Non-invasive Detection of Immunotherapy-Induced Adverse Events

Significant advances in understanding the role that immune checkpoints play in downregulating the immune response have led to a revolution in the field of immuno-oncology. The development of immune-checkpoint inhibitors (ICI) that activate cytotoxic T cells has demonstrated strikingly positive clinical outcomes across multiple tumor types (1, 2). With the growing clinical indications for immunotherapy agents, the number of patients with cancer eligible to receive checkpoint inhibitor drugs went from 1.54% in 2011 to nearly 44% in 2018 (3). Because there are more than 1,700,000 new cases of cancer diagnosed per year (4), nearly 760,000 new patients would be eligible to receive immunotherapy in the United States alone. As immune checkpoints are partly responsible for immune homeostasis, disturbing these pathways can contribute to the loss of immunological self-tolerance in patients (5). These therapy-induced immune-related adverse events (irAE) have hindered the safe administration of checkpoint inhibitors. irAEs are frequent, with astonishing occurrence that varies between 15% and up to 91% of patients with combination treatment (6, 7). irAEs may present within a few days of immunotherapy initiation and up to a year after completion of therapy (8). irAEs can occur with different severity levels in any organ system with dermatitis, colitis, hepatitis, pneumonitis, and nephritis being the most common (7, 9). irAEs may also be severe and life-threatening, often requiring prompt patient management with appropriate therapeutic decisions (10). In fact, approximately one third of patients will need to eventually stop immunotherapy treatment due to irAE onset (11–14), and patients may also need to initiate steroids or other immuno-modulatory drugs such as infliximab.

Even though the underlying mechanisms of irAEs remain unknown, they have been linked to overactivation of T cells resulting in healthy tissue damage (15, 16). It is known that cytotoxic CD8⁺ T lymphocytes are directly responsible for killing tumor cells upon recognition by releasing granzyme B and perforin (17–19). In fact, these lymphocytes are the main effector cells in clinically approved immunotherapy drugs, as well as many under development (20–22). The link between anticancer effects and adverse events is more evident with reports that patients who presented with irAEs also presented improved outcomes such as higher response rates and overall survival (7).

Early evaluation and diagnosis are critical to achieve safe resolution of irAEs. However, the diagnosis of irAEs can be challenging, and currently relies primarily on invasive biopsies with their attendant risks (23). Furthermore, treatment of severe irAEs not only involves the discontinuation of therapy, but also treatment with immunomodulators that not only treat the irAE but can also stop the antitumoral immune response. Considering the fact that the time course for developing irAEs is variable, can occur after one dose or after several months of therapy, and can affect multiple organs (8), a non-invasive approach of detecting and monitoring irAEs is urgently needed. Our group has developed a PET imaging tracer that specifically targets the active extracellular form of granzyme B and has previously demonstrated successful use of granzyme B PET imaging as a biomarker for predicting tumoral immunotherapy response (24–26). Herein, we hypothesized that granzyme B release could potentially be harnessed as a biomarker to allow non-invasive detection of irAEs. In this study, we demonstrate the use of a granzyme B–targeted peptide (GZP), radiolabeled with a PET isotope (⁶⁸Ga), as an agent for irAE identification in a mouse model of irAEs. Furthermore, we validated the usefulness of granzyme B for the assessment of irAEs by staining tissue specimens of human subjects with pathology-confirmed checkpoint inhibitor–associated irAEs. To the best of our knowledge, this is the first study reporting non-invasive imaging of irAEs.

All experimental procedures and animal studies were approved by the Institutional Animal Care and Use Committee (IACUC). The Murine MC38 cell line derived from C57BL6 murine colon adenocarcinoma cells were obtained from Kerafast. Cells were grown using DMEM medium supplemented with 10% FBS at 37°C and 5% CO₂. Mycoplasma testing was carried out by PCR screening on a monthly basis and cells were discarded after 15 passages. All cell-based experiments were done with cells acquired within 6 months to ensure fidelity of the cell line identity. Allograft tumors were implanted with the subcutaneous injection of MC38 cells (1 × 10⁶) in a 1:1 (v:v) ratio in Matrigel (Thermo Fisher Scientific) into the right shoulder of animals.

We addressed the challenge that mice treated with ICIs usually do not develop irAEs by using a previously reported mouse model in which clinical and pathologic equivalents of irAEs could be generated in animals by combining Treg depletion, which lowers self-tolerance, with checkpoint inhibitor therapy. In the Foxp3–GFP–DTR animal model, transient Treg depletion induced in these mice by administration of diphtheria toxin (DT), followed by administration of monoclonal antibody targeting immunomodulatory CD137 receptor, induced severe irAEs (27).

