries, capillaries, and conduction system. These toxicities can manifest either slowly over time, leading to significantly increased long-term cardiovascular morbidity and mortality, or acutely in the days to weeks after treatment.^2,3 Long-term follow-up of patients with Hodgkin’s lymphoma who received large doses of radiation to the thorax revealed greatly increased incidence of coronary artery disease, congestive heart failure, valvular disease, and conduction abnormalities.⁴

Of note, subclinical cardiac abnormalities can be common even in relatively asymptomatic patient cohorts.⁵ Because it can often be asymptomatic and insidious, RICT is likely underrecognized after thoracic radiation. In one prospective analysis of patients with locally advanced lung cancer who were receiving high doses of cardiac radiation, approximately half developed conduction abnormalities by electrocardiographic screening.⁶

Recently, RICT has received more attention as a significant cause of morbidity and mortality in the setting of lung cancer after the dose escalation arm of RTOG 0617, which had survival inferior to that of the control arm per an earlier report.⁷ Secondary analysis of this trial found that higher cardiac dose was associated with inferior survival, indicating that RICT may be resulting in increased mortality.⁸ A recent update on RTOG 0617 demonstrated reduction of survival by 9 months from high-dose radiotherapy (RT), which is associated with higher cardiac dose.⁹ Other events can also occur, including effusion, myocardial infarction, pericarditis, and heart failure. Risk factors for cardiac events include mean heart dose and prior cardiac disease.¹⁰

Recently, immune checkpoint inhibitors have seen increasing clinical use with promising initial results.^11,12 By inhibiting programmed death ligand 1 (PD-1), novel drugs such as pembrolizumab and nivolumab increase host recognition of malignant tissue to enhance T-cell–mediated immunity in the hope of generating a durable and lasting antitumor response.¹³ However, only a fraction of patients experience objective clinical benefit after monotherapy. For this reason, clinical trials combining multiple checkpoint blockade have explored the potential for synergistic immune activation. Although these trials achieved increased immune-mediated antitumor activity, they also resulted in increased rates of autoimmune-related toxicity.¹⁴ Already, fatal cases of fulminant autoimmune-mediated myocarditis with lymphocytic infiltration have been reported in patients receiving combined anti–cytotoxic T-lymphocyte–associated protein 4 and anti–PD-1 therapy. Of note, these severe cardiac events appear to occur more frequently in patients treated with combined immune checkpoint blockade than in those treated with monotherapy.¹⁵

In addition to the generation of free radicals and double-strand DNA breaks, the antitumor effect of radiation therapy is partially immune mediated.¹⁶ Ionizing radiation leads to neoantigen generation, increases MHC class I expression, and increases proinflammatory cytokine release.^17,18 Cytotoxic T cells have a critical role, as CD8-positive T cells are necessary for the ionizing radiation to be effective in murine models.¹⁹ Preclinical and clinical evidence exists to suggest that combining radiation therapy with immune checkpoint blockade may result in greater response rates than with either modality alone.^20–23 For this reason, many clinical trials combining radiation therapy with immune checkpoint blockade are ongoing.²⁴ However, enhanced immune activation is a double-edged sword; although it may enhance antitumor response, it can also result in autoimmune toxicities.

Programmed death ligand 1 (PD-L1) is enriched in cardiac tissue to maintain self-tolerance,²⁵ and an autoimmune-mediated cardiomyopathy develops in PD-1–deficient mice.²⁶ Further, patients with a history of autoimmune disease seem to be predisposed to various radiation toxicities.²⁷ We therefore hypothesized that PD-1 would exacerbate RICT. By using a murine model and acute mortality as an end point, we demonstrated that anti–PD-1 therapy exacerbated RICT if administered concurrently. We also observed changes in physiological testing, immunohistochemistry, and the degree of cardiac fibrosis. These changes were observed whether radiation was delivered in a single fraction of 20 Gy or 30 Gy in five daily fractions.

