Volume 123, Issue 19 p. 3807-3815
Original Article
Free Access

Assessment of programmed death-ligand 1 expression and tumor-associated immune cells in pediatric cancer tissues

Robbie G. Majzner MD

Robbie G. Majzner MD

Department of Pediatrics, Stanford University, Stanford, California

Search for more papers by this author
Jason S. Simon PhD

Jason S. Simon PhD

Bristol-Myers Squibb, Princeton, New Jersey

Search for more papers by this author
Joseph F. Grosso PhD

Joseph F. Grosso PhD

Bristol-Myers Squibb, Princeton, New Jersey

Search for more papers by this author
Daniel Martinez BS

Daniel Martinez BS

Department of Pathology and Laboratory Medicine, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania

Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

Search for more papers by this author
Bruce R. Pawel MD

Bruce R. Pawel MD

Department of Pathology and Laboratory Medicine, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania

Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

Search for more papers by this author
Mariarita Santi MD, PhD

Mariarita Santi MD, PhD

Department of Pathology and Laboratory Medicine, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania

Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

Search for more papers by this author
Melinda S. Merchant MD, PhD

Melinda S. Merchant MD, PhD

Pediatric Oncology Branch, National Cancer Institute, Bethesda, Maryland

Search for more papers by this author
Birgit Geoerger MD, PhD

Birgit Geoerger MD, PhD

Department of Pediatric and Adolescent Medicine, Gustave Roussy Institute, Villejuif, France

Search for more papers by this author
Imene Hezam

Imene Hezam

Department of Pediatric and Adolescent Medicine, Gustave Roussy Institute, Villejuif, France

Search for more papers by this author
Virginie Marty

Virginie Marty

Department of Medical Biology and Pathology, Gustave Roussy Institute, Villejuif, France

Search for more papers by this author
Phillippe Vielh MD, PhD

Phillippe Vielh MD, PhD

Department of Medical Biology and Pathology, Gustave Roussy Institute, Villejuif, France

Search for more papers by this author
Mads Daugaard PhD

Mads Daugaard PhD

Vancouver Prostate Center, Vancouver, British Columbia, Canada

Department of Urologic Sciences, University of British Columbia, Vancouver, British Columbia, Canada

Search for more papers by this author
Poul H. Sorensen MD, PhD

Poul H. Sorensen MD, PhD

British Columbia Cancer Agency, Vancouver, British Columbia, Canada

Search for more papers by this author
Crystal L. Mackall MD

Crystal L. Mackall MD

Department of Pediatrics, Stanford University, Stanford, California

Search for more papers by this author
John M. Maris MD

Corresponding Author

John M. Maris MD

Department of Pediatrics, University of Pennsylvania, Philadelphia, Pennsylvania

Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

Corresponding author: John M. Maris, MD, Department of Pediatrics, Division of Oncology, Children's Hospital of Philadelphia, Perelman School of Medicine at the University of Pennsylvania, Colket Translational Research Building, 3501 Civic Center Boulevard, No. 3060 4399, Philadelphia, PA 19104; Fax: (215) 426-0685; [email protected]Search for more papers by this author
First published: 13 June 2017
Citations: 136

We thank Hector Inzunza for his help in organizing pathology slides and samples.

Abstract

BACKGROUND

Programmed death 1 (PD-1) signaling in the tumor microenvironment dampens immune responses to cancer, and blocking this axis induces antitumor effects in several malignancies. Clinical studies of PD-1 blockade are only now being initiated in pediatric patients, and little is known regarding programmed death-ligand 1 (PD-L1) expression in common childhood cancers. The authors characterized PD-L1 expression and tumor-associated immune cells (TAICs) (lymphocytes and macrophages) in common pediatric cancers.

METHODS

Whole slide sections and tissue microarrays were evaluated by immunohistochemistry for PD-L1 expression and for the presence of TAICs. TAICs were also screened for PD-L1 expression.

RESULTS

Thirty-nine of 451 evaluable tumors (9%) expressed PD-L1 in at least 1% of tumor cells. The highest frequency histotypes comprised Burkitt lymphoma (80%; 8 of 10 tumors), glioblastoma multiforme (36%; 5 of 14 tumors), and neuroblastoma (14%; 17 of 118 tumors). PD-L1 staining was associated with inferior survival among patients with neuroblastoma (P = .004). Seventy-four percent of tumors contained lymphocytes and/or macrophages. Macrophages were significantly more likely to be identified in PD-L1–positive versus PD-L1–negative tumors (P < .001).

