Skip to main content
NCBI home page
JCO Precision Oncology logoLink to JCO Precision Oncology
. 2019 Jun 12;3:PO.19.00054. doi: 10.1200/PO.19.00054

Biologic Significance of Magnetic Resonance Imaging Invisibility in Localized Prostate Cancer

Simpa S Salami 1,2,✉, Jeremy B Kaplan 1, Srinivas Nallandhighal 1, Mandeep Takhar 3, Jeffrey J Tosoian 1, Matthew Lee 1, Junhee Yoon 4, Daniel H Hovelson 1, Komal R Plouffe 1, Samuel D Kaffenberger 1,2, Edward M Schaeffer 5, R Jeffrey Karnes 6, Tamara L Lotan 7, Todd M Morgan 1,2, Arvin K George 1,2, Jeffrey S Montgomery 1,2, Matthew S Davenport 1, Sungyong You 4, Scott A Tomlins 1,2, Nicole E Curci 1, Hyung L Kim 4, Daniel E Spratt 2,1, Aaron M Udager 1,2, Ganesh S Palapattu 1,2,8
PMCID: PMC7446477  PMID: 32914029

Abstract

PURPOSE

Multiparametric magnetic resonance imaging (mpMRI) is used widely for prostate cancer (PCa) evaluation. Approximately 35% of aggressive tumors, however, are not visible on mpMRI. We sought to identify the molecular alterations associated with mpMRI-invisible tumors and determine whether mpMRI visibility is associated with PCa prognosis.

METHODS

Discovery and validation cohorts included patients who underwent mpMRI before radical prostatectomy and were found to harbor both mpMRI-visible (Prostate Imaging and Reporting Data System 3 to 5) and -invisible (Prostate Imaging and Reporting Data System 1 or 2) foci on surgical pathology. Next-generation sequencing was performed to determine differential gene expression between mpMRI-visible and -invisible foci. A genetic signature for tumor mpMRI visibility was derived in the discovery cohort and assessed in an independent validation cohort. Its association with long-term oncologic outcomes was evaluated in a separate testing cohort.

RESULTS

The discovery cohort included 10 patients with 26 distinct PCa foci on surgical pathology, of which 12 (46%) were visible and 14 (54%) were invisible on preoperative mpMRI. Next-generation sequencing detected prioritized genetic mutations in 14 (54%) tumor foci (n = 8 mpMRI visible, n = 6 mpMRI invisible). A nine-gene signature (composed largely of cell organization/structure genes) associated with mpMRI visibility was derived (area under the curve = 0.89), and the signature predicted MRI visibility with 75% sensitivity and 100% specificity (area under the curve = 0.88) in the validation cohort. In the testing cohort (n = 375, median follow-up 8 years) there was no significant difference in biochemical recurrence, distant metastasis, or cancer-specific mortality in patients with predicted mpMRI-visible versus -invisible tumors (all P > .05).

CONCLUSION

Compared with mpMRI-invisible disease, mpMRI-visible tumors are associated with underexpression of cellular organization genes. mpMRI visibility does not seem to be predictive of long-term cancer outcomes, highlighting the need for biopsy strategies that detect mpMRI-invisible tumors.

INTRODUCTION

Distinguishing aggressive from indolent clinically localized prostate cancer (PCa) continues to pose a significant clinical challenge. Recent efforts to overcome this have involved the development and optimization of several diagnostic strategies, including multiparametric magnetic resonance imaging (mpMRI). mpMRI permits visual identification of areas that are suggestive for intermediate to high-grade cancer. The emergence of various MRI/ultrasound fusion biopsy platforms has led to increased detection of aggressive PCa by facilitating targeted biopsy of visible lesions.1-6 As a result, mpMRI is now widely used in guiding treatment decisions in men with clinically localized disease, especially when selecting patients suitable for active surveillance or potentially focal therapy.7-10 The prevailing view is that only mpMRI-visible cancers require clinical action.

CONTEXT

  • Key Objective

  • What is the molecular basis for prostate cancer visibility on multiparametric magnetic resonance imaging (mpMRI), and do mpMRI-invisible tumors harbor any clinical or biologic significance compared with visible tumors?

  • Knowledge Generated

  • mpMRI-visible tumors demonstrated underexpression of genes associated with cellular organization and structure. Using a novel genetic signature for tumor visibility on mpMRI, patients with predicted mpMRI-visible and -invisible tumors did not experience significant differences in biochemical recurrence, distant metastasis, or cancer-specific mortality during follow-up.

  • Relevance

  • Prostate cancers that are mpMRI invisible have similar clinical behavior to mpMRI-visible tumors. Negative mpMRI seems insufficient to rule out clinically relevant prostate cancer, and patients at increased risk should be considered for additional testing or systematic prostate biopsy.

However, use of mpMRI in the evaluation of men with PCa is limited by cancer multifocality and interfocal disease heterogeneity. Individual patients are known to harbor multiple spatially distinct PCa foci with varying clinical, radiographic, and pathologic characteristics.11-15 Up to 55% of all PCa foci and 35% of clinically significant foci are not visible on mpMRI.3,16,17 Furthermore, more than 35% of lesions 1 cm or larger are missed by mpMRI.17 Although some studies have demonstrated that up to 50% of mpMRI-invisible PCa may harbor relevant genomic alterations, the clinical and prognostic significance of mpMRI-invisible PCa remains unknown.18 An improved understanding of the molecular characteristics and clinical trajectories of mpMRI-visible and -invisible cancers could facilitate more optimal treatment allocation. For example, if mpMRI-invisible foci are found to be biologically indolent, those with a known diagnosis of low-grade disease and a negative mpMRI could be directed toward active surveillance. Similarly, those with a single lesion detected on mpMRI could be more confidently directed toward focal therapy, with low concern for missing a clinically relevant lesion. We herein sought to characterize the molecular profile of mpMRI-visible and -invisible PCa foci using next-generation sequencing (NGS). In addition, we test the prognostic significance of our mpMRI-derived genomic signature after radical prostatectomy (RP).