Material and Methods

SARRP for CIR

Mice were immobilized and irradiated per the manufacturer’s instructions (Xstrahl, Surrey, UK). Briefly, the small animal radiation research platform (SARRP) for cardiac irradiation (CIR) involved cone beam computed tomography scan, which was performed to allow target (heart) contouring and radiation planning using Muriplan (Camberley, UK). Clinical guidelines were used to evaluate the various plans. An optimal plan was selected and used to treat mice using 220-KV and 13-mA photon beams.

In Vivo Treatments

Eight- to 12-week-old C57BL/6 mice (Taconic Biosciences, Rensselaer, NY) were purchased and randomly assigned to various experimental groups. Echocardiography was performed before killing of the animals and tissue collection. Before CIR, mice were intraperitoneally injected with 10 mg/kg of anti–PD-1 or 10 mg/kg of control immunoglobulin G (IgG), respectively,²⁸ starting 1 day before the cardiac irradiation. To deplete CD4-positive and CD8-positive T cells, mice were given 100 μg of anti-CD4 (Bio X Cell, West Lebanon, NH) or anti-CD8 (Bio X Cell) every 3 days for 4 weeks, beginning with the first dose of anti–PD-1, as described in Figure 1B.²⁹

Figure 1.

CD8 lymphocytes mediate acute mortality from cardiac irradiation (CIR) plus anti-programmed anti-programmed death 1 (anti-PD-1). (A). Immune cell infiltration was analyzed by immunofluorescence. CD45 was used as a marker for lymphocytes. CD4 and CD8 were used to identify CD4-positive T cells and CD8-positive T cells, respectively. F4/80 was used a marker for macrophages. (B) Schema of the T-lymphocyte subset deletion experiment. Mice receiving anti-PD-1 (black arrow) and 20 Gy of CIR in a single fraction (lightning bolt) were randomized to receive either anti-CD4 or anti-CD8 (black triangle). Survival curves for T-cell depletion experiment. Anti-CD8 reversed mortality from CIR plus anti-PD-1. A log-rank test was used for all statistical analyses (p < 0.05 [n = 10]). IgG, immunoglobulin G; mAb, monoclonal antibody; NS, not significant; H&E, hematoxylin and eosin; DAPI, 4′,6-diamidino-2-phenylindole.

Echocardiography

When mice displayed lethargy or weight loss greater than 20%, they underwent echocardiography. Transthoracic echocardiography was performed using the Vevo 2100 System equipped with a 40-MHz probe (Fujifilm VisualSonics, Toronto, Ontario, Canada) as previously described³⁰ at the Jefferson Translational Medicine Core. After the mice were immobilized under anesthesia, acoustic coupling gel was applied to the chest after hair removal. Gains were adjusted to eliminate background noise and five to 10 cycles were recorded to ensure accurate readings.

Histological Analysis and Immunofluorescence

Murine hearts were collected for histological analysis and immunochemistry as described.³¹ For immunofluorescence, paraffin-embedded sections were deparaffinized for antigen retrieval by using Antigen Unmasking Solution (H-3300, Vector Laboratories, Burlingame, CA). Frozen sections were fixed with acetone and rinsed with phosphate-buffered saline. Nonspecific binding was blocked by incubating sections in the supernatant of 0.5% casein in phosphate-buffered saline stirred for 1 hour. Sections were incubated overnight at 4°C with the following primary antibodies: CD45 (1:100) (R&D Systems, Minneapolis, MN), CD4 (1:50) (Thermo Fisher Scientific, Waltham, MA), CD8 (1:50) (R&D Systems), and F4/80 (1:200) (Abcam, Cambridge, MA). Slides were washed and incubated with appropriate fluorochrome-conjugated secondary antibodies for 1 hour and nuclei were counterstained with 4′,6-diamidino-2-phenylindole (SigmaAldrich, St. Louis, MO). Stained sections were imaged at × 40 on a confocal microscope (Leica, Buffalo Grove, IL). For quantitation, eight to 10 images of duplicate tissue sections per heart were acquired. Stained cells per high-powered field were counted from two nonoverlapping images by an observer without knowledge of the treatment groups by using the Fiji-Image analysis program (National Institutes of Health, Bethesda, MD).