CONCLUSIONS

A subset of diagnostic pediatric cancers exhibit PD-L1 expression, whereas a much larger fraction demonstrates infiltration with tumor-associated lymphocytes. PD-L1 expression may be a biomarker for poor outcome in neuroblastoma. Further preclinical and clinical investigation will define the predictive nature of PD-L1 expression in childhood cancers both at diagnosis and after exposure to chemoradiotherapy. Cancer 2017;123:3807–3815. © 2017 American Cancer Society

INTRODUCTION

Immune checkpoint blockade has emerged as a major tool for harnessing the immune system to control malignancy.1 Three agents have already been approved by the US Food and Drug Administration in several adult malignancies,2-4 and several more are in clinical trials.5, 6 Phase 3 clinical trials have demonstrated that agents that disrupt the programmed death 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis have significant efficacy, including the ability to induce complete responses and prolong survival in previously untreatable cancers.7-9 PD-1 blocking agents have exhibited activity across a wide range of histologies, from hematologic malignancies10, 11 to solid tumors.7, 12, 13 By blocking the interaction of the T-cell coreceptor PD-1 to its ligands, PD-L1 and PD-L2, these agents disinhibit T cells and promote an antitumor immune response.

Pediatric oncology has witnessed great improvements in survival largely through the administration and optimization of combination chemotherapy. Impressive reductions in cancer deaths and treatment-related mortality have been reported with leukemias, lymphomas, and localized sarcomas.14 However, limited progress has occurred for specific clinical groups, including metastatic solid tumors and brain tumors, related in part to de novo and acquired resistance to chemotherapy.15-17 Immunotherapy presents a promising modality in pediatric oncology that may offer a new therapeutic option for patients with chemotherapy-refractory disease.18-20 Ch14.18, a chimeric anti-GD2 monoclonal antibody, increases survival in pediatric patients with metastatic neuroblastoma,21 and chimeric antigen receptor-modified T cells have shown promise in early phase clinical trials among children with relapsed leukemia.22, 23 However, to date, there have been limited preclinical or clinical data focused on checkpoint blockade for pediatric cancers. A phase 1 pediatric trial of ipilimumab was completed but did not produce objective responses.24 Clinical studies of PD-1 blockade have not yet been completed in pediatric patients, and little is known regarding PD-L1 expression in common childhood cancers. Some studies have implicated tumor expression of PD-L1 as a negative prognostic factor in cancer7, 25, 26 and as a predictor of response after treatment with PD-1–blocking antibodies.27, 28

In this study, we characterized PD-L1 expression in a wide variety of pediatric tumor samples. In addition, to further understand the immune microenvironment in pediatric cancer, we identified tumor associated immune cells (TAICs) (lymphocytes and macrophages) in these samples by immunohistochemistry (IHC).

MATERIALS AND METHODS

Tumor Samples

Archived samples representing a variety of pediatric tumor types were obtained from multiple sources. Whole slide sections of osteosarcoma (N = 20) as well as several tumor microarrays (TMAs) were obtained from the Children's Hospital of Philadelphia. Each TMA was comprised of 0.6-mm cores in duplicate, triplicate, or greater. Whole slide sections (N = 71) of Ewing sarcoma, neuroblastoma, rhabdomyosarcoma, and Burkitt lymphoma were obtained from Gustave Roussy Comprehensive Cancer Center (Villejuif, France). Tumors were diagnostic samples with the exception of 5 recurrent ependymoma samples and 7 post-therapy neuroblastoma samples. Informed consent was obtained from all patients or their guardians for use of their samples for research, and local institutional review boards confirmed that this analysis did not constitute human subjects research.

Assessment of PD-L1 and TAICs by Immunohistochemistry

An automated and validated PD-L1 IHC assay co-developed by Bristol-Myers Squibb (New York, NY) and Dako (Carpinteria, Calif) with a rabbit antihuman PD-L1 antibody (clone 28-8) from Epitomics Inc (Burlingame, Calif) was used to assess PD-L1 expression in archived tumor samples.29 Analyses were performed at Mosaic Laboratories (Lake Forest, Calif) by personnel trained in the use of the assay. Deparaffinization, rehydration, and target retrieval were performed in the target-retrieval solution on the Dako PT Link pretreatment module. PD-L1 staining was conducted on a Dako Autostainer Link 48 using the following general procedure, with a buffer rinse performed after each step: peroxidase-blocking, followed by PD-L1 antibody application, followed by antirabbit linker application, followed by visualization reagent application, followed by diaminobenzidine application, followed by diaminobenzidine enhancer application, followed by hematoxylin application.

