TGF-ß neutralization attenuates tumor residency of activated T cells to enhance systemic immunity in mice.

A tissue resident-like phenotype in tumor infiltrating T cells can limit systemic anti-tumor immunity. Enhanced systemic anti-tumor immunity is observed in head and neck cancer patients after neoadjuvant PD-L1 immune checkpoint blockade (ICB) and transforming growth factor ß (TGF-ß) neutralization. Using T cell receptor (TCR) sequencing and functional immunity assays in a syngeneic model of oral cancer, we dissect the relative contribution of these treatments to enhanced systemic immunity. The addition of TGF-ß neutralization to ICB resulted in the egress of expanded and exhausted CD8+ tumor infiltrating lymphocytes (TILs) into circulation and greater systemic anti-tumor immunity. This enhanced egress associated with reduced expression of Itgae (CD103) and its upstream regulator Znf683. Circulating CD8+ T cells expressed higher Cxcr3 after treatment, an observation also made in samples from patients treated with dual TGF-ß neutralization and ICB. These findings provide the scientific rationale for the use of PD-L1 ICB and TGF-ß neutralization in newly diagnosed patients with carcinomas prior to definitive treatment of locoregional disease.

Efficacy and safety of bintrafusp alfa in 2 phase I expansion cohorts with advanced HCC.

BACKGROUND AND AIMS: Simultaneous inhibition of the TGF-ß and programmed cell death 1 ligand 1 pathways provides a potential novel treatment approach. Bintrafusp alfa, a first-in-class bifunctional fusion protein composed of the extracellular domain of TGF-ßRII (a TGF-ß "trap") fused to a human IgG1 monoclonal antibody blocking programmed cell death 1 ligand 1, was evaluated in patients with advanced HCC. APPROACH AND RESULTS: In this global, open-label, phase I study (NCT02517398), patients with programmed cell death 1 ligand 1-unselected HCC who failed or were intolerant to ≥1 line of sorafenib received bintrafusp alfa 1200 mg every 2 weeks in a dose-escalation (n = 38) or dose-expansion (n = 68) cohort until confirmed progression, unacceptable toxicity, or trial withdrawal. The primary endpoint was the best overall response per Response Evaluation Criteria in Solid Tumors version 1.1 by an independent review committee. Secondary endpoints included investigator-assessed best overall response, safety, and pharmacokinetics. Median follow-up times (range) were 41.4 (39.8-44.2) and 38.6 (33.5-39.7) months in the dose-escalation and dose-expansion cohorts, respectively. The objective response rate was below the prespecified 20% objective response rate threshold set to evaluate the efficacy of bintrafusp alfa in both cohorts (10.5% and 8.8%, respectively). Median overall survival and progression-free survival, respectively, were 13.8 and 1.5 months in the dose-escalation cohort and 13.5 and 1.4 months in the dose-expansion cohort. Treatment-related adverse events occurred in 78.9% and 64.7% of patients in the respective cohorts (grade ≥3 in 18.4% and 25.0% of patients). CONCLUSIONS: Bintrafusp alfa showed moderate clinical activity and a safety profile consistent with previous studies of bintrafusp alfa in patients with advanced HCC.

Final Results From a Phase I Trial and Expansion Cohorts of Cabozantinib and Nivolumab Alone or With Ipilimumab for Advanced/Metastatic Genitourinary Tumors.

