The development of new drugs and focused treatment strategies is significantly hampered by the inadequate understanding of the complex pathophysiology of esophageal cancer (EC). Significant risk factors for the development of EC include imbalances in the oxidative stress and anti-oxidative response pathways. It has been discovered that therapeutic interventions that target the ubiquitin-specific protease 10 (USP10) and its targeted response element in response to oxidative stress can inhibit tumour growth and provide notable clinical benefits for cancer patients. To determine if USP10 is a ubiquitin-specific protease that controls oxidative stress, which in turn stabilises NOTCH1 and deactivates the anti-oxidative transcription-related pathway, and to investigate the impact of USP10 on the development of EC. As a major endogenous stabiliser of NOTCH1 intracellular domain (NICD1) function, we recognized USP10 as a protein that interacts with NICD1 and catalyses the deubiquitination of NICD1 in EC cells. Loss of USP10 specific to epithelial cells interferes with NICD1 stabilisation, which in turn stimulates the production of antioxidant genes to prevent ferroptosis caused by lipid peroxidation and significantly accelerates the development of esophageal cancer. On the other hand, in animal models treated with 4-NQO, transgenic overexpression-mediated USP10 gene therapy reduces the development of esophageal cancer. Mechanistically, USP10 binds to NICD1 in response to oxidative stress, deconjugates ubiquitination chains, and increases NICD1 abundance and the activation of its downstream signalling cascade. In vivo investigations have also shown that inhibiting USP10 expression significantly reduces the impacts that cause cancer. Importantly, a bad prognosis is indicated by reduced expression of USP10, which is linked to the severity of EC. According to these results, USP10 might be a suitable therapeutic target for the treatment of esophageal cancer.
Recurrence of IDH-wild-type glioblastoma (GBM) remains the major obstacle to improving patient survival. Current recurrence models emphasize tumor-intrinsic mechanisms, including therapeutic resistance, cellular plasticity, and clonal evolution, but do not fully explain how residual tumor cells reconstruct recurrent disease after surgery. Here, we propose an NGS-centered seed-soil framework that considers surgical resection as the starting point of recurrence evolution: residual NEU/OPC/NPC-like tumor cells represent potential recurrence "seed", the postsurgical brain constitutes a dynamically remodeled "soil", and neuron-glioma synapse (NGS) provides a potential interface between residual tumor cells and neurons of this environment. At the recurrence endpoint, recurrent GBM exhibits enhanced neuronal and NGS-associated characteristics, providing an observable phenotype with which this framework can be evaluated. We then examine the potential processes linking these two states, including therapy-associated NGS remodeling, NGS-associated network reconstruction, and neuronal-like migration, which may contribute to the development of local and distant recurrence. Throughout, established observations are distinguished from hypothesis-driven interpretations and untested predictions. Finally, we discuss critical experimental requirements, including longitudinal post-resection models and NGS-specific perturbation, as well as emerging translational opportunities such as neuromodulation and perioperative intervention. This framework shifts the perspective on GBM recurrence from residual tumor-cell survival alone toward dynamic interactions between residual tumor cells and the postsurgical neural environment.
Colorectal cancer (CRC) is the third most commonly diagnosed cancer worldwide, accounting for approximately 10% of all cases, and is the second leading cause of cancer-related deaths in the United States. In patients with Familial Adenomatous Polyposis (FAP), APC mutations drive c-MYC overexpression, promoting metabolic and immune dysregulation that contributes to CRC development, making c-MYC a potential target for FAP-associated chemoprevention. Although no c-MYC inhibitors are currently available, recent evidence indicates that a uropathogenic E. coli protease can degrade c-MYC protein. This study evaluated recombinant LON protease (rLONP) for chemopreventive efficacy in PIRC rats, a model of FAP. Rat pups were genotyped, baseline polyp burden was established by colonoscopy, and animals were randomized to receive saline or rLONP twice daily (2.5 mg/kg, p.o.) for 2 or 4 weeks. Serum analyses showed no significant changes in liver or kidney function markers in rLONP-treated rats compared with controls. rLONP-treated male rats exhibited a ∼40% reduction in colonic polyps at both time points, and small intestinal polyp counts were also lower in treated males and females. Combined analysis demonstrated a ≥60% reduction in total intestinal polyp multiplicity after 2-4 weeks of treatment (p<0.001). Protein expression analyses of colonic polyps revealed decreased c-MYC protein levels following rLONP treatment. Immunohistochemistry further showed reduced Ki-67, Cyclin D1, and c-MYC expression in treated polyps compared with controls. Overall, rLONP was well tolerated and significantly reduced intestinal polyp burden in PIRC rats in this short-term study, supporting further long-term efficacy evaluation.
