Molecular Hallmarks of Multiparametric Magnetic Resonance Imaging Visibility in Prostate Cancer

Abstract

Multiparametric magnetic resonance imaging (mpMRI) has transformed the management of localized prostate cancer by improving identification of clinically significant disease at diagnosis. Approximately 20% of primary prostate tumors are invisible to mpMRI, and we hypothesize that this invisibility reflects fundamental molecular properties of the tumor. We therefore profiled the genomes and transcriptomes of 40 International Society of Urological Pathology grade 2 tumors: 20 mpMRI-invisible (Prostate Imaging-Reporting and Data System [PI-RADS] v2 <3) and 20 mpMRI-visible (PI-RADS v2 5) tumors. mpMRI-visible tumors were enriched in hallmarks of nimbosus, an aggressive pathological, molecular, and microenvironmental phenomenon in prostate cancer. These hallmarks included genomes with increased mutation density, a higher prevalence of intraductal carcinoma/cribriform architecture pathology, and altered abundance of 102 transcripts, including overexpression of noncoding RNAs such as SCHLAP1. Multiple small nucleolar RNAs (snoRNAs) were identified, and a snoRNA signature synergized with nimbosus hallmarks to discriminate visible from invisible tumors. These data suggest a confluence of aggressive molecular and microenvironmental phenomena underlie mpMRI visibility of localized prostate cancer. PATIENT SUMMARY: We examined the correlation between tumor biology and magnetic resonance imaging (MRI) visibility in a group of patients with low- intermediate-risk prostate cancer. We observed that MRI findings are associated with biological features of aggressive prostate cancer.

Keywords: Multiparametric magnetic resonance imaging visibility; Nimbosus; Prostate cancer; Radiogenomics; Transcriptomics.

Figure 1 -. Transcriptomic basis of mpMRI visibility

A) Outline of study design, highlighting criteria for patient inclusion and exclusion. B) Differentially abundant RNAs in PI-RADSv2 5 vs 1–2 tumors as determined from negative binomial generalized log-linear model. Horizontal dotted line indicates FDR=0.05. ANKRD30A (C) and RLN1 (D) are upregulated in PI-RADSv2 5 tumors. Fold change and FDR from negative binomial generalized log-linear model and boxplots show 25% quartile, median and 75% quartile. E) snoRNAs were enriched in the subset of differentially abundant RNAs. The top barplot indicates the number of each non-coding RNA in the subset and the dotmap indicates the enrichment of each RNA. The size of the circle reflects the odds ratio in log₂ space for easier visualization while the colour of the circle reflects if the raw odds ratio is > 1 (red) or < 1 (blue). Finally, the background shading indicates the FDR (Fisher’s Exact test). F) A snoRNA signature accurately predicted PI-RADSv2 5 tumors with 76% accuracy.

Figure 2 -. Nimbosus drives mpMRI visibility

A) SCHLAP1 was upregulated in PI-RADSv2 5 tumors as determined from negative binomial generalized log-linear model. Boxplot shows 25% quartile, median and 75% quartile. B) PGA was upregulated in PI-RADSv2 5 tumors. P-value from logistic regression adjusting for tumour purity. C) IDC/CA was enriched in PI-RADSv2 5 tumors. OR and p-value from Fisher’s Exact test. D) Hypoxia, quantified using the Ragnum signature³⁴, was not significantly different between conditions. P-value from Mann-Whitney test. E) The co-occurrence of two or more nimbosus hallmarks increased the odds of PI-RADSv2 5 tumor. OR and p-value from Fisher’s Exact test. F) Nimbosus hallmarks synergize with snoRNA signature to accurately predict PI-RADSv2 5 tumours with 87% accuracy.