All animal studies were conducted using 8- to 12-week-old male B6.129(Cg)-Foxp3^{tm3(DTR/GFP)Ayr} (Foxp3–DTR–GFP) mice, purchased from The Jackson Laboratory. Male mice were used to ensure uniform knockout of the X-linked Foxp3 allele. Mice were housed and maintained by the Center for Comparative Medicine at Massachusetts General Hospital. To induce Treg depletion in vivo, tumor-bearing animals were injected intraperitoneally with 250 ng of DT (Sigma-Aldrich) diluted in PBS. This procedure was carried out three times, 3 days apart from each injection, that is, on days 3, 6, and 9 following MC38 tumor inoculation. The animals where then separated into four groups (n = 5). Each group received intraperitoneally three doses at 3 days apart (days 12, 15, and 19) of either PBS (DT-only group), both 200 μg of anti–PD-1 (clone RMP1–14; Bioxcell) and 100 μg of anti–CTLA-4 (clone 9D9; Bioxcell), or 250 μg of anti-CD137 (clone 3H3; Bioxcell). An additional group of five mice was treated with intraperitoneal injection of 250 μg of anti-CD137 (clone 3H3) followed by five daily doses of dexamethasone (intraperitoneally; 5 mg/kg dissolved in saline; Sigma-Aldrich). A control group of animals that did not receive DT and only received PBS was also used. The humane endpoint criteria were selected as tumor diameter > 20 mm [Mean = (d+D)/2, where d and D are the shortest and longest diameter in mm], a body weight decrease of > 20%, or if general health was deemed to be too poor to continue. Animals were monitored every other day and measurements of tumor volume and body weight were collected.

⁶⁸Ga was eluted from a ⁶⁸Ge/⁶⁸Ga generator (Eckert & Ziegler) with 0.1 mol/L HCl. Radiolabeling of NOTA-GZP was performed as previously described (25). In brief, 100 μg of NOTA-GZP (in 100 μL of PBS) was mixed with 2 mol/L HEPES buffer until pH was 3.5–4.0. Then, approximately 370 MBq of eluted ⁶⁸Ga was added to the mixture and the reaction proceeded for 10 minutes at room temperature. The reaction product was purified using a reverse-phase C18 Sep-Pak mini cartridge, eluted with 70% ethanol and diluted with saline before administration. Radiolabeling yield was calculated through instant thin-layer chromatography (iTLC) using two solvent systems as reported elsewhere (28, 29).

PET imaging was performed 1 hour post-injection of approximately 7.4 MBq of ⁶⁸Ga-NOTA-GZP via tail-vein injection. Tail-vein injection was performed on awake mice. For imaging, mice were anesthetized (isoflurane inhalation) and PET/CT images were acquired on a rodent Triumph PET/CT (Trident Healthcare). PET acquisition was performed in static mode for 15 minutes with a single-bed position, followed by CT acquisition. Images were reconstructed using 3D-MLEM algorithm (4 iterations and 20 subsets) and corrected for scatter and randoms. Uptake values presented as a percentage of injected dose per gram (%ID/g) for each organ was calculated in a 3D region of interest manually drawn around the tumor using CT images. Images were postprocessed using VivoQuant software (InviCRO). After the CT scan, the animals were euthanized, and major organs were harvested and decay-corrected retained activity per organ weight was measured using a gamma counter detector (Wizard², PerkinElmer). To validate the specificity of our tracer, we have also performed ex vivo biodistribution studies with a non-specific peptide (⁶⁸Ga-NSP) in animals with the generated irAEs model (DT + anti-CD137). In addition, we have performed PET imaging after administration of the specific (⁶⁸Ga-GZP) and non-specific (⁶⁸Ga-NSP) peptides in a local inflammation animal model. To induce local inflammation, a subcutaneous injection of lipopolysaccharide (LPS; 100 μL of 1 mg/mL LPS with 100 μL Matrigel) was administered into the right posterior shoulder of each mouse and the same volume of PBS into the contralateral shoulder as a control. Radiotracers were injected 24 hours after administration of LPS and PET imaging was performed at 1 hour post-injection time.

A separate cohort of animals paralleling the different groups above was used for flow cytometric analysis. Spleen, liver, kidney, and lungs were excised, minced, and strained through a 70-μm nylon mesh and incubated with RPMI with 1 mg/mL collagenase type IV for 35 minutes at 37°C for the generation of single-cell suspension. Cells were then stained with live/dead fixable violet dead cell stain kit (ThermoFisher) followed by staining against murine anti-CD3 (clone17A2; ThermoFisher), anti-CD4 (clone GK1.5; ThermoFisher), anti-CD8 (clone 53–6.7; ThermoFisher), and anti-CD107a (clone HA3; ThermoFisher) according to the manufacturer's specifications. Flow cytometry was performed on a cytometry cell analyzer BD LSR II and gating established using FlowJo software (version 8.7).

