Urothelial Carcinoma – Ureteral and Renal Pelvic Tumors
Overview
Ureteropelvic and renal pelvic carcinoma have similar origins, however, variations in genetic and clinical parameters makes them two distinct conditions. The single-center data obtained for the molecular subtypes and characteristics of the two muscle-invasive tumors were validated using clinical data from the Surveillance, Epidemiology, and End Results (SEER) database. Also, a deep learning algorithm was used for the accurate classification of molecular subtypes seen in histology to increase the sample size for more verification. It was proposed that there may be a predilection of renal pelvic carcinoma towards the lumen and urethral carcinoma to be inclined towards the basal membrane and p53-like protein. Additionally, the diversity of the immune tumor microenvironment and matrix was assessed which showed that there was higher stiffness and more immune cell filtration among the ureteral group. Furthermore, the results showed that distinctive molecular subtypes, clinical features and tumor microenvironment existed in the muscle-invasive urothelial carcinomas.
Introduction
Upper tract urothelial cancers are neoplastic outgrowths occurring originating from the urothelial lining of the upper urinary tract (which includes the ureter and renal pelvis). They are highly invasive and relate to a poorer prognosis when compared with lower tract urothelial carcinomas (like bladder cancers). In upper tract urothelial carcinomas, their differing clinical features and epigenetic factors reveal that ureteral tumors and renal pelvic tumors may be two distinct conditions. Recent studies have shown that patients who have ureteral tumors usually present with a poor prognosis. The molecular subtypes of urothelial carcinomas were seen to be correlated. Deep learning algorithm used to determine molecular subtypes on histopathological pictures was preferred over the current classification system on the basis of molecular processes, as the latter is more expensive and time-consuming. For instance, Woerl and colleagues predicted the molecular features of muscle-invasive bladder carcinoma (MIBC) using a deep learning algorithm with hematoxylin and eosin stain, and the result from this is most likely to be applicable for upper tract urothelial carcinoma.
Tumor microenvironment characteristics such as matrix stiffness and immune infiltration differ based on their anatomic site. It was noted that an increased matrix stiffness could be a vital signal indicating tumor formation and also, a thin submucosal layer led to increased level of muscle invasion. Additionally, the formation of an immune microenvironment with certain possible immune pathways may have a role to play in cancer behavior.
Method
18 registries from the Surveillance, Epidemiology, and End Results (SEER) database were used to identify patients diagnosed with ureter carcinoma or renal pelvis carcinoma between the years 2004 and 2015. The SEER Stat software was employed in this study.
Chinese patients with upper tract urothelial carcinoma were evaluated using data obtained from the Dalian Medical University First Hospital from the years 2013 to 2021.
Ten patients with upper tract urothelial carcinoma who had surgery between the 1st January 2020 and 31st June 2020 were retrospectively selected. The Whole Human Genome Oligo Microarray, RNA quality and quantity, RNA array hybridization and RNA labeling data were obtained. R language with packages available publicly was used to process genetic data.
A BLCA cohort from TCGA was used to develop the MDA subtype classification model, which gives enough room to download images of digitized hematoxylin and eosin stained histopathological slides for model training and assessment, and also the genetic data for MDA label acquisition.
Some preprocessing steps were carried out before the model training which included deletion, tiling and color normalization using the Macenko technique. An inception V3 model was trained to identify tumor patches against stroma and necrosis using 1000 patches of necrosis, stroma and tumor characterized by an expert. Another inception model was built to classify the MDA subtype using tumor patches. Patches from each image in certain categories were randomly selected, making sure that the total number of included patches in each category amounted to 50,000, to lower the effect of possible data imbalance during tests.
Finally, the average predicted value of tumor patches was used as the final prediction result in the calculation of the patient level’s prediction value.
Inclusion Criteria
The ICD-O-3/WHO 2008 ureter/kidney and renal pelvis was used as a searching criteria. Samples that were included in this study were patients who had tumors only in the ureter or renal pelvis, patients with unilateral muscle-invasive cancer, patients who did not have any evidence of metastasis at diagnosis and patients who had sufficient frozen tissue specimens for additional research.
Exclusion Criteria
Patients were excluded using the ICD-O-3. Patients who had unknown survival time and unknown or insufficient clinical features were excluded from the study. Patients with non muscle-invasive data and synchronous bilateral upper tract urothelial carcinoma were also excluded from the study.
Statistical Analysis
χ2 -tests were used to test categorical variables, while the Mann-Whitney U test was used to test the continuous distribution. The relationship between certain factors and survival results were investigated using a Multivariate Cox proportional hazard model. The overall survival rates obtained using Kaplan-Meier curves were compared using a log-rank test. mRNA expression was compared using the Wilcoxon test. A significant threshold for the study was set at p < 0.005 and were all conducted on a two-sided basis.