Malignant solid tumors comprise not only cancer cells but also diverse non-cancerous stromal cells that shape the tumor microenvironment. The tricarboxylic acid (TCA) cycle has an overarching presence in providing substrates needed to drive the electron transport chain and, ultimately, ATP synthesis. However, it remains unclear which stromal cell lineages influence tumor growth through TCA-dependent mitochondrial function, and whether such activities act in a tumor-promoting or tumor-suppressive manner. Isocitrate dehydrogenase 3 (IDH3), a rate-limiting TCA cycle enzyme that generates NADH to support mitochondrial respiration, provides a genetic entry point to interrogate mitochondrial TCA-dependent function in stromal cells. In this study, we established a mouse model in which tamoxifen administration induces CreERT2-dependent knockout of the α subunit of IDH3 (IDH3α) in all somatic cells. Using this model with transplantation of Idh3a-intact murine cancer cells, we found that host Idh3a deficiency accelerated growth of murine MC38 tumors in a cancer cell line-dependent manner. Bone marrow chimera experiments indicated that hematopoietic lineages were not responsible for this phenotype, suggesting a contribution from tissue-resident non-hematopoietic stromal cells that are not replaced by bone marrow transplantation. Single-cell RNA sequencing of human tumor specimens revealed broad IDH3A expression across multiple tumor microenvironment compartments, including fibroblasts. Consistently, in vitro co-culture assays demonstrated that Idh3a-intact, but not Idh3a-KO, fibroblasts suppressed cancer cell proliferation in a contact-dependent manner. Together, these findings identify IDH3α-dependent mitochondrial function in fibroblasts as a critical determinant of tumor progression and suggest that stromal mitochondrial metabolism represents an important axis for modulating tumor behavior.
Background The relative contribution of HLA class I molecules—including classical HLA class Ia (HLA-A, -B, -C) and non-classical HLA class Ib (HLA-E, -F, -G)—to shaping the tumor immune microenvironment in colon cancer remains insufficiently defined. We investigated how their expression patterns relate to immune infiltration and clinical outcome. Methods In a retrospective cohort of 280 colon cancers, we assessed HLA class Ia and class Ib expression and quantified CD45-positive immune cell infiltration by immunohistochemistry. These features were correlated with clinicopathological variables, microsatellite instability (MSI) status, and previously established genomic immune signatures. Results High HLA class Ia expression and increased CD45-positive cell infiltration were each associated with improved overall, disease-specific, and progression-free survival. CD45-positive density correlated strongly with Immunologic Constant of Rejection scores. HLA class Ia loss was more frequent in advanced stages and in MSI-H tumors. Among HLA class Ib molecules, only HLA-E expression was associated with favorable disease-specific and progression-free survival. Integrative analysis identified three immune phenotypes with distinct prognostic profiles; tumors characterized by high HLA class Ia expression, low HLA class Ib expression, and high CD45-positive infiltration had the best outcomes. Conclusion Colon cancer immunogenicity is shaped by coordinated patterns of HLA class I expression and immune infiltration. Integrating HLA class Ia/Ib expression with immune cell density provides a refined stratification of tumor immune phenotypes and may support personalized immunotherapeutic decision-making.
Immune checkpoint blockade has transformed cancer therapy, yet primary and acquired resistance remain major clinical challenges. Increasing evidence indicates that immunotherapy resistance cannot be fully explained by tumor-intrinsic alterations or conventional biomarkers such as PD-L1 expression, tumor mutational burden, or microsatellite instability. Instead, therapeutic response is shaped by the tumor immune microenvironment (TIME) as a heterogeneous, spatially organized, and dynamically evolving ecosystem. Single-cell omics has revealed diverse immune and stromal cell states, including progenitor and terminally exhausted T cells, suppressive myeloid programs, B-cell/TLS-associated immune-reactive states, and CAF-mediated exclusion phenotypes. Spatial transcriptomics, spatial proteomics, and imaging-based approaches further demonstrate that these cell states assemble into distinct immune niches, including immune-inflamed, T-cell-excluded, myeloid-suppressive, metabolic/hypoxic, and TLS-associated niches. These spatial ecosystems determine whether antitumor immune cells can access malignant cells, receive antigen-presenting support, or become restrained by stromal, vascular, metabolic, and myeloid barriers. In this review, we summarize how single-cell and spatial multi-omics redefine TIME heterogeneity in immunotherapy resistance, highlight ligand–receptor communication networks linking cell states to spatial immune dysfunction, and discuss emerging translational biomarkers for patient stratification. We further propose that future immunotherapy biomarkers should evolve from static single-marker assays toward longitudinal, spatially resolved, and interpretable multi-omics models that guide precision combination immunotherapy.