PURPOSE: Cabozantinib and nivolumab (CaboNivo) alone or with ipilimumab (CaboNivoIpi) have shown promising efficacy and safety in patients with metastatic urothelial carcinoma (mUC), metastatic renal cell carcinoma (mRCC), and rare genitourinary (GU) tumors in a dose-escalation phase I study. We report the final data analysis of the safety, overall response rate (ORR), progression-free survival (PFS), and overall survival (OS) of the phase I patients and seven expansion cohorts. METHODS: This is an investigator-initiated, multicenter, phase I trial. CaboNivo doublet expansion cohorts included (1) mUC, (2) mRCC, and (3) adenocarcinoma of the bladder/urachal; CaboNivoIpi triplet expansion cohorts included (1) mUC, (2) mRCC, (3) penile cancer, and (4) squamous cell carcinoma of the bladder and other rare GU tumors (ClinicalTrials.gov identifier: NCT02496208). RESULTS: The study enrolled 120 patients treated with CaboNivo (n = 64) or CaboNivoIpi (n = 56), with a median follow-up of 49.2 months. In 108 evaluable patients (CaboNivo n = 59; CaboNivoIpi n = 49), the ORR was 38% (complete response rate 11%) and the median duration of response was 20 months. The ORR was 42.4% for mUC, 62.5% for mRCC (n = 16), 85.7% for squamous cell carcinoma of the bladder (n = 7), 44.4% for penile cancer (n = 9), and 50.0% for renal medullary carcinoma (n = 2). Grade ≥ 3 treatment-related adverse events occurred in 84% of CaboNivo patients and 80% of CaboNivoIpi patients. CONCLUSION: CaboNivo and CaboNivoIpi demonstrated clinical activity and safety in patients with multiple GU malignancies, especially clear cell RCC, urothelial carcinoma, and rare GU tumors such as squamous cell carcinoma of the bladder, small cell carcinoma of the bladder, adenocarcinoma of the bladder, renal medullary carcinoma, and penile cancer.

Hepatic artery infusion pump (HAIP) therapy in combination with targeted delivery of IL-12 for patients with metastatic colorectal cancer or intrahepatic cholangiocarcinoma: a phase II trial protocol.

Background: Treatment of advanced liver tumors remains challenging. Although immune checkpoint inhibition has revolutionized treatment for many cancers, responses in colorectal liver metastases and biliary tract cancers remain suboptimal. Investigation into additional immunomodulatory therapies for these cancers is needed. Interleukin-12 (IL-12) is a pro-inflammatory cytokine with robust anti-tumor activity, but systemic adverse effects largely terminated therapeutic development of recombinant human IL-12 (rhIL-12). PDS01ADC is a novel human monoclonal antibody (NHS76) conjugated to two IL-12 heterodimers with established safety in phase I trials. The NHS76 antibody specifically targets histone/DNA complexes which are accessible only in regions of cell death and this antibody has been shown to accumulate locally in tumors. Methods: Patients with unresectable metastatic colorectal cancer (mCRC) or unresectable intrahepatic cholangiocarcinoma (ICC) will receive synchronization of subcutaneous PDS01ADC with floxuridine delivered via a hepatic artery infusion pump (HAIP). The primary outcome measured in this study will be overall response rate as measured by Response Evaluation Criteria in Solid Tumors (RECIST) criteria. Secondary outcomes measured in this study will include hepatic and non-hepatic progression-free survival (PFS), overall survival, and safety of PDS01ADC combination therapy with HAIP. Discussion: Poor clinical response of these liver tumors to immunotherapy is likely due to various factors, including poor immune infiltrate into the tumor and immunosuppression by the tumor microenvironment. By exploiting the tumor cell death induced by HAIP locoregional therapy in combination with systemic chemotherapy, PDS01ADC is poised to modulate the tumor immune microenvironment to improve outcomes for patients undergoing HAIP therapy. Trial Registration: ClinicalTrials.gov (ID NCT05286814 version 2023-10-18); https://clinicaltrials.gov/study/NCT05286814?term=NCT05286814&rank=1.

Hypothesis: the generation of T cells directed against neoepitopes employing immune-mediating agents other than neoepitope vaccines.

The development of vaccines, especially RNA-based, directed against patient-specific tumor neoepitopes is an active and productive area of cancer immunotherapy. Promising clinical results in melanoma and other solid tumor types are emerging. As with all cancer therapy modalities, neoepitope vaccine development and delivery also has some drawbacks, including the level of effort to develop a patient-specific product, accuracy of algorithms to predict neoepitopes, and with the exception of melanoma and some other tumor types, biopsies of metastatic lesions of solid tumors are often not available. We hypothesize that in some circumstances the use of rationally designed combinations of "off-the-shelf" agents may prove an additional path to enable the patient to produce his/her own "neoepitope vaccine" in situ. These combination therapies may consist of agents to activate a tumor-associated T-cell response, potentiate that response, reduce or eliminate immunosuppressive entities in the tumor microenvironment, and/or alter the phenotype of tumor cells to render them more susceptible to immune-mediated lysis. Examples are provided in both preclinical and clinical studies in which combinations of "off-the-shelf" agents lead to the generation of T cells directed against tumor-derived neoepitopes with consequent antitumor activity.