Ex vivo tissue staining was performed on animals from the different groups following Treg depletion and therapeutic protocols. Three days after the last therapeutic doses were given, animals were euthanized and liver, spleen, colon, kidney, lungs, and skin were harvested, fixed in 10% neutral-buffered formalin overnight, and submitted to the Specialized Histopathology Services of the MGH Pathology Core for processing, sectioning, and H&E staining. Before immunofluorescence staining against granzyme B, tissue was deparaffinized and rehydrated. Antigen retrieval was carried out using standard heat-based antigen retrieval techniques (30). Blocking was carried out for 1 hour at room temperature in PBS containing 5% goat serum. Primary antibody incubation was performed overnight at 4°C with the addition of rabbit anti-mouse granzyme B antibody (ab255598; Abcam). On the following day, stained slides were washed and incubated with AlexaFluor 647 conjugated goat anti-rabbit IgG secondary antibody (A32733; ThermoFisher) at room temperature in a moist dark chamber for 1 hour. A coverslip was applied to each slide using Vectashield mounting medium for fluorescence microscopy with DAPI (4′, 6-diamidino-2-phenylindole). Semiquantitative data of immunofluorescence were carried out by measuring total fluorescence in the red channel (granzyme B) divided by total fluorescence in the blue channel (DAPI). All images were acquired using Biotek Cytation 5 Cell Imaging Multi-Mode Reader and analyzed through Biotek Gen5 software.

Samples were provided by the Severe Immunotherapy Complications Service of Massachusetts General Hospital Cancer Center. All specimens were acquired from patients under an institutional review board–approved clinical protocol. Colon samples were obtained from patients undergoing checkpoint inhibitor treatment with or without immune-related colitis. The samples of patients that did not develop colitis were assigned as control. Kidney samples were obtained from patients who presented with checkpoint inhibitor–associated nephritis. Healthy kidney samples were used as control. Immunofluorescence staining against granzyme B was performed on either formalin-fixed samples (colon samples) as described above or on frozen samples (kidney samples). For these experiments, rabbit anti-human granzyme B antibody (ab243879; Abcam) was used as primary antibody. Frozen-tissue slices were fixed with 4% paraformaldehyde for 10 minutes, rinsed with PBS, and blocked with 5% goat serum before addition of primary and secondary antibodies. Rabbit anti-human granzyme B (ab243879; Abcam) primary antibody was incubated overnight at 4°C and then Alexa Fluor Plus 647–conjugated goat anti-rabbit IgG (A32733; ThermoFisher) secondary antibody was added. Cell nuclei were stained with DAPI. Fluorescence images were acquired with Cytation 5 Cell Imaging Multi-Mode Reader equipped with three excitation lasers (488, 546, and 633 nm).

All quantitative data were analyzed using GraphPad Prism (version 8.4.2) and are presented as the mean ± standard deviation. Comparisons between groups were made using two-way ANOVA or unpaired t test, where P < 0.05 was considered statistically significant.

Schematics of animal studies workflow can be found in Supplementary Fig. S1 and Supplementary Table S1. The peptide was radiolabeled with ⁶⁸Ga with a radiolabeling yield of 74.2% ± 3.6% and radiochemical purity of >97% (data not shown), and an average calculated specific activity of 5,180 ± 155 MBq/mg. PET/CT imaging of PBS and DT treatment control groups demonstrated low uptake in normal organs with tracer in the bladder reflecting its renal excretion; this is in line with our previous studies of the same probe in murine models (24). In contrast, in mice that received checkpoint inhibitors, there was high uptake in the abdominal area localizing to liver, kidney, spleen, and colon (Fig. 1A). The group injected with DT + anti-CD137 had the highest uptake values for all organs when compared with all other groups, with %ID/g values of 4.63 ± 0.77, 2.86 ± 1.69, 2.25 ± 1.33, and 1.58 ± 1.00 for colon, kidney, spleen, and liver, respectively. Those values were statistically higher than the PBS-injected group for spleen (P = 0.0034), colon (P < 0.0001), kidney (P = 0.0471), and tumor (P = 0.0063) and statistically higher than DT-only–injected group for spleen (P = 0.0140), colon (P < 0.0001), kidney (P = 0.0133), and tumor (P = 0.0322). %ID/g values for the group of animals injected with DT + anti–PD-1 + anti–CTLA-4 were 2.78 ± 1.98, 1.48 ± 1.03, and 1.17 ± 0.27 for colon, kidney, and spleen, respectively. Colon was the organ with the highest %ID/g for animals injected with DT + anti–PD-1 + anti–CTLA-4, which was statistically higher than the colon uptake of animals injected with PBS (P = 0.0062) and DT only (P = 0.0018). Of note, the tumor in the group injected with DT + anti-CD137 showed markedly increased uptake with values of 2.14 ± 0.29, six times higher than the tumor uptake of the PBS-injected group, which likely indicates response to immunotherapy (25). Ex vivo biodistribution data, shown in Supplementary Fig. S2, were in agreement with findings from PET and showed that for the animals that received DT + anti-CD137, multiple organs including kidney, spleen, colon, and tumor had significantly higher uptakes (P < 0.005) than the same organs in PBS-only injected mice. For the groups that received DT + anti–PD-1 + anti–CTLA-4, significantly (P = 0.0395) higher uptake in the colon was found when compared with the PBS-only–treated group. The organ with the highest uptake among the groups that received immunotherapy was found to be the colon with values of 10.8 ± 6.1 and 4.8 ± 1.5%ID/g for DT + anti-CD137 and DT + anti–PD-1 + anti–CTLA-4 groups, respectively; this was significantly higher (P < 0.05) than 0.7 ± 0.5%ID/g found for the PBS group. Of note, uptake values for blood pool in PET images and biodistribution studies were not statistically different for any of the groups investigated. Secondary analysis of organ-to-blood pool ratios from ex vivo biodistribution data (Supplementary Fig. S3) was performed. Although the same trends were observed as Supplementary Fig. S2, kidney-to-blood ratios of animals that received anti–PD-1 + anti–CTLA-4 were higher than that of the anti-CD137 group, suggesting that the effects in the kidney might be more prominent in animals that received combination therapy and further investigation is necessary.