METHODS

Study Design

The study used three independent patient populations: discovery, validation, and testing cohorts. Institutional review board approval was obtained for each cohort. First, we identified patients with clinically localized disease who underwent preoperative mpMRI at the University of Michigan in 2015 to 2016 and were subsequently found to harbor multifocal PCa at RP. We enriched for patients with both mpMRI visible and invisible PCa (Figs 1A and 1B) to constitute the discovery cohort. The validation cohort from Cedars-Sinai Medical Center included patients with either mpMRI-visible or -invisible foci, as previously described.19 The testing cohort was composed of patients from the Decipher GRID PCa database treated at Johns Hopkins Medical Institute and Mayo Clinic (ClinicalTrials.gov identifier: NCT02609269) who underwent genome-wide expression profiling after RP.20,21

FIG 1.

FIG 1.

Open in a new tab

Radiogenomic characterization of multifocal prostate cancer. (A) Cartoon depicting multifocal prostate cancer (PCa) with both multiparametric magnetic resonance imaging (mpMRI)–visible (solid black, left) and invisible (gray with black discontinuous borders, right) lesions. (B) Coregistration of axial mpMRI images with whole-mount histopathology. (1) Axial high-resolution T2, (2) axial diffusion-weighted imaging (b-value = 1,600), and (3) axial apparent diffusion coefficient map shows a visible lesion corresponding to cancer focus P1 (grade group [GG] 4; arrows) on the radical prostatectomy specimen (hematoxylin and eosin, panel 4). Cancer foci P2 (GG 2) and P3 (GG 1) were both mpMRI invisible. (C) Integrative summary of the primary multifocal PCa cohort. Ten patients comprising 26 distinct PCa foci were evaluated. Two samples (patient 7) did not pass initial RNA quality thresholds and were thus omitted. Each patient had at least one MRI-visible and one MRI-invisible cancer focus. Recurrent DNA variants are shown. Log2 copy-number ratio for PTEN is also shown. PTEN one copy number loss was observed in 25% (three of 12) and 14.3% (two of 14) of mpMRI-visible and invisible cancer foci, respectively (false discovery rate, less than 5%). Expression of SPINK1, ERG, and ETV1, as well as expressed isoforms of TMPRSS2-ERG are shown.

Preoperative Prostate mpMRI and Pathologic Evaluation

In the discovery and validation cohorts, mpMRI comprising T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced imaging was obtained. All mpMRI results were re-reviewed and coregistered with whole-mount formalin-fixed paraffin-embedded RP specimens to delineate mpMRI-visible (Prostate Imaging and Reporting Data System [PI-RADS] version 2; score, 3 to 5) and -invisible foci. Additional procedural details are described in the Data Supplement. Data on mpMRI were not available for the testing cohort.22

Targeted DNA and RNA NGS

In the discovery cohort, DNA and RNA from each focus were co-isolated for targeted multiplex NGS as previously described23 and detailed in the Data Supplement. Our targeted NGS assays were designed to assess relevant PCa genomic and transcriptomic alterations and derive clinically available prognostic tests.15 The details of RNA sequencing in the validation cohort and genome-wide expression profiling in the testing cohorts have been previously described.19,24

Bioinformatic Analysis of the Discovery Cohort

NGS data analysis was performed using Torrent Suite (4.2.0; Thermo Fisher Scientfic, Waltham, MA) and the Coverage Analysis Plug-ins v.5.0.4. (Thermo Fisher Scientific), along with the Ion Reporter (4.2.0; Thermo Fisher Scientific). All other analyses were performed using R Project for Statistical Computing v.3.2.3. Details regarding targeted NGS techniques, quality control parameters, DNA copy number alterations and variant calls, fusion isoform and partner level analysis, androgen receptor (AR) and AR-splice variants detection, and prognostic scores derivation have been previously described and summarized in the Data Supplement.15,25,26

Differential Gene Expression Analysis of mpMRI-Visible and -Invisible Cancer Foci

To determine gene expression differences between mpMRI-visible and -invisible tumors, we analyzed RNAseq data from the discovery cohort using two approaches—differential expression (DE) analysis and random forest (RF) classifier—as described in the Data Supplement. From these two approaches, a gene expression signature comprising independent differentially expressed genes was developed to predict mpMRI tumor visibility status.

Prognostic Significance of mpMRI-Based Gene Expression Signature

A total of 375 patients with genome-wide expression profiles were pooled from two independent case-cohort studies20,21 to constitute the testing cohort. The mpMRI-based nine-gene expression signature was applied to the testing cohort to predict mpMRI visibility status. Kaplan-Meier curves and Cox proportional hazard regression were used to evaluate the performance of this signature in predicting oncological outcomes: biochemical recurrence-free survival (BFS), distant metastasis–free survival (DMFS), and PCa-specific mortality (PCSM). Multivariable analyses were performed to evaluate this signature as an independent predictor of oncological outcomes after adjusting for relevant clinicopathological variables, including preoperative prostate-specific antigen, pathologic grade group (GG), surgical margins, extraprostatic extension, seminal vesicle invasion, and lymph node invasion. Spearman correlation analysis was performed to measure the association of the gene signature with cellular organization pathway activity. Mean expression of genes involved in the cellular organization pathway on the Oncotype Dx genomic prostate score (GPS; Genomic Health, Redwood City, CA) assay was correlated with the mpMRI-based gene expression signature.27 Statistical analyses were performed in R version 3.3.3, and all statistical tests were two-sided using a .05 significance level.