Results
A total of 9012 patients recruited from the Surveillance, Epidemiology, and End Results (SEER) database were included in the study. 65.73% (5924 out of 9012) had renal pelvic tumor, while 34.27% (3088 out of 9012) had ureteral tumors. The total patient cohort had a median age of 73 years. The median survival duration was 51 months. The 5-year overall survival prediction was 41% for patients with renal pelvic tumors and 37% for patients with ureteral tumors. Univariate analysis showed that age, location, age, histologic type, grade and TNM stage were notably significant. The location of the tumor in the ureter was seen to be an independent risk factor using a multivariable cox regression model.
107 (43.1%) out of 248 patients in the hospital had renal pelvic tumors while 141 (45.9%) out of 248 patients in the hospital had ureteral tumors. It was discovered on evaluation of tumor location that patients who had ureteral tumors were more likely to have hydronephrosis, hematuria and smoke more compared to patients who had renal pelvic tumors. However, patients with renal pelvic tumors were likely to test positive in urine cytology with higher N/M stage. Ureteral tumors were hypothesized to be likely diagnosed earlier than renal pelvic tumors due to the presence of ureteral symptoms, which leads to the detection of ureteral tumors at its early stage, with a 69-month mean follow-up duration. Multivariate analysis was carried out following the identification of significant variables in univariate analysis, and it established that ureteral tumors had worse prognosis compared to renal pelvic cancers.
Molecular subtypes associated with muscle-invasive bladder carcinoma include UNC, TCGA, Lund and MDA. These subtypes are seen to be the reason for the different patterns of progression, biochemical processes and clinical features of the disease. Higher expression of ECM, luminal, epithelial-mesenchymal transition, squamous and basal markers was observed in the renal pelvis group. The analysis was unable to accurately confirm the transitional subtypes, probably due to the low sample size. The Lund system was used to validate findings that were noted, and it was discovered that the ureter tumor group were fortified with immune markers which is similar to the infiltrating subtype and characterized by the expression of stromal cell and immune biomarkers which indicated the presence of myofibroblast and immunologic cells. Thus, starting with the matrix and immune cells are thought to be a critical step.
Aside from the immunological markers, other identifying genetic components were seen to be highly expressed among the renal pelvis tumor group.
422 TCGA slide images were labeled via the MDA subtype and further divided into validation, testing and training datasets in a ratio of 2:2:6. A Cyclegan1o was trained by data from patients with upper tract urothelial carcinoma who had radical nephroureterectomy at the Dalian Medical University First Hospital from the years 2016 to 2021 to carry out color transformations from this study’s images to the TCGA images to better collate accurate prediction result for the study’s model. 44 patients were predicted to be luminal, 32 patients were predicted to be basal, while 4 patients were predicted to be p53-like. In all, the MDA-labeled deep learning algorithm in upper tract urothelial cancers was completely built and validated the mRNA results.
Genetic ontology (GO) analysis and molecular subtype analysis showed significant variations in the immune microenvironment. Tumor immunogenicity, immune checkpoints and tumor microenvironment (TME) are immunological factors that have been listed to be associated with a poor disease prognosis. The ureter tumor group had lower tumor purity, higher stromal scores, higher immune scores and higher estimate scores compared to the renal pelvis tumor group. Release of chemokines as a result of infiltration of the tumor microenvironment was seen more in the ureteral tumor group. The ureteral tumor group also had a higher rate of secretion of certain factors such as anti apoptotic factors and immunosuppressive factors. The renal pelvis tumor group showed lesser class 1 HLA-related antigen presenting molecules compared to the ureter tumor group. The ureteral tumor group expressed more immune checkpoint and costimulatory molecules compared to the renal pelvis tumors. This showed that the immune checkpoints expressed by ureteral tumors were to deflect immune killing.
Conclusion
It was determined in this study from multicenter data obtained from the Chinese cohorts and SEER database that patients with ureteral tumors had worse prognoses among all muscle-invasive upper tract urothelial carcinomas. Additionally, MDA clustering showed that ureteral tumors were basal and p53-type while renal pelvis tumors were luminal. Also, the deep learning model broadened the capacity of the small sample number. Patients with ureteral carcinomas are more likely to have hydronephrosis compared to patients with renal pelvis carcinoma, which is admittedly a predictor of poor prognosis of upper tract urothelial carcinomas. The submucosa of the ureter is thin which enhances muscular spread and invasion of carcinoma compared to the renal pelvis. The matrix and stiffness of ureteral tumors are more which contributes to the progression and recurrence of cancer compared to that of renal pelvis tumors.
Anatomical and mechanical parameters such as submucosa thickness, extracellular matrix stiffness and elements of the immunological elements like immune checkpoints play a role in tumor behaviors and patterns.
From this study, therapy targeted towards blocking CTLA-4 and preventing extracellular matrix stiffening may be valuable for treatment.
Oncology Related Tools
Other
Latest Research