To validate the specificity of our tracer and rule out differences in tracer uptake due to perfusion and/or inflammation, we performed ex vivo biodistribution studies after administration of a non-specific DOTA-conjugated peptide radiolabeled with ⁶⁸Ga (⁶⁸Ga-NSP) in animals treated with DT + anti-CD137. Non-specific radiotracer uptake was negligible in major organs investigated (Supplementary Fig. S4A) except the kidneys, which is in agreement with the elimination pathway of the probe. Uptake in the kidneys, spleen, colon, and tumors was significantly lower than uptake in the organs of animals injected with ⁶⁸Ga-GZP (Supplementary Fig. S4B), indicating that inflammatory status was not the driver of radiotracer accumulation in those organs and that ⁶⁸Ga-GZP uptake was indeed specific to granzyme B presence. Furthermore, as an additional investigation into the specificity of our tracer, we evaluated through PET imaging differences in uptake of our granzyme B–specific (⁶⁸Ga-GZP) and the non-specific peptide (⁶⁸Ga-GZP) in local inflammation. Results (Supplementary Fig. S5) demonstrate specific and significantly higher (P < 0.001) ⁶⁸Ga-GZP uptake in the region of LPS injection when compared with the PBS-injected contralateral region, as seen in PET images (Supplementary Fig. S5A) and ROI analysis (Supplementary Fig. S5B). Imaging with ⁶⁸Ga-NSP demonstrates negligible uptake in both LPS- and PBS-injected regions. More importantly, ⁶⁸Ga-GZP uptake was significantly higher (P < 0.0001) than ⁶⁸Ga-NSP uptake in the LPS-injected region. This experiment demonstrates the specificity of our radiotracer in the context of granzyme B activity.

Because corticosteroids are first-line treatment for the management of irAEs, we aimed to evaluate the impact of dexamethasone on granzyme B presence as well as on uptake of our imaging agent as a surrogate for whether resolution of irAEs can also be non-invasively assessed with granzyme B PET imaging. For that, we selected mice treated with DT + anti-CD137 that had exhibited the most severe and abundant adverse events and administered five doses of dexamethasone (5 mg/kg) on 5 consecutive days intraperitoneally 3 days after the last anti-CD137 treatment in 5 mice. We then performed PET imaging using the same protocol as specified before at 1 hour post-injection of ⁶⁸Ga-NOTA-GZP. As observed in Fig. 2A, overall PET signal in the abdominal area diminished considerably. Relatively high signal is still seen in bladder and kidney that is consistent with the elimination pathway of the probe. Values of ROI calculated from the PET images (Fig. 2B) demonstrated that although heart and liver uptake did not significantly change in the group that receive dexamethasone, uptake values were significantly lower in spleen (P = 0.0015), colon (P < 0.0001), kidneys (P = 00085), and tumor (P = 001) when compared with the group that did not receive dexamethasone. For the group that received DT + anti-CD137+ dexamethasone, %ID/g values were found to be 1.4 ± 1.01, 1.09 ± 0.59, 0.40 ± 0.17, and 0.43 ± 0.11 for colon, kidney, liver, and spleen respectively. Of note, for the group injected with dexamethasone, colon had the highest uptake value among the organs investigated (except bladder) and uptake values were still higher than that found for PBS- and DT-only groups. Biodistribution data (Supplementary Fig. S2) validated findings from PET imaging.

After PET/CT imaging, major organs were dissected for ex vivo analysis. Immunofluorescence staining for granzyme B demonstrated that, when compared with animals from the PBS-injected group, a higher granzyme B staining is seen for colon, liver, lung, kidney, and skin samples of animals injected with DT + anti-CD137 or DT + anti–PD-1 + anti–CTLA-4, indicating that some extent of immunotherapy-associated inflammation occurred in those organs (Fig. 3). The fluorescence staining was the strongest in specimens from mice injected with DT + anti-CD137, in concordance with findings from PET/CT scans. For the animals injected with anti-CD137, granzyme B staining was significantly higher in colon (P < 0.001), skin (P < 0.0001), and kidney (P < 0.05) when compared with the PBS-injected group (Supplementary Fig. S6). The group of animals that received DT + anti–PD-1 + anti–CTLA-4 had significantly higher staining in skin (P < 0.001) and colon samples (P < 0.05), when compared with the PBS-injected group.