A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics.

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT-DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT-DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

Challenges and opportunities in cancer immunotherapy: a Society for Immunotherapy of Cancer (SITC) strategic vision.

Cancer immunotherapy has flourished over the last 10-15 years, transforming the practice of oncology and providing long-term clinical benefit to some patients. During this time, three distinct classes of immune checkpoint inhibitors, chimeric antigen receptor-T cell therapies specific for two targets, and two distinct classes of bispecific T cell engagers, a vaccine, and an oncolytic virus have joined cytokines as a standard of cancer care. At the same time, scientific progress has delivered vast amounts of new knowledge. For example, advances in technologies such as single-cell sequencing and spatial transcriptomics have provided deep insights into the immunobiology of the tumor microenvironment. With this rapid clinical and scientific progress, the field of cancer immunotherapy is currently at a critical inflection point, with potential for exponential growth over the next decade. Recognizing this, the Society for Immunotherapy of Cancer convened a diverse group of experts in cancer immunotherapy representing academia, the pharmaceutical and biotechnology industries, patient advocacy, and the regulatory community to identify current opportunities and challenges with the goal of prioritizing areas with the highest potential for clinical impact. The consensus group identified seven high-priority areas of current opportunity for the field: mechanisms of antitumor activity and toxicity; mechanisms of drug resistance; biomarkers and biospecimens; unique aspects of novel therapeutics; host and environmental interactions; premalignant immunity, immune interception, and immunoprevention; and clinical trial design, endpoints, and conduct. Additionally, potential roadblocks to progress were discussed, and several topics were identified as cross-cutting tools for optimization, each with potential to impact multiple scientific priority areas. These cross-cutting tools include preclinical models, data curation and sharing, biopsies and biospecimens, diversification of funding sources, definitions and standards, and patient engagement. Finally, three key guiding principles were identified that will both optimize and maximize progress in the field. These include engaging the patient community; cultivating diversity, equity, inclusion, and accessibility; and leveraging the power of artificial intelligence to accelerate progress. Here, we present the outcomes of these discussions as a strategic vision to galvanize the field for the next decade of exponential progress in cancer immunotherapy.

Chordoma cancer stem cell subpopulation characterization may guide targeted immunotherapy approaches to reduce disease recurrence.

Introduction: Cancer stem cells (CSCs), a group of tumor-initiating and tumor-maintaining cells, may be major players in the treatment resistance and recurrence distinctive of chordoma. Characterizing CSCs is crucial to better targeting this subpopulation. Methods: Using flow cytometry, six chordoma cell lines were evaluated for CSC composition. In vitro, cell lines were stained for B7H6, HER2, MICA-B, ULBP1, EGFR, and PD-L1 surface markers. Eighteen resected chordomas were stained using a multispectral immunofluorescence (mIF) antibody panel to identify CSCs in vivo. HALO software was used for quantitative CSC density and spatial analysis. Results: In vitro, chordoma CSCs express more B7H6, MICA-B, and ULBP1, assessed by percent positivity and mean fluorescence intensity (MFI), as compared to non-CSCs in all cell lines. PD- L1 percent positivity is increased by >20% in CSCs compared to non-CSCs in all cell lines except CH22. In vivo, CSCs comprise 1.39% of chordoma cells and most are PD-L1+ (75.18%). A spatial analysis suggests that chordoma CSCs cluster at an average distance of 71.51 mm (SD 73.40 mm) from stroma. Discussion: To our knowledge, this study is the first to identify individual chordoma CSCs and describe their surface phenotypes using in vitro and in vivo methods. PD-L1 is overexpressed on CSCs in chordoma human cell lines and operative tumor samples. Similarly, potential immunotherapeutic targets on CSCs, including B7H6, MICA-B, ULBP1, EGFR, and HER2 are overexpressed across cell lines. Targeting these markers may have a preferential role in combating CSCs, an aggressive subpopulation likely consequential to chordoma's high recurrence rate.

Deep learning-based whole-body PSMA PET/CT attenuation correction utilizing Pix-2-Pix GAN.

