Heidegger Isabel, Frantzi Maria, Salcher Stefan, Tymoszuk Piotr, Martowicz Agnieszka, Gomez-Gomez Enrique, Blanca Ana, Lendinez Cano Guillermo, Latosinska Agnieszka, Mischak Harald, Vlahou Antonia, Langer Christian, Aigner Friedrich, Puhr Martin, Krogsdam Anne, Trajanoski Zlatko, Wolf Dominik, Pircher Andreas
Department of Urology, Medical University of Innsbruck, Innsbruck, Austria.
Department of Biomarker Research, Mosaiques Diagnostics GmbH, Hannover, Germany.
Eur Urol Oncol. 2025 Jun;8(3):652-662. doi: 10.1016/j.euo.2024.05.014. Epub 2024 Jun 8.
While collagen density has been associated with poor outcomes in various cancers, its role in prostate cancer (PCa) remains elusive. Our aim was to analyze collagen-related transcriptomic, proteomic, and urinome alterations in the context of detection of clinically significant PCa (csPCa, International Society of Urological Pathology [ISUP] grade group ≥2).
Comprehensive analyses for PCa transcriptome (n = 1393), proteome (n = 104), and urinome (n = 923) data sets focused on 55 collagen-related genes. Investigation of the cellular source of collagen-related transcripts via single-cell RNA sequencing was conducted. Statistical evaluations, clustering, and machine learning models were used for data analysis to identify csPCa signatures.
Differential expression of 30 of 55 collagen-related genes and 34 proteins was confirmed in csPCa in comparison to benign prostate tissue or ISUP 1 cancer. A collagen-high cancer cluster exhibited distinct cellular and molecular characteristics, including fibroblast and endothelial cell infiltration, intense extracellular matrix turnover, and enhanced growth factor and inflammatory signaling. Robust collagen-based machine learning models were established to identify csPCa. The models outcompeted prostate-specific antigen (PSA) and age, showing comparable performance to multiparametric magnetic resonance imaging (mpMRI) in predicting csPCa. Of note, the urinome-based collagen model identified four of five csPCa cases among patients with Prostate Imaging-Reporting and Data System (PI-IRADS) 3 lesions, for which the presence of csPCa is considered equivocal. The retrospective character of the study is a limitation.
Collagen-related transcriptome, proteome, and urinome signatures exhibited superior accuracy in detecting csPCa in comparison to PSA and age. The collagen signatures, especially in cases of ambiguous lesions on mpMRI, successfully identified csPCa and could potentially reduce unnecessary biopsies. The urinome-based collagen signature represents a promising liquid biopsy tool that requires prospective evaluation to improve the potential of this collagen-based approach to enhance diagnostic precision in PCa for risk stratification and guiding personalized interventions.
In our study, collagen-related alterations in tissue, and urine were able to predict the presence of clinically significant prostate cancer at primary diagnosis.
虽然胶原蛋白密度与多种癌症的不良预后相关，但其在前列腺癌（PCa）中的作用仍不明确。我们的目的是在检测临床显著性前列腺癌（csPCa，国际泌尿病理学会[ISUP]分级组≥2）的背景下，分析与胶原蛋白相关的转录组、蛋白质组和尿液组改变。
对PCa转录组（n = 1393）、蛋白质组（n = 104）和尿液组（n = 923）数据集进行综合分析，重点关注55个与胶原蛋白相关的基因。通过单细胞RNA测序研究与胶原蛋白相关转录本的细胞来源。使用统计评估、聚类和机器学习模型进行数据分析，以识别csPCa特征。
与良性前列腺组织或ISUP 1级癌症相比，在csPCa中证实了55个与胶原蛋白相关基因中的30个以及34种蛋白质的差异表达。一个高胶原蛋白癌症簇表现出独特的细胞和分子特征，包括成纤维细胞和内皮细胞浸润、强烈的细胞外基质周转以及增强的生长因子和炎症信号传导。建立了强大的基于胶原蛋白的机器学习模型来识别csPCa。这些模型优于前列腺特异性抗原（PSA）和年龄，在预测csPCa方面表现出与多参数磁共振成像（mpMRI）相当的性能。值得注意的是，基于尿液组的胶原蛋白模型在前列腺影像报告和数据系统（PI-IRADS）3类病变患者中识别出了五例csPCa病例中的四例，而这些病例中csPCa的存在被认为是不明确的。本研究的回顾性特点是一个局限性。
与PSA和年龄相比，与胶原蛋白相关的转录组、蛋白质组和尿液组特征在检测csPCa方面表现出更高的准确性。胶原蛋白特征，特别是在mpMRI上病变不明确的情况下，成功识别了csPCa，并可能减少不必要的活检。基于尿液组的胶原蛋白特征代表了一种有前景的液体活检工具，需要进行前瞻性评估，以提高这种基于胶原蛋白的方法在PCa风险分层和指导个性化干预中提高诊断准确性的潜力。
在我们的研究中，组织和尿液中与胶原蛋白相关的改变能够在初次诊断时预测临床显著性前列腺癌的存在。