In addition, a visual evaluation of hematoxylin & eosin (H&E)-stained samples was carried out to qualitatively assess the degree of inflammatory infiltration. As can be seen in Fig. 4, immune infiltrates are present in colon, liver, lungs, kidney, and skin for the animals injected with DT + anti-CD137 as well as in colon, liver, lungs, and skin of the animals injected with DT + anti–PD-1 + anti–CTLA-4. Of the organs analyzed, colon and skin had a higher degree of immune infiltrate presence for the mice treated with DT + anti-CD137 whereas the lungs and skin had the higher degree of infiltration for those treated with DT + anti–PD-1 + anti–CTLA-4. The presence of immune infiltration is not visualized in the organs of animals injected with PBS only. Of note, immunofluorescence staining against granzyme B for spleen and tumor revealed high granzyme B signal in spleen and tumor samples of animals from both groups injected with DT + immunotherapy that diminished after administration of dexamethasone, whereas granzyme B signal was not observed in heart samples of any of the groups investigated (Supplementary Fig. S7).

We proceeded to investigate effects of dexamethasone on histological evaluation of organ samples as well as on granzyme B expression within the tissues. It was possible to observe through immunofluorescence staining that granzyme B signal decreased in all organs investigated for animals injected with DT + anti-CD137 + dexamethasone when compared with those administered with DT + anti-CD137 only (Fig. 3; Supplementary Fig. S6). In a similar manner, diminished presence of immune infiltrates is found on the dexamethason-administered group as confirmed by H&E staining (Fig. 4).

In addition, mice were monitored and body weight as well as tumor dimensions were measured throughout the course of the study (Supplementary Fig. S8). It is possible to observe that the tumor volumes decreased significantly in the groups that received immunotherapy when compared with the PBS-only–treated group. Of note, after administration of dexamethasone, tumor volumes started to increase again, which can indicate that the immunosuppression affected immunotherapy efficacy. None of the animals from any of the groups investigated reached the humane endpoint criteria.

Flow cytometry was performed in excised tissue samples of all groups studied to quantify levels of CD3-, CD8-, and CD107a-positive cells as a separate means of analysis of cytotoxic T-cell infiltrate and activity state of infiltrating cells. Results can be seen in Fig. 5 and Supplementary Fig. S9. CD3-positive cells were highest for the animals injected with DT + anti-CD137 with values of 14.35% ± 4.45% and 7.78% ± 1.94% of total cells in kidney and liver samples, respectively, and highest in the animals injected with DT + anti–PD-1 + anti–CTLA-4 with values of 39.10% ± 9.47% and 27.95% ± 5.59% of total cells in lung and spleen samples, respectively. This reveals that CD3-positive cells were more prevalent in animals treated with DT + anti-CD137 or DT + anti–PD-1 + anti–CTLA-4 when compared with all other groups investigated. Levels of CD3-positive cells in animals treated with only PBS was found to be 2.55% ± 1.68%, 3.01% ± 1.37%, 1.83% ± 0.92%, and 5.54% ± 0.14% of total cells for kidney, liver, lungs, and spleen, respectively. Of note, when dexamethasone was administered, CD3-positive cells in kidney, liver, and lungs diminished and reached values of 6.62% ± 0.17%, 0.64% ± 0.08%, and 25.7% ± 0.70% of total cells, respectively, similar to the levels of the DT-only group in all organs investigated. This same trend is seen for CD8-positive cells, with levels at least 4 times higher for animals injected with either DT + anti-CD137 or DT + anti–PD-1 + anti–CTLA-4 when compared with the PBS-only–injected group. In kidney and liver samples, the highest values of CD8-positive cells were found for the group that received DT + anti-CD137 (5.69% ± 4.79% and 2.22% ± 1.36% of total cells, respectively). The group injected with DT + anti–PD-1 + anti–CTLA-4 had the highest values of CD8-positive cells with 28.5% ± 1.41% of total cells in the lungs. In addition, injection of dexamethasone decreased CD8-positive cell level in all organs investigated and reached values similar or lower than those found for the PBS-injected animals in kidney and liver samples. Most importantly, in the animals that received DT + anti-CD137, levels of CD8-positive cells that also had the granzyme B release marker (CD107a) were significantly higher (P < 0.05) in kidney and lungs with values of 3.63% ± 1.95% and 5.55% ± 0.74% of total cells, respectively, when compared with those animals that received PBS only (0.07% ± 0.03% and 1.25% ± 0.34% and 1.71% ± 0.01% of total cells for kidney and lungs, respectively). CD107a-positive cells were also more prevalent in all organ samples of the animals injected with DT + anti–PD-1 + anti–CTLA-4 when compared with PBS- and DT-only groups. As seen with CD3 and CD8 markers, administration of dexamethasone significantly decreased levels of CD107a-positive cells.

Immunofluorescence staining to verify the presence of granzyme B was carried out in human colon and kidney samples of patients that presented with ICI-associated colitis and nephritis, respectively (clinical information found on Supplementary Tables S2 and S3). As observed in Fig. 6A, granzyme B staining (in red) was increased in all patient kidney samples (n = 4) when compared with healthy kidney tissue sample (control). In addition, granzyme B signal was also identified in colon samples of all patients investigated (n = 3) in higher levels when compared with control colon samples of patients that did not develop colitis as a result of receiving checkpoint inhibitor drugs.