RESULTS

Study Cohorts

The discovery cohort included 10 patients from the University of Michigan PCa database with both mpMRI-visible and -invisible lesions (Fig 1). The clinicopathological characteristics of the discovery cohort are shown in the Data Supplement. Of the 26 cancer foci identified on surgical pathology specimens, 12 foci (46%) were visible on mpMRI. Among the 14 mpMRI-invisible foci (54%), five (36%) were GG2 and the remainder were GG1 (Fig 1C and Data Supplement). There were 16 patients in the validation cohort, of whom eight (50%) had mpMRI-invisible cancer lesions, and two of these (25%) were GG2 (Data Supplement). A summary of patient-level characteristics of the testing cohort (n = 375) stratified by predicted mpMRI visibility status is shown in the Data Supplement. The median age at RP was 62 years, and median follow-up time for censored patients was 8 years. During follow-up, 136 (36.3%) patients experienced biochemical recurrence, 55 (14.7%) developed metastasis, and 28 (7.5%) died as a result of PCa (Data Supplement).

Detection of Mutations and Copy Number Alterations in the Discovery Cohort

We detected high-confidence mutations in 14 of 26 (54%) tumor foci; six (43%) of the mutations were identified in mpMRI-invisible lesions. Notable somatic point mutations were in APC, ARID1B, ATM, NOTCH1, and SPOP. We detected PTEN one copy number loss in 25% (three of 12) and 14.3% (two of 14) of mpMRI-visible and -invisible foci, respectively (Fig 1C).

Discovery and Validation of a Nine-Gene Expression Signature for mpMRI Visibility

Of the 26 total tumor foci in the discovery cohort and 306 amplicons on the RNAseq panel, 24 samples and 74 amplicons, respectively, passed quality control parameters and underwent DE analysis (Data Supplement). Using DE analysis (Data Supplement) and RF classifier (Data Supplement) to identify candidate differentially expressed genes, we interrogated four separate logistic regression models for predicting mpMRI tumor visibility status using the 19 DE analysis genes, 20 RF genes, 11 shared genes between the DE analysis and RF gene sets, and 11 shared genes combined with the mutually exclusive genes (Data Supplement). A multivariable RNAseq-based logistic regression model with the best performance for predicting mpMRI visibility status, comprising a nine-gene expression signature, was developed from the intersection of the DE analysis and RF gene sets (Fig 2A; Data Supplement). This signature correctly predicted seven (70%) of the mpMRI-visible and 13 (93%) of the mpMRI-invisible foci in the discovery cohort, yielding an area under the curve of 0.89. The optimal probability cutoff for predicting mpMRI-visible tumor was greater than 0.46, with a sensitivity and specificity of 80% and 86%, respectively, in the discovery cohort (Figs 2A and 2B). We observed underexpression of seven of the nine genes in mpMRI-visible tumors, the majority of which were stromal, cellular organization, and structure genes (Fig 2A; Data Supplement).

FIG 2.

FIG 2.

Open in a new tab

Development and validation of a nine-gene signature to predict multiparametric magnetic resonance imaging (mpMRI)–visible tumors. (A) Annotated heat map of differentially expressed genes in the training cohort. Differentially expressed genes were identified with the EdgeR package using a Benjamini-Hochberg procedure adjusted false discovery rate cutoff of less than 0.05. Targets and samples were clustered using hierarchical clustering on the basis of Euclidian distances. Comparisons of observed (first row) versus predicted (second row) mpMRI visibility using the nine-gene signature are shown in the annotation, as well as International Society of Urological Pathology grade group. (B) Receiver operating characteristic curves for the signature in the discovery (University of Michigan [UM]) versus the validation (Cedars-Sinai Medical Center [CSMC]) cohorts. The signature was developed with multivariable ridge logistic-regression model using cross-validation for λ hyperparameter selection. The area under the curve (AUC) for the signature was not significantly different between the discovery and the validation cohorts (0.89 v 0.88, Delong’s unpaired t test, P = .877). The optimal probability cutoff for predicting mpMRI-visible tumor was greater than 0.46, with a sensitivity and specificity of 75% and 100% in the validation cohort, respectively.

The nine-gene expression signature was then evaluated in the independent validation cohort (Cedars-Sinai Medical Center) using the predetermined optimal probability cutoff (from the discovery cohort) to predict mpMRI visibility status. The receiver operating characteristic curve in the validation cohort is shown in Fig 2B, with an area under the curve of 0.88. The sensitivity and specificity of the signature for predicting mpMRI visibility status were 75% and 100%, respectively, in the validation cohort. Notably, the signature correctly predicted two GG2 cancers that were mpMRI invisible in the validation cohort.

Prognostic Significance of the Nine-Gene mpMRI Visibility Expression Signature

The distribution of each gene composing the nine-gene signature in the normalized microarray data (testing cohort) from the Decipher GRID mapped to The Cancer Genome Atlas Prostate Adenocarcinoma (TCGA-PRAD) RNAseq closely resemble that of the discovery cohort (Data Supplement). We applied the expression signature to the testing cohort as a proxy for mpMRI tumor visibility. Of the 375 patients in the testing cohort, 177 (47.2%) were classified as mpMRI visible (Data Supplement). Using the predicted probability as a surrogate for mpMRI, we found that the mpMRI visibility signature was not a predictor of BFS, DMFS, or PCSM (Fig 3; all log-rank P > .05). Similar findings were observed when the testing cohort data were not mapped to the TCGA-PRAD RNAseq cohort (Data Supplement; all log-rank P > .05). Adjusting for relevant clinicopathological variables on multivariable analysis, we found that genomic signature–determined mpMRI visibility status was not an independent predictor of BCR, metastasis, or PCSM (Fig 4; Data Supplement; all P > .05). Similar findings were observed when the testing cohort data were not mapped to the TCGA-PRAD RNAseq data (Data Supplement; all P > .05)

FIG 3.