SMARCA4-deficient thoracic malignancies comprise biologically heterogeneous tumors, ranging from conventional non-small cell lung cancer with SMARCA4 alterations to thoracic SMARCA4-deficient undifferentiated tumor (SMARCA4-UT), an aggressive entity frequently associated with concomitant SMARCA2 loss. However, the extent to which SMARCA4-deficient lung cancer cell lines recapitulate SMARCA4-UT-like biology remains incompletely defined. Here, we characterized lung cancer cell lines across distinct SMARCA4 and SMARCA2 states and identified a subgroup with SMARCA4/2 co-deficiency that exhibited reduced expression of epithelial lineage markers and transcriptional similarity to SMARCA4-UT and other SWI/SNF-deficient malignancies. The EZH1/2 inhibitor HM97662 selectively suppressed growth in SMARCA4/2-deficient cells, with limited effects in SMARCA2-proficient cells. EZH1/2 inhibition broadly reduced H3K27me3 and induced derepression of PRC2 targets regardless of drug sensitivity. However, its biological effects were most pronounced in SMARCA4/2-deficient cells, where it promoted apoptosis, reduced stemness marker expression, attenuated the SMARCA4-UT-associated transcriptional signature, and suppressed proliferative and mTORC1-related programs. Chromatin accessibility analysis further revealed cell-line-specific patterns of accessibility loss, with reduced accessibility at stemness-associated transcription factor motif-enriched regions coupled with transcriptional repression of nearby genes in SMARCA4/2-deficient cells. These findings support dual EZH1/2 inhibition as a potential therapeutic vulnerability in SMARCA4/2-deficient, SMARCA4-UT-like lung cancer cells.
Dual-target c-Met/VEGFR kinase inhibitors have emerged as promising therapeutic agents for cancer therapy, although their underlying antitumor mechanisms remain poorly defined. In this study, we demonstrate that foretinib, a selective c-Met/VEGFR2 inhibitor, triggers immunogenic cell death (ICD) in colorectal cancer (CRC) cells. This process is characterized by the surface exposure of calreticulin (CRT) and extracellular secretion of ATP and high-mobility group box 1 (HMGB1), which subsequently drives the maturation and activation of dendritic cells (DCs). Mechanistically, foretinib activates the p53 signaling pathway, thereby transcriptionally upregulating Death Receptor 5 (DR5), a critical mediator of the extrinsic apoptotic cascade. Furthermore, foretinib remodels the tumor immune microenvironment (TIME) by facilitating DCs maturation and potentiating the infiltration and activation of T cells. Moreover, combined treatment with foretinib and anti-CD47 antibody exerts synergistic antitumor effects. Together, these findings uncover a previously unreported immunomodulatory function of c-Met/VEGFR2 inhibition, demonstrating that foretinib elicits potent anti-CRC efficacy via direct cytotoxicity and immunogenic regulation, and highlighting its considerable clinical therapeutic potential.
Aberrant mucin-type O-glycosylation is a common feature of pancreatic ductal adenocarcinoma (PDAC), yet its functional contribution to pancreatic tumorigenesis remains incompletely defined. In particular, whether defective core 1 O-glycan maturation cooperates with oncogenic KRAS during early pancreatic neoplasia has not been fully resolved. Here, we addressed this question by deleting Cosmc (C1galt1c1), the obligate molecular chaperone for T-synthase (C1galt1), in a pancreas-specific Kras-driven mouse model. Analysis of human PDAC tissue microarrays revealed frequent expression of the truncated O-glycan epitopes Tn and sialyl-Tn, supporting the clinical relevance of impaired O-glycan elongation in pancreatic cancer. In vivo, pancreas-specific Cosmc loss in the setting of oncogenic Kras accelerated pancreatic tumor progression, enhanced tissue proliferation, altered acidic mucin-associated glycosylation, and promoted pronounced stromal remodeling. Lectin histochemistry and biochemical analyses confirmed robust accumulation of Tn antigen in Cosmc-deficient pancreatic tumors. To define molecular programs associated with this phenotype, we performed transcriptomic and proteomic profiling of tumor-derived PDAC cell lines together with VVA-enriched glycoproteomic analysis of pancreatic tissues and cell lines. These approaches identified coordinated remodeling of pathways linked to extracellular matrix organization, adhesion, cytoskeletal regulation, stress adaptation, and metabolic reprogramming. Together, these data identify Cosmc-dependent O-glycosylation as a regulator of Kras-driven pancreatic carcinogenesis. Our findings support a model in which truncated O-glycans contribute to PDAC progression in association with remodeling of the tumor glycoproteome and stromal architecture during Kras-driven pancreatic tumor evolution.