PURPOSE: Sequential PET/CT studies oncology patients can undergo during their treatment follow-up course is limited by radiation dosage. We propose an artificial intelligence (AI) tool to produce attenuation-corrected PET (AC-PET) images from non-attenuation-corrected PET (NAC-PET) images to reduce need for low-dose CT scans. METHODS: A deep learning algorithm based on 2D Pix-2-Pix generative adversarial network (GAN) architecture was developed from paired AC-PET and NAC-PET images. 18F-DCFPyL PSMA PET-CT studies from 302 prostate cancer patients, split into training, validation, and testing cohorts (n = 183, 60, 59, respectively). Models were trained with two normalization strategies: Standard Uptake Value (SUV)-based and SUV-Nyul-based. Scan-level performance was evaluated by normalized mean square error (NMSE), mean absolute error (MAE), structural similarity index (SSIM), and peak signal-to-noise ratio (PSNR). Lesion-level analysis was performed in regions-of-interest prospectively from nuclear medicine physicians. SUV metrics were evaluated using intraclass correlation coefficient (ICC), repeatability coefficient (RC), and linear mixed-effects modeling. RESULTS: Median NMSE, MAE, SSIM, and PSNR were 13.26%, 3.59%, 0.891, and 26.82, respectively, in the independent test cohort. ICC for SUVmax and SUVmean were 0.88 and 0.89, which indicated a high correlation between original and AI-generated quantitative imaging markers. Lesion location, density (Hounsfield units), and lesion uptake were all shown to impact relative error in generated SUV metrics (all p < 0.05). CONCLUSION: The Pix-2-Pix GAN model for generating AC-PET demonstrates SUV metrics that highly correlate with original images. AI-generated PET images show clinical potential for reducing the need for CT scans for attenuation correction while preserving quantitative markers and image quality.

Utility of Large Language Models for Health Care Professionals and Patients in Navigating Hematopoietic Stem Cell Transplantation: Comparison of the Performance of ChatGPT-3.5, ChatGPT-4, and Bard.

BACKGROUND: Artificial intelligence is increasingly being applied to many workflows. Large language models (LLMs) are publicly accessible platforms trained to understand, interact with, and produce human-readable text; their ability to deliver relevant and reliable information is also of particular interest for the health care providers and the patients. Hematopoietic stem cell transplantation (HSCT) is a complex medical field requiring extensive knowledge, background, and training to practice successfully and can be challenging for the nonspecialist audience to comprehend. OBJECTIVE: We aimed to test the applicability of 3 prominent LLMs, namely ChatGPT-3.5 (OpenAI), ChatGPT-4 (OpenAI), and Bard (Google AI), in guiding nonspecialist health care professionals and advising patients seeking information regarding HSCT. METHODS: We submitted 72 open-ended HSCT-related questions of variable difficulty to the LLMs and rated their responses based on consistency-defined as replicability of the response-response veracity, language comprehensibility, specificity to the topic, and the presence of hallucinations. We then rechallenged the 2 best performing chatbots by resubmitting the most difficult questions and prompting to respond as if communicating with either a health care professional or a patient and to provide verifiable sources of information. Responses were then rerated with the additional criterion of language appropriateness, defined as language adaptation for the intended audience. RESULTS: ChatGPT-4 outperformed both ChatGPT-3.5 and Bard in terms of response consistency (66/72, 92%; 54/72, 75%; and 63/69, 91%, respectively; P=.007), response veracity (58/66, 88%; 40/54, 74%; and 16/63, 25%, respectively; P<.001), and specificity to the topic (60/66, 91%; 43/54, 80%; and 27/63, 43%, respectively; P<.001). Both ChatGPT-4 and ChatGPT-3.5 outperformed Bard in terms of language comprehensibility (64/66, 97%; 53/54, 98%; and 52/63, 83%, respectively; P=.002). All displayed episodes of hallucinations. ChatGPT-3.5 and ChatGPT-4 were then rechallenged with a prompt to adapt their language to the audience and to provide source of information, and responses were rated. ChatGPT-3.5 showed better ability to adapt its language to nonmedical audience than ChatGPT-4 (17/21, 81% and 10/22, 46%, respectively; P=.03); however, both failed to consistently provide correct and up-to-date information resources, reporting either out-of-date materials, incorrect URLs, or unfocused references, making their output not verifiable by the reader. CONCLUSIONS: In conclusion, despite LLMs' potential capability in confronting challenging medical topics such as HSCT, the presence of mistakes and lack of clear references make them not yet appropriate for routine, unsupervised clinical use, or patient counseling. Implementation of LLMs' ability to access and to reference current and updated websites and research papers, as well as development of LLMs trained in specialized domain knowledge data sets, may offer potential solutions for their future clinical application.