The therapeutic benefits of checkpoint inhibitor drugs are indisputable. However, the overall safety of ICIs is concerning, with an incidence of between 54% and 76% for the emergence of adverse events after monotherapy, with grade ≥3 adverse events (severe, life-threatening, or death; ref. 31) representing around 60% of the total irAEs (6, 32). Substantial individual variations in the risk of irAE development among patients and the idiosyncratic nature of irAEs contribute to difficulty in management (8). Diagnosis and management of irAEs should be done early to ensure minimal morbidity, to prevent life-threatening complications, and to continue immunotherapy when the diagnosis of an irAE is excluded (1, 23). However, diagnosis of irAEs relies mostly on biopsies and their inherent complications and usually requires multidisciplinary efforts, with the clinical expertise of multiple specialties as well as multiple laboratory exams and imaging techniques (33). For that reason, high-quality studies with a focus on irAEs are extremely important to improve early management and recognition of irAEs. Although the precise mechanisms of irAEs remain unclear and may differ depending on the drug used and the organ affected, it is known that checkpoint inhibitors increase cytotoxic T-cell activation and proliferation while diminishing regulatory T-cell functions, which can promote autoimmunity and, in turn, disturb immune homeostasis (34). In analysis of clinical specimens, CD8⁺ T-cell abundance has been linked to the development of irAEs (35, 36), and multiple reports have demonstrated intense CD8⁺ T-cell infiltrates in biopsy specimens of patients with irAEs (37).

Granzyme B, a serine-protease released by CD8⁺ T cells and natural killer cells during the cellular immune response, represents one of the two dominant mechanisms by which cytotoxic T cells mediate cancer cell death (38–40) and has been used as a T-cell activation marker (24–26, 41, 42). Our group has previously demonstrated the ability of a peptide-targeting granzyme B (GZP) imaging agent to predict immunotherapy response by identifying those tumors with high levels of activated T cells (24, 25). With the establishment of radiolabeled GZP as an indicator of T-cell activation, and the fact that irAEs are linked to CD8⁺ T-cell–mediated organ infiltration (43), we explored the use of that imaging agent in identifying irAEs in vivo.

We first demonstrated the ability of radiolabeled GZP to be used as a non-invasive imaging agent for the identification of irAEs. PET imaging revealed the establishment of irAEs in multiple organs after administration of DT followed by administration of either anti-CD137 or anti–PD-1 + anti–CTLA-4. The development of irAEs such as dermatitis and enteritis was clear upon observation of overall health conditions of animals that received DT + anti-CD137 and DT + anti–PD-1 + anti–CTLA-4. In our study, the organ demonstrating the highest granzyme B–positive signal was the colon. The tracer uptake in the colon reached more than 10% ID/g as calculated in biodistribution studies, even though clear quantification of signal in the colon was hampered by overlapping strong bladder signal in preclinical PET imaging. These findings are consistent with the fact that checkpoint inhibitor–induced colitis (CPI colitis) has been correlated with augmented activation and proliferation of cytotoxic effector CD8-positive T cells (37). In fact, these increased levels are so high that CD8⁺ T-cell levels have been used to differentiate CPI colitis from other colon diseases such as ulcerative colitis and Crohn's disease. GZP uptake levels can therefore potentially be used to discriminate CPI colitis from other inflammatory colon diseases and assist in therapeutic management decisions.

We also observed significantly higher tracer uptake in the kidneys of animals that received DT + anti-CD137, which could indicate the development of inflammatory disease related to the immunotherapy treatment. Although the most common tissues involved in irAEs are the skin, gastrointestinal tract, and endocrine system, ICI-associated acute kidney injury (CPI-AKI) has been reported (44). Non-invasive detection of CPI-AKI would be highly beneficial, as differentiating CPI-AKI from other causes of AKI is challenging and often requires a high-risk percutaneous renal biopsy (45).

We then sought to determine whether the organs with higher GZP uptake had signs of immunotherapy-induced toxicity and corresponding granzyme B presence. Histologic and immunofluorescent assessment of the major organs confirmed the presence of immune infiltrates, and granzyme B presence. Immune infiltrates were found in colon, kidney, skin, and lung tissues of the animals injected with DT + anti-CD137 and colon, skin, and lung tissues of animals injected with DT + anti–PD-1 + anti–CTLA-4. Immunofluorescent staining showed granzyme B expression in those same organs, confirming radiotracer specificity. Colon, as correlated with PET uptake, had the highest degree of inflammatory infiltration and granzyme B expression. Interestingly, skin and lung samples of both groups of animals also had very high levels of immune cell infiltrates. Even though tracer uptake in skin and lungs was not apparent in the PET scans, higher uptake levels were found for those groups when compared with all other groups when investigated through biodistribution studies, which further corroborates to the connection between granzyme B upregulation and adverse event establishment. Of note, although there is no specific irAE associated with increased T-cell infiltrate in the spleen, we noted a marked increase in T-cell infiltration and radiotracer accumulation in the spleen of mice treated with DT + anti-CD137, which speaks to overall upregulation of the immune system. We also noted a high degree of granzyme B signal variability (both by PET tracer uptake and immunofluorescence) across the livers of animals injected with DT + anti-CD137. We did see a correlation between PET uptake and immunofluorescence staining in the liver of animals from that group (i.e., the same animals that had higher immunofluorescence signal also had higher PET tracer uptake). The high degree of variability led to no statistically significant difference between this group and other treatment groups. Although investigation of variability within each group is beyond the scope of this article, we posit this could be due to a variable incidence of CPI hepatitis within this treatment group.