Scoring for PD-L1 expression was performed manually by a pathologist trained in the assessment of the Dako PD-L1 IHC assay. Positive cellular staining was defined as complete circumferential or partial linear plasma membrane staining. Negative cellular staining was defined as no staining of the plasma membrane. Positive tumor samples were scored according to the percentage of tumor cells that had 1 + or greater plasma membrane staining, with a minimum of 100 cells evaluated. Samples were considered uninterpretable if <100 cells were scored.

Identification of TAICs

All samples were assessed for the presence of lymphocytes and macrophages (eg, TAICs) by histology, and TAICs were visually assessed for the presence of PD-L1 expression. A sample was considered positive for TAICs if it contained any lymphocytes or macrophages by hematoxylin and eosin (H&E) staining. TAICs were defined as PD-L1–positive if there was complete circumferential or partial linear staining of the plasma membrane of any cells.

RESULTS

PD-L1 Expression

All 91 whole slide samples and at least 1 core from 360 individual tumors on the TMAs had sufficient tissue present to allow counts of >100 cells and thus were judged as adequate for analysis (N = 451). The source of all samples (TMA or whole slides) as well as the details of their acquisition before therapy (N = 439) or after therapy (N = 12) are provided in Supporting Table 1 (see online supporting information). Tumor cells in 39 samples (9%) (Table 1) stained positive for PD-L1 above a 1% threshold30, 31 (range, 1%-50% of tumor cells; median, 2%). The highest proportion of positive samples was observed in Burkitt lymphoma, in which 8 of 10 samples (80%) were positive for PD-L1. PD-L1 staining also was observed frequently in glioblastoma multiforme (GBM) (5 of 14 samples; 36%) and neuroblastoma (17 of 118 samples; 14%). Other histotypes that demonstrated PD-L1 positivity included ganglioneuroblastoma (2 of 18 samples; 11%), ependymoma (2 of 42 samples; 5%), osteosarcoma (1 of 20 samples; 5%), rhabdomyosarcoma (1 of 53 samples; 2%), and synovial sarcoma (1 of 1 sample; 100%). Representative images are provided in Figure 1. Details of all PD-L1-positie samples, including PD-L1 scoring, are available in Supporting Table 2 (see online supporting information). There was no statistically significant difference between PD-L1 staining scores derived from the TMAs or the whole slides. PD-L1 staining was not observed in Ewing sarcoma (N = 25) or medulloblastoma (N = 40).

Details are in the caption following the image

Representative samples of programmed death-ligand 1 (PD-L1) staining are shown. (a) Burkitt lymphoma (from a whole slide) exhibits partial PD-L1 plasma membrane-positive staining of tumor cells (black arrows) and tumor-associated macrophages (red arrow). (b) Glioblastoma multiforme (from a tumor microarray) exhibits positive circumferential PD-L1 membrane staining of tumor cells (arrows). (c) Neuroblastoma (from a tumor microarray) has negative PD-L1 staining of tumor cells (green arrows), but PD-L1–positive lymphocytes (black arrow) and macrophages (red arrow) are observed.

Table 1. Programmed Death-Ligand 1 Expression and Tumor-Associated Immune Cell Analysis According to Tumor Type
TAIC Analysis
Diagnosis (No.) Percentage of Samples With PD-L1–Positive Tumors (No.) Percentage of Samples Containing Lymphocytes (No.) Percentage of Samples Containing Macrophages (No.) Percentage of TAIC-Positive Tumors Positive for PD-L1 on TAICs
Burkitt lymphoma (10) 80 (8) 0 (0) 100 (10) 80 (8)
Glioblastoma multiforme (14) 36 (5) 71 (10) 7 (1) 20 (2)
Neuroblastoma (118) 14 (17) 83 (98) 32 (38) 24 (24)
Ganglioneuroblastoma (18) 11 (2) 89 (16) 0 (0) 0 (0)
Atypical teratoid/rhabdoid tumor (16) 6 (1) 94 (15) 0 (0) 13 (2)
Ependymoma (42) 5 (2) 62 (26) 0 (0) 4 (1)
Osteosarcoma (20) 5 (1) 100 (20) 70 (14) 45 (9)
Rhabdomyosarcoma (53) 2 (1) 89 (47) 28 (15) 30 (14)
Synovial sarcoma (1) 100 (1) 100 (1) 100 (1) 100 (1)
Supratentorial PNET (5) 20 (1) 60 (3) 0 (0) 0 (0)
Ewing sarcoma and peripheral PNET (25) 0 (0) 92 (23) 44 (11) 35 (8)
Medulloblastoma (40) 0 (0) 63 (25) 3 (1) 0 (0)
Astrocytoma, WHO Grades 1-2 (63) 0 (0) 38 (24) 0 (0) 0 (0)
Anaplastic astrocytoma and anaplastic oligodendroglioma (7) 0 (0) 14 (1) 0 (0) 0 (0)
Wilms tumor (4) 0 (0) 100 (4) 25 (1) 25 (1)
Ganglioneuroma (10) 0 (0) 70 (7) 0 (0) 0 (0)
Other (5) 0 (0) 60 (3) 0 (0) 33 (1)
Total (451) 9 (39) 72 (323) 20 (92) 21 (71)
  • Abbreviations: PD-L1, programmed death-ligand 1; PNET, primitive neuroectodermal tumor; TAICs, tumor-associated immune cells; WHO, World Health Organization.