Reverse-Transcriptase PCR

RNA was extracted from heart tissue by using TRIzol (Thermo Fisher Scientific). The RNA amount and its purity were determined by optical density at 260 and 280 nm by using Nanodrop (ThermoFisher Scientific, USA). Complementary DNA was synthesized with 1 mg of RNA in a final volume of 20 μL using Superscript III (Invitrogen, Carlsbad, CA) and oligomers. Quantitative polymerase chain reaction (PCR) was performed in a 20-μL reaction volume containing 100 μg of complementary DNA, RNase-free water, Fast SYBR Green Master Mix (Applied Biosystems, Foster City, CA), and 25 pmol of each primer. The primer sequences for mouse interferon gamma (IFN-γ) and tumor necrosis factor-α TNF-α were as follows: 5’-AACGCTACACACTGCATCTTGGG-3’ (forward) and 5’-GCCGTGGCAGTAACAGCCGCC-3’ (reverse) and 5’-GCACAGAAAGCATGACCCGG-3’ (forward) and 5’-GCCCCCCATCTTTTGGGGG-3’ (reverse) (Integrated DNA Technologies, Coralville, IA), respectively. The PCR setup was as follows: after 10 minutes at 95°C, 30 cycles of amplification were followed by a 10-minute extension at 72°C. Each cycle included denaturation at 94°C for 30 seconds, annealing at 58°C for 30 seconds, and extension at 72°C for 1 minute. As a control, β-actin mRNA was analyzed by using the following primers: 5’-TCCTTCGTTGCCGGTCCACCA-3’ (forward) and 5’-ACCAGCGCAGCGATATCGTCTC-3’ (reverse).

Flow Cytometry

All antibodies were purchased from eBioscience (San Diego, CA) and Biolegend (San Diego, CA) and titrated against isotype controls. Cells were stained with optimal amounts of each antibody for 30 minutes at 4°C in staining buffer per the standard protocol provided by manufacturers. For intracellular staining, cells were fixed for 20 minutes at 4°C by using a Fix/Perm Kit, (BD Biosciences, San Jose, CA) per the manufacturers’ protocol. Data (10 000 events) were acquired on a LSR II and BD Fortessa flow cytometers (BD Biosciences, San Diego, CA) and analyzed with Flowjo software (FlowJo LLC, Ashland, OR).

Masson’s Trichrome Staining and Hydroxyproline Assay for Collagen Content

Masson’s trichrome staining was performed by the tissue core laboratory at Thomas Jefferson University. Stained sections were imaged at ×40 on a confocal microscope (Leica). Hydroxyproline was measured to determine the total collagen content in the cardiac tissue by using a Hydroxyproline Assay Kit (SigmaAldrich). Tissue homogenates (100 μL) were added with 100 μL of 12 M hydrochloric acid and hydrolyzed at 120°C for 3 hours. Next, 50 μL of supernatant was transferred to a 96-well plate. After the samples were dried under a vacuum, 100 μL of the chloramine T/oxidation buffer mixture was added and incubated at room temperature for 5 minutes. Then, 100 μL of diluted DMAB reagent was added and incubated for 90 minutes at 60°C. Absorbance was measured at 560 nm in a microplate reader.

Statistical Analysis

A two-tailed Student’s t-test was used to determine p values between two groups. For multiple groups, analysis of variance was used to determine p values, followed by Tukey or Bonferroni correction. Differences in survival were determined by a log-rank test. Error bars represented SD unless otherwise indicated. All the statistical analyses were done with GraphPad Prism 6 (GraphPad Software, La Jolla, CA). A probability value of p less than 0.05 was considered statistically significant.