FIG 3.

Open in a new tab

Prognostic significance of predicted multiparametric magnetic resonance imaging (mpMRI) visibility status. Patients (n = 375) in the testing cohort were pooled from two independent case-cohort studies (Johns Hopkins Medical Institute [n = 260] and Mayo Clinic [n = 235]) to test the capacity of predicted mpMRI visibility status to predict (A) biochemical recurrence-free survival (BFS), (B) distant metastasis-free survival (DMFS), and (C) prostate cancer–specific mortality (PCSM). The expression data in this cohort were generated using Affymetrix human exon 1.0 ST array (Santa Clara, CA). Normalization was performed to match the distribution of the genomic data from this cohort to The Cancer Genome Atlas Prostate Adenocarcinoma RNAseq data, as described in Methods, to facilitate testing of the RNASeq-based nine-gene signature to predict mpMRI-visible tumors (Data Supplement). mpMRI visibility status was computed using the signature: high score denotes mpMRI-visible and low score denotes mpMRI-invisible tumor. Kaplan-Meier survival curves were plotted and compared between predicted mpMRI-visible and -invisible tumor using log-rank test. There were no significant differences in BFS, DMFS, and PCSM between predicted mpMRI-visible and -invisible tumors (all P > .05). Similar results were obtained using the Affymetrix microarray data that were not matched to the distribution of the The Cancer Genome Atlas Prostate Adenocarcinoma RNAseq data (Data Supplement).

FIG 4.

FIG 4.

Open in a new tab

Multivariable analysis to assess the prognostic significance of predicted multiparametric magnetic resonance imaging (mpMRI) visibility status. Using data from the testing cohort described in Figure 3 (Affymetrix microarray data matched to the distribution of the The Cancer Genome Atlas Prostate Adenocarcinoma RNAseq data; n = 375), multivariable Cox proportional hazard models were developed to assess the capacity of predicted mpMRI visibility status to predict: (A) biochemical recurrence-free survival (BFS), (B) distant metastasis–free survival (DMFS), and (C) prostate cancer–specific mortality (PCSM), adjusting for relevant clinicopathological variables. mpMRI visibility status was not an independent predictor of BFS, DMFS, and PCSM (all adjusted P > .05). Similar results were obtained when the Affymetrix microarray data were not matched to the distribution of The Cancer Genome Atlas Prostate Adenocarcinoma RNAseq data (Data Supplement). EPE, extraprostatic extension; GG, grade group; HR, hazard ratio; LNI, lymph node invasion; PSA, prostate-specific antigen; SM, surgical margins; SVI, seminal vesical invasion.

Molecular Basis of Cancer Visibility on mpMRI

Using our multiplex (mx) RNAseq data from the discovery cohort, we derived commercially available tissue-based prognostic biomarker test scores (Myriad Prolaris cell cycle progression [mxCCP] score, Oncotype DX [mxGPS], and the GenomeDX genomic classifier [mxGC]) for each cancer focus, as previously described.15 We found no significant difference in the mxCCP, mxGPS, and mxGC scores between mpMRI-visible and -invisible foci (Fig 5A; all P > .05). However, as described above, we observed underexpression of seven of the nine genes in mpMRI-visible tumors, the majority of which were stromal, cellular organization, and structure genes (Fig 2A; Data Supplement). We then computed three subcomponents of the OncotypeDx GPS, as previously described,15,27 and compared these between mpMRI-visible and -invisible tumors. There were no significant differences in the expression of OncotypeDx GPS androgen signaling and stromal response submodules between mpMRI-visible and -invisible tumors (Fig 5B; both P > .05). However, we found underexpression of the cellular organization submodule of the OncotypeDx GPS panel in mpMRI-visible tumors consistent with the results of the nine-gene signature (Fig 5B; all P = .014). Similarly, using data from the testing cohort, we found underexpression of the OncotypeDx GPS cellular organization module in predicted mpMRI-visible compared with -invisible foci (Data Supplement; all P < .05). Taken together, these findings suggest that loss of cellular organization and structure contributes to PCa visibility on mpMRI.

FIG 5.

FIG 5.

Open in a new tab

Derivation and comparison of expression-based prognostic scores between multiparametric magnetic resonance imaging (mpMRI)–visible and -invisible lesions. (A) Box plots of derived Prolaris cell cycle progression (mxCCP) score, Oncotype DX genomic prostate score (mxGPS), and Decipher genomic classifier (mxGC) stratified by mpMRI visibility status in the discovery cohort (n = 10 patients; 26 cancer foci). Points represent individual cancer focus colored according to International Society of Urological Pathology grade group (GG). Unpaired t tests were used to test for significant differences in mean score. There was no statistically significant difference between the derived prognostic scores of mpMRI-visible and -invisible lesions (P > .05). (B) Comparisons of derived mxGPS submodules stratified by mpMRI visibility status. Box plots of derived mxGPS androgen, cellular organization, and stromal submodules stratified by mpMRI visibility status are shown, with each point representing an individual cancer focus colored according to GG. Unpaired t tests were used to compare mean subcomponent scores. Only the cellular organization submodule had a significant difference in mean expression (P = .014), suggesting that at the RNA expression level mpMRI visibility is related to underlying cellular organization of the tumor.