TAICs

TAICs were identified using standard H&E microscopy in the majority of samples (333 of 451; 74%). Twenty percent (92 of 451) of all samples contained macrophages, and 72% (323 of 451 samples) contained lymphocytes. Of the TAIC-containing samples, the majority contained lymphocytes only (242 of 333 samples; 72%), whereas 24% (81 of 333 samples) contained lymphocytes and macrophages, and only 3% (11 of 333 samples) contained macrophages in the absence of lymphocytes.

The majority of PD-L1–positive samples were infiltrated by TAICs (34 of 39 samples; 87%). Lymphocytes were present in similar percentages of both PD-L1–positive (27 of 39 samples; 69%) and negative (296 of 412 samples; 72%) samples. However, PD-L1–positive samples were significantly more likely to contain macrophages (20 of 39 samples; 51%) than PD-L1–negative samples (72 of 412 samples; 17%; P < .001; Fisher exact test). This is represented in Figure 2a.

Details are in the caption following the image

(a) Tumor-associated immune cells are compared between programmed death-ligand 1 (PD-L1)-positive and PD-L1 negative tumors according to the rates of macrophage and lymphocyte infiltration. PD-L1–positive tumors were significantly more likely to contain macrophages than PD-L1–negative tumors. (b) The percentage of samples that contained PD-L1–positive and PD-L1–negative macrophages and lymphocytes indicate that most infiltrating macrophages were PD-L1 positive, whereas infiltrating lymphocytes were largely PD-L1 negative.

TAICs also stained positive for PD-L1 (71 of 333 samples; 21% of those with TAIC). When macrophages were present in samples, they were mostly positive for PD-L1 (60 of 92 samples; 65%), whereas lymphocytes rarely stained positive for PD-L1 (22 of 323 samples; 7%; P < .001; Fisher exact test). This is represented in Figure 2b. Taken together, PD-L1 was expressed on tumor cells and/or immune cells in 20% of samples (89 of 451 of all samples).

All osteosarcoma samples (20 of 20; 100%) were infiltrated by lymphocytes. Similarly, the majority of Ewing sarcoma (92%), rhabdomyosarcoma (89%), neuroblastoma (83%), glioblastoma (71%), medulloblastoma (63%), and ependymoma (62%) samples contained lymphocytes. In addition, a high percentage of Burkitt lymphoma (100%), osteosarcoma (70%), Ewing sarcoma (44%), and rhabdomyosarcoma (72%) samples contained macrophages.

Survival Data for Neuroblastoma Samples

Survival data were available for 94% (104 of 111 tumors) of the pretreatment neuroblastoma samples but not for other histologies. The other 7 neuroblastoma samples were obtained after systemic treatment and were not included in the analysis. Kaplan-Meier survival curves differed significantly by PD-L1 expression on the tumor cells and/or TAICs, with poorer survival for PD-L1–positive patients (P = .004; log-rank test) (Fig. 3a). A trend toward statistical significance was maintained when evaluating samples that expressed PD-L1 only on tumor cells (P = .06; log-rank test). No significant differences in survival were observed between samples that did and did not contain TAICs (data not shown). International Neuroblastoma Staging System staging and risk group classification were available for 98 of the 104 samples in the survival analysis. There was a significant decrease in survival probability for patients with stage IV disease (N = 29) whose tumors expressed PD-L1 (P = .04; log-rank test) (Fig. 3b) and for high-risk patients (N = 30) whose tumors expressed PD-L1 (P = .05; log-rank test) (Fig. 3c). No survival differences by PD-L1 expression were observed for patients with stage I, II, or III disease or for those with intermediate-risk and low-risk disease, although the analysis was underpowered because there were only 2 deaths among these patients. International Neuroblastoma Staging System stage, risk group classification, and PD-L1 expression in samples from the data set are provided in Figure 3d,e. No statistically significant differences were observed in the frequency of PD-L1 expression between any stages or any risk groups. There was no relation in chi-square analysis between V-Myc avian myelocytomatosis viral oncogene neuroblastoma-derived homolog (MYCN) amplification (N = 10 positive, N = 74 negative) and PD-L1 expression.