Our flow cytometry analysis demonstrated a higher prevalence of CD3⁺, CD8⁺ and CD8⁺CD107a⁺ cells for the animals treated with DT + immunotherapy in all organs investigated when compared with the PBS-injected group. Levels of CD107a, which is a marker of CD8⁺ degranulation following stimulation (46), reached values 51 times higher for kidney and 10 times higher for liver of the animals that received DT + anti-CD137 when compared with those found for the PBS-injected group. In a similar fashion, the numbers of CD107a-positive cells were 27 and 10 times higher for kidney and lungs of animals injected with DT + anti–PD-1 + anti–CTLA-4 when compared with the PBS-only group, respectively. Unfortunately, due to the difficulty in sample preparation for flow cytometry acquisition, colon and skin samples could not be analyzed. Altogether, findings from PET, biodistribution data, and ex vivo studies demonstrate a direct association between granzyme B upregulation and irAEs and suggest its potential use as a marker of irAE development

Notably, we have found more robust and severe signs of irAEs in animals that received DT + anti-CD137 than DT + anti–PD-1 + anti–CTLA-4. These differences between treatment groups in this mouse model have been previously described (27), and can be explained by significant differences in proliferating CD8⁺ T cells and serum levels of IFNγ and TNF found among the groups. In fact, in the clinical settings, agonistic antibodies targeting CD137 induced dose-dependent fatal hepatotoxicity and other toxic effects that led to the termination of a number of clinical trials (47).

Seeking to understand the effects of immunosuppressant drugs on GZP tracer uptake, we selected the group demonstrating the most severe adverse events (DT + anti-CD137) and treated those animals with the steroid dexamethasone. Treatment with dexamethasone lowered radiotracer uptake in all organs investigated. This same pattern was observed for granzyme B expression, as assessed by immunofluorescence staining. This finding is especially important in the kidneys, because urinary excretion of tracer can partially obscure calculation of organ-specific radiotracer uptake. The fact that, upon dexamethasone administration, tracer uptake significantly diminished in the kidneys suggests that ⁶⁸Ga-GZP can still potentially be used for the detection of irAEs in the kidneys. However, further studies focused on the pharmacokinetics of this radiotracer, including different imaging timepoints, are warranted to establish an optimal protocol for the detection of renal irAEs. Steroid-dependent decrease in organ immune infiltrate was also observed through flow cytometry analysis and histological staining for all organs investigated. Interestingly, although granzyme B signal and the presence of immune infiltrates in colon diminished after administration of dexamethasone, granzyme B presence and immune infiltration can still be found in the colon samples of those animals. This may be consistent with the fact that dexamethasone is at times insufficient to resolve irAEs in clinical practice (48, 49). Furthermore, we chose to evaluate the effect of steroids on resolving irAEs induced by a therapeutic regimen, anti-CD137, that has not been thoroughly evaluated in clinical practice, and the intensity of GZB uptake we observed may correlate to more severe irAEs than typically observed with PD-1/CTLA-4 inhibition. These can have important implications in the clinical practice because persistent high levels of granzyme B signal after steroids can indicate steroid resistance and suggest potential benefits in the utilization of other therapeutic targets such as anti-TNF drugs (50, 51). Interestingly, through our tumor volume growth curve, it was possible to observe that after injection of dexamethasone, tumor size started to increase again, which may indicate that the antitumor efficacy of checkpoint inhibition was affected by immunosuppression. That hypothesis was enriched by reduced tumor tracer uptake as seen by PET and biodistribution data for the group injected with dexamethasone when compared with the group injected with DT and anti-CD137. The ability to monitor the immune response in both target tumor as well in affected organ during irAE treatment may potentially be helpful in guiding the decision to stop ICI drug administration and be invaluable in the clinical management of immunotherapy patients.

Finally, we sought to demonstrate the clinical correlate of our preclinical imaging findings by analyzing the presence of granzyme B expression in biopsy samples from ICI-treated patients presenting with irAEs. The presence of granzyme B in those samples could indicate the clinical potential of the use of radiolabeled GZP for non-invasive detection and monitoring of irAEs. In fact, immunofluorescence staining of colon and kidney samples of patients presenting with immunotherapy-induced colitis and nephritis demonstrates very high levels of granzyme B in all investigated specimens, with negligible staining of healthy kidney and colon samples. These findings further suggest that GZP could be used to non-invasively detect irAEs in CPI-treated patients. Altogether, we believe that radiolabeled GZP has enormous potential in aiding clinicians through management of irAEs in two significant ways: (i) through diagnosis and monitoring of irAEs as herein demonstrated and (ii) monitoring of tumor response throughout the course of irAE treatment to potentially identify patients that will experience reduced antitumor efficacy due to the use of immunosuppressants.