DISCUSSION

To better understand the molecular alterations associated with mpMRI visibility and prognostic significance of mpMRI-invisible disease, we performed a comprehensive molecular characterization of primary multifocal PCa inclusive of both mpMRI-visible and -invisible tumor foci using a targeted multiplex NGS approach. We observed that mpMRI-invisible cancer may possess mutations in known cancer-associated genes, with close to 15% harboring PTEN one copy number loss. Using robust biostatistic methods, we developed and validated a novel nine-gene signature to predict PCa mpMRI visibility status. Interrogation of this signature in a distinct cohort with long-term follow-up revealed no significant association with BFS, DMFS, or PCSM. Intriguingly, additional analyses revealed that underexpression of genes associated with cellular organization and structure may underpin the molecular basis of PCa visibility on mpMRI. Taken together, these findings indicate that mpMRI-invisible cancer foci harbor many of the same aggressive molecular features as mpMRI-visible foci and may also be clinically significant.

The molecular basis of PCa visibility on mpMRI is poorly understood. Although tumor size and grade contribute to cancer visibility on mpMRI, the architecture of the glands may play an important role.28-31 For example, tumors harboring cribriform Gleason pattern 4 were less likely to be detected by mpMRI compared with poorly formed or fused glands, suggesting that tumor size and grade alone do not explain cancer visibility on mpMRI.28 In a recent report, Li et al19 observed significant fold changes of differentially expressed genes on the basis of mpMRI visibility regardless of Gleason score or tumor size, including genes involved in cytoskeleton organization. Similarly, the majority of genes composing our novel mpMRI visibility signature are involved in cytoskeletal organization and structure. Other smaller-scale studies have reported the possible role of CHD1 deletion32 and PTEN loss33,34 in PCa mpMRI visibility. In the current study, we found that 25% of mpMRI-visible foci in the discovery cohort demonstrated PTEN one copy number loss compared with 14% in mpMRI-invisible foci. In aggregate, our work and that of others suggests that cellular (dis)organization contributes significantly to the underlying basis of PCa visibility on mpMRI. Additional studies are needed to further characterize the fundamental basis of PCa visibility on mpMRI.

The prognostic significance of mpMRI-invisible PCa foci is unknown. Although PCa is multifocal and mpMRI may miss up to 35% of intermediate- to high-grade PCa, the absence of visible lesions on mpMRI has been proposed as a reason to defer confirmatory biopsy when considering active surveillance.1,3,16,17 In addition, mpMRI is increasingly being used to identify the index or dominant cancer foci for focal therapy. To be sure, although size and grade are believed to be important, how best to define the biologically dominant cancer in multifocal disease is not known.35 To date, no study has demonstrated the clinical trajectory of mpMRI-visualized lesions. Such a study would be a challenge to perform, given the long duration of follow-up required and the multifocal nature of PCa, with frequent coexistence of mpMRI-visible and -invisible cancers within the same gland.31

Salmasi et al36 reported that the PI-RADS (a grading system for mpMRI lesion visibility) was not a significant predictor of adverse pathology at the time of RP. Similarly, Parry et al18 found that 50% of mpMRI-invisible cancers harbored one or more genetic alterations commonly observed in metastatic castrate-resistant PCa, suggesting that mpMRI-invisible tumors may be as important as visible ones. In the study by Li et al,19 a four-gene signature comprising genes differentially expressed between mpMRI-visible and -invisible PCa was shown to predict BFS in two external data sets. However, this signature was not developed as a predictor of or a surrogate for mpMRI cancer visibility, but rather it was selected on the basis of their common association with mpMRI visibility and metastasis. By contrast, in this first study of its kind to our knowledge, using a validated novel mpMRI-based RNAseq signature as a surrogate instrument for mpMRI visibility status, we have demonstrated that predicted mpMRI visibility status was not associated with BFS, DMFS, or PCSM during long-term follow-up. Put another way, mpMRI-invisible PCa does not seem to represent purely indolent disease; mpMRI-invisible lesions may be just as clinically relevant as mpMRI-visible disease. Future studies aimed at better defining the biologically dominant nodule and prognostic significance of mpMRI are warranted.

Our findings have significant clinical implications in the management of PCa. First, in the diagnostic setting, these data corroborate findings from several institutions indicating that a negative mpMRI does not rule out the presence of clinically significant PCa3,17,31 and should therefore not preclude a prostate biopsy without consideration of clinical risk.37,38 Second, in the setting of active surveillance, our findings underscore the potential for mpMRI-invisible cancer foci to harbor similar biologic trajectories as mpMRI-visible disease. Although additional studies are needed to delineate the utility of mpMRI in reducing the frequency of surveillance biopsies, the current literature supports systematic in addition to targeted biopsies in men undergoing active surveillance.3,6 Last, for men considering focal therapy, our data demonstrate that mpMRI alone is not sufficient to rule out the presence of a potentially lethal, nondominant cancer focus.