Details are in the caption following the image

Overall survival is illustrated in patients with neuroblastoma according to programmed death-ligand 1 (PD-L1) expression. Kaplan-Meir curves were generated for patients who had tumor cells or tumor-associated immune cells (TAICs) that expressed PD-L1. (a) Survival was significantly better for patients who were negative for PD-L1. This difference was maintained for both (b) patients with International Neuroblastoma Staging System (INSS) stage 4 neuroblastoma and (c) patients classified as high risk. Summaries of PD-L1 expression are also provided for all neuroblastoma samples divided by (d) INSS stage and (e) risk group.

DISCUSSION

Here, we report the first broad screen of primary pediatric tumor samples for expression of PD-L1 and the presence of TAICs. Although multiple studies have investigated at PD-L1 expression in adult tumors,32-35 only limited, small studies of PD-L1 expression have been conducted in pediatric cancer.36-39 We identified PD-L1 in the majority of Burkitt lymphoma samples (80%) and in lower but significant fractions of GBM (36%), neuroblastoma (14%), and ganglioneuroblastoma (11%) samples. PD-L1 expression in Burkitt lymphoma has not previously been reported, but this observation parallels reports of PD-L1 expression in other B-cell lymphomas.40-42 High rates of PD-L1 expression have previously been reported in adult GBM, and the level of expression has been correlated with the grade of the tumor.43, 44 The results presented here raise the prospect of a similar immunobiology in pediatric glioblastoma, despite differing oncologic profiles.45-47 PD-L1 expression in a large number of neuroblastoma biopsies has not previously been reported. One previous study reported no PD-L1 expression in a limited data set of 18 patients.38 In our data set of more than 100 samples, patients whose samples stained positive for PD-L1 had inferior survival compared with those whose samples were PD-L1–negative. To our knowledge, this is the first report of a survival detriment for PD-L1 expression in neuroblastoma and complements what has been observed with other cancers.7, 25, 26, 34, 37 Outside of these select histologies, the expression of PD-L1 in our series of pediatric tumors was low, and few sarcoma samples expressed PD-L1. This is not entirely surprising given recent reports of limited PD-L1 expression in adult sarcomas,48 and it may relate to the relatively low mutational burden of pediatric cancers and translocation-associated sarcomas.49-53 We report a rate of PD-L1 positivity in osteosarcoma (5%) that is lower than that in some previous studies, which included both adult and pediatric samples,31, 54 but it compares favorably with data obtained from TMAs in other studies.31 When considering expression on TAICs, mainly macrophages, in addition to tumor cells, 20% of all samples were PD-L1–positive. This rate was even higher in certain subgroups, such as osteosarcoma (45%), rhabdomyosarcoma (30%), and Ewing sarcoma (32%). A study by Chowdhury et al found higher rates of PD-L1 expression in a series of 115 pediatric tumors,55 but their study relied on an antibody that was not validated for the detection of PD-L1.56

Despite the modest levels of PD-L1 expression observed in this study, lymphocytic infiltration was observed in a large majority of our samples. Although current concepts hold that PD-L1 expression may serve as a marker of lymphocyte activation and production of interferon-γ within the tumor microenvironment,57 we did not observe a correlation between lymphocyte infiltration and PD-L1 expression in this data set. Rather, PD-L1 expression correlated with macrophage infiltration, suggesting the possibility that mechanisms of PD-L1 expression in pediatric tumors may be associated with a unique biology. Previous studies have reported a predominance of macrophages in pediatric tumors58, 59; whereas, in our samples, a minority of tumors demonstrated macrophage infiltration, and these macrophages often expressed PD-L1. In mouse models, pediatric tumors induced myeloid-derived suppressor cells, which can aid in immune escape,60 and pediatric sarcomas induced the expansion of such cells in patients.61 We hypothesize that these tumors can signal myeloid cells, such as macrophages, to migrate to the tumor and aid in immune escape by expressing PD-L1. Despite the high rates of lymphocytic infiltration observed in a subset of pediatric tumors, attempts at growing tumor-infiltrating lymphocytes from pediatric tumors have largely been unsuccessful.62