Contrasting our preclinical GZP PET findings with the known clinical findings of FDG PET may be helpful to highlight GZP PET's future potential role for irAE assessment. GZP PET shows uniform low background uptake across organs susceptible to irAEs, with the exception of kidneys due to urinary excretion. This stands in contrast with variable FDG uptake seen within organs, including the colon and heart, that may make diagnosis of irAEs more challenging. From a mechanistic standpoint, numerous pathologies have increased glucose uptake and utilization, potentially confounding FDG PET in the assessment of adverse events and lowering specificity of imaged FDG signal, whereas granzyme B PET signal should be specific to cytotoxic immune cells. Nonetheless, future clinical studies comparing these two imaging agents are warranted.

Even though the overarching goal of this article was to non-invasively detect irAEs rather than to focus on the mechanistic details of irAE development, we understand that using preclinical mouse models of irAEs imposes limitations. Unfortunately, the development of irAEs is uncommon in preclinical mouse models, even when multiple checkpoint blockade pathways are targeted. This might be explained by the fact that different mouse strains are more resistant to developing irAEs when compared with humans or simply by the fact that experiments using animals are usually carried out and monitored over shorter periods of time (43). To our knowledge, the only murine model specifically investigated for its ability to recapitulate symptomatology similar to clinical irAEs (27) is the Foxp3–DTR conditional knockout mouse model, in which transient Treg depletion is induced by intraperitoneal administration of diphtheria toxin, lowering self-tolerance and thus allowing irAEs to be more easily induced following treatment with immunomodulatory antibodies. The authors found a significant increase in proliferating CD8⁺ T cells and serum levels of IFNγ and TNFalpha after administration of immunomodulatory antibodies in multiple organs, which correlated with irAE development and severity as well as antitumor effects. The correlation between CD8⁺ T-cell abundance and irAE development has been described in the clinical settings (35, 36, 52), and the known correlation between activated T cells and granzyme B presence (41, 42, 53) warranted our investigation of the use of a granzyme B–targeted peptide to non-invasively detect irAEs. However, the mechanisms of irAE development in these animals and how they correlate with clinical irAEs still needs to be investigated, as this mouse model differs meaningfully from the clinical environment where no specific intervention on regulatory T-cell activity is performed. In addition, clinical mechanistic details (54, 55) and identification of potential biomarkers (36, 56) of irAEs still need to be elucidated and efforts are being made to fully address this challenge.

In this proof-of-concept study, we show in vitro, in vivo, and ex vivo data that in aggregate suggest the possibility of detecting irAEs using a non-invasive imaging agent. This potential is further demonstrated by the fact that increased granzyme B expression was found in tissue samples of clinical patients that presented with CPI colitis and nephritis. More work remains to be done to determine optimal imaging timepoints for irAE-specific imaging, and to understand whether granzyme B expression can serve as a biomarker in human tissues in other organ-specific irAEs. We do believe that clinical investigation of granzyme B PET imaging to evaluate clinical irAE development is warranted to further answer these questions and determine the potential clinical value of this imaging agent.

In this study, we demonstrate an interconnection between the establishment of irAEs and granzyme B presence. Herein, to the best of our knowledge, we show for the first time the visualization of such events through PET imaging. We further demonstrate the presence of granzyme B in human tissue of clinical patients that presented with checkpoint inhibitor–associated adverse events, warranting the possibility of direct clinical translation. Taken together, our data suggest a direct role for granzyme B imaging in the clinical management of immunotherapy and immunotherapy-induced adverse events.

E. Wehrenberg-Klee reports grants, personal fees, and other support from CytoSite outside the submitted work, as well as a patent for Granzyme B imaging issued, licensed, and with royalties paid from CytoSite Biopharma. U. Mahmood reports grants from NIH and Melanoma Research Alliance during the conduct of the study, as well as grants, personal fees, and other support from CytoSite BioPharma outside the submitted work; U. Mahmood also reports a patent for Granzyme B imaging issued, licensed, and with royalties paid from CytoSite Biopharma. No disclosures were reported by the other authors.

C.A. Ferreira: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. P. Heidari: Conceptualization, resources, supervision, funding acquisition, validation, methodology, project administration, writing–review and editing. B. Ataeinia: Investigation. N. Sinevici: Software, methodology. M.E. Sise: Investigation. R.B. Colvin: Investigation. E. Wehrenberg-Klee: Conceptualization, supervision, writing–review and editing. U. Mahmood: Conceptualization, resources, software, supervision, funding acquisition, project administration, writing–review and editing.

This study was funded by a Melanoma Research Alliance award, NIH-R01CA214744 and NIH-R01DK123143 (to U. Mahmood). P. Heidari is supported by K08-CA249047.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.