Highlighted Research Articles

Deng and colleagues showed that intratumoral BO-112, a nanoplexed form of a noncoding double-stranded RNA based on poly I:C, cooperates with hypofractionated radiotherapy to engage persistent myeloid cells and promote T cell–dependent tumor control. In mouse models as well as a phase I neoadjuvant trial, treatment reprogrammed tumor-associated myeloid cells toward antigen-presenting states, promoted clonal replacement by less exhausted T cells, and enhanced sarcoma depletion. These immunologic changes coincided with systemic T-cell immunity and an encouraging degree of disease control, while occurrence of rare immune-related adverse events were mitigated by dose adjustment. These findings support myeloid-directed innate immune activation as a strategy to enhance radiotherapy and immunotherapy responses.See article, p. 1280.Early detection of hepatocellular carcinoma (HCC) remains limited by incomplete recognition of individuals at risk, particularly before cirrhosis is diagnosed. Clusmann and colleagues developed PRE-Screen-HCC, an interpretable machine learning framework that predicts future HCC risk from routinely available clinical data. Using population-scale data from UK Biobank and external validation in the All of Us Research Program, the authors demonstrated that random-forest models integrating demographics, lifestyle, electronic health records, and blood tests outperform established HCC risk scores. Additional genomic and metabolomic data provided limited incremental benefit, highlighting the value of simple clinical data for scalable HCC pre-screening and targeted surveillance.See article, p. 1304.Pancreatic intraepithelial neoplasia (PanIN) is the noninvasive precursor of pancreatic ductal adenocarcinoma (PDAC), yet the protein programs driving its progression remain poorly understood. Min, Schweizer, and colleagues applied Deep Visual Proteomics, comprising AI-guided annotation, laser microdissection, and mass spectrometry, to profile normal ducts, acinar-to-ductal metaplasia, low- and high-grade PanINs, and invasive cancer. Quantification of 9,181 proteins revealed that normal-looking ducts near tumors already carried stress, immune, and metabolic signatures, that mitochondrial remodeling intensified before invasion, and that KRAS-mutant peptides appeared in low-grade PanINs from cancer-free individuals. These findings indicate that molecular reprogramming precedes visible transformation, opening an early window for interception.See article, p. 1323.Qian, Deng, Xi, Zhang, Sun, and colleagues established a risk assessment framework across the therapeutic continuum for esophageal squamous cell carcinoma (ESCC) chemoimmunotherapy: a baseline predictor for initial response, an on-treatment predictor for long-term response based on metabolic shift patterns, and a real-time progression risk model using sphingolipid and glycerophospholipid alterations. Additionally, two dietary metabolites, garlic-derived S-allyl-L-cysteine and cruciferous vegetable-derived indole-3-carbinol, enhanced outcomes in preclinical models. This provides the first metabolomic roadmap unifying baseline prediction, longitudinal surveillance, and dietary modulation for ESCC chemoimmunotherapy.See article, p. 1341.Neoadjuvant PD-1 blockade benefits only ∼30% of patients with esophageal squamous cell carcinoma (ESCC), highlighting the urgent need for reliable predictive biomarkers. Zhu, Wang, and colleagues showed that cancer-associated fibroblast–epithelial (CAF-Epi) niches activate SOX2 in tumor cells, which transcriptionally upregulates GPNMB, driving shedding of soluble GPNMB (sGPNMB) into circulation. Tumor-derived sGPNMB suppresses CD8+ T-cell function via the SDC4–CD148 axis, promoting exhaustion and immunotherapy resistance. Integrating plasma GPNMB levels, CAF-Epi niche detection in endoscopic biopsies, and clinical staging, the authors developed and prospectively validated a multimodal model, termed CANP, that robustly predicted immunotherapy response and survival across independent ESCC cohorts.See article, p. 1360.Liu, Tan, Wu, Huang, Cheng, and colleagues identified circulating tumor cell (CTC)–secreted GPNMB as a key driver of brain metastasis. Multi-omic profiling of patient CTCs and brain metastases uncovered CBX3-dependent upregulation of GPNMB. Mechanistically, GPNMB binds endothelial EGFR and promotes CBL-mediated EGFR degradation, thereby suppressing FTO and destabilizing TJP1 through YTHDF2-dependent m6A regulation, weakening tight junctions and enhancing blood–brain barrier permeability. GPNMB also induced endothelial CXCL12 to recruit CXCR4+ immune cells and led to progressive T-cell exhaustion within the brain microenvironment. Translationally, CBX3+GPNMB+ CTCs and plasma CXCL12 abundance were significantly associated with brain metastasis progression, and combined GPNMB/PD1 blockade improved therapeutic efficacy in vivo, supporting a biomarker-guided immunotherapeutic strategy.See article, p. 1382.Metastatic breast cancer cells that colonize the lung require metabolic support from the local microenvironment. Using spatial analyses of mouse and patient samples, Liu and colleagues showed that metastatic cancer cells induce the proliferation and metabolic reprogramming of neighboring AT2 cells into lipid feeder cells through SREBP1 activation. This program enhanced the expression of key lipid synthesis genes, including FASN and GPAM in AT2 cells, creating a lipid-rich environment that promotes metastatic growth. Inhibition of FASN and GPAM suppressed lung metastasis growth, identifying AT2 cell lipid metabolism as a therapeutic vulnerability in metastatic breast cancer.See article, p. 1412.LKB1-mutant lung tumors are highly inflammatory and poorly responsive to immune checkpoint blockade. Pillai, Rashidfarrokhi, and colleagues demonstrated that LKB1 loss in lung adenocarcinoma leads to the upregulation of LIF. Tumor-derived LIF acted in an autocrine manner to drive the emergence of dedifferentiated and inflammatory cancer cell states. These inflammatory cancer cell states recruited immunosuppressive macrophages and neutrophils associated with LKB1-mutant lung adenocarcinoma. Antibody-mediated blockade of LIF depleted inflammatory cancer cell states, leading to reversal of the immunosuppressive microenvironment and enhanced antitumor immune responses. These findings provide a rationale for targeting LIF in solid tumors.See article, p. 1436.CAR T-cell therapies can fail to achieve durable clinical responses owing to T-cell exhaustion under chronic stimulation. Integrating an in vivo chronic stimulation CRISPR screen with a multiparametric in vitro screen, Jain, Shi, May, and colleagues identified NFIL3 as a key driver of T-cell dysfunction. Disruption of NFIL3 enhanced CAR T-cell expansion, sustained cytotoxicity, reduced inhibitory receptor expression, and promoted a transcriptional program associated with favorable clinical outcomes. Consistent with these findings, NFIL3-deficient CAR T cells demonstrated superior antitumor activity across multiple hematologic and solid tumor models, establishing NFIL3 as a promising target for next-generation immunotherapies.See article, p. 1456.