Results

Dosimetric Planning of CIR

To establish a mouse model of investigating radiation injury in an organ-specific manner, we used SARRP to target the murine heart in an attempt to minimize radiation exposure in the surrounding organs (e.g., lungs, esophagus, and spinal cord). Pretreatment onboard cone beam computed tomography with 360 projections acquired at approximately 1-degree angular increments were obtained to contour hearts and surrounding organs, and different radiation plans were designed for testing. To conform the radiation portal to the murine heart, a 10-mm circular collimator was selected. Four separate radiation plans were designed. Plan A used two diagonal crossing fixed beams delivered through the anterior of the heart (Fig. 2A) with the aim to simulate clinical 3-dimensional conformal RT (see Fig. 2A). Plan B used two anterior-posterior fixed beams to simulate clinical two-dimensional RT (Fig. 2B). Plan C used one rotating beam angled from 120 degrees to −120 degrees clockwise to simulate partial single-arc RT (Fig. 2C). Plan D used one rotating-beam from −179 degrees to 179 degrees clockwise to simulate whole single-arc RT (Fig. 2D). Plan A had the shortest treatment time, whereas plan D has the longest treatment time. The target coverage and homogeneity were similar among the four plans, but the conformity was the worst in plan B. Furthermore, compared with plan A, plan B had higher doses to the surrounding normal tissue whereas plans C and D had a low integral dose given to a larger volume of the surrounding normal tissue (see Fig. 2). Plan A produced favorable target coverage with satisfactory surrounding normal tissue sparing and minimum treatment time; thus, plan A was selected for CIR in this study.

Figure 2.

Establishment of a mouse model for studying the toxicities from cardiac irradiation (CIR). The dose distributions of transverse, sagittal, and coronal sections and dose-volume histogram (DVHs) of different CIR plans with a 10 × 10-mm circular collimator. (A) Two diagonal crossing beams. (B) Two anterior-posterior (AP-PA) beams. (C) One partial arc (from 120 degrees to −120 degrees clockwise. (D) One single arc (from −179 degrees to 179 degrees clockwise). Plan Awas used in this clinical experiment.

Increased Acute Mortality and Reduced Cardiac Output after CIR with Concurrent anti-PD-1

We have previously demonstrated increased acute mortality from thoracic radiation that targets both the heart and the bilateral lungs when anti–PD-1 antibody was concurrently administered.²⁸ To determine whether the acute mortality was mainly a consequence of cardiac toxicity rather than a consequence of pulmonary toxicity, we performed heart-targeting versus lung-targeting irradiation by using a SARRP irradiator. This allows targeted irradiation toward either organ without a significant radiation dose to the other as already described. To determine whether CIR plus anti–PD-1 antibody results in acute mortality, C57BL/6 mice were randomly stratified into four groups receiving control IgG (n = 20), anti–PD-1 (n = 20), CIR plus IgG (n = 20), or CIR plus anti–PD-1 (n = 20) (Fig. 3A). As shown in Figure 3B, mice receiving CIR with concurrent anti–PD-1 have 40% acute mortality compared with 0% mortality in mice receiving CIR plus IgG, anti–PD-1 alone, or IgG alone. In contrast, a similar experiment was performed by using lung-targeting irradiation with heart sparing to determine whether lung irradiation had a role in acute mortality. All groups including mice receiving lung-targeting irradiation with concurrent anti–PD-1 had 0% acute mortality (data not shown).

Figure 3.

Increased acute mortality and reduced cardiac output after cardiac irradiation (CIR) with concurrent anti-programmed death 1 (anti-PD-1). (A) The schedule of the experiment. (groups 1-4) anti-PD-1 monoclonal antibody (mAb) or immunoglobulin G (IgG) control was intraperitoneally administrated 24 hours before of CIR (black lightning bolt) with a loading dose of 200 μg, then every other day with 100 μg per mouse. (B) Survival curves for each group. Mice receiving combined treatment had an inferior survival (p = 0.023). (C) Ejection fraction and fractional shortening decreased significantly in the group receiving CIR plus anti-PD-1, whereas LVID-s increased significantly in the same group. **p < 0.01, *p < 0.05. NS, not significant.