Augmentation of tumor expression of HLA-DR, CXCL9, and CXCL10 may improve olfactory neuroblastoma immunotherapeutic responses.

BACKGROUND: Olfactory neuroblastoma is a rare malignancy of the anterior skull base typically treated with surgery and adjuvant radiation. Although outcomes are fair for low-grade disease, patients with high-grade, recurrent, or metastatic disease oftentimes respond poorly to standard treatment methods. We hypothesized that an in-depth evaluation of the olfactory neuroblastoma tumor immune microenvironment would identify mechanisms of immune evasion in high-grade olfactory neuroblastoma as well as rational targetable mechanisms for future translational immunotherapeutic approaches. METHODS: Multispectral immunofluorescence and RNAScope evaluation of the tumor immune microenvironment was performed on forty-seven clinically annotated olfactory neuroblastoma samples. A retrospective chart review was performed and clinical correlations assessed. RESULTS: A significant T cell infiltration was noted in olfactory neuroblastoma samples with a stromal predilection, presence of myeloid-derived suppressor cells, and sparse natural killer cells. A striking decrease was observed in MHC-I expression in high-grade olfactory neuroblastoma compared to low-grade disease, representing a mechanism of immune evasion in high-grade disease. Mechanistically, the immune effector stromal predilection appears driven by low tumor cell MHC class II (HLA-DR), CXCL9, and CXCL10 expression as those tumors with increased tumor cell expression of each of these mediators correlated with significant increases in T cell infiltration. CONCLUSION: These data suggest that immunotherapeutic strategies that augment tumor cell expression of MHC class II, CXCL9, and CXCL10 may improve parenchymal trafficking of immune effector cells in olfactory neuroblastoma and augment immunotherapeutic responses.

Efficacy, safety, and biomarker analyses of bintrafusp alfa, a bifunctional fusion protein targeting TGF-ß and PD-L1, in patients with advanced non-small cell lung cancer.

BACKGROUND: Bintrafusp alfa, a first-in-class bifunctional fusion protein targeting transforming growth factor-ß (TGF-ß) and programmed cell death ligand 1, has demonstrated encouraging efficacy as second-line treatment in patients with non-small cell lung cancer (NSCLC) in a dose expansion cohort of the phase 1, open-label clinical trial (NCT02517398). Here, we report the safety, efficacy, and biomarker analysis of bintrafusp alfa in a second expansion cohort of the same trial (biomarker cohort). METHODS: Patients with stage IIIb/IV NSCLC who were either immune checkpoint inhibitor (ICI)-naïve (n=18) or ICI-experienced (n=23) were enrolled. The primary endpoint was the best overall response. Paired biopsies (n=9/41) and peripheral blood (n=14/41) pretreatment and on-treatment were studied to determine the immunological effects of treatment and for associations with clinical activity. RESULTS: Per independent review committee assessment, objective responses were observed in the ICI-naïve group (overall response rate, 27.8%). No new or unexpected safety signals were identified. Circulating TGF-ß levels were reduced (>97%; p<0.001) 2 weeks after initiation of treatment with bintrafusp alfa and remained reduced up to 12 weeks. Increases in lymphocytes and tumor-associated macrophages (TAMs) were observed in on-treatment biospies, with an increase in the M2 (tumor trophic TAMs)/M1 (inflammatory TAMs) ratio associated with poor outcomes. Specific peripheral immune analytes at baseline and early changes after treatment were associated with clinical response. CONCLUSIONS: Bintrafusp alfa was observed to have modest clinical activity and manageable safety, and was associated with notable immunologic changes involving modulation of the tumor immune microenvironment in patients with advanced NSCLC.

Increasing diversity in clinical trials: demographic trends at the National Cancer Institute, 2005-2020.

