Introduction

Esophageal squamous cell carcinoma (ESCC) is associated with poor prognosis despite advances in neoadjuvant immunochemotherapy (NICT). With highly variable treatment responses, identifying reliable predictive biomarkers remains crucial. While gut microbiome influences on immunotherapy efficacy are increasingly recognized, the specific role of the fungal microbiome (mycobiome) in NICT outcomes for ESCC remains unexplored, representing a significant gap in developing personalized treatment strategies.

Aims & Methods

This study aimed to characterize gut mycobiome signatures in ESCC patients and develop fungal biomarker-based predictive models for NICT response. We conducted ITS2 sequencing on 136 fecal samples collected from 68 ESCC patients (pre- and post-NICT) and 19 healthy controls (HC). Patients were stratified as responders (R, n=26) or non-responders (NR, n=42) based on tumor regression grade (TRG) scores. We performed comprehensive analyses of mycobiome diversity, composition, co-occurrence networks, and functional pathways across groups. The Boruta algorithm was employed for feature selection to identify fungal biomarkers, followed by multilayer perceptron (MLP) machine learning with five-fold cross-validation to develop predictive models. Additionally, ESCC xenograft models were established to investigate gut mycobiome-immune interactions through fecal microbiota transplantation (FMT) and targeted fungal manipulation experiments.

Results

ESCC patients demonstrated significant mycobiome dysbiosis compared to healthy controls, characterized by reduced fungal alpha diversity and distinct beta diversity patterns. At baseline, ESCC patients showed enrichment of Rhodotorula minuta, Actinomucor elegans, and several Candida species. Responders exhibited significantly higher baseline fungal diversity than non-responders, with enrichment of Meyerozyma, Candida boidinii, Trichosporon dermatis, and Cryptococcus species, whereas non-responders showed predominance of Saccharomyces, Nakaseomyces, and pathogenic Candida species (C. glabrata, C. parapsilosis, C. albicans). Responders also displayed more complex, stable fungal co-occurrence networks with higher density and clustering coefficients. Post-NICT, both groups showed partial recovery of mycobiome diversity, with significant increases in beneficial fungal taxa. Functional pathway analysis revealed distinct metabolic profiles between responders and non-responders. Using the Boruta algorithm, we identified 23 genus-level and 22 species-level fungal biomarkers that demonstrated high predictive value. The MLP model based on genus-level markers achieved AUCs of 90.2% (training) and 91.9% (test), while the species-level model showed comparable performance. In animal experiments, mice receiving R-FMT exhibited reduced tumor growth and enhanced CD8+ T cell infiltration compared to NR-FMT recipients. Notably, Candida boidinii supplementation synergized with anti-PD-1 therapy, while Saccharomyces depletion enhanced treatment sensitivity.

Conclusion

Our findings establish the gut mycobiome as a powerful predictor of NICT response in ESCC, with machine learning models achieving >90% accuracy using fungal biomarkers. Experimental validation not only confirms mechanistic links between specific fungi and antitumor immunity but also demonstrates that targeted mycobiome modulation can enhance immunotherapy efficacy. This dual approach of prediction and intervention offers a novel framework for precision oncology in ESCC, potentially transforming clinical decision-making and treatment outcomes in this challenging malignancy.