Mice receiving CIR were monitored daily and subjected to echocardiography to determine their cardiac output before they were humanely killed when presenting with lethargy or weight loss of greater than 20%. Parameters of left ventricle systolic function such as left ventricle ejection fraction (LVEF), fraction shortening (FS), and left ventricular internal diameter at end-systole (LVID-s) were measured. As shown in Figure 3C, group 3 (CIR + IgG) exhibited a significant decline in both LVEF and FS compared with group 1 (IgG alone) (LVEF = 50.85% ± 5.06% versus 62.43% ± 7.54% [p < 0.01] and FS = 25.50% ± 3.13% versus 33.48% ± 5.48% [p < 0.01]), indicating an impact on cardiac function from CIR. Further impact from concurrent anti-PD-1 was observed from a significant decline in both EF and FS and an increase in LVID-s in group 4 (CIR + anti-PD-1) compared with in group 3 (CIR + IgG) (LVEF = 38.84% ± 6.06% versus 50.85% ± 5.06% [p < 0.01], FS = 18.58% ± 3.28% versus 25.50% ± 3.13% [p < 0.01], and LVID-s = 3.22 ± 0.29 versus 2.86 ± 0.32 [p = 0.021]). These data suggest that CIR decreased the cardiac contractility and output, and concurrent PD-1 inhibition exacerbated this side effect.

CD8 Lymphocytes Mediate Acute Mortality from CIR plus anti-PD-1

After the echocardiography, mice were humanely killed for histological analyses. As shown in Figure 1A, infiltrating immune cells in the cardiac tissue were examined by immunofluorescence analyses and found to be significantly more numerous in group 4 (CIR + anti–PD-1) than in group 3 (CIR + IgG) (CD45-positive cells = 90.11 ± 4.58 versus 50.89 ± 4.54 [p < 0.00002], CD4-positive T cells = 7.00 ± 0.91 versus 5.56 ± 0.97 [p = 0.30], CD8-positive T cells = 8.11 ± 1.30 versus 4.11 ± 0.72 [p = 0.0003], and F4/80-positive macrophages = 24.89 ± 3.35 versus 15.89 ± 1.67 [p = 0.029]). A significant increase was also present in group 3 versus in group 1 (CD45-positive lymphocytes = 50.89 ± 4.54 versus 30.78 ± 3.54 [p < 0.003], CD4-positive T cells = 5.56 ± 0. 97 versus 1.33 ± 0.41 [p < 0.001], CD8-positive T cells = 4.11 ± 0.72 versus 0.89 ± 0.30 [p < 0.0008], and F4/80-positive macrophages = 15.89 ± 1.67 versus 5.22 ± 0.74 [p < 0.00002]). In addition, CD8-positive cell infiltrates were more numerous in group 2, than in group 1 (2.11 ± 0.48 versus 0.89 ± 0.30 [p = 0.049]). These results suggest that cytotoxic CD8-positive T cells are likely a key mediator for acute cardiac toxicities. To validate this hypothesis, depleting antibodies, either anti-CD4 or anti-CD8, were given with anti–PD-1 before CIR, as outlined in Figure 1B. As shown, depletion of CD8-positive cells (group 6) reversed the acute mortality from CIR plus anti–PD-1, whereas depletion of CD4 did not (p < 0.05 [n = 10]).

Acute Cardiac Mortality and Active T-Cell Response from fCIR plus anti-PD-1

To determine whether the observed cardiotoxicity is independent of fractionation of CIR, we repeated the study using a biologically equivalent fractionated regimen (6 Gy delivered over five fractions) as outlined in Figure 4A. Again, acute mortality of 40% was observed in the group (n = 10) receiving fractionated CIR (fCIR) plus anti–PD-1 compared with in the mice receiving fCIR plus IgG (p = 0.027), as shown in Figure 4B. Irradiated cardiac tissue samples were collected for flow cytometry and reverse-transcriptase PCR analyses to examine the presence of active CD4-positive and CD8-positive T cells and cytotoxic T-cell response (IFN-γ and TNF-α) in Figure 4C. Flow cytometry analyses as shown in Figure 5A and B demonstrated significantly more active CD4-positive and CD8-positive T cells that were also positive for IFN-γ and TNF-α. In addition, levels of IFN-γ and TNF-α transcripts by reverse-transcriptase PCR were significantly higher in the cardiac tissue from mice receiving fCIR plus anti-PD-1, as shown in Figure 5C.

Figure 4.