During carcinoma progression, CD8+ T cells mediate tumor-cytolytic functions following tumor-antigen sensitization, typically by antigen cross-presenting dendritic cells. Lymphatic endothelial cells (LECs) are not typically endowed with such immune-effector capacity, and often promote tolerance. Inspired by insights from dendritic-cell glycan targeting, we examined how antigen-pulsed lymph node LECs genetically deficient in Ndst1, expressing under-sulfated heparan sulfate, might affect antigen-sensitized CD8+ T cell responses. Mutation significantly augmented responses to Ova-peptide pulsed LECs, including marked Nur77 activation along with more modest IFN-γ induction in co-cultured Ova-responsive CD8+ T cells, and increased antigen presentation relative to wildtype LECs (without marked differences in MHC-I, CD86 or PD-L1 immune checkpoint ligands). Tumor draining lymph nodes of lymphatic targeted tumor-bearing Ndst1f/f Prox1Cre+ mutant mice showed reduced metastatic loads accompanied by augmented CD8+ T cell densities compared to Ndst1f/f Prox1Cre- controls. These findings provide mechanistic insight on harnessing glycan-targeted lymphatic endothelium to boost effector T cell immunity.
Cancer cells can exploit developmental lineage programs to generate phenotypic heterogeneity under therapeutic pressure. Although a subset of resistant tumors preserves its founding lineage, receptor, or oncogenic dependency, accumulating evidence shows that others enter reversible persister states or stabilize alternative lineage programs. Here we frame malignant plasticity as developmentally constrained, not limitless: cell of origin, lineage history, injury memory, genetic gates, chromatin state, transcription-factor circuits, and tumor-ecosystem feedback help delimit and probabilistically bias which state transitions are accessible under therapy. We distinguish physiological expansion of state space during repair from premalignant permissiveness and malignant fixation, and classify resistance into three modes: lineage-maintained resistance, adaptive reversible plasticity, and fixed reprogramming through lineage switching or histological transformation. These modes should be read as diagnoses rather than rigid therapeutic silos. Prime-then-kill strategies are most defensible when a resistant state is reversible, targetable, and paired with readouts; they may also be considered as biomarker-defined add-on hypotheses in lineage-maintained or lineage-rerouted disease when an evidence-supported state or immune-visibility module is present. Conversely, tumors that retain driver, receptor, or lineage dependency should keep the preserved axis or bypass pathway as the therapeutic backbone, and fixed histological transformation often requires treatment according to the new lineage. We also discuss how single-cell and spatial multi-omics can map state-space breadth and ecosystem context and, when paired with perturbational designs, help test transition capacity and reversibility; static marker expression alone cannot establish plasticity. A resistance-mode-guided approach can sharpen therapeutic hypotheses and limit overgeneralization across tumor types.
Chronic inflammation increases cancer risk, yet the mechanisms linking transient injury to long-term susceptibility remain elusive. Nagaraja et al [1] demonstrate that intestinal stem cells retain durable epigenetic memory of colitis, thereby priming tumorigenesis. This discovery reframes cancer initiation as a consequence of heritable chromatin states, introducing an additional regulatory dimension that operates alongside, and beyond, genetic aberrations. By positioning inflammation induced epigenetic memory as a novel axis for cancer predisposition, this study open avenues for biomarker development and preventive therapeutic strategies.