BACKGROUND: We described participant demographics for National Cancer Institute (NCI) clinical trials at the clinical center (NCI-CC participants) of the National Institutes of Health to identify enrollment disparities. METHODS: We analyzed NCI-CC data from 2005 to 2020, calculated enrollment fractions, compared with the US cancer population represented by the Surveillance, Epidemiology, and End Results cancer incidence data (2018) and the Cancer in North America database (2018), and compared further with clinical trial disparities data from the NCI Community Oncology Research Program and National Clinical Trials Network (2005-2019), and from ClinicalTrials.gov (2003-2016). RESULTS: NCI-CC (38 531 participants) had higher enrollment fractions for older adults (8.5%), male (5.6%), non-Hispanic (5.1%), and Black or African American (5.3%) participants; lower women proportion across race and ethnicity; and fewer female sex-specific cancer (6.8%) than male sex-specific cancer (11.7%) participants. NCI-CC had lower median age than Surveillance, Epidemiology, and End Results (54.0 vs 65.4); more Black or African American participants (12.0% vs 11.1%); and fewer women (41.7% vs 49.5%), White (76.1% vs 80.5%), Asian or Pacific Islander (4.6% vs 6.0%), American Indian or Alaska Native (0.3% vs 0.5%), and Hispanic participants (7.1% vs 13%). NCI-CC had more Black or African American and Asian or Pacific Islander participants; fewer Hispanic participants than the NCI Community Oncology Research Program and National Clinical Trials Network; more Black or African American and Hispanic participants; fewer Asian or Pacific Islander participants than ClinicalTrials.gov data. Improvement was noted for NCI-CC (older adults, Black or African American, Asian or Pacific Islander, Hispanic participants). CONCLUSION: We found lower representation of older adults, women, Asian or Pacific Islander, American Indian or Alaska Native, and Hispanic participants vs the US cancer population and higher representation of Black or African American vs US cancer population and oncology clinical trials. Multifaceted efforts are underway to reduce disparities in cancer clinical trials at the NCI-CC.

Comparison of Large Language Models in Answering Immuno-Oncology Questions: A Cross-Sectional Study.

BACKGROUND: The capability of large language models (LLMs) to understand and generate human-readable text has prompted the investigation of their potential as educational and management tools for patients with cancer and healthcare providers. MATERIALS AND METHODS: We conducted a cross-sectional study aimed at evaluating the ability of ChatGPT-4, ChatGPT-3.5, and Google Bard to answer questions related to 4 domains of immuno-oncology (Mechanisms, Indications, Toxicities, and Prognosis). We generated 60 open-ended questions (15 for each section). Questions were manually submitted to LLMs, and responses were collected on June 30, 2023. Two reviewers evaluated the answers independently. RESULTS: ChatGPT-4 and ChatGPT-3.5 answered all questions, whereas Google Bard answered only 53.3% (P < .0001). The number of questions with reproducible answers was higher for ChatGPT-4 (95%) and ChatGPT3.5 (88.3%) than for Google Bard (50%) (P < .0001). In terms of accuracy, the number of answers deemed fully correct were 75.4%, 58.5%, and 43.8% for ChatGPT-4, ChatGPT-3.5, and Google Bard, respectively (P = .03). Furthermore, the number of responses deemed highly relevant was 71.9%, 77.4%, and 43.8% for ChatGPT-4, ChatGPT-3.5, and Google Bard, respectively (P = .04). Regarding readability, the number of highly readable was higher for ChatGPT-4 and ChatGPT-3.5 (98.1%) and (100%) compared to Google Bard (87.5%) (P = .02). CONCLUSION: ChatGPT-4 and ChatGPT-3.5 are potentially powerful tools in immuno-oncology, whereas Google Bard demonstrated relatively poorer performance. However, the risk of inaccuracy or incompleteness in the responses was evident in all 3 LLMs, highlighting the importance of expert-driven verification of the outputs returned by these technologies.

Deep Learning-Based Detection and Classification of Bone Lesions on Staging Computed Tomography in Prostate Cancer: A Development Study.