References

1. Li J, Xu J, Zheng Y, et al. Esophageal cancer: Epidemiology, risk factors and screening. Chin J Cancer Res. 2021; 33(5): 535-547.
2. van Hagen P, Hulshof MC, van Lanschot JJ, et al. Preoperative chemoradiotherapy for esophageal or junctional cancer. N Engl J Med. 2012; 366(22): 2074-2084.
3. Schroeder W, Ghadimi M, Schloesser H, et al. Long-Term Outcome After Histopathological Complete Response with and Without Nodal Metastases Following Multimodal Treatment of Esophageal Cancer. Ann Surg Oncol. 2022.
4. Rosner S, Liu C, Forde PM, Hu C. Association of Pathologic Complete Response and Long-Term Survival Outcomes Among Patients Treated With Neoadjuvant Chemotherapy or Chemoradiotherapy for NSCLC: A Meta-Analysis. JTO Clin Res Rep. 2022; 3(9): 100384.
5. Coker OO, Nakatsu G, Dai RZ, et al. Enteric fungal microbiota dysbiosis and ecological alterations in colorectal cancer. Gut. 2019; 68(4): 654-662.
6. Dohlman AB, Klug J, Mesko M, et al. A pan-cancer mycobiome analysis reveals fungal involvement in gastrointestinal and lung tumors. Cell. 2022; 185(20): 3807-3822.
7. Wurster S, Albert ND, Kontoyiannis DP. Candida auris Bloodstream Infection Induces Upregulation of the PD-1/PD-L1 Immune Checkpoint Pathway in an Immunocompetent Mouse Model. Msphere. 2022; 7(2): e81721.

Introduction

Endoscopic ultrasound (EUS) is the gold standard for evaluating pancreatic lesions, allowing for real-time imaging and biopsy sampling. Nevertheless, the differentiation between solid pancreatic lesions — such as pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors (PNETs) — as well as cystic lesions is often troublesome. Specifically, identifying intraductal papillary mucinous neoplasms (IPMNs) with varying degrees of dysplasia is crucial for guiding clinical management. However, the diagnostic accuracy of current methods remains suboptimal. This multicenter study aimed to develop and validate convolutional neural networks (CNNs) to characterize solid and cystic pancreatic lesions and stratify IPMNs into high-grade dysplasia/carcinoma (HGD/C) and low-grade dysplasia (LGD).

Aims & Methods

This multicenter study included EUS images from 12 centers in Spain, Brazil, the United Kingdom and the United States of America. Solid lesion differentiation (PDAC vs. PNET) was achieved using a CNN trained on 49034 images from 95 patients (n=69 for PDAC and n=26 for PNET), with diagnoses confirmed via EUS-guided fine-needle biopsy or surgical specimen analysis. A second CNN was trained on 79395 images from 82 patients (n=64 for mucinous cystic lesions and n=18 for non-mucinous cystic lesions). Finally, a third model used the subset of IPMNs (n=51046 frames, 30 patients) to distinguish between HGD/C (n=5 patients) and LGD (n=25 patients). Histological confirmation was used as the gold standard. Performance metrics included sensitivity, specificity, accuracy, and area under the precision-recall curve (AUC-PR). To comprehensively determine the model's performance, we evaluated multiple thresholds for the confidence score and the IoU, between 0.45 and 0.55 with a step of 0.02 ([0.45, 0.47, 0.50, 0.52, 0.55]). The datasets were split into training (70%), validation (20%), and testing (10%) subsets.

Results

For solid lesions, the model differentiated PDAC from PNETs with an accuracy of 91.8%, with sensitivities of 96.2% and 97.8%, respectively. The AUC-PR was 0.982 for PDAC and 0.977 for PNET. Regarding cystic lesions the model detected mucinous cystic lesions with a sensitivity of 97.5%, an overall accuracy of 89.7% and an AUC-PR of 0.977. When the subset of IPMNs was specifically analyzed, the model distinguished IPMNs with HGD/C from those with LGD with a sensitivity of 97.5%, specificity of 94.3%, and an overall accuracy of 89.8%. The AUC was 0.988.

Conclusion

This is the first multicenter, transatlantic AI study aiming to provide a holistic approach to both solid and cystic pancreatic lesions, focusing on those with a significantly higher malignant potential. Our models achieved high accuracy in identifying and differentiating these conditions. Furthermore, to our knowledge this is the first integrative model to include the ability to stratify IPMNs into LGD and HGD/C, which has significant clinical implications, potentially reducing unnecessary surgeries and ensuring timely intervention for high-risk patients. The inclusion of data from diverse regions reduces demographic bias and enhances model generalizability. Real-time implementation of these solutions could significantly improve decision-making in EUS and optimize patient outcomes.