Introduction Synthetic lethal interactions with IDH1 and IDH2 (IDH) mutations were identified in non-endogenous IDH mutant (IDHMUT) AML and glioma models, but are absent in endogenous IDHMUT chondrosarcoma cell lines. The translation into successful clinical applications has remained challenging, implying artificially created models do not fully recapitulate endogenous IDHMUT tumour biology. The aim of this study was to elucidate if the model system is indeed an important factor to consider when studying therapeutic vulnerabilities in IDHMUT tumours. Methods Vector-based and CRISPR-Cas9 approaches were used to introduce or revert the IDH1 mutation in chondrosarcoma cell lines. These isogenic cell line pairs were used to examine the presence of known therapeutic vulnerabilities and their underlying biological mechanisms. Results Vector-based IDHMUT chondrosarcoma models showed the previously reported synthetic lethal interactions, but these treatment sensitivities were absent in the CRISPR-edited models. Interestingly, not all vector-based IDHMUT cell lines displayed the same therapeutic vulnerabilities. Differences in treatment response were associated with multiple factors, including IDHMUT protein expression and D-2-HG levels, in line with the fact that therapeutic vulnerabilities could be induced in the CRISPR-edited models by enhancing D-2-HG levels. Conclusion Our findings demonstrate that synthetic lethal interactions observed in vector-based models are often a consequence of IDHMUT protein overexpression and supra-physiological D-2-HG levels. These results highlight that relying on artificially created IDHMUT models may lead to the identification of therapeutic vulnerabilities that are not present in IDHMUT tumours, potentially explaining the poor translation of preclinical findings to clinical trials.
Cervical squamous cell carcinoma (CESC) is frequently complicated by cisplatin (CDDP) resistance, which is a primary cause of treatment failure and poor prognosis. Here, we identified a CLCF1-ZIC5-CNTFR feedback loop that contributes to CDDP resistance in CESC. We found that ZIC5 was significantly upregulated in CDDP-resistant CESC and correlated with worse overall survival. Functionally, ZIC5 promoted CDDP resistance, in part by enhancing properties associated with cancer stem cells (CSCs) through the transcriptional activation of CNTFR. CLCF1, the ligand for CNTFR, was found to induce ZIC5 expression via STAT3 signaling, thereby forming a CLCF1–ZIC5–CNTFR positive feedback loop. Targeting CNTFR with a neutralizing antibody disrupted this regulatory circuit, leading to a reversal of CDDP resistance and suppression of CSC-associated phenotypes in vitro. Furthermore, combining an anti-CNTFR antibody with CDDP resulted in enhanced antitumor effects in patient-derived xenograft (PDX) and organoid (PDO) models. In summary, our findings suggest that the CLCF1-ZIC5-CNTFR positive feedback loop promotes CDDP resistance in CESC, potentially by sustaining CSC properties. Targeting this pathway, particularly CNTFR, may represent a potential therapeutic strategy for patients with CDDP-resistant CESC.
ABL-rearranged (ABLr) acute lymphoblastic leukemia (ALL) is associated with treatment failure and relapse and novel treatments are required. We investigated asciminib efficacy against NUP214::ABL1 ALL, the second most common ABL1 rearrangement associated with aggressive disease. Asciminib activity was established in three patient derived xenograft models of NUP214::ABL1 ALL with different ABL1 breakpoints (e31a2, e32a3, e34a3). In all models, treatment with asciminib reduced NUP214::ABL1 leukemic burden and increased survival outcomes compared with control mice. These results contrast with recent in vitro studies in the setting of BCR::ABL1 leukemia, where ABL1 exon 3 breakpoints (e13a3, e14a3) result in asciminib resistance due to incomplete SH3 domain. Conversely, in silico modeling of NUP214::ABL1 e32a3 predicted NUP214 exon 32 mimics the missing ABL1 exon 2, forming a chimeric SH3 domain that rescues asciminib sensitivity. This prediction was supported by viability assays of Ba/F3 cells expressing NUP214::ABL1 with exon 32 present and absent. Additional viability assays and structural modeling of NUP214::ABL1 Ba/F3 cells expressing various ABL1 deletions defined a region of the SH3 domain critical for asciminib efficacy and necessary for allosteric inhibition. Our findings establish asciminib as a potential treatment for NUP214::ABL1 ALL and reveal a critical role for the fusion partner gene. Asciminib efficacy in ABLr leukemia cannot be predicted by evaluating ABL1 exon 2 inclusion/exclusion in isolation. This is of clinical importance as first-line use of STAMP inhibitors such as asciminib becomes more common.