RATIONALE AND OBJECTIVES: Efficiently detecting and characterizing metastatic bone lesions on staging CT is crucial for prostate cancer (PCa) care. However, it demands significant expert time and additional imaging such as PET/CT. We aimed to develop an ensemble of two automated deep learning AI models for 1) bone lesion detection and segmentation and 2) benign vs. metastatic lesion classification on staging CTs and to compare its performance with radiologists. MATERIALS AND METHODS: This retrospective study developed two AI models using 297 staging CT scans (81 metastatic) with 4601 benign and 1911 metastatic lesions in PCa patients. Metastases were validated by follow-up scans, bone biopsy, or PET/CT. Segmentation AI (3DAISeg) was developed using the lesion contours delineated by a radiologist. 3DAISeg performance was evaluated with the Dice similarity coefficient, and classification AI (3DAIClass) performance on AI and radiologist contours was assessed with F1-score and accuracy. Training/validation/testing data partitions of 70:15:15 were used. A multi-reader study was performed with two junior and two senior radiologists within a subset of the testing dataset (n = 36). RESULTS: In 45 unseen staging CT scans (12 metastatic PCa) with 669 benign and 364 metastatic lesions, 3DAISeg detected 73.1% of metastatic (266/364) and 72.4% of benign lesions (484/669). Each scan averaged 12 extra segmentations (range: 1-31). All metastatic scans had at least one detected metastatic lesion, achieving a 100% patient-level detection. The mean Dice score for 3DAISeg was 0.53 (median: 0.59, range: 0-0.87). The F1 for 3DAIClass was 94.8% (radiologist contours) and 92.4% (3DAISeg contours), with a median false positive of 0 (range: 0-3). Using radiologist contours, 3DAIClass had PPV and NPV rates comparable to junior and senior radiologists: PPV (semi-automated approach AI 40.0% vs. Juniors 32.0% vs. Seniors 50.0%) and NPV (AI 96.2% vs. Juniors 95.7% vs. Seniors 91.9%). When using 3DAISeg, 3DAIClass mimicked junior radiologists in PPV (pure-AI 20.0% vs. Juniors 32.0% vs. Seniors 50.0%) but surpassed seniors in NPV (pure-AI 93.8% vs. Juniors 95.7% vs. Seniors 91.9%). CONCLUSION: Our lesion detection and classification AI model performs on par with junior and senior radiologists in discerning benign and metastatic lesions on staging CTs obtained for PCa.

Cooperative Armoring of CAR and TCR T Cells by T Cell-Restricted IL15 and IL21 Universally Enhances Solid Tumor Efficacy.

PURPOSE: Chimeric antigen receptor (CAR) and T-cell receptor (TCR) T-cell therapies are effective in a subset of patients with solid tumors, but new approaches are needed to universally improve patient outcomes. Here, we developed a technology to leverage the cooperative effects of IL15 and IL21, two common cytokine-receptor gamma chain family members with distinct, pleiotropic effects on T cells and other lymphocytes, to enhance the efficacy of adoptive T cells. EXPERIMENTAL DESIGN: We designed vectors that induce the constitutive expression of either membrane-tethered IL15, IL21, or IL15/IL21. We used clinically relevant preclinical models of transgenic CARs and TCRs against pediatric and adult solid tumors to determine the effect of the membrane-tethered cytokines on engineered T cells for human administration. RESULTS: We found that self-delivery of these cytokines by CAR or TCR T cells prevents functional exhaustion by repeated stimulation and limits the emergence of dysfunctional natural killer (NK)-like T cells. Across different preclinical murine solid tumor models, we observed enhanced regression with each individual cytokine but the greatest antitumor efficacy when T cells were armored with both. CONCLUSIONS: The coexpression of membrane-tethered IL15 and IL21 represents a technology to enhance the resilience and function of engineered T cells against solid tumors and could be applicable to multiple therapy platforms and diseases. See related commentary by Ruffin et al., p. 1431.

Comparison of Large Language Models in Answering Immuno-Oncology Questions: A Cross-Sectional Study.