Increased cytotoxic T-cell response and acute cardiac mortality from fractionated cardiac irradiation (fCIR) plus anti-PD-1. (A) Mice were subjected to 30Gy delivered over 5 fractions with either control immunoglobulin G (IgG) or anti-programmed death 1 (anti-PD-1). (B) Survival curves were determined by a log-rank test (p < 0.05 [n = 10]). (C) Immune infiltration was analyzed by immunofluorescence. CD45 was used as a marker for lymphocytes. CD4 and CD8 were used to identify CD4-positive T cells and CD8-positive T cells, respectively. F4/80 was used a marker for macrophages. H.E., hematoxylin and eosin; IFNγ, interferon gamma; TNFα, tumor necrosis factor-α.

Figure 5.

(A and B) Flow cytometry analyses of cardiac tissue. (C) Tumor necrosis factor-α (TNF-α) and interferon gamma (IFN-γ) expression by real-time polymerase chain reaction using irradiated cardiac tissue. **p < 0.01, *p < 0.05 (n = 3 or 4 in each group). anti-PD-1 (anti-programmed death 1); fCRT, fractionated conformal radiotherapy.

Acute Fibrosis in Cardiac Tissue Treated with fCIR Concurrently with PD-1 Blockade

To determine whether the reduced cardiac output is secondary to injury to cardiomyocytes and the subsequent fibrosis, we performed Masson’s trichrome staining on the irradiated cardiac tissue treated with either IgG or anti–PD-1. Nonirradiated heart was used as a negative control. As shown in Figure 6A, significantly more collagen deposits were found in irradiated cardiac tissue that were exposed to anti–PD-1. Consistently, collagen content as determined by hydroxyproline assay was significantly higher in the cardiac tissue from mice treated with fCIR plus anti–PD-1 (as shown in Figure 6B). fCIR plus anti–PD-1 appeared to induce rapid replacement fibrosis after death of cardiomyocytes, which can lead to stiffening of cardiac tissue as well as to arrhythmia. Both can lead to reduced cardiac output and acute mortality.

Figure 6.

Acute fibrosis in cardiac tissues treated with fractionated cardiac irradiation (fCIR) concurrently with programmed death 1 (PD-1) blockade. (A) Cardiac tissue samples were stained with Masson’s trichrome. (B) Hydroxyproline was measured by using the same cardiac tissue samples with the Hydroxyproline Assay Kit (SigmaAldrich) per the manufacturer’s instruction. **p < 0.01, *p < 0.05. (n = 3). IgG, immunoglobulin G.

Discussion

This study provides strong preclinical evidence to support the conclusion that PD-1 modulates R1CT and that concurrent PD-1 inhibition and C1R increases R1CT. We observed increased acute morality from anti–PD-1 treatment in mice treated with radiation whether it was delivered in a single dose of 20 Gy (as commonly utilized in published preclinical models³¹) or in a biologically equivalent fCIR regimen (30 Gy delivered over five fractions). The observed acute mortality was associated with decreased cardiac output, increased lymphocytic infiltration, and fibrosis. Although our experimental model may not truly reflect how patients with lung cancer are treated (as the entire heart does not generally receive high doses of radiation), our data raise the concern for cardiotoxicity when patients with lung cancer receive a significant cardiac dose in the setting of concurrent anti–PD-1 immunotherapy.

R1CT has been viewed as a late toxicity after treatment of certain cancers (e.g., Hodgkin’s lymphoma, breast cancer, lung cancer, esophageal cancer), and most patients have no clinical symptoms of cardiac injury for months to years after the treatment.^32–34 Only patients (especially those with Hodgkin’s lymphoma) who have survived decades allow these late toxicities to be quantified, which makes such quantification challenging among patients with locally advanced lung cancer with a median survival of 20 to 30 months. Although acute RICT is generally clinically silent,³⁵ acute cardiac death after RT has also been reported.³⁶

It is possible that the surviving mice in this preclinical study likely suffer from subclinical cardiac toxicity, which could lead to significant late sequelae given evidence of cardiac fibrosis. The pathogenesis of the observed acute sudden death in mice remains to be investigated. It is thought that intimal hyperplasia of coronary arteries may result in sudden death,³⁷ although cardiac arrhythmia is also a plausible cause. Here, we have demonstrated that a PD-1-mediated immune mechanism contributes to the exacerbation of RICT through increased fibrotic changes.