Our study has several limitations. First, we used a targeted NGS approach; thus it is conceivable that other potential alterations implicated in mpMRI cancer visibility may have been missed. Notwithstanding, our novel RNAseq signature developed from a targeted NGS approach demonstrated high fidelity for predicting mpMRI visibility in the validation cohort, where tumors underwent whole-transcriptome profiling. Second, we did not use the commercially available platforms for Oncotype Dx, Prolaris, and Decipher assays in the discovery cohort. The validity and consistency of deriving these scores from RNAseq data has been previously reported.15 Third, there were no GG4 and 5 lesions in our cohort. However, our novel RNAseq signature demonstrated high accuracy for predicting mpMRI visibility in the validation cohort, including 19% GG5 lesions. Moreover, GG4 and 5 lesions are generally mpMRI visible, and such patients routinely undergo whole-gland therapy. Fourth, the discovery cohort was made up of a relatively small sample, with low proportion of ERG-positive tumors. Nonetheless, we similarly observed high test performance in the validation cohort with ERG overexpression in 31% of samples. Fifth, what constitutes mpMRI-visible or -invisible lesions is not purely objective. To facilitate reproducibility, all lesions in the current study were scored according to the validated PI-RADS v2 system. PI-RADS 1 and 2 lesions were classified as mpMRI invisible, and PI-RADS 3 to 5 were classified as mpMRI visible. Finally, the prognostic significance of mpMRI-invisible cancer was evaluated in the testing cohort using a surrogate molecular marker for mpMRI visibility status. Thus, additional studies are needed to delineate the prognostic significance of mpMRI-invisible PCa in a prospective clinical cohort.

Discerning aggressive from indolent disease remains a significant clinical challenge in the evaluation and management of men with primary PCa. Our findings indicate that mpMRI-invisible cancers were no less likely to harbor lethal biologic potential than visible tumors, highlighting the limitation of using mpMRI alone to guide patient management or delineate specific index cancer foci for ablative therapy. Our results also highlight the continued need for biopsy strategies that detect mpMRI-invisible tumors. Future PCa molecular studies are needed to further characterize the molecular basis of cancer visibility on mpMRI and determine its prognostic significance.

Footnotes

Presented in part at the ASCO Genitourinary Cancer Symposium, San Francisco, CA, February 14-16, 2019; the European Association of Urology Annual Meeting, Barcelona, Spain, March 15-19, 2019; and the American Urological Association Annual Meeting, Chicago, IL, May 3-6, 2019.

Supported in part by the Prostate Cancer Foundation (S.S.S., T.M.M., S.A.T., D.E.S., and G.S.P.); the National Institutes of Health Grant No. R01 CA183857 (S.A.T.); the University of Michigan Prostate Specialized Program of Research Excellence Grant No. P50 CA186786-05; the Men of Michigan Prostate Cancer Research Fund, the University of Michigan Comprehensive Cancer Center core Grant No. 2-P30-CA-046592-24; the A. Alfred Taubman Biomedical Research Institute (T.M.M. and S.A.T.); and the Department of Defense (S.S.S., T.M.M., and D.E.S.).

AUTHOR CONTRIBUTIONS

Conception and design: Simpa S. Salami, Edward M. Schaeffer, Matthew S. Davenport, Scott A. Tomlins, Daniel E. Spratt, Aaron M. Udager, Ganesh S. Palapattu

Financial support: Scott A. Tomlins, Ganesh S. Palapattu

Administrative support: Ganesh S. Palapattu

Provision of study material or patients: R. Jeffrey Karnes, Arvin K. George, Daniel E. Spratt

Collection and assembly of data: Simpa S. Salami, Jeremy B. Kaplan, Matthew Lee, Komal R. Plouffe, R. Jeffrey Karnes, Tamara L. Lotan, Arvin K. George, Nicole E. Curci, Daniel E. Spratt, Aaron M. Udager

Data analysis and interpretation: Simpa S. Salami, Jeremy B. Kaplan, Srinivas Nallandhighal, Mandeep Takhar, Jeffrey J. Tosoian, Matthew Lee, Junhee Yoon, Daniel H. Hovelson, Samuel D. Kaffenberger, Edward M. Schaeffer, R. Jeffrey Karnes, Todd M. Morgan, Arvin K. George, Jeffrey S. Montgomery, Matthew S. Davenport, Sungyong You, Hyung L. Kim, Daniel E. Spratt, Aaron M. Udager

Manuscript writing: All authors

Final approval of manuscript: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Jeremy B. Kaplan

Employment: Tempus Labs

Mandeep Takhar

Employment: GenomeDx

Daniel H. Hovelson

Employment: Strata Oncology

Consulting or Advisory Role: Terumo

Travel, Accommodations, Expenses: Thermo Fisher Scientific

Samuel D. Kaffenberger

Consulting or Advisory Role: MDxHealth, Clovis Oncology

Travel, Accommodations, Expenses: Bristol-Myers Squibb

Edward M. Schaeffer

Consulting or Advisory Role: OPKO Diagnostics, AbbVie

R. Jeffrey Karnes

Research Funding: GenomeDx

Patents, Royalties, Other Intellectual Property: GenomeDx

Travel, Accommodations, Expenses: GenomeDx

Tamara L. Lotan

Consulting or Advisory Role: Janssen Oncology

Research Funding: Ventana Medical Systems

Todd M. Morgan

Consulting or Advisory Role: Myriad Genetics, Terumo BCT

Research Funding: Myriad Genetics (Inst), MDxHealth (Inst), GenomeDx (Inst)

Arvin K. George

Consulting or Advisory Role: Trod Medical

Research Funding: Nanospectra Biosciences (Inst)

Travel, Accommodations, Expenses: Profound Medical, Trod Medical

Matthew S. Davenport

Patents, Royalties, Other Intellectual Property: Royalties from Wolters Kluwer, royalties from uptodate.com

Scott A. Tomlins

Employment: Strata Oncology

Leadership: Strata Oncology

Stock and Other Ownership Interests: Strata Oncology

Consulting or Advisory Role: AbbVie, Janssen, Astellas Medivation, Strata Oncology, Sanofi, Almac Diagnostics

Research Funding: Astellas Medivation (Inst), GenomeDx (Inst)

Patents, Royalties, Other Intellectual Property: I am a coauthor on a patent issued to the University of Michigan on ETS gene fusions in prostate cancer. The diagnostic field of use has been licensed to Hologic/Gen-Probe, who has sublicensed some rights to Ventana Medical Systems/Roche.