This is the first large study to systemically interrogate PD-L1 expression and TAICs in childhood cancer, but several limitations of the study should be noted. First, although we screened a large series of primary tumor samples, the numbers of samples available for some histologic subsets were small, and clinical correlation was not available for most disease types, precluding an assessment of the potential prognostic impact of PD-L1 for diseases other than neuroblastoma. Although the vast majority of samples were diagnostic and likely came from primary tumor sites, because they are the usual source of tissue in pediatric cancer, we could not confirm the exact source (primary vs metastatic) for individual cases. Second, given the low frequency of tumor cells that express PD-L1 and the heterogeneous distribution observed in many cancers,63 the use of TMAs likely underestimates the true rate of PD-L1 expression given the limited sample size for each tumor studied. We used a cutoff of 1% of cells expressing PD-L1 as positive given that we stained TMAs, which capture just a small section of the tumor that often does not include the margin where PD-L1–positive cells are more likely to be located.30, 31, 54 Third, emerging evidence suggests that pediatric cancer genomes evolve extensively after exposure to chemoradiotherapy and attain significantly larger mutational burdens.53 Given that PD-L1 expression may be correlated with mutational burden,64, 65 analysis of samples obtained at the time of diagnosis may underestimate PD-L1 expression at the time of recurrent disease.54

Together, our data demonstrate that significant subsets of pediatric tumors express PD-L1, either on the tumor itself or in the microenvironment. The novel finding of frequent PD-L1 expression in Burkitt lymphoma suggests that further studies are warranted to determine whether the PD-1/PD-L1 axis is a good therapeutic target as has been established with other types of lymphoma.10 We also observed that a significant fraction of GBM and neuroblastoma samples express PD-L1, raising the prospect of the utility of anti-PD1 therapies in these difficult to treat diseases. Finally, this is the first work to demonstrate differential outcomes based on PD-L1 expression in neuroblastoma. Further studies are needed to better delineate the immune cells present within the tumor microenvironment of pediatric cancers and to assess the role of anti–PD-1 therapies in the treatment of pediatric malignancies.

FUNDING SUPPORT

This research project was funded by Bristol-Myers Squibb and was supported by a Stand Up To Cancer-St. Baldrick's Pediatric Dream Team Translational Research Grant (SU2C-AACR-DT1113). Stand Up To Cancer is a program of the Entertainment Industry Foundation administered by the American Association for Cancer Research.

CONFLICT OF INTEREST DISCLOSURES

Jason S. Simon and Joseph F. Grosso were employed by Bristol-Myers Squibb at the time of the current study and received both salary and stock as compensation for their employment. Melinda S. Merchant is an employee of AstraZeneca. Imene Hezam reports grants from Bristol-Myers Squibb during the conduct of the study. Brigit Geoerger reports grants from Bristol-Myers Squibb during the conduct of the study. The remaining authors made no disclosures.

AUTHOR CONTRIBUTIONS

Robbie G. Majzner: Conceptualization, methodology, validation, formal analysis, investigation, data curation, writing–original draft, writing–review and editing, and visualization. Jason S. Simon: Conceptualization, methodology, investigation, resources, and writing–review and editing. Joseph F. Grosso: Conceptualization, methodology, investigation, resources, and writing–review and editing. Daniel Martinez: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Bruce R. Pawel: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Mariarita Santi: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Melinda S. Merchant: Conceptualization and writing–review and editing. Birgit Georger: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Imene Hezam: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Virginie Marty: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Philippe Vielh: Methodology, validation, investigation, resources, data curation, and writing–review and editing. Mads Daugaard: Investigation, resources, data curation, and writing–review and editing. Poul H. Sorensen: Conceptualization, methodology, validation, investigation, resources, data curation, and writing–review and editing. Crystal L. Mackall: Conceptualization, methodology, validation, formal analysis, investigation, resources, data curation, writing–original draft, writing–review and editing, visualization, supervision, project administration, and funding acquisition. John M. Maris: Conceptualization, methodology, validation, formal analysis, investigation, resources, data curation, writing–original draft, writing–review and editing, visualization, supervision, project administration, and funding acquisition.