RT changes tumor immunity with increased numbers of CD8-positive T cells in the tumor microenvironment after ablative RT, which mediates therapeutic synergism between RT and anti–PD-1 therapy.^19,29 Radiation to the normal tissue, including the heart, can potentially trigger an autoimmune response through a similar mechanism. In the present study, infiltration with inflammatory cells, including CD8-positive T cells, was also observed in cardiac tissue from the group receiving CIR alone compared with in cardiac tissue from the control group, which corroborates previous findings. This was further enhanced by anti–PD-1 treatment. These immune infiltrates may have led to subsequent fibrosis and impaired contractibility of the heart. The increased fibrosis is of particular importance, as it is thought to be responsible for a variety of radiation induced cardiac toxicities.^1,2

PD-1 is induced in activated T lymphocytes, B lymphocytes, and myeloid cells. PD-1 plays a major role in attenuating autoimmune responses in the setting of inflammation as well as tumor immune escape.^38,39 For example, in myocardial disease, PD-L1 is localized on endothelium and cardiomyocytes, where it is critical for control of immune-mediated cardiac injury and polymorphonuclear leukocyte inflammation.²⁵ Consistently, PD-1-deficient mice have significant atrial myocardial fibrosis, oxidative stress, and structural remodeling that leads to atrial fibrillation.⁴⁰

Like cardiomyocytes, cancer cells may express high levels of PD-L1, which binds PD-1 to inactivate T-cell-mediated responses. New immunomodulatory antibodies have been designed against the PD-1 pathway to impair this negative immune checkpoint with promising clinical results¹³; however, only a fraction of patients respond to monotherapy. Trials combining multiple immune checkpoints are ongoing in the hope of increasing the response rate through the generation of increased antitumor response.²⁴ Although results have been promising, they are not without risk. Already, various types of autoimmune-mediated cardiotoxicity have been reported after combined blockade of PD-1 and cytotoxic T-lymphocyte-associated protein 4.¹⁵

Because RT activates the immune system by nonredundant mechanisms,⁴¹ combination therapies will likely see increasing use. The abscopal effect, as seen in combined RT with immune checkpoint inhibition, may enable systemic regression of all malignant lesions with the sensitized immune system,^20,42–44 providing a powerful anticancer combination. However, the sensitized immune reaction is a double-edged sword because it may also worsen normal tissue toxicities. Currently, clinical data suggest that the combination of RT with PD-1/PD-L1 inhibition may improve survival, but little attention is given to cardiac toxicity, and self-reporting toxicity assessment in clinical trials may contribute to an underassessment of this side effect.

Radiation dose to the heart poses a clinical challenge. Although radiation dose to the heart can be reduced through the use of volumetric modulated arc therapy or intensity-modulated radiation therapy in clinical practice, it remains challenging when the tumor is in close proximity to the heart, contributing to a higher mean heart dose.^8,45 Furthermore, sparing heart tissue can inadvertently increase lung dose, which can lead to increased pulmonary toxicity. Most importantly, disease progression is much more common than adverse cardiac events in patients receiving concurrent chemo-radiation,¹⁰ so adequate tumor coverage is paramount. As a consequence, the heart dose cannot always be avoided when delivering thoracic radiation for lung cancer.

In summary, the present study establishes a preclinical murine RICT model with dosimetric, functional, and immunohistologic analyses to investigate how PD-1 inhibition increases RICT. The present study provides evidence that the combination of PD-1 inhibition increases the cardiac toxicity caused by CID through cytotoxic CD8-positive T-cytotoxic cell-mediated myocarditis with associated fibrosis. Our data provide strong preclinical evidence that immune-related adverse events should be considered when combining PD-1 inhibitors and RT.

We therefore anticipate an increased incidence of cardiac toxicity in patients receiving thoracic radiation with concurrent PD-1 inhibition during the coming years, especially in those who are receiving high doses of radiation to cardiac tissue. This potentially lethal side effect requires clinical vigilance and a novel therapeutic approach to minimize this potentially life-threating toxicity.