Travel, Accommodations, Expenses: Strata Oncology

Hyung L. Kim

Research Funding: GenomeDx

Patents, Royalties, Other Intellectual Property: I am an inventor of a prognostic biomarker signature and immunotherapies.

Daniel E. Spratt

Consulting or Advisory Role: Blue Earth Diagnostics, Janssen Oncology

Aaron M. Udager

Research Funding: Ventana Medical Systems

Ganesh S. Palapattu

Stock and Other Ownership Interests: NantKwest

Consulting or Advisory Role: Janssen Scientific Affairs

Research Funding: Minomic

Patents, Royalties, Other Intellectual Property: Implantable nanotechnology for long-term testosterone delivery functionalized fiducial marker for drug delivery

No other potential conflicts of interest were reported.

REFERENCES

  • 1.Kasivisvanathan V, Rannikko AS, Borghi M, et al. MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med. 2018;378:1767–1777. doi: 10.1056/NEJMoa1801993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Ahmed HU, El-Shater Bosaily A, Brown LC, et al. Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): A paired validating confirmatory study. Lancet. 2017;389:815–822. doi: 10.1016/S0140-6736(16)32401-1. [DOI] [PubMed] [Google Scholar]
  • 3.Filson CP, Natarajan S, Margolis DJA, et al. Prostate cancer detection with magnetic resonance-ultrasound fusion biopsy: The role of systematic and targeted biopsies. Cancer. 2016;122:884–892. doi: 10.1002/cncr.29874. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Salami SS, Ben-Levi E, Yaskiv O, et al: In patients with a previous negative prostate biopsy and a suspicious lesion on magnetic resonance imaging, is a 12-core biopsy still necessary in addition to a targeted biopsy? BJU Int 115:562-570, 2015. [DOI] [PubMed]
  • 5.Salami SS, Vira MA, Turkbey B, et al. Multiparametric magnetic resonance imaging outperforms the Prostate Cancer Prevention Trial risk calculator in predicting clinically significant prostate cancer. Cancer. 2014;120:2876–2882. doi: 10.1002/cncr.28790. [DOI] [PubMed] [Google Scholar]
  • 6.Siddiqui MM, Rais-Bahrami S, Turkbey B, et al. Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA. 2015;313:390–397. doi: 10.1001/jama.2014.17942. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Ahmed HU, Hindley RG, Dickinson L, et al. Focal therapy for localised unifocal and multifocal prostate cancer: A prospective development study. Lancet Oncol. 2012;13:622–632. doi: 10.1016/S1470-2045(12)70121-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Natarajan S, Raman S, Priester AM, et al. Focal laser ablation of prostate cancer: Phase I clinical trial. J Urol. 2016;196:68–75. doi: 10.1016/j.juro.2015.12.083. [DOI] [PubMed] [Google Scholar]
  • 9.Ahmed HU, Dickinson L, Charman S, et al. Focal ablation targeted to the index lesion in multifocal localised prostate cancer: A prospective development study. Eur Urol. 2015;68:927–936. doi: 10.1016/j.eururo.2015.01.030. [DOI] [PubMed] [Google Scholar]
  • 10.Guillaumier S, Peters M, Arya M, et al. A multicentre study of 5-year outcomes following focal therapy in treating clinically significant nonmetastatic prostate cancer. Eur Urol. 2018;74:422–429. doi: 10.1016/j.eururo.2018.06.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Boutros PC, Fraser M, Harding NJ, et al. Spatial genomic heterogeneity within localized, multifocal prostate cancer. Nat Genet. 2015;47:736–745. doi: 10.1038/ng.3315. [DOI] [PubMed] [Google Scholar]
  • 12.Cooper CS, Eeles R, Wedge DC, et al. Analysis of the genetic phylogeny of multifocal prostate cancer identifies multiple independent clonal expansions in neoplastic and morphologically normal prostate tissue. Nat Genet. 2015;47:367–372. doi: 10.1038/ng.3221. Erratum: Nat Genet 47:689, 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Cancer Genome Atlas Research Network The molecular taxonomy of primary prostate cancer. Cell. 2015;163:1011–1025. doi: 10.1016/j.cell.2015.10.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kumar A, Coleman I, Morrissey C, et al. Substantial interindividual and limited intraindividual genomic diversity among tumors from men with metastatic prostate cancer. Nat Med. 2016;22:369–378. doi: 10.1038/nm.4053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. doi: 10.1172/jci.insight.123468. Salami SS, Hovelson DH, Kaplan JB, et al: Transcriptomic heterogeneity in multifocal prostate cancer. JCI Insight 3:123468, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Radtke JP, Kuru TH, Boxler S, et al. Comparative analysis of transperineal template saturation prostate biopsy versus magnetic resonance imaging targeted biopsy with magnetic resonance imaging-ultrasound fusion guidance. J Urol 193:87-94, 2015. [DOI] [PubMed]
  • 17.Johnson DC, Raman SS, Mirak SA, et al. Detection of individual prostate cancer foci via multiparametric magnetic resonance imaging. Eur Urol. 2019;75:712–720. doi: 10.1016/j.eururo.2018.11.031. [DOI] [PubMed] [Google Scholar]