Antibody-drug conjugate (ADC) efficacy in solid tumors is often limited by low or heterogeneous antigen expression. Here, we investigated low-dose radiotherapy (LDRT) as a noninvasive sensitization strategy to enhance tumor surface antigen expression and improve ADC activity. Transcriptomic profiling and flow cytometry identified ICAM1 as the most strongly radiation-inducible antigen after 2 Gy irradiation, showing an approximately 4-fold increase and exceeding several established ADC targets. Mechanistic studies revealed that radiation-induced reactive oxygen species (ROS) and oxidative stress signaling upregulated ICAM1, enhanced antibody internalization, and improved ICAM1-targeted ADC delivery. In xenograft models, LDRT combined with ICAM1-targeted ADCs achieved >70% tumor growth inhibition with minimal toxicity. Transcriptomic and immunohistochemical (IHC) staining analyses further showed that LDRT induced recruitment of immunosuppressive CCL8high M2-like tumor-associated macrophages, whereas ADC-mediated bystander killing attenuated this adaptive feedback and promoted a more immunoactive tumor microenvironment. Together, these findings support LDRT as a clinically translatable strategy to increase ICAM1 expression, overcome antigen-low barriers, and enhance ADC efficacy in refractory solid tumors.
Pharmacotherapy induces complex molecular reprogramming in cancer, driving transcriptome-wide alterations and widespread dysregulation of alternative splicing. Despite these profound changes, there remain limited resources characterizing drug-induced whole-transcriptomic responses in cancer. Furthermore, while aberrant splicing can generate immunogenic neoantigens, existing resources fail to systematically integrate drug perturbations, splicing dynamics, and neoantigen landscapes. To address this gap, the DRIVE database was constructed as a comprehensive resource detailing drug-induced transcriptomic and splicing responses. Utilizing the large language models for rigorous metadata curation and construct the standardized processing pipeline, thousands of publicly available raw transcriptomic datasets from drug-treated and control cancer cell lines were systematically processed. The resulting repository encompasses 3,911 samples, involving 278 drugs and 272 cell lines, enabling the precise quantification of differential gene expression, differential alternative splicing events, and the prediction of splicing-derived human leukocyte antigen-binding peptides. Analysis of the data revealed that drug-induced transcriptomic reprogramming is highly context-dependent and correlated with chemical structural similarity. We identified Osimertinib as a potential immunomodulatory agent associated with transcriptional signatures of an activated tumor microenvironment, while KB-0742 emerged as an unappreciated candidate global splicing modulator. Furthermore, our large-scale prediction of differential splicing-derived neoantigens uncovered several drugs that warrant further investigation as candidates for combination immunotherapy. DRIVE also provides a user-friendly interface to browse datasets, perform drug enrichment and connectivity analysis. (https://componclab.com/DRIVE). This database could improve our understanding of molecular reprogramming under pharmacotherapy, and serve as a valuable platform for deciphering drug mechanisms, promoting virtual cell modeling and discovering novel strategies of drug repurposing.
Background: HPV-positive (HPV+) oropharyngeal squamous cell carcinoma (OPSCC) patients have improved survival compared to HPV-negative OPSCC (HPV–) patients (5-year survival rates of ∼80% versus ∼40%). The predictive value of p16 immunohistochemistry (IHC) for HPV+ OPSCC, the most widely used clinical test for HPV status, varies by HPV+ OPSCC prevalence in geographically different populations, but variation within a region is not well described. Methods: We analyzed p16 IHC and survival in two geographically and demographically diverse OPSCC cohorts of African American (AA, n = 177, Louisiana Tumor Registry) and non-Hispanic white (NHW, n = 392, University of Michigan) patients. Results: HPV+ prevalence was 29.3% in AA versus 86.8% in NHW. p16+ OPSCC was associated with significantly better survival in NHW than in AA patients (HR=4.65, 95% CI: 2.88-7.51, Cox p < 0.0001), adjusting for covariates. AA p16– had worse survival than NHW p16– (log-rank p < 0.001). p16 was less predictive of HPV RNA status for AA than NHW patients (positive predictive value = 65.4% vs. 94.9%, p < 0.05). Among HPV RNA+ patients (79 NHW,17 AA), AA had worse survival than NHW (log-rank p < 0.001). Conclusions: Survival disparities exist after accounting for heterogeneous HPV+ tumor rates. The PPV of p16 for HPV RNA–defined OPSCC is substantially lower in AA vs NHW patients. Guidelines for HPV testing in OPSCC should be attentive to factors beyond geography.