Background: The capability of large language models (LLMs) to understand and generate human-readable text has prompted the investigation of their potential as educational and management tools for cancer patients and healthcare providers. Materials and Methods: We conducted a cross-sectional study aimed at evaluating the ability of ChatGPT-4, ChatGPT-3.5, and Google Bard to answer questions related to four domains of immuno-oncology (Mechanisms, Indications, Toxicities, and Prognosis). We generated 60 open-ended questions (15 for each section). Questions were manually submitted to LLMs, and responses were collected on June 30th, 2023. Two reviewers evaluated the answers independently. Results: ChatGPT-4 and ChatGPT-3.5 answered all questions, whereas Google Bard answered only 53.3% (p <0.0001). The number of questions with reproducible answers was higher for ChatGPT-4 (95%) and ChatGPT3.5 (88.3%) than for Google Bard (50%) (p <0.0001). In terms of accuracy, the number of answers deemed fully correct were 75.4%, 58.5%, and 43.8% for ChatGPT-4, ChatGPT-3.5, and Google Bard, respectively (p = 0.03). Furthermore, the number of responses deemed highly relevant was 71.9%, 77.4%, and 43.8% for ChatGPT-4, ChatGPT-3.5, and Google Bard, respectively (p = 0.04). Regarding readability, the number of highly readable was higher for ChatGPT-4 and ChatGPT-3.5 (98.1%) and (100%) compared to Google Bard (87.5%) (p = 0.02). Conclusion: ChatGPT-4 and ChatGPT-3.5 are potentially powerful tools in immuno-oncology, whereas Google Bard demonstrated relatively poorer performance. However, the risk of inaccuracy or incompleteness in the responses was evident in all three LLMs, highlighting the importance of expert-driven verification of the outputs returned by these technologies.

A T cell receptor ß chain-directed antibody fusion molecule activates and expands subsets of T cells to promote antitumor activity.

Despite the success of programmed cell death-1 (PD-1) and PD-1 ligand (PD-L1) inhibitors in treating solid tumors, only a proportion of patients respond. Here, we describe a first-in-class bifunctional therapeutic molecule, STAR0602, that comprises an antibody targeting germline Vß6 and Vß10 T cell receptors (TCRs) fused to human interleukin-2 (IL-2) and simultaneously engages a nonclonal mode of TCR activation with costimulation to promote activation and expansion of αß T cell subsets expressing distinct variable ß (Vß) TCR chains. In solution, STAR0602 binds IL-2 receptors in cis with Vß6/Vß10 TCRs on the same T cell, promoting expansion of human Vß6 and Vß10 CD4+ and CD8+ T cells that acquire an atypical central memory phenotype. Monotherapy with a mouse surrogate molecule induced durable tumor regression across six murine solid tumor models, including several refractory to anti-PD-1. Analysis of murine tumor-infiltrating lymphocyte (TIL) transcriptomes revealed that expanded Vß T cells acquired a distinct effector memory phenotype with suppression of genes associated with T cell exhaustion and TCR signaling repression. Sequencing of TIL TCRs also revealed an increased T cell repertoire diversity within targeted Vß T cell subsets, suggesting clonal revival of tumor T cell responses. These immunological and antitumor effects in mice were recapitulated in studies of STAR0602 in nonhuman primates and human ex vivo models, wherein STAR0602 boosted human antigen-specific T cell responses and killing of tumor organoids. Thus, STAR0602 represents a distinct class of T cell-activating molecules with the potential to deliver enhanced antitumor activity in checkpoint inhibitor-refractory settings.

Prediction of cancer treatment response from histopathology images through imputed transcriptomics.

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin (H&E)-stained tumor slides for precision oncology. We present ENLIGHT-DeepPT, an approach for predicting response to multiple targeted and immunotherapies from H&E-slides. In difference from existing approaches that aim to predict treatment response directly from the slides, ENLIGHT-DeepPT is an indirect two-step approach consisting of (1) DeepPT, a new deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response based on the DeepPT inferred expression values. DeepPT successfully predicts transcriptomics in all 16 TCGA cohorts tested and generalizes well to two independent datasets. Our key contribution is showing that ENLIGHT-DeepPT successfully predicts true responders in five independent patients' cohorts involving four different treatments spanning six cancer types with an overall odds ratio of 2.44, increasing the baseline response rate by 43.47% among predicted responders, without the need for any treatment data for training. Furthermore, its prediction accuracy on these datasets is comparable to a supervised approach predicting the response directly from the images, which needs to be trained and tested on the same cohort. ENLIGHT-DeepPT future application could provide clinicians with rapid treatment recommendations to an array of different therapies and importantly, may contribute to advancing precision oncology in developing countries.