  • 18. Parry MA, Srivastava S, Ali A, et al. Genomic evaluation of multiparametric magnetic resonance imaging-visible and -nonvisible lesions in clinically localised prostate cancer. Eur Urol Oncol 2:1-11, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Li P, You S, Nguyen C, et al. Genes involved in prostate cancer progression determine MRI visibility. Theranostics. 2018;8:1752–1765. doi: 10.7150/thno.23180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Ross AE, Johnson MH, Yousefi K, et al. Tissue-based genomics augments post-prostatectomy risk stratification in a natural history cohort of intermediate- and high-risk men. Eur Urol. 2016;69:157–165. doi: 10.1016/j.eururo.2015.05.042. [DOI] [PubMed] [Google Scholar]
  • 21.Karnes RJ, Bergstralh EJ, Davicioni E, et al. Validation of a genomic classifier that predicts metastasis following radical prostatectomy in an at risk patient population. J Urol. 2013;190:2047–2053. doi: 10.1016/j.juro.2013.06.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Weinreb JC, Barentsz JO, Choyke PL, et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol. 2016;69:16–40. doi: 10.1016/j.eururo.2015.08.052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Hovelson DH, McDaniel AS, Cani AK, et al. Development and validation of a scalable next-generation sequencing system for assessing relevant somatic variants in solid tumors. Neoplasia. 2015;17:385–399. doi: 10.1016/j.neo.2015.03.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Erho N, Crisan A, Vergara IA, et al: Discovery and validation of a prostate cancer genomic classifier that predicts early metastasis following radical prostatectomy. PLoS One 8:e66855, 2013 . [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Warrick JI, Hovelson DH, Amin A, et al. Tumor evolution and progression in multifocal and paired non-invasive/invasive urothelial carcinoma. Virchows Arch. 2015;466:297–311. doi: 10.1007/s00428-014-1699-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Palapattu GS, Salami SS, Cani AK, et al. Molecular profiling to determine clonality of serial magnetic resonance imaging/ultrasound fusion biopsies from men on active surveillance for low-risk prostate cancer. Clin Cancer Res. 2017;23:985–991. doi: 10.1158/1078-0432.CCR-16-1454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Klein EA, Cooperberg MR, Magi-Galluzzi C, et al. A 17-gene assay to predict prostate cancer aggressiveness in the context of Gleason grade heterogeneity, tumor multifocality, and biopsy undersampling. Eur Urol. 2014;66:550–560. doi: 10.1016/j.eururo.2014.05.004. [DOI] [PubMed] [Google Scholar]
  • 28.Truong M, Hollenberg G, Weinberg E, et al. Impact of Gleason subtype on prostate cancer detection using multiparametric magnetic resonance imaging: Correlation with final histopathology. J Urol. 2017;198:316–321. doi: 10.1016/j.juro.2017.01.077. [DOI] [PubMed] [Google Scholar]
  • 29.Vargas HA, Akin O, Shukla-Dave A, et al. Performance characteristics of MR imaging in the evaluation of clinically low-risk prostate cancer: A prospective study. Radiology. 2012;265:478–487. doi: 10.1148/radiol.12120041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Hurrell SL, McGarry SD, Kaczmarowski A, et al. Optimized b-value selection for the discrimination of prostate cancer grades, including the cribriform pattern, using diffusion weighted imaging. J Med Imaging (Bellingham) 2018;5:011004. doi: 10.1117/1.JMI.5.1.011004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Le JD, Tan N, Shkolyar E, et al. Multifocality and prostate cancer detection by multiparametric magnetic resonance imaging: Correlation with whole-mount histopathology. Eur Urol. 2015;67:569–576. doi: 10.1016/j.eururo.2014.08.079. [DOI] [PubMed] [Google Scholar]
  • 32.Lee D, Fontugne J, Gumpeni N, et al. Molecular alterations in prostate cancer and association with MRI features. Prostate Cancer Prostatic Dis. 2017;20:430–435. doi: 10.1038/pcan.2017.33. [DOI] [PubMed] [Google Scholar]
  • 33. McCann SM, Fan X, Wang J, et al: Quantitative multiparametric MRI features and PTEN expression of peripheral zone prostate cancer: A pilot study. 206:559-565, 2016. [DOI] [PubMed]
  • 34.Zundel W, Schindler C, Haas-Kogan D, et al. Loss of PTEN facilitates HIF-1-mediated gene expression. Genes Dev. 2000;14:391–396. [PMC free article] [PubMed] [Google Scholar]
  • 35.Haffner MC, Mosbruger T, Esopi DM, et al. Tracking the clonal origin of lethal prostate cancer. J Clin Invest. 2013;123:4918–4922. doi: 10.1172/JCI70354. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Salmasi A, Khoshnoodi P, Felker ER, et al: A 17-gene genomic prostate score assay provides independent information on adverse pathology in the setting of combined multiparametric magnetic resonance imaging fusion targeted and systematic prostate biopsy. J Urol 200:564-572, 2018. [DOI] [PubMed]
  • 37.Nassiri N, Natarajan S, Margolis DJ, et al. Targeted prostate biopsy: Lessons learned midst the evolution of a disruptive technology. Urology. 2015;86:432–438. doi: 10.1016/j.urology.2015.07.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Panebianco V, Barchetti G, Simone G, et al. Negative multiparametric magnetic resonance imaging for prostate cancer: What’s next? Eur Urol. 2018;74:48–54. doi: 10.1016/j.eururo.2018.03.007. [DOI] [PubMed] [Google Scholar]

Articles from JCO Precision Oncology are provided here courtesy of American Society of Clinical Oncology

